{"hparams": {"dataset": {"align_stage_components": ["download/llava-laion-cc-sbu-558k/chat.json", "download/llava-laion-cc-sbu-558k"], "dataset_id": "llava-v15", "dataset_root_dir": "data", "finetune_stage_components": ["download/llava-v1.5-instruct/llava_v1_5_mix665k.json", "download/llava-v1.5-instruct"], "type": "llava-v15"}, "hf_token": ".hf_token", "model": {"align_epochs": 1, "align_global_batch_size": 256, "align_learning_rate": 0.001, "align_lr_scheduler_type": "linear-warmup+cosine-decay", "align_max_grad_norm": 1.0, "align_max_steps": null, "align_per_device_batch_size": 16, "align_train_strategy": "fsdp-shard-grad-op", "align_warmup_ratio": 0.03, "align_weight_decay": 0.0, "arch_specifier": "gelu-mlp", "enable_gradient_checkpointing": true, "enable_mixed_precision_training": true, "finetune_epochs": 1, "finetune_global_batch_size": 128, "finetune_learning_rate": 2e-05, "finetune_lr_scheduler_type": "linear-warmup+cosine-decay", "finetune_max_grad_norm": 1.0, "finetune_max_steps": null, "finetune_per_device_batch_size": 16, "finetune_train_strategy": "fsdp-full-shard", "finetune_warmup_ratio": 0.03, "finetune_weight_decay": 0.1, "image_resize_strategy": "letterbox", "llm_backbone_id": "vicuna-v15-7b", "llm_max_length": 2048, "model_id": "full-ft-multi-stage+7b", "reduce_in_full_precision": false, "type": "full-ft-multi-stage+7b", "vision_backbone_id": "clip-vit-l-336px"}, "pretrained_checkpoint": null, "run_id": "full-ft-multi-stage+7b", "run_root_dir": "runs", "seed": 7, "stage": "full-finetune", "trackers": ["jsonl", "wandb"], "wandb_entity": null, "wandb_project": "prismatic"}, "run_id": "full-ft-multi-stage+7b+stage-full-finetune+x7"} {"Full-finetune/Learning Rate": 1.2903225806451614e-07, "Full-finetune/Loss": 1.0748499631881714, "Full-finetune/Loss (Raw)": 1.0748499631881714, "Full-finetune/Step": 1, "Full-finetune/Step Time": 24.772311687469482} {"Full-finetune/Learning Rate": 2.580645161290323e-07, "Full-finetune/Loss": 1.2188456058502197, "Full-finetune/Loss (Raw)": 1.3628413677215576, "Full-finetune/Step": 2, "Full-finetune/Step Time": 16.984344005584717} {"Full-finetune/Learning Rate": 3.870967741935484e-07, "Full-finetune/Loss": 1.1900430917739868, "Full-finetune/Loss (Raw)": 1.1324377059936523, "Full-finetune/Step": 3, "Full-finetune/Step Time": 13.21389921506246} {"Full-finetune/Learning Rate": 5.161290322580646e-07, "Full-finetune/Loss": 1.1564662456512451, "Full-finetune/Loss (Raw)": 1.0557360649108887, "Full-finetune/Step": 4, "Full-finetune/Step Time": 11.370725929737091} {"Full-finetune/Learning Rate": 6.451612903225807e-07, "Full-finetune/Loss": 1.1510323286056519, "Full-finetune/Loss (Raw)": 1.1292961835861206, "Full-finetune/Step": 5, "Full-finetune/Step Time": 10.467435789108276} {"Full-finetune/Learning Rate": 7.741935483870968e-07, "Full-finetune/Loss": 1.144124150276184, "Full-finetune/Loss (Raw)": 1.1095831394195557, "Full-finetune/Step": 6, "Full-finetune/Step Time": 9.774863600730896} {"Full-finetune/Learning Rate": 9.032258064516129e-07, "Full-finetune/Loss": 1.1417956352233887, "Full-finetune/Loss (Raw)": 1.1278249025344849, "Full-finetune/Step": 7, "Full-finetune/Step Time": 9.277042354856219} {"Full-finetune/Learning Rate": 1.0322580645161291e-06, "Full-finetune/Loss": 1.1565089225769043, "Full-finetune/Loss (Raw)": 1.2595027685165405, "Full-finetune/Step": 8, "Full-finetune/Step Time": 9.065148532390594} {"Full-finetune/Learning Rate": 1.1612903225806454e-06, "Full-finetune/Loss": 1.1348953247070312, "Full-finetune/Loss (Raw)": 0.9619851112365723, "Full-finetune/Step": 9, "Full-finetune/Step Time": 8.776648971769545} {"Full-finetune/Learning Rate": 1.2903225806451614e-06, "Full-finetune/Loss": 1.1524587869644165, "Full-finetune/Loss (Raw)": 1.3105307817459106, "Full-finetune/Step": 10, "Full-finetune/Step Time": 8.782447147369385} {"Full-finetune/Learning Rate": 1.4193548387096776e-06, "Full-finetune/Loss": 1.1618678569793701, "Full-finetune/Loss (Raw)": 1.2559579610824585, "Full-finetune/Step": 11, "Full-finetune/Step Time": 8.537609577178955} {"Full-finetune/Learning Rate": 1.5483870967741937e-06, "Full-finetune/Loss": 1.1595168113708496, "Full-finetune/Loss (Raw)": 1.1336543560028076, "Full-finetune/Step": 12, "Full-finetune/Step Time": 8.34317578872045} {"Full-finetune/Learning Rate": 1.67741935483871e-06, "Full-finetune/Loss": 1.154473900794983, "Full-finetune/Loss (Raw)": 1.0939580202102661, "Full-finetune/Step": 13, "Full-finetune/Step Time": 8.184080600738525} {"Full-finetune/Learning Rate": 1.8064516129032258e-06, "Full-finetune/Loss": 1.1655008792877197, "Full-finetune/Loss (Raw)": 1.3088511228561401, "Full-finetune/Step": 14, "Full-finetune/Step Time": 8.013396569660731} {"Full-finetune/Learning Rate": 1.935483870967742e-06, "Full-finetune/Loss": 1.1691389083862305, "Full-finetune/Loss (Raw)": 1.2200725078582764, "Full-finetune/Step": 15, "Full-finetune/Step Time": 7.964829095204672} {"Full-finetune/Learning Rate": 2.0645161290322582e-06, "Full-finetune/Loss": 1.1490130424499512, "Full-finetune/Loss (Raw)": 0.8471276164054871, "Full-finetune/Step": 16, "Full-finetune/Step Time": 7.930087745189667} {"Full-finetune/Learning Rate": 2.1935483870967745e-06, "Full-finetune/Loss": 1.1455979347229004, "Full-finetune/Loss (Raw)": 1.0909544229507446, "Full-finetune/Step": 17, "Full-finetune/Step Time": 7.8366144544938034} {"Full-finetune/Learning Rate": 2.3225806451612907e-06, "Full-finetune/Loss": 1.132299780845642, "Full-finetune/Loss (Raw)": 0.9062319397926331, "Full-finetune/Step": 18, "Full-finetune/Step Time": 7.77835410171085} {"Full-finetune/Learning Rate": 2.4516129032258066e-06, "Full-finetune/Loss": 1.1302481889724731, "Full-finetune/Loss (Raw)": 1.0933202505111694, "Full-finetune/Step": 19, "Full-finetune/Step Time": 7.678843385294864} {"Full-finetune/Learning Rate": 2.580645161290323e-06, "Full-finetune/Loss": 1.1264359951019287, "Full-finetune/Loss (Raw)": 1.0540025234222412, "Full-finetune/Step": 20, "Full-finetune/Step Time": 7.578145635128021} {"Full-finetune/Learning Rate": 2.709677419354839e-06, "Full-finetune/Loss": 1.118780493736267, "Full-finetune/Loss (Raw)": 0.9656715393066406, "Full-finetune/Step": 21, "Full-finetune/Step Time": 7.570456538881574} {"Full-finetune/Learning Rate": 2.8387096774193553e-06, "Full-finetune/Loss": 1.1135525703430176, "Full-finetune/Loss (Raw)": 1.00376558303833, "Full-finetune/Step": 22, "Full-finetune/Step Time": 7.514203082431447} {"Full-finetune/Learning Rate": 2.967741935483871e-06, "Full-finetune/Loss": 1.075116515159607, "Full-finetune/Loss (Raw)": 0.2295239269733429, "Full-finetune/Step": 23, "Full-finetune/Step Time": 7.578380149343739} {"Full-finetune/Learning Rate": 3.0967741935483874e-06, "Full-finetune/Loss": 1.0773600339889526, "Full-finetune/Loss (Raw)": 1.1289600133895874, "Full-finetune/Step": 24, "Full-finetune/Step Time": 7.506330172220866} {"Full-finetune/Learning Rate": 3.225806451612903e-06, "Full-finetune/Loss": 1.071800947189331, "Full-finetune/Loss (Raw)": 0.938385009765625, "Full-finetune/Step": 25, "Full-finetune/Step Time": 7.457449102401734} {"Full-finetune/Learning Rate": 3.35483870967742e-06, "Full-finetune/Loss": 1.041022777557373, "Full-finetune/Loss (Raw)": 0.27156585454940796, "Full-finetune/Step": 26, "Full-finetune/Step Time": 7.529557814964881} {"Full-finetune/Learning Rate": 3.4838709677419357e-06, "Full-finetune/Loss": 1.0453684329986572, "Full-finetune/Loss (Raw)": 1.1583571434020996, "Full-finetune/Step": 27, "Full-finetune/Step Time": 7.459796031316121} {"Full-finetune/Learning Rate": 3.6129032258064515e-06, "Full-finetune/Loss": 1.0449161529541016, "Full-finetune/Loss (Raw)": 1.0327024459838867, "Full-finetune/Step": 28, "Full-finetune/Step Time": 7.39781299659184} {"Full-finetune/Learning Rate": 3.741935483870968e-06, "Full-finetune/Loss": 1.0463663339614868, "Full-finetune/Loss (Raw)": 1.0869718790054321, "Full-finetune/Step": 29, "Full-finetune/Step Time": 7.389850764439024} {"Full-finetune/Learning Rate": 3.870967741935484e-06, "Full-finetune/Loss": 1.043196439743042, "Full-finetune/Loss (Raw)": 0.9512673616409302, "Full-finetune/Step": 30, "Full-finetune/Step Time": 7.386123029390971} {"Full-finetune/Learning Rate": 4.000000000000001e-06, "Full-finetune/Loss": 1.043619155883789, "Full-finetune/Loss (Raw)": 1.0563048124313354, "Full-finetune/Step": 31, "Full-finetune/Step Time": 7.357477564965525} {"Full-finetune/Learning Rate": 4.1290322580645165e-06, "Full-finetune/Loss": 1.0397331714630127, "Full-finetune/Loss (Raw)": 0.919269323348999, "Full-finetune/Step": 32, "Full-finetune/Step Time": 7.385377995669842} {"Full-finetune/Learning Rate": 4.258064516129032e-06, "Full-finetune/Loss": 1.0327354669570923, "Full-finetune/Loss (Raw)": 0.8088052868843079, "Full-finetune/Step": 33, "Full-finetune/Step Time": 7.367329980387832} {"Full-finetune/Learning Rate": 4.387096774193549e-06, "Full-finetune/Loss": 1.029011607170105, "Full-finetune/Loss (Raw)": 0.9061237573623657, "Full-finetune/Step": 34, "Full-finetune/Step Time": 7.332466980990241} {"Full-finetune/Learning Rate": 4.516129032258065e-06, "Full-finetune/Loss": 1.0323480367660522, "Full-finetune/Loss (Raw)": 1.1457891464233398, "Full-finetune/Step": 35, "Full-finetune/Step Time": 7.30841076033456} {"Full-finetune/Learning Rate": 4.6451612903225815e-06, "Full-finetune/Loss": 1.0312546491622925, "Full-finetune/Loss (Raw)": 0.9929848313331604, "Full-finetune/Step": 36, "Full-finetune/Step Time": 7.321942501597935} {"Full-finetune/Learning Rate": 4.774193548387097e-06, "Full-finetune/Loss": 1.0325647592544556, "Full-finetune/Loss (Raw)": 1.0797258615493774, "Full-finetune/Step": 37, "Full-finetune/Step Time": 7.374752933914597} {"Full-finetune/Learning Rate": 4.903225806451613e-06, "Full-finetune/Loss": 1.032148003578186, "Full-finetune/Loss (Raw)": 1.016729474067688, "Full-finetune/Step": 38, "Full-finetune/Step Time": 7.3316365543164705} {"Full-finetune/Learning Rate": 5.032258064516129e-06, "Full-finetune/Loss": 1.0302692651748657, "Full-finetune/Loss (Raw)": 0.9588775634765625, "Full-finetune/Step": 39, "Full-finetune/Step Time": 7.292703597973555} {"Full-finetune/Learning Rate": 5.161290322580646e-06, "Full-finetune/Loss": 1.0299100875854492, "Full-finetune/Loss (Raw)": 1.015904426574707, "Full-finetune/Step": 40, "Full-finetune/Step Time": 7.300389111042023} {"Full-finetune/Learning Rate": 5.290322580645162e-06, "Full-finetune/Loss": 1.0321282148361206, "Full-finetune/Loss (Raw)": 1.1208574771881104, "Full-finetune/Step": 41, "Full-finetune/Step Time": 7.251489627652052} {"Full-finetune/Learning Rate": 5.419354838709678e-06, "Full-finetune/Loss": 1.0343416929244995, "Full-finetune/Loss (Raw)": 1.1250859498977661, "Full-finetune/Step": 42, "Full-finetune/Step Time": 7.215615550676982} {"Full-finetune/Learning Rate": 5.548387096774194e-06, "Full-finetune/Loss": 1.0332951545715332, "Full-finetune/Loss (Raw)": 0.9893417954444885, "Full-finetune/Step": 43, "Full-finetune/Step Time": 7.214148538057194} {"Full-finetune/Learning Rate": 5.677419354838711e-06, "Full-finetune/Loss": 1.033769965171814, "Full-finetune/Loss (Raw)": 1.0541847944259644, "Full-finetune/Step": 44, "Full-finetune/Step Time": 7.20595558123155} {"Full-finetune/Learning Rate": 5.806451612903226e-06, "Full-finetune/Loss": 1.0307555198669434, "Full-finetune/Loss (Raw)": 0.8981221914291382, "Full-finetune/Step": 45, "Full-finetune/Step Time": 7.2641798178354895} {"Full-finetune/Learning Rate": 5.935483870967742e-06, "Full-finetune/Loss": 1.013856291770935, "Full-finetune/Loss (Raw)": 0.2533896565437317, "Full-finetune/Step": 46, "Full-finetune/Step Time": 7.314600846041804} {"Full-finetune/Learning Rate": 6.064516129032259e-06, "Full-finetune/Loss": 1.0143729448318481, "Full-finetune/Loss (Raw)": 1.038142204284668, "Full-finetune/Step": 47, "Full-finetune/Step Time": 7.293464462807838} {"Full-finetune/Learning Rate": 6.193548387096775e-06, "Full-finetune/Loss": 1.0132503509521484, "Full-finetune/Loss (Raw)": 0.9604852795600891, "Full-finetune/Step": 48, "Full-finetune/Step Time": 7.354513853788376} {"Full-finetune/Learning Rate": 6.3225806451612906e-06, "Full-finetune/Loss": 1.0141305923461914, "Full-finetune/Loss (Raw)": 1.0563864707946777, "Full-finetune/Step": 49, "Full-finetune/Step Time": 7.330218295661771} {"Full-finetune/Learning Rate": 6.451612903225806e-06, "Full-finetune/Loss": 1.012505054473877, "Full-finetune/Loss (Raw)": 0.9328505992889404, "Full-finetune/Step": 50, "Full-finetune/Step Time": 7.295989618301392} {"Full-finetune/Learning Rate": 6.580645161290323e-06, "Full-finetune/Loss": 1.0138188600540161, "Full-finetune/Loss (Raw)": 1.0795097351074219, "Full-finetune/Step": 51, "Full-finetune/Step Time": 7.271105200636621} {"Full-finetune/Learning Rate": 6.70967741935484e-06, "Full-finetune/Loss": 1.013770341873169, "Full-finetune/Loss (Raw)": 1.0112948417663574, "Full-finetune/Step": 52, "Full-finetune/Step Time": 7.263315769342276} {"Full-finetune/Learning Rate": 6.838709677419355e-06, "Full-finetune/Loss": 1.0091780424118042, "Full-finetune/Loss (Raw)": 0.770380437374115, "Full-finetune/Step": 53, "Full-finetune/Step Time": 7.276447943921359} {"Full-finetune/Learning Rate": 6.967741935483871e-06, "Full-finetune/Loss": 1.0061982870101929, "Full-finetune/Loss (Raw)": 0.8482712507247925, "Full-finetune/Step": 54, "Full-finetune/Step Time": 7.293160875638326} {"Full-finetune/Learning Rate": 7.096774193548388e-06, "Full-finetune/Loss": 1.0054984092712402, "Full-finetune/Loss (Raw)": 0.9677059054374695, "Full-finetune/Step": 55, "Full-finetune/Step Time": 7.2942792805758385} {"Full-finetune/Learning Rate": 7.225806451612903e-06, "Full-finetune/Loss": 1.004417061805725, "Full-finetune/Loss (Raw)": 0.9449381828308105, "Full-finetune/Step": 56, "Full-finetune/Step Time": 7.326093903609684} {"Full-finetune/Learning Rate": 7.35483870967742e-06, "Full-finetune/Loss": 1.002458095550537, "Full-finetune/Loss (Raw)": 0.8927604556083679, "Full-finetune/Step": 57, "Full-finetune/Step Time": 7.30268470864547} {"Full-finetune/Learning Rate": 7.483870967741936e-06, "Full-finetune/Loss": 1.0008751153945923, "Full-finetune/Loss (Raw)": 0.9106411337852478, "Full-finetune/Step": 58, "Full-finetune/Step Time": 7.327035953258646} {"Full-finetune/Learning Rate": 7.612903225806451e-06, "Full-finetune/Loss": 0.999348521232605, "Full-finetune/Loss (Raw)": 0.9108092784881592, "Full-finetune/Step": 59, "Full-finetune/Step Time": 7.332933583502042} {"Full-finetune/Learning Rate": 7.741935483870968e-06, "Full-finetune/Loss": 0.9988264441490173, "Full-finetune/Loss (Raw)": 0.9680201411247253, "Full-finetune/Step": 60, "Full-finetune/Step Time": 7.3075037439664206} {"Full-finetune/Learning Rate": 7.870967741935484e-06, "Full-finetune/Loss": 0.999118447303772, "Full-finetune/Loss (Raw)": 1.0166420936584473, "Full-finetune/Step": 61, "Full-finetune/Step Time": 7.2927525512507705} {"Full-finetune/Learning Rate": 8.000000000000001e-06, "Full-finetune/Loss": 0.9998980760574341, "Full-finetune/Loss (Raw)": 1.0474579334259033, "Full-finetune/Step": 62, "Full-finetune/Step Time": 7.266778561376756} {"Full-finetune/Learning Rate": 8.129032258064517e-06, "Full-finetune/Loss": 1.0011004209518433, "Full-finetune/Loss (Raw)": 1.0756367444992065, "Full-finetune/Step": 63, "Full-finetune/Step Time": 7.246067516387455} {"Full-finetune/Learning Rate": 8.258064516129033e-06, "Full-finetune/Loss": 0.998633086681366, "Full-finetune/Loss (Raw)": 0.8431961536407471, "Full-finetune/Step": 64, "Full-finetune/Step Time": 7.278901435434818} {"Full-finetune/Learning Rate": 8.387096774193549e-06, "Full-finetune/Loss": 0.9981203675270081, "Full-finetune/Loss (Raw)": 0.96530681848526, "Full-finetune/Step": 65, "Full-finetune/Step Time": 7.271674688045795} {"Full-finetune/Learning Rate": 8.516129032258065e-06, "Full-finetune/Loss": 0.9955158829689026, "Full-finetune/Loss (Raw)": 0.8262230157852173, "Full-finetune/Step": 66, "Full-finetune/Step Time": 7.256814205285274} {"Full-finetune/Learning Rate": 8.64516129032258e-06, "Full-finetune/Loss": 0.9921262264251709, "Full-finetune/Loss (Raw)": 0.768412172794342, "Full-finetune/Step": 67, "Full-finetune/Step Time": 7.23314786313185} {"Full-finetune/Learning Rate": 8.774193548387098e-06, "Full-finetune/Loss": 0.9920028448104858, "Full-finetune/Loss (Raw)": 0.9837366342544556, "Full-finetune/Step": 68, "Full-finetune/Step Time": 7.230101003366358} {"Full-finetune/Learning Rate": 8.903225806451614e-06, "Full-finetune/Loss": 0.9937366247177124, "Full-finetune/Loss (Raw)": 1.111633539199829, "Full-finetune/Step": 69, "Full-finetune/Step Time": 7.2065410406693164} {"Full-finetune/Learning Rate": 9.03225806451613e-06, "Full-finetune/Loss": 0.995074987411499, "Full-finetune/Loss (Raw)": 1.0874241590499878, "Full-finetune/Step": 70, "Full-finetune/Step Time": 7.186733191353934} {"Full-finetune/Learning Rate": 9.161290322580645e-06, "Full-finetune/Loss": 0.993073582649231, "Full-finetune/Loss (Raw)": 0.8529723286628723, "Full-finetune/Step": 71, "Full-finetune/Step Time": 7.1692207329709765} {"Full-finetune/Learning Rate": 9.290322580645163e-06, "Full-finetune/Loss": 0.9924492239952087, "Full-finetune/Loss (Raw)": 0.9481185674667358, "Full-finetune/Step": 72, "Full-finetune/Step Time": 7.14642596244812} {"Full-finetune/Learning Rate": 9.419354838709677e-06, "Full-finetune/Loss": 0.9904162287712097, "Full-finetune/Loss (Raw)": 0.8440384268760681, "Full-finetune/Step": 73, "Full-finetune/Step Time": 7.1286577394563855} {"Full-finetune/Learning Rate": 9.548387096774195e-06, "Full-finetune/Loss": 0.9880989789962769, "Full-finetune/Loss (Raw)": 0.8189401626586914, "Full-finetune/Step": 74, "Full-finetune/Step Time": 7.112955666877128} {"Full-finetune/Learning Rate": 9.67741935483871e-06, "Full-finetune/Loss": 0.9842184782028198, "Full-finetune/Loss (Raw)": 0.6970617771148682, "Full-finetune/Step": 75, "Full-finetune/Step Time": 7.106799243291219} {"Full-finetune/Learning Rate": 9.806451612903226e-06, "Full-finetune/Loss": 0.982440173625946, "Full-finetune/Loss (Raw)": 0.8490703105926514, "Full-finetune/Step": 76, "Full-finetune/Step Time": 7.100879534294731} {"Full-finetune/Learning Rate": 9.935483870967742e-06, "Full-finetune/Loss": 0.982176661491394, "Full-finetune/Loss (Raw)": 0.9621464610099792, "Full-finetune/Step": 77, "Full-finetune/Step Time": 7.079773014241999} {"Full-finetune/Learning Rate": 1.0064516129032258e-05, "Full-finetune/Loss": 0.9834824800491333, "Full-finetune/Loss (Raw)": 1.0840355157852173, "Full-finetune/Step": 78, "Full-finetune/Step Time": 7.060343565084995} {"Full-finetune/Learning Rate": 1.0193548387096774e-05, "Full-finetune/Loss": 0.9849904775619507, "Full-finetune/Loss (Raw)": 1.1026101112365723, "Full-finetune/Step": 79, "Full-finetune/Step Time": 7.045474523230444} {"Full-finetune/Learning Rate": 1.0322580645161291e-05, "Full-finetune/Loss": 0.9838593602180481, "Full-finetune/Loss (Raw)": 0.8945001363754272, "Full-finetune/Step": 80, "Full-finetune/Step Time": 7.051542875170708} {"Full-finetune/Learning Rate": 1.0451612903225807e-05, "Full-finetune/Loss": 0.9836124181747437, "Full-finetune/Loss (Raw)": 0.9638618230819702, "Full-finetune/Step": 81, "Full-finetune/Step Time": 7.049215113675153} {"Full-finetune/Learning Rate": 1.0580645161290325e-05, "Full-finetune/Loss": 0.9836103320121765, "Full-finetune/Loss (Raw)": 0.9834486246109009, "Full-finetune/Step": 82, "Full-finetune/Step Time": 7.033918511576768} {"Full-finetune/Learning Rate": 1.070967741935484e-05, "Full-finetune/Loss": 0.9815067052841187, "Full-finetune/Loss (Raw)": 0.8090009689331055, "Full-finetune/Step": 83, "Full-finetune/Step Time": 7.042287654187306} {"Full-finetune/Learning Rate": 1.0838709677419356e-05, "Full-finetune/Loss": 0.9819548726081848, "Full-finetune/Loss (Raw)": 1.0191444158554077, "Full-finetune/Step": 84, "Full-finetune/Step Time": 7.044385274251302} {"Full-finetune/Learning Rate": 1.096774193548387e-05, "Full-finetune/Loss": 0.9816893935203552, "Full-finetune/Loss (Raw)": 0.9593958854675293, "Full-finetune/Step": 85, "Full-finetune/Step Time": 7.031096853929408} {"Full-finetune/Learning Rate": 1.1096774193548388e-05, "Full-finetune/Loss": 0.9738984107971191, "Full-finetune/Loss (Raw)": 0.3116708993911743, "Full-finetune/Step": 86, "Full-finetune/Step Time": 7.044165636217872} {"Full-finetune/Learning Rate": 1.1225806451612904e-05, "Full-finetune/Loss": 0.972963809967041, "Full-finetune/Loss (Raw)": 0.8925825357437134, "Full-finetune/Step": 87, "Full-finetune/Step Time": 7.033307086462262} {"Full-finetune/Learning Rate": 1.1354838709677421e-05, "Full-finetune/Loss": 0.974043607711792, "Full-finetune/Loss (Raw)": 1.0679893493652344, "Full-finetune/Step": 88, "Full-finetune/Step Time": 7.041798743334684} {"Full-finetune/Learning Rate": 1.1483870967741937e-05, "Full-finetune/Loss": 0.973778486251831, "Full-finetune/Loss (Raw)": 0.9504423141479492, "Full-finetune/Step": 89, "Full-finetune/Step Time": 7.029265004597353} {"Full-finetune/Learning Rate": 1.1612903225806453e-05, "Full-finetune/Loss": 0.9743667840957642, "Full-finetune/Loss (Raw)": 1.0267212390899658, "Full-finetune/Step": 90, "Full-finetune/Step Time": 7.037873387336731} {"Full-finetune/Learning Rate": 1.1741935483870967e-05, "Full-finetune/Loss": 0.974793016910553, "Full-finetune/Loss (Raw)": 1.0131582021713257, "Full-finetune/Step": 91, "Full-finetune/Step Time": 7.040951723580832} {"Full-finetune/Learning Rate": 1.1870967741935484e-05, "Full-finetune/Loss": 0.9765245318412781, "Full-finetune/Loss (Raw)": 1.1340969800949097, "Full-finetune/Step": 92, "Full-finetune/Step Time": 7.030435683934585} {"Full-finetune/Learning Rate": 1.2e-05, "Full-finetune/Loss": 0.9749904870986938, "Full-finetune/Loss (Raw)": 0.833860456943512, "Full-finetune/Step": 93, "Full-finetune/Step Time": 7.023786675545477} {"Full-finetune/Learning Rate": 1.2129032258064518e-05, "Full-finetune/Loss": 0.9757617712020874, "Full-finetune/Loss (Raw)": 1.0474916696548462, "Full-finetune/Step": 94, "Full-finetune/Step Time": 7.00760069806525} {"Full-finetune/Learning Rate": 1.2258064516129034e-05, "Full-finetune/Loss": 0.9754664897918701, "Full-finetune/Loss (Raw)": 0.947699785232544, "Full-finetune/Step": 95, "Full-finetune/Step Time": 6.996767392911409} {"Full-finetune/Learning Rate": 1.238709677419355e-05, "Full-finetune/Loss": 0.9750109910964966, "Full-finetune/Loss (Raw)": 0.9317392110824585, "Full-finetune/Step": 96, "Full-finetune/Step Time": 7.000046956042449} {"Full-finetune/Learning Rate": 1.2516129032258067e-05, "Full-finetune/Loss": 0.9745097160339355, "Full-finetune/Loss (Raw)": 0.9263947010040283, "Full-finetune/Step": 97, "Full-finetune/Step Time": 6.996707495954848} {"Full-finetune/Learning Rate": 1.2645161290322581e-05, "Full-finetune/Loss": 0.9748122096061707, "Full-finetune/Loss (Raw)": 1.0041544437408447, "Full-finetune/Step": 98, "Full-finetune/Step Time": 6.985527092096757} {"Full-finetune/Learning Rate": 1.2774193548387097e-05, "Full-finetune/Loss": 0.9668284058570862, "Full-finetune/Loss (Raw)": 0.18441376090049744, "Full-finetune/Step": 99, "Full-finetune/Step Time": 6.995671065166743} {"Full-finetune/Learning Rate": 1.2903225806451613e-05, "Full-finetune/Loss": 0.9684154391288757, "Full-finetune/Loss (Raw)": 1.125534176826477, "Full-finetune/Step": 100, "Full-finetune/Step Time": 6.991500644683838} {"Full-finetune/Learning Rate": 1.303225806451613e-05, "Full-finetune/Loss": 0.9676183462142944, "Full-finetune/Loss (Raw)": 0.8879057168960571, "Full-finetune/Step": 101, "Full-finetune/Step Time": 6.993475392313287} {"Full-finetune/Learning Rate": 1.3161290322580646e-05, "Full-finetune/Loss": 0.969302773475647, "Full-finetune/Loss (Raw)": 1.1394262313842773, "Full-finetune/Step": 102, "Full-finetune/Step Time": 6.977866413546543} {"Full-finetune/Learning Rate": 1.3290322580645164e-05, "Full-finetune/Loss": 0.9625870585441589, "Full-finetune/Loss (Raw)": 0.27759358286857605, "Full-finetune/Step": 103, "Full-finetune/Step Time": 6.989318546739597} {"Full-finetune/Learning Rate": 1.341935483870968e-05, "Full-finetune/Loss": 0.963596522808075, "Full-finetune/Loss (Raw)": 1.0675575733184814, "Full-finetune/Step": 104, "Full-finetune/Step Time": 6.978739050718454} {"Full-finetune/Learning Rate": 1.3548387096774194e-05, "Full-finetune/Loss": 0.9634713530540466, "Full-finetune/Loss (Raw)": 0.9504600763320923, "Full-finetune/Step": 105, "Full-finetune/Step Time": 6.963155714670817} {"Full-finetune/Learning Rate": 1.367741935483871e-05, "Full-finetune/Loss": 0.9634217023849487, "Full-finetune/Loss (Raw)": 0.9582082033157349, "Full-finetune/Step": 106, "Full-finetune/Step Time": 6.9520741791095375} {"Full-finetune/Learning Rate": 1.3806451612903227e-05, "Full-finetune/Loss": 0.9638165831565857, "Full-finetune/Loss (Raw)": 1.0056811571121216, "Full-finetune/Step": 107, "Full-finetune/Step Time": 6.942386313019512} {"Full-finetune/Learning Rate": 1.3935483870967743e-05, "Full-finetune/Loss": 0.963653564453125, "Full-finetune/Loss (Raw)": 0.9462096095085144, "Full-finetune/Step": 108, "Full-finetune/Step Time": 6.946097482133795} {"Full-finetune/Learning Rate": 1.406451612903226e-05, "Full-finetune/Loss": 0.9639303088188171, "Full-finetune/Loss (Raw)": 0.9938147664070129, "Full-finetune/Step": 109, "Full-finetune/Step Time": 6.93528139919316} {"Full-finetune/Learning Rate": 1.4193548387096776e-05, "Full-finetune/Loss": 0.9638679027557373, "Full-finetune/Loss (Raw)": 0.957071840763092, "Full-finetune/Step": 110, "Full-finetune/Step Time": 6.923566454107111} {"Full-finetune/Learning Rate": 1.4322580645161292e-05, "Full-finetune/Loss": 0.963413417339325, "Full-finetune/Loss (Raw)": 0.9134123921394348, "Full-finetune/Step": 111, "Full-finetune/Step Time": 6.9293148367254584} {"Full-finetune/Learning Rate": 1.4451612903225806e-05, "Full-finetune/Loss": 0.963842511177063, "Full-finetune/Loss (Raw)": 1.0114660263061523, "Full-finetune/Step": 112, "Full-finetune/Step Time": 6.93735426238605} {"Full-finetune/Learning Rate": 1.4580645161290324e-05, "Full-finetune/Loss": 0.9639149308204651, "Full-finetune/Loss (Raw)": 0.9720304012298584, "Full-finetune/Step": 113, "Full-finetune/Step Time": 6.939865838109919} {"Full-finetune/Learning Rate": 1.470967741935484e-05, "Full-finetune/Loss": 0.9574784636497498, "Full-finetune/Loss (Raw)": 0.23015490174293518, "Full-finetune/Step": 114, "Full-finetune/Step Time": 6.959814546401041} {"Full-finetune/Learning Rate": 1.4838709677419357e-05, "Full-finetune/Loss": 0.9578284025192261, "Full-finetune/Loss (Raw)": 0.9977245330810547, "Full-finetune/Step": 115, "Full-finetune/Step Time": 6.963750105318816} {"Full-finetune/Learning Rate": 1.4967741935483873e-05, "Full-finetune/Loss": 0.9563643336296082, "Full-finetune/Loss (Raw)": 0.7879948019981384, "Full-finetune/Step": 116, "Full-finetune/Step Time": 6.968543032120014} {"Full-finetune/Learning Rate": 1.5096774193548389e-05, "Full-finetune/Loss": 0.9568753242492676, "Full-finetune/Loss (Raw)": 1.0161470174789429, "Full-finetune/Step": 117, "Full-finetune/Step Time": 6.9634617495740585} {"Full-finetune/Learning Rate": 1.5225806451612903e-05, "Full-finetune/Loss": 0.9576597213745117, "Full-finetune/Loss (Raw)": 1.0494389533996582, "Full-finetune/Step": 118, "Full-finetune/Step Time": 6.952924411175615} {"Full-finetune/Learning Rate": 1.535483870967742e-05, "Full-finetune/Loss": 0.9575682878494263, "Full-finetune/Loss (Raw)": 0.9467727541923523, "Full-finetune/Step": 119, "Full-finetune/Step Time": 6.9441152821068} {"Full-finetune/Learning Rate": 1.5483870967741936e-05, "Full-finetune/Loss": 0.957754373550415, "Full-finetune/Loss (Raw)": 0.9799000024795532, "Full-finetune/Step": 120, "Full-finetune/Step Time": 6.940392043193182} {"Full-finetune/Learning Rate": 1.5612903225806454e-05, "Full-finetune/Loss": 0.9576630592346191, "Full-finetune/Loss (Raw)": 0.9467158317565918, "Full-finetune/Step": 121, "Full-finetune/Step Time": 6.949061332655347} {"Full-finetune/Learning Rate": 1.5741935483870968e-05, "Full-finetune/Loss": 0.9571056365966797, "Full-finetune/Loss (Raw)": 0.8896539211273193, "Full-finetune/Step": 122, "Full-finetune/Step Time": 6.939542678535962} {"Full-finetune/Learning Rate": 1.5870967741935485e-05, "Full-finetune/Loss": 0.9570546746253967, "Full-finetune/Loss (Raw)": 0.9508354663848877, "Full-finetune/Step": 123, "Full-finetune/Step Time": 6.93090089356027} {"Full-finetune/Learning Rate": 1.6000000000000003e-05, "Full-finetune/Loss": 0.9573003649711609, "Full-finetune/Loss (Raw)": 0.9875242710113525, "Full-finetune/Step": 124, "Full-finetune/Step Time": 6.924705153511416} {"Full-finetune/Learning Rate": 1.6129032258064517e-05, "Full-finetune/Loss": 0.9567270874977112, "Full-finetune/Loss (Raw)": 0.8856337070465088, "Full-finetune/Step": 125, "Full-finetune/Step Time": 6.918248474121094} {"Full-finetune/Learning Rate": 1.6258064516129034e-05, "Full-finetune/Loss": 0.95702064037323, "Full-finetune/Loss (Raw)": 0.9937176704406738, "Full-finetune/Step": 126, "Full-finetune/Step Time": 6.929446434217786} {"Full-finetune/Learning Rate": 1.638709677419355e-05, "Full-finetune/Loss": 0.9564632177352905, "Full-finetune/Loss (Raw)": 0.8862341642379761, "Full-finetune/Step": 127, "Full-finetune/Step Time": 6.919635218898143} {"Full-finetune/Learning Rate": 1.6516129032258066e-05, "Full-finetune/Loss": 0.9570351243019104, "Full-finetune/Loss (Raw)": 1.0296612977981567, "Full-finetune/Step": 128, "Full-finetune/Step Time": 6.9126957803964615} {"Full-finetune/Learning Rate": 1.6645161290322583e-05, "Full-finetune/Loss": 0.9565653800964355, "Full-finetune/Loss (Raw)": 1.014727234840393, "Full-finetune/Step": 129, "Full-finetune/Step Time": 6.764907423406839} {"Full-finetune/Learning Rate": 1.6774193548387098e-05, "Full-finetune/Loss": 0.9539847373962402, "Full-finetune/Loss (Raw)": 1.0325212478637695, "Full-finetune/Step": 130, "Full-finetune/Step Time": 6.746865669265389} {"Full-finetune/Learning Rate": 1.6903225806451615e-05, "Full-finetune/Loss": 0.9531371593475342, "Full-finetune/Loss (Raw)": 1.023948311805725, "Full-finetune/Step": 131, "Full-finetune/Step Time": 6.746084110811353} {"Full-finetune/Learning Rate": 1.703225806451613e-05, "Full-finetune/Loss": 0.9526897072792053, "Full-finetune/Loss (Raw)": 0.9984552264213562, "Full-finetune/Step": 132, "Full-finetune/Step Time": 6.74577914737165} {"Full-finetune/Learning Rate": 1.7161290322580647e-05, "Full-finetune/Loss": 0.9498162269592285, "Full-finetune/Loss (Raw)": 0.7614976167678833, "Full-finetune/Step": 133, "Full-finetune/Step Time": 6.754295429214835} {"Full-finetune/Learning Rate": 1.729032258064516e-05, "Full-finetune/Loss": 0.9497991800308228, "Full-finetune/Loss (Raw)": 1.1073938608169556, "Full-finetune/Step": 134, "Full-finetune/Step Time": 6.750473467633128} {"Full-finetune/Learning Rate": 1.741935483870968e-05, "Full-finetune/Loss": 0.948068380355835, "Full-finetune/Loss (Raw)": 0.9062848687171936, "Full-finetune/Step": 135, "Full-finetune/Step Time": 6.751739677041769} {"Full-finetune/Learning Rate": 1.7548387096774196e-05, "Full-finetune/Loss": 0.945603609085083, "Full-finetune/Loss (Raw)": 0.9440091848373413, "Full-finetune/Step": 136, "Full-finetune/Step Time": 6.737373588606715} {"Full-finetune/Learning Rate": 1.7677419354838713e-05, "Full-finetune/Loss": 0.944964587688446, "Full-finetune/Loss (Raw)": 0.8801961541175842, "Full-finetune/Step": 137, "Full-finetune/Step Time": 6.738711144775152} {"Full-finetune/Learning Rate": 1.7806451612903228e-05, "Full-finetune/Loss": 0.9420092701911926, "Full-finetune/Loss (Raw)": 0.9322510361671448, "Full-finetune/Step": 138, "Full-finetune/Step Time": 6.711120467633009} {"Full-finetune/Learning Rate": 1.7935483870967742e-05, "Full-finetune/Loss": 0.9387122988700867, "Full-finetune/Loss (Raw)": 0.8339397311210632, "Full-finetune/Step": 139, "Full-finetune/Step Time": 6.712137499824166} {"Full-finetune/Learning Rate": 1.806451612903226e-05, "Full-finetune/Loss": 0.9393205642700195, "Full-finetune/Loss (Raw)": 1.2115103006362915, "Full-finetune/Step": 140, "Full-finetune/Step Time": 6.70636759698391} {"Full-finetune/Learning Rate": 1.8193548387096777e-05, "Full-finetune/Loss": 0.9381213188171387, "Full-finetune/Loss (Raw)": 0.9404536485671997, "Full-finetune/Step": 141, "Full-finetune/Step Time": 6.699104169383645} {"Full-finetune/Learning Rate": 1.832258064516129e-05, "Full-finetune/Loss": 0.9348430633544922, "Full-finetune/Loss (Raw)": 0.8892448544502258, "Full-finetune/Step": 142, "Full-finetune/Step Time": 6.738710146397352} {"Full-finetune/Learning Rate": 1.845161290322581e-05, "Full-finetune/Loss": 0.9317542314529419, "Full-finetune/Loss (Raw)": 0.8246999382972717, "Full-finetune/Step": 143, "Full-finetune/Step Time": 6.752726452425122} {"Full-finetune/Learning Rate": 1.8580645161290326e-05, "Full-finetune/Loss": 0.9277099370956421, "Full-finetune/Loss (Raw)": 0.3294568359851837, "Full-finetune/Step": 144, "Full-finetune/Step Time": 6.762337855994701} {"Full-finetune/Learning Rate": 1.870967741935484e-05, "Full-finetune/Loss": 0.921695351600647, "Full-finetune/Loss (Raw)": 0.32109296321868896, "Full-finetune/Step": 145, "Full-finetune/Step Time": 6.784402972087264} {"Full-finetune/Learning Rate": 1.8838709677419354e-05, "Full-finetune/Loss": 0.9215710759162903, "Full-finetune/Loss (Raw)": 0.890320360660553, "Full-finetune/Step": 146, "Full-finetune/Step Time": 6.776618713513017} {"Full-finetune/Learning Rate": 1.896774193548387e-05, "Full-finetune/Loss": 0.9193120002746582, "Full-finetune/Loss (Raw)": 0.8041523694992065, "Full-finetune/Step": 147, "Full-finetune/Step Time": 6.7826149463653564} {"Full-finetune/Learning Rate": 1.909677419354839e-05, "Full-finetune/Loss": 0.918674111366272, "Full-finetune/Loss (Raw)": 0.9723562598228455, "Full-finetune/Step": 148, "Full-finetune/Step Time": 6.785621227696538} {"Full-finetune/Learning Rate": 1.9225806451612907e-05, "Full-finetune/Loss": 0.9192275404930115, "Full-finetune/Loss (Raw)": 1.0365084409713745, "Full-finetune/Step": 149, "Full-finetune/Step Time": 6.775341242551804} {"Full-finetune/Learning Rate": 1.935483870967742e-05, "Full-finetune/Loss": 0.9190558791160583, "Full-finetune/Loss (Raw)": 0.9817941784858704, "Full-finetune/Step": 150, "Full-finetune/Step Time": 6.77921224758029} {"Full-finetune/Learning Rate": 1.948387096774194e-05, "Full-finetune/Loss": 0.924938976764679, "Full-finetune/Loss (Raw)": 0.982559323310852, "Full-finetune/Step": 151, "Full-finetune/Step Time": 6.759449653327465} {"Full-finetune/Learning Rate": 1.9612903225806452e-05, "Full-finetune/Loss": 0.9237385392189026, "Full-finetune/Loss (Raw)": 0.9753028750419617, "Full-finetune/Step": 152, "Full-finetune/Step Time": 6.763370027765632} {"Full-finetune/Learning Rate": 1.974193548387097e-05, "Full-finetune/Loss": 0.924388587474823, "Full-finetune/Loss (Raw)": 1.0215953588485718, "Full-finetune/Step": 153, "Full-finetune/Step Time": 6.773370580747724} {"Full-finetune/Learning Rate": 1.9870967741935484e-05, "Full-finetune/Loss": 0.9302299618721008, "Full-finetune/Loss (Raw)": 1.0192649364471436, "Full-finetune/Step": 154, "Full-finetune/Step Time": 6.747875854372978} {"Full-finetune/Learning Rate": 2e-05, "Full-finetune/Loss": 0.9296713471412659, "Full-finetune/Loss (Raw)": 1.086848258972168, "Full-finetune/Step": 155, "Full-finetune/Step Time": 6.755222564563155} {"Full-finetune/Learning Rate": 1.9999998059597617e-05, "Full-finetune/Loss": 0.9241287708282471, "Full-finetune/Loss (Raw)": 0.32325318455696106, "Full-finetune/Step": 156, "Full-finetune/Step Time": 6.775068501010537} {"Full-finetune/Learning Rate": 1.9999992238391222e-05, "Full-finetune/Loss": 0.923346221446991, "Full-finetune/Loss (Raw)": 0.9868060350418091, "Full-finetune/Step": 157, "Full-finetune/Step Time": 6.766845853999257} {"Full-finetune/Learning Rate": 1.999998253638307e-05, "Full-finetune/Loss": 0.9236559867858887, "Full-finetune/Loss (Raw)": 0.9909191727638245, "Full-finetune/Step": 158, "Full-finetune/Step Time": 6.758726852014661} {"Full-finetune/Learning Rate": 1.999996895357693e-05, "Full-finetune/Loss": 0.9229277968406677, "Full-finetune/Loss (Raw)": 0.9630942940711975, "Full-finetune/Step": 159, "Full-finetune/Step Time": 6.750818375498056} {"Full-finetune/Learning Rate": 1.9999951489978074e-05, "Full-finetune/Loss": 0.9226013422012329, "Full-finetune/Loss (Raw)": 0.8774774074554443, "Full-finetune/Step": 160, "Full-finetune/Step Time": 6.739903051406145} {"Full-finetune/Learning Rate": 1.9999930145593275e-05, "Full-finetune/Loss": 0.9233502149581909, "Full-finetune/Loss (Raw)": 0.904670238494873, "Full-finetune/Step": 161, "Full-finetune/Step Time": 6.735902845859528} {"Full-finetune/Learning Rate": 1.9999904920430818e-05, "Full-finetune/Loss": 0.9247357845306396, "Full-finetune/Loss (Raw)": 1.083476185798645, "Full-finetune/Step": 162, "Full-finetune/Step Time": 6.733394036069512} {"Full-finetune/Learning Rate": 1.9999875814500492e-05, "Full-finetune/Loss": 0.9235052466392517, "Full-finetune/Loss (Raw)": 0.9882760643959045, "Full-finetune/Step": 163, "Full-finetune/Step Time": 6.73304270580411} {"Full-finetune/Learning Rate": 1.99998428278136e-05, "Full-finetune/Loss": 0.9220646619796753, "Full-finetune/Loss (Raw)": 0.8085857629776001, "Full-finetune/Step": 164, "Full-finetune/Step Time": 6.729569127783179} {"Full-finetune/Learning Rate": 1.9999805960382928e-05, "Full-finetune/Loss": 0.9212220311164856, "Full-finetune/Loss (Raw)": 0.9718731045722961, "Full-finetune/Step": 165, "Full-finetune/Step Time": 6.70564348064363} {"Full-finetune/Learning Rate": 1.99997652122228e-05, "Full-finetune/Loss": 0.919964611530304, "Full-finetune/Loss (Raw)": 0.855783224105835, "Full-finetune/Step": 166, "Full-finetune/Step Time": 6.716977460309863} {"Full-finetune/Learning Rate": 1.9999720583349015e-05, "Full-finetune/Loss": 0.9207637906074524, "Full-finetune/Loss (Raw)": 1.0611670017242432, "Full-finetune/Step": 167, "Full-finetune/Step Time": 6.727708661928773} {"Full-finetune/Learning Rate": 1.9999672073778905e-05, "Full-finetune/Loss": 0.9197950959205627, "Full-finetune/Loss (Raw)": 0.8919165730476379, "Full-finetune/Step": 168, "Full-finetune/Step Time": 6.727622048929334} {"Full-finetune/Learning Rate": 1.999961968353129e-05, "Full-finetune/Loss": 0.9180228114128113, "Full-finetune/Loss (Raw)": 0.8940050601959229, "Full-finetune/Step": 169, "Full-finetune/Step Time": 6.731695141643286} {"Full-finetune/Learning Rate": 1.9999563412626503e-05, "Full-finetune/Loss": 0.9120199680328369, "Full-finetune/Loss (Raw)": 0.35672083497047424, "Full-finetune/Step": 170, "Full-finetune/Step Time": 6.748176135122776} {"Full-finetune/Learning Rate": 1.9999503261086378e-05, "Full-finetune/Loss": 0.91151362657547, "Full-finetune/Loss (Raw)": 0.9245266318321228, "Full-finetune/Step": 171, "Full-finetune/Step Time": 6.7466455809772015} {"Full-finetune/Learning Rate": 1.999943922893426e-05, "Full-finetune/Loss": 0.9099066257476807, "Full-finetune/Loss (Raw)": 0.8484889268875122, "Full-finetune/Step": 172, "Full-finetune/Step Time": 6.748537803068757} {"Full-finetune/Learning Rate": 1.9999371316195005e-05, "Full-finetune/Loss": 0.9113742113113403, "Full-finetune/Loss (Raw)": 1.085973858833313, "Full-finetune/Step": 173, "Full-finetune/Step Time": 6.716775711625814} {"Full-finetune/Learning Rate": 1.9999299522894957e-05, "Full-finetune/Loss": 0.9157512187957764, "Full-finetune/Loss (Raw)": 0.8136558532714844, "Full-finetune/Step": 174, "Full-finetune/Step Time": 6.692830469459295} {"Full-finetune/Learning Rate": 1.9999223849061987e-05, "Full-finetune/Loss": 0.9158567190170288, "Full-finetune/Loss (Raw)": 1.051636815071106, "Full-finetune/Step": 175, "Full-finetune/Step Time": 6.687473848462105} {"Full-finetune/Learning Rate": 1.999914429472546e-05, "Full-finetune/Loss": 0.9111005067825317, "Full-finetune/Loss (Raw)": 0.3516935408115387, "Full-finetune/Step": 176, "Full-finetune/Step Time": 6.670448558405042} {"Full-finetune/Learning Rate": 1.9999060859916247e-05, "Full-finetune/Loss": 0.9097375869750977, "Full-finetune/Loss (Raw)": 0.8819300532341003, "Full-finetune/Step": 177, "Full-finetune/Step Time": 6.673865446820855} {"Full-finetune/Learning Rate": 1.9998973544666734e-05, "Full-finetune/Loss": 0.9109088182449341, "Full-finetune/Loss (Raw)": 1.0827674865722656, "Full-finetune/Step": 178, "Full-finetune/Step Time": 6.680773597210646} {"Full-finetune/Learning Rate": 1.9998882349010798e-05, "Full-finetune/Loss": 0.908954381942749, "Full-finetune/Loss (Raw)": 0.8293370604515076, "Full-finetune/Step": 179, "Full-finetune/Step Time": 6.684984304010868} {"Full-finetune/Learning Rate": 1.999878727298384e-05, "Full-finetune/Loss": 0.9084576964378357, "Full-finetune/Loss (Raw)": 0.9477265477180481, "Full-finetune/Step": 180, "Full-finetune/Step Time": 6.680968303233385} {"Full-finetune/Learning Rate": 1.999868831662274e-05, "Full-finetune/Loss": 0.9098941087722778, "Full-finetune/Loss (Raw)": 0.954241931438446, "Full-finetune/Step": 181, "Full-finetune/Step Time": 6.661890517920256} {"Full-finetune/Learning Rate": 1.9998585479965918e-05, "Full-finetune/Loss": 0.9108947515487671, "Full-finetune/Loss (Raw)": 0.9763504862785339, "Full-finetune/Step": 182, "Full-finetune/Step Time": 6.645757479593158} {"Full-finetune/Learning Rate": 1.999847876305328e-05, "Full-finetune/Loss": 0.9107030630111694, "Full-finetune/Loss (Raw)": 0.9431716799736023, "Full-finetune/Step": 183, "Full-finetune/Step Time": 6.6292021702975035} {"Full-finetune/Learning Rate": 1.9998368165926237e-05, "Full-finetune/Loss": 0.9117724895477295, "Full-finetune/Loss (Raw)": 1.0818276405334473, "Full-finetune/Step": 184, "Full-finetune/Step Time": 6.606178872287273} {"Full-finetune/Learning Rate": 1.9998253688627706e-05, "Full-finetune/Loss": 0.9115622043609619, "Full-finetune/Loss (Raw)": 0.8658368587493896, "Full-finetune/Step": 185, "Full-finetune/Step Time": 6.604103999212384} {"Full-finetune/Learning Rate": 1.9998135331202124e-05, "Full-finetune/Loss": 0.9105817079544067, "Full-finetune/Loss (Raw)": 0.7851399183273315, "Full-finetune/Step": 186, "Full-finetune/Step Time": 6.595173871144652} {"Full-finetune/Learning Rate": 1.9998013093695413e-05, "Full-finetune/Loss": 0.9105815887451172, "Full-finetune/Loss (Raw)": 0.9107906818389893, "Full-finetune/Step": 187, "Full-finetune/Step Time": 6.582041436806321} {"Full-finetune/Learning Rate": 1.9997886976155018e-05, "Full-finetune/Loss": 0.9107246398925781, "Full-finetune/Loss (Raw)": 0.9863322377204895, "Full-finetune/Step": 188, "Full-finetune/Step Time": 6.5811746548861265} {"Full-finetune/Learning Rate": 1.9997756978629876e-05, "Full-finetune/Loss": 0.9096635580062866, "Full-finetune/Loss (Raw)": 0.8808251023292542, "Full-finetune/Step": 189, "Full-finetune/Step Time": 6.591538244858384} {"Full-finetune/Learning Rate": 1.999762310117044e-05, "Full-finetune/Loss": 0.9090718030929565, "Full-finetune/Loss (Raw)": 0.9717198610305786, "Full-finetune/Step": 190, "Full-finetune/Step Time": 6.598607322201133} {"Full-finetune/Learning Rate": 1.999748534382867e-05, "Full-finetune/Loss": 0.907798707485199, "Full-finetune/Loss (Raw)": 0.9126778244972229, "Full-finetune/Step": 191, "Full-finetune/Step Time": 6.603796947747469} {"Full-finetune/Learning Rate": 1.9997343706658023e-05, "Full-finetune/Loss": 0.9087910056114197, "Full-finetune/Loss (Raw)": 0.9702120423316956, "Full-finetune/Step": 192, "Full-finetune/Step Time": 6.584450040012598} {"Full-finetune/Learning Rate": 1.9997198189713462e-05, "Full-finetune/Loss": 0.907927930355072, "Full-finetune/Loss (Raw)": 0.8548282384872437, "Full-finetune/Step": 193, "Full-finetune/Step Time": 6.579204671084881} {"Full-finetune/Learning Rate": 1.9997048793051463e-05, "Full-finetune/Loss": 0.9094301462173462, "Full-finetune/Loss (Raw)": 1.0185061693191528, "Full-finetune/Step": 194, "Full-finetune/Step Time": 6.578965630382299} {"Full-finetune/Learning Rate": 1.9996895516730005e-05, "Full-finetune/Loss": 0.9104863405227661, "Full-finetune/Loss (Raw)": 0.9036012887954712, "Full-finetune/Step": 195, "Full-finetune/Step Time": 6.585415288805962} {"Full-finetune/Learning Rate": 1.9996738360808566e-05, "Full-finetune/Loss": 0.9091886281967163, "Full-finetune/Loss (Raw)": 0.8176345229148865, "Full-finetune/Step": 196, "Full-finetune/Step Time": 6.581457952037454} {"Full-finetune/Learning Rate": 1.999657732534814e-05, "Full-finetune/Loss": 0.9082285165786743, "Full-finetune/Loss (Raw)": 0.9887439012527466, "Full-finetune/Step": 197, "Full-finetune/Step Time": 6.588599069043994} {"Full-finetune/Learning Rate": 1.9996412410411224e-05, "Full-finetune/Loss": 0.906355619430542, "Full-finetune/Loss (Raw)": 0.8476905226707458, "Full-finetune/Step": 198, "Full-finetune/Step Time": 6.588031576946378} {"Full-finetune/Learning Rate": 1.9996243616061812e-05, "Full-finetune/Loss": 0.9069034457206726, "Full-finetune/Loss (Raw)": 0.9230934977531433, "Full-finetune/Step": 199, "Full-finetune/Step Time": 6.593569101765752} {"Full-finetune/Learning Rate": 1.9996070942365416e-05, "Full-finetune/Loss": 0.9069085717201233, "Full-finetune/Loss (Raw)": 0.9487758278846741, "Full-finetune/Step": 200, "Full-finetune/Step Time": 6.597384126856923} {"Full-finetune/Learning Rate": 1.9995894389389038e-05, "Full-finetune/Loss": 0.9080861806869507, "Full-finetune/Loss (Raw)": 0.9947690963745117, "Full-finetune/Step": 201, "Full-finetune/Step Time": 6.602986466139555} {"Full-finetune/Learning Rate": 1.9995713957201208e-05, "Full-finetune/Loss": 0.9099019169807434, "Full-finetune/Loss (Raw)": 1.0513546466827393, "Full-finetune/Step": 202, "Full-finetune/Step Time": 6.605599544942379} {"Full-finetune/Learning Rate": 1.9995529645871938e-05, "Full-finetune/Loss": 0.912196159362793, "Full-finetune/Loss (Raw)": 0.9907227158546448, "Full-finetune/Step": 203, "Full-finetune/Step Time": 6.601684084162116} {"Full-finetune/Learning Rate": 1.999534145547276e-05, "Full-finetune/Loss": 0.9133191108703613, "Full-finetune/Loss (Raw)": 0.9928078651428223, "Full-finetune/Step": 204, "Full-finetune/Step Time": 6.595645699650049} {"Full-finetune/Learning Rate": 1.99951493860767e-05, "Full-finetune/Loss": 0.9082553386688232, "Full-finetune/Loss (Raw)": 0.31399455666542053, "Full-finetune/Step": 205, "Full-finetune/Step Time": 6.617781225591898} {"Full-finetune/Learning Rate": 1.9994953437758305e-05, "Full-finetune/Loss": 0.9073164463043213, "Full-finetune/Loss (Raw)": 0.9638490080833435, "Full-finetune/Step": 206, "Full-finetune/Step Time": 6.629290007054806} {"Full-finetune/Learning Rate": 1.999475361059362e-05, "Full-finetune/Loss": 0.9053667783737183, "Full-finetune/Loss (Raw)": 0.8530582189559937, "Full-finetune/Step": 207, "Full-finetune/Step Time": 6.6520010605454445} {"Full-finetune/Learning Rate": 1.9994549904660187e-05, "Full-finetune/Loss": 0.9052721261978149, "Full-finetune/Loss (Raw)": 0.8823841214179993, "Full-finetune/Step": 208, "Full-finetune/Step Time": 6.637942833825946} {"Full-finetune/Learning Rate": 1.9994342320037065e-05, "Full-finetune/Loss": 0.9051157236099243, "Full-finetune/Loss (Raw)": 0.9438334703445435, "Full-finetune/Step": 209, "Full-finetune/Step Time": 6.632992392405868} {"Full-finetune/Learning Rate": 1.9994130856804808e-05, "Full-finetune/Loss": 0.8998560905456543, "Full-finetune/Loss (Raw)": 0.3102196753025055, "Full-finetune/Step": 210, "Full-finetune/Step Time": 6.655909547582269} {"Full-finetune/Learning Rate": 1.9993915515045485e-05, "Full-finetune/Loss": 0.8998896479606628, "Full-finetune/Loss (Raw)": 0.8132931590080261, "Full-finetune/Step": 211, "Full-finetune/Step Time": 6.6504122242331505} {"Full-finetune/Learning Rate": 1.999369629484267e-05, "Full-finetune/Loss": 0.8992623686790466, "Full-finetune/Loss (Raw)": 0.9388539791107178, "Full-finetune/Step": 212, "Full-finetune/Step Time": 6.6483092196285725} {"Full-finetune/Learning Rate": 1.999347319628143e-05, "Full-finetune/Loss": 0.900144636631012, "Full-finetune/Loss (Raw)": 1.0723289251327515, "Full-finetune/Step": 213, "Full-finetune/Step Time": 6.651337970048189} {"Full-finetune/Learning Rate": 1.9993246219448347e-05, "Full-finetune/Loss": 0.9045109152793884, "Full-finetune/Loss (Raw)": 0.8705553412437439, "Full-finetune/Step": 214, "Full-finetune/Step Time": 6.651495469734073} {"Full-finetune/Learning Rate": 1.999301536443151e-05, "Full-finetune/Loss": 0.9044420719146729, "Full-finetune/Loss (Raw)": 0.8837732076644897, "Full-finetune/Step": 215, "Full-finetune/Step Time": 6.650199860334396} {"Full-finetune/Learning Rate": 1.9992780631320508e-05, "Full-finetune/Loss": 0.9026119709014893, "Full-finetune/Loss (Raw)": 0.833727240562439, "Full-finetune/Step": 216, "Full-finetune/Step Time": 6.648782709613442} {"Full-finetune/Learning Rate": 1.9992542020206432e-05, "Full-finetune/Loss": 0.9028927087783813, "Full-finetune/Loss (Raw)": 0.9863842725753784, "Full-finetune/Step": 217, "Full-finetune/Step Time": 6.646829603239894} {"Full-finetune/Learning Rate": 1.9992299531181892e-05, "Full-finetune/Loss": 0.9028013348579407, "Full-finetune/Loss (Raw)": 1.015026330947876, "Full-finetune/Step": 218, "Full-finetune/Step Time": 6.6363740637898445} {"Full-finetune/Learning Rate": 1.999205316434098e-05, "Full-finetune/Loss": 0.903418242931366, "Full-finetune/Loss (Raw)": 1.0921176671981812, "Full-finetune/Step": 219, "Full-finetune/Step Time": 6.625265691429377} {"Full-finetune/Learning Rate": 1.9991802919779318e-05, "Full-finetune/Loss": 0.901362955570221, "Full-finetune/Loss (Raw)": 0.8710247278213501, "Full-finetune/Step": 220, "Full-finetune/Step Time": 6.622826676815748} {"Full-finetune/Learning Rate": 1.9991548797594017e-05, "Full-finetune/Loss": 0.9018178582191467, "Full-finetune/Loss (Raw)": 0.892088770866394, "Full-finetune/Step": 221, "Full-finetune/Step Time": 6.625060610473156} {"Full-finetune/Learning Rate": 1.9991290797883693e-05, "Full-finetune/Loss": 0.9003720283508301, "Full-finetune/Loss (Raw)": 0.8624264001846313, "Full-finetune/Step": 222, "Full-finetune/Step Time": 6.6319467425346375} {"Full-finetune/Learning Rate": 1.9991028920748475e-05, "Full-finetune/Loss": 0.8995591402053833, "Full-finetune/Loss (Raw)": 0.8436499238014221, "Full-finetune/Step": 223, "Full-finetune/Step Time": 6.64307639375329} {"Full-finetune/Learning Rate": 1.9990763166289993e-05, "Full-finetune/Loss": 0.8949329853057861, "Full-finetune/Loss (Raw)": 0.33958759903907776, "Full-finetune/Step": 224, "Full-finetune/Step Time": 6.649550214409828} {"Full-finetune/Learning Rate": 1.9990493534611376e-05, "Full-finetune/Loss": 0.8953085541725159, "Full-finetune/Loss (Raw)": 0.9744657278060913, "Full-finetune/Step": 225, "Full-finetune/Step Time": 6.650606248527765} {"Full-finetune/Learning Rate": 1.9990220025817268e-05, "Full-finetune/Loss": 0.8959765434265137, "Full-finetune/Loss (Raw)": 1.0896612405776978, "Full-finetune/Step": 226, "Full-finetune/Step Time": 6.647647723555565} {"Full-finetune/Learning Rate": 1.998994264001381e-05, "Full-finetune/Loss": 0.9027684926986694, "Full-finetune/Loss (Raw)": 1.0537798404693604, "Full-finetune/Step": 227, "Full-finetune/Step Time": 6.6399069968611} {"Full-finetune/Learning Rate": 1.9989661377308654e-05, "Full-finetune/Loss": 0.9017539024353027, "Full-finetune/Loss (Raw)": 0.9956645965576172, "Full-finetune/Step": 228, "Full-finetune/Step Time": 6.632987601682544} {"Full-finetune/Learning Rate": 1.9989376237810944e-05, "Full-finetune/Loss": 0.9015589356422424, "Full-finetune/Loss (Raw)": 0.8629491329193115, "Full-finetune/Step": 229, "Full-finetune/Step Time": 6.6375632137060165} {"Full-finetune/Learning Rate": 1.9989087221631344e-05, "Full-finetune/Loss": 0.9004309177398682, "Full-finetune/Loss (Raw)": 0.9950437545776367, "Full-finetune/Step": 230, "Full-finetune/Step Time": 6.639797326177359} {"Full-finetune/Learning Rate": 1.9988794328882017e-05, "Full-finetune/Loss": 0.9057509303092957, "Full-finetune/Loss (Raw)": 0.9585555791854858, "Full-finetune/Step": 231, "Full-finetune/Step Time": 6.62315733730793} {"Full-finetune/Learning Rate": 1.9988497559676622e-05, "Full-finetune/Loss": 0.9050499200820923, "Full-finetune/Loss (Raw)": 0.9778191447257996, "Full-finetune/Step": 232, "Full-finetune/Step Time": 6.6248432490974665} {"Full-finetune/Learning Rate": 1.9988196914130332e-05, "Full-finetune/Loss": 0.9001317620277405, "Full-finetune/Loss (Raw)": 0.32094526290893555, "Full-finetune/Step": 233, "Full-finetune/Step Time": 6.6461428459733725} {"Full-finetune/Learning Rate": 1.9987892392359827e-05, "Full-finetune/Loss": 0.9004842638969421, "Full-finetune/Loss (Raw)": 1.0033295154571533, "Full-finetune/Step": 234, "Full-finetune/Step Time": 6.6659004762768745} {"Full-finetune/Learning Rate": 1.998758399448328e-05, "Full-finetune/Loss": 0.9007989168167114, "Full-finetune/Loss (Raw)": 1.045951008796692, "Full-finetune/Step": 235, "Full-finetune/Step Time": 6.70449367724359} {"Full-finetune/Learning Rate": 1.9987271720620376e-05, "Full-finetune/Loss": 0.9007975459098816, "Full-finetune/Loss (Raw)": 0.9460386037826538, "Full-finetune/Step": 236, "Full-finetune/Step Time": 6.694931831210852} {"Full-finetune/Learning Rate": 1.9986955570892302e-05, "Full-finetune/Loss": 0.9000850915908813, "Full-finetune/Loss (Raw)": 0.9026182293891907, "Full-finetune/Step": 237, "Full-finetune/Step Time": 6.7283530086278915} {"Full-finetune/Learning Rate": 1.998663554542175e-05, "Full-finetune/Loss": 0.9000149965286255, "Full-finetune/Loss (Raw)": 0.9481015205383301, "Full-finetune/Step": 238, "Full-finetune/Step Time": 6.732249336317182} {"Full-finetune/Learning Rate": 1.9986311644332913e-05, "Full-finetune/Loss": 0.8991938829421997, "Full-finetune/Loss (Raw)": 0.8083089590072632, "Full-finetune/Step": 239, "Full-finetune/Step Time": 6.716862382367253} {"Full-finetune/Learning Rate": 1.9985983867751497e-05, "Full-finetune/Loss": 0.8979437351226807, "Full-finetune/Loss (Raw)": 0.8514484167098999, "Full-finetune/Step": 240, "Full-finetune/Step Time": 6.704780016094446} {"Full-finetune/Learning Rate": 1.99856522158047e-05, "Full-finetune/Loss": 0.898238480091095, "Full-finetune/Loss (Raw)": 1.0097570419311523, "Full-finetune/Step": 241, "Full-finetune/Step Time": 6.692677363753319} {"Full-finetune/Learning Rate": 1.998531668862123e-05, "Full-finetune/Loss": 0.9038187265396118, "Full-finetune/Loss (Raw)": 0.9444314241409302, "Full-finetune/Step": 242, "Full-finetune/Step Time": 6.669197706505656} {"Full-finetune/Learning Rate": 1.9984977286331302e-05, "Full-finetune/Loss": 0.9035976529121399, "Full-finetune/Loss (Raw)": 0.9694271087646484, "Full-finetune/Step": 243, "Full-finetune/Step Time": 6.666766867041588} {"Full-finetune/Learning Rate": 1.998463400906663e-05, "Full-finetune/Loss": 0.9037692546844482, "Full-finetune/Loss (Raw)": 0.8099515438079834, "Full-finetune/Step": 244, "Full-finetune/Step Time": 6.652097906917334} {"Full-finetune/Learning Rate": 1.998428685696043e-05, "Full-finetune/Loss": 0.9047119617462158, "Full-finetune/Loss (Raw)": 1.1368147134780884, "Full-finetune/Step": 245, "Full-finetune/Step Time": 6.644708210602403} {"Full-finetune/Learning Rate": 1.998393583014743e-05, "Full-finetune/Loss": 0.9042953252792358, "Full-finetune/Loss (Raw)": 0.996111273765564, "Full-finetune/Step": 246, "Full-finetune/Step Time": 6.653563167899847} {"Full-finetune/Learning Rate": 1.9983580928763852e-05, "Full-finetune/Loss": 0.9041476249694824, "Full-finetune/Loss (Raw)": 0.927866518497467, "Full-finetune/Step": 247, "Full-finetune/Step Time": 6.654508659616113} {"Full-finetune/Learning Rate": 1.998322215294743e-05, "Full-finetune/Loss": 0.9038627743721008, "Full-finetune/Loss (Raw)": 0.9434366822242737, "Full-finetune/Step": 248, "Full-finetune/Step Time": 6.657641023397446} {"Full-finetune/Learning Rate": 1.9982859502837395e-05, "Full-finetune/Loss": 0.9037082195281982, "Full-finetune/Loss (Raw)": 0.9269397854804993, "Full-finetune/Step": 249, "Full-finetune/Step Time": 6.63830553740263} {"Full-finetune/Learning Rate": 1.998249297857449e-05, "Full-finetune/Loss": 0.9041764736175537, "Full-finetune/Loss (Raw)": 0.9495856165885925, "Full-finetune/Step": 250, "Full-finetune/Step Time": 6.638495232909918} {"Full-finetune/Learning Rate": 1.9982122580300947e-05, "Full-finetune/Loss": 0.9045826196670532, "Full-finetune/Loss (Raw)": 1.0028287172317505, "Full-finetune/Step": 251, "Full-finetune/Step Time": 6.652047770097852} {"Full-finetune/Learning Rate": 1.9981748308160517e-05, "Full-finetune/Loss": 0.9043673276901245, "Full-finetune/Loss (Raw)": 0.9599608778953552, "Full-finetune/Step": 252, "Full-finetune/Step Time": 6.650822747498751} {"Full-finetune/Learning Rate": 1.9981370162298444e-05, "Full-finetune/Loss": 0.9054560661315918, "Full-finetune/Loss (Raw)": 1.0249884128570557, "Full-finetune/Step": 253, "Full-finetune/Step Time": 6.645860519260168} {"Full-finetune/Learning Rate": 1.9980988142861486e-05, "Full-finetune/Loss": 0.9044834971427917, "Full-finetune/Loss (Raw)": 0.8692328333854675, "Full-finetune/Step": 254, "Full-finetune/Step Time": 6.627499107271433} {"Full-finetune/Learning Rate": 1.998060224999789e-05, "Full-finetune/Loss": 0.9004297852516174, "Full-finetune/Loss (Raw)": 0.3673621118068695, "Full-finetune/Step": 255, "Full-finetune/Step Time": 6.644711401313543} {"Full-finetune/Learning Rate": 1.9980212483857412e-05, "Full-finetune/Loss": 0.899097204208374, "Full-finetune/Loss (Raw)": 0.859093427658081, "Full-finetune/Step": 256, "Full-finetune/Step Time": 6.646650919690728} {"Full-finetune/Learning Rate": 1.9979818844591318e-05, "Full-finetune/Loss": 0.8995288610458374, "Full-finetune/Loss (Raw)": 1.0699691772460938, "Full-finetune/Step": 257, "Full-finetune/Step Time": 6.646302795037627} {"Full-finetune/Learning Rate": 1.9979421332352372e-05, "Full-finetune/Loss": 0.8981334567070007, "Full-finetune/Loss (Raw)": 0.8539193272590637, "Full-finetune/Step": 258, "Full-finetune/Step Time": 6.646224664524198} {"Full-finetune/Learning Rate": 1.9979019947294836e-05, "Full-finetune/Loss": 0.8983558416366577, "Full-finetune/Loss (Raw)": 1.0524120330810547, "Full-finetune/Step": 259, "Full-finetune/Step Time": 6.648243233561516} {"Full-finetune/Learning Rate": 1.9978614689574486e-05, "Full-finetune/Loss": 0.8968260884284973, "Full-finetune/Loss (Raw)": 0.8026416301727295, "Full-finetune/Step": 260, "Full-finetune/Step Time": 6.667564941570163} {"Full-finetune/Learning Rate": 1.9978205559348586e-05, "Full-finetune/Loss": 0.8969736099243164, "Full-finetune/Loss (Raw)": 0.780381441116333, "Full-finetune/Step": 261, "Full-finetune/Step Time": 6.651375921443105} {"Full-finetune/Learning Rate": 1.997779255677592e-05, "Full-finetune/Loss": 0.8941152095794678, "Full-finetune/Loss (Raw)": 0.7415185570716858, "Full-finetune/Step": 262, "Full-finetune/Step Time": 6.661528909578919} {"Full-finetune/Learning Rate": 1.997737568201676e-05, "Full-finetune/Loss": 0.8947526812553406, "Full-finetune/Loss (Raw)": 0.9878831505775452, "Full-finetune/Step": 263, "Full-finetune/Step Time": 6.660405183210969} {"Full-finetune/Learning Rate": 1.9976954935232892e-05, "Full-finetune/Loss": 0.8938229084014893, "Full-finetune/Loss (Raw)": 0.8249988555908203, "Full-finetune/Step": 264, "Full-finetune/Step Time": 6.673907918855548} {"Full-finetune/Learning Rate": 1.99765303165876e-05, "Full-finetune/Loss": 0.8937443494796753, "Full-finetune/Loss (Raw)": 0.8701408505439758, "Full-finetune/Step": 265, "Full-finetune/Step Time": 6.670555191114545} {"Full-finetune/Learning Rate": 1.997610182624566e-05, "Full-finetune/Loss": 0.8951027989387512, "Full-finetune/Loss (Raw)": 1.1061322689056396, "Full-finetune/Step": 266, "Full-finetune/Step Time": 6.680833904072642} {"Full-finetune/Learning Rate": 1.9975669464373377e-05, "Full-finetune/Loss": 0.8950064182281494, "Full-finetune/Loss (Raw)": 0.8216100931167603, "Full-finetune/Step": 267, "Full-finetune/Step Time": 6.678271232172847} {"Full-finetune/Learning Rate": 1.9975233231138525e-05, "Full-finetune/Loss": 0.8936055898666382, "Full-finetune/Loss (Raw)": 1.032195806503296, "Full-finetune/Step": 268, "Full-finetune/Step Time": 6.681466897949576} {"Full-finetune/Learning Rate": 1.997479312671041e-05, "Full-finetune/Loss": 0.8938634395599365, "Full-finetune/Loss (Raw)": 0.9734659194946289, "Full-finetune/Step": 269, "Full-finetune/Step Time": 6.6882818434387445} {"Full-finetune/Learning Rate": 1.9974349151259824e-05, "Full-finetune/Loss": 0.89413982629776, "Full-finetune/Loss (Raw)": 0.9246136546134949, "Full-finetune/Step": 270, "Full-finetune/Step Time": 6.653600137680769} {"Full-finetune/Learning Rate": 1.9973901304959066e-05, "Full-finetune/Loss": 0.8945574760437012, "Full-finetune/Loss (Raw)": 0.8781704306602478, "Full-finetune/Step": 271, "Full-finetune/Step Time": 6.644402708858252} {"Full-finetune/Learning Rate": 1.9973449587981935e-05, "Full-finetune/Loss": 0.8997032046318054, "Full-finetune/Loss (Raw)": 0.9881010055541992, "Full-finetune/Step": 272, "Full-finetune/Step Time": 6.62607422657311} {"Full-finetune/Learning Rate": 1.9972994000503734e-05, "Full-finetune/Loss": 0.9051328301429749, "Full-finetune/Loss (Raw)": 1.016081690788269, "Full-finetune/Step": 273, "Full-finetune/Step Time": 6.603170154616237} {"Full-finetune/Learning Rate": 1.997253454270127e-05, "Full-finetune/Loss": 0.9058436751365662, "Full-finetune/Loss (Raw)": 0.9813127517700195, "Full-finetune/Step": 274, "Full-finetune/Step Time": 6.608448943123221} {"Full-finetune/Learning Rate": 1.9972071214752843e-05, "Full-finetune/Loss": 0.9064862728118896, "Full-finetune/Loss (Raw)": 0.8864076733589172, "Full-finetune/Step": 275, "Full-finetune/Step Time": 6.60461812838912} {"Full-finetune/Learning Rate": 1.9971604016838264e-05, "Full-finetune/Loss": 0.9043987989425659, "Full-finetune/Loss (Raw)": 0.7051569223403931, "Full-finetune/Step": 276, "Full-finetune/Step Time": 6.635383736342192} {"Full-finetune/Learning Rate": 1.997113294913885e-05, "Full-finetune/Loss": 0.9038266539573669, "Full-finetune/Loss (Raw)": 0.9632741212844849, "Full-finetune/Step": 277, "Full-finetune/Step Time": 6.634491648525} {"Full-finetune/Learning Rate": 1.9970658011837404e-05, "Full-finetune/Loss": 0.9037030935287476, "Full-finetune/Loss (Raw)": 0.9659738540649414, "Full-finetune/Step": 278, "Full-finetune/Step Time": 6.626538313925266} {"Full-finetune/Learning Rate": 1.997017920511825e-05, "Full-finetune/Loss": 0.9035943150520325, "Full-finetune/Loss (Raw)": 0.9686367511749268, "Full-finetune/Step": 279, "Full-finetune/Step Time": 6.641559321433306} {"Full-finetune/Learning Rate": 1.9969696529167197e-05, "Full-finetune/Loss": 0.9028613567352295, "Full-finetune/Loss (Raw)": 0.8814879059791565, "Full-finetune/Step": 280, "Full-finetune/Step Time": 6.640814613550901} {"Full-finetune/Learning Rate": 1.996920998417156e-05, "Full-finetune/Loss": 0.9024634957313538, "Full-finetune/Loss (Raw)": 0.9706717729568481, "Full-finetune/Step": 281, "Full-finetune/Step Time": 6.625171512365341} {"Full-finetune/Learning Rate": 1.996871957032016e-05, "Full-finetune/Loss": 0.9019684195518494, "Full-finetune/Loss (Raw)": 0.9558922052383423, "Full-finetune/Step": 282, "Full-finetune/Step Time": 6.624199526384473} {"Full-finetune/Learning Rate": 1.996822528780332e-05, "Full-finetune/Loss": 0.9020342826843262, "Full-finetune/Loss (Raw)": 1.0952776670455933, "Full-finetune/Step": 283, "Full-finetune/Step Time": 6.631087580695748} {"Full-finetune/Learning Rate": 1.996772713681286e-05, "Full-finetune/Loss": 0.9077056050300598, "Full-finetune/Loss (Raw)": 1.049180507659912, "Full-finetune/Step": 284, "Full-finetune/Step Time": 6.622323682531714} {"Full-finetune/Learning Rate": 1.9967225117542097e-05, "Full-finetune/Loss": 0.9061218500137329, "Full-finetune/Loss (Raw)": 0.7840882539749146, "Full-finetune/Step": 285, "Full-finetune/Step Time": 6.637686720117927} {"Full-finetune/Learning Rate": 1.9966719230185864e-05, "Full-finetune/Loss": 0.9054627418518066, "Full-finetune/Loss (Raw)": 0.906552255153656, "Full-finetune/Step": 286, "Full-finetune/Step Time": 6.650789592415094} {"Full-finetune/Learning Rate": 1.9966209474940477e-05, "Full-finetune/Loss": 0.9057233929634094, "Full-finetune/Loss (Raw)": 0.9964628219604492, "Full-finetune/Step": 287, "Full-finetune/Step Time": 6.658339252695441} {"Full-finetune/Learning Rate": 1.996569585200377e-05, "Full-finetune/Loss": 0.9050329327583313, "Full-finetune/Loss (Raw)": 0.7890934348106384, "Full-finetune/Step": 288, "Full-finetune/Step Time": 6.665074935182929} {"Full-finetune/Learning Rate": 1.9965178361575068e-05, "Full-finetune/Loss": 0.9059349298477173, "Full-finetune/Loss (Raw)": 1.0201252698898315, "Full-finetune/Step": 289, "Full-finetune/Step Time": 6.668170882388949} {"Full-finetune/Learning Rate": 1.9964657003855197e-05, "Full-finetune/Loss": 0.9039586782455444, "Full-finetune/Loss (Raw)": 0.8305127620697021, "Full-finetune/Step": 290, "Full-finetune/Step Time": 6.679182847961783} {"Full-finetune/Learning Rate": 1.9964131779046484e-05, "Full-finetune/Loss": 0.9042650461196899, "Full-finetune/Loss (Raw)": 1.0274983644485474, "Full-finetune/Step": 291, "Full-finetune/Step Time": 6.673331132158637} {"Full-finetune/Learning Rate": 1.996360268735276e-05, "Full-finetune/Loss": 0.9059627056121826, "Full-finetune/Loss (Raw)": 1.0258915424346924, "Full-finetune/Step": 292, "Full-finetune/Step Time": 6.662802381440997} {"Full-finetune/Learning Rate": 1.9963069728979357e-05, "Full-finetune/Loss": 0.9062204957008362, "Full-finetune/Loss (Raw)": 1.0048632621765137, "Full-finetune/Step": 293, "Full-finetune/Step Time": 6.659440217539668} {"Full-finetune/Learning Rate": 1.9962532904133107e-05, "Full-finetune/Loss": 0.9043554067611694, "Full-finetune/Loss (Raw)": 0.6170479655265808, "Full-finetune/Step": 294, "Full-finetune/Step Time": 6.655414605513215} {"Full-finetune/Learning Rate": 1.996199221302234e-05, "Full-finetune/Loss": 0.9038873910903931, "Full-finetune/Loss (Raw)": 1.0012600421905518, "Full-finetune/Step": 295, "Full-finetune/Step Time": 6.651950506493449} {"Full-finetune/Learning Rate": 1.9961447655856883e-05, "Full-finetune/Loss": 0.9038841724395752, "Full-finetune/Loss (Raw)": 0.891508162021637, "Full-finetune/Step": 296, "Full-finetune/Step Time": 6.64456339366734} {"Full-finetune/Learning Rate": 1.9960899232848074e-05, "Full-finetune/Loss": 0.9047813415527344, "Full-finetune/Loss (Raw)": 1.008845329284668, "Full-finetune/Step": 297, "Full-finetune/Step Time": 6.647968541830778} {"Full-finetune/Learning Rate": 1.9960346944208743e-05, "Full-finetune/Loss": 0.9079539775848389, "Full-finetune/Loss (Raw)": 0.7628152966499329, "Full-finetune/Step": 298, "Full-finetune/Step Time": 6.655766936019063} {"Full-finetune/Learning Rate": 1.995979079015322e-05, "Full-finetune/Loss": 0.9032740592956543, "Full-finetune/Loss (Raw)": 0.3254985213279724, "Full-finetune/Step": 299, "Full-finetune/Step Time": 6.665863344445825} {"Full-finetune/Learning Rate": 1.995923077089734e-05, "Full-finetune/Loss": 0.8995139598846436, "Full-finetune/Loss (Raw)": 0.3671930432319641, "Full-finetune/Step": 300, "Full-finetune/Step Time": 6.674387188628316} {"Full-finetune/Learning Rate": 1.9958666886658436e-05, "Full-finetune/Loss": 0.8941202163696289, "Full-finetune/Loss (Raw)": 0.395578533411026, "Full-finetune/Step": 301, "Full-finetune/Step Time": 6.694067737087607} {"Full-finetune/Learning Rate": 1.995809913765534e-05, "Full-finetune/Loss": 0.8941515684127808, "Full-finetune/Loss (Raw)": 0.817671537399292, "Full-finetune/Step": 302, "Full-finetune/Step Time": 6.70395671389997} {"Full-finetune/Learning Rate": 1.9957527524108386e-05, "Full-finetune/Loss": 0.8884307146072388, "Full-finetune/Loss (Raw)": 0.3193637728691101, "Full-finetune/Step": 303, "Full-finetune/Step Time": 6.7227579560130835} {"Full-finetune/Learning Rate": 1.99569520462394e-05, "Full-finetune/Loss": 0.8932708501815796, "Full-finetune/Loss (Raw)": 0.9712350964546204, "Full-finetune/Step": 304, "Full-finetune/Step Time": 6.708472283557057} {"Full-finetune/Learning Rate": 1.995637270427172e-05, "Full-finetune/Loss": 0.8939758539199829, "Full-finetune/Loss (Raw)": 0.9721659421920776, "Full-finetune/Step": 305, "Full-finetune/Step Time": 6.70036693289876} {"Full-finetune/Learning Rate": 1.9955789498430178e-05, "Full-finetune/Loss": 0.8924107551574707, "Full-finetune/Loss (Raw)": 0.8824366331100464, "Full-finetune/Step": 306, "Full-finetune/Step Time": 6.702454118058085} {"Full-finetune/Learning Rate": 1.99552024289411e-05, "Full-finetune/Loss": 0.8944301605224609, "Full-finetune/Loss (Raw)": 1.0878171920776367, "Full-finetune/Step": 307, "Full-finetune/Step Time": 6.70418775267899} {"Full-finetune/Learning Rate": 1.995461149603232e-05, "Full-finetune/Loss": 0.8945414423942566, "Full-finetune/Loss (Raw)": 0.9619770050048828, "Full-finetune/Step": 308, "Full-finetune/Step Time": 6.7059789802879095} {"Full-finetune/Learning Rate": 1.9954016699933163e-05, "Full-finetune/Loss": 0.894819438457489, "Full-finetune/Loss (Raw)": 0.9898187518119812, "Full-finetune/Step": 309, "Full-finetune/Step Time": 6.720190569758415} {"Full-finetune/Learning Rate": 1.995341804087446e-05, "Full-finetune/Loss": 0.8946045637130737, "Full-finetune/Loss (Raw)": 0.9488458633422852, "Full-finetune/Step": 310, "Full-finetune/Step Time": 6.721303775906563} {"Full-finetune/Learning Rate": 1.9952815519088544e-05, "Full-finetune/Loss": 0.8946717381477356, "Full-finetune/Loss (Raw)": 0.951772928237915, "Full-finetune/Step": 311, "Full-finetune/Step Time": 6.728057142347097} {"Full-finetune/Learning Rate": 1.9952209134809234e-05, "Full-finetune/Loss": 0.8931257724761963, "Full-finetune/Loss (Raw)": 0.8839437961578369, "Full-finetune/Step": 312, "Full-finetune/Step Time": 6.7285071816295385} {"Full-finetune/Learning Rate": 1.995159888827186e-05, "Full-finetune/Loss": 0.8945096731185913, "Full-finetune/Loss (Raw)": 1.0429741144180298, "Full-finetune/Step": 313, "Full-finetune/Step Time": 6.737145185470581} {"Full-finetune/Learning Rate": 1.9950984779713244e-05, "Full-finetune/Loss": 0.8966540694236755, "Full-finetune/Loss (Raw)": 1.0596247911453247, "Full-finetune/Step": 314, "Full-finetune/Step Time": 6.720779264345765} {"Full-finetune/Learning Rate": 1.9950366809371713e-05, "Full-finetune/Loss": 0.8962603807449341, "Full-finetune/Loss (Raw)": 0.8604056239128113, "Full-finetune/Step": 315, "Full-finetune/Step Time": 6.726627925410867} {"Full-finetune/Learning Rate": 1.9949744977487087e-05, "Full-finetune/Loss": 0.891564130783081, "Full-finetune/Loss (Raw)": 0.3852052092552185, "Full-finetune/Step": 316, "Full-finetune/Step Time": 6.746108271181583} {"Full-finetune/Learning Rate": 1.994911928430069e-05, "Full-finetune/Loss": 0.8916763067245483, "Full-finetune/Loss (Raw)": 0.8951838612556458, "Full-finetune/Step": 317, "Full-finetune/Step Time": 6.739397967234254} {"Full-finetune/Learning Rate": 1.994848973005534e-05, "Full-finetune/Loss": 0.8917837142944336, "Full-finetune/Loss (Raw)": 0.985465943813324, "Full-finetune/Step": 318, "Full-finetune/Step Time": 6.731959482654929} {"Full-finetune/Learning Rate": 1.994785631499535e-05, "Full-finetune/Loss": 0.8914734721183777, "Full-finetune/Loss (Raw)": 0.8729690313339233, "Full-finetune/Step": 319, "Full-finetune/Step Time": 6.739876313135028} {"Full-finetune/Learning Rate": 1.994721903936654e-05, "Full-finetune/Loss": 0.8906209468841553, "Full-finetune/Loss (Raw)": 0.8610880374908447, "Full-finetune/Step": 320, "Full-finetune/Step Time": 6.736588686704636} {"Full-finetune/Learning Rate": 1.9946577903416224e-05, "Full-finetune/Loss": 0.8865275979042053, "Full-finetune/Loss (Raw)": 0.33088016510009766, "Full-finetune/Step": 321, "Full-finetune/Step Time": 6.761194134131074} {"Full-finetune/Learning Rate": 1.994593290739322e-05, "Full-finetune/Loss": 0.8857210874557495, "Full-finetune/Loss (Raw)": 0.9152770042419434, "Full-finetune/Step": 322, "Full-finetune/Step Time": 6.762438774108887} {"Full-finetune/Learning Rate": 1.9945284051547823e-05, "Full-finetune/Loss": 0.8860365748405457, "Full-finetune/Loss (Raw)": 0.9439837336540222, "Full-finetune/Step": 323, "Full-finetune/Step Time": 6.757622737437487} {"Full-finetune/Learning Rate": 1.9944631336131856e-05, "Full-finetune/Loss": 0.8876587152481079, "Full-finetune/Loss (Raw)": 1.0252658128738403, "Full-finetune/Step": 324, "Full-finetune/Step Time": 6.757266346365213} {"Full-finetune/Learning Rate": 1.994397476139862e-05, "Full-finetune/Loss": 0.8874931931495667, "Full-finetune/Loss (Raw)": 0.9675559997558594, "Full-finetune/Step": 325, "Full-finetune/Step Time": 6.753306744620204} {"Full-finetune/Learning Rate": 1.994331432760292e-05, "Full-finetune/Loss": 0.888706386089325, "Full-finetune/Loss (Raw)": 1.002975583076477, "Full-finetune/Step": 326, "Full-finetune/Step Time": 6.754448257386684} {"Full-finetune/Learning Rate": 1.9942650035001053e-05, "Full-finetune/Loss": 0.8884382247924805, "Full-finetune/Loss (Raw)": 0.8887684941291809, "Full-finetune/Step": 327, "Full-finetune/Step Time": 6.747615082189441} {"Full-finetune/Learning Rate": 1.994198188385082e-05, "Full-finetune/Loss": 0.8884967565536499, "Full-finetune/Loss (Raw)": 0.9562679529190063, "Full-finetune/Step": 328, "Full-finetune/Step Time": 6.755168933421373} {"Full-finetune/Learning Rate": 1.9941309874411525e-05, "Full-finetune/Loss": 0.8871995210647583, "Full-finetune/Loss (Raw)": 0.8287248611450195, "Full-finetune/Step": 329, "Full-finetune/Step Time": 6.768862770870328} {"Full-finetune/Learning Rate": 1.9940634006943952e-05, "Full-finetune/Loss": 0.8854843378067017, "Full-finetune/Loss (Raw)": 0.8318130373954773, "Full-finetune/Step": 330, "Full-finetune/Step Time": 6.7731624115258455} {"Full-finetune/Learning Rate": 1.9939954281710395e-05, "Full-finetune/Loss": 0.8853213787078857, "Full-finetune/Loss (Raw)": 0.9698691368103027, "Full-finetune/Step": 331, "Full-finetune/Step Time": 6.776064813137054} {"Full-finetune/Learning Rate": 1.9939270698974643e-05, "Full-finetune/Loss": 0.8834512233734131, "Full-finetune/Loss (Raw)": 0.7534251809120178, "Full-finetune/Step": 332, "Full-finetune/Step Time": 6.79126344434917} {"Full-finetune/Learning Rate": 1.9938583259001983e-05, "Full-finetune/Loss": 0.888312816619873, "Full-finetune/Loss (Raw)": 0.9362793564796448, "Full-finetune/Step": 333, "Full-finetune/Step Time": 6.777330396696925} {"Full-finetune/Learning Rate": 1.993789196205919e-05, "Full-finetune/Loss": 0.8881466388702393, "Full-finetune/Loss (Raw)": 0.9425718784332275, "Full-finetune/Step": 334, "Full-finetune/Step Time": 6.768501123413444} {"Full-finetune/Learning Rate": 1.9937196808414552e-05, "Full-finetune/Loss": 0.8880946040153503, "Full-finetune/Loss (Raw)": 0.8464014530181885, "Full-finetune/Step": 335, "Full-finetune/Step Time": 6.755562616512179} {"Full-finetune/Learning Rate": 1.993649779833784e-05, "Full-finetune/Loss": 0.8891141414642334, "Full-finetune/Loss (Raw)": 1.0128810405731201, "Full-finetune/Step": 336, "Full-finetune/Step Time": 6.758185626938939} {"Full-finetune/Learning Rate": 1.993579493210032e-05, "Full-finetune/Loss": 0.8905378580093384, "Full-finetune/Loss (Raw)": 1.1260740756988525, "Full-finetune/Step": 337, "Full-finetune/Step Time": 6.763698648661375} {"Full-finetune/Learning Rate": 1.9935088209974774e-05, "Full-finetune/Loss": 0.8945612907409668, "Full-finetune/Loss (Raw)": 0.8252128958702087, "Full-finetune/Step": 338, "Full-finetune/Step Time": 6.748664410784841} {"Full-finetune/Learning Rate": 1.9934377632235455e-05, "Full-finetune/Loss": 0.894659161567688, "Full-finetune/Loss (Raw)": 0.8258301019668579, "Full-finetune/Step": 339, "Full-finetune/Step Time": 6.749287405982614} {"Full-finetune/Learning Rate": 1.9933663199158135e-05, "Full-finetune/Loss": 0.8938204050064087, "Full-finetune/Loss (Raw)": 0.8314855694770813, "Full-finetune/Step": 340, "Full-finetune/Step Time": 6.747009918093681} {"Full-finetune/Learning Rate": 1.993294491102006e-05, "Full-finetune/Loss": 0.8927075862884521, "Full-finetune/Loss (Raw)": 0.9298990368843079, "Full-finetune/Step": 341, "Full-finetune/Step Time": 6.759865202009678} {"Full-finetune/Learning Rate": 1.9932222768099992e-05, "Full-finetune/Loss": 0.8932532072067261, "Full-finetune/Loss (Raw)": 0.9403813481330872, "Full-finetune/Step": 342, "Full-finetune/Step Time": 6.740991797298193} {"Full-finetune/Learning Rate": 1.9931496770678178e-05, "Full-finetune/Loss": 0.8936193585395813, "Full-finetune/Loss (Raw)": 0.9306431412696838, "Full-finetune/Step": 343, "Full-finetune/Step Time": 6.748847499489784} {"Full-finetune/Learning Rate": 1.993076691903636e-05, "Full-finetune/Loss": 0.8947070837020874, "Full-finetune/Loss (Raw)": 0.9729552865028381, "Full-finetune/Step": 344, "Full-finetune/Step Time": 6.733932446688414} {"Full-finetune/Learning Rate": 1.993003321345779e-05, "Full-finetune/Loss": 0.8948416709899902, "Full-finetune/Loss (Raw)": 1.0036170482635498, "Full-finetune/Step": 345, "Full-finetune/Step Time": 6.7363551165908575} {"Full-finetune/Learning Rate": 1.9929295654227193e-05, "Full-finetune/Loss": 0.8946108222007751, "Full-finetune/Loss (Raw)": 0.9854713082313538, "Full-finetune/Step": 346, "Full-finetune/Step Time": 6.742497969418764} {"Full-finetune/Learning Rate": 1.9928554241630804e-05, "Full-finetune/Loss": 0.8943219184875488, "Full-finetune/Loss (Raw)": 1.055144190788269, "Full-finetune/Step": 347, "Full-finetune/Step Time": 6.74670073390007} {"Full-finetune/Learning Rate": 1.9927808975956354e-05, "Full-finetune/Loss": 0.894838809967041, "Full-finetune/Loss (Raw)": 0.9371800422668457, "Full-finetune/Step": 348, "Full-finetune/Step Time": 6.765411784872413} {"Full-finetune/Learning Rate": 1.9927059857493068e-05, "Full-finetune/Loss": 0.895771324634552, "Full-finetune/Loss (Raw)": 1.011456847190857, "Full-finetune/Step": 349, "Full-finetune/Step Time": 6.773761790245771} {"Full-finetune/Learning Rate": 1.9926306886531658e-05, "Full-finetune/Loss": 0.8953638672828674, "Full-finetune/Loss (Raw)": 0.8102690577507019, "Full-finetune/Step": 350, "Full-finetune/Step Time": 6.7879037875682116} {"Full-finetune/Learning Rate": 1.9925550063364337e-05, "Full-finetune/Loss": 0.8956368565559387, "Full-finetune/Loss (Raw)": 0.8785939812660217, "Full-finetune/Step": 351, "Full-finetune/Step Time": 6.790967218577862} {"Full-finetune/Learning Rate": 1.992478938828482e-05, "Full-finetune/Loss": 0.8997019529342651, "Full-finetune/Loss (Raw)": 0.8599220514297485, "Full-finetune/Step": 352, "Full-finetune/Step Time": 6.783113045617938} {"Full-finetune/Learning Rate": 1.9924024861588307e-05, "Full-finetune/Loss": 0.8985942602157593, "Full-finetune/Loss (Raw)": 0.8326769471168518, "Full-finetune/Step": 353, "Full-finetune/Step Time": 6.780518701300025} {"Full-finetune/Learning Rate": 1.9923256483571495e-05, "Full-finetune/Loss": 0.8963354825973511, "Full-finetune/Loss (Raw)": 0.8005367517471313, "Full-finetune/Step": 354, "Full-finetune/Step Time": 6.788108289241791} {"Full-finetune/Learning Rate": 1.9922484254532578e-05, "Full-finetune/Loss": 0.8958300352096558, "Full-finetune/Loss (Raw)": 0.9890828728675842, "Full-finetune/Step": 355, "Full-finetune/Step Time": 6.787630984559655} {"Full-finetune/Learning Rate": 1.9921708174771237e-05, "Full-finetune/Loss": 0.8942201137542725, "Full-finetune/Loss (Raw)": 0.7895893454551697, "Full-finetune/Step": 356, "Full-finetune/Step Time": 6.794950347393751} {"Full-finetune/Learning Rate": 1.9920928244588662e-05, "Full-finetune/Loss": 0.8933614492416382, "Full-finetune/Loss (Raw)": 0.7530422806739807, "Full-finetune/Step": 357, "Full-finetune/Step Time": 6.798035377636552} {"Full-finetune/Learning Rate": 1.9920144464287527e-05, "Full-finetune/Loss": 0.8884908556938171, "Full-finetune/Loss (Raw)": 0.37160852551460266, "Full-finetune/Step": 358, "Full-finetune/Step Time": 6.815744061022997} {"Full-finetune/Learning Rate": 1.9919356834171996e-05, "Full-finetune/Loss": 0.8871418833732605, "Full-finetune/Loss (Raw)": 0.7858887314796448, "Full-finetune/Step": 359, "Full-finetune/Step Time": 6.817963263019919} {"Full-finetune/Learning Rate": 1.9918565354547738e-05, "Full-finetune/Loss": 0.8865948915481567, "Full-finetune/Loss (Raw)": 0.9078083038330078, "Full-finetune/Step": 360, "Full-finetune/Step Time": 6.819164520129561} {"Full-finetune/Learning Rate": 1.991777002572191e-05, "Full-finetune/Loss": 0.8903378248214722, "Full-finetune/Loss (Raw)": 0.8000375032424927, "Full-finetune/Step": 361, "Full-finetune/Step Time": 6.801150877028704} {"Full-finetune/Learning Rate": 1.9916970848003166e-05, "Full-finetune/Loss": 0.8889902234077454, "Full-finetune/Loss (Raw)": 0.8308395147323608, "Full-finetune/Step": 362, "Full-finetune/Step Time": 6.7925436701625586} {"Full-finetune/Learning Rate": 1.9916167821701645e-05, "Full-finetune/Loss": 0.888475239276886, "Full-finetune/Loss (Raw)": 0.9800301194190979, "Full-finetune/Step": 363, "Full-finetune/Step Time": 6.752805748954415} {"Full-finetune/Learning Rate": 1.991536094712899e-05, "Full-finetune/Loss": 0.8890389800071716, "Full-finetune/Loss (Raw)": 1.0181995630264282, "Full-finetune/Step": 364, "Full-finetune/Step Time": 6.751762894913554} {"Full-finetune/Learning Rate": 1.991455022459833e-05, "Full-finetune/Loss": 0.8900251984596252, "Full-finetune/Loss (Raw)": 1.0288572311401367, "Full-finetune/Step": 365, "Full-finetune/Step Time": 6.72206699475646} {"Full-finetune/Learning Rate": 1.9913735654424297e-05, "Full-finetune/Loss": 0.8891383409500122, "Full-finetune/Loss (Raw)": 0.8345774412155151, "Full-finetune/Step": 366, "Full-finetune/Step Time": 6.727357426658273} {"Full-finetune/Learning Rate": 1.9912917236923003e-05, "Full-finetune/Loss": 0.8902988433837891, "Full-finetune/Loss (Raw)": 0.9568575024604797, "Full-finetune/Step": 367, "Full-finetune/Step Time": 6.732612641528249} {"Full-finetune/Learning Rate": 1.9912094972412067e-05, "Full-finetune/Loss": 0.8907946944236755, "Full-finetune/Loss (Raw)": 0.9149157404899597, "Full-finetune/Step": 368, "Full-finetune/Step Time": 6.732090713456273} {"Full-finetune/Learning Rate": 1.9911268861210584e-05, "Full-finetune/Loss": 0.8899993896484375, "Full-finetune/Loss (Raw)": 0.9079594612121582, "Full-finetune/Step": 369, "Full-finetune/Step Time": 6.73228707164526} {"Full-finetune/Learning Rate": 1.991043890363916e-05, "Full-finetune/Loss": 0.8910107612609863, "Full-finetune/Loss (Raw)": 1.0738835334777832, "Full-finetune/Step": 370, "Full-finetune/Step Time": 6.731262676417828} {"Full-finetune/Learning Rate": 1.9909605100019883e-05, "Full-finetune/Loss": 0.8918558359146118, "Full-finetune/Loss (Raw)": 1.0775998830795288, "Full-finetune/Step": 371, "Full-finetune/Step Time": 6.722605546936393} {"Full-finetune/Learning Rate": 1.9908767450676336e-05, "Full-finetune/Loss": 0.892877995967865, "Full-finetune/Loss (Raw)": 0.9407830238342285, "Full-finetune/Step": 372, "Full-finetune/Step Time": 6.7316763661801815} {"Full-finetune/Learning Rate": 1.9907925955933593e-05, "Full-finetune/Loss": 0.8874167203903198, "Full-finetune/Loss (Raw)": 0.43777427077293396, "Full-finetune/Step": 373, "Full-finetune/Step Time": 6.7519064005464315} {"Full-finetune/Learning Rate": 1.9907080616118223e-05, "Full-finetune/Loss": 0.8823869228363037, "Full-finetune/Loss (Raw)": 0.3522981107234955, "Full-finetune/Step": 374, "Full-finetune/Step Time": 6.762609938159585} {"Full-finetune/Learning Rate": 1.9906231431558283e-05, "Full-finetune/Loss": 0.8829250335693359, "Full-finetune/Loss (Raw)": 0.9967413544654846, "Full-finetune/Step": 375, "Full-finetune/Step Time": 6.760600881651044} {"Full-finetune/Learning Rate": 1.990537840258333e-05, "Full-finetune/Loss": 0.8816341161727905, "Full-finetune/Loss (Raw)": 0.7782057523727417, "Full-finetune/Step": 376, "Full-finetune/Step Time": 6.759801331907511} {"Full-finetune/Learning Rate": 1.9904521529524402e-05, "Full-finetune/Loss": 0.8812863826751709, "Full-finetune/Loss (Raw)": 0.8824179172515869, "Full-finetune/Step": 377, "Full-finetune/Step Time": 6.760665934532881} {"Full-finetune/Learning Rate": 1.990366081271404e-05, "Full-finetune/Loss": 0.8825165629386902, "Full-finetune/Loss (Raw)": 1.1070629358291626, "Full-finetune/Step": 378, "Full-finetune/Step Time": 6.759022321552038} {"Full-finetune/Learning Rate": 1.990279625248627e-05, "Full-finetune/Loss": 0.8828896284103394, "Full-finetune/Loss (Raw)": 1.0505715608596802, "Full-finetune/Step": 379, "Full-finetune/Step Time": 6.7469776682555676} {"Full-finetune/Learning Rate": 1.990192784917661e-05, "Full-finetune/Loss": 0.882805347442627, "Full-finetune/Loss (Raw)": 0.9491789937019348, "Full-finetune/Step": 380, "Full-finetune/Step Time": 6.744139444082975} {"Full-finetune/Learning Rate": 1.9901055603122067e-05, "Full-finetune/Loss": 0.8822747468948364, "Full-finetune/Loss (Raw)": 0.9570719003677368, "Full-finetune/Step": 381, "Full-finetune/Step Time": 6.751949975267053} {"Full-finetune/Learning Rate": 1.9900179514661148e-05, "Full-finetune/Loss": 0.8819049596786499, "Full-finetune/Loss (Raw)": 0.8218984007835388, "Full-finetune/Step": 382, "Full-finetune/Step Time": 6.789140146225691} {"Full-finetune/Learning Rate": 1.9899299584133848e-05, "Full-finetune/Loss": 0.8849591612815857, "Full-finetune/Loss (Raw)": 0.758296012878418, "Full-finetune/Step": 383, "Full-finetune/Step Time": 6.804581258445978} {"Full-finetune/Learning Rate": 1.9898415811881643e-05, "Full-finetune/Loss": 0.8815447092056274, "Full-finetune/Loss (Raw)": 0.42204347252845764, "Full-finetune/Step": 384, "Full-finetune/Step Time": 6.817155351862311} {"Full-finetune/Learning Rate": 1.989752819824751e-05, "Full-finetune/Loss": 0.8800852298736572, "Full-finetune/Loss (Raw)": 0.883155107498169, "Full-finetune/Step": 385, "Full-finetune/Step Time": 6.815813671797514} {"Full-finetune/Learning Rate": 1.989663674357592e-05, "Full-finetune/Loss": 0.8796895742416382, "Full-finetune/Loss (Raw)": 0.8032830953598022, "Full-finetune/Step": 386, "Full-finetune/Step Time": 6.815465480089188} {"Full-finetune/Learning Rate": 1.9895741448212822e-05, "Full-finetune/Loss": 0.8777076005935669, "Full-finetune/Loss (Raw)": 0.798721969127655, "Full-finetune/Step": 387, "Full-finetune/Step Time": 6.817551739513874} {"Full-finetune/Learning Rate": 1.9894842312505673e-05, "Full-finetune/Loss": 0.8774740695953369, "Full-finetune/Loss (Raw)": 0.772742509841919, "Full-finetune/Step": 388, "Full-finetune/Step Time": 6.795850837603211} {"Full-finetune/Learning Rate": 1.9893939336803396e-05, "Full-finetune/Loss": 0.8781099915504456, "Full-finetune/Loss (Raw)": 0.8617855310440063, "Full-finetune/Step": 389, "Full-finetune/Step Time": 6.794759852811694} {"Full-finetune/Learning Rate": 1.989303252145643e-05, "Full-finetune/Loss": 0.8777351379394531, "Full-finetune/Loss (Raw)": 0.6935316920280457, "Full-finetune/Step": 390, "Full-finetune/Step Time": 6.782363837584853} {"Full-finetune/Learning Rate": 1.9892121866816682e-05, "Full-finetune/Loss": 0.8772783279418945, "Full-finetune/Loss (Raw)": 0.9294090270996094, "Full-finetune/Step": 391, "Full-finetune/Step Time": 6.777201751247048} {"Full-finetune/Learning Rate": 1.989120737323757e-05, "Full-finetune/Loss": 0.8773036003112793, "Full-finetune/Loss (Raw)": 0.8282399773597717, "Full-finetune/Step": 392, "Full-finetune/Step Time": 6.774181192740798} {"Full-finetune/Learning Rate": 1.9890289041073984e-05, "Full-finetune/Loss": 0.8760576248168945, "Full-finetune/Loss (Raw)": 0.710645854473114, "Full-finetune/Step": 393, "Full-finetune/Step Time": 6.797886857762933} {"Full-finetune/Learning Rate": 1.988936687068231e-05, "Full-finetune/Loss": 0.8733141422271729, "Full-finetune/Loss (Raw)": 0.7549776434898376, "Full-finetune/Step": 394, "Full-finetune/Step Time": 6.8131161238998175} {"Full-finetune/Learning Rate": 1.9888440862420433e-05, "Full-finetune/Loss": 0.8757203817367554, "Full-finetune/Loss (Raw)": 1.129601001739502, "Full-finetune/Step": 395, "Full-finetune/Step Time": 6.8233722653239965} {"Full-finetune/Learning Rate": 1.988751101664771e-05, "Full-finetune/Loss": 0.8757253885269165, "Full-finetune/Loss (Raw)": 1.0328441858291626, "Full-finetune/Step": 396, "Full-finetune/Step Time": 6.824095634743571} {"Full-finetune/Learning Rate": 1.9886577333724995e-05, "Full-finetune/Loss": 0.8731271624565125, "Full-finetune/Loss (Raw)": 0.6408914923667908, "Full-finetune/Step": 397, "Full-finetune/Step Time": 6.8386649042367935} {"Full-finetune/Learning Rate": 1.9885639814014642e-05, "Full-finetune/Loss": 0.8741245269775391, "Full-finetune/Loss (Raw)": 1.052274227142334, "Full-finetune/Step": 398, "Full-finetune/Step Time": 6.832696825265884} {"Full-finetune/Learning Rate": 1.9884698457880476e-05, "Full-finetune/Loss": 0.8728380799293518, "Full-finetune/Loss (Raw)": 0.7135040760040283, "Full-finetune/Step": 399, "Full-finetune/Step Time": 6.820469943806529} {"Full-finetune/Learning Rate": 1.988375326568782e-05, "Full-finetune/Loss": 0.8717279434204102, "Full-finetune/Loss (Raw)": 0.8459991216659546, "Full-finetune/Step": 400, "Full-finetune/Step Time": 6.818475142121315} {"Full-finetune/Learning Rate": 1.9882804237803487e-05, "Full-finetune/Loss": 0.8715608716011047, "Full-finetune/Loss (Raw)": 0.994698703289032, "Full-finetune/Step": 401, "Full-finetune/Step Time": 6.831592930480838} {"Full-finetune/Learning Rate": 1.9881851374595774e-05, "Full-finetune/Loss": 0.8710336089134216, "Full-finetune/Loss (Raw)": 0.9138243198394775, "Full-finetune/Step": 402, "Full-finetune/Step Time": 6.842268528416753} {"Full-finetune/Learning Rate": 1.988089467643447e-05, "Full-finetune/Loss": 0.8712906837463379, "Full-finetune/Loss (Raw)": 0.9193198680877686, "Full-finetune/Step": 403, "Full-finetune/Step Time": 6.842598794028163} {"Full-finetune/Learning Rate": 1.9879934143690848e-05, "Full-finetune/Loss": 0.8736340999603271, "Full-finetune/Loss (Raw)": 1.0051101446151733, "Full-finetune/Step": 404, "Full-finetune/Step Time": 6.810928743332624} {"Full-finetune/Learning Rate": 1.9878969776737675e-05, "Full-finetune/Loss": 0.8713205456733704, "Full-finetune/Loss (Raw)": 0.6671410799026489, "Full-finetune/Step": 405, "Full-finetune/Step Time": 6.822892777621746} {"Full-finetune/Learning Rate": 1.9878001575949205e-05, "Full-finetune/Loss": 0.8674418330192566, "Full-finetune/Loss (Raw)": 0.4694903492927551, "Full-finetune/Step": 406, "Full-finetune/Step Time": 6.841367546468973} {"Full-finetune/Learning Rate": 1.9877029541701174e-05, "Full-finetune/Loss": 0.8630915880203247, "Full-finetune/Loss (Raw)": 0.4118146598339081, "Full-finetune/Step": 407, "Full-finetune/Step Time": 6.839423812925816} {"Full-finetune/Learning Rate": 1.9876053674370806e-05, "Full-finetune/Loss": 0.8633474707603455, "Full-finetune/Loss (Raw)": 0.9142358899116516, "Full-finetune/Step": 408, "Full-finetune/Step Time": 6.840685935690999} {"Full-finetune/Learning Rate": 1.9875073974336827e-05, "Full-finetune/Loss": 0.8643237352371216, "Full-finetune/Loss (Raw)": 1.0956372022628784, "Full-finetune/Step": 409, "Full-finetune/Step Time": 6.842590477317572} {"Full-finetune/Learning Rate": 1.9874090441979433e-05, "Full-finetune/Loss": 0.8654109239578247, "Full-finetune/Loss (Raw)": 1.0950534343719482, "Full-finetune/Step": 410, "Full-finetune/Step Time": 6.840435525402427} {"Full-finetune/Learning Rate": 1.987310307768031e-05, "Full-finetune/Loss": 0.8640859723091125, "Full-finetune/Loss (Raw)": 0.925675630569458, "Full-finetune/Step": 411, "Full-finetune/Step Time": 6.833287984132767} {"Full-finetune/Learning Rate": 1.987211188182264e-05, "Full-finetune/Loss": 0.858802080154419, "Full-finetune/Loss (Raw)": 0.3728482127189636, "Full-finetune/Step": 412, "Full-finetune/Step Time": 6.84223828278482} {"Full-finetune/Learning Rate": 1.9871116854791085e-05, "Full-finetune/Loss": 0.8602249622344971, "Full-finetune/Loss (Raw)": 0.966218888759613, "Full-finetune/Step": 413, "Full-finetune/Step Time": 6.822481840848923} {"Full-finetune/Learning Rate": 1.9870117996971797e-05, "Full-finetune/Loss": 0.8597190380096436, "Full-finetune/Loss (Raw)": 0.8417890667915344, "Full-finetune/Step": 414, "Full-finetune/Step Time": 6.808685902506113} {"Full-finetune/Learning Rate": 1.986911530875241e-05, "Full-finetune/Loss": 0.8593553304672241, "Full-finetune/Loss (Raw)": 0.9499143362045288, "Full-finetune/Step": 415, "Full-finetune/Step Time": 6.805954717099667} {"Full-finetune/Learning Rate": 1.9868108790522052e-05, "Full-finetune/Loss": 0.8613641858100891, "Full-finetune/Loss (Raw)": 1.046216368675232, "Full-finetune/Step": 416, "Full-finetune/Step Time": 6.793289855122566} {"Full-finetune/Learning Rate": 1.986709844267133e-05, "Full-finetune/Loss": 0.8607706427574158, "Full-finetune/Loss (Raw)": 0.9441553354263306, "Full-finetune/Step": 417, "Full-finetune/Step Time": 6.789519490674138} {"Full-finetune/Learning Rate": 1.9866084265592343e-05, "Full-finetune/Loss": 0.862108588218689, "Full-finetune/Loss (Raw)": 1.001766562461853, "Full-finetune/Step": 418, "Full-finetune/Step Time": 6.777720043435693} {"Full-finetune/Learning Rate": 1.986506625967867e-05, "Full-finetune/Loss": 0.8604899644851685, "Full-finetune/Loss (Raw)": 0.8203144073486328, "Full-finetune/Step": 419, "Full-finetune/Step Time": 6.796251630410552} {"Full-finetune/Learning Rate": 1.9864044425325382e-05, "Full-finetune/Loss": 0.8553367853164673, "Full-finetune/Loss (Raw)": 0.3662920296192169, "Full-finetune/Step": 420, "Full-finetune/Step Time": 6.812775894999504} {"Full-finetune/Learning Rate": 1.986301876292903e-05, "Full-finetune/Loss": 0.8547987937927246, "Full-finetune/Loss (Raw)": 0.9359956383705139, "Full-finetune/Step": 421, "Full-finetune/Step Time": 6.814361492171884} {"Full-finetune/Learning Rate": 1.986198927288766e-05, "Full-finetune/Loss": 0.8571464419364929, "Full-finetune/Loss (Raw)": 0.9175435304641724, "Full-finetune/Step": 422, "Full-finetune/Step Time": 6.82126634195447} {"Full-finetune/Learning Rate": 1.9860955955600784e-05, "Full-finetune/Loss": 0.8573254346847534, "Full-finetune/Loss (Raw)": 1.0241755247116089, "Full-finetune/Step": 423, "Full-finetune/Step Time": 6.815098937600851} {"Full-finetune/Learning Rate": 1.9859918811469424e-05, "Full-finetune/Loss": 0.8581044673919678, "Full-finetune/Loss (Raw)": 0.9912198781967163, "Full-finetune/Step": 424, "Full-finetune/Step Time": 6.815299091860652} {"Full-finetune/Learning Rate": 1.9858877840896076e-05, "Full-finetune/Loss": 0.8551669716835022, "Full-finetune/Loss (Raw)": 0.632847011089325, "Full-finetune/Step": 425, "Full-finetune/Step Time": 6.85036813095212} {"Full-finetune/Learning Rate": 1.9857833044284714e-05, "Full-finetune/Loss": 0.8562166690826416, "Full-finetune/Loss (Raw)": 0.897178053855896, "Full-finetune/Step": 426, "Full-finetune/Step Time": 6.831234868615866} {"Full-finetune/Learning Rate": 1.9856784422040802e-05, "Full-finetune/Loss": 0.8605788946151733, "Full-finetune/Loss (Raw)": 0.8838657140731812, "Full-finetune/Step": 427, "Full-finetune/Step Time": 6.820652695372701} {"Full-finetune/Learning Rate": 1.98557319745713e-05, "Full-finetune/Loss": 0.8655630350112915, "Full-finetune/Loss (Raw)": 1.005164623260498, "Full-finetune/Step": 428, "Full-finetune/Step Time": 6.8013960942626} {"Full-finetune/Learning Rate": 1.985467570228463e-05, "Full-finetune/Loss": 0.870466411113739, "Full-finetune/Loss (Raw)": 1.0232120752334595, "Full-finetune/Step": 429, "Full-finetune/Step Time": 6.786688087508082} {"Full-finetune/Learning Rate": 1.985361560559072e-05, "Full-finetune/Loss": 0.8703767657279968, "Full-finetune/Loss (Raw)": 0.8061909079551697, "Full-finetune/Step": 430, "Full-finetune/Step Time": 6.77941650711} {"Full-finetune/Learning Rate": 1.9852551684900964e-05, "Full-finetune/Loss": 0.8750859498977661, "Full-finetune/Loss (Raw)": 0.9221420288085938, "Full-finetune/Step": 431, "Full-finetune/Step Time": 6.770402094349265} {"Full-finetune/Learning Rate": 1.9851483940628256e-05, "Full-finetune/Loss": 0.8742101192474365, "Full-finetune/Loss (Raw)": 0.8591254353523254, "Full-finetune/Step": 432, "Full-finetune/Step Time": 6.775878839194775} {"Full-finetune/Learning Rate": 1.9850412373186967e-05, "Full-finetune/Loss": 0.873084306716919, "Full-finetune/Loss (Raw)": 0.8280628323554993, "Full-finetune/Step": 433, "Full-finetune/Step Time": 6.779075795784593} {"Full-finetune/Learning Rate": 1.9849336982992948e-05, "Full-finetune/Loss": 0.8731234073638916, "Full-finetune/Loss (Raw)": 0.8874490261077881, "Full-finetune/Step": 434, "Full-finetune/Step Time": 6.779279110953212} {"Full-finetune/Learning Rate": 1.984825777046354e-05, "Full-finetune/Loss": 0.8706604838371277, "Full-finetune/Loss (Raw)": 0.7725595831871033, "Full-finetune/Step": 435, "Full-finetune/Step Time": 6.770365951582789} {"Full-finetune/Learning Rate": 1.9847174736017557e-05, "Full-finetune/Loss": 0.8695716857910156, "Full-finetune/Loss (Raw)": 0.8226102590560913, "Full-finetune/Step": 436, "Full-finetune/Step Time": 6.767364194616675} {"Full-finetune/Learning Rate": 1.9846087880075314e-05, "Full-finetune/Loss": 0.8690468072891235, "Full-finetune/Loss (Raw)": 0.922641396522522, "Full-finetune/Step": 437, "Full-finetune/Step Time": 6.758450010791421} {"Full-finetune/Learning Rate": 1.984499720305859e-05, "Full-finetune/Loss": 0.8674604296684265, "Full-finetune/Loss (Raw)": 0.745781660079956, "Full-finetune/Step": 438, "Full-finetune/Step Time": 6.756050581112504} {"Full-finetune/Learning Rate": 1.984390270539066e-05, "Full-finetune/Loss": 0.8667727708816528, "Full-finetune/Loss (Raw)": 0.8637498021125793, "Full-finetune/Step": 439, "Full-finetune/Step Time": 6.762443833053112} {"Full-finetune/Learning Rate": 1.9842804387496276e-05, "Full-finetune/Loss": 0.8671936988830566, "Full-finetune/Loss (Raw)": 0.9378201961517334, "Full-finetune/Step": 440, "Full-finetune/Step Time": 6.762241415679455} {"Full-finetune/Learning Rate": 1.9841702249801677e-05, "Full-finetune/Loss": 0.8665732741355896, "Full-finetune/Loss (Raw)": 0.9635642170906067, "Full-finetune/Step": 441, "Full-finetune/Step Time": 6.759015696123242} {"Full-finetune/Learning Rate": 1.9840596292734573e-05, "Full-finetune/Loss": 0.861793041229248, "Full-finetune/Loss (Raw)": 0.4477575719356537, "Full-finetune/Step": 442, "Full-finetune/Step Time": 6.77952342107892} {"Full-finetune/Learning Rate": 1.9839486516724173e-05, "Full-finetune/Loss": 0.8627387285232544, "Full-finetune/Loss (Raw)": 0.981454074382782, "Full-finetune/Step": 443, "Full-finetune/Step Time": 6.78612857311964} {"Full-finetune/Learning Rate": 1.983837292220115e-05, "Full-finetune/Loss": 0.8676878213882446, "Full-finetune/Loss (Raw)": 1.0186835527420044, "Full-finetune/Step": 444, "Full-finetune/Step Time": 6.76809417270124} {"Full-finetune/Learning Rate": 1.983725550959768e-05, "Full-finetune/Loss": 0.8676170706748962, "Full-finetune/Loss (Raw)": 0.8861348628997803, "Full-finetune/Step": 445, "Full-finetune/Step Time": 6.761145729571581} {"Full-finetune/Learning Rate": 1.98361342793474e-05, "Full-finetune/Loss": 0.8674513101577759, "Full-finetune/Loss (Raw)": 0.9642457365989685, "Full-finetune/Step": 446, "Full-finetune/Step Time": 6.764543864876032} {"Full-finetune/Learning Rate": 1.9835009231885444e-05, "Full-finetune/Loss": 0.868064284324646, "Full-finetune/Loss (Raw)": 0.9514287114143372, "Full-finetune/Step": 447, "Full-finetune/Step Time": 6.754450315609574} {"Full-finetune/Learning Rate": 1.9833880367648412e-05, "Full-finetune/Loss": 0.8684152960777283, "Full-finetune/Loss (Raw)": 0.9060176014900208, "Full-finetune/Step": 448, "Full-finetune/Step Time": 6.753965973854065} {"Full-finetune/Learning Rate": 1.9832747687074403e-05, "Full-finetune/Loss": 0.874122142791748, "Full-finetune/Loss (Raw)": 1.0613574981689453, "Full-finetune/Step": 449, "Full-finetune/Step Time": 6.7316091768443584} {"Full-finetune/Learning Rate": 1.9831611190602984e-05, "Full-finetune/Loss": 0.873629093170166, "Full-finetune/Loss (Raw)": 0.8521697521209717, "Full-finetune/Step": 450, "Full-finetune/Step Time": 6.730118984356523} {"Full-finetune/Learning Rate": 1.983047087867521e-05, "Full-finetune/Loss": 0.8730868101119995, "Full-finetune/Loss (Raw)": 0.8745740056037903, "Full-finetune/Step": 451, "Full-finetune/Step Time": 6.755243832245469} {"Full-finetune/Learning Rate": 1.9829326751733607e-05, "Full-finetune/Loss": 0.8718768954277039, "Full-finetune/Loss (Raw)": 0.8703912496566772, "Full-finetune/Step": 452, "Full-finetune/Step Time": 6.748833375051618} {"Full-finetune/Learning Rate": 1.9828178810222197e-05, "Full-finetune/Loss": 0.8727195858955383, "Full-finetune/Loss (Raw)": 1.0754220485687256, "Full-finetune/Step": 453, "Full-finetune/Step Time": 6.744616709649563} {"Full-finetune/Learning Rate": 1.9827027054586468e-05, "Full-finetune/Loss": 0.871766984462738, "Full-finetune/Loss (Raw)": 0.8810415863990784, "Full-finetune/Step": 454, "Full-finetune/Step Time": 6.759086420759559} {"Full-finetune/Learning Rate": 1.9825871485273395e-05, "Full-finetune/Loss": 0.8726078867912292, "Full-finetune/Loss (Raw)": 0.9964050650596619, "Full-finetune/Step": 455, "Full-finetune/Step Time": 6.75872896052897} {"Full-finetune/Learning Rate": 1.9824712102731435e-05, "Full-finetune/Loss": 0.867922306060791, "Full-finetune/Loss (Raw)": 0.3565080165863037, "Full-finetune/Step": 456, "Full-finetune/Step Time": 6.764236804097891} {"Full-finetune/Learning Rate": 1.982354890741052e-05, "Full-finetune/Loss": 0.8675457239151001, "Full-finetune/Loss (Raw)": 0.7805339097976685, "Full-finetune/Step": 457, "Full-finetune/Step Time": 6.747944302856922} {"Full-finetune/Learning Rate": 1.982238189976206e-05, "Full-finetune/Loss": 0.8693850040435791, "Full-finetune/Loss (Raw)": 1.0672338008880615, "Full-finetune/Step": 458, "Full-finetune/Step Time": 6.745510993525386} {"Full-finetune/Learning Rate": 1.9821211080238953e-05, "Full-finetune/Loss": 0.8685637712478638, "Full-finetune/Loss (Raw)": 0.8647480607032776, "Full-finetune/Step": 459, "Full-finetune/Step Time": 6.748456306755543} {"Full-finetune/Learning Rate": 1.9820036449295564e-05, "Full-finetune/Loss": 0.8681899905204773, "Full-finetune/Loss (Raw)": 0.7055850028991699, "Full-finetune/Step": 460, "Full-finetune/Step Time": 6.74900296702981} {"Full-finetune/Learning Rate": 1.9818858007387753e-05, "Full-finetune/Loss": 0.8675492405891418, "Full-finetune/Loss (Raw)": 0.8542662858963013, "Full-finetune/Step": 461, "Full-finetune/Step Time": 6.758447732776403} {"Full-finetune/Learning Rate": 1.9817675754972846e-05, "Full-finetune/Loss": 0.8670641779899597, "Full-finetune/Loss (Raw)": 0.8804813623428345, "Full-finetune/Step": 462, "Full-finetune/Step Time": 6.758911821991205} {"Full-finetune/Learning Rate": 1.9816489692509654e-05, "Full-finetune/Loss": 0.8688997030258179, "Full-finetune/Loss (Raw)": 1.081347942352295, "Full-finetune/Step": 463, "Full-finetune/Step Time": 6.754336142912507} {"Full-finetune/Learning Rate": 1.9815299820458458e-05, "Full-finetune/Loss": 0.863324761390686, "Full-finetune/Loss (Raw)": 0.2992832362651825, "Full-finetune/Step": 464, "Full-finetune/Step Time": 6.770878350362182} {"Full-finetune/Learning Rate": 1.9814106139281033e-05, "Full-finetune/Loss": 0.8618561625480652, "Full-finetune/Loss (Raw)": 0.9381020069122314, "Full-finetune/Step": 465, "Full-finetune/Step Time": 6.76867289096117} {"Full-finetune/Learning Rate": 1.9812908649440618e-05, "Full-finetune/Loss": 0.8608249425888062, "Full-finetune/Loss (Raw)": 0.6932145953178406, "Full-finetune/Step": 466, "Full-finetune/Step Time": 6.773336725309491} {"Full-finetune/Learning Rate": 1.981170735140194e-05, "Full-finetune/Loss": 0.8616851568222046, "Full-finetune/Loss (Raw)": 0.935943067073822, "Full-finetune/Step": 467, "Full-finetune/Step Time": 6.773710971698165} {"Full-finetune/Learning Rate": 1.9810502245631194e-05, "Full-finetune/Loss": 0.8606287837028503, "Full-finetune/Loss (Raw)": 0.6962639689445496, "Full-finetune/Step": 468, "Full-finetune/Step Time": 6.7859058156609535} {"Full-finetune/Learning Rate": 1.980929333259606e-05, "Full-finetune/Loss": 0.8560936450958252, "Full-finetune/Loss (Raw)": 0.34940195083618164, "Full-finetune/Step": 469, "Full-finetune/Step Time": 6.787550184875727} {"Full-finetune/Learning Rate": 1.9808080612765694e-05, "Full-finetune/Loss": 0.8556640148162842, "Full-finetune/Loss (Raw)": 0.8853901624679565, "Full-finetune/Step": 470, "Full-finetune/Step Time": 6.789210809394717} {"Full-finetune/Learning Rate": 1.980686408661073e-05, "Full-finetune/Loss": 0.8552858829498291, "Full-finetune/Loss (Raw)": 0.8822380304336548, "Full-finetune/Step": 471, "Full-finetune/Step Time": 6.7941777762025595} {"Full-finetune/Learning Rate": 1.9805643754603275e-05, "Full-finetune/Loss": 0.8557525277137756, "Full-finetune/Loss (Raw)": 1.0326828956604004, "Full-finetune/Step": 472, "Full-finetune/Step Time": 6.796312415972352} {"Full-finetune/Learning Rate": 1.9804419617216922e-05, "Full-finetune/Loss": 0.8542740345001221, "Full-finetune/Loss (Raw)": 0.814369261264801, "Full-finetune/Step": 473, "Full-finetune/Step Time": 6.794303568080068} {"Full-finetune/Learning Rate": 1.9803191674926727e-05, "Full-finetune/Loss": 0.8540316820144653, "Full-finetune/Loss (Raw)": 0.9544541835784912, "Full-finetune/Step": 474, "Full-finetune/Step Time": 6.791029579937458} {"Full-finetune/Learning Rate": 1.9801959928209236e-05, "Full-finetune/Loss": 0.8530581593513489, "Full-finetune/Loss (Raw)": 0.9305358529090881, "Full-finetune/Step": 475, "Full-finetune/Step Time": 6.789423454552889} {"Full-finetune/Learning Rate": 1.9800724377542463e-05, "Full-finetune/Loss": 0.8529809713363647, "Full-finetune/Loss (Raw)": 0.9272933006286621, "Full-finetune/Step": 476, "Full-finetune/Step Time": 6.774793637916446} {"Full-finetune/Learning Rate": 1.9799485023405904e-05, "Full-finetune/Loss": 0.8518904447555542, "Full-finetune/Loss (Raw)": 0.8718752861022949, "Full-finetune/Step": 477, "Full-finetune/Step Time": 6.780901987105608} {"Full-finetune/Learning Rate": 1.9798241866280523e-05, "Full-finetune/Loss": 0.85234534740448, "Full-finetune/Loss (Raw)": 0.868491530418396, "Full-finetune/Step": 478, "Full-finetune/Step Time": 6.758765609934926} {"Full-finetune/Learning Rate": 1.979699490664877e-05, "Full-finetune/Loss": 0.8527611494064331, "Full-finetune/Loss (Raw)": 0.9318223595619202, "Full-finetune/Step": 479, "Full-finetune/Step Time": 6.7448755372315645} {"Full-finetune/Learning Rate": 1.9795744144994562e-05, "Full-finetune/Loss": 0.8534315824508667, "Full-finetune/Loss (Raw)": 0.9457296133041382, "Full-finetune/Step": 480, "Full-finetune/Step Time": 6.738165911287069} {"Full-finetune/Learning Rate": 1.97944895818033e-05, "Full-finetune/Loss": 0.8540959358215332, "Full-finetune/Loss (Raw)": 0.9177232980728149, "Full-finetune/Step": 481, "Full-finetune/Step Time": 6.742464780807495} {"Full-finetune/Learning Rate": 1.979323121756185e-05, "Full-finetune/Loss": 0.8551322817802429, "Full-finetune/Loss (Raw)": 0.9331863522529602, "Full-finetune/Step": 482, "Full-finetune/Step Time": 6.7515335585922} {"Full-finetune/Learning Rate": 1.9791969052758563e-05, "Full-finetune/Loss": 0.8543548583984375, "Full-finetune/Loss (Raw)": 0.8895784616470337, "Full-finetune/Step": 483, "Full-finetune/Step Time": 6.76275154016912} {"Full-finetune/Learning Rate": 1.9790703087883258e-05, "Full-finetune/Loss": 0.854193925857544, "Full-finetune/Loss (Raw)": 0.7689856886863708, "Full-finetune/Step": 484, "Full-finetune/Step Time": 6.766958478838205} {"Full-finetune/Learning Rate": 1.9789433323427233e-05, "Full-finetune/Loss": 0.8533401489257812, "Full-finetune/Loss (Raw)": 0.6437649130821228, "Full-finetune/Step": 485, "Full-finetune/Step Time": 6.77108271792531} {"Full-finetune/Learning Rate": 1.9788159759883258e-05, "Full-finetune/Loss": 0.8570677042007446, "Full-finetune/Loss (Raw)": 0.8487318754196167, "Full-finetune/Step": 486, "Full-finetune/Step Time": 6.7662524953484535} {"Full-finetune/Learning Rate": 1.9786882397745573e-05, "Full-finetune/Loss": 0.8581828474998474, "Full-finetune/Loss (Raw)": 0.9286216497421265, "Full-finetune/Step": 487, "Full-finetune/Step Time": 6.765228729695082} {"Full-finetune/Learning Rate": 1.9785601237509907e-05, "Full-finetune/Loss": 0.8586373329162598, "Full-finetune/Loss (Raw)": 0.9659839868545532, "Full-finetune/Step": 488, "Full-finetune/Step Time": 6.761727347970009} {"Full-finetune/Learning Rate": 1.978431627967345e-05, "Full-finetune/Loss": 0.858536958694458, "Full-finetune/Loss (Raw)": 0.7871877551078796, "Full-finetune/Step": 489, "Full-finetune/Step Time": 6.766592400148511} {"Full-finetune/Learning Rate": 1.9783027524734865e-05, "Full-finetune/Loss": 0.8578993082046509, "Full-finetune/Loss (Raw)": 0.7492197155952454, "Full-finetune/Step": 490, "Full-finetune/Step Time": 6.773052414879203} {"Full-finetune/Learning Rate": 1.9781734973194294e-05, "Full-finetune/Loss": 0.854088306427002, "Full-finetune/Loss (Raw)": 0.49222761392593384, "Full-finetune/Step": 491, "Full-finetune/Step Time": 6.801379783079028} {"Full-finetune/Learning Rate": 1.9780438625553352e-05, "Full-finetune/Loss": 0.8524438738822937, "Full-finetune/Loss (Raw)": 0.8077088594436646, "Full-finetune/Step": 492, "Full-finetune/Step Time": 6.807049797847867} {"Full-finetune/Learning Rate": 1.9779138482315128e-05, "Full-finetune/Loss": 0.8515840768814087, "Full-finetune/Loss (Raw)": 0.918804943561554, "Full-finetune/Step": 493, "Full-finetune/Step Time": 6.809383915737271} {"Full-finetune/Learning Rate": 1.977783454398418e-05, "Full-finetune/Loss": 0.8517440557479858, "Full-finetune/Loss (Raw)": 0.8550524115562439, "Full-finetune/Step": 494, "Full-finetune/Step Time": 6.8203789703547955} {"Full-finetune/Learning Rate": 1.9776526811066543e-05, "Full-finetune/Loss": 0.8509534597396851, "Full-finetune/Loss (Raw)": 0.8556551933288574, "Full-finetune/Step": 495, "Full-finetune/Step Time": 6.8208932019770145} {"Full-finetune/Learning Rate": 1.977521528406972e-05, "Full-finetune/Loss": 0.8510674238204956, "Full-finetune/Loss (Raw)": 0.9295064210891724, "Full-finetune/Step": 496, "Full-finetune/Step Time": 6.834685346111655} {"Full-finetune/Learning Rate": 1.9773899963502693e-05, "Full-finetune/Loss": 0.8506592512130737, "Full-finetune/Loss (Raw)": 0.8557149767875671, "Full-finetune/Step": 497, "Full-finetune/Step Time": 6.838587747886777} {"Full-finetune/Learning Rate": 1.9772580849875905e-05, "Full-finetune/Loss": 0.8496821522712708, "Full-finetune/Loss (Raw)": 0.9488168358802795, "Full-finetune/Step": 498, "Full-finetune/Step Time": 6.837885746732354} {"Full-finetune/Learning Rate": 1.9771257943701284e-05, "Full-finetune/Loss": 0.8476810455322266, "Full-finetune/Loss (Raw)": 0.821458101272583, "Full-finetune/Step": 499, "Full-finetune/Step Time": 6.840178081765771} {"Full-finetune/Learning Rate": 1.9769931245492225e-05, "Full-finetune/Loss": 0.8469817638397217, "Full-finetune/Loss (Raw)": 0.8512739539146423, "Full-finetune/Step": 500, "Full-finetune/Step Time": 6.844889482483268} {"Full-finetune/Learning Rate": 1.976860075576359e-05, "Full-finetune/Loss": 0.8488014936447144, "Full-finetune/Loss (Raw)": 0.6707003116607666, "Full-finetune/Step": 501, "Full-finetune/Step Time": 6.839469639584422} {"Full-finetune/Learning Rate": 1.9767266475031716e-05, "Full-finetune/Loss": 0.851677656173706, "Full-finetune/Loss (Raw)": 0.7204511165618896, "Full-finetune/Step": 502, "Full-finetune/Step Time": 6.850758943706751} {"Full-finetune/Learning Rate": 1.9765928403814416e-05, "Full-finetune/Loss": 0.8497805595397949, "Full-finetune/Loss (Raw)": 0.7539039850234985, "Full-finetune/Step": 503, "Full-finetune/Step Time": 6.866408292204142} {"Full-finetune/Learning Rate": 1.9764586542630965e-05, "Full-finetune/Loss": 0.8507993817329407, "Full-finetune/Loss (Raw)": 0.9086229801177979, "Full-finetune/Step": 504, "Full-finetune/Step Time": 6.862910026684403} {"Full-finetune/Learning Rate": 1.9763240892002113e-05, "Full-finetune/Loss": 0.8510371446609497, "Full-finetune/Loss (Raw)": 0.9128457903862, "Full-finetune/Step": 505, "Full-finetune/Step Time": 6.863893559202552} {"Full-finetune/Learning Rate": 1.9761891452450082e-05, "Full-finetune/Loss": 0.8470311164855957, "Full-finetune/Loss (Raw)": 0.5942979454994202, "Full-finetune/Step": 506, "Full-finetune/Step Time": 6.876032562926412} {"Full-finetune/Learning Rate": 1.9760538224498563e-05, "Full-finetune/Loss": 0.84602290391922, "Full-finetune/Loss (Raw)": 0.9215152263641357, "Full-finetune/Step": 507, "Full-finetune/Step Time": 6.871278369799256} {"Full-finetune/Learning Rate": 1.975918120867272e-05, "Full-finetune/Loss": 0.8454217910766602, "Full-finetune/Loss (Raw)": 0.8722418546676636, "Full-finetune/Step": 508, "Full-finetune/Step Time": 6.871926534920931} {"Full-finetune/Learning Rate": 1.975782040549918e-05, "Full-finetune/Loss": 0.8444690704345703, "Full-finetune/Loss (Raw)": 0.835125207901001, "Full-finetune/Step": 509, "Full-finetune/Step Time": 6.876395620405674} {"Full-finetune/Learning Rate": 1.9756455815506045e-05, "Full-finetune/Loss": 0.8444445133209229, "Full-finetune/Loss (Raw)": 0.8187507390975952, "Full-finetune/Step": 510, "Full-finetune/Step Time": 6.83901322260499} {"Full-finetune/Learning Rate": 1.9755087439222888e-05, "Full-finetune/Loss": 0.8459362983703613, "Full-finetune/Loss (Raw)": 0.9492430686950684, "Full-finetune/Step": 511, "Full-finetune/Step Time": 6.811461308971047} {"Full-finetune/Learning Rate": 1.975371527718075e-05, "Full-finetune/Loss": 0.8494393229484558, "Full-finetune/Loss (Raw)": 0.8704264760017395, "Full-finetune/Step": 512, "Full-finetune/Step Time": 6.7957648523151875} {"Full-finetune/Learning Rate": 1.9752339329912137e-05, "Full-finetune/Loss": 0.8502232432365417, "Full-finetune/Loss (Raw)": 0.9834993481636047, "Full-finetune/Step": 513, "Full-finetune/Step Time": 6.799134708940983} {"Full-finetune/Learning Rate": 1.9750959597951026e-05, "Full-finetune/Loss": 0.851041316986084, "Full-finetune/Loss (Raw)": 0.9080019593238831, "Full-finetune/Step": 514, "Full-finetune/Step Time": 6.812850469723344} {"Full-finetune/Learning Rate": 1.9749576081832865e-05, "Full-finetune/Loss": 0.8479632139205933, "Full-finetune/Loss (Raw)": 0.40472179651260376, "Full-finetune/Step": 515, "Full-finetune/Step Time": 6.829552920535207} {"Full-finetune/Learning Rate": 1.9748188782094576e-05, "Full-finetune/Loss": 0.8493454456329346, "Full-finetune/Loss (Raw)": 0.9496632218360901, "Full-finetune/Step": 516, "Full-finetune/Step Time": 6.855080017820001} {"Full-finetune/Learning Rate": 1.9746797699274535e-05, "Full-finetune/Loss": 0.8500251770019531, "Full-finetune/Loss (Raw)": 0.948795735836029, "Full-finetune/Step": 517, "Full-finetune/Step Time": 6.854865845292807} {"Full-finetune/Learning Rate": 1.9745402833912598e-05, "Full-finetune/Loss": 0.8517321348190308, "Full-finetune/Loss (Raw)": 0.9120250940322876, "Full-finetune/Step": 518, "Full-finetune/Step Time": 6.857889778912067} {"Full-finetune/Learning Rate": 1.9744004186550086e-05, "Full-finetune/Loss": 0.8518056869506836, "Full-finetune/Loss (Raw)": 0.9388202428817749, "Full-finetune/Step": 519, "Full-finetune/Step Time": 6.872604366391897} {"Full-finetune/Learning Rate": 1.974260175772978e-05, "Full-finetune/Loss": 0.8512333631515503, "Full-finetune/Loss (Raw)": 0.7549782991409302, "Full-finetune/Step": 520, "Full-finetune/Step Time": 6.872511832043529} {"Full-finetune/Learning Rate": 1.9741195547995944e-05, "Full-finetune/Loss": 0.8530784249305725, "Full-finetune/Loss (Raw)": 0.9468168616294861, "Full-finetune/Step": 521, "Full-finetune/Step Time": 6.852850545197725} {"Full-finetune/Learning Rate": 1.9739785557894296e-05, "Full-finetune/Loss": 0.8538832664489746, "Full-finetune/Loss (Raw)": 0.8580044507980347, "Full-finetune/Step": 522, "Full-finetune/Step Time": 6.8345822505652905} {"Full-finetune/Learning Rate": 1.9738371787972025e-05, "Full-finetune/Loss": 0.8502562046051025, "Full-finetune/Loss (Raw)": 0.6653326153755188, "Full-finetune/Step": 523, "Full-finetune/Step Time": 6.829447655007243} {"Full-finetune/Learning Rate": 1.9736954238777793e-05, "Full-finetune/Loss": 0.8492028713226318, "Full-finetune/Loss (Raw)": 0.8980220556259155, "Full-finetune/Step": 524, "Full-finetune/Step Time": 6.825569933280349} {"Full-finetune/Learning Rate": 1.9735532910861713e-05, "Full-finetune/Loss": 0.8514754772186279, "Full-finetune/Loss (Raw)": 0.9317812323570251, "Full-finetune/Step": 525, "Full-finetune/Step Time": 6.8086032792925835} {"Full-finetune/Learning Rate": 1.9734107804775385e-05, "Full-finetune/Loss": 0.849614143371582, "Full-finetune/Loss (Raw)": 0.8140237331390381, "Full-finetune/Step": 526, "Full-finetune/Step Time": 6.820596186444163} {"Full-finetune/Learning Rate": 1.9732678921071858e-05, "Full-finetune/Loss": 0.8504157066345215, "Full-finetune/Loss (Raw)": 0.8160995244979858, "Full-finetune/Step": 527, "Full-finetune/Step Time": 6.814693298190832} {"Full-finetune/Learning Rate": 1.9731246260305657e-05, "Full-finetune/Loss": 0.8509807586669922, "Full-finetune/Loss (Raw)": 0.9183230996131897, "Full-finetune/Step": 528, "Full-finetune/Step Time": 6.809406392276287} {"Full-finetune/Learning Rate": 1.9729809823032767e-05, "Full-finetune/Loss": 0.8501825928688049, "Full-finetune/Loss (Raw)": 0.8925414681434631, "Full-finetune/Step": 529, "Full-finetune/Step Time": 6.7982428297400475} {"Full-finetune/Learning Rate": 1.9728369609810648e-05, "Full-finetune/Loss": 0.8505464792251587, "Full-finetune/Loss (Raw)": 0.9604039788246155, "Full-finetune/Step": 530, "Full-finetune/Step Time": 6.783430064097047} {"Full-finetune/Learning Rate": 1.972692562119821e-05, "Full-finetune/Loss": 0.8513273000717163, "Full-finetune/Loss (Raw)": 1.0192582607269287, "Full-finetune/Step": 531, "Full-finetune/Step Time": 6.779041476547718} {"Full-finetune/Learning Rate": 1.9725477857755842e-05, "Full-finetune/Loss": 0.8498361110687256, "Full-finetune/Loss (Raw)": 0.8142422437667847, "Full-finetune/Step": 532, "Full-finetune/Step Time": 6.776784049347043} {"Full-finetune/Learning Rate": 1.9724026320045393e-05, "Full-finetune/Loss": 0.8514449596405029, "Full-finetune/Loss (Raw)": 0.8730660080909729, "Full-finetune/Step": 533, "Full-finetune/Step Time": 6.772769613191485} {"Full-finetune/Learning Rate": 1.9722571008630172e-05, "Full-finetune/Loss": 0.8545573949813843, "Full-finetune/Loss (Raw)": 0.867888331413269, "Full-finetune/Step": 534, "Full-finetune/Step Time": 6.762080915272236} {"Full-finetune/Learning Rate": 1.9721111924074958e-05, "Full-finetune/Loss": 0.8575267791748047, "Full-finetune/Loss (Raw)": 0.7918878197669983, "Full-finetune/Step": 535, "Full-finetune/Step Time": 6.7543195225298405} {"Full-finetune/Learning Rate": 1.9719649066946e-05, "Full-finetune/Loss": 0.8562816381454468, "Full-finetune/Loss (Raw)": 0.7548679709434509, "Full-finetune/Step": 536, "Full-finetune/Step Time": 6.774115294218063} {"Full-finetune/Learning Rate": 1.9718182437810996e-05, "Full-finetune/Loss": 0.8548426032066345, "Full-finetune/Loss (Raw)": 0.9114368557929993, "Full-finetune/Step": 537, "Full-finetune/Step Time": 6.775852270424366} {"Full-finetune/Learning Rate": 1.971671203723912e-05, "Full-finetune/Loss": 0.8534011840820312, "Full-finetune/Loss (Raw)": 0.9105561375617981, "Full-finetune/Step": 538, "Full-finetune/Step Time": 6.785061461851001} {"Full-finetune/Learning Rate": 1.9715237865801007e-05, "Full-finetune/Loss": 0.8539713621139526, "Full-finetune/Loss (Raw)": 0.9986538290977478, "Full-finetune/Step": 539, "Full-finetune/Step Time": 6.777717180550098} {"Full-finetune/Learning Rate": 1.971375992406875e-05, "Full-finetune/Loss": 0.8578363060951233, "Full-finetune/Loss (Raw)": 0.8675618767738342, "Full-finetune/Step": 540, "Full-finetune/Step Time": 6.757005665451288} {"Full-finetune/Learning Rate": 1.9712278212615913e-05, "Full-finetune/Loss": 0.8571957349777222, "Full-finetune/Loss (Raw)": 0.8842234015464783, "Full-finetune/Step": 541, "Full-finetune/Step Time": 6.762261934578419} {"Full-finetune/Learning Rate": 1.9710792732017517e-05, "Full-finetune/Loss": 0.8575694561004639, "Full-finetune/Loss (Raw)": 0.8896186947822571, "Full-finetune/Step": 542, "Full-finetune/Step Time": 6.760118409991264} {"Full-finetune/Learning Rate": 1.9709303482850047e-05, "Full-finetune/Loss": 0.8580394387245178, "Full-finetune/Loss (Raw)": 1.010076642036438, "Full-finetune/Step": 543, "Full-finetune/Step Time": 6.759334675967693} {"Full-finetune/Learning Rate": 1.9707810465691456e-05, "Full-finetune/Loss": 0.857255220413208, "Full-finetune/Loss (Raw)": 0.9458363056182861, "Full-finetune/Step": 544, "Full-finetune/Step Time": 6.759829103946686} {"Full-finetune/Learning Rate": 1.970631368112115e-05, "Full-finetune/Loss": 0.8570383787155151, "Full-finetune/Loss (Raw)": 0.9164025783538818, "Full-finetune/Step": 545, "Full-finetune/Step Time": 6.756319468840957} {"Full-finetune/Learning Rate": 1.970481312972001e-05, "Full-finetune/Loss": 0.8559366464614868, "Full-finetune/Loss (Raw)": 0.8607416152954102, "Full-finetune/Step": 546, "Full-finetune/Step Time": 6.757489884272218} {"Full-finetune/Learning Rate": 1.9703308812070358e-05, "Full-finetune/Loss": 0.8572338223457336, "Full-finetune/Loss (Raw)": 0.986356258392334, "Full-finetune/Step": 547, "Full-finetune/Step Time": 6.747988421469927} {"Full-finetune/Learning Rate": 1.9701800728756e-05, "Full-finetune/Loss": 0.860248327255249, "Full-finetune/Loss (Raw)": 0.7521503567695618, "Full-finetune/Step": 548, "Full-finetune/Step Time": 6.744529051706195} {"Full-finetune/Learning Rate": 1.970028888036219e-05, "Full-finetune/Loss": 0.857864499092102, "Full-finetune/Loss (Raw)": 0.6308625340461731, "Full-finetune/Step": 549, "Full-finetune/Step Time": 6.751341626048088} {"Full-finetune/Learning Rate": 1.9698773267475652e-05, "Full-finetune/Loss": 0.8572391271591187, "Full-finetune/Loss (Raw)": 0.8375006318092346, "Full-finetune/Step": 550, "Full-finetune/Step Time": 6.733296513557434} {"Full-finetune/Learning Rate": 1.9697253890684556e-05, "Full-finetune/Loss": 0.8556032776832581, "Full-finetune/Loss (Raw)": 0.8147837519645691, "Full-finetune/Step": 551, "Full-finetune/Step Time": 6.733733274042606} {"Full-finetune/Learning Rate": 1.969573075057855e-05, "Full-finetune/Loss": 0.8557596206665039, "Full-finetune/Loss (Raw)": 1.0112360715866089, "Full-finetune/Step": 552, "Full-finetune/Step Time": 6.729352749884129} {"Full-finetune/Learning Rate": 1.9694203847748734e-05, "Full-finetune/Loss": 0.8561813831329346, "Full-finetune/Loss (Raw)": 0.6868278980255127, "Full-finetune/Step": 553, "Full-finetune/Step Time": 6.692717760801315} {"Full-finetune/Learning Rate": 1.969267318278767e-05, "Full-finetune/Loss": 0.856716513633728, "Full-finetune/Loss (Raw)": 0.9656776189804077, "Full-finetune/Step": 554, "Full-finetune/Step Time": 6.689015321433544} {"Full-finetune/Learning Rate": 1.969113875628937e-05, "Full-finetune/Loss": 0.856596827507019, "Full-finetune/Loss (Raw)": 0.8685477375984192, "Full-finetune/Step": 555, "Full-finetune/Step Time": 6.702583137899637} {"Full-finetune/Learning Rate": 1.9689600568849328e-05, "Full-finetune/Loss": 0.8554889559745789, "Full-finetune/Loss (Raw)": 0.8633522391319275, "Full-finetune/Step": 556, "Full-finetune/Step Time": 6.714841101318598} {"Full-finetune/Learning Rate": 1.968805862106448e-05, "Full-finetune/Loss": 0.8540947437286377, "Full-finetune/Loss (Raw)": 0.8447552919387817, "Full-finetune/Step": 557, "Full-finetune/Step Time": 6.710983823984861} {"Full-finetune/Learning Rate": 1.968651291353322e-05, "Full-finetune/Loss": 0.8499782085418701, "Full-finetune/Loss (Raw)": 0.2792690694332123, "Full-finetune/Step": 558, "Full-finetune/Step Time": 6.7214689161628485} {"Full-finetune/Learning Rate": 1.968496344685541e-05, "Full-finetune/Loss": 0.8495961427688599, "Full-finetune/Loss (Raw)": 0.8732436299324036, "Full-finetune/Step": 559, "Full-finetune/Step Time": 6.716664893552661} {"Full-finetune/Learning Rate": 1.9683410221632374e-05, "Full-finetune/Loss": 0.8488257527351379, "Full-finetune/Loss (Raw)": 0.760513186454773, "Full-finetune/Step": 560, "Full-finetune/Step Time": 6.710738496854901} {"Full-finetune/Learning Rate": 1.9681853238466883e-05, "Full-finetune/Loss": 0.8485778570175171, "Full-finetune/Loss (Raw)": 0.7963350415229797, "Full-finetune/Step": 561, "Full-finetune/Step Time": 6.722447955980897} {"Full-finetune/Learning Rate": 1.9680292497963172e-05, "Full-finetune/Loss": 0.8478841781616211, "Full-finetune/Loss (Raw)": 0.7986553907394409, "Full-finetune/Step": 562, "Full-finetune/Step Time": 6.719476286321878} {"Full-finetune/Learning Rate": 1.9678728000726936e-05, "Full-finetune/Loss": 0.8492087125778198, "Full-finetune/Loss (Raw)": 0.9421043395996094, "Full-finetune/Step": 563, "Full-finetune/Step Time": 6.719978598877788} {"Full-finetune/Learning Rate": 1.967715974736532e-05, "Full-finetune/Loss": 0.8498852252960205, "Full-finetune/Loss (Raw)": 0.9092035293579102, "Full-finetune/Step": 564, "Full-finetune/Step Time": 6.7280819825828075} {"Full-finetune/Learning Rate": 1.9675587738486935e-05, "Full-finetune/Loss": 0.848436176776886, "Full-finetune/Loss (Raw)": 0.7371588349342346, "Full-finetune/Step": 565, "Full-finetune/Step Time": 6.7367756720632315} {"Full-finetune/Learning Rate": 1.9674011974701855e-05, "Full-finetune/Loss": 0.8503854870796204, "Full-finetune/Loss (Raw)": 0.9952961802482605, "Full-finetune/Step": 566, "Full-finetune/Step Time": 6.7375597804784775} {"Full-finetune/Learning Rate": 1.9672432456621597e-05, "Full-finetune/Loss": 0.8499390482902527, "Full-finetune/Loss (Raw)": 0.8066076636314392, "Full-finetune/Step": 567, "Full-finetune/Step Time": 6.738045785576105} {"Full-finetune/Learning Rate": 1.9670849184859136e-05, "Full-finetune/Loss": 0.8447728753089905, "Full-finetune/Loss (Raw)": 0.2765476703643799, "Full-finetune/Step": 568, "Full-finetune/Step Time": 6.751306783407927} {"Full-finetune/Learning Rate": 1.9669262160028916e-05, "Full-finetune/Loss": 0.8436645269393921, "Full-finetune/Loss (Raw)": 0.8216933012008667, "Full-finetune/Step": 569, "Full-finetune/Step Time": 6.768104093149304} {"Full-finetune/Learning Rate": 1.966767138274683e-05, "Full-finetune/Loss": 0.8471577167510986, "Full-finetune/Loss (Raw)": 0.8948879241943359, "Full-finetune/Step": 570, "Full-finetune/Step Time": 6.754412867128849} {"Full-finetune/Learning Rate": 1.966607685363022e-05, "Full-finetune/Loss": 0.8456921577453613, "Full-finetune/Loss (Raw)": 0.7938580513000488, "Full-finetune/Step": 571, "Full-finetune/Step Time": 6.756874464452267} {"Full-finetune/Learning Rate": 1.96644785732979e-05, "Full-finetune/Loss": 0.8448470234870911, "Full-finetune/Loss (Raw)": 0.9105080366134644, "Full-finetune/Step": 572, "Full-finetune/Step Time": 6.755230700597167} {"Full-finetune/Learning Rate": 1.966287654237013e-05, "Full-finetune/Loss": 0.8458472490310669, "Full-finetune/Loss (Raw)": 1.0141655206680298, "Full-finetune/Step": 573, "Full-finetune/Step Time": 6.759978290647268} {"Full-finetune/Learning Rate": 1.9661270761468625e-05, "Full-finetune/Loss": 0.846169114112854, "Full-finetune/Loss (Raw)": 1.0054479837417603, "Full-finetune/Step": 574, "Full-finetune/Step Time": 6.755960987880826} {"Full-finetune/Learning Rate": 1.965966123121656e-05, "Full-finetune/Loss": 0.8464570045471191, "Full-finetune/Loss (Raw)": 0.9882778525352478, "Full-finetune/Step": 575, "Full-finetune/Step Time": 6.77332373522222} {"Full-finetune/Learning Rate": 1.965804795223856e-05, "Full-finetune/Loss": 0.8462698459625244, "Full-finetune/Loss (Raw)": 0.8820601105690002, "Full-finetune/Step": 576, "Full-finetune/Step Time": 6.773895459249616} {"Full-finetune/Learning Rate": 1.9656430925160704e-05, "Full-finetune/Loss": 0.8448362350463867, "Full-finetune/Loss (Raw)": 0.8778485059738159, "Full-finetune/Step": 577, "Full-finetune/Step Time": 6.769428480416536} {"Full-finetune/Learning Rate": 1.9654810150610536e-05, "Full-finetune/Loss": 0.8444334864616394, "Full-finetune/Loss (Raw)": 0.8006228804588318, "Full-finetune/Step": 578, "Full-finetune/Step Time": 6.764866279438138} {"Full-finetune/Learning Rate": 1.9653185629217038e-05, "Full-finetune/Loss": 0.8434976935386658, "Full-finetune/Loss (Raw)": 0.7547894716262817, "Full-finetune/Step": 579, "Full-finetune/Step Time": 6.7658243626356125} {"Full-finetune/Learning Rate": 1.9651557361610663e-05, "Full-finetune/Loss": 0.8440634608268738, "Full-finetune/Loss (Raw)": 0.9428096413612366, "Full-finetune/Step": 580, "Full-finetune/Step Time": 6.767331078648567} {"Full-finetune/Learning Rate": 1.9649925348423308e-05, "Full-finetune/Loss": 0.8436201810836792, "Full-finetune/Loss (Raw)": 1.018681526184082, "Full-finetune/Step": 581, "Full-finetune/Step Time": 6.772041732445359} {"Full-finetune/Learning Rate": 1.964828959028832e-05, "Full-finetune/Loss": 0.8445247411727905, "Full-finetune/Loss (Raw)": 0.9968283772468567, "Full-finetune/Step": 582, "Full-finetune/Step Time": 6.76743695512414} {"Full-finetune/Learning Rate": 1.964665008784051e-05, "Full-finetune/Loss": 0.8428062200546265, "Full-finetune/Loss (Raw)": 0.7764323949813843, "Full-finetune/Step": 583, "Full-finetune/Step Time": 6.7714497316628695} {"Full-finetune/Learning Rate": 1.9645006841716137e-05, "Full-finetune/Loss": 0.8468440175056458, "Full-finetune/Loss (Raw)": 0.8733484745025635, "Full-finetune/Step": 584, "Full-finetune/Step Time": 6.7637848276644945} {"Full-finetune/Learning Rate": 1.9643359852552912e-05, "Full-finetune/Loss": 0.8482156991958618, "Full-finetune/Loss (Raw)": 0.9561106562614441, "Full-finetune/Step": 585, "Full-finetune/Step Time": 6.764152703806758} {"Full-finetune/Learning Rate": 1.9641709120989996e-05, "Full-finetune/Loss": 0.8425959944725037, "Full-finetune/Loss (Raw)": 0.3479079008102417, "Full-finetune/Step": 586, "Full-finetune/Step Time": 6.776525564491749} {"Full-finetune/Learning Rate": 1.9640054647668015e-05, "Full-finetune/Loss": 0.8440989255905151, "Full-finetune/Loss (Raw)": 1.057125449180603, "Full-finetune/Step": 587, "Full-finetune/Step Time": 6.769946355372667} {"Full-finetune/Learning Rate": 1.9638396433229025e-05, "Full-finetune/Loss": 0.8446764945983887, "Full-finetune/Loss (Raw)": 0.7795169353485107, "Full-finetune/Step": 588, "Full-finetune/Step Time": 6.777659198269248} {"Full-finetune/Learning Rate": 1.963673447831656e-05, "Full-finetune/Loss": 0.8457842469215393, "Full-finetune/Loss (Raw)": 0.9960535168647766, "Full-finetune/Step": 589, "Full-finetune/Step Time": 6.762253940105438} {"Full-finetune/Learning Rate": 1.9635068783575577e-05, "Full-finetune/Loss": 0.8467946648597717, "Full-finetune/Loss (Raw)": 1.0098224878311157, "Full-finetune/Step": 590, "Full-finetune/Step Time": 6.763452421873808} {"Full-finetune/Learning Rate": 1.9633399349652512e-05, "Full-finetune/Loss": 0.8446098566055298, "Full-finetune/Loss (Raw)": 0.8016886115074158, "Full-finetune/Step": 591, "Full-finetune/Step Time": 6.764831254258752} {"Full-finetune/Learning Rate": 1.9631726177195235e-05, "Full-finetune/Loss": 0.8492188453674316, "Full-finetune/Loss (Raw)": 0.8892337083816528, "Full-finetune/Step": 592, "Full-finetune/Step Time": 6.747284421697259} {"Full-finetune/Learning Rate": 1.9630049266853076e-05, "Full-finetune/Loss": 0.8486972451210022, "Full-finetune/Loss (Raw)": 0.8713310360908508, "Full-finetune/Step": 593, "Full-finetune/Step Time": 6.760449443012476} {"Full-finetune/Learning Rate": 1.96283686192768e-05, "Full-finetune/Loss": 0.8505195379257202, "Full-finetune/Loss (Raw)": 0.9264763593673706, "Full-finetune/Step": 594, "Full-finetune/Step Time": 6.74910294637084} {"Full-finetune/Learning Rate": 1.9626684235118646e-05, "Full-finetune/Loss": 0.8506187200546265, "Full-finetune/Loss (Raw)": 0.9486281275749207, "Full-finetune/Step": 595, "Full-finetune/Step Time": 6.73971282504499} {"Full-finetune/Learning Rate": 1.9624996115032287e-05, "Full-finetune/Loss": 0.8516428470611572, "Full-finetune/Loss (Raw)": 0.8273608088493347, "Full-finetune/Step": 596, "Full-finetune/Step Time": 6.722826723009348} {"Full-finetune/Learning Rate": 1.9623304259672844e-05, "Full-finetune/Loss": 0.8549612760543823, "Full-finetune/Loss (Raw)": 0.7741585969924927, "Full-finetune/Step": 597, "Full-finetune/Step Time": 6.7082689348608255} {"Full-finetune/Learning Rate": 1.96216086696969e-05, "Full-finetune/Loss": 0.8549000024795532, "Full-finetune/Loss (Raw)": 0.8775458931922913, "Full-finetune/Step": 598, "Full-finetune/Step Time": 6.728741832077503} {"Full-finetune/Learning Rate": 1.9619909345762474e-05, "Full-finetune/Loss": 0.8556118011474609, "Full-finetune/Loss (Raw)": 0.973353922367096, "Full-finetune/Step": 599, "Full-finetune/Step Time": 6.711104687303305} {"Full-finetune/Learning Rate": 1.961820628852905e-05, "Full-finetune/Loss": 0.8550897836685181, "Full-finetune/Loss (Raw)": 0.9658644795417786, "Full-finetune/Step": 600, "Full-finetune/Step Time": 6.715431185439229} {"Full-finetune/Learning Rate": 1.9616499498657537e-05, "Full-finetune/Loss": 0.8549827337265015, "Full-finetune/Loss (Raw)": 0.8006584048271179, "Full-finetune/Step": 601, "Full-finetune/Step Time": 6.739268507808447} {"Full-finetune/Learning Rate": 1.961478897681032e-05, "Full-finetune/Loss": 0.8548610806465149, "Full-finetune/Loss (Raw)": 0.9388828873634338, "Full-finetune/Step": 602, "Full-finetune/Step Time": 6.732328426092863} {"Full-finetune/Learning Rate": 1.9613074723651214e-05, "Full-finetune/Loss": 0.8532311916351318, "Full-finetune/Loss (Raw)": 0.7219125032424927, "Full-finetune/Step": 603, "Full-finetune/Step Time": 6.737209422513843} {"Full-finetune/Learning Rate": 1.9611356739845484e-05, "Full-finetune/Loss": 0.8535343408584595, "Full-finetune/Loss (Raw)": 0.9660966396331787, "Full-finetune/Step": 604, "Full-finetune/Step Time": 6.7373118959367275} {"Full-finetune/Learning Rate": 1.9609635026059855e-05, "Full-finetune/Loss": 0.8526926040649414, "Full-finetune/Loss (Raw)": 0.7641310691833496, "Full-finetune/Step": 605, "Full-finetune/Step Time": 6.732475332915783} {"Full-finetune/Learning Rate": 1.9607909582962478e-05, "Full-finetune/Loss": 0.852078914642334, "Full-finetune/Loss (Raw)": 0.7899448275566101, "Full-finetune/Step": 606, "Full-finetune/Step Time": 6.748494891449809} {"Full-finetune/Learning Rate": 1.9606180411222974e-05, "Full-finetune/Loss": 0.8498038649559021, "Full-finetune/Loss (Raw)": 0.6406141519546509, "Full-finetune/Step": 607, "Full-finetune/Step Time": 6.77698602154851} {"Full-finetune/Learning Rate": 1.9604447511512396e-05, "Full-finetune/Loss": 0.8503211736679077, "Full-finetune/Loss (Raw)": 1.0119491815567017, "Full-finetune/Step": 608, "Full-finetune/Step Time": 6.773429347202182} {"Full-finetune/Learning Rate": 1.960271088450325e-05, "Full-finetune/Loss": 0.8494217395782471, "Full-finetune/Loss (Raw)": 0.802591860294342, "Full-finetune/Step": 609, "Full-finetune/Step Time": 6.769134879112244} {"Full-finetune/Learning Rate": 1.960097053086949e-05, "Full-finetune/Loss": 0.8492318391799927, "Full-finetune/Loss (Raw)": 0.9088798761367798, "Full-finetune/Step": 610, "Full-finetune/Step Time": 6.759534329175949} {"Full-finetune/Learning Rate": 1.9599226451286504e-05, "Full-finetune/Loss": 0.8493921756744385, "Full-finetune/Loss (Raw)": 0.9101032018661499, "Full-finetune/Step": 611, "Full-finetune/Step Time": 6.737362211570144} {"Full-finetune/Learning Rate": 1.9597478646431147e-05, "Full-finetune/Loss": 0.8518803119659424, "Full-finetune/Loss (Raw)": 1.0874649286270142, "Full-finetune/Step": 612, "Full-finetune/Step Time": 6.735465880483389} {"Full-finetune/Learning Rate": 1.95957271169817e-05, "Full-finetune/Loss": 0.8525323867797852, "Full-finetune/Loss (Raw)": 0.7272348999977112, "Full-finetune/Step": 613, "Full-finetune/Step Time": 6.726886166259646} {"Full-finetune/Learning Rate": 1.95939718636179e-05, "Full-finetune/Loss": 0.8532407283782959, "Full-finetune/Loss (Raw)": 0.9393907189369202, "Full-finetune/Step": 614, "Full-finetune/Step Time": 6.71239578537643} {"Full-finetune/Learning Rate": 1.9592212887020923e-05, "Full-finetune/Loss": 0.8529422283172607, "Full-finetune/Loss (Raw)": 0.8904194235801697, "Full-finetune/Step": 615, "Full-finetune/Step Time": 6.717648329213262} {"Full-finetune/Learning Rate": 1.9590450187873403e-05, "Full-finetune/Loss": 0.85252845287323, "Full-finetune/Loss (Raw)": 0.9130229949951172, "Full-finetune/Step": 616, "Full-finetune/Step Time": 6.719782751053572} {"Full-finetune/Learning Rate": 1.95886837668594e-05, "Full-finetune/Loss": 0.8536311388015747, "Full-finetune/Loss (Raw)": 0.9283221364021301, "Full-finetune/Step": 617, "Full-finetune/Step Time": 6.723567631095648} {"Full-finetune/Learning Rate": 1.958691362466443e-05, "Full-finetune/Loss": 0.8548146486282349, "Full-finetune/Loss (Raw)": 0.9007179737091064, "Full-finetune/Step": 618, "Full-finetune/Step Time": 6.7172815185040236} {"Full-finetune/Learning Rate": 1.958513976197545e-05, "Full-finetune/Loss": 0.8572883605957031, "Full-finetune/Loss (Raw)": 0.8088608384132385, "Full-finetune/Step": 619, "Full-finetune/Step Time": 6.711498998105526} {"Full-finetune/Learning Rate": 1.9583362179480866e-05, "Full-finetune/Loss": 0.858267605304718, "Full-finetune/Loss (Raw)": 0.9330494403839111, "Full-finetune/Step": 620, "Full-finetune/Step Time": 6.702445484697819} {"Full-finetune/Learning Rate": 1.9581580877870514e-05, "Full-finetune/Loss": 0.8583498597145081, "Full-finetune/Loss (Raw)": 0.9293292164802551, "Full-finetune/Step": 621, "Full-finetune/Step Time": 6.720837343484163} {"Full-finetune/Learning Rate": 1.9579795857835692e-05, "Full-finetune/Loss": 0.858471155166626, "Full-finetune/Loss (Raw)": 0.8705815076828003, "Full-finetune/Step": 622, "Full-finetune/Step Time": 6.718366550281644} {"Full-finetune/Learning Rate": 1.9578007120069123e-05, "Full-finetune/Loss": 0.8593345284461975, "Full-finetune/Loss (Raw)": 0.9661669731140137, "Full-finetune/Step": 623, "Full-finetune/Step Time": 6.7195282317698} {"Full-finetune/Learning Rate": 1.957621466526499e-05, "Full-finetune/Loss": 0.8592514395713806, "Full-finetune/Loss (Raw)": 0.9188725352287292, "Full-finetune/Step": 624, "Full-finetune/Step Time": 6.705359498038888} {"Full-finetune/Learning Rate": 1.9574418494118905e-05, "Full-finetune/Loss": 0.8599972128868103, "Full-finetune/Loss (Raw)": 0.9511741995811462, "Full-finetune/Step": 625, "Full-finetune/Step Time": 6.704165609553456} {"Full-finetune/Learning Rate": 1.9572618607327928e-05, "Full-finetune/Loss": 0.8601869344711304, "Full-finetune/Loss (Raw)": 0.9731085896492004, "Full-finetune/Step": 626, "Full-finetune/Step Time": 6.70162994787097} {"Full-finetune/Learning Rate": 1.9570815005590556e-05, "Full-finetune/Loss": 0.8609497547149658, "Full-finetune/Loss (Raw)": 0.9190852642059326, "Full-finetune/Step": 627, "Full-finetune/Step Time": 6.702975522726774} {"Full-finetune/Learning Rate": 1.9569007689606735e-05, "Full-finetune/Loss": 0.8610227108001709, "Full-finetune/Loss (Raw)": 0.8606173396110535, "Full-finetune/Step": 628, "Full-finetune/Step Time": 6.693527106195688} {"Full-finetune/Learning Rate": 1.9567196660077848e-05, "Full-finetune/Loss": 0.8631223440170288, "Full-finetune/Loss (Raw)": 0.9394561648368835, "Full-finetune/Step": 629, "Full-finetune/Step Time": 6.682243958115578} {"Full-finetune/Learning Rate": 1.9565381917706726e-05, "Full-finetune/Loss": 0.8660023808479309, "Full-finetune/Loss (Raw)": 1.0890932083129883, "Full-finetune/Step": 630, "Full-finetune/Step Time": 6.651925805956125} {"Full-finetune/Learning Rate": 1.9563563463197625e-05, "Full-finetune/Loss": 0.8658947348594666, "Full-finetune/Loss (Raw)": 0.7401291728019714, "Full-finetune/Step": 631, "Full-finetune/Step Time": 6.637669790536165} {"Full-finetune/Learning Rate": 1.9561741297256257e-05, "Full-finetune/Loss": 0.8666543364524841, "Full-finetune/Loss (Raw)": 1.005852460861206, "Full-finetune/Step": 632, "Full-finetune/Step Time": 6.63489911518991} {"Full-finetune/Learning Rate": 1.955991542058977e-05, "Full-finetune/Loss": 0.86835777759552, "Full-finetune/Loss (Raw)": 1.1308770179748535, "Full-finetune/Step": 633, "Full-finetune/Step Time": 6.6370096653699875} {"Full-finetune/Learning Rate": 1.9558085833906753e-05, "Full-finetune/Loss": 0.8700895309448242, "Full-finetune/Loss (Raw)": 0.8159685730934143, "Full-finetune/Step": 634, "Full-finetune/Step Time": 6.627021772786975} {"Full-finetune/Learning Rate": 1.9556252537917228e-05, "Full-finetune/Loss": 0.8701005578041077, "Full-finetune/Loss (Raw)": 0.9229241609573364, "Full-finetune/Step": 635, "Full-finetune/Step Time": 6.630589794367552} {"Full-finetune/Learning Rate": 1.955441553333266e-05, "Full-finetune/Loss": 0.870232343673706, "Full-finetune/Loss (Raw)": 0.8891075849533081, "Full-finetune/Step": 636, "Full-finetune/Step Time": 6.654697638005018} {"Full-finetune/Learning Rate": 1.955257482086596e-05, "Full-finetune/Loss": 0.8713796734809875, "Full-finetune/Loss (Raw)": 0.9819875359535217, "Full-finetune/Step": 637, "Full-finetune/Step Time": 6.643902441486716} {"Full-finetune/Learning Rate": 1.955073040123147e-05, "Full-finetune/Loss": 0.8727753162384033, "Full-finetune/Loss (Raw)": 0.9973928332328796, "Full-finetune/Step": 638, "Full-finetune/Step Time": 6.644949954003096} {"Full-finetune/Learning Rate": 1.9548882275144974e-05, "Full-finetune/Loss": 0.8723738789558411, "Full-finetune/Loss (Raw)": 0.8978611826896667, "Full-finetune/Step": 639, "Full-finetune/Step Time": 6.640916086733341} {"Full-finetune/Learning Rate": 1.9547030443323695e-05, "Full-finetune/Loss": 0.8726650476455688, "Full-finetune/Loss (Raw)": 0.9077001214027405, "Full-finetune/Step": 640, "Full-finetune/Step Time": 6.65097551047802} {"Full-finetune/Learning Rate": 1.9545174906486287e-05, "Full-finetune/Loss": 0.8714081048965454, "Full-finetune/Loss (Raw)": 0.8226015567779541, "Full-finetune/Step": 641, "Full-finetune/Step Time": 6.657209811732173} {"Full-finetune/Learning Rate": 1.9543315665352855e-05, "Full-finetune/Loss": 0.8671703338623047, "Full-finetune/Loss (Raw)": 0.36557716131210327, "Full-finetune/Step": 642, "Full-finetune/Step Time": 6.653022713959217} {"Full-finetune/Learning Rate": 1.954145272064493e-05, "Full-finetune/Loss": 0.8710308074951172, "Full-finetune/Loss (Raw)": 0.8988599181175232, "Full-finetune/Step": 643, "Full-finetune/Step Time": 6.635888032615185} {"Full-finetune/Learning Rate": 1.9539586073085482e-05, "Full-finetune/Loss": 0.8709171414375305, "Full-finetune/Loss (Raw)": 0.9351073503494263, "Full-finetune/Step": 644, "Full-finetune/Step Time": 6.618638850748539} {"Full-finetune/Learning Rate": 1.9537715723398926e-05, "Full-finetune/Loss": 0.8713061809539795, "Full-finetune/Loss (Raw)": 0.9985901117324829, "Full-finetune/Step": 645, "Full-finetune/Step Time": 6.616385743021965} {"Full-finetune/Learning Rate": 1.9535841672311106e-05, "Full-finetune/Loss": 0.8713122010231018, "Full-finetune/Loss (Raw)": 0.9128015041351318, "Full-finetune/Step": 646, "Full-finetune/Step Time": 6.626788990572095} {"Full-finetune/Learning Rate": 1.9533963920549307e-05, "Full-finetune/Loss": 0.8705600500106812, "Full-finetune/Loss (Raw)": 0.8425478339195251, "Full-finetune/Step": 647, "Full-finetune/Step Time": 6.625856192782521} {"Full-finetune/Learning Rate": 1.9532082468842243e-05, "Full-finetune/Loss": 0.872038722038269, "Full-finetune/Loss (Raw)": 0.9442506432533264, "Full-finetune/Step": 648, "Full-finetune/Step Time": 6.623714132234454} {"Full-finetune/Learning Rate": 1.953019731792007e-05, "Full-finetune/Loss": 0.8709365129470825, "Full-finetune/Loss (Raw)": 0.8057281970977783, "Full-finetune/Step": 649, "Full-finetune/Step Time": 6.618163088336587} {"Full-finetune/Learning Rate": 1.9528308468514383e-05, "Full-finetune/Loss": 0.8703867793083191, "Full-finetune/Loss (Raw)": 0.7876431345939636, "Full-finetune/Step": 650, "Full-finetune/Step Time": 6.631102377548814} {"Full-finetune/Learning Rate": 1.9526415921358202e-05, "Full-finetune/Loss": 0.8723310232162476, "Full-finetune/Loss (Raw)": 0.9141985177993774, "Full-finetune/Step": 651, "Full-finetune/Step Time": 6.639563020318747} {"Full-finetune/Learning Rate": 1.952451967718599e-05, "Full-finetune/Loss": 0.8721847534179688, "Full-finetune/Loss (Raw)": 0.8792902827262878, "Full-finetune/Step": 652, "Full-finetune/Step Time": 6.643463084474206} {"Full-finetune/Learning Rate": 1.952261973673364e-05, "Full-finetune/Loss": 0.8715882897377014, "Full-finetune/Loss (Raw)": 0.855440080165863, "Full-finetune/Step": 653, "Full-finetune/Step Time": 6.646065900102258} {"Full-finetune/Learning Rate": 1.9520716100738487e-05, "Full-finetune/Loss": 0.8742015361785889, "Full-finetune/Loss (Raw)": 1.148519515991211, "Full-finetune/Step": 654, "Full-finetune/Step Time": 6.638360822573304} {"Full-finetune/Learning Rate": 1.9518808769939287e-05, "Full-finetune/Loss": 0.8744193315505981, "Full-finetune/Loss (Raw)": 0.8439809679985046, "Full-finetune/Step": 655, "Full-finetune/Step Time": 6.646764263510704} {"Full-finetune/Learning Rate": 1.9516897745076243e-05, "Full-finetune/Loss": 0.8737502694129944, "Full-finetune/Loss (Raw)": 0.8326743841171265, "Full-finetune/Step": 656, "Full-finetune/Step Time": 6.65016426704824} {"Full-finetune/Learning Rate": 1.9514983026890988e-05, "Full-finetune/Loss": 0.8730371594429016, "Full-finetune/Loss (Raw)": 0.8012666702270508, "Full-finetune/Step": 657, "Full-finetune/Step Time": 6.6703645046800375} {"Full-finetune/Learning Rate": 1.9513064616126588e-05, "Full-finetune/Loss": 0.8727271556854248, "Full-finetune/Loss (Raw)": 0.9207251667976379, "Full-finetune/Step": 658, "Full-finetune/Step Time": 6.67217449657619} {"Full-finetune/Learning Rate": 1.951114251352753e-05, "Full-finetune/Loss": 0.8700894117355347, "Full-finetune/Loss (Raw)": 0.6816310286521912, "Full-finetune/Step": 659, "Full-finetune/Step Time": 6.693783299997449} {"Full-finetune/Learning Rate": 1.950921671983976e-05, "Full-finetune/Loss": 0.871740460395813, "Full-finetune/Loss (Raw)": 1.0255661010742188, "Full-finetune/Step": 660, "Full-finetune/Step Time": 6.699140714481473} {"Full-finetune/Learning Rate": 1.9507287235810625e-05, "Full-finetune/Loss": 0.8720272779464722, "Full-finetune/Loss (Raw)": 0.9097810387611389, "Full-finetune/Step": 661, "Full-finetune/Step Time": 6.691034900024533} {"Full-finetune/Learning Rate": 1.9505354062188934e-05, "Full-finetune/Loss": 0.8728032112121582, "Full-finetune/Loss (Raw)": 0.9672086238861084, "Full-finetune/Step": 662, "Full-finetune/Step Time": 6.680789150297642} {"Full-finetune/Learning Rate": 1.9503417199724904e-05, "Full-finetune/Loss": 0.874262809753418, "Full-finetune/Loss (Raw)": 0.9787219762802124, "Full-finetune/Step": 663, "Full-finetune/Step Time": 6.670794360339642} {"Full-finetune/Learning Rate": 1.95014766491702e-05, "Full-finetune/Loss": 0.8747463226318359, "Full-finetune/Loss (Raw)": 0.8167538642883301, "Full-finetune/Step": 664, "Full-finetune/Step Time": 6.658007802441716} {"Full-finetune/Learning Rate": 1.9499532411277907e-05, "Full-finetune/Loss": 0.8753151297569275, "Full-finetune/Loss (Raw)": 0.9842438697814941, "Full-finetune/Step": 665, "Full-finetune/Step Time": 6.664836773648858} {"Full-finetune/Learning Rate": 1.9497584486802548e-05, "Full-finetune/Loss": 0.8765174746513367, "Full-finetune/Loss (Raw)": 1.0644549131393433, "Full-finetune/Step": 666, "Full-finetune/Step Time": 6.657277325168252} {"Full-finetune/Learning Rate": 1.9495632876500076e-05, "Full-finetune/Loss": 0.8762816190719604, "Full-finetune/Loss (Raw)": 0.9684615731239319, "Full-finetune/Step": 667, "Full-finetune/Step Time": 6.670623013749719} {"Full-finetune/Learning Rate": 1.949367758112787e-05, "Full-finetune/Loss": 0.8753494024276733, "Full-finetune/Loss (Raw)": 0.7482433915138245, "Full-finetune/Step": 668, "Full-finetune/Step Time": 6.675702635198832} {"Full-finetune/Learning Rate": 1.949171860144474e-05, "Full-finetune/Loss": 0.8757643699645996, "Full-finetune/Loss (Raw)": 0.9373337030410767, "Full-finetune/Step": 669, "Full-finetune/Step Time": 6.669774521142244} {"Full-finetune/Learning Rate": 1.9489755938210936e-05, "Full-finetune/Loss": 0.8753643035888672, "Full-finetune/Loss (Raw)": 0.8384150266647339, "Full-finetune/Step": 670, "Full-finetune/Step Time": 6.672911899164319} {"Full-finetune/Learning Rate": 1.9487789592188123e-05, "Full-finetune/Loss": 0.8743525147438049, "Full-finetune/Loss (Raw)": 0.8805673718452454, "Full-finetune/Step": 671, "Full-finetune/Step Time": 6.672137528657913} {"Full-finetune/Learning Rate": 1.9485819564139404e-05, "Full-finetune/Loss": 0.8743999004364014, "Full-finetune/Loss (Raw)": 0.9518944025039673, "Full-finetune/Step": 672, "Full-finetune/Step Time": 6.668563993647695} {"Full-finetune/Learning Rate": 1.9483845854829304e-05, "Full-finetune/Loss": 0.8754441738128662, "Full-finetune/Loss (Raw)": 1.0500832796096802, "Full-finetune/Step": 673, "Full-finetune/Step Time": 6.670172480866313} {"Full-finetune/Learning Rate": 1.9481868465023788e-05, "Full-finetune/Loss": 0.8769041299819946, "Full-finetune/Loss (Raw)": 1.0476093292236328, "Full-finetune/Step": 674, "Full-finetune/Step Time": 6.670855583623052} {"Full-finetune/Learning Rate": 1.9479887395490234e-05, "Full-finetune/Loss": 0.8766140937805176, "Full-finetune/Loss (Raw)": 0.9492359757423401, "Full-finetune/Step": 675, "Full-finetune/Step Time": 6.673138957470655} {"Full-finetune/Learning Rate": 1.9477902646997464e-05, "Full-finetune/Loss": 0.8782373666763306, "Full-finetune/Loss (Raw)": 0.9599217772483826, "Full-finetune/Step": 676, "Full-finetune/Step Time": 6.667989902198315} {"Full-finetune/Learning Rate": 1.947591422031572e-05, "Full-finetune/Loss": 0.8790719509124756, "Full-finetune/Loss (Raw)": 0.7376959323883057, "Full-finetune/Step": 677, "Full-finetune/Step Time": 6.708986742421985} {"Full-finetune/Learning Rate": 1.9473922116216664e-05, "Full-finetune/Loss": 0.8803682327270508, "Full-finetune/Loss (Raw)": 1.0034228563308716, "Full-finetune/Step": 678, "Full-finetune/Step Time": 6.7319683115929365} {"Full-finetune/Learning Rate": 1.9471926335473396e-05, "Full-finetune/Loss": 0.8818721771240234, "Full-finetune/Loss (Raw)": 1.0072929859161377, "Full-finetune/Step": 679, "Full-finetune/Step Time": 6.736992973834276} {"Full-finetune/Learning Rate": 1.9469926878860447e-05, "Full-finetune/Loss": 0.8805195689201355, "Full-finetune/Loss (Raw)": 0.8380932807922363, "Full-finetune/Step": 680, "Full-finetune/Step Time": 6.734489481896162} {"Full-finetune/Learning Rate": 1.9467923747153756e-05, "Full-finetune/Loss": 0.8827036619186401, "Full-finetune/Loss (Raw)": 0.9663926362991333, "Full-finetune/Step": 681, "Full-finetune/Step Time": 6.733202317729592} {"Full-finetune/Learning Rate": 1.9465916941130705e-05, "Full-finetune/Loss": 0.8827311992645264, "Full-finetune/Loss (Raw)": 0.9692046046257019, "Full-finetune/Step": 682, "Full-finetune/Step Time": 6.730264902114868} {"Full-finetune/Learning Rate": 1.9463906461570098e-05, "Full-finetune/Loss": 0.8839001655578613, "Full-finetune/Loss (Raw)": 1.018175482749939, "Full-finetune/Step": 683, "Full-finetune/Step Time": 6.711742339655757} {"Full-finetune/Learning Rate": 1.9461892309252158e-05, "Full-finetune/Loss": 0.8835985064506531, "Full-finetune/Loss (Raw)": 0.8247389793395996, "Full-finetune/Step": 684, "Full-finetune/Step Time": 6.700812220573425} {"Full-finetune/Learning Rate": 1.945987448495854e-05, "Full-finetune/Loss": 0.8820213079452515, "Full-finetune/Loss (Raw)": 0.642874002456665, "Full-finetune/Step": 685, "Full-finetune/Step Time": 6.716233558952808} {"Full-finetune/Learning Rate": 1.945785298947233e-05, "Full-finetune/Loss": 0.8854691982269287, "Full-finetune/Loss (Raw)": 0.7205937504768372, "Full-finetune/Step": 686, "Full-finetune/Step Time": 6.712009411305189} {"Full-finetune/Learning Rate": 1.9455827823578017e-05, "Full-finetune/Loss": 0.8854728937149048, "Full-finetune/Loss (Raw)": 0.8737141489982605, "Full-finetune/Step": 687, "Full-finetune/Step Time": 6.707953190430999} {"Full-finetune/Learning Rate": 1.9453798988061535e-05, "Full-finetune/Loss": 0.8865221738815308, "Full-finetune/Loss (Raw)": 0.8948253393173218, "Full-finetune/Step": 688, "Full-finetune/Step Time": 6.707727540284395} {"Full-finetune/Learning Rate": 1.9451766483710236e-05, "Full-finetune/Loss": 0.8843337297439575, "Full-finetune/Loss (Raw)": 0.5162238478660583, "Full-finetune/Step": 689, "Full-finetune/Step Time": 6.709850030019879} {"Full-finetune/Learning Rate": 1.94497303113129e-05, "Full-finetune/Loss": 0.8834297060966492, "Full-finetune/Loss (Raw)": 0.6829272508621216, "Full-finetune/Step": 690, "Full-finetune/Step Time": 6.744597801938653} {"Full-finetune/Learning Rate": 1.9447690471659715e-05, "Full-finetune/Loss": 0.8837008476257324, "Full-finetune/Loss (Raw)": 0.9768184423446655, "Full-finetune/Step": 691, "Full-finetune/Step Time": 6.756783649325371} {"Full-finetune/Learning Rate": 1.944564696554231e-05, "Full-finetune/Loss": 0.8827176094055176, "Full-finetune/Loss (Raw)": 0.7833421230316162, "Full-finetune/Step": 692, "Full-finetune/Step Time": 6.775100288912654} {"Full-finetune/Learning Rate": 1.944359979375373e-05, "Full-finetune/Loss": 0.8829619288444519, "Full-finetune/Loss (Raw)": 0.7684372067451477, "Full-finetune/Step": 693, "Full-finetune/Step Time": 6.775052251294255} {"Full-finetune/Learning Rate": 1.9441548957088438e-05, "Full-finetune/Loss": 0.8783353567123413, "Full-finetune/Loss (Raw)": 0.4030952751636505, "Full-finetune/Step": 694, "Full-finetune/Step Time": 6.79166716709733} {"Full-finetune/Learning Rate": 1.9439494456342327e-05, "Full-finetune/Loss": 0.8774660229682922, "Full-finetune/Loss (Raw)": 0.6953359246253967, "Full-finetune/Step": 695, "Full-finetune/Step Time": 6.79971625469625} {"Full-finetune/Learning Rate": 1.943743629231271e-05, "Full-finetune/Loss": 0.8830149173736572, "Full-finetune/Loss (Raw)": 0.9867987632751465, "Full-finetune/Step": 696, "Full-finetune/Step Time": 6.788089141249657} {"Full-finetune/Learning Rate": 1.9435374465798318e-05, "Full-finetune/Loss": 0.8828064203262329, "Full-finetune/Loss (Raw)": 0.795012354850769, "Full-finetune/Step": 697, "Full-finetune/Step Time": 6.782833576202393} {"Full-finetune/Learning Rate": 1.9433308977599306e-05, "Full-finetune/Loss": 0.8839653730392456, "Full-finetune/Loss (Raw)": 1.0432231426239014, "Full-finetune/Step": 698, "Full-finetune/Step Time": 6.77685571834445} {"Full-finetune/Learning Rate": 1.943123982851725e-05, "Full-finetune/Loss": 0.886295735836029, "Full-finetune/Loss (Raw)": 1.0921558141708374, "Full-finetune/Step": 699, "Full-finetune/Step Time": 6.773602697998285} {"Full-finetune/Learning Rate": 1.9429167019355143e-05, "Full-finetune/Loss": 0.8854070901870728, "Full-finetune/Loss (Raw)": 0.7967560291290283, "Full-finetune/Step": 700, "Full-finetune/Step Time": 6.792885737493634} {"Full-finetune/Learning Rate": 1.9427090550917407e-05, "Full-finetune/Loss": 0.8799651861190796, "Full-finetune/Loss (Raw)": 0.31759828329086304, "Full-finetune/Step": 701, "Full-finetune/Step Time": 6.804998628795147} {"Full-finetune/Learning Rate": 1.9425010424009876e-05, "Full-finetune/Loss": 0.8789148330688477, "Full-finetune/Loss (Raw)": 0.871009111404419, "Full-finetune/Step": 702, "Full-finetune/Step Time": 6.81356742978096} {"Full-finetune/Learning Rate": 1.9422926639439806e-05, "Full-finetune/Loss": 0.8767334818840027, "Full-finetune/Loss (Raw)": 0.7090597748756409, "Full-finetune/Step": 703, "Full-finetune/Step Time": 6.803621098399162} {"Full-finetune/Learning Rate": 1.9420839198015878e-05, "Full-finetune/Loss": 0.8772122859954834, "Full-finetune/Loss (Raw)": 0.9433513283729553, "Full-finetune/Step": 704, "Full-finetune/Step Time": 6.815958766266704} {"Full-finetune/Learning Rate": 1.941874810054818e-05, "Full-finetune/Loss": 0.8775926828384399, "Full-finetune/Loss (Raw)": 0.9265382885932922, "Full-finetune/Step": 705, "Full-finetune/Step Time": 6.815025890246034} {"Full-finetune/Learning Rate": 1.941665334784823e-05, "Full-finetune/Loss": 0.876539945602417, "Full-finetune/Loss (Raw)": 0.6658656001091003, "Full-finetune/Step": 706, "Full-finetune/Step Time": 6.816932071000338} {"Full-finetune/Learning Rate": 1.9414554940728963e-05, "Full-finetune/Loss": 0.8768561482429504, "Full-finetune/Loss (Raw)": 0.795272946357727, "Full-finetune/Step": 707, "Full-finetune/Step Time": 6.792011586949229} {"Full-finetune/Learning Rate": 1.9412452880004728e-05, "Full-finetune/Loss": 0.877673864364624, "Full-finetune/Loss (Raw)": 1.0474746227264404, "Full-finetune/Step": 708, "Full-finetune/Step Time": 6.797527505084872} {"Full-finetune/Learning Rate": 1.941034716649129e-05, "Full-finetune/Loss": 0.8758761882781982, "Full-finetune/Loss (Raw)": 0.7885755896568298, "Full-finetune/Step": 709, "Full-finetune/Step Time": 6.797278286889195} {"Full-finetune/Learning Rate": 1.9408237801005837e-05, "Full-finetune/Loss": 0.8744134306907654, "Full-finetune/Loss (Raw)": 0.8096034526824951, "Full-finetune/Step": 710, "Full-finetune/Step Time": 6.785404995083809} {"Full-finetune/Learning Rate": 1.9406124784366976e-05, "Full-finetune/Loss": 0.8755061626434326, "Full-finetune/Loss (Raw)": 0.9163022041320801, "Full-finetune/Step": 711, "Full-finetune/Step Time": 6.798868993297219} {"Full-finetune/Learning Rate": 1.9404008117394726e-05, "Full-finetune/Loss": 0.8734726905822754, "Full-finetune/Loss (Raw)": 0.6130661964416504, "Full-finetune/Step": 712, "Full-finetune/Step Time": 6.808555863797665} {"Full-finetune/Learning Rate": 1.9401887800910525e-05, "Full-finetune/Loss": 0.873281717300415, "Full-finetune/Loss (Raw)": 0.9316616654396057, "Full-finetune/Step": 713, "Full-finetune/Step Time": 6.808987472206354} {"Full-finetune/Learning Rate": 1.939976383573722e-05, "Full-finetune/Loss": 0.8763734102249146, "Full-finetune/Loss (Raw)": 0.743635892868042, "Full-finetune/Step": 714, "Full-finetune/Step Time": 6.796059675514698} {"Full-finetune/Learning Rate": 1.9397636222699088e-05, "Full-finetune/Loss": 0.875914454460144, "Full-finetune/Loss (Raw)": 0.9983857274055481, "Full-finetune/Step": 715, "Full-finetune/Step Time": 6.7929502800107} {"Full-finetune/Learning Rate": 1.939550496262181e-05, "Full-finetune/Loss": 0.8759675025939941, "Full-finetune/Loss (Raw)": 0.7863060832023621, "Full-finetune/Step": 716, "Full-finetune/Step Time": 6.768048355355859} {"Full-finetune/Learning Rate": 1.939337005633249e-05, "Full-finetune/Loss": 0.8756173849105835, "Full-finetune/Loss (Raw)": 0.9512439966201782, "Full-finetune/Step": 717, "Full-finetune/Step Time": 6.779432523995638} {"Full-finetune/Learning Rate": 1.939123150465964e-05, "Full-finetune/Loss": 0.8748366832733154, "Full-finetune/Loss (Raw)": 0.90989089012146, "Full-finetune/Step": 718, "Full-finetune/Step Time": 6.782678401097655} {"Full-finetune/Learning Rate": 1.938908930843319e-05, "Full-finetune/Loss": 0.8758590221405029, "Full-finetune/Loss (Raw)": 0.932544469833374, "Full-finetune/Step": 719, "Full-finetune/Step Time": 6.782498311251402} {"Full-finetune/Learning Rate": 1.9386943468484487e-05, "Full-finetune/Loss": 0.877136766910553, "Full-finetune/Loss (Raw)": 1.0527862310409546, "Full-finetune/Step": 720, "Full-finetune/Step Time": 6.783980106934905} {"Full-finetune/Learning Rate": 1.938479398564629e-05, "Full-finetune/Loss": 0.8768285512924194, "Full-finetune/Loss (Raw)": 0.831876814365387, "Full-finetune/Step": 721, "Full-finetune/Step Time": 6.786798670887947} {"Full-finetune/Learning Rate": 1.938264086075277e-05, "Full-finetune/Loss": 0.8761755228042603, "Full-finetune/Loss (Raw)": 0.8428950309753418, "Full-finetune/Step": 722, "Full-finetune/Step Time": 6.811811093240976} {"Full-finetune/Learning Rate": 1.938048409463951e-05, "Full-finetune/Loss": 0.876754641532898, "Full-finetune/Loss (Raw)": 1.0227553844451904, "Full-finetune/Step": 723, "Full-finetune/Step Time": 6.808055836707354} {"Full-finetune/Learning Rate": 1.9378323688143515e-05, "Full-finetune/Loss": 0.8775078058242798, "Full-finetune/Loss (Raw)": 0.9237650036811829, "Full-finetune/Step": 724, "Full-finetune/Step Time": 6.822295628488064} {"Full-finetune/Learning Rate": 1.937615964210319e-05, "Full-finetune/Loss": 0.8798485398292542, "Full-finetune/Loss (Raw)": 1.07377028465271, "Full-finetune/Step": 725, "Full-finetune/Step Time": 6.8208964597433805} {"Full-finetune/Learning Rate": 1.9373991957358362e-05, "Full-finetune/Loss": 0.8776437044143677, "Full-finetune/Loss (Raw)": 0.5953267216682434, "Full-finetune/Step": 726, "Full-finetune/Step Time": 6.80818597599864} {"Full-finetune/Learning Rate": 1.9371820634750266e-05, "Full-finetune/Loss": 0.8773103356361389, "Full-finetune/Loss (Raw)": 0.9306823015213013, "Full-finetune/Step": 727, "Full-finetune/Step Time": 6.8140164371579885} {"Full-finetune/Learning Rate": 1.9369645675121554e-05, "Full-finetune/Loss": 0.8776887059211731, "Full-finetune/Loss (Raw)": 1.014297366142273, "Full-finetune/Step": 728, "Full-finetune/Step Time": 6.812305293977261} {"Full-finetune/Learning Rate": 1.936746707931628e-05, "Full-finetune/Loss": 0.8753164410591125, "Full-finetune/Loss (Raw)": 0.4970037639141083, "Full-finetune/Step": 729, "Full-finetune/Step Time": 6.811735715717077} {"Full-finetune/Learning Rate": 1.9365284848179917e-05, "Full-finetune/Loss": 0.8754279613494873, "Full-finetune/Loss (Raw)": 0.9531574249267578, "Full-finetune/Step": 730, "Full-finetune/Step Time": 6.814428508281708} {"Full-finetune/Learning Rate": 1.9363098982559346e-05, "Full-finetune/Loss": 0.8771618604660034, "Full-finetune/Loss (Raw)": 0.9438592791557312, "Full-finetune/Step": 731, "Full-finetune/Step Time": 6.81048228032887} {"Full-finetune/Learning Rate": 1.9360909483302862e-05, "Full-finetune/Loss": 0.8756953477859497, "Full-finetune/Loss (Raw)": 0.7783855199813843, "Full-finetune/Step": 732, "Full-finetune/Step Time": 6.839879456907511} {"Full-finetune/Learning Rate": 1.9358716351260164e-05, "Full-finetune/Loss": 0.8764939308166504, "Full-finetune/Loss (Raw)": 0.8663469552993774, "Full-finetune/Step": 733, "Full-finetune/Step Time": 6.836892496794462} {"Full-finetune/Learning Rate": 1.935651958728236e-05, "Full-finetune/Loss": 0.8763123750686646, "Full-finetune/Loss (Raw)": 0.766697883605957, "Full-finetune/Step": 734, "Full-finetune/Step Time": 6.824986677616835} {"Full-finetune/Learning Rate": 1.9354319192221975e-05, "Full-finetune/Loss": 0.877083957195282, "Full-finetune/Loss (Raw)": 0.7393813133239746, "Full-finetune/Step": 735, "Full-finetune/Step Time": 6.8013945966959} {"Full-finetune/Learning Rate": 1.9352115166932943e-05, "Full-finetune/Loss": 0.8759324550628662, "Full-finetune/Loss (Raw)": 0.8645594120025635, "Full-finetune/Step": 736, "Full-finetune/Step Time": 6.835632938891649} {"Full-finetune/Learning Rate": 1.9349907512270597e-05, "Full-finetune/Loss": 0.8775752782821655, "Full-finetune/Loss (Raw)": 1.0128722190856934, "Full-finetune/Step": 737, "Full-finetune/Step Time": 6.8307027611881495} {"Full-finetune/Learning Rate": 1.9347696229091685e-05, "Full-finetune/Loss": 0.8777766823768616, "Full-finetune/Loss (Raw)": 0.9346588850021362, "Full-finetune/Step": 738, "Full-finetune/Step Time": 6.827082805335522} {"Full-finetune/Learning Rate": 1.934548131825437e-05, "Full-finetune/Loss": 0.8743009567260742, "Full-finetune/Loss (Raw)": 0.46520864963531494, "Full-finetune/Step": 739, "Full-finetune/Step Time": 6.848149701952934} {"Full-finetune/Learning Rate": 1.9343262780618206e-05, "Full-finetune/Loss": 0.8738194704055786, "Full-finetune/Loss (Raw)": 1.0258440971374512, "Full-finetune/Step": 740, "Full-finetune/Step Time": 6.850126096978784} {"Full-finetune/Learning Rate": 1.934104061704417e-05, "Full-finetune/Loss": 0.8733808994293213, "Full-finetune/Loss (Raw)": 0.6710876226425171, "Full-finetune/Step": 741, "Full-finetune/Step Time": 6.877922158688307} {"Full-finetune/Learning Rate": 1.9338814828394638e-05, "Full-finetune/Loss": 0.8728783130645752, "Full-finetune/Loss (Raw)": 0.8750598430633545, "Full-finetune/Step": 742, "Full-finetune/Step Time": 6.882995277643204} {"Full-finetune/Learning Rate": 1.9336585415533398e-05, "Full-finetune/Loss": 0.8736903667449951, "Full-finetune/Loss (Raw)": 0.9943608641624451, "Full-finetune/Step": 743, "Full-finetune/Step Time": 6.8770137540996075} {"Full-finetune/Learning Rate": 1.9334352379325638e-05, "Full-finetune/Loss": 0.8739532828330994, "Full-finetune/Loss (Raw)": 0.9466770887374878, "Full-finetune/Step": 744, "Full-finetune/Step Time": 6.877208847552538} {"Full-finetune/Learning Rate": 1.9332115720637957e-05, "Full-finetune/Loss": 0.8725030422210693, "Full-finetune/Loss (Raw)": 0.74269038438797, "Full-finetune/Step": 745, "Full-finetune/Step Time": 6.893472108989954} {"Full-finetune/Learning Rate": 1.9329875440338363e-05, "Full-finetune/Loss": 0.8734130859375, "Full-finetune/Loss (Raw)": 1.017203450202942, "Full-finetune/Step": 746, "Full-finetune/Step Time": 6.883887369185686} {"Full-finetune/Learning Rate": 1.9327631539296258e-05, "Full-finetune/Loss": 0.8749535083770752, "Full-finetune/Loss (Raw)": 1.0060464143753052, "Full-finetune/Step": 747, "Full-finetune/Step Time": 6.863486586138606} {"Full-finetune/Learning Rate": 1.932538401838246e-05, "Full-finetune/Loss": 0.8756179809570312, "Full-finetune/Loss (Raw)": 1.0181000232696533, "Full-finetune/Step": 748, "Full-finetune/Step Time": 6.8645870219916105} {"Full-finetune/Learning Rate": 1.9323132878469185e-05, "Full-finetune/Loss": 0.8747838139533997, "Full-finetune/Loss (Raw)": 0.8225482106208801, "Full-finetune/Step": 749, "Full-finetune/Step Time": 6.853953847661614} {"Full-finetune/Learning Rate": 1.9320878120430063e-05, "Full-finetune/Loss": 0.874285876750946, "Full-finetune/Loss (Raw)": 0.8068439364433289, "Full-finetune/Step": 750, "Full-finetune/Step Time": 6.849300151690841} {"Full-finetune/Learning Rate": 1.9318619745140115e-05, "Full-finetune/Loss": 0.8725593090057373, "Full-finetune/Loss (Raw)": 0.7451683282852173, "Full-finetune/Step": 751, "Full-finetune/Step Time": 6.851885927841067} {"Full-finetune/Learning Rate": 1.9316357753475776e-05, "Full-finetune/Loss": 0.8731423020362854, "Full-finetune/Loss (Raw)": 0.9935020208358765, "Full-finetune/Step": 752, "Full-finetune/Step Time": 6.85291544161737} {"Full-finetune/Learning Rate": 1.9314092146314872e-05, "Full-finetune/Loss": 0.8689035773277283, "Full-finetune/Loss (Raw)": 0.4086095094680786, "Full-finetune/Step": 753, "Full-finetune/Step Time": 6.868752151727676} {"Full-finetune/Learning Rate": 1.9311822924536656e-05, "Full-finetune/Loss": 0.8674704432487488, "Full-finetune/Loss (Raw)": 0.7896722555160522, "Full-finetune/Step": 754, "Full-finetune/Step Time": 6.870030228048563} {"Full-finetune/Learning Rate": 1.930955008902176e-05, "Full-finetune/Loss": 0.8660470247268677, "Full-finetune/Loss (Raw)": 0.7368857860565186, "Full-finetune/Step": 755, "Full-finetune/Step Time": 6.891907079145312} {"Full-finetune/Learning Rate": 1.9307273640652228e-05, "Full-finetune/Loss": 0.8660409450531006, "Full-finetune/Loss (Raw)": 0.859844982624054, "Full-finetune/Step": 756, "Full-finetune/Step Time": 6.888336345553398} {"Full-finetune/Learning Rate": 1.9304993580311505e-05, "Full-finetune/Loss": 0.865162193775177, "Full-finetune/Loss (Raw)": 0.8269674777984619, "Full-finetune/Step": 757, "Full-finetune/Step Time": 6.90130134485662} {"Full-finetune/Learning Rate": 1.9302709908884437e-05, "Full-finetune/Loss": 0.8643548488616943, "Full-finetune/Loss (Raw)": 0.9857580661773682, "Full-finetune/Step": 758, "Full-finetune/Step Time": 6.909598836675286} {"Full-finetune/Learning Rate": 1.9300422627257275e-05, "Full-finetune/Loss": 0.8660046458244324, "Full-finetune/Loss (Raw)": 0.9513011574745178, "Full-finetune/Step": 759, "Full-finetune/Step Time": 6.9090047013014555} {"Full-finetune/Learning Rate": 1.9298131736317666e-05, "Full-finetune/Loss": 0.8652386665344238, "Full-finetune/Loss (Raw)": 0.9078124761581421, "Full-finetune/Step": 760, "Full-finetune/Step Time": 6.903124153614044} {"Full-finetune/Learning Rate": 1.929583723695466e-05, "Full-finetune/Loss": 0.8635446429252625, "Full-finetune/Loss (Raw)": 0.9140324592590332, "Full-finetune/Step": 761, "Full-finetune/Step Time": 6.906578805297613} {"Full-finetune/Learning Rate": 1.9293539130058708e-05, "Full-finetune/Loss": 0.8601487874984741, "Full-finetune/Loss (Raw)": 0.38129886984825134, "Full-finetune/Step": 762, "Full-finetune/Step Time": 6.930897718295455} {"Full-finetune/Learning Rate": 1.9291237416521662e-05, "Full-finetune/Loss": 0.8587383031845093, "Full-finetune/Loss (Raw)": 0.7423863410949707, "Full-finetune/Step": 763, "Full-finetune/Step Time": 6.936092654243112} {"Full-finetune/Learning Rate": 1.9288932097236772e-05, "Full-finetune/Loss": 0.8585456609725952, "Full-finetune/Loss (Raw)": 0.8644512891769409, "Full-finetune/Step": 764, "Full-finetune/Step Time": 6.915059959515929} {"Full-finetune/Learning Rate": 1.9286623173098683e-05, "Full-finetune/Loss": 0.8580049276351929, "Full-finetune/Loss (Raw)": 0.9127653241157532, "Full-finetune/Step": 765, "Full-finetune/Step Time": 6.918447785079479} {"Full-finetune/Learning Rate": 1.928431064500345e-05, "Full-finetune/Loss": 0.8579095005989075, "Full-finetune/Loss (Raw)": 0.9851849675178528, "Full-finetune/Step": 766, "Full-finetune/Step Time": 6.936460888013244} {"Full-finetune/Learning Rate": 1.9281994513848515e-05, "Full-finetune/Loss": 0.8578459024429321, "Full-finetune/Loss (Raw)": 0.8897149562835693, "Full-finetune/Step": 767, "Full-finetune/Step Time": 6.945336852222681} {"Full-finetune/Learning Rate": 1.9279674780532726e-05, "Full-finetune/Loss": 0.8572169542312622, "Full-finetune/Loss (Raw)": 0.8272027373313904, "Full-finetune/Step": 768, "Full-finetune/Step Time": 6.945415696129203} {"Full-finetune/Learning Rate": 1.9277351445956323e-05, "Full-finetune/Loss": 0.85649573802948, "Full-finetune/Loss (Raw)": 0.7302806973457336, "Full-finetune/Step": 769, "Full-finetune/Step Time": 6.954076550900936} {"Full-finetune/Learning Rate": 1.927502451102095e-05, "Full-finetune/Loss": 0.8601592183113098, "Full-finetune/Loss (Raw)": 0.8345038890838623, "Full-finetune/Step": 770, "Full-finetune/Step Time": 6.9448275454342365} {"Full-finetune/Learning Rate": 1.9272693976629642e-05, "Full-finetune/Loss": 0.8609788417816162, "Full-finetune/Loss (Raw)": 1.0037750005722046, "Full-finetune/Step": 771, "Full-finetune/Step Time": 6.94840044900775} {"Full-finetune/Learning Rate": 1.9270359843686837e-05, "Full-finetune/Loss": 0.8608450889587402, "Full-finetune/Loss (Raw)": 0.9179845452308655, "Full-finetune/Step": 772, "Full-finetune/Step Time": 6.9418217074126005} {"Full-finetune/Learning Rate": 1.9268022113098366e-05, "Full-finetune/Loss": 0.8605741262435913, "Full-finetune/Loss (Raw)": 0.963899552822113, "Full-finetune/Step": 773, "Full-finetune/Step Time": 6.940766213461757} {"Full-finetune/Learning Rate": 1.9265680785771455e-05, "Full-finetune/Loss": 0.8591388463973999, "Full-finetune/Loss (Raw)": 0.7290963530540466, "Full-finetune/Step": 774, "Full-finetune/Step Time": 6.9507998041808605} {"Full-finetune/Learning Rate": 1.926333586261473e-05, "Full-finetune/Loss": 0.8599504828453064, "Full-finetune/Loss (Raw)": 0.9464355111122131, "Full-finetune/Step": 775, "Full-finetune/Step Time": 6.9542189706116915} {"Full-finetune/Learning Rate": 1.9260987344538205e-05, "Full-finetune/Loss": 0.857780933380127, "Full-finetune/Loss (Raw)": 0.6665441393852234, "Full-finetune/Step": 776, "Full-finetune/Step Time": 6.955545462667942} {"Full-finetune/Learning Rate": 1.9258635232453296e-05, "Full-finetune/Loss": 0.8592219948768616, "Full-finetune/Loss (Raw)": 0.9901831150054932, "Full-finetune/Step": 777, "Full-finetune/Step Time": 6.972141223028302} {"Full-finetune/Learning Rate": 1.9256279527272815e-05, "Full-finetune/Loss": 0.8608005046844482, "Full-finetune/Loss (Raw)": 0.9896934628486633, "Full-finetune/Step": 778, "Full-finetune/Step Time": 6.9573184587061405} {"Full-finetune/Learning Rate": 1.9253920229910966e-05, "Full-finetune/Loss": 0.8603682518005371, "Full-finetune/Loss (Raw)": 0.8588733673095703, "Full-finetune/Step": 779, "Full-finetune/Step Time": 6.954144036397338} {"Full-finetune/Learning Rate": 1.925155734128334e-05, "Full-finetune/Loss": 0.8561220169067383, "Full-finetune/Loss (Raw)": 0.335769921541214, "Full-finetune/Step": 780, "Full-finetune/Step Time": 6.97085483931005} {"Full-finetune/Learning Rate": 1.9249190862306935e-05, "Full-finetune/Loss": 0.8559045791625977, "Full-finetune/Loss (Raw)": 0.8276070952415466, "Full-finetune/Step": 781, "Full-finetune/Step Time": 6.972007716074586} {"Full-finetune/Learning Rate": 1.9246820793900127e-05, "Full-finetune/Loss": 0.8513734340667725, "Full-finetune/Loss (Raw)": 0.5685426592826843, "Full-finetune/Step": 782, "Full-finetune/Step Time": 7.000972516834736} {"Full-finetune/Learning Rate": 1.92444471369827e-05, "Full-finetune/Loss": 0.8514195680618286, "Full-finetune/Loss (Raw)": 0.849876880645752, "Full-finetune/Step": 783, "Full-finetune/Step Time": 7.0328835770487785} {"Full-finetune/Learning Rate": 1.924206989247582e-05, "Full-finetune/Loss": 0.8496930003166199, "Full-finetune/Loss (Raw)": 0.6116737723350525, "Full-finetune/Step": 784, "Full-finetune/Step Time": 7.0495696645230055} {"Full-finetune/Learning Rate": 1.9239689061302053e-05, "Full-finetune/Loss": 0.8508061170578003, "Full-finetune/Loss (Raw)": 0.9437447786331177, "Full-finetune/Step": 785, "Full-finetune/Step Time": 7.045004377141595} {"Full-finetune/Learning Rate": 1.923730464438535e-05, "Full-finetune/Loss": 0.8513965606689453, "Full-finetune/Loss (Raw)": 0.9963010549545288, "Full-finetune/Step": 786, "Full-finetune/Step Time": 7.0421470664441586} {"Full-finetune/Learning Rate": 1.923491664265106e-05, "Full-finetune/Loss": 0.8530924320220947, "Full-finetune/Loss (Raw)": 0.898703396320343, "Full-finetune/Step": 787, "Full-finetune/Step Time": 7.046683989465237} {"Full-finetune/Learning Rate": 1.9232525057025915e-05, "Full-finetune/Loss": 0.8530179262161255, "Full-finetune/Loss (Raw)": 1.0160367488861084, "Full-finetune/Step": 788, "Full-finetune/Step Time": 7.042856523767114} {"Full-finetune/Learning Rate": 1.9230129888438046e-05, "Full-finetune/Loss": 0.8531444668769836, "Full-finetune/Loss (Raw)": 0.9259696006774902, "Full-finetune/Step": 789, "Full-finetune/Step Time": 7.041921420022845} {"Full-finetune/Learning Rate": 1.922773113781697e-05, "Full-finetune/Loss": 0.8498996496200562, "Full-finetune/Loss (Raw)": 0.5518810153007507, "Full-finetune/Step": 790, "Full-finetune/Step Time": 7.082252886146307} {"Full-finetune/Learning Rate": 1.9225328806093596e-05, "Full-finetune/Loss": 0.8489482402801514, "Full-finetune/Loss (Raw)": 0.8569417595863342, "Full-finetune/Step": 791, "Full-finetune/Step Time": 7.082897998392582} {"Full-finetune/Learning Rate": 1.922292289420022e-05, "Full-finetune/Loss": 0.8485971689224243, "Full-finetune/Loss (Raw)": 0.7718077301979065, "Full-finetune/Step": 792, "Full-finetune/Step Time": 7.074338033795357} {"Full-finetune/Learning Rate": 1.9220513403070534e-05, "Full-finetune/Loss": 0.8476954698562622, "Full-finetune/Loss (Raw)": 0.8688238263130188, "Full-finetune/Step": 793, "Full-finetune/Step Time": 7.076788701117039} {"Full-finetune/Learning Rate": 1.9218100333639608e-05, "Full-finetune/Loss": 0.8458455801010132, "Full-finetune/Loss (Raw)": 0.8276816010475159, "Full-finetune/Step": 794, "Full-finetune/Step Time": 7.078368430957198} {"Full-finetune/Learning Rate": 1.9215683686843912e-05, "Full-finetune/Loss": 0.8454616069793701, "Full-finetune/Loss (Raw)": 0.9193048477172852, "Full-finetune/Step": 795, "Full-finetune/Step Time": 7.082121705636382} {"Full-finetune/Learning Rate": 1.92132634636213e-05, "Full-finetune/Loss": 0.8463842272758484, "Full-finetune/Loss (Raw)": 0.8663380146026611, "Full-finetune/Step": 796, "Full-finetune/Step Time": 7.0888414066284895} {"Full-finetune/Learning Rate": 1.9210839664911013e-05, "Full-finetune/Loss": 0.8458585143089294, "Full-finetune/Loss (Raw)": 0.8700424432754517, "Full-finetune/Step": 797, "Full-finetune/Step Time": 7.095957648009062} {"Full-finetune/Learning Rate": 1.9208412291653674e-05, "Full-finetune/Loss": 0.8467021584510803, "Full-finetune/Loss (Raw)": 0.9464048743247986, "Full-finetune/Step": 798, "Full-finetune/Step Time": 7.096753677353263} {"Full-finetune/Learning Rate": 1.9205981344791307e-05, "Full-finetune/Loss": 0.8460354804992676, "Full-finetune/Loss (Raw)": 0.7952314615249634, "Full-finetune/Step": 799, "Full-finetune/Step Time": 7.103649130091071} {"Full-finetune/Learning Rate": 1.9203546825267313e-05, "Full-finetune/Loss": 0.8455812931060791, "Full-finetune/Loss (Raw)": 0.893754243850708, "Full-finetune/Step": 800, "Full-finetune/Step Time": 7.108833795413375} {"Full-finetune/Learning Rate": 1.9201108734026477e-05, "Full-finetune/Loss": 0.8446508646011353, "Full-finetune/Loss (Raw)": 0.9309909343719482, "Full-finetune/Step": 801, "Full-finetune/Step Time": 7.131795847788453} {"Full-finetune/Learning Rate": 1.9198667072014978e-05, "Full-finetune/Loss": 0.8426843881607056, "Full-finetune/Loss (Raw)": 0.795907735824585, "Full-finetune/Step": 802, "Full-finetune/Step Time": 7.134946070611477} {"Full-finetune/Learning Rate": 1.9196221840180378e-05, "Full-finetune/Loss": 0.8408884406089783, "Full-finetune/Loss (Raw)": 0.7193518877029419, "Full-finetune/Step": 803, "Full-finetune/Step Time": 7.140093920752406} {"Full-finetune/Learning Rate": 1.9193773039471622e-05, "Full-finetune/Loss": 0.8404322862625122, "Full-finetune/Loss (Raw)": 0.9015244841575623, "Full-finetune/Step": 804, "Full-finetune/Step Time": 7.141254980117083} {"Full-finetune/Learning Rate": 1.9191320670839047e-05, "Full-finetune/Loss": 0.8415651917457581, "Full-finetune/Loss (Raw)": 0.8827137351036072, "Full-finetune/Step": 805, "Full-finetune/Step Time": 7.102225547656417} {"Full-finetune/Learning Rate": 1.918886473523436e-05, "Full-finetune/Loss": 0.8390974998474121, "Full-finetune/Loss (Raw)": 0.6875609159469604, "Full-finetune/Step": 806, "Full-finetune/Step Time": 7.104455275461078} {"Full-finetune/Learning Rate": 1.918640523361067e-05, "Full-finetune/Loss": 0.8380627632141113, "Full-finetune/Loss (Raw)": 0.8748438954353333, "Full-finetune/Step": 807, "Full-finetune/Step Time": 7.096772614866495} {"Full-finetune/Learning Rate": 1.9183942166922455e-05, "Full-finetune/Loss": 0.8378922939300537, "Full-finetune/Loss (Raw)": 0.8162752389907837, "Full-finetune/Step": 808, "Full-finetune/Step Time": 7.116413455456495} {"Full-finetune/Learning Rate": 1.9181475536125588e-05, "Full-finetune/Loss": 0.8374289274215698, "Full-finetune/Loss (Raw)": 0.9070751667022705, "Full-finetune/Step": 809, "Full-finetune/Step Time": 7.120090110227466} {"Full-finetune/Learning Rate": 1.917900534217732e-05, "Full-finetune/Loss": 0.8376128673553467, "Full-finetune/Loss (Raw)": 0.9927586317062378, "Full-finetune/Step": 810, "Full-finetune/Step Time": 7.122763043269515} {"Full-finetune/Learning Rate": 1.9176531586036282e-05, "Full-finetune/Loss": 0.8355997800827026, "Full-finetune/Loss (Raw)": 0.7604901790618896, "Full-finetune/Step": 811, "Full-finetune/Step Time": 7.119444513693452} {"Full-finetune/Learning Rate": 1.917405426866249e-05, "Full-finetune/Loss": 0.8375270366668701, "Full-finetune/Loss (Raw)": 1.0714300870895386, "Full-finetune/Step": 812, "Full-finetune/Step Time": 7.116523083299398} {"Full-finetune/Learning Rate": 1.9171573391017348e-05, "Full-finetune/Loss": 0.8394027948379517, "Full-finetune/Loss (Raw)": 0.8829758167266846, "Full-finetune/Step": 813, "Full-finetune/Step Time": 7.103979168459773} {"Full-finetune/Learning Rate": 1.9169088954063634e-05, "Full-finetune/Loss": 0.8405081033706665, "Full-finetune/Loss (Raw)": 0.8620678782463074, "Full-finetune/Step": 814, "Full-finetune/Step Time": 7.092550570145249} {"Full-finetune/Learning Rate": 1.9166600958765508e-05, "Full-finetune/Loss": 0.8404194116592407, "Full-finetune/Loss (Raw)": 0.8623708486557007, "Full-finetune/Step": 815, "Full-finetune/Step Time": 7.096737304702401} {"Full-finetune/Learning Rate": 1.916410940608851e-05, "Full-finetune/Loss": 0.8408545851707458, "Full-finetune/Loss (Raw)": 0.9505213499069214, "Full-finetune/Step": 816, "Full-finetune/Step Time": 7.095227817073464} {"Full-finetune/Learning Rate": 1.9161614296999565e-05, "Full-finetune/Loss": 0.8441210389137268, "Full-finetune/Loss (Raw)": 0.9343318939208984, "Full-finetune/Step": 817, "Full-finetune/Step Time": 7.084493411704898} {"Full-finetune/Learning Rate": 1.9159115632466978e-05, "Full-finetune/Loss": 0.8414667844772339, "Full-finetune/Loss (Raw)": 0.34318047761917114, "Full-finetune/Step": 818, "Full-finetune/Step Time": 7.062947055324912} {"Full-finetune/Learning Rate": 1.9156613413460432e-05, "Full-finetune/Loss": 0.8394861221313477, "Full-finetune/Loss (Raw)": 0.7232955098152161, "Full-finetune/Step": 819, "Full-finetune/Step Time": 7.053463505581021} {"Full-finetune/Learning Rate": 1.9154107640950986e-05, "Full-finetune/Loss": 0.8381549715995789, "Full-finetune/Loss (Raw)": 0.6129511594772339, "Full-finetune/Step": 820, "Full-finetune/Step Time": 7.034302424639463} {"Full-finetune/Learning Rate": 1.9151598315911088e-05, "Full-finetune/Loss": 0.8385113477706909, "Full-finetune/Loss (Raw)": 0.8140587210655212, "Full-finetune/Step": 821, "Full-finetune/Step Time": 7.034596653655171} {"Full-finetune/Learning Rate": 1.9149085439314548e-05, "Full-finetune/Loss": 0.8418198823928833, "Full-finetune/Loss (Raw)": 0.8265862464904785, "Full-finetune/Step": 822, "Full-finetune/Step Time": 7.021464316174388} {"Full-finetune/Learning Rate": 1.9146569012136572e-05, "Full-finetune/Loss": 0.844273567199707, "Full-finetune/Loss (Raw)": 1.0094060897827148, "Full-finetune/Step": 823, "Full-finetune/Step Time": 7.006789894774556} {"Full-finetune/Learning Rate": 1.914404903535373e-05, "Full-finetune/Loss": 0.8437198400497437, "Full-finetune/Loss (Raw)": 0.9159232378005981, "Full-finetune/Step": 824, "Full-finetune/Step Time": 7.004946708679199} {"Full-finetune/Learning Rate": 1.9141525509943987e-05, "Full-finetune/Loss": 0.8440592288970947, "Full-finetune/Loss (Raw)": 0.8384556174278259, "Full-finetune/Step": 825, "Full-finetune/Step Time": 6.996508494019508} {"Full-finetune/Learning Rate": 1.9138998436886664e-05, "Full-finetune/Loss": 0.8416767120361328, "Full-finetune/Loss (Raw)": 0.7382584810256958, "Full-finetune/Step": 826, "Full-finetune/Step Time": 7.002477295696735} {"Full-finetune/Learning Rate": 1.913646781716247e-05, "Full-finetune/Loss": 0.8396120071411133, "Full-finetune/Loss (Raw)": 0.8278672695159912, "Full-finetune/Step": 827, "Full-finetune/Step Time": 6.990068137645721} {"Full-finetune/Learning Rate": 1.9133933651753493e-05, "Full-finetune/Loss": 0.8363897800445557, "Full-finetune/Loss (Raw)": 0.3843143582344055, "Full-finetune/Step": 828, "Full-finetune/Step Time": 6.988292410969734} {"Full-finetune/Learning Rate": 1.913139594164319e-05, "Full-finetune/Loss": 0.8403944373130798, "Full-finetune/Loss (Raw)": 0.8301956653594971, "Full-finetune/Step": 829, "Full-finetune/Step Time": 6.98570491746068} {"Full-finetune/Learning Rate": 1.9128854687816394e-05, "Full-finetune/Loss": 0.8406046032905579, "Full-finetune/Loss (Raw)": 0.8979150056838989, "Full-finetune/Step": 830, "Full-finetune/Step Time": 6.983384259045124} {"Full-finetune/Learning Rate": 1.9126309891259324e-05, "Full-finetune/Loss": 0.841545581817627, "Full-finetune/Loss (Raw)": 0.8295037150382996, "Full-finetune/Step": 831, "Full-finetune/Step Time": 6.989523679018021} {"Full-finetune/Learning Rate": 1.9123761552959562e-05, "Full-finetune/Loss": 0.8418785929679871, "Full-finetune/Loss (Raw)": 0.9859772324562073, "Full-finetune/Step": 832, "Full-finetune/Step Time": 6.971842253580689} {"Full-finetune/Learning Rate": 1.9121209673906064e-05, "Full-finetune/Loss": 0.8379371166229248, "Full-finetune/Loss (Raw)": 0.4220294654369354, "Full-finetune/Step": 833, "Full-finetune/Step Time": 6.991148652508855} {"Full-finetune/Learning Rate": 1.911865425508917e-05, "Full-finetune/Loss": 0.8396616578102112, "Full-finetune/Loss (Raw)": 0.886602520942688, "Full-finetune/Step": 834, "Full-finetune/Step Time": 7.001258995383978} {"Full-finetune/Learning Rate": 1.9116095297500588e-05, "Full-finetune/Loss": 0.840186357498169, "Full-finetune/Loss (Raw)": 0.8624319434165955, "Full-finetune/Step": 835, "Full-finetune/Step Time": 7.007116002961993} {"Full-finetune/Learning Rate": 1.9113532802133394e-05, "Full-finetune/Loss": 0.8389033675193787, "Full-finetune/Loss (Raw)": 0.8832544684410095, "Full-finetune/Step": 836, "Full-finetune/Step Time": 7.001259272918105} {"Full-finetune/Learning Rate": 1.911096676998205e-05, "Full-finetune/Loss": 0.8385360240936279, "Full-finetune/Loss (Raw)": 0.7415536642074585, "Full-finetune/Step": 837, "Full-finetune/Step Time": 7.011283110827208} {"Full-finetune/Learning Rate": 1.9108397202042376e-05, "Full-finetune/Loss": 0.8384395837783813, "Full-finetune/Loss (Raw)": 0.7972689270973206, "Full-finetune/Step": 838, "Full-finetune/Step Time": 7.016455624252558} {"Full-finetune/Learning Rate": 1.910582409931158e-05, "Full-finetune/Loss": 0.8373203873634338, "Full-finetune/Loss (Raw)": 0.7730414271354675, "Full-finetune/Step": 839, "Full-finetune/Step Time": 7.009134827181697} {"Full-finetune/Learning Rate": 1.910324746278822e-05, "Full-finetune/Loss": 0.8406350612640381, "Full-finetune/Loss (Raw)": 1.0373393297195435, "Full-finetune/Step": 840, "Full-finetune/Step Time": 6.998676462098956} {"Full-finetune/Learning Rate": 1.9100667293472252e-05, "Full-finetune/Loss": 0.8395496606826782, "Full-finetune/Loss (Raw)": 0.7927359342575073, "Full-finetune/Step": 841, "Full-finetune/Step Time": 7.002787442877889} {"Full-finetune/Learning Rate": 1.9098083592364976e-05, "Full-finetune/Loss": 0.8410133719444275, "Full-finetune/Loss (Raw)": 0.930984616279602, "Full-finetune/Step": 842, "Full-finetune/Step Time": 7.00369717925787} {"Full-finetune/Learning Rate": 1.9095496360469087e-05, "Full-finetune/Loss": 0.8402048349380493, "Full-finetune/Loss (Raw)": 0.8949021697044373, "Full-finetune/Step": 843, "Full-finetune/Step Time": 7.017299592494965} {"Full-finetune/Learning Rate": 1.9092905598788633e-05, "Full-finetune/Loss": 0.8407580852508545, "Full-finetune/Loss (Raw)": 0.8571159243583679, "Full-finetune/Step": 844, "Full-finetune/Step Time": 7.030216995626688} {"Full-finetune/Learning Rate": 1.9090311308329045e-05, "Full-finetune/Loss": 0.840518593788147, "Full-finetune/Loss (Raw)": 0.9205960631370544, "Full-finetune/Step": 845, "Full-finetune/Step Time": 7.018086014315486} {"Full-finetune/Learning Rate": 1.9087713490097108e-05, "Full-finetune/Loss": 0.8409504294395447, "Full-finetune/Loss (Raw)": 0.9651639461517334, "Full-finetune/Step": 846, "Full-finetune/Step Time": 7.011145168915391} {"Full-finetune/Learning Rate": 1.9085112145100987e-05, "Full-finetune/Loss": 0.8373553156852722, "Full-finetune/Loss (Raw)": 0.4723665416240692, "Full-finetune/Step": 847, "Full-finetune/Step Time": 7.029544815421104} {"Full-finetune/Learning Rate": 1.9082507274350216e-05, "Full-finetune/Loss": 0.8350597023963928, "Full-finetune/Loss (Raw)": 0.758946418762207, "Full-finetune/Step": 848, "Full-finetune/Step Time": 7.03405430726707} {"Full-finetune/Learning Rate": 1.9079898878855695e-05, "Full-finetune/Loss": 0.835649847984314, "Full-finetune/Loss (Raw)": 0.9074123501777649, "Full-finetune/Step": 849, "Full-finetune/Step Time": 7.012464871630073} {"Full-finetune/Learning Rate": 1.9077286959629684e-05, "Full-finetune/Loss": 0.8351452350616455, "Full-finetune/Loss (Raw)": 0.7783032655715942, "Full-finetune/Step": 850, "Full-finetune/Step Time": 6.99787887185812} {"Full-finetune/Learning Rate": 1.907467151768583e-05, "Full-finetune/Loss": 0.8342117071151733, "Full-finetune/Loss (Raw)": 0.9032650589942932, "Full-finetune/Step": 851, "Full-finetune/Step Time": 7.006201623007655} {"Full-finetune/Learning Rate": 1.9072052554039123e-05, "Full-finetune/Loss": 0.8330166339874268, "Full-finetune/Loss (Raw)": 0.7708020210266113, "Full-finetune/Step": 852, "Full-finetune/Step Time": 7.001327712088823} {"Full-finetune/Learning Rate": 1.9069430069705936e-05, "Full-finetune/Loss": 0.8313915729522705, "Full-finetune/Loss (Raw)": 0.8657590746879578, "Full-finetune/Step": 853, "Full-finetune/Step Time": 6.998177649453282} {"Full-finetune/Learning Rate": 1.9066804065704006e-05, "Full-finetune/Loss": 0.8330310583114624, "Full-finetune/Loss (Raw)": 0.8051835298538208, "Full-finetune/Step": 854, "Full-finetune/Step Time": 7.001021018251777} {"Full-finetune/Learning Rate": 1.9064174543052435e-05, "Full-finetune/Loss": 0.8305748701095581, "Full-finetune/Loss (Raw)": 0.6162881851196289, "Full-finetune/Step": 855, "Full-finetune/Step Time": 7.037064144387841} {"Full-finetune/Learning Rate": 1.9061541502771682e-05, "Full-finetune/Loss": 0.8294084072113037, "Full-finetune/Loss (Raw)": 0.8649865388870239, "Full-finetune/Step": 856, "Full-finetune/Step Time": 7.035886192694306} {"Full-finetune/Learning Rate": 1.905890494588359e-05, "Full-finetune/Loss": 0.8315602540969849, "Full-finetune/Loss (Raw)": 0.7724402546882629, "Full-finetune/Step": 857, "Full-finetune/Step Time": 7.016004469245672} {"Full-finetune/Learning Rate": 1.9056264873411343e-05, "Full-finetune/Loss": 0.8326927423477173, "Full-finetune/Loss (Raw)": 1.098114013671875, "Full-finetune/Step": 858, "Full-finetune/Step Time": 7.013933103531599} {"Full-finetune/Learning Rate": 1.905362128637951e-05, "Full-finetune/Loss": 0.8326584100723267, "Full-finetune/Loss (Raw)": 0.9394685626029968, "Full-finetune/Step": 859, "Full-finetune/Step Time": 7.028228014707565} {"Full-finetune/Learning Rate": 1.905097418581401e-05, "Full-finetune/Loss": 0.8329815864562988, "Full-finetune/Loss (Raw)": 0.8197501301765442, "Full-finetune/Step": 860, "Full-finetune/Step Time": 6.998204048722982} {"Full-finetune/Learning Rate": 1.904832357274214e-05, "Full-finetune/Loss": 0.830875039100647, "Full-finetune/Loss (Raw)": 0.5967084765434265, "Full-finetune/Step": 861, "Full-finetune/Step Time": 6.998950397595763} {"Full-finetune/Learning Rate": 1.9045669448192537e-05, "Full-finetune/Loss": 0.8317713141441345, "Full-finetune/Loss (Raw)": 0.8814242482185364, "Full-finetune/Step": 862, "Full-finetune/Step Time": 7.002302773296833} {"Full-finetune/Learning Rate": 1.904301181319523e-05, "Full-finetune/Loss": 0.8314591646194458, "Full-finetune/Loss (Raw)": 0.6994230151176453, "Full-finetune/Step": 863, "Full-finetune/Step Time": 7.008468072861433} {"Full-finetune/Learning Rate": 1.9040350668781584e-05, "Full-finetune/Loss": 0.8320447206497192, "Full-finetune/Loss (Raw)": 0.9395087361335754, "Full-finetune/Step": 864, "Full-finetune/Step Time": 6.977388449013233} {"Full-finetune/Learning Rate": 1.9037686015984343e-05, "Full-finetune/Loss": 0.8304752111434937, "Full-finetune/Loss (Raw)": 0.811980128288269, "Full-finetune/Step": 865, "Full-finetune/Step Time": 6.999551840126514} {"Full-finetune/Learning Rate": 1.9035017855837604e-05, "Full-finetune/Loss": 0.8309170007705688, "Full-finetune/Loss (Raw)": 0.9912059307098389, "Full-finetune/Step": 866, "Full-finetune/Step Time": 7.001979062333703} {"Full-finetune/Learning Rate": 1.9032346189376828e-05, "Full-finetune/Loss": 0.8348236083984375, "Full-finetune/Loss (Raw)": 0.9652586579322815, "Full-finetune/Step": 867, "Full-finetune/Step Time": 6.984702484682202} {"Full-finetune/Learning Rate": 1.902967101763884e-05, "Full-finetune/Loss": 0.8330828547477722, "Full-finetune/Loss (Raw)": 0.803025484085083, "Full-finetune/Step": 868, "Full-finetune/Step Time": 6.97757557593286} {"Full-finetune/Learning Rate": 1.902699234166182e-05, "Full-finetune/Loss": 0.8354567885398865, "Full-finetune/Loss (Raw)": 0.9749497771263123, "Full-finetune/Step": 869, "Full-finetune/Step Time": 6.959476022049785} {"Full-finetune/Learning Rate": 1.9024310162485308e-05, "Full-finetune/Loss": 0.8346986174583435, "Full-finetune/Loss (Raw)": 0.7780157923698425, "Full-finetune/Step": 870, "Full-finetune/Step Time": 6.963062912225723} {"Full-finetune/Learning Rate": 1.9021624481150208e-05, "Full-finetune/Loss": 0.8322819471359253, "Full-finetune/Loss (Raw)": 0.6850185394287109, "Full-finetune/Step": 871, "Full-finetune/Step Time": 6.966392083093524} {"Full-finetune/Learning Rate": 1.9018935298698777e-05, "Full-finetune/Loss": 0.8316956758499146, "Full-finetune/Loss (Raw)": 0.8716432452201843, "Full-finetune/Step": 872, "Full-finetune/Step Time": 6.968307236209512} {"Full-finetune/Learning Rate": 1.9016242616174637e-05, "Full-finetune/Loss": 0.8332284092903137, "Full-finetune/Loss (Raw)": 0.9388778209686279, "Full-finetune/Step": 873, "Full-finetune/Step Time": 6.941496290266514} {"Full-finetune/Learning Rate": 1.9013546434622766e-05, "Full-finetune/Loss": 0.8309899568557739, "Full-finetune/Loss (Raw)": 0.7306773066520691, "Full-finetune/Step": 874, "Full-finetune/Step Time": 6.947094466537237} {"Full-finetune/Learning Rate": 1.90108467550895e-05, "Full-finetune/Loss": 0.8290808796882629, "Full-finetune/Loss (Raw)": 0.7616892457008362, "Full-finetune/Step": 875, "Full-finetune/Step Time": 6.959142774343491} {"Full-finetune/Learning Rate": 1.9008143578622527e-05, "Full-finetune/Loss": 0.8277624845504761, "Full-finetune/Loss (Raw)": 0.8493419885635376, "Full-finetune/Step": 876, "Full-finetune/Step Time": 6.9606507662683725} {"Full-finetune/Learning Rate": 1.9005436906270902e-05, "Full-finetune/Loss": 0.8280189037322998, "Full-finetune/Loss (Raw)": 0.8553745746612549, "Full-finetune/Step": 877, "Full-finetune/Step Time": 6.959171602502465} {"Full-finetune/Learning Rate": 1.9002726739085027e-05, "Full-finetune/Loss": 0.8283772468566895, "Full-finetune/Loss (Raw)": 0.8527082204818726, "Full-finetune/Step": 878, "Full-finetune/Step Time": 6.952828094363213} {"Full-finetune/Learning Rate": 1.900001307811667e-05, "Full-finetune/Loss": 0.8286736607551575, "Full-finetune/Loss (Raw)": 0.7831145524978638, "Full-finetune/Step": 879, "Full-finetune/Step Time": 6.973204597830772} {"Full-finetune/Learning Rate": 1.899729592441895e-05, "Full-finetune/Loss": 0.8274912238121033, "Full-finetune/Loss (Raw)": 0.8421483635902405, "Full-finetune/Step": 880, "Full-finetune/Step Time": 6.976979022845626} {"Full-finetune/Learning Rate": 1.899457527904634e-05, "Full-finetune/Loss": 0.8294978141784668, "Full-finetune/Loss (Raw)": 0.665448784828186, "Full-finetune/Step": 881, "Full-finetune/Step Time": 6.973554307594895} {"Full-finetune/Learning Rate": 1.8991851143054662e-05, "Full-finetune/Loss": 0.830077052116394, "Full-finetune/Loss (Raw)": 0.8638210296630859, "Full-finetune/Step": 882, "Full-finetune/Step Time": 6.972131876274943} {"Full-finetune/Learning Rate": 1.8989123517501112e-05, "Full-finetune/Loss": 0.8319202661514282, "Full-finetune/Loss (Raw)": 0.9728127717971802, "Full-finetune/Step": 883, "Full-finetune/Step Time": 6.943881947547197} {"Full-finetune/Learning Rate": 1.8986392403444222e-05, "Full-finetune/Loss": 0.8328969478607178, "Full-finetune/Loss (Raw)": 0.9848546385765076, "Full-finetune/Step": 884, "Full-finetune/Step Time": 6.944347547367215} {"Full-finetune/Learning Rate": 1.8983657801943884e-05, "Full-finetune/Loss": 0.8282018899917603, "Full-finetune/Loss (Raw)": 0.2260057032108307, "Full-finetune/Step": 885, "Full-finetune/Step Time": 6.949551334604621} {"Full-finetune/Learning Rate": 1.8980919714061343e-05, "Full-finetune/Loss": 0.8271675109863281, "Full-finetune/Loss (Raw)": 0.8533636927604675, "Full-finetune/Step": 886, "Full-finetune/Step Time": 6.95460407435894} {"Full-finetune/Learning Rate": 1.89781781408592e-05, "Full-finetune/Loss": 0.8229238390922546, "Full-finetune/Loss (Raw)": 0.4081066846847534, "Full-finetune/Step": 887, "Full-finetune/Step Time": 6.980935070663691} {"Full-finetune/Learning Rate": 1.8975433083401403e-05, "Full-finetune/Loss": 0.8233171701431274, "Full-finetune/Loss (Raw)": 0.9581623673439026, "Full-finetune/Step": 888, "Full-finetune/Step Time": 6.988064816221595} {"Full-finetune/Learning Rate": 1.897268454275326e-05, "Full-finetune/Loss": 0.8232641220092773, "Full-finetune/Loss (Raw)": 0.9072368144989014, "Full-finetune/Step": 889, "Full-finetune/Step Time": 6.993245754390955} {"Full-finetune/Learning Rate": 1.896993251998142e-05, "Full-finetune/Loss": 0.8270197510719299, "Full-finetune/Loss (Raw)": 0.8620219826698303, "Full-finetune/Step": 890, "Full-finetune/Step Time": 6.971292477101088} {"Full-finetune/Learning Rate": 1.8967177016153895e-05, "Full-finetune/Loss": 0.8280029892921448, "Full-finetune/Loss (Raw)": 0.8682417869567871, "Full-finetune/Step": 891, "Full-finetune/Step Time": 6.9740795604884624} {"Full-finetune/Learning Rate": 1.8964418032340037e-05, "Full-finetune/Loss": 0.827478289604187, "Full-finetune/Loss (Raw)": 0.7972815036773682, "Full-finetune/Step": 892, "Full-finetune/Step Time": 6.977118065580726} {"Full-finetune/Learning Rate": 1.8961655569610557e-05, "Full-finetune/Loss": 0.8278340101242065, "Full-finetune/Loss (Raw)": 0.9583017826080322, "Full-finetune/Step": 893, "Full-finetune/Step Time": 6.973756916821003} {"Full-finetune/Learning Rate": 1.8958889629037514e-05, "Full-finetune/Loss": 0.8273070454597473, "Full-finetune/Loss (Raw)": 0.9177353382110596, "Full-finetune/Step": 894, "Full-finetune/Step Time": 6.958527486771345} {"Full-finetune/Learning Rate": 1.895612021169431e-05, "Full-finetune/Loss": 0.8264962434768677, "Full-finetune/Loss (Raw)": 0.7859386205673218, "Full-finetune/Step": 895, "Full-finetune/Step Time": 6.965308282524347} {"Full-finetune/Learning Rate": 1.8953347318655704e-05, "Full-finetune/Loss": 0.8271673917770386, "Full-finetune/Loss (Raw)": 0.9131026864051819, "Full-finetune/Step": 896, "Full-finetune/Step Time": 6.958182217553258} {"Full-finetune/Learning Rate": 1.8950570950997808e-05, "Full-finetune/Loss": 0.8266905546188354, "Full-finetune/Loss (Raw)": 0.6692429780960083, "Full-finetune/Step": 897, "Full-finetune/Step Time": 6.961697211489081} {"Full-finetune/Learning Rate": 1.894779110979807e-05, "Full-finetune/Loss": 0.8230935335159302, "Full-finetune/Loss (Raw)": 0.37409254908561707, "Full-finetune/Step": 898, "Full-finetune/Step Time": 6.971315650269389} {"Full-finetune/Learning Rate": 1.894500779613529e-05, "Full-finetune/Loss": 0.8210088610649109, "Full-finetune/Loss (Raw)": 0.7369386553764343, "Full-finetune/Step": 899, "Full-finetune/Step Time": 7.0066194869577885} {"Full-finetune/Learning Rate": 1.8942221011089623e-05, "Full-finetune/Loss": 0.8211416602134705, "Full-finetune/Loss (Raw)": 0.9349780082702637, "Full-finetune/Step": 900, "Full-finetune/Step Time": 7.007576061412692} {"Full-finetune/Learning Rate": 1.8939430755742564e-05, "Full-finetune/Loss": 0.8213908672332764, "Full-finetune/Loss (Raw)": 0.9957916736602783, "Full-finetune/Step": 901, "Full-finetune/Step Time": 7.0138602908700705} {"Full-finetune/Learning Rate": 1.8936637031176957e-05, "Full-finetune/Loss": 0.8221864700317383, "Full-finetune/Loss (Raw)": 0.8309436440467834, "Full-finetune/Step": 902, "Full-finetune/Step Time": 6.9974539671093225} {"Full-finetune/Learning Rate": 1.893383983847699e-05, "Full-finetune/Loss": 0.8215435743331909, "Full-finetune/Loss (Raw)": 0.8641391396522522, "Full-finetune/Step": 903, "Full-finetune/Step Time": 7.007861452177167} {"Full-finetune/Learning Rate": 1.89310391787282e-05, "Full-finetune/Loss": 0.8224198222160339, "Full-finetune/Loss (Raw)": 0.7787076234817505, "Full-finetune/Step": 904, "Full-finetune/Step Time": 7.0047866478562355} {"Full-finetune/Learning Rate": 1.892823505301747e-05, "Full-finetune/Loss": 0.8221803903579712, "Full-finetune/Loss (Raw)": 0.9595305323600769, "Full-finetune/Step": 905, "Full-finetune/Step Time": 6.985535154119134} {"Full-finetune/Learning Rate": 1.8925427462433027e-05, "Full-finetune/Loss": 0.8205949068069458, "Full-finetune/Loss (Raw)": 0.7867614030838013, "Full-finetune/Step": 906, "Full-finetune/Step Time": 6.983884196728468} {"Full-finetune/Learning Rate": 1.8922616408064436e-05, "Full-finetune/Loss": 0.8209636807441711, "Full-finetune/Loss (Raw)": 0.9060705900192261, "Full-finetune/Step": 907, "Full-finetune/Step Time": 6.972250146791339} {"Full-finetune/Learning Rate": 1.8919801891002615e-05, "Full-finetune/Loss": 0.8252788782119751, "Full-finetune/Loss (Raw)": 0.888120174407959, "Full-finetune/Step": 908, "Full-finetune/Step Time": 6.955918738618493} {"Full-finetune/Learning Rate": 1.891698391233983e-05, "Full-finetune/Loss": 0.8251849412918091, "Full-finetune/Loss (Raw)": 0.8155818581581116, "Full-finetune/Step": 909, "Full-finetune/Step Time": 6.971075598150492} {"Full-finetune/Learning Rate": 1.8914162473169676e-05, "Full-finetune/Loss": 0.827283501625061, "Full-finetune/Loss (Raw)": 0.837155282497406, "Full-finetune/Step": 910, "Full-finetune/Step Time": 6.943731604143977} {"Full-finetune/Learning Rate": 1.8911337574587098e-05, "Full-finetune/Loss": 0.8272143602371216, "Full-finetune/Loss (Raw)": 0.8410289883613586, "Full-finetune/Step": 911, "Full-finetune/Step Time": 6.90799318253994} {"Full-finetune/Learning Rate": 1.8908509217688387e-05, "Full-finetune/Loss": 0.830070972442627, "Full-finetune/Loss (Raw)": 0.9773091077804565, "Full-finetune/Step": 912, "Full-finetune/Step Time": 6.910651648417115} {"Full-finetune/Learning Rate": 1.8905677403571177e-05, "Full-finetune/Loss": 0.827519416809082, "Full-finetune/Loss (Raw)": 0.617156982421875, "Full-finetune/Step": 913, "Full-finetune/Step Time": 6.905449867248535} {"Full-finetune/Learning Rate": 1.890284213333443e-05, "Full-finetune/Loss": 0.8279099464416504, "Full-finetune/Loss (Raw)": 1.0462865829467773, "Full-finetune/Step": 914, "Full-finetune/Step Time": 6.919557206332684} {"Full-finetune/Learning Rate": 1.8900003408078468e-05, "Full-finetune/Loss": 0.8280211091041565, "Full-finetune/Loss (Raw)": 0.9129374623298645, "Full-finetune/Step": 915, "Full-finetune/Step Time": 6.892753316089511} {"Full-finetune/Learning Rate": 1.889716122890494e-05, "Full-finetune/Loss": 0.8272186517715454, "Full-finetune/Loss (Raw)": 0.9133139252662659, "Full-finetune/Step": 916, "Full-finetune/Step Time": 6.899811636656523} {"Full-finetune/Learning Rate": 1.889431559691684e-05, "Full-finetune/Loss": 0.8279542922973633, "Full-finetune/Loss (Raw)": 1.020138144493103, "Full-finetune/Step": 917, "Full-finetune/Step Time": 6.940941963344812} {"Full-finetune/Learning Rate": 1.8891466513218507e-05, "Full-finetune/Loss": 0.830169677734375, "Full-finetune/Loss (Raw)": 0.8354435563087463, "Full-finetune/Step": 918, "Full-finetune/Step Time": 6.907914079725742} {"Full-finetune/Learning Rate": 1.888861397891561e-05, "Full-finetune/Loss": 0.8281124234199524, "Full-finetune/Loss (Raw)": 0.5936170220375061, "Full-finetune/Step": 919, "Full-finetune/Step Time": 6.922546494752169} {"Full-finetune/Learning Rate": 1.888575799511516e-05, "Full-finetune/Loss": 0.8277386426925659, "Full-finetune/Loss (Raw)": 0.7239669561386108, "Full-finetune/Step": 920, "Full-finetune/Step Time": 6.930243082344532} {"Full-finetune/Learning Rate": 1.888289856292552e-05, "Full-finetune/Loss": 0.8283074498176575, "Full-finetune/Loss (Raw)": 0.94162517786026, "Full-finetune/Step": 921, "Full-finetune/Step Time": 6.9205319583415985} {"Full-finetune/Learning Rate": 1.8880035683456365e-05, "Full-finetune/Loss": 0.8283392190933228, "Full-finetune/Loss (Raw)": 0.8317462801933289, "Full-finetune/Step": 922, "Full-finetune/Step Time": 6.920331489294767} {"Full-finetune/Learning Rate": 1.887716935781873e-05, "Full-finetune/Loss": 0.8284159898757935, "Full-finetune/Loss (Raw)": 0.9291374087333679, "Full-finetune/Step": 923, "Full-finetune/Step Time": 6.912339612841606} {"Full-finetune/Learning Rate": 1.887429958712498e-05, "Full-finetune/Loss": 0.8277262449264526, "Full-finetune/Loss (Raw)": 0.7780520915985107, "Full-finetune/Step": 924, "Full-finetune/Step Time": 6.910878766328096} {"Full-finetune/Learning Rate": 1.8871426372488818e-05, "Full-finetune/Loss": 0.8265759944915771, "Full-finetune/Loss (Raw)": 0.7228044867515564, "Full-finetune/Step": 925, "Full-finetune/Step Time": 6.9300217144191265} {"Full-finetune/Learning Rate": 1.886854971502528e-05, "Full-finetune/Loss": 0.8238468170166016, "Full-finetune/Loss (Raw)": 0.597064197063446, "Full-finetune/Step": 926, "Full-finetune/Step Time": 6.935602581128478} {"Full-finetune/Learning Rate": 1.886566961585074e-05, "Full-finetune/Loss": 0.8245177865028381, "Full-finetune/Loss (Raw)": 0.8811225295066833, "Full-finetune/Step": 927, "Full-finetune/Step Time": 6.9289875868707895} {"Full-finetune/Learning Rate": 1.886278607608291e-05, "Full-finetune/Loss": 0.8245362043380737, "Full-finetune/Loss (Raw)": 0.8961071372032166, "Full-finetune/Step": 928, "Full-finetune/Step Time": 6.921262705698609} {"Full-finetune/Learning Rate": 1.8859899096840836e-05, "Full-finetune/Loss": 0.8240039348602295, "Full-finetune/Loss (Raw)": 0.8628628849983215, "Full-finetune/Step": 929, "Full-finetune/Step Time": 6.899677876383066} {"Full-finetune/Learning Rate": 1.8857008679244897e-05, "Full-finetune/Loss": 0.8240443468093872, "Full-finetune/Loss (Raw)": 0.8010820150375366, "Full-finetune/Step": 930, "Full-finetune/Step Time": 6.90425275452435} {"Full-finetune/Learning Rate": 1.885411482441681e-05, "Full-finetune/Loss": 0.8255670070648193, "Full-finetune/Loss (Raw)": 0.9142545461654663, "Full-finetune/Step": 931, "Full-finetune/Step Time": 6.8867872301489115} {"Full-finetune/Learning Rate": 1.885121753347962e-05, "Full-finetune/Loss": 0.8258055448532104, "Full-finetune/Loss (Raw)": 0.932063639163971, "Full-finetune/Step": 932, "Full-finetune/Step Time": 6.883003994822502} {"Full-finetune/Learning Rate": 1.884831680755771e-05, "Full-finetune/Loss": 0.8254342675209045, "Full-finetune/Loss (Raw)": 0.8351837396621704, "Full-finetune/Step": 933, "Full-finetune/Step Time": 6.892765760421753} {"Full-finetune/Learning Rate": 1.8845412647776795e-05, "Full-finetune/Loss": 0.8274257779121399, "Full-finetune/Loss (Raw)": 0.9424723982810974, "Full-finetune/Step": 934, "Full-finetune/Step Time": 6.880229180678725} {"Full-finetune/Learning Rate": 1.8842505055263928e-05, "Full-finetune/Loss": 0.8257942199707031, "Full-finetune/Loss (Raw)": 0.6660106182098389, "Full-finetune/Step": 935, "Full-finetune/Step Time": 6.880052056163549} {"Full-finetune/Learning Rate": 1.883959403114748e-05, "Full-finetune/Loss": 0.8255108594894409, "Full-finetune/Loss (Raw)": 0.7800026535987854, "Full-finetune/Step": 936, "Full-finetune/Step Time": 6.872114073485136} {"Full-finetune/Learning Rate": 1.8836679576557165e-05, "Full-finetune/Loss": 0.8262763023376465, "Full-finetune/Loss (Raw)": 1.005053997039795, "Full-finetune/Step": 937, "Full-finetune/Step Time": 6.866974901407957} {"Full-finetune/Learning Rate": 1.8833761692624034e-05, "Full-finetune/Loss": 0.8265314698219299, "Full-finetune/Loss (Raw)": 1.025417447090149, "Full-finetune/Step": 938, "Full-finetune/Step Time": 6.8713839296251535} {"Full-finetune/Learning Rate": 1.883084038048045e-05, "Full-finetune/Loss": 0.8272241353988647, "Full-finetune/Loss (Raw)": 0.8491522073745728, "Full-finetune/Step": 939, "Full-finetune/Step Time": 6.87212710082531} {"Full-finetune/Learning Rate": 1.8827915641260127e-05, "Full-finetune/Loss": 0.8261977434158325, "Full-finetune/Loss (Raw)": 0.9400491714477539, "Full-finetune/Step": 940, "Full-finetune/Step Time": 6.875156790018082} {"Full-finetune/Learning Rate": 1.882498747609809e-05, "Full-finetune/Loss": 0.8261393904685974, "Full-finetune/Loss (Raw)": 0.8755058646202087, "Full-finetune/Step": 941, "Full-finetune/Step Time": 6.873218659311533} {"Full-finetune/Learning Rate": 1.8822055886130708e-05, "Full-finetune/Loss": 0.8247177004814148, "Full-finetune/Loss (Raw)": 0.6800919771194458, "Full-finetune/Step": 942, "Full-finetune/Step Time": 6.879565557464957} {"Full-finetune/Learning Rate": 1.8819120872495675e-05, "Full-finetune/Loss": 0.8248603343963623, "Full-finetune/Loss (Raw)": 0.8806313276290894, "Full-finetune/Step": 943, "Full-finetune/Step Time": 6.878583313897252} {"Full-finetune/Learning Rate": 1.8816182436332006e-05, "Full-finetune/Loss": 0.8237406015396118, "Full-finetune/Loss (Raw)": 0.8071982264518738, "Full-finetune/Step": 944, "Full-finetune/Step Time": 6.8737878277897835} {"Full-finetune/Learning Rate": 1.881324057878006e-05, "Full-finetune/Loss": 0.8223932981491089, "Full-finetune/Loss (Raw)": 0.7618702054023743, "Full-finetune/Step": 945, "Full-finetune/Step Time": 6.88283178023994} {"Full-finetune/Learning Rate": 1.8810295300981504e-05, "Full-finetune/Loss": 0.825221061706543, "Full-finetune/Loss (Raw)": 0.7051364183425903, "Full-finetune/Step": 946, "Full-finetune/Step Time": 6.8817126378417015} {"Full-finetune/Learning Rate": 1.8807346604079352e-05, "Full-finetune/Loss": 0.8258436322212219, "Full-finetune/Loss (Raw)": 0.8029857873916626, "Full-finetune/Step": 947, "Full-finetune/Step Time": 6.899116376414895} {"Full-finetune/Learning Rate": 1.880439448921793e-05, "Full-finetune/Loss": 0.8240069150924683, "Full-finetune/Loss (Raw)": 0.3778539299964905, "Full-finetune/Step": 948, "Full-finetune/Step Time": 6.906644064933062} {"Full-finetune/Learning Rate": 1.88014389575429e-05, "Full-finetune/Loss": 0.8254486322402954, "Full-finetune/Loss (Raw)": 0.9985930323600769, "Full-finetune/Step": 949, "Full-finetune/Step Time": 6.893201233819127} {"Full-finetune/Learning Rate": 1.879848001020124e-05, "Full-finetune/Loss": 0.8256819844245911, "Full-finetune/Loss (Raw)": 0.8564573526382446, "Full-finetune/Step": 950, "Full-finetune/Step Time": 6.889946794137359} {"Full-finetune/Learning Rate": 1.8795517648341267e-05, "Full-finetune/Loss": 0.8260163068771362, "Full-finetune/Loss (Raw)": 1.0521950721740723, "Full-finetune/Step": 951, "Full-finetune/Step Time": 6.883186973631382} {"Full-finetune/Learning Rate": 1.8792551873112614e-05, "Full-finetune/Loss": 0.8249177932739258, "Full-finetune/Loss (Raw)": 0.7753232717514038, "Full-finetune/Step": 952, "Full-finetune/Step Time": 6.887532100081444} {"Full-finetune/Learning Rate": 1.8789582685666237e-05, "Full-finetune/Loss": 0.8250224590301514, "Full-finetune/Loss (Raw)": 0.8518526554107666, "Full-finetune/Step": 953, "Full-finetune/Step Time": 6.87926110252738} {"Full-finetune/Learning Rate": 1.878661008715442e-05, "Full-finetune/Loss": 0.8255903720855713, "Full-finetune/Loss (Raw)": 0.810947597026825, "Full-finetune/Step": 954, "Full-finetune/Step Time": 6.875480700284243} {"Full-finetune/Learning Rate": 1.8783634078730774e-05, "Full-finetune/Loss": 0.8260214328765869, "Full-finetune/Loss (Raw)": 0.8830431699752808, "Full-finetune/Step": 955, "Full-finetune/Step Time": 6.874122163280845} {"Full-finetune/Learning Rate": 1.878065466155023e-05, "Full-finetune/Loss": 0.8300806283950806, "Full-finetune/Loss (Raw)": 0.903891384601593, "Full-finetune/Step": 956, "Full-finetune/Step Time": 6.859022453427315} {"Full-finetune/Learning Rate": 1.877767183676904e-05, "Full-finetune/Loss": 0.8299555778503418, "Full-finetune/Loss (Raw)": 0.8141909837722778, "Full-finetune/Step": 957, "Full-finetune/Step Time": 6.854391377419233} {"Full-finetune/Learning Rate": 1.8774685605544776e-05, "Full-finetune/Loss": 0.8292496204376221, "Full-finetune/Loss (Raw)": 0.8075535893440247, "Full-finetune/Step": 958, "Full-finetune/Step Time": 6.8609109073877335} {"Full-finetune/Learning Rate": 1.8771695969036342e-05, "Full-finetune/Loss": 0.828445553779602, "Full-finetune/Loss (Raw)": 0.726576566696167, "Full-finetune/Step": 959, "Full-finetune/Step Time": 6.842561725527048} {"Full-finetune/Learning Rate": 1.8768702928403954e-05, "Full-finetune/Loss": 0.8274216651916504, "Full-finetune/Loss (Raw)": 0.8549246788024902, "Full-finetune/Step": 960, "Full-finetune/Step Time": 6.8447381146252155} {"Full-finetune/Learning Rate": 1.8765706484809155e-05, "Full-finetune/Loss": 0.8300366401672363, "Full-finetune/Loss (Raw)": 0.7567445039749146, "Full-finetune/Step": 961, "Full-finetune/Step Time": 6.827705653384328} {"Full-finetune/Learning Rate": 1.8762706639414806e-05, "Full-finetune/Loss": 0.8304225206375122, "Full-finetune/Loss (Raw)": 0.9359900951385498, "Full-finetune/Step": 962, "Full-finetune/Step Time": 6.823193488642573} {"Full-finetune/Learning Rate": 1.8759703393385087e-05, "Full-finetune/Loss": 0.8302847146987915, "Full-finetune/Loss (Raw)": 0.8447920680046082, "Full-finetune/Step": 963, "Full-finetune/Step Time": 6.816393885761499} {"Full-finetune/Learning Rate": 1.87566967478855e-05, "Full-finetune/Loss": 0.8299587965011597, "Full-finetune/Loss (Raw)": 0.8415364623069763, "Full-finetune/Step": 964, "Full-finetune/Step Time": 6.8161857556551695} {"Full-finetune/Learning Rate": 1.8753686704082868e-05, "Full-finetune/Loss": 0.8298718929290771, "Full-finetune/Loss (Raw)": 0.7304328083992004, "Full-finetune/Step": 965, "Full-finetune/Step Time": 6.811124997213483} {"Full-finetune/Learning Rate": 1.8750673263145323e-05, "Full-finetune/Loss": 0.8299523591995239, "Full-finetune/Loss (Raw)": 0.8075742125511169, "Full-finetune/Step": 966, "Full-finetune/Step Time": 6.8125046994537115} {"Full-finetune/Learning Rate": 1.874765642624233e-05, "Full-finetune/Loss": 0.8300855755805969, "Full-finetune/Loss (Raw)": 0.7900884747505188, "Full-finetune/Step": 967, "Full-finetune/Step Time": 6.836098339408636} {"Full-finetune/Learning Rate": 1.874463619454466e-05, "Full-finetune/Loss": 0.829138994216919, "Full-finetune/Loss (Raw)": 0.9161784648895264, "Full-finetune/Step": 968, "Full-finetune/Step Time": 6.834129532799125} {"Full-finetune/Learning Rate": 1.874161256922441e-05, "Full-finetune/Loss": 0.8292638659477234, "Full-finetune/Loss (Raw)": 0.8087196350097656, "Full-finetune/Step": 969, "Full-finetune/Step Time": 6.852787202224135} {"Full-finetune/Learning Rate": 1.8738585551454982e-05, "Full-finetune/Loss": 0.8261287212371826, "Full-finetune/Loss (Raw)": 0.5296845436096191, "Full-finetune/Step": 970, "Full-finetune/Step Time": 6.858442002907395} {"Full-finetune/Learning Rate": 1.873555514241111e-05, "Full-finetune/Loss": 0.8257885575294495, "Full-finetune/Loss (Raw)": 0.8513662815093994, "Full-finetune/Step": 971, "Full-finetune/Step Time": 6.872262895107269} {"Full-finetune/Learning Rate": 1.8732521343268838e-05, "Full-finetune/Loss": 0.8222014904022217, "Full-finetune/Loss (Raw)": 0.39796870946884155, "Full-finetune/Step": 972, "Full-finetune/Step Time": 6.877135444432497} {"Full-finetune/Learning Rate": 1.8729484155205517e-05, "Full-finetune/Loss": 0.8210874199867249, "Full-finetune/Loss (Raw)": 0.77799391746521, "Full-finetune/Step": 973, "Full-finetune/Step Time": 6.882992018014193} {"Full-finetune/Learning Rate": 1.8726443579399828e-05, "Full-finetune/Loss": 0.8197251558303833, "Full-finetune/Loss (Raw)": 0.7907953262329102, "Full-finetune/Step": 974, "Full-finetune/Step Time": 6.89814268425107} {"Full-finetune/Learning Rate": 1.8723399617031754e-05, "Full-finetune/Loss": 0.8237398862838745, "Full-finetune/Loss (Raw)": 0.9862502813339233, "Full-finetune/Step": 975, "Full-finetune/Step Time": 6.892061265185475} {"Full-finetune/Learning Rate": 1.8720352269282592e-05, "Full-finetune/Loss": 0.8242604732513428, "Full-finetune/Loss (Raw)": 0.8255776762962341, "Full-finetune/Step": 976, "Full-finetune/Step Time": 6.88597696274519} {"Full-finetune/Learning Rate": 1.8717301537334975e-05, "Full-finetune/Loss": 0.8246818780899048, "Full-finetune/Loss (Raw)": 0.9613597989082336, "Full-finetune/Step": 977, "Full-finetune/Step Time": 6.892581801861525} {"Full-finetune/Learning Rate": 1.8714247422372816e-05, "Full-finetune/Loss": 0.8248913288116455, "Full-finetune/Loss (Raw)": 0.8051061630249023, "Full-finetune/Step": 978, "Full-finetune/Step Time": 6.876992076635361} {"Full-finetune/Learning Rate": 1.8711189925581366e-05, "Full-finetune/Loss": 0.8243184089660645, "Full-finetune/Loss (Raw)": 0.8299325108528137, "Full-finetune/Step": 979, "Full-finetune/Step Time": 6.876371122896671} {"Full-finetune/Learning Rate": 1.8708129048147176e-05, "Full-finetune/Loss": 0.8264354467391968, "Full-finetune/Loss (Raw)": 1.0417789220809937, "Full-finetune/Step": 980, "Full-finetune/Step Time": 6.864093754440546} {"Full-finetune/Learning Rate": 1.8705064791258118e-05, "Full-finetune/Loss": 0.8262958526611328, "Full-finetune/Loss (Raw)": 0.8478946685791016, "Full-finetune/Step": 981, "Full-finetune/Step Time": 6.864495541900396} {"Full-finetune/Learning Rate": 1.870199715610336e-05, "Full-finetune/Loss": 0.8273216485977173, "Full-finetune/Loss (Raw)": 0.9364945888519287, "Full-finetune/Step": 982, "Full-finetune/Step Time": 6.853270823135972} {"Full-finetune/Learning Rate": 1.8698926143873404e-05, "Full-finetune/Loss": 0.8295036554336548, "Full-finetune/Loss (Raw)": 0.8955729007720947, "Full-finetune/Step": 983, "Full-finetune/Step Time": 6.817045021802187} {"Full-finetune/Learning Rate": 1.8695851755760042e-05, "Full-finetune/Loss": 0.8286389112472534, "Full-finetune/Loss (Raw)": 0.7543030977249146, "Full-finetune/Step": 984, "Full-finetune/Step Time": 6.8260905761271715} {"Full-finetune/Learning Rate": 1.8692773992956383e-05, "Full-finetune/Loss": 0.8293912410736084, "Full-finetune/Loss (Raw)": 0.8687452673912048, "Full-finetune/Step": 985, "Full-finetune/Step Time": 6.824075857177377} {"Full-finetune/Learning Rate": 1.8689692856656854e-05, "Full-finetune/Loss": 0.8268144726753235, "Full-finetune/Loss (Raw)": 0.768284261226654, "Full-finetune/Step": 986, "Full-finetune/Step Time": 6.82729903049767} {"Full-finetune/Learning Rate": 1.8686608348057178e-05, "Full-finetune/Loss": 0.8263022899627686, "Full-finetune/Loss (Raw)": 0.8739070296287537, "Full-finetune/Step": 987, "Full-finetune/Step Time": 6.823937380686402} {"Full-finetune/Learning Rate": 1.868352046835439e-05, "Full-finetune/Loss": 0.8247060179710388, "Full-finetune/Loss (Raw)": 0.6154243350028992, "Full-finetune/Step": 988, "Full-finetune/Step Time": 6.825729196891189} {"Full-finetune/Learning Rate": 1.8680429218746842e-05, "Full-finetune/Loss": 0.8273525238037109, "Full-finetune/Loss (Raw)": 0.9354609847068787, "Full-finetune/Step": 989, "Full-finetune/Step Time": 6.819322908297181} {"Full-finetune/Learning Rate": 1.867733460043419e-05, "Full-finetune/Loss": 0.8252060413360596, "Full-finetune/Loss (Raw)": 0.6066824197769165, "Full-finetune/Step": 990, "Full-finetune/Step Time": 6.822222612798214} {"Full-finetune/Learning Rate": 1.8674236614617386e-05, "Full-finetune/Loss": 0.8270887136459351, "Full-finetune/Loss (Raw)": 0.9403951168060303, "Full-finetune/Step": 991, "Full-finetune/Step Time": 6.813003819435835} {"Full-finetune/Learning Rate": 1.86711352624987e-05, "Full-finetune/Loss": 0.8264526128768921, "Full-finetune/Loss (Raw)": 0.8580964803695679, "Full-finetune/Step": 992, "Full-finetune/Step Time": 6.809998346492648} {"Full-finetune/Learning Rate": 1.866803054528171e-05, "Full-finetune/Loss": 0.8263756036758423, "Full-finetune/Loss (Raw)": 0.802122950553894, "Full-finetune/Step": 993, "Full-finetune/Step Time": 6.819384295493364} {"Full-finetune/Learning Rate": 1.8664922464171295e-05, "Full-finetune/Loss": 0.8237596750259399, "Full-finetune/Loss (Raw)": 0.6563684344291687, "Full-finetune/Step": 994, "Full-finetune/Step Time": 6.825812250375748} {"Full-finetune/Learning Rate": 1.866181102037364e-05, "Full-finetune/Loss": 0.8229775428771973, "Full-finetune/Loss (Raw)": 0.8651363849639893, "Full-finetune/Step": 995, "Full-finetune/Step Time": 6.824024388566613} {"Full-finetune/Learning Rate": 1.8658696215096235e-05, "Full-finetune/Loss": 0.8225690126419067, "Full-finetune/Loss (Raw)": 0.750738263130188, "Full-finetune/Step": 996, "Full-finetune/Step Time": 6.82705725543201} {"Full-finetune/Learning Rate": 1.8655578049547878e-05, "Full-finetune/Loss": 0.8215627670288086, "Full-finetune/Loss (Raw)": 0.8461509346961975, "Full-finetune/Step": 997, "Full-finetune/Step Time": 6.8063141237944365} {"Full-finetune/Learning Rate": 1.8652456524938665e-05, "Full-finetune/Loss": 0.8213570713996887, "Full-finetune/Loss (Raw)": 0.7516878247261047, "Full-finetune/Step": 998, "Full-finetune/Step Time": 6.818476650863886} {"Full-finetune/Learning Rate": 1.864933164248e-05, "Full-finetune/Loss": 0.8234431743621826, "Full-finetune/Loss (Raw)": 0.952038586139679, "Full-finetune/Step": 999, "Full-finetune/Step Time": 6.811865676194429} {"Full-finetune/Learning Rate": 1.8646203403384583e-05, "Full-finetune/Loss": 0.8244320154190063, "Full-finetune/Loss (Raw)": 0.9982147812843323, "Full-finetune/Step": 1000, "Full-finetune/Step Time": 6.803461333736777} {"Full-finetune/Learning Rate": 1.8643071808866434e-05, "Full-finetune/Loss": 0.8196685910224915, "Full-finetune/Loss (Raw)": 0.3291611075401306, "Full-finetune/Step": 1001, "Full-finetune/Step Time": 6.82446070574224} {"Full-finetune/Learning Rate": 1.8639936860140856e-05, "Full-finetune/Loss": 0.8213875889778137, "Full-finetune/Loss (Raw)": 0.950712263584137, "Full-finetune/Step": 1002, "Full-finetune/Step Time": 6.825658055022359} {"Full-finetune/Learning Rate": 1.8636798558424464e-05, "Full-finetune/Loss": 0.8229163885116577, "Full-finetune/Loss (Raw)": 0.957374095916748, "Full-finetune/Step": 1003, "Full-finetune/Step Time": 6.828083358705044} {"Full-finetune/Learning Rate": 1.8633656904935168e-05, "Full-finetune/Loss": 0.82389235496521, "Full-finetune/Loss (Raw)": 0.974258303642273, "Full-finetune/Step": 1004, "Full-finetune/Step Time": 6.82766636647284} {"Full-finetune/Learning Rate": 1.863051190089219e-05, "Full-finetune/Loss": 0.824171781539917, "Full-finetune/Loss (Raw)": 0.8911375403404236, "Full-finetune/Step": 1005, "Full-finetune/Step Time": 6.820665972307324} {"Full-finetune/Learning Rate": 1.8627363547516032e-05, "Full-finetune/Loss": 0.8228859305381775, "Full-finetune/Loss (Raw)": 0.6881258487701416, "Full-finetune/Step": 1006, "Full-finetune/Step Time": 6.8350014965981245} {"Full-finetune/Learning Rate": 1.8624211846028522e-05, "Full-finetune/Loss": 0.8232110142707825, "Full-finetune/Loss (Raw)": 0.8247281312942505, "Full-finetune/Step": 1007, "Full-finetune/Step Time": 6.816436665132642} {"Full-finetune/Learning Rate": 1.862105679765277e-05, "Full-finetune/Loss": 0.8236746788024902, "Full-finetune/Loss (Raw)": 0.9015010595321655, "Full-finetune/Step": 1008, "Full-finetune/Step Time": 6.815405899658799} {"Full-finetune/Learning Rate": 1.8617898403613183e-05, "Full-finetune/Loss": 0.8218858242034912, "Full-finetune/Loss (Raw)": 0.4364727735519409, "Full-finetune/Step": 1009, "Full-finetune/Step Time": 6.819347694516182} {"Full-finetune/Learning Rate": 1.8614736665135476e-05, "Full-finetune/Loss": 0.821157693862915, "Full-finetune/Loss (Raw)": 0.7706162929534912, "Full-finetune/Step": 1010, "Full-finetune/Step Time": 6.822826834395528} {"Full-finetune/Learning Rate": 1.861157158344666e-05, "Full-finetune/Loss": 0.8198049068450928, "Full-finetune/Loss (Raw)": 0.7996577620506287, "Full-finetune/Step": 1011, "Full-finetune/Step Time": 6.821754205971956} {"Full-finetune/Learning Rate": 1.8608403159775037e-05, "Full-finetune/Loss": 0.819258451461792, "Full-finetune/Loss (Raw)": 0.9149050712585449, "Full-finetune/Step": 1012, "Full-finetune/Step Time": 6.827597415074706} {"Full-finetune/Learning Rate": 1.8605231395350213e-05, "Full-finetune/Loss": 0.8241996765136719, "Full-finetune/Loss (Raw)": 0.8584864139556885, "Full-finetune/Step": 1013, "Full-finetune/Step Time": 6.808351134881377} {"Full-finetune/Learning Rate": 1.860205629140309e-05, "Full-finetune/Loss": 0.8252211213111877, "Full-finetune/Loss (Raw)": 0.9841081500053406, "Full-finetune/Step": 1014, "Full-finetune/Step Time": 6.802894167602062} {"Full-finetune/Learning Rate": 1.8598877849165857e-05, "Full-finetune/Loss": 0.8244463801383972, "Full-finetune/Loss (Raw)": 0.30893880128860474, "Full-finetune/Step": 1015, "Full-finetune/Step Time": 6.795878831297159} {"Full-finetune/Learning Rate": 1.8595696069872013e-05, "Full-finetune/Loss": 0.8239551186561584, "Full-finetune/Loss (Raw)": 0.8952811360359192, "Full-finetune/Step": 1016, "Full-finetune/Step Time": 6.794076794758439} {"Full-finetune/Learning Rate": 1.859251095475634e-05, "Full-finetune/Loss": 0.8244805335998535, "Full-finetune/Loss (Raw)": 0.9744917750358582, "Full-finetune/Step": 1017, "Full-finetune/Step Time": 6.785905554890633} {"Full-finetune/Learning Rate": 1.858932250505492e-05, "Full-finetune/Loss": 0.8241525292396545, "Full-finetune/Loss (Raw)": 0.8200377225875854, "Full-finetune/Step": 1018, "Full-finetune/Step Time": 6.822807909920812} {"Full-finetune/Learning Rate": 1.8586130722005128e-05, "Full-finetune/Loss": 0.8237905502319336, "Full-finetune/Loss (Raw)": 0.8219074606895447, "Full-finetune/Step": 1019, "Full-finetune/Step Time": 6.820123156532645} {"Full-finetune/Learning Rate": 1.8582935606845636e-05, "Full-finetune/Loss": 0.8243733048439026, "Full-finetune/Loss (Raw)": 0.8718705177307129, "Full-finetune/Step": 1020, "Full-finetune/Step Time": 6.816047141328454} {"Full-finetune/Learning Rate": 1.8579737160816397e-05, "Full-finetune/Loss": 0.8223451375961304, "Full-finetune/Loss (Raw)": 0.6986987590789795, "Full-finetune/Step": 1021, "Full-finetune/Step Time": 6.838103560730815} {"Full-finetune/Learning Rate": 1.8576535385158675e-05, "Full-finetune/Loss": 0.8216575980186462, "Full-finetune/Loss (Raw)": 0.8297306895256042, "Full-finetune/Step": 1022, "Full-finetune/Step Time": 6.843875806778669} {"Full-finetune/Learning Rate": 1.8573330281115008e-05, "Full-finetune/Loss": 0.820892870426178, "Full-finetune/Loss (Raw)": 0.6880531907081604, "Full-finetune/Step": 1023, "Full-finetune/Step Time": 6.847586886957288} {"Full-finetune/Learning Rate": 1.8570121849929245e-05, "Full-finetune/Loss": 0.8198764324188232, "Full-finetune/Loss (Raw)": 0.7829993367195129, "Full-finetune/Step": 1024, "Full-finetune/Step Time": 6.849450413137674} {"Full-finetune/Learning Rate": 1.8566910092846508e-05, "Full-finetune/Loss": 0.820744514465332, "Full-finetune/Loss (Raw)": 0.7803614139556885, "Full-finetune/Step": 1025, "Full-finetune/Step Time": 6.8446856290102005} {"Full-finetune/Learning Rate": 1.8563695011113217e-05, "Full-finetune/Loss": 0.8244116306304932, "Full-finetune/Loss (Raw)": 0.8434833288192749, "Full-finetune/Step": 1026, "Full-finetune/Step Time": 6.834914011880755} {"Full-finetune/Learning Rate": 1.8560476605977085e-05, "Full-finetune/Loss": 0.821546196937561, "Full-finetune/Loss (Raw)": 0.3701547086238861, "Full-finetune/Step": 1027, "Full-finetune/Step Time": 6.813421119004488} {"Full-finetune/Learning Rate": 1.8557254878687114e-05, "Full-finetune/Loss": 0.8214836120605469, "Full-finetune/Loss (Raw)": 0.9269654154777527, "Full-finetune/Step": 1028, "Full-finetune/Step Time": 6.817304255440831} {"Full-finetune/Learning Rate": 1.8554029830493587e-05, "Full-finetune/Loss": 0.8197776675224304, "Full-finetune/Loss (Raw)": 0.7774362564086914, "Full-finetune/Step": 1029, "Full-finetune/Step Time": 6.824261894449592} {"Full-finetune/Learning Rate": 1.855080146264809e-05, "Full-finetune/Loss": 0.8194333910942078, "Full-finetune/Loss (Raw)": 0.7868773937225342, "Full-finetune/Step": 1030, "Full-finetune/Step Time": 6.8278668858110905} {"Full-finetune/Learning Rate": 1.854756977640348e-05, "Full-finetune/Loss": 0.819148063659668, "Full-finetune/Loss (Raw)": 0.8276129364967346, "Full-finetune/Step": 1031, "Full-finetune/Step Time": 6.814949968829751} {"Full-finetune/Learning Rate": 1.8544334773013924e-05, "Full-finetune/Loss": 0.8158248662948608, "Full-finetune/Loss (Raw)": 0.35334762930870056, "Full-finetune/Step": 1032, "Full-finetune/Step Time": 6.837506264448166} {"Full-finetune/Learning Rate": 1.8541096453734852e-05, "Full-finetune/Loss": 0.8153319358825684, "Full-finetune/Loss (Raw)": 0.8964277505874634, "Full-finetune/Step": 1033, "Full-finetune/Step Time": 6.864924110472202} {"Full-finetune/Learning Rate": 1.8537854819822998e-05, "Full-finetune/Loss": 0.8150502443313599, "Full-finetune/Loss (Raw)": 0.7507116198539734, "Full-finetune/Step": 1034, "Full-finetune/Step Time": 6.871875699609518} {"Full-finetune/Learning Rate": 1.853460987253638e-05, "Full-finetune/Loss": 0.8148715496063232, "Full-finetune/Loss (Raw)": 0.8831921219825745, "Full-finetune/Step": 1035, "Full-finetune/Step Time": 6.883642161265016} {"Full-finetune/Learning Rate": 1.8531361613134286e-05, "Full-finetune/Loss": 0.8144866228103638, "Full-finetune/Loss (Raw)": 0.8388485908508301, "Full-finetune/Step": 1036, "Full-finetune/Step Time": 6.888007199391723} {"Full-finetune/Learning Rate": 1.8528110042877316e-05, "Full-finetune/Loss": 0.8135033249855042, "Full-finetune/Loss (Raw)": 0.6897222995758057, "Full-finetune/Step": 1037, "Full-finetune/Step Time": 6.879349971190095} {"Full-finetune/Learning Rate": 1.8524855163027337e-05, "Full-finetune/Loss": 0.8151658177375793, "Full-finetune/Loss (Raw)": 1.04995596408844, "Full-finetune/Step": 1038, "Full-finetune/Step Time": 6.873574588447809} {"Full-finetune/Learning Rate": 1.85215969748475e-05, "Full-finetune/Loss": 0.8152157068252563, "Full-finetune/Loss (Raw)": 0.8474058508872986, "Full-finetune/Step": 1039, "Full-finetune/Step Time": 6.872621634975076} {"Full-finetune/Learning Rate": 1.851833547960225e-05, "Full-finetune/Loss": 0.8130775690078735, "Full-finetune/Loss (Raw)": 0.7036362886428833, "Full-finetune/Step": 1040, "Full-finetune/Step Time": 6.85775132663548} {"Full-finetune/Learning Rate": 1.8515070678557304e-05, "Full-finetune/Loss": 0.8161671161651611, "Full-finetune/Loss (Raw)": 1.0126135349273682, "Full-finetune/Step": 1041, "Full-finetune/Step Time": 6.8404395543038845} {"Full-finetune/Learning Rate": 1.8511802572979672e-05, "Full-finetune/Loss": 0.8158045411109924, "Full-finetune/Loss (Raw)": 0.9998817443847656, "Full-finetune/Step": 1042, "Full-finetune/Step Time": 6.831885311752558} {"Full-finetune/Learning Rate": 1.8508531164137644e-05, "Full-finetune/Loss": 0.8116069436073303, "Full-finetune/Loss (Raw)": 0.3756401836872101, "Full-finetune/Step": 1043, "Full-finetune/Step Time": 6.855102725327015} {"Full-finetune/Learning Rate": 1.8505256453300778e-05, "Full-finetune/Loss": 0.8093127012252808, "Full-finetune/Loss (Raw)": 0.6196442246437073, "Full-finetune/Step": 1044, "Full-finetune/Step Time": 6.856586059555411} {"Full-finetune/Learning Rate": 1.850197844173994e-05, "Full-finetune/Loss": 0.8076662421226501, "Full-finetune/Loss (Raw)": 0.8094022274017334, "Full-finetune/Step": 1045, "Full-finetune/Step Time": 6.834213128313422} {"Full-finetune/Learning Rate": 1.8498697130727255e-05, "Full-finetune/Loss": 0.8077011108398438, "Full-finetune/Loss (Raw)": 0.8399085998535156, "Full-finetune/Step": 1046, "Full-finetune/Step Time": 6.836855294182897} {"Full-finetune/Learning Rate": 1.8495412521536135e-05, "Full-finetune/Loss": 0.8099257946014404, "Full-finetune/Loss (Raw)": 0.878372848033905, "Full-finetune/Step": 1047, "Full-finetune/Step Time": 6.824079565703869} {"Full-finetune/Learning Rate": 1.8492124615441276e-05, "Full-finetune/Loss": 0.8121384978294373, "Full-finetune/Loss (Raw)": 1.0071948766708374, "Full-finetune/Step": 1048, "Full-finetune/Step Time": 6.815332846716046} {"Full-finetune/Learning Rate": 1.8488833413718646e-05, "Full-finetune/Loss": 0.8106319904327393, "Full-finetune/Loss (Raw)": 0.7487928867340088, "Full-finetune/Step": 1049, "Full-finetune/Step Time": 6.813211089000106} {"Full-finetune/Learning Rate": 1.84855389176455e-05, "Full-finetune/Loss": 0.8099417686462402, "Full-finetune/Loss (Raw)": 0.7433910965919495, "Full-finetune/Step": 1050, "Full-finetune/Step Time": 6.825237330049276} {"Full-finetune/Learning Rate": 1.8482241128500366e-05, "Full-finetune/Loss": 0.8078383207321167, "Full-finetune/Loss (Raw)": 0.659902811050415, "Full-finetune/Step": 1051, "Full-finetune/Step Time": 6.829756649211049} {"Full-finetune/Learning Rate": 1.847894004756305e-05, "Full-finetune/Loss": 0.8095349073410034, "Full-finetune/Loss (Raw)": 0.995212197303772, "Full-finetune/Step": 1052, "Full-finetune/Step Time": 6.828097429126501} {"Full-finetune/Learning Rate": 1.8475635676114643e-05, "Full-finetune/Loss": 0.8104509711265564, "Full-finetune/Loss (Raw)": 0.8400630950927734, "Full-finetune/Step": 1053, "Full-finetune/Step Time": 6.815299652516842} {"Full-finetune/Learning Rate": 1.8472328015437503e-05, "Full-finetune/Loss": 0.8126887679100037, "Full-finetune/Loss (Raw)": 0.883501410484314, "Full-finetune/Step": 1054, "Full-finetune/Step Time": 6.811433758586645} {"Full-finetune/Learning Rate": 1.8469017066815265e-05, "Full-finetune/Loss": 0.8117082118988037, "Full-finetune/Loss (Raw)": 0.7556091547012329, "Full-finetune/Step": 1055, "Full-finetune/Step Time": 6.811309263110161} {"Full-finetune/Learning Rate": 1.8465702831532848e-05, "Full-finetune/Loss": 0.8126639127731323, "Full-finetune/Loss (Raw)": 1.0184376239776611, "Full-finetune/Step": 1056, "Full-finetune/Step Time": 6.813162598758936} {"Full-finetune/Learning Rate": 1.8462385310876444e-05, "Full-finetune/Loss": 0.8122013211250305, "Full-finetune/Loss (Raw)": 0.8036521673202515, "Full-finetune/Step": 1057, "Full-finetune/Step Time": 6.807393930852413} {"Full-finetune/Learning Rate": 1.845906450613351e-05, "Full-finetune/Loss": 0.8117837905883789, "Full-finetune/Loss (Raw)": 0.7476388812065125, "Full-finetune/Step": 1058, "Full-finetune/Step Time": 6.801025098189712} {"Full-finetune/Learning Rate": 1.8455740418592793e-05, "Full-finetune/Loss": 0.8105921745300293, "Full-finetune/Loss (Raw)": 0.7617289423942566, "Full-finetune/Step": 1059, "Full-finetune/Step Time": 6.819356299936771} {"Full-finetune/Learning Rate": 1.8452413049544304e-05, "Full-finetune/Loss": 0.8094245195388794, "Full-finetune/Loss (Raw)": 0.7826026678085327, "Full-finetune/Step": 1060, "Full-finetune/Step Time": 6.818149274215102} {"Full-finetune/Learning Rate": 1.844908240027933e-05, "Full-finetune/Loss": 0.8079659342765808, "Full-finetune/Loss (Raw)": 0.6484867930412292, "Full-finetune/Step": 1061, "Full-finetune/Step Time": 6.799623303115368} {"Full-finetune/Learning Rate": 1.8445748472090432e-05, "Full-finetune/Loss": 0.8071867227554321, "Full-finetune/Loss (Raw)": 0.8427348732948303, "Full-finetune/Step": 1062, "Full-finetune/Step Time": 6.794862424954772} {"Full-finetune/Learning Rate": 1.844241126627144e-05, "Full-finetune/Loss": 0.8088022470474243, "Full-finetune/Loss (Raw)": 0.8727983236312866, "Full-finetune/Step": 1063, "Full-finetune/Step Time": 6.803505409508944} {"Full-finetune/Learning Rate": 1.843907078411746e-05, "Full-finetune/Loss": 0.8112509846687317, "Full-finetune/Loss (Raw)": 1.0934339761734009, "Full-finetune/Step": 1064, "Full-finetune/Step Time": 6.792748684063554} {"Full-finetune/Learning Rate": 1.8435727026924864e-05, "Full-finetune/Loss": 0.8114243745803833, "Full-finetune/Loss (Raw)": 1.0272469520568848, "Full-finetune/Step": 1065, "Full-finetune/Step Time": 6.7942780666053295} {"Full-finetune/Learning Rate": 1.8432379995991307e-05, "Full-finetune/Loss": 0.8095381259918213, "Full-finetune/Loss (Raw)": 0.7839781045913696, "Full-finetune/Step": 1066, "Full-finetune/Step Time": 6.789412397891283} {"Full-finetune/Learning Rate": 1.8429029692615702e-05, "Full-finetune/Loss": 0.8107279539108276, "Full-finetune/Loss (Raw)": 1.0014514923095703, "Full-finetune/Step": 1067, "Full-finetune/Step Time": 6.798199202865362} {"Full-finetune/Learning Rate": 1.8425676118098236e-05, "Full-finetune/Loss": 0.809465765953064, "Full-finetune/Loss (Raw)": 0.7784844040870667, "Full-finetune/Step": 1068, "Full-finetune/Step Time": 6.802076855674386} {"Full-finetune/Learning Rate": 1.8422319273740364e-05, "Full-finetune/Loss": 0.8090641498565674, "Full-finetune/Loss (Raw)": 0.8241072297096252, "Full-finetune/Step": 1069, "Full-finetune/Step Time": 6.829316282644868} {"Full-finetune/Learning Rate": 1.8418959160844813e-05, "Full-finetune/Loss": 0.8106387853622437, "Full-finetune/Loss (Raw)": 0.8816436529159546, "Full-finetune/Step": 1070, "Full-finetune/Step Time": 6.8193950317800045} {"Full-finetune/Learning Rate": 1.8415595780715583e-05, "Full-finetune/Loss": 0.8110252022743225, "Full-finetune/Loss (Raw)": 0.9300920963287354, "Full-finetune/Step": 1071, "Full-finetune/Step Time": 6.817642629146576} {"Full-finetune/Learning Rate": 1.8412229134657926e-05, "Full-finetune/Loss": 0.8120896816253662, "Full-finetune/Loss (Raw)": 0.9434527158737183, "Full-finetune/Step": 1072, "Full-finetune/Step Time": 6.838476039469242} {"Full-finetune/Learning Rate": 1.8408859223978378e-05, "Full-finetune/Loss": 0.8126440644264221, "Full-finetune/Loss (Raw)": 0.8328328132629395, "Full-finetune/Step": 1073, "Full-finetune/Step Time": 6.822415942326188} {"Full-finetune/Learning Rate": 1.840548604998474e-05, "Full-finetune/Loss": 0.8146512508392334, "Full-finetune/Loss (Raw)": 0.962051272392273, "Full-finetune/Step": 1074, "Full-finetune/Step Time": 6.808841748163104} {"Full-finetune/Learning Rate": 1.840210961398606e-05, "Full-finetune/Loss": 0.815035879611969, "Full-finetune/Loss (Raw)": 0.8522206544876099, "Full-finetune/Step": 1075, "Full-finetune/Step Time": 6.804525235667825} {"Full-finetune/Learning Rate": 1.8398729917292683e-05, "Full-finetune/Loss": 0.8195856809616089, "Full-finetune/Loss (Raw)": 0.9602245688438416, "Full-finetune/Step": 1076, "Full-finetune/Step Time": 6.7877823654562235} {"Full-finetune/Learning Rate": 1.8395346961216193e-05, "Full-finetune/Loss": 0.8193005919456482, "Full-finetune/Loss (Raw)": 0.962108314037323, "Full-finetune/Step": 1077, "Full-finetune/Step Time": 6.790930645540357} {"Full-finetune/Learning Rate": 1.8391960747069453e-05, "Full-finetune/Loss": 0.8200461268424988, "Full-finetune/Loss (Raw)": 0.9518846273422241, "Full-finetune/Step": 1078, "Full-finetune/Step Time": 6.7940877471119165} {"Full-finetune/Learning Rate": 1.8388571276166588e-05, "Full-finetune/Loss": 0.8176723122596741, "Full-finetune/Loss (Raw)": 0.7483420968055725, "Full-finetune/Step": 1079, "Full-finetune/Step Time": 6.797388881444931} {"Full-finetune/Learning Rate": 1.8385178549822982e-05, "Full-finetune/Loss": 0.817017674446106, "Full-finetune/Loss (Raw)": 0.6915338039398193, "Full-finetune/Step": 1080, "Full-finetune/Step Time": 6.797821309417486} {"Full-finetune/Learning Rate": 1.8381782569355288e-05, "Full-finetune/Loss": 0.8174389600753784, "Full-finetune/Loss (Raw)": 0.9057695865631104, "Full-finetune/Step": 1081, "Full-finetune/Step Time": 6.802969539538026} {"Full-finetune/Learning Rate": 1.837838333608142e-05, "Full-finetune/Loss": 0.816373348236084, "Full-finetune/Loss (Raw)": 0.674554705619812, "Full-finetune/Step": 1082, "Full-finetune/Step Time": 6.823176940903068} {"Full-finetune/Learning Rate": 1.8374980851320555e-05, "Full-finetune/Loss": 0.8166418075561523, "Full-finetune/Loss (Raw)": 0.9174075722694397, "Full-finetune/Step": 1083, "Full-finetune/Step Time": 6.822609966620803} {"Full-finetune/Learning Rate": 1.8371575116393125e-05, "Full-finetune/Loss": 0.8167939782142639, "Full-finetune/Loss (Raw)": 0.9233668446540833, "Full-finetune/Step": 1084, "Full-finetune/Step Time": 6.847221080213785} {"Full-finetune/Learning Rate": 1.8368166132620838e-05, "Full-finetune/Loss": 0.8173760175704956, "Full-finetune/Loss (Raw)": 0.8886982202529907, "Full-finetune/Step": 1085, "Full-finetune/Step Time": 6.837440704926848} {"Full-finetune/Learning Rate": 1.8364753901326647e-05, "Full-finetune/Loss": 0.8186399936676025, "Full-finetune/Loss (Raw)": 0.9693328142166138, "Full-finetune/Step": 1086, "Full-finetune/Step Time": 6.8253932520747185} {"Full-finetune/Learning Rate": 1.8361338423834776e-05, "Full-finetune/Loss": 0.8202906250953674, "Full-finetune/Loss (Raw)": 0.9378609657287598, "Full-finetune/Step": 1087, "Full-finetune/Step Time": 6.828574884682894} {"Full-finetune/Learning Rate": 1.8357919701470702e-05, "Full-finetune/Loss": 0.8186435699462891, "Full-finetune/Loss (Raw)": 0.6440997123718262, "Full-finetune/Step": 1088, "Full-finetune/Step Time": 6.835414417088032} {"Full-finetune/Learning Rate": 1.835449773556117e-05, "Full-finetune/Loss": 0.8194853067398071, "Full-finetune/Loss (Raw)": 0.8644971251487732, "Full-finetune/Step": 1089, "Full-finetune/Step Time": 6.832508804276586} {"Full-finetune/Learning Rate": 1.8351072527434173e-05, "Full-finetune/Loss": 0.8196709156036377, "Full-finetune/Loss (Raw)": 0.9597410559654236, "Full-finetune/Step": 1090, "Full-finetune/Step Time": 6.842625260353088} {"Full-finetune/Learning Rate": 1.8347644078418968e-05, "Full-finetune/Loss": 0.8194948434829712, "Full-finetune/Loss (Raw)": 0.8222501277923584, "Full-finetune/Step": 1091, "Full-finetune/Step Time": 6.852398360148072} {"Full-finetune/Learning Rate": 1.834421238984607e-05, "Full-finetune/Loss": 0.8203098773956299, "Full-finetune/Loss (Raw)": 0.9458693861961365, "Full-finetune/Step": 1092, "Full-finetune/Step Time": 6.851125031709671} {"Full-finetune/Learning Rate": 1.8340777463047253e-05, "Full-finetune/Loss": 0.8205541372299194, "Full-finetune/Loss (Raw)": 0.7616891264915466, "Full-finetune/Step": 1093, "Full-finetune/Step Time": 6.849527440965176} {"Full-finetune/Learning Rate": 1.833733929935554e-05, "Full-finetune/Loss": 0.820798397064209, "Full-finetune/Loss (Raw)": 0.8388399481773376, "Full-finetune/Step": 1094, "Full-finetune/Step Time": 6.843209994956851} {"Full-finetune/Learning Rate": 1.833389790010522e-05, "Full-finetune/Loss": 0.8222523927688599, "Full-finetune/Loss (Raw)": 0.9762001633644104, "Full-finetune/Step": 1095, "Full-finetune/Step Time": 6.807608006522059} {"Full-finetune/Learning Rate": 1.8330453266631828e-05, "Full-finetune/Loss": 0.8222432136535645, "Full-finetune/Loss (Raw)": 0.9150063991546631, "Full-finetune/Step": 1096, "Full-finetune/Step Time": 6.804110512137413} {"Full-finetune/Learning Rate": 1.8327005400272165e-05, "Full-finetune/Loss": 0.8225517272949219, "Full-finetune/Loss (Raw)": 0.8482133746147156, "Full-finetune/Step": 1097, "Full-finetune/Step Time": 6.793566334992647} {"Full-finetune/Learning Rate": 1.8323554302364273e-05, "Full-finetune/Loss": 0.8248562812805176, "Full-finetune/Loss (Raw)": 0.8246634006500244, "Full-finetune/Step": 1098, "Full-finetune/Step Time": 6.80183364264667} {"Full-finetune/Learning Rate": 1.8320099974247462e-05, "Full-finetune/Loss": 0.825253963470459, "Full-finetune/Loss (Raw)": 0.9022682905197144, "Full-finetune/Step": 1099, "Full-finetune/Step Time": 6.77519872225821} {"Full-finetune/Learning Rate": 1.831664241726229e-05, "Full-finetune/Loss": 0.8285229206085205, "Full-finetune/Loss (Raw)": 0.8163994550704956, "Full-finetune/Step": 1100, "Full-finetune/Step Time": 6.762967858463526} {"Full-finetune/Learning Rate": 1.831318163275056e-05, "Full-finetune/Loss": 0.8294764757156372, "Full-finetune/Loss (Raw)": 0.9000434875488281, "Full-finetune/Step": 1101, "Full-finetune/Step Time": 6.761390808969736} {"Full-finetune/Learning Rate": 1.8309717622055343e-05, "Full-finetune/Loss": 0.8289138078689575, "Full-finetune/Loss (Raw)": 0.718775749206543, "Full-finetune/Step": 1102, "Full-finetune/Step Time": 6.74797934666276} {"Full-finetune/Learning Rate": 1.830625038652095e-05, "Full-finetune/Loss": 0.8290145397186279, "Full-finetune/Loss (Raw)": 0.9991439580917358, "Full-finetune/Step": 1103, "Full-finetune/Step Time": 6.73541153781116} {"Full-finetune/Learning Rate": 1.8302779927492945e-05, "Full-finetune/Loss": 0.8303428888320923, "Full-finetune/Loss (Raw)": 0.9956021308898926, "Full-finetune/Step": 1104, "Full-finetune/Step Time": 6.745363058522344} {"Full-finetune/Learning Rate": 1.829930624631815e-05, "Full-finetune/Loss": 0.8294072151184082, "Full-finetune/Loss (Raw)": 0.8415994048118591, "Full-finetune/Step": 1105, "Full-finetune/Step Time": 6.748989699408412} {"Full-finetune/Learning Rate": 1.829582934434463e-05, "Full-finetune/Loss": 0.8290443420410156, "Full-finetune/Loss (Raw)": 0.7586549520492554, "Full-finetune/Step": 1106, "Full-finetune/Step Time": 6.761886095628142} {"Full-finetune/Learning Rate": 1.829234922292171e-05, "Full-finetune/Loss": 0.8298748731613159, "Full-finetune/Loss (Raw)": 0.9362434148788452, "Full-finetune/Step": 1107, "Full-finetune/Step Time": 6.757267223671079} {"Full-finetune/Learning Rate": 1.8288865883399946e-05, "Full-finetune/Loss": 0.8296767473220825, "Full-finetune/Loss (Raw)": 1.0164166688919067, "Full-finetune/Step": 1108, "Full-finetune/Step Time": 6.759433763101697} {"Full-finetune/Learning Rate": 1.8285379327131157e-05, "Full-finetune/Loss": 0.8277825117111206, "Full-finetune/Loss (Raw)": 0.6054293513298035, "Full-finetune/Step": 1109, "Full-finetune/Step Time": 6.770921662449837} {"Full-finetune/Learning Rate": 1.828188955546841e-05, "Full-finetune/Loss": 0.8273911476135254, "Full-finetune/Loss (Raw)": 0.8863999247550964, "Full-finetune/Step": 1110, "Full-finetune/Step Time": 6.769920961931348} {"Full-finetune/Learning Rate": 1.827839656976602e-05, "Full-finetune/Loss": 0.828514039516449, "Full-finetune/Loss (Raw)": 1.0393128395080566, "Full-finetune/Step": 1111, "Full-finetune/Step Time": 6.766155743971467} {"Full-finetune/Learning Rate": 1.8274900371379543e-05, "Full-finetune/Loss": 0.8289239406585693, "Full-finetune/Loss (Raw)": 0.8067588806152344, "Full-finetune/Step": 1112, "Full-finetune/Step Time": 6.755734188482165} {"Full-finetune/Learning Rate": 1.8271400961665784e-05, "Full-finetune/Loss": 0.8288664817810059, "Full-finetune/Loss (Raw)": 0.8613949418067932, "Full-finetune/Step": 1113, "Full-finetune/Step Time": 6.753122933208942} {"Full-finetune/Learning Rate": 1.8267898341982796e-05, "Full-finetune/Loss": 0.8288698792457581, "Full-finetune/Loss (Raw)": 0.7687235474586487, "Full-finetune/Step": 1114, "Full-finetune/Step Time": 6.777233220636845} {"Full-finetune/Learning Rate": 1.8264392513689882e-05, "Full-finetune/Loss": 0.8290035724639893, "Full-finetune/Loss (Raw)": 0.8910186290740967, "Full-finetune/Step": 1115, "Full-finetune/Step Time": 6.774971999228001} {"Full-finetune/Learning Rate": 1.8260883478147578e-05, "Full-finetune/Loss": 0.8317872285842896, "Full-finetune/Loss (Raw)": 0.9717321395874023, "Full-finetune/Step": 1116, "Full-finetune/Step Time": 6.771239468827844} {"Full-finetune/Learning Rate": 1.8257371236717682e-05, "Full-finetune/Loss": 0.830703616142273, "Full-finetune/Loss (Raw)": 0.7967520356178284, "Full-finetune/Step": 1117, "Full-finetune/Step Time": 6.770203281193972} {"Full-finetune/Learning Rate": 1.8253855790763215e-05, "Full-finetune/Loss": 0.8330274820327759, "Full-finetune/Loss (Raw)": 0.9041367173194885, "Full-finetune/Step": 1118, "Full-finetune/Step Time": 6.758937759324908} {"Full-finetune/Learning Rate": 1.8250337141648464e-05, "Full-finetune/Loss": 0.8303622007369995, "Full-finetune/Loss (Raw)": 0.5992515683174133, "Full-finetune/Step": 1119, "Full-finetune/Step Time": 6.769889844581485} {"Full-finetune/Learning Rate": 1.8246815290738937e-05, "Full-finetune/Loss": 0.8303903937339783, "Full-finetune/Loss (Raw)": 0.8616952896118164, "Full-finetune/Step": 1120, "Full-finetune/Step Time": 6.773591730743647} {"Full-finetune/Learning Rate": 1.8243290239401402e-05, "Full-finetune/Loss": 0.8318976163864136, "Full-finetune/Loss (Raw)": 0.9950513243675232, "Full-finetune/Step": 1121, "Full-finetune/Step Time": 6.744763879105449} {"Full-finetune/Learning Rate": 1.8239761989003866e-05, "Full-finetune/Loss": 0.8337839841842651, "Full-finetune/Loss (Raw)": 0.8978233337402344, "Full-finetune/Step": 1122, "Full-finetune/Step Time": 6.752312701195478} {"Full-finetune/Learning Rate": 1.8236230540915567e-05, "Full-finetune/Loss": 0.8328829407691956, "Full-finetune/Loss (Raw)": 0.7498013377189636, "Full-finetune/Step": 1123, "Full-finetune/Step Time": 6.770509300753474} {"Full-finetune/Learning Rate": 1.8232695896506995e-05, "Full-finetune/Loss": 0.8353996872901917, "Full-finetune/Loss (Raw)": 1.0728839635849, "Full-finetune/Step": 1124, "Full-finetune/Step Time": 6.767975823953748} {"Full-finetune/Learning Rate": 1.8229158057149875e-05, "Full-finetune/Loss": 0.8353710770606995, "Full-finetune/Loss (Raw)": 0.8424872159957886, "Full-finetune/Step": 1125, "Full-finetune/Step Time": 6.764323445037007} {"Full-finetune/Learning Rate": 1.8225617024217174e-05, "Full-finetune/Loss": 0.8364608883857727, "Full-finetune/Loss (Raw)": 0.8911803960800171, "Full-finetune/Step": 1126, "Full-finetune/Step Time": 6.747046526521444} {"Full-finetune/Learning Rate": 1.8222072799083096e-05, "Full-finetune/Loss": 0.8357590436935425, "Full-finetune/Loss (Raw)": 0.862205982208252, "Full-finetune/Step": 1127, "Full-finetune/Step Time": 6.746403058990836} {"Full-finetune/Learning Rate": 1.821852538312309e-05, "Full-finetune/Loss": 0.8340801000595093, "Full-finetune/Loss (Raw)": 0.7833117246627808, "Full-finetune/Step": 1128, "Full-finetune/Step Time": 6.759508941322565} {"Full-finetune/Learning Rate": 1.8214974777713834e-05, "Full-finetune/Loss": 0.8389055728912354, "Full-finetune/Loss (Raw)": 0.9468210339546204, "Full-finetune/Step": 1129, "Full-finetune/Step Time": 6.736449943855405} {"Full-finetune/Learning Rate": 1.821142098423325e-05, "Full-finetune/Loss": 0.8365417718887329, "Full-finetune/Loss (Raw)": 0.6481473445892334, "Full-finetune/Step": 1130, "Full-finetune/Step Time": 6.743616860359907} {"Full-finetune/Learning Rate": 1.8207864004060497e-05, "Full-finetune/Loss": 0.8362449407577515, "Full-finetune/Loss (Raw)": 0.9193755388259888, "Full-finetune/Step": 1131, "Full-finetune/Step Time": 6.726412897929549} {"Full-finetune/Learning Rate": 1.820430383857597e-05, "Full-finetune/Loss": 0.8360155820846558, "Full-finetune/Loss (Raw)": 0.9449005722999573, "Full-finetune/Step": 1132, "Full-finetune/Step Time": 6.728002525866032} {"Full-finetune/Learning Rate": 1.8200740489161296e-05, "Full-finetune/Loss": 0.8359408378601074, "Full-finetune/Loss (Raw)": 0.8815672397613525, "Full-finetune/Step": 1133, "Full-finetune/Step Time": 6.7231814451515675} {"Full-finetune/Learning Rate": 1.819717395719935e-05, "Full-finetune/Loss": 0.8373029232025146, "Full-finetune/Loss (Raw)": 0.8624700307846069, "Full-finetune/Step": 1134, "Full-finetune/Step Time": 6.705178780481219} {"Full-finetune/Learning Rate": 1.819360424407422e-05, "Full-finetune/Loss": 0.8382123112678528, "Full-finetune/Loss (Raw)": 0.9411399960517883, "Full-finetune/Step": 1135, "Full-finetune/Step Time": 6.700193854048848} {"Full-finetune/Learning Rate": 1.8190031351171256e-05, "Full-finetune/Loss": 0.8376739621162415, "Full-finetune/Loss (Raw)": 0.8325856924057007, "Full-finetune/Step": 1136, "Full-finetune/Step Time": 6.697129230946302} {"Full-finetune/Learning Rate": 1.818645527987702e-05, "Full-finetune/Loss": 0.8418418765068054, "Full-finetune/Loss (Raw)": 0.9699673652648926, "Full-finetune/Step": 1137, "Full-finetune/Step Time": 6.67940484918654} {"Full-finetune/Learning Rate": 1.8182876031579312e-05, "Full-finetune/Loss": 0.842698335647583, "Full-finetune/Loss (Raw)": 0.8802462816238403, "Full-finetune/Step": 1138, "Full-finetune/Step Time": 6.696882296353579} {"Full-finetune/Learning Rate": 1.8179293607667177e-05, "Full-finetune/Loss": 0.8428293466567993, "Full-finetune/Loss (Raw)": 0.8164219260215759, "Full-finetune/Step": 1139, "Full-finetune/Step Time": 6.710067758336663} {"Full-finetune/Learning Rate": 1.817570800953088e-05, "Full-finetune/Loss": 0.8425843715667725, "Full-finetune/Loss (Raw)": 0.8835552930831909, "Full-finetune/Step": 1140, "Full-finetune/Step Time": 6.711951585486531} {"Full-finetune/Learning Rate": 1.8172119238561924e-05, "Full-finetune/Loss": 0.8424671292304993, "Full-finetune/Loss (Raw)": 0.8434761166572571, "Full-finetune/Step": 1141, "Full-finetune/Step Time": 6.71813371963799} {"Full-finetune/Learning Rate": 1.8168527296153037e-05, "Full-finetune/Loss": 0.8371255993843079, "Full-finetune/Loss (Raw)": 0.3003893792629242, "Full-finetune/Step": 1142, "Full-finetune/Step Time": 6.727314626798034} {"Full-finetune/Learning Rate": 1.8164932183698184e-05, "Full-finetune/Loss": 0.8412870168685913, "Full-finetune/Loss (Raw)": 0.8416012525558472, "Full-finetune/Step": 1143, "Full-finetune/Step Time": 6.722357980906963} {"Full-finetune/Learning Rate": 1.8161333902592554e-05, "Full-finetune/Loss": 0.8410120606422424, "Full-finetune/Loss (Raw)": 0.8600898385047913, "Full-finetune/Step": 1144, "Full-finetune/Step Time": 6.7244868110865355} {"Full-finetune/Learning Rate": 1.8157732454232577e-05, "Full-finetune/Loss": 0.840073823928833, "Full-finetune/Loss (Raw)": 0.8543968796730042, "Full-finetune/Step": 1145, "Full-finetune/Step Time": 6.72447738237679} {"Full-finetune/Learning Rate": 1.81541278400159e-05, "Full-finetune/Loss": 0.8410515785217285, "Full-finetune/Loss (Raw)": 0.9451931715011597, "Full-finetune/Step": 1146, "Full-finetune/Step Time": 6.690820315852761} {"Full-finetune/Learning Rate": 1.81505200613414e-05, "Full-finetune/Loss": 0.8399257659912109, "Full-finetune/Loss (Raw)": 0.6778002381324768, "Full-finetune/Step": 1147, "Full-finetune/Step Time": 6.692672343924642} {"Full-finetune/Learning Rate": 1.8146909119609196e-05, "Full-finetune/Loss": 0.8359005451202393, "Full-finetune/Loss (Raw)": 0.3566405177116394, "Full-finetune/Step": 1148, "Full-finetune/Step Time": 6.716809758916497} {"Full-finetune/Learning Rate": 1.8143295016220612e-05, "Full-finetune/Loss": 0.8386884927749634, "Full-finetune/Loss (Raw)": 1.0555622577667236, "Full-finetune/Step": 1149, "Full-finetune/Step Time": 6.718703454360366} {"Full-finetune/Learning Rate": 1.813967775257822e-05, "Full-finetune/Loss": 0.8382201790809631, "Full-finetune/Loss (Raw)": 0.7697797417640686, "Full-finetune/Step": 1150, "Full-finetune/Step Time": 6.708571610972285} {"Full-finetune/Learning Rate": 1.8136057330085805e-05, "Full-finetune/Loss": 0.838909387588501, "Full-finetune/Loss (Raw)": 0.7762826681137085, "Full-finetune/Step": 1151, "Full-finetune/Step Time": 6.692635636776686} {"Full-finetune/Learning Rate": 1.8132433750148383e-05, "Full-finetune/Loss": 0.8392270803451538, "Full-finetune/Loss (Raw)": 0.8236548900604248, "Full-finetune/Step": 1152, "Full-finetune/Step Time": 6.7084655705839396} {"Full-finetune/Learning Rate": 1.8128807014172197e-05, "Full-finetune/Loss": 0.8402527570724487, "Full-finetune/Loss (Raw)": 0.9116494059562683, "Full-finetune/Step": 1153, "Full-finetune/Step Time": 6.710927112028003} {"Full-finetune/Learning Rate": 1.812517712356471e-05, "Full-finetune/Loss": 0.8362404108047485, "Full-finetune/Loss (Raw)": 0.32990485429763794, "Full-finetune/Step": 1154, "Full-finetune/Step Time": 6.720969893038273} {"Full-finetune/Learning Rate": 1.812154407973461e-05, "Full-finetune/Loss": 0.8360123634338379, "Full-finetune/Loss (Raw)": 0.340971440076828, "Full-finetune/Step": 1155, "Full-finetune/Step Time": 6.719927180558443} {"Full-finetune/Learning Rate": 1.8117907884091815e-05, "Full-finetune/Loss": 0.8366096615791321, "Full-finetune/Loss (Raw)": 1.0034112930297852, "Full-finetune/Step": 1156, "Full-finetune/Step Time": 6.715689919888973} {"Full-finetune/Learning Rate": 1.8114268538047457e-05, "Full-finetune/Loss": 0.8379757404327393, "Full-finetune/Loss (Raw)": 0.9522897005081177, "Full-finetune/Step": 1157, "Full-finetune/Step Time": 6.707686202600598} {"Full-finetune/Learning Rate": 1.81106260430139e-05, "Full-finetune/Loss": 0.8380211591720581, "Full-finetune/Loss (Raw)": 0.7926910519599915, "Full-finetune/Step": 1158, "Full-finetune/Step Time": 6.698079971596599} {"Full-finetune/Learning Rate": 1.810698040040472e-05, "Full-finetune/Loss": 0.8376107811927795, "Full-finetune/Loss (Raw)": 0.7750918865203857, "Full-finetune/Step": 1159, "Full-finetune/Step Time": 6.686000583693385} {"Full-finetune/Learning Rate": 1.810333161163472e-05, "Full-finetune/Loss": 0.8413832187652588, "Full-finetune/Loss (Raw)": 0.8362162709236145, "Full-finetune/Step": 1160, "Full-finetune/Step Time": 6.680517680943012} {"Full-finetune/Learning Rate": 1.809967967811993e-05, "Full-finetune/Loss": 0.8405992984771729, "Full-finetune/Loss (Raw)": 0.7960866689682007, "Full-finetune/Step": 1161, "Full-finetune/Step Time": 6.67169701308012} {"Full-finetune/Learning Rate": 1.8096024601277586e-05, "Full-finetune/Loss": 0.8421241044998169, "Full-finetune/Loss (Raw)": 0.9458951950073242, "Full-finetune/Step": 1162, "Full-finetune/Step Time": 6.669560795649886} {"Full-finetune/Learning Rate": 1.8092366382526157e-05, "Full-finetune/Loss": 0.8400595188140869, "Full-finetune/Loss (Raw)": 0.6189182996749878, "Full-finetune/Step": 1163, "Full-finetune/Step Time": 6.6775633823126554} {"Full-finetune/Learning Rate": 1.8088705023285324e-05, "Full-finetune/Loss": 0.8417820930480957, "Full-finetune/Loss (Raw)": 1.0593441724777222, "Full-finetune/Step": 1164, "Full-finetune/Step Time": 6.675841121003032} {"Full-finetune/Learning Rate": 1.8085040524975994e-05, "Full-finetune/Loss": 0.8392245769500732, "Full-finetune/Loss (Raw)": 0.3623475134372711, "Full-finetune/Step": 1165, "Full-finetune/Step Time": 6.68138231895864} {"Full-finetune/Learning Rate": 1.8081372889020282e-05, "Full-finetune/Loss": 0.8372579216957092, "Full-finetune/Loss (Raw)": 0.7982293963432312, "Full-finetune/Step": 1166, "Full-finetune/Step Time": 6.689845331013203} {"Full-finetune/Learning Rate": 1.8077702116841524e-05, "Full-finetune/Loss": 0.8354306221008301, "Full-finetune/Loss (Raw)": 0.6135178804397583, "Full-finetune/Step": 1167, "Full-finetune/Step Time": 6.704206025227904} {"Full-finetune/Learning Rate": 1.807402820986428e-05, "Full-finetune/Loss": 0.8360259532928467, "Full-finetune/Loss (Raw)": 0.779832124710083, "Full-finetune/Step": 1168, "Full-finetune/Step Time": 6.702466247603297} {"Full-finetune/Learning Rate": 1.8070351169514324e-05, "Full-finetune/Loss": 0.8352733850479126, "Full-finetune/Loss (Raw)": 0.9162846207618713, "Full-finetune/Step": 1169, "Full-finetune/Step Time": 6.704055491834879} {"Full-finetune/Learning Rate": 1.8066670997218633e-05, "Full-finetune/Loss": 0.8311881422996521, "Full-finetune/Loss (Raw)": 0.4769696593284607, "Full-finetune/Step": 1170, "Full-finetune/Step Time": 6.716562740504742} {"Full-finetune/Learning Rate": 1.806298769440542e-05, "Full-finetune/Loss": 0.8359495997428894, "Full-finetune/Loss (Raw)": 0.9851148724555969, "Full-finetune/Step": 1171, "Full-finetune/Step Time": 6.69349261932075} {"Full-finetune/Learning Rate": 1.8059301262504097e-05, "Full-finetune/Loss": 0.83815997838974, "Full-finetune/Loss (Raw)": 0.902568519115448, "Full-finetune/Step": 1172, "Full-finetune/Step Time": 6.6861315704882145} {"Full-finetune/Learning Rate": 1.80556117029453e-05, "Full-finetune/Loss": 0.8386267423629761, "Full-finetune/Loss (Raw)": 0.8691415786743164, "Full-finetune/Step": 1173, "Full-finetune/Step Time": 6.669123584404588} {"Full-finetune/Learning Rate": 1.8051919017160875e-05, "Full-finetune/Loss": 0.8349025249481201, "Full-finetune/Loss (Raw)": 0.36320799589157104, "Full-finetune/Step": 1174, "Full-finetune/Step Time": 6.677503157407045} {"Full-finetune/Learning Rate": 1.804822320658388e-05, "Full-finetune/Loss": 0.8341286182403564, "Full-finetune/Loss (Raw)": 0.7793207764625549, "Full-finetune/Step": 1175, "Full-finetune/Step Time": 6.691387727856636} {"Full-finetune/Learning Rate": 1.8044524272648588e-05, "Full-finetune/Loss": 0.8326324224472046, "Full-finetune/Loss (Raw)": 0.815678596496582, "Full-finetune/Step": 1176, "Full-finetune/Step Time": 6.685921851545572} {"Full-finetune/Learning Rate": 1.8040822216790475e-05, "Full-finetune/Loss": 0.8290731906890869, "Full-finetune/Loss (Raw)": 0.293208509683609, "Full-finetune/Step": 1177, "Full-finetune/Step Time": 6.70521118119359} {"Full-finetune/Learning Rate": 1.803711704044625e-05, "Full-finetune/Loss": 0.8258200883865356, "Full-finetune/Loss (Raw)": 0.326995313167572, "Full-finetune/Step": 1178, "Full-finetune/Step Time": 6.708815727382898} {"Full-finetune/Learning Rate": 1.803340874505381e-05, "Full-finetune/Loss": 0.8286992311477661, "Full-finetune/Loss (Raw)": 1.0284353494644165, "Full-finetune/Step": 1179, "Full-finetune/Step Time": 6.703137136995792} {"Full-finetune/Learning Rate": 1.8029697332052277e-05, "Full-finetune/Loss": 0.8269634246826172, "Full-finetune/Loss (Raw)": 0.7730284333229065, "Full-finetune/Step": 1180, "Full-finetune/Step Time": 6.718207489699125} {"Full-finetune/Learning Rate": 1.8025982802881976e-05, "Full-finetune/Loss": 0.8281204700469971, "Full-finetune/Loss (Raw)": 0.9881628155708313, "Full-finetune/Step": 1181, "Full-finetune/Step Time": 6.705786334350705} {"Full-finetune/Learning Rate": 1.802226515898444e-05, "Full-finetune/Loss": 0.825424313545227, "Full-finetune/Loss (Raw)": 0.5383933186531067, "Full-finetune/Step": 1182, "Full-finetune/Step Time": 6.70606292784214} {"Full-finetune/Learning Rate": 1.8018544401802422e-05, "Full-finetune/Loss": 0.826938807964325, "Full-finetune/Loss (Raw)": 0.9494701027870178, "Full-finetune/Step": 1183, "Full-finetune/Step Time": 6.707919294014573} {"Full-finetune/Learning Rate": 1.801482053277987e-05, "Full-finetune/Loss": 0.8243534564971924, "Full-finetune/Loss (Raw)": 0.6875022649765015, "Full-finetune/Step": 1184, "Full-finetune/Step Time": 6.73595405369997} {"Full-finetune/Learning Rate": 1.8011093553361944e-05, "Full-finetune/Loss": 0.8249869346618652, "Full-finetune/Loss (Raw)": 0.8847417831420898, "Full-finetune/Step": 1185, "Full-finetune/Step Time": 6.738002650439739} {"Full-finetune/Learning Rate": 1.8007363464995015e-05, "Full-finetune/Loss": 0.827072262763977, "Full-finetune/Loss (Raw)": 1.014563798904419, "Full-finetune/Step": 1186, "Full-finetune/Step Time": 6.737592468038201} {"Full-finetune/Learning Rate": 1.800363026912665e-05, "Full-finetune/Loss": 0.823837161064148, "Full-finetune/Loss (Raw)": 0.34763476252555847, "Full-finetune/Step": 1187, "Full-finetune/Step Time": 6.740827118977904} {"Full-finetune/Learning Rate": 1.7999893967205643e-05, "Full-finetune/Loss": 0.8248364329338074, "Full-finetune/Loss (Raw)": 0.9105114936828613, "Full-finetune/Step": 1188, "Full-finetune/Step Time": 6.741301203146577} {"Full-finetune/Learning Rate": 1.7996154560681963e-05, "Full-finetune/Loss": 0.8273671269416809, "Full-finetune/Loss (Raw)": 0.9724116921424866, "Full-finetune/Step": 1189, "Full-finetune/Step Time": 6.74639911018312} {"Full-finetune/Learning Rate": 1.7992412051006817e-05, "Full-finetune/Loss": 0.8276597857475281, "Full-finetune/Loss (Raw)": 0.8801994919776917, "Full-finetune/Step": 1190, "Full-finetune/Step Time": 6.74391532689333} {"Full-finetune/Learning Rate": 1.7988666439632593e-05, "Full-finetune/Loss": 0.8280210494995117, "Full-finetune/Loss (Raw)": 0.9190410375595093, "Full-finetune/Step": 1191, "Full-finetune/Step Time": 6.737332681193948} {"Full-finetune/Learning Rate": 1.7984917728012886e-05, "Full-finetune/Loss": 0.8270575404167175, "Full-finetune/Loss (Raw)": 0.9700981974601746, "Full-finetune/Step": 1192, "Full-finetune/Step Time": 6.74437490850687} {"Full-finetune/Learning Rate": 1.7981165917602502e-05, "Full-finetune/Loss": 0.8261001110076904, "Full-finetune/Loss (Raw)": 0.9046974778175354, "Full-finetune/Step": 1193, "Full-finetune/Step Time": 6.743031445890665} {"Full-finetune/Learning Rate": 1.7977411009857446e-05, "Full-finetune/Loss": 0.8270954489707947, "Full-finetune/Loss (Raw)": 0.911384105682373, "Full-finetune/Step": 1194, "Full-finetune/Step Time": 6.754769759252667} {"Full-finetune/Learning Rate": 1.797365300623492e-05, "Full-finetune/Loss": 0.8260141015052795, "Full-finetune/Loss (Raw)": 0.8630411028862, "Full-finetune/Step": 1195, "Full-finetune/Step Time": 6.743480164557695} {"Full-finetune/Learning Rate": 1.7969891908193338e-05, "Full-finetune/Loss": 0.8269399404525757, "Full-finetune/Loss (Raw)": 0.8969884514808655, "Full-finetune/Step": 1196, "Full-finetune/Step Time": 6.7509354166686535} {"Full-finetune/Learning Rate": 1.7966127717192305e-05, "Full-finetune/Loss": 0.828060507774353, "Full-finetune/Loss (Raw)": 0.9675436019897461, "Full-finetune/Step": 1197, "Full-finetune/Step Time": 6.7180074490606785} {"Full-finetune/Learning Rate": 1.796236043469263e-05, "Full-finetune/Loss": 0.8282082676887512, "Full-finetune/Loss (Raw)": 0.9005506038665771, "Full-finetune/Step": 1198, "Full-finetune/Step Time": 6.723042264580727} {"Full-finetune/Learning Rate": 1.7958590062156318e-05, "Full-finetune/Loss": 0.8285043239593506, "Full-finetune/Loss (Raw)": 0.9679970741271973, "Full-finetune/Step": 1199, "Full-finetune/Step Time": 6.718681775033474} {"Full-finetune/Learning Rate": 1.7954816601046587e-05, "Full-finetune/Loss": 0.8285355567932129, "Full-finetune/Loss (Raw)": 0.9474356174468994, "Full-finetune/Step": 1200, "Full-finetune/Step Time": 6.705989899113774} {"Full-finetune/Learning Rate": 1.7951040052827836e-05, "Full-finetune/Loss": 0.8297207355499268, "Full-finetune/Loss (Raw)": 0.9845446348190308, "Full-finetune/Step": 1201, "Full-finetune/Step Time": 6.7188157476484776} {"Full-finetune/Learning Rate": 1.794726041896567e-05, "Full-finetune/Loss": 0.8306817412376404, "Full-finetune/Loss (Raw)": 1.0850586891174316, "Full-finetune/Step": 1202, "Full-finetune/Step Time": 6.745147189125419} {"Full-finetune/Learning Rate": 1.7943477700926895e-05, "Full-finetune/Loss": 0.8321030139923096, "Full-finetune/Loss (Raw)": 1.034141182899475, "Full-finetune/Step": 1203, "Full-finetune/Step Time": 6.734803715720773} {"Full-finetune/Learning Rate": 1.7939691900179507e-05, "Full-finetune/Loss": 0.8321366310119629, "Full-finetune/Loss (Raw)": 0.964527428150177, "Full-finetune/Step": 1204, "Full-finetune/Step Time": 6.742459682747722} {"Full-finetune/Learning Rate": 1.79359030181927e-05, "Full-finetune/Loss": 0.8307008743286133, "Full-finetune/Loss (Raw)": 0.7783346176147461, "Full-finetune/Step": 1205, "Full-finetune/Step Time": 6.759529115632176} {"Full-finetune/Learning Rate": 1.7932111056436868e-05, "Full-finetune/Loss": 0.8313626050949097, "Full-finetune/Loss (Raw)": 1.0365877151489258, "Full-finetune/Step": 1206, "Full-finetune/Step Time": 6.751500571146607} {"Full-finetune/Learning Rate": 1.79283160163836e-05, "Full-finetune/Loss": 0.8324292302131653, "Full-finetune/Loss (Raw)": 0.8848704099655151, "Full-finetune/Step": 1207, "Full-finetune/Step Time": 6.759453263133764} {"Full-finetune/Learning Rate": 1.792451789950567e-05, "Full-finetune/Loss": 0.8341647386550903, "Full-finetune/Loss (Raw)": 0.913671612739563, "Full-finetune/Step": 1208, "Full-finetune/Step Time": 6.759712150320411} {"Full-finetune/Learning Rate": 1.7920716707277054e-05, "Full-finetune/Loss": 0.8345195055007935, "Full-finetune/Loss (Raw)": 0.9511884450912476, "Full-finetune/Step": 1209, "Full-finetune/Step Time": 6.756709551438689} {"Full-finetune/Learning Rate": 1.7916912441172924e-05, "Full-finetune/Loss": 0.8359311819076538, "Full-finetune/Loss (Raw)": 0.8552477359771729, "Full-finetune/Step": 1210, "Full-finetune/Step Time": 6.735058780759573} {"Full-finetune/Learning Rate": 1.7913105102669643e-05, "Full-finetune/Loss": 0.8356270790100098, "Full-finetune/Loss (Raw)": 0.8784775137901306, "Full-finetune/Step": 1211, "Full-finetune/Step Time": 6.737832522019744} {"Full-finetune/Learning Rate": 1.790929469324476e-05, "Full-finetune/Loss": 0.8358842134475708, "Full-finetune/Loss (Raw)": 0.9562774896621704, "Full-finetune/Step": 1212, "Full-finetune/Step Time": 6.715773778036237} {"Full-finetune/Learning Rate": 1.7905481214377027e-05, "Full-finetune/Loss": 0.8355699777603149, "Full-finetune/Loss (Raw)": 0.8484848141670227, "Full-finetune/Step": 1213, "Full-finetune/Step Time": 6.7147651091217995} {"Full-finetune/Learning Rate": 1.790166466754637e-05, "Full-finetune/Loss": 0.8339612483978271, "Full-finetune/Loss (Raw)": 0.7634066343307495, "Full-finetune/Step": 1214, "Full-finetune/Step Time": 6.728257050737739} {"Full-finetune/Learning Rate": 1.7897845054233927e-05, "Full-finetune/Loss": 0.8338395357131958, "Full-finetune/Loss (Raw)": 0.9222880005836487, "Full-finetune/Step": 1215, "Full-finetune/Step Time": 6.728354902938008} {"Full-finetune/Learning Rate": 1.7894022375922006e-05, "Full-finetune/Loss": 0.8316035270690918, "Full-finetune/Loss (Raw)": 0.3578820824623108, "Full-finetune/Step": 1216, "Full-finetune/Step Time": 6.737409679219127} {"Full-finetune/Learning Rate": 1.7890196634094122e-05, "Full-finetune/Loss": 0.8328895568847656, "Full-finetune/Loss (Raw)": 1.0291107892990112, "Full-finetune/Step": 1217, "Full-finetune/Step Time": 6.741345563903451} {"Full-finetune/Learning Rate": 1.7886367830234968e-05, "Full-finetune/Loss": 0.8329145908355713, "Full-finetune/Loss (Raw)": 0.9629554748535156, "Full-finetune/Step": 1218, "Full-finetune/Step Time": 6.726916961371899} {"Full-finetune/Learning Rate": 1.788253596583043e-05, "Full-finetune/Loss": 0.8289002180099487, "Full-finetune/Loss (Raw)": 0.30841153860092163, "Full-finetune/Step": 1219, "Full-finetune/Step Time": 6.735321717336774} {"Full-finetune/Learning Rate": 1.787870104236757e-05, "Full-finetune/Loss": 0.8282544612884521, "Full-finetune/Loss (Raw)": 0.8632020950317383, "Full-finetune/Step": 1220, "Full-finetune/Step Time": 6.736089704558253} {"Full-finetune/Learning Rate": 1.7874863061334658e-05, "Full-finetune/Loss": 0.8288156986236572, "Full-finetune/Loss (Raw)": 0.8335400819778442, "Full-finetune/Step": 1221, "Full-finetune/Step Time": 6.745435729622841} {"Full-finetune/Learning Rate": 1.7871022024221133e-05, "Full-finetune/Loss": 0.8286302089691162, "Full-finetune/Loss (Raw)": 0.8150904774665833, "Full-finetune/Step": 1222, "Full-finetune/Step Time": 6.756137624382973} {"Full-finetune/Learning Rate": 1.7867177932517634e-05, "Full-finetune/Loss": 0.8280878663063049, "Full-finetune/Loss (Raw)": 0.9067807793617249, "Full-finetune/Step": 1223, "Full-finetune/Step Time": 6.757837243378162} {"Full-finetune/Learning Rate": 1.786333078771597e-05, "Full-finetune/Loss": 0.8281283378601074, "Full-finetune/Loss (Raw)": 0.9201835989952087, "Full-finetune/Step": 1224, "Full-finetune/Step Time": 6.758669059723616} {"Full-finetune/Learning Rate": 1.785948059130915e-05, "Full-finetune/Loss": 0.827362596988678, "Full-finetune/Loss (Raw)": 0.7501983046531677, "Full-finetune/Step": 1225, "Full-finetune/Step Time": 6.74387495405972} {"Full-finetune/Learning Rate": 1.785562734479135e-05, "Full-finetune/Loss": 0.8271534442901611, "Full-finetune/Loss (Raw)": 0.7978994846343994, "Full-finetune/Step": 1226, "Full-finetune/Step Time": 6.734506446868181} {"Full-finetune/Learning Rate": 1.785177104965795e-05, "Full-finetune/Loss": 0.8255414962768555, "Full-finetune/Loss (Raw)": 0.6959323287010193, "Full-finetune/Step": 1227, "Full-finetune/Step Time": 6.742479845881462} {"Full-finetune/Learning Rate": 1.7847911707405494e-05, "Full-finetune/Loss": 0.8253306150436401, "Full-finetune/Loss (Raw)": 0.7894026637077332, "Full-finetune/Step": 1228, "Full-finetune/Step Time": 6.747828816995025} {"Full-finetune/Learning Rate": 1.7844049319531725e-05, "Full-finetune/Loss": 0.8260998725891113, "Full-finetune/Loss (Raw)": 0.998511016368866, "Full-finetune/Step": 1229, "Full-finetune/Step Time": 6.743715872988105} {"Full-finetune/Learning Rate": 1.7840183887535558e-05, "Full-finetune/Loss": 0.8237625360488892, "Full-finetune/Loss (Raw)": 0.41960129141807556, "Full-finetune/Step": 1230, "Full-finetune/Step Time": 6.761575093492866} {"Full-finetune/Learning Rate": 1.7836315412917088e-05, "Full-finetune/Loss": 0.8238472938537598, "Full-finetune/Loss (Raw)": 1.0099866390228271, "Full-finetune/Step": 1231, "Full-finetune/Step Time": 6.755648506805301} {"Full-finetune/Learning Rate": 1.78324438971776e-05, "Full-finetune/Loss": 0.8226736783981323, "Full-finetune/Loss (Raw)": 0.845384418964386, "Full-finetune/Step": 1232, "Full-finetune/Step Time": 6.743765562772751} {"Full-finetune/Learning Rate": 1.782856934181955e-05, "Full-finetune/Loss": 0.8233851194381714, "Full-finetune/Loss (Raw)": 0.9326591491699219, "Full-finetune/Step": 1233, "Full-finetune/Step Time": 6.730071501806378} {"Full-finetune/Learning Rate": 1.782469174834658e-05, "Full-finetune/Loss": 0.8229869604110718, "Full-finetune/Loss (Raw)": 0.7076968550682068, "Full-finetune/Step": 1234, "Full-finetune/Step Time": 6.737978756427765} {"Full-finetune/Learning Rate": 1.7820811118263503e-05, "Full-finetune/Loss": 0.8225604295730591, "Full-finetune/Loss (Raw)": 0.8816477656364441, "Full-finetune/Step": 1235, "Full-finetune/Step Time": 6.736470513045788} {"Full-finetune/Learning Rate": 1.7816927453076324e-05, "Full-finetune/Loss": 0.8224449157714844, "Full-finetune/Loss (Raw)": 1.0016281604766846, "Full-finetune/Step": 1236, "Full-finetune/Step Time": 6.731997575610876} {"Full-finetune/Learning Rate": 1.781304075429221e-05, "Full-finetune/Loss": 0.8253227472305298, "Full-finetune/Loss (Raw)": 0.9737921953201294, "Full-finetune/Step": 1237, "Full-finetune/Step Time": 6.733645109459758} {"Full-finetune/Learning Rate": 1.7809151023419517e-05, "Full-finetune/Loss": 0.8263275623321533, "Full-finetune/Loss (Raw)": 1.0150212049484253, "Full-finetune/Step": 1238, "Full-finetune/Step Time": 6.736425893381238} {"Full-finetune/Learning Rate": 1.7805258261967774e-05, "Full-finetune/Loss": 0.82526695728302, "Full-finetune/Loss (Raw)": 0.9035535454750061, "Full-finetune/Step": 1239, "Full-finetune/Step Time": 6.743290381506085} {"Full-finetune/Learning Rate": 1.780136247144768e-05, "Full-finetune/Loss": 0.8258042335510254, "Full-finetune/Loss (Raw)": 0.87552410364151, "Full-finetune/Step": 1240, "Full-finetune/Step Time": 6.7405263017863035} {"Full-finetune/Learning Rate": 1.7797463653371125e-05, "Full-finetune/Loss": 0.825329065322876, "Full-finetune/Loss (Raw)": 0.80057293176651, "Full-finetune/Step": 1241, "Full-finetune/Step Time": 6.744484722614288} {"Full-finetune/Learning Rate": 1.7793561809251153e-05, "Full-finetune/Loss": 0.8263732194900513, "Full-finetune/Loss (Raw)": 0.9023850560188293, "Full-finetune/Step": 1242, "Full-finetune/Step Time": 6.717799125239253} {"Full-finetune/Learning Rate": 1.7789656940602003e-05, "Full-finetune/Loss": 0.8266794681549072, "Full-finetune/Loss (Raw)": 0.930213987827301, "Full-finetune/Step": 1243, "Full-finetune/Step Time": 6.711444944143295} {"Full-finetune/Learning Rate": 1.7785749048939076e-05, "Full-finetune/Loss": 0.8258261680603027, "Full-finetune/Loss (Raw)": 0.8625098466873169, "Full-finetune/Step": 1244, "Full-finetune/Step Time": 6.713574381545186} {"Full-finetune/Learning Rate": 1.7781838135778945e-05, "Full-finetune/Loss": 0.8256184458732605, "Full-finetune/Loss (Raw)": 0.7701602578163147, "Full-finetune/Step": 1245, "Full-finetune/Step Time": 6.7037772461771965} {"Full-finetune/Learning Rate": 1.777792420263936e-05, "Full-finetune/Loss": 0.8250315189361572, "Full-finetune/Loss (Raw)": 0.829007625579834, "Full-finetune/Step": 1246, "Full-finetune/Step Time": 6.7250226475298405} {"Full-finetune/Learning Rate": 1.7774007251039246e-05, "Full-finetune/Loss": 0.8280658721923828, "Full-finetune/Loss (Raw)": 0.9876561760902405, "Full-finetune/Step": 1247, "Full-finetune/Step Time": 6.710148394107819} {"Full-finetune/Learning Rate": 1.777008728249869e-05, "Full-finetune/Loss": 0.8269539475440979, "Full-finetune/Loss (Raw)": 0.7193640470504761, "Full-finetune/Step": 1248, "Full-finetune/Step Time": 6.714562086388469} {"Full-finetune/Learning Rate": 1.776616429853896e-05, "Full-finetune/Loss": 0.8256511688232422, "Full-finetune/Loss (Raw)": 0.8283026814460754, "Full-finetune/Step": 1249, "Full-finetune/Step Time": 6.723332293331623} {"Full-finetune/Learning Rate": 1.7762238300682484e-05, "Full-finetune/Loss": 0.8258306980133057, "Full-finetune/Loss (Raw)": 0.920802116394043, "Full-finetune/Step": 1250, "Full-finetune/Step Time": 6.702801490202546} {"Full-finetune/Learning Rate": 1.775830929045287e-05, "Full-finetune/Loss": 0.8271770477294922, "Full-finetune/Loss (Raw)": 0.9221222400665283, "Full-finetune/Step": 1251, "Full-finetune/Step Time": 6.688932858407497} {"Full-finetune/Learning Rate": 1.7754377269374893e-05, "Full-finetune/Loss": 0.8249531984329224, "Full-finetune/Loss (Raw)": 0.7882331609725952, "Full-finetune/Step": 1252, "Full-finetune/Step Time": 6.684126930311322} {"Full-finetune/Learning Rate": 1.7750442238974483e-05, "Full-finetune/Loss": 0.8245184421539307, "Full-finetune/Loss (Raw)": 0.786846399307251, "Full-finetune/Step": 1253, "Full-finetune/Step Time": 6.702727423980832} {"Full-finetune/Learning Rate": 1.7746504200778756e-05, "Full-finetune/Loss": 0.8243666887283325, "Full-finetune/Loss (Raw)": 0.8717493414878845, "Full-finetune/Step": 1254, "Full-finetune/Step Time": 6.706728370860219} {"Full-finetune/Learning Rate": 1.7742563156315987e-05, "Full-finetune/Loss": 0.8252925872802734, "Full-finetune/Loss (Raw)": 0.9807261824607849, "Full-finetune/Step": 1255, "Full-finetune/Step Time": 6.7079990934580564} {"Full-finetune/Learning Rate": 1.773861910711562e-05, "Full-finetune/Loss": 0.8262895345687866, "Full-finetune/Loss (Raw)": 0.9109138250350952, "Full-finetune/Step": 1256, "Full-finetune/Step Time": 6.7039246298372746} {"Full-finetune/Learning Rate": 1.773467205470826e-05, "Full-finetune/Loss": 0.8234465718269348, "Full-finetune/Loss (Raw)": 0.5829275846481323, "Full-finetune/Step": 1257, "Full-finetune/Step Time": 6.729599867016077} {"Full-finetune/Learning Rate": 1.7730722000625683e-05, "Full-finetune/Loss": 0.8252567052841187, "Full-finetune/Loss (Raw)": 0.8798483610153198, "Full-finetune/Step": 1258, "Full-finetune/Step Time": 6.718114383518696} {"Full-finetune/Learning Rate": 1.7726768946400825e-05, "Full-finetune/Loss": 0.8240150213241577, "Full-finetune/Loss (Raw)": 0.7604371309280396, "Full-finetune/Step": 1259, "Full-finetune/Step Time": 6.7289905324578285} {"Full-finetune/Learning Rate": 1.772281289356779e-05, "Full-finetune/Loss": 0.8251733183860779, "Full-finetune/Loss (Raw)": 1.0931605100631714, "Full-finetune/Step": 1260, "Full-finetune/Step Time": 6.741729129105806} {"Full-finetune/Learning Rate": 1.771885384366185e-05, "Full-finetune/Loss": 0.8241493701934814, "Full-finetune/Loss (Raw)": 0.7505043745040894, "Full-finetune/Step": 1261, "Full-finetune/Step Time": 6.7406144961714745} {"Full-finetune/Learning Rate": 1.7714891798219432e-05, "Full-finetune/Loss": 0.8245564699172974, "Full-finetune/Loss (Raw)": 0.9145756959915161, "Full-finetune/Step": 1262, "Full-finetune/Step Time": 6.73677215538919} {"Full-finetune/Learning Rate": 1.7710926758778123e-05, "Full-finetune/Loss": 0.8245221376419067, "Full-finetune/Loss (Raw)": 0.9367495179176331, "Full-finetune/Step": 1263, "Full-finetune/Step Time": 6.733882075175643} {"Full-finetune/Learning Rate": 1.7706958726876685e-05, "Full-finetune/Loss": 0.8248502016067505, "Full-finetune/Loss (Raw)": 0.8745775818824768, "Full-finetune/Step": 1264, "Full-finetune/Step Time": 6.732710208743811} {"Full-finetune/Learning Rate": 1.770298770405503e-05, "Full-finetune/Loss": 0.8203476667404175, "Full-finetune/Loss (Raw)": 0.39364054799079895, "Full-finetune/Step": 1265, "Full-finetune/Step Time": 6.750730399042368} {"Full-finetune/Learning Rate": 1.7699013691854238e-05, "Full-finetune/Loss": 0.8199383020401001, "Full-finetune/Loss (Raw)": 0.8278496265411377, "Full-finetune/Step": 1266, "Full-finetune/Step Time": 6.758351827040315} {"Full-finetune/Learning Rate": 1.769503669181654e-05, "Full-finetune/Loss": 0.8209260702133179, "Full-finetune/Loss (Raw)": 0.9428470134735107, "Full-finetune/Step": 1267, "Full-finetune/Step Time": 6.7445562947541475} {"Full-finetune/Learning Rate": 1.7691056705485333e-05, "Full-finetune/Loss": 0.8217921257019043, "Full-finetune/Loss (Raw)": 0.9944141507148743, "Full-finetune/Step": 1268, "Full-finetune/Step Time": 6.738754190504551} {"Full-finetune/Learning Rate": 1.7687073734405177e-05, "Full-finetune/Loss": 0.821552574634552, "Full-finetune/Loss (Raw)": 0.812816858291626, "Full-finetune/Step": 1269, "Full-finetune/Step Time": 6.730441357940435} {"Full-finetune/Learning Rate": 1.768308778012178e-05, "Full-finetune/Loss": 0.8263476490974426, "Full-finetune/Loss (Raw)": 0.9141586422920227, "Full-finetune/Step": 1270, "Full-finetune/Step Time": 6.722844084724784} {"Full-finetune/Learning Rate": 1.767909884418202e-05, "Full-finetune/Loss": 0.8275967836380005, "Full-finetune/Loss (Raw)": 1.0014876127243042, "Full-finetune/Step": 1271, "Full-finetune/Step Time": 6.707379575818777} {"Full-finetune/Learning Rate": 1.7675106928133915e-05, "Full-finetune/Loss": 0.8265013098716736, "Full-finetune/Loss (Raw)": 0.7198726534843445, "Full-finetune/Step": 1272, "Full-finetune/Step Time": 6.716323859989643} {"Full-finetune/Learning Rate": 1.7671112033526655e-05, "Full-finetune/Loss": 0.8257529735565186, "Full-finetune/Loss (Raw)": 0.7586050629615784, "Full-finetune/Step": 1273, "Full-finetune/Step Time": 6.725765714421868} {"Full-finetune/Learning Rate": 1.7667114161910586e-05, "Full-finetune/Loss": 0.8247269988059998, "Full-finetune/Loss (Raw)": 0.8138783574104309, "Full-finetune/Step": 1274, "Full-finetune/Step Time": 6.717977747321129} {"Full-finetune/Learning Rate": 1.7663113314837192e-05, "Full-finetune/Loss": 0.8274576663970947, "Full-finetune/Loss (Raw)": 1.0273205041885376, "Full-finetune/Step": 1275, "Full-finetune/Step Time": 6.711460407823324} {"Full-finetune/Learning Rate": 1.7659109493859134e-05, "Full-finetune/Loss": 0.8312746286392212, "Full-finetune/Loss (Raw)": 0.8452091217041016, "Full-finetune/Step": 1276, "Full-finetune/Step Time": 6.688759833574295} {"Full-finetune/Learning Rate": 1.7655102700530212e-05, "Full-finetune/Loss": 0.8300097584724426, "Full-finetune/Loss (Raw)": 0.8936628103256226, "Full-finetune/Step": 1277, "Full-finetune/Step Time": 6.6800464149564505} {"Full-finetune/Learning Rate": 1.7651092936405385e-05, "Full-finetune/Loss": 0.8286296725273132, "Full-finetune/Loss (Raw)": 0.5931284427642822, "Full-finetune/Step": 1278, "Full-finetune/Step Time": 6.6869701743125916} {"Full-finetune/Learning Rate": 1.7647080203040765e-05, "Full-finetune/Loss": 0.8294407725334167, "Full-finetune/Loss (Raw)": 0.8801037669181824, "Full-finetune/Step": 1279, "Full-finetune/Step Time": 6.689104883000255} {"Full-finetune/Learning Rate": 1.7643064501993614e-05, "Full-finetune/Loss": 0.8308577537536621, "Full-finetune/Loss (Raw)": 1.0050276517868042, "Full-finetune/Step": 1280, "Full-finetune/Step Time": 6.669648837298155} {"Full-finetune/Learning Rate": 1.763904583482235e-05, "Full-finetune/Loss": 0.830854058265686, "Full-finetune/Loss (Raw)": 0.9111778140068054, "Full-finetune/Step": 1281, "Full-finetune/Step Time": 6.655025748535991} {"Full-finetune/Learning Rate": 1.7635024203086532e-05, "Full-finetune/Loss": 0.835099995136261, "Full-finetune/Loss (Raw)": 0.873386561870575, "Full-finetune/Step": 1282, "Full-finetune/Step Time": 6.639076625928283} {"Full-finetune/Learning Rate": 1.7630999608346888e-05, "Full-finetune/Loss": 0.8377906084060669, "Full-finetune/Loss (Raw)": 0.6853682398796082, "Full-finetune/Step": 1283, "Full-finetune/Step Time": 6.632803399115801} {"Full-finetune/Learning Rate": 1.7626972052165275e-05, "Full-finetune/Loss": 0.8365921378135681, "Full-finetune/Loss (Raw)": 0.850001871585846, "Full-finetune/Step": 1284, "Full-finetune/Step Time": 6.635957973077893} {"Full-finetune/Learning Rate": 1.7622941536104715e-05, "Full-finetune/Loss": 0.8367196321487427, "Full-finetune/Loss (Raw)": 0.9686108231544495, "Full-finetune/Step": 1285, "Full-finetune/Step Time": 6.636353705078363} {"Full-finetune/Learning Rate": 1.7618908061729373e-05, "Full-finetune/Loss": 0.8371241092681885, "Full-finetune/Loss (Raw)": 0.8444668650627136, "Full-finetune/Step": 1286, "Full-finetune/Step Time": 6.652760643512011} {"Full-finetune/Learning Rate": 1.7614871630604554e-05, "Full-finetune/Loss": 0.8378567695617676, "Full-finetune/Loss (Raw)": 0.8688725829124451, "Full-finetune/Step": 1287, "Full-finetune/Step Time": 6.652588313445449} {"Full-finetune/Learning Rate": 1.761083224429672e-05, "Full-finetune/Loss": 0.8378515839576721, "Full-finetune/Loss (Raw)": 0.8355498313903809, "Full-finetune/Step": 1288, "Full-finetune/Step Time": 6.6354229766875505} {"Full-finetune/Learning Rate": 1.760678990437349e-05, "Full-finetune/Loss": 0.8380911350250244, "Full-finetune/Loss (Raw)": 0.8267464637756348, "Full-finetune/Step": 1289, "Full-finetune/Step Time": 6.620211264118552} {"Full-finetune/Learning Rate": 1.76027446124036e-05, "Full-finetune/Loss": 0.838585376739502, "Full-finetune/Loss (Raw)": 1.0091660022735596, "Full-finetune/Step": 1290, "Full-finetune/Step Time": 6.614463813602924} {"Full-finetune/Learning Rate": 1.759869636995696e-05, "Full-finetune/Loss": 0.8398328423500061, "Full-finetune/Loss (Raw)": 0.7785916328430176, "Full-finetune/Step": 1291, "Full-finetune/Step Time": 6.6127242762595415} {"Full-finetune/Learning Rate": 1.7594645178604614e-05, "Full-finetune/Loss": 0.8380088210105896, "Full-finetune/Loss (Raw)": 0.8258658647537231, "Full-finetune/Step": 1292, "Full-finetune/Step Time": 6.607618583366275} {"Full-finetune/Learning Rate": 1.7590591039918744e-05, "Full-finetune/Loss": 0.8421859741210938, "Full-finetune/Loss (Raw)": 0.8970291614532471, "Full-finetune/Step": 1293, "Full-finetune/Step Time": 6.591998394578695} {"Full-finetune/Learning Rate": 1.7586533955472686e-05, "Full-finetune/Loss": 0.8432904481887817, "Full-finetune/Loss (Raw)": 0.9395909905433655, "Full-finetune/Step": 1294, "Full-finetune/Step Time": 6.595469566062093} {"Full-finetune/Learning Rate": 1.7582473926840916e-05, "Full-finetune/Loss": 0.844917356967926, "Full-finetune/Loss (Raw)": 0.8217697739601135, "Full-finetune/Step": 1295, "Full-finetune/Step Time": 6.594668377190828} {"Full-finetune/Learning Rate": 1.7578410955599046e-05, "Full-finetune/Loss": 0.8450653553009033, "Full-finetune/Loss (Raw)": 0.7987710237503052, "Full-finetune/Step": 1296, "Full-finetune/Step Time": 6.600637098774314} {"Full-finetune/Learning Rate": 1.7574345043323843e-05, "Full-finetune/Loss": 0.8442975282669067, "Full-finetune/Loss (Raw)": 0.8180038928985596, "Full-finetune/Step": 1297, "Full-finetune/Step Time": 6.615323139354587} {"Full-finetune/Learning Rate": 1.7570276191593202e-05, "Full-finetune/Loss": 0.8464308977127075, "Full-finetune/Loss (Raw)": 0.7500502467155457, "Full-finetune/Step": 1298, "Full-finetune/Step Time": 6.594546312466264} {"Full-finetune/Learning Rate": 1.7566204401986172e-05, "Full-finetune/Loss": 0.8439292907714844, "Full-finetune/Loss (Raw)": 0.664905309677124, "Full-finetune/Step": 1299, "Full-finetune/Step Time": 6.59855748899281} {"Full-finetune/Learning Rate": 1.7562129676082933e-05, "Full-finetune/Loss": 0.8432006239891052, "Full-finetune/Loss (Raw)": 0.8092948794364929, "Full-finetune/Step": 1300, "Full-finetune/Step Time": 6.596305202692747} {"Full-finetune/Learning Rate": 1.75580520154648e-05, "Full-finetune/Loss": 0.8429827690124512, "Full-finetune/Loss (Raw)": 0.8412618041038513, "Full-finetune/Step": 1301, "Full-finetune/Step Time": 6.614184292033315} {"Full-finetune/Learning Rate": 1.755397142171424e-05, "Full-finetune/Loss": 0.8465826511383057, "Full-finetune/Loss (Raw)": 0.8239796161651611, "Full-finetune/Step": 1302, "Full-finetune/Step Time": 6.617646615952253} {"Full-finetune/Learning Rate": 1.7549887896414853e-05, "Full-finetune/Loss": 0.8470179438591003, "Full-finetune/Loss (Raw)": 0.8350473642349243, "Full-finetune/Step": 1303, "Full-finetune/Step Time": 6.600030928850174} {"Full-finetune/Learning Rate": 1.7545801441151368e-05, "Full-finetune/Loss": 0.8479142785072327, "Full-finetune/Loss (Raw)": 0.9304107427597046, "Full-finetune/Step": 1304, "Full-finetune/Step Time": 6.605286655947566} {"Full-finetune/Learning Rate": 1.7541712057509666e-05, "Full-finetune/Loss": 0.8521291613578796, "Full-finetune/Loss (Raw)": 0.8327091932296753, "Full-finetune/Step": 1305, "Full-finetune/Step Time": 6.592254294082522} {"Full-finetune/Learning Rate": 1.7537619747076752e-05, "Full-finetune/Loss": 0.8560246229171753, "Full-finetune/Loss (Raw)": 0.8256189227104187, "Full-finetune/Step": 1306, "Full-finetune/Step Time": 6.589477887377143} {"Full-finetune/Learning Rate": 1.7533524511440777e-05, "Full-finetune/Loss": 0.8557535409927368, "Full-finetune/Loss (Raw)": 0.9937397241592407, "Full-finetune/Step": 1307, "Full-finetune/Step Time": 6.591942995786667} {"Full-finetune/Learning Rate": 1.7529426352191014e-05, "Full-finetune/Loss": 0.8576581478118896, "Full-finetune/Loss (Raw)": 1.0168166160583496, "Full-finetune/Step": 1308, "Full-finetune/Step Time": 6.575252944603562} {"Full-finetune/Learning Rate": 1.752532527091789e-05, "Full-finetune/Loss": 0.8566005229949951, "Full-finetune/Loss (Raw)": 0.852791965007782, "Full-finetune/Step": 1309, "Full-finetune/Step Time": 6.592788223177195} {"Full-finetune/Learning Rate": 1.7521221269212944e-05, "Full-finetune/Loss": 0.8588896989822388, "Full-finetune/Loss (Raw)": 0.8314002156257629, "Full-finetune/Step": 1310, "Full-finetune/Step Time": 6.5908685475587845} {"Full-finetune/Learning Rate": 1.751711434866886e-05, "Full-finetune/Loss": 0.8585265278816223, "Full-finetune/Loss (Raw)": 0.9029839038848877, "Full-finetune/Step": 1311, "Full-finetune/Step Time": 6.603576624765992} {"Full-finetune/Learning Rate": 1.751300451087946e-05, "Full-finetune/Loss": 0.8577656745910645, "Full-finetune/Loss (Raw)": 0.5901167392730713, "Full-finetune/Step": 1312, "Full-finetune/Step Time": 6.597805881872773} {"Full-finetune/Learning Rate": 1.7508891757439687e-05, "Full-finetune/Loss": 0.8577159643173218, "Full-finetune/Loss (Raw)": 0.8783745169639587, "Full-finetune/Step": 1313, "Full-finetune/Step Time": 6.604581482708454} {"Full-finetune/Learning Rate": 1.7504776089945626e-05, "Full-finetune/Loss": 0.8571882247924805, "Full-finetune/Loss (Raw)": 0.9470111727714539, "Full-finetune/Step": 1314, "Full-finetune/Step Time": 6.601312423124909} {"Full-finetune/Learning Rate": 1.7500657509994478e-05, "Full-finetune/Loss": 0.861773669719696, "Full-finetune/Loss (Raw)": 0.934572696685791, "Full-finetune/Step": 1315, "Full-finetune/Step Time": 6.585236370563507} {"Full-finetune/Learning Rate": 1.7496536019184588e-05, "Full-finetune/Loss": 0.8629283308982849, "Full-finetune/Loss (Raw)": 1.0583109855651855, "Full-finetune/Step": 1316, "Full-finetune/Step Time": 6.587247300893068} {"Full-finetune/Learning Rate": 1.749241161911543e-05, "Full-finetune/Loss": 0.8628264665603638, "Full-finetune/Loss (Raw)": 0.9593700170516968, "Full-finetune/Step": 1317, "Full-finetune/Step Time": 6.579271649941802} {"Full-finetune/Learning Rate": 1.7488284311387597e-05, "Full-finetune/Loss": 0.8626400232315063, "Full-finetune/Loss (Raw)": 0.8563376665115356, "Full-finetune/Step": 1318, "Full-finetune/Step Time": 6.579592077061534} {"Full-finetune/Learning Rate": 1.7484154097602822e-05, "Full-finetune/Loss": 0.8623778820037842, "Full-finetune/Loss (Raw)": 0.8854849338531494, "Full-finetune/Step": 1319, "Full-finetune/Step Time": 6.577220702543855} {"Full-finetune/Learning Rate": 1.7480020979363956e-05, "Full-finetune/Loss": 0.8607906699180603, "Full-finetune/Loss (Raw)": 0.7669383883476257, "Full-finetune/Step": 1320, "Full-finetune/Step Time": 6.588234415277839} {"Full-finetune/Learning Rate": 1.747588495827498e-05, "Full-finetune/Loss": 0.8607730269432068, "Full-finetune/Loss (Raw)": 0.9024336338043213, "Full-finetune/Step": 1321, "Full-finetune/Step Time": 6.59084996022284} {"Full-finetune/Learning Rate": 1.7471746035941007e-05, "Full-finetune/Loss": 0.8605934381484985, "Full-finetune/Loss (Raw)": 0.8884052634239197, "Full-finetune/Step": 1322, "Full-finetune/Step Time": 6.580518539994955} {"Full-finetune/Learning Rate": 1.7467604213968274e-05, "Full-finetune/Loss": 0.8611131310462952, "Full-finetune/Loss (Raw)": 0.9295578598976135, "Full-finetune/Step": 1323, "Full-finetune/Step Time": 6.5808030590415} {"Full-finetune/Learning Rate": 1.746345949396414e-05, "Full-finetune/Loss": 0.8616416454315186, "Full-finetune/Loss (Raw)": 0.9646430611610413, "Full-finetune/Step": 1324, "Full-finetune/Step Time": 6.572730397805572} {"Full-finetune/Learning Rate": 1.7459311877537085e-05, "Full-finetune/Loss": 0.861749529838562, "Full-finetune/Loss (Raw)": 0.9813515543937683, "Full-finetune/Step": 1325, "Full-finetune/Step Time": 6.589130885899067} {"Full-finetune/Learning Rate": 1.745516136629672e-05, "Full-finetune/Loss": 0.8622149229049683, "Full-finetune/Loss (Raw)": 0.960111141204834, "Full-finetune/Step": 1326, "Full-finetune/Step Time": 6.582172483205795} {"Full-finetune/Learning Rate": 1.7451007961853778e-05, "Full-finetune/Loss": 0.8604669570922852, "Full-finetune/Loss (Raw)": 0.7442650198936462, "Full-finetune/Step": 1327, "Full-finetune/Step Time": 6.618308763951063} {"Full-finetune/Learning Rate": 1.7446851665820118e-05, "Full-finetune/Loss": 0.861413836479187, "Full-finetune/Loss (Raw)": 1.0686368942260742, "Full-finetune/Step": 1328, "Full-finetune/Step Time": 6.6123841274529696} {"Full-finetune/Learning Rate": 1.744269247980871e-05, "Full-finetune/Loss": 0.8622233867645264, "Full-finetune/Loss (Raw)": 1.0881603956222534, "Full-finetune/Step": 1329, "Full-finetune/Step Time": 6.605286166071892} {"Full-finetune/Learning Rate": 1.7438530405433663e-05, "Full-finetune/Loss": 0.8592442870140076, "Full-finetune/Loss (Raw)": 0.7037379741668701, "Full-finetune/Step": 1330, "Full-finetune/Step Time": 6.57491304166615} {"Full-finetune/Learning Rate": 1.7434365444310185e-05, "Full-finetune/Loss": 0.8584474921226501, "Full-finetune/Loss (Raw)": 0.932151734828949, "Full-finetune/Step": 1331, "Full-finetune/Step Time": 6.568942567333579} {"Full-finetune/Learning Rate": 1.7430197598054622e-05, "Full-finetune/Loss": 0.8568496704101562, "Full-finetune/Loss (Raw)": 0.7600030899047852, "Full-finetune/Step": 1332, "Full-finetune/Step Time": 6.571942120790482} {"Full-finetune/Learning Rate": 1.7426026868284438e-05, "Full-finetune/Loss": 0.8579171895980835, "Full-finetune/Loss (Raw)": 0.9149764776229858, "Full-finetune/Step": 1333, "Full-finetune/Step Time": 6.558370498940349} {"Full-finetune/Learning Rate": 1.7421853256618204e-05, "Full-finetune/Loss": 0.8571069240570068, "Full-finetune/Loss (Raw)": 0.9328803420066833, "Full-finetune/Step": 1334, "Full-finetune/Step Time": 6.561941096559167} {"Full-finetune/Learning Rate": 1.741767676467562e-05, "Full-finetune/Loss": 0.8573697805404663, "Full-finetune/Loss (Raw)": 0.9185110926628113, "Full-finetune/Step": 1335, "Full-finetune/Step Time": 6.553549457341433} {"Full-finetune/Learning Rate": 1.7413497394077502e-05, "Full-finetune/Loss": 0.8574050664901733, "Full-finetune/Loss (Raw)": 0.9181907773017883, "Full-finetune/Step": 1336, "Full-finetune/Step Time": 6.549793964251876} {"Full-finetune/Learning Rate": 1.7409315146445783e-05, "Full-finetune/Loss": 0.857207715511322, "Full-finetune/Loss (Raw)": 0.9259286522865295, "Full-finetune/Step": 1337, "Full-finetune/Step Time": 6.552031772211194} {"Full-finetune/Learning Rate": 1.7405130023403513e-05, "Full-finetune/Loss": 0.8531565070152283, "Full-finetune/Loss (Raw)": 0.3366911709308624, "Full-finetune/Step": 1338, "Full-finetune/Step Time": 6.572264784947038} {"Full-finetune/Learning Rate": 1.740094202657485e-05, "Full-finetune/Loss": 0.8522400259971619, "Full-finetune/Loss (Raw)": 0.761168360710144, "Full-finetune/Step": 1339, "Full-finetune/Step Time": 6.579096348956227} {"Full-finetune/Learning Rate": 1.7396751157585078e-05, "Full-finetune/Loss": 0.8513733148574829, "Full-finetune/Loss (Raw)": 0.8453392386436462, "Full-finetune/Step": 1340, "Full-finetune/Step Time": 6.585227318108082} {"Full-finetune/Learning Rate": 1.739255741806059e-05, "Full-finetune/Loss": 0.851249098777771, "Full-finetune/Loss (Raw)": 0.8325870633125305, "Full-finetune/Step": 1341, "Full-finetune/Step Time": 6.586529619991779} {"Full-finetune/Learning Rate": 1.73883608096289e-05, "Full-finetune/Loss": 0.8525323867797852, "Full-finetune/Loss (Raw)": 0.9276672005653381, "Full-finetune/Step": 1342, "Full-finetune/Step Time": 6.584016751497984} {"Full-finetune/Learning Rate": 1.738416133391862e-05, "Full-finetune/Loss": 0.8516884446144104, "Full-finetune/Loss (Raw)": 0.8142594695091248, "Full-finetune/Step": 1343, "Full-finetune/Step Time": 6.606106782332063} {"Full-finetune/Learning Rate": 1.7379958992559494e-05, "Full-finetune/Loss": 0.855118989944458, "Full-finetune/Loss (Raw)": 0.7970010638237, "Full-finetune/Step": 1344, "Full-finetune/Step Time": 6.594143519178033} {"Full-finetune/Learning Rate": 1.7375753787182364e-05, "Full-finetune/Loss": 0.8549147844314575, "Full-finetune/Loss (Raw)": 1.002966284751892, "Full-finetune/Step": 1345, "Full-finetune/Step Time": 6.592660054564476} {"Full-finetune/Learning Rate": 1.7371545719419188e-05, "Full-finetune/Loss": 0.8543246984481812, "Full-finetune/Loss (Raw)": 0.8874238133430481, "Full-finetune/Step": 1346, "Full-finetune/Step Time": 6.59440129622817} {"Full-finetune/Learning Rate": 1.7367334790903034e-05, "Full-finetune/Loss": 0.8593623638153076, "Full-finetune/Loss (Raw)": 0.9532288908958435, "Full-finetune/Step": 1347, "Full-finetune/Step Time": 6.593676812946796} {"Full-finetune/Learning Rate": 1.736312100326808e-05, "Full-finetune/Loss": 0.8560687899589539, "Full-finetune/Loss (Raw)": 0.4416283071041107, "Full-finetune/Step": 1348, "Full-finetune/Step Time": 6.612559923902154} {"Full-finetune/Learning Rate": 1.7358904358149625e-05, "Full-finetune/Loss": 0.8570454120635986, "Full-finetune/Loss (Raw)": 0.9585469365119934, "Full-finetune/Step": 1349, "Full-finetune/Step Time": 6.595018669962883} {"Full-finetune/Learning Rate": 1.7354684857184056e-05, "Full-finetune/Loss": 0.8577477931976318, "Full-finetune/Loss (Raw)": 0.9049950242042542, "Full-finetune/Step": 1350, "Full-finetune/Step Time": 6.5934843104332685} {"Full-finetune/Learning Rate": 1.735046250200888e-05, "Full-finetune/Loss": 0.8578568696975708, "Full-finetune/Loss (Raw)": 0.9207379221916199, "Full-finetune/Step": 1351, "Full-finetune/Step Time": 6.597434284165502} {"Full-finetune/Learning Rate": 1.7346237294262717e-05, "Full-finetune/Loss": 0.8540557622909546, "Full-finetune/Loss (Raw)": 0.43364447355270386, "Full-finetune/Step": 1352, "Full-finetune/Step Time": 6.614454988390207} {"Full-finetune/Learning Rate": 1.734200923558528e-05, "Full-finetune/Loss": 0.8548890352249146, "Full-finetune/Loss (Raw)": 0.8568681478500366, "Full-finetune/Step": 1353, "Full-finetune/Step Time": 6.615857692435384} {"Full-finetune/Learning Rate": 1.7337778327617397e-05, "Full-finetune/Loss": 0.8555519580841064, "Full-finetune/Loss (Raw)": 0.8827441334724426, "Full-finetune/Step": 1354, "Full-finetune/Step Time": 6.612778753042221} {"Full-finetune/Learning Rate": 1.733354457200101e-05, "Full-finetune/Loss": 0.8580355644226074, "Full-finetune/Loss (Raw)": 1.0138375759124756, "Full-finetune/Step": 1355, "Full-finetune/Step Time": 6.606165062636137} {"Full-finetune/Learning Rate": 1.7329307970379145e-05, "Full-finetune/Loss": 0.8597267270088196, "Full-finetune/Loss (Raw)": 1.0058661699295044, "Full-finetune/Step": 1356, "Full-finetune/Step Time": 6.600450089201331} {"Full-finetune/Learning Rate": 1.732506852439595e-05, "Full-finetune/Loss": 0.8589030504226685, "Full-finetune/Loss (Raw)": 0.8930805921554565, "Full-finetune/Step": 1357, "Full-finetune/Step Time": 6.613481568172574} {"Full-finetune/Learning Rate": 1.7320826235696672e-05, "Full-finetune/Loss": 0.8627623319625854, "Full-finetune/Loss (Raw)": 0.913595974445343, "Full-finetune/Step": 1358, "Full-finetune/Step Time": 6.5961162354797125} {"Full-finetune/Learning Rate": 1.7316581105927658e-05, "Full-finetune/Loss": 0.8610086441040039, "Full-finetune/Loss (Raw)": 0.7855168581008911, "Full-finetune/Step": 1359, "Full-finetune/Step Time": 6.609850041568279} {"Full-finetune/Learning Rate": 1.731233313673636e-05, "Full-finetune/Loss": 0.8617959022521973, "Full-finetune/Loss (Raw)": 0.946140706539154, "Full-finetune/Step": 1360, "Full-finetune/Step Time": 6.613604590296745} {"Full-finetune/Learning Rate": 1.7308082329771335e-05, "Full-finetune/Loss": 0.8623278141021729, "Full-finetune/Loss (Raw)": 1.0007559061050415, "Full-finetune/Step": 1361, "Full-finetune/Step Time": 6.625249680131674} {"Full-finetune/Learning Rate": 1.7303828686682233e-05, "Full-finetune/Loss": 0.8636466264724731, "Full-finetune/Loss (Raw)": 0.87650066614151, "Full-finetune/Step": 1362, "Full-finetune/Step Time": 6.609607387334108} {"Full-finetune/Learning Rate": 1.7299572209119813e-05, "Full-finetune/Loss": 0.8634197115898132, "Full-finetune/Loss (Raw)": 0.8526017069816589, "Full-finetune/Step": 1363, "Full-finetune/Step Time": 6.610224729403853} {"Full-finetune/Learning Rate": 1.7295312898735933e-05, "Full-finetune/Loss": 0.8617799878120422, "Full-finetune/Loss (Raw)": 0.7917412519454956, "Full-finetune/Step": 1364, "Full-finetune/Step Time": 6.648326514288783} {"Full-finetune/Learning Rate": 1.7291050757183546e-05, "Full-finetune/Loss": 0.8602943420410156, "Full-finetune/Loss (Raw)": 0.7836249470710754, "Full-finetune/Step": 1365, "Full-finetune/Step Time": 6.6395533457398415} {"Full-finetune/Learning Rate": 1.7286785786116707e-05, "Full-finetune/Loss": 0.8593816757202148, "Full-finetune/Loss (Raw)": 0.8982067704200745, "Full-finetune/Step": 1366, "Full-finetune/Step Time": 6.633127016946673} {"Full-finetune/Learning Rate": 1.7282517987190567e-05, "Full-finetune/Loss": 0.8586974143981934, "Full-finetune/Loss (Raw)": 0.8159669637680054, "Full-finetune/Step": 1367, "Full-finetune/Step Time": 6.652221692726016} {"Full-finetune/Learning Rate": 1.7278247362061372e-05, "Full-finetune/Loss": 0.8586894273757935, "Full-finetune/Loss (Raw)": 0.8744969964027405, "Full-finetune/Step": 1368, "Full-finetune/Step Time": 6.6523496843874454} {"Full-finetune/Learning Rate": 1.7273973912386473e-05, "Full-finetune/Loss": 0.8601197004318237, "Full-finetune/Loss (Raw)": 0.9836512207984924, "Full-finetune/Step": 1369, "Full-finetune/Step Time": 6.66155363433063} {"Full-finetune/Learning Rate": 1.726969763982431e-05, "Full-finetune/Loss": 0.8589637279510498, "Full-finetune/Loss (Raw)": 0.7544257044792175, "Full-finetune/Step": 1370, "Full-finetune/Step Time": 6.662854660302401} {"Full-finetune/Learning Rate": 1.7265418546034424e-05, "Full-finetune/Loss": 0.8587117791175842, "Full-finetune/Loss (Raw)": 0.8979617953300476, "Full-finetune/Step": 1371, "Full-finetune/Step Time": 6.663562191650271} {"Full-finetune/Learning Rate": 1.7261136632677445e-05, "Full-finetune/Loss": 0.8550355434417725, "Full-finetune/Loss (Raw)": 0.391949862241745, "Full-finetune/Step": 1372, "Full-finetune/Step Time": 6.678598614409566} {"Full-finetune/Learning Rate": 1.7256851901415102e-05, "Full-finetune/Loss": 0.8566745519638062, "Full-finetune/Loss (Raw)": 0.9799543023109436, "Full-finetune/Step": 1373, "Full-finetune/Step Time": 6.6799891367554665} {"Full-finetune/Learning Rate": 1.725256435391021e-05, "Full-finetune/Loss": 0.8559452295303345, "Full-finetune/Loss (Raw)": 0.7356583476066589, "Full-finetune/Step": 1374, "Full-finetune/Step Time": 6.671337043866515} {"Full-finetune/Learning Rate": 1.7248273991826692e-05, "Full-finetune/Loss": 0.8556886911392212, "Full-finetune/Loss (Raw)": 0.954820454120636, "Full-finetune/Step": 1375, "Full-finetune/Step Time": 6.680504949763417} {"Full-finetune/Learning Rate": 1.7243980816829543e-05, "Full-finetune/Loss": 0.8526185750961304, "Full-finetune/Loss (Raw)": 0.32638677954673767, "Full-finetune/Step": 1376, "Full-finetune/Step Time": 6.68970543704927} {"Full-finetune/Learning Rate": 1.723968483058487e-05, "Full-finetune/Loss": 0.8525999784469604, "Full-finetune/Loss (Raw)": 0.8259256482124329, "Full-finetune/Step": 1377, "Full-finetune/Step Time": 6.689697153866291} {"Full-finetune/Learning Rate": 1.7235386034759853e-05, "Full-finetune/Loss": 0.8528009057044983, "Full-finetune/Loss (Raw)": 0.9465134143829346, "Full-finetune/Step": 1378, "Full-finetune/Step Time": 6.708094859495759} {"Full-finetune/Learning Rate": 1.723108443102278e-05, "Full-finetune/Loss": 0.8503302335739136, "Full-finetune/Loss (Raw)": 0.60587477684021, "Full-finetune/Step": 1379, "Full-finetune/Step Time": 6.723160237073898} {"Full-finetune/Learning Rate": 1.722678002104301e-05, "Full-finetune/Loss": 0.8508661985397339, "Full-finetune/Loss (Raw)": 0.8568346500396729, "Full-finetune/Step": 1380, "Full-finetune/Step Time": 6.722498554736376} {"Full-finetune/Learning Rate": 1.722247280649101e-05, "Full-finetune/Loss": 0.8520006537437439, "Full-finetune/Loss (Raw)": 0.932058572769165, "Full-finetune/Step": 1381, "Full-finetune/Step Time": 6.708150874823332} {"Full-finetune/Learning Rate": 1.7218162789038312e-05, "Full-finetune/Loss": 0.8520415425300598, "Full-finetune/Loss (Raw)": 0.8769847750663757, "Full-finetune/Step": 1382, "Full-finetune/Step Time": 6.707054430618882} {"Full-finetune/Learning Rate": 1.7213849970357565e-05, "Full-finetune/Loss": 0.8502542972564697, "Full-finetune/Loss (Raw)": 0.7519626021385193, "Full-finetune/Step": 1383, "Full-finetune/Step Time": 6.712083304300904} {"Full-finetune/Learning Rate": 1.7209534352122484e-05, "Full-finetune/Loss": 0.8498498201370239, "Full-finetune/Loss (Raw)": 0.8591333031654358, "Full-finetune/Step": 1384, "Full-finetune/Step Time": 6.7131766732782125} {"Full-finetune/Learning Rate": 1.720521593600787e-05, "Full-finetune/Loss": 0.852735698223114, "Full-finetune/Loss (Raw)": 0.9523270726203918, "Full-finetune/Step": 1385, "Full-finetune/Step Time": 6.6900032088160515} {"Full-finetune/Learning Rate": 1.7200894723689626e-05, "Full-finetune/Loss": 0.8525473475456238, "Full-finetune/Loss (Raw)": 0.8557367324829102, "Full-finetune/Step": 1386, "Full-finetune/Step Time": 6.6906165312975645} {"Full-finetune/Learning Rate": 1.7196570716844724e-05, "Full-finetune/Loss": 0.8524134755134583, "Full-finetune/Loss (Raw)": 0.7433018088340759, "Full-finetune/Step": 1387, "Full-finetune/Step Time": 6.68543235398829} {"Full-finetune/Learning Rate": 1.7192243917151228e-05, "Full-finetune/Loss": 0.851220965385437, "Full-finetune/Loss (Raw)": 0.940517008304596, "Full-finetune/Step": 1388, "Full-finetune/Step Time": 6.669788330793381} {"Full-finetune/Learning Rate": 1.7187914326288286e-05, "Full-finetune/Loss": 0.8523002862930298, "Full-finetune/Loss (Raw)": 0.8886541724205017, "Full-finetune/Step": 1389, "Full-finetune/Step Time": 6.671091435477138} {"Full-finetune/Learning Rate": 1.718358194593612e-05, "Full-finetune/Loss": 0.8533384799957275, "Full-finetune/Loss (Raw)": 1.0474722385406494, "Full-finetune/Step": 1390, "Full-finetune/Step Time": 6.6798708364367485} {"Full-finetune/Learning Rate": 1.7179246777776053e-05, "Full-finetune/Loss": 0.8524624705314636, "Full-finetune/Loss (Raw)": 0.824622392654419, "Full-finetune/Step": 1391, "Full-finetune/Step Time": 6.695663565769792} {"Full-finetune/Learning Rate": 1.7174908823490475e-05, "Full-finetune/Loss": 0.8519526720046997, "Full-finetune/Loss (Raw)": 0.8093162775039673, "Full-finetune/Step": 1392, "Full-finetune/Step Time": 6.727265806868672} {"Full-finetune/Learning Rate": 1.717056808476286e-05, "Full-finetune/Loss": 0.8549310564994812, "Full-finetune/Loss (Raw)": 0.7748775482177734, "Full-finetune/Step": 1393, "Full-finetune/Step Time": 6.746606655418873} {"Full-finetune/Learning Rate": 1.716622456327776e-05, "Full-finetune/Loss": 0.8558240532875061, "Full-finetune/Loss (Raw)": 0.9421480894088745, "Full-finetune/Step": 1394, "Full-finetune/Step Time": 6.721229979768395} {"Full-finetune/Learning Rate": 1.7161878260720826e-05, "Full-finetune/Loss": 0.8549488186836243, "Full-finetune/Loss (Raw)": 0.830821692943573, "Full-finetune/Step": 1395, "Full-finetune/Step Time": 6.723448997363448} {"Full-finetune/Learning Rate": 1.7157529178778757e-05, "Full-finetune/Loss": 0.8519785404205322, "Full-finetune/Loss (Raw)": 0.6142195463180542, "Full-finetune/Step": 1396, "Full-finetune/Step Time": 6.730912961065769} {"Full-finetune/Learning Rate": 1.7153177319139352e-05, "Full-finetune/Loss": 0.852439284324646, "Full-finetune/Loss (Raw)": 0.8717873096466064, "Full-finetune/Step": 1397, "Full-finetune/Step Time": 6.754715422168374} {"Full-finetune/Learning Rate": 1.7148822683491483e-05, "Full-finetune/Loss": 0.8521000146865845, "Full-finetune/Loss (Raw)": 0.8707332015037537, "Full-finetune/Step": 1398, "Full-finetune/Step Time": 6.744888352230191} {"Full-finetune/Learning Rate": 1.7144465273525105e-05, "Full-finetune/Loss": 0.8490474820137024, "Full-finetune/Loss (Raw)": 0.6107686161994934, "Full-finetune/Step": 1399, "Full-finetune/Step Time": 6.751206401735544} {"Full-finetune/Learning Rate": 1.7140105090931232e-05, "Full-finetune/Loss": 0.8507933616638184, "Full-finetune/Loss (Raw)": 0.9433404207229614, "Full-finetune/Step": 1400, "Full-finetune/Step Time": 6.746903048828244} {"Full-finetune/Learning Rate": 1.7135742137401976e-05, "Full-finetune/Loss": 0.8512657284736633, "Full-finetune/Loss (Raw)": 0.8190720677375793, "Full-finetune/Step": 1401, "Full-finetune/Step Time": 6.744719855487347} {"Full-finetune/Learning Rate": 1.7131376414630506e-05, "Full-finetune/Loss": 0.8479940891265869, "Full-finetune/Loss (Raw)": 0.39510056376457214, "Full-finetune/Step": 1402, "Full-finetune/Step Time": 6.764249177649617} {"Full-finetune/Learning Rate": 1.712700792431108e-05, "Full-finetune/Loss": 0.8465129733085632, "Full-finetune/Loss (Raw)": 0.8377415537834167, "Full-finetune/Step": 1403, "Full-finetune/Step Time": 6.7653588112443686} {"Full-finetune/Learning Rate": 1.7122636668139022e-05, "Full-finetune/Loss": 0.8468546867370605, "Full-finetune/Loss (Raw)": 0.888949453830719, "Full-finetune/Step": 1404, "Full-finetune/Step Time": 6.761978922411799} {"Full-finetune/Learning Rate": 1.711826264781073e-05, "Full-finetune/Loss": 0.8453354835510254, "Full-finetune/Loss (Raw)": 0.6992055177688599, "Full-finetune/Step": 1405, "Full-finetune/Step Time": 6.7602002546191216} {"Full-finetune/Learning Rate": 1.711388586502368e-05, "Full-finetune/Loss": 0.8473126888275146, "Full-finetune/Loss (Raw)": 0.8462049961090088, "Full-finetune/Step": 1406, "Full-finetune/Step Time": 6.753778211772442} {"Full-finetune/Learning Rate": 1.710950632147641e-05, "Full-finetune/Loss": 0.8434076309204102, "Full-finetune/Loss (Raw)": 0.38026776909828186, "Full-finetune/Step": 1407, "Full-finetune/Step Time": 6.766615556553006} {"Full-finetune/Learning Rate": 1.710512401886854e-05, "Full-finetune/Loss": 0.841294527053833, "Full-finetune/Loss (Raw)": 0.7345421314239502, "Full-finetune/Step": 1408, "Full-finetune/Step Time": 6.7706585098057985} {"Full-finetune/Learning Rate": 1.7100738958900752e-05, "Full-finetune/Loss": 0.8413970470428467, "Full-finetune/Loss (Raw)": 0.9243031740188599, "Full-finetune/Step": 1409, "Full-finetune/Step Time": 6.7687899842858315} {"Full-finetune/Learning Rate": 1.7096351143274804e-05, "Full-finetune/Loss": 0.8405333757400513, "Full-finetune/Loss (Raw)": 0.7628319263458252, "Full-finetune/Step": 1410, "Full-finetune/Step Time": 6.796666543930769} {"Full-finetune/Learning Rate": 1.7091960573693523e-05, "Full-finetune/Loss": 0.8419880867004395, "Full-finetune/Loss (Raw)": 0.8715724349021912, "Full-finetune/Step": 1411, "Full-finetune/Step Time": 6.7855886202305555} {"Full-finetune/Learning Rate": 1.70875672518608e-05, "Full-finetune/Loss": 0.8405964374542236, "Full-finetune/Loss (Raw)": 0.6718734502792358, "Full-finetune/Step": 1412, "Full-finetune/Step Time": 6.784793592989445} {"Full-finetune/Learning Rate": 1.7083171179481603e-05, "Full-finetune/Loss": 0.8401895761489868, "Full-finetune/Loss (Raw)": 0.916528582572937, "Full-finetune/Step": 1413, "Full-finetune/Step Time": 6.788396371528506} {"Full-finetune/Learning Rate": 1.7078772358261957e-05, "Full-finetune/Loss": 0.8415772318840027, "Full-finetune/Loss (Raw)": 1.0220885276794434, "Full-finetune/Step": 1414, "Full-finetune/Step Time": 6.777937991544604} {"Full-finetune/Learning Rate": 1.707437078990896e-05, "Full-finetune/Loss": 0.8414231538772583, "Full-finetune/Loss (Raw)": 0.8491496443748474, "Full-finetune/Step": 1415, "Full-finetune/Step Time": 6.781442563980818} {"Full-finetune/Learning Rate": 1.7069966476130773e-05, "Full-finetune/Loss": 0.8424656391143799, "Full-finetune/Loss (Raw)": 0.9689974784851074, "Full-finetune/Step": 1416, "Full-finetune/Step Time": 6.77950114198029} {"Full-finetune/Learning Rate": 1.706555941863663e-05, "Full-finetune/Loss": 0.8425916433334351, "Full-finetune/Loss (Raw)": 0.8428713083267212, "Full-finetune/Step": 1417, "Full-finetune/Step Time": 6.788850395008922} {"Full-finetune/Learning Rate": 1.7061149619136818e-05, "Full-finetune/Loss": 0.8403363227844238, "Full-finetune/Loss (Raw)": 0.7204886078834534, "Full-finetune/Step": 1418, "Full-finetune/Step Time": 6.795631369575858} {"Full-finetune/Learning Rate": 1.7056737079342694e-05, "Full-finetune/Loss": 0.8399441242218018, "Full-finetune/Loss (Raw)": 0.7283852696418762, "Full-finetune/Step": 1419, "Full-finetune/Step Time": 6.779248919337988} {"Full-finetune/Learning Rate": 1.7052321800966684e-05, "Full-finetune/Loss": 0.8406345248222351, "Full-finetune/Loss (Raw)": 0.9142399430274963, "Full-finetune/Step": 1420, "Full-finetune/Step Time": 6.789309734478593} {"Full-finetune/Learning Rate": 1.7047903785722268e-05, "Full-finetune/Loss": 0.8400479555130005, "Full-finetune/Loss (Raw)": 0.8219432234764099, "Full-finetune/Step": 1421, "Full-finetune/Step Time": 6.788613004609942} {"Full-finetune/Learning Rate": 1.704348303532399e-05, "Full-finetune/Loss": 0.8387680053710938, "Full-finetune/Loss (Raw)": 0.7757604718208313, "Full-finetune/Step": 1422, "Full-finetune/Step Time": 6.783939568325877} {"Full-finetune/Learning Rate": 1.7039059551487457e-05, "Full-finetune/Loss": 0.8397545218467712, "Full-finetune/Loss (Raw)": 0.9480419754981995, "Full-finetune/Step": 1423, "Full-finetune/Step Time": 6.773942271247506} {"Full-finetune/Learning Rate": 1.703463333592934e-05, "Full-finetune/Loss": 0.8396544456481934, "Full-finetune/Loss (Raw)": 0.7859583497047424, "Full-finetune/Step": 1424, "Full-finetune/Step Time": 6.76697893999517} {"Full-finetune/Learning Rate": 1.7030204390367368e-05, "Full-finetune/Loss": 0.839832067489624, "Full-finetune/Loss (Raw)": 0.8407446146011353, "Full-finetune/Step": 1425, "Full-finetune/Step Time": 6.760678477585316} {"Full-finetune/Learning Rate": 1.7025772716520324e-05, "Full-finetune/Loss": 0.8401103019714355, "Full-finetune/Loss (Raw)": 0.7856655716896057, "Full-finetune/Step": 1426, "Full-finetune/Step Time": 6.773373626172543} {"Full-finetune/Learning Rate": 1.702133831610805e-05, "Full-finetune/Loss": 0.8407180309295654, "Full-finetune/Loss (Raw)": 0.7426857948303223, "Full-finetune/Step": 1427, "Full-finetune/Step Time": 6.77203444391489} {"Full-finetune/Learning Rate": 1.701690119085146e-05, "Full-finetune/Loss": 0.8403184413909912, "Full-finetune/Loss (Raw)": 0.7581593990325928, "Full-finetune/Step": 1428, "Full-finetune/Step Time": 6.779091138392687} {"Full-finetune/Learning Rate": 1.701246134247251e-05, "Full-finetune/Loss": 0.8393821716308594, "Full-finetune/Loss (Raw)": 0.72141033411026, "Full-finetune/Step": 1429, "Full-finetune/Step Time": 6.769316149875522} {"Full-finetune/Learning Rate": 1.700801877269422e-05, "Full-finetune/Loss": 0.8394581079483032, "Full-finetune/Loss (Raw)": 0.8336971998214722, "Full-finetune/Step": 1430, "Full-finetune/Step Time": 6.7607788648456335} {"Full-finetune/Learning Rate": 1.7003573483240666e-05, "Full-finetune/Loss": 0.8388639688491821, "Full-finetune/Loss (Raw)": 0.759007453918457, "Full-finetune/Step": 1431, "Full-finetune/Step Time": 6.771494675427675} {"Full-finetune/Learning Rate": 1.6999125475836973e-05, "Full-finetune/Loss": 0.8376983404159546, "Full-finetune/Loss (Raw)": 0.7811995148658752, "Full-finetune/Step": 1432, "Full-finetune/Step Time": 6.7746544517576694} {"Full-finetune/Learning Rate": 1.6994674752209334e-05, "Full-finetune/Loss": 0.8386988639831543, "Full-finetune/Loss (Raw)": 0.9607856273651123, "Full-finetune/Step": 1433, "Full-finetune/Step Time": 6.785534922033548} {"Full-finetune/Learning Rate": 1.6990221314084976e-05, "Full-finetune/Loss": 0.835206925868988, "Full-finetune/Loss (Raw)": 0.3786492645740509, "Full-finetune/Step": 1434, "Full-finetune/Step Time": 6.7876840848475695} {"Full-finetune/Learning Rate": 1.6985765163192204e-05, "Full-finetune/Loss": 0.8345619440078735, "Full-finetune/Loss (Raw)": 0.9111759066581726, "Full-finetune/Step": 1435, "Full-finetune/Step Time": 6.786262273788452} {"Full-finetune/Learning Rate": 1.698130630126036e-05, "Full-finetune/Loss": 0.8313491344451904, "Full-finetune/Loss (Raw)": 0.6055846810340881, "Full-finetune/Step": 1436, "Full-finetune/Step Time": 6.7960823494941} {"Full-finetune/Learning Rate": 1.697684473001983e-05, "Full-finetune/Loss": 0.8316994309425354, "Full-finetune/Loss (Raw)": 0.8976248502731323, "Full-finetune/Step": 1437, "Full-finetune/Step Time": 6.7804602067917585} {"Full-finetune/Learning Rate": 1.6972380451202083e-05, "Full-finetune/Loss": 0.8322535157203674, "Full-finetune/Loss (Raw)": 0.9023259282112122, "Full-finetune/Step": 1438, "Full-finetune/Step Time": 6.775767847895622} {"Full-finetune/Learning Rate": 1.69679134665396e-05, "Full-finetune/Loss": 0.8326253890991211, "Full-finetune/Loss (Raw)": 0.9505869150161743, "Full-finetune/Step": 1439, "Full-finetune/Step Time": 6.762165121734142} {"Full-finetune/Learning Rate": 1.696344377776594e-05, "Full-finetune/Loss": 0.8328889608383179, "Full-finetune/Loss (Raw)": 0.6238462924957275, "Full-finetune/Step": 1440, "Full-finetune/Step Time": 6.767244076356292} {"Full-finetune/Learning Rate": 1.69589713866157e-05, "Full-finetune/Loss": 0.8328176140785217, "Full-finetune/Loss (Raw)": 0.8692421317100525, "Full-finetune/Step": 1441, "Full-finetune/Step Time": 6.764039324596524} {"Full-finetune/Learning Rate": 1.6954496294824526e-05, "Full-finetune/Loss": 0.8318967819213867, "Full-finetune/Loss (Raw)": 0.8291444182395935, "Full-finetune/Step": 1442, "Full-finetune/Step Time": 6.769420202821493} {"Full-finetune/Learning Rate": 1.6950018504129112e-05, "Full-finetune/Loss": 0.8311822414398193, "Full-finetune/Loss (Raw)": 0.8431093692779541, "Full-finetune/Step": 1443, "Full-finetune/Step Time": 6.771991541609168} {"Full-finetune/Learning Rate": 1.6945538016267206e-05, "Full-finetune/Loss": 0.8290609121322632, "Full-finetune/Loss (Raw)": 0.7867897748947144, "Full-finetune/Step": 1444, "Full-finetune/Step Time": 6.770158030092716} {"Full-finetune/Learning Rate": 1.6941054832977598e-05, "Full-finetune/Loss": 0.82826828956604, "Full-finetune/Loss (Raw)": 0.8579078316688538, "Full-finetune/Step": 1445, "Full-finetune/Step Time": 6.772257726639509} {"Full-finetune/Learning Rate": 1.6936568956000123e-05, "Full-finetune/Loss": 0.8271976709365845, "Full-finetune/Loss (Raw)": 0.7192971706390381, "Full-finetune/Step": 1446, "Full-finetune/Step Time": 6.787114104256034} {"Full-finetune/Learning Rate": 1.693208038707566e-05, "Full-finetune/Loss": 0.8273932337760925, "Full-finetune/Loss (Raw)": 0.9105185270309448, "Full-finetune/Step": 1447, "Full-finetune/Step Time": 6.7918768636882305} {"Full-finetune/Learning Rate": 1.6927589127946136e-05, "Full-finetune/Loss": 0.8285608291625977, "Full-finetune/Loss (Raw)": 0.9163910150527954, "Full-finetune/Step": 1448, "Full-finetune/Step Time": 6.771929165348411} {"Full-finetune/Learning Rate": 1.692309518035452e-05, "Full-finetune/Loss": 0.8273192644119263, "Full-finetune/Loss (Raw)": 0.7435147166252136, "Full-finetune/Step": 1449, "Full-finetune/Step Time": 6.775132084265351} {"Full-finetune/Learning Rate": 1.691859854604483e-05, "Full-finetune/Loss": 0.8259894251823425, "Full-finetune/Loss (Raw)": 0.7181859612464905, "Full-finetune/Step": 1450, "Full-finetune/Step Time": 6.809885388240218} {"Full-finetune/Learning Rate": 1.6914099226762116e-05, "Full-finetune/Loss": 0.8253230452537537, "Full-finetune/Loss (Raw)": 0.8442619442939758, "Full-finetune/Step": 1451, "Full-finetune/Step Time": 6.812814857810736} {"Full-finetune/Learning Rate": 1.6909597224252478e-05, "Full-finetune/Loss": 0.8250601291656494, "Full-finetune/Loss (Raw)": 0.9309918284416199, "Full-finetune/Step": 1452, "Full-finetune/Step Time": 6.8043675273656845} {"Full-finetune/Learning Rate": 1.6905092540263055e-05, "Full-finetune/Loss": 0.8235543966293335, "Full-finetune/Loss (Raw)": 0.7886095643043518, "Full-finetune/Step": 1453, "Full-finetune/Step Time": 6.799530183896422} {"Full-finetune/Learning Rate": 1.690058517654203e-05, "Full-finetune/Loss": 0.8243187069892883, "Full-finetune/Loss (Raw)": 1.0579462051391602, "Full-finetune/Step": 1454, "Full-finetune/Step Time": 6.802391609176993} {"Full-finetune/Learning Rate": 1.689607513483862e-05, "Full-finetune/Loss": 0.8258326053619385, "Full-finetune/Loss (Raw)": 0.9380426406860352, "Full-finetune/Step": 1455, "Full-finetune/Step Time": 6.770067688077688} {"Full-finetune/Learning Rate": 1.6891562416903083e-05, "Full-finetune/Loss": 0.8239091634750366, "Full-finetune/Loss (Raw)": 0.8224435448646545, "Full-finetune/Step": 1456, "Full-finetune/Step Time": 6.771917782723904} {"Full-finetune/Learning Rate": 1.688704702448672e-05, "Full-finetune/Loss": 0.818355143070221, "Full-finetune/Loss (Raw)": 0.3772351145744324, "Full-finetune/Step": 1457, "Full-finetune/Step Time": 6.787416459992528} {"Full-finetune/Learning Rate": 1.6882528959341865e-05, "Full-finetune/Loss": 0.819614589214325, "Full-finetune/Loss (Raw)": 0.8649510145187378, "Full-finetune/Step": 1458, "Full-finetune/Step Time": 6.7867976147681475} {"Full-finetune/Learning Rate": 1.6878008223221895e-05, "Full-finetune/Loss": 0.8199407458305359, "Full-finetune/Loss (Raw)": 0.97390216588974, "Full-finetune/Step": 1459, "Full-finetune/Step Time": 6.794954273849726} {"Full-finetune/Learning Rate": 1.687348481788121e-05, "Full-finetune/Loss": 0.8218817114830017, "Full-finetune/Loss (Raw)": 1.0084491968154907, "Full-finetune/Step": 1460, "Full-finetune/Step Time": 6.785054253414273} {"Full-finetune/Learning Rate": 1.6868958745075262e-05, "Full-finetune/Loss": 0.822946310043335, "Full-finetune/Loss (Raw)": 1.0512408018112183, "Full-finetune/Step": 1461, "Full-finetune/Step Time": 6.778563588857651} {"Full-finetune/Learning Rate": 1.6864430006560527e-05, "Full-finetune/Loss": 0.8226858377456665, "Full-finetune/Loss (Raw)": 0.8995434641838074, "Full-finetune/Step": 1462, "Full-finetune/Step Time": 6.780956458300352} {"Full-finetune/Learning Rate": 1.685989860409453e-05, "Full-finetune/Loss": 0.823036789894104, "Full-finetune/Loss (Raw)": 0.9634246826171875, "Full-finetune/Step": 1463, "Full-finetune/Step Time": 6.782602168619633} {"Full-finetune/Learning Rate": 1.6855364539435805e-05, "Full-finetune/Loss": 0.8222006559371948, "Full-finetune/Loss (Raw)": 0.8111688494682312, "Full-finetune/Step": 1464, "Full-finetune/Step Time": 6.786290613934398} {"Full-finetune/Learning Rate": 1.685082781434395e-05, "Full-finetune/Loss": 0.8217041492462158, "Full-finetune/Loss (Raw)": 0.8623808026313782, "Full-finetune/Step": 1465, "Full-finetune/Step Time": 6.779843417927623} {"Full-finetune/Learning Rate": 1.6846288430579565e-05, "Full-finetune/Loss": 0.8258578181266785, "Full-finetune/Loss (Raw)": 0.8683620095252991, "Full-finetune/Step": 1466, "Full-finetune/Step Time": 6.76637271605432} {"Full-finetune/Learning Rate": 1.6841746389904306e-05, "Full-finetune/Loss": 0.8265435099601746, "Full-finetune/Loss (Raw)": 0.848935604095459, "Full-finetune/Step": 1467, "Full-finetune/Step Time": 6.782651465386152} {"Full-finetune/Learning Rate": 1.683720169408085e-05, "Full-finetune/Loss": 0.8270039558410645, "Full-finetune/Loss (Raw)": 0.9042763113975525, "Full-finetune/Step": 1468, "Full-finetune/Step Time": 6.775372629985213} {"Full-finetune/Learning Rate": 1.6832654344872893e-05, "Full-finetune/Loss": 0.8235421776771545, "Full-finetune/Loss (Raw)": 0.3894784450531006, "Full-finetune/Step": 1469, "Full-finetune/Step Time": 6.811668949201703} {"Full-finetune/Learning Rate": 1.6828104344045186e-05, "Full-finetune/Loss": 0.8225898146629333, "Full-finetune/Loss (Raw)": 0.8057613372802734, "Full-finetune/Step": 1470, "Full-finetune/Step Time": 6.801408641040325} {"Full-finetune/Learning Rate": 1.682355169336349e-05, "Full-finetune/Loss": 0.8229010105133057, "Full-finetune/Loss (Raw)": 0.8540966510772705, "Full-finetune/Step": 1471, "Full-finetune/Step Time": 6.775692939758301} {"Full-finetune/Learning Rate": 1.6818996394594603e-05, "Full-finetune/Loss": 0.82281494140625, "Full-finetune/Loss (Raw)": 0.7859780788421631, "Full-finetune/Step": 1472, "Full-finetune/Step Time": 6.768879346549511} {"Full-finetune/Learning Rate": 1.681443844950634e-05, "Full-finetune/Loss": 0.8196033239364624, "Full-finetune/Loss (Raw)": 0.5918769240379333, "Full-finetune/Step": 1473, "Full-finetune/Step Time": 6.774972232058644} {"Full-finetune/Learning Rate": 1.6809877859867557e-05, "Full-finetune/Loss": 0.8207082748413086, "Full-finetune/Loss (Raw)": 1.0288584232330322, "Full-finetune/Step": 1474, "Full-finetune/Step Time": 6.76967834122479} {"Full-finetune/Learning Rate": 1.680531462744813e-05, "Full-finetune/Loss": 0.8186333179473877, "Full-finetune/Loss (Raw)": 0.6876440644264221, "Full-finetune/Step": 1475, "Full-finetune/Step Time": 6.761941296979785} {"Full-finetune/Learning Rate": 1.680074875401896e-05, "Full-finetune/Loss": 0.8224668502807617, "Full-finetune/Loss (Raw)": 0.9323232769966125, "Full-finetune/Step": 1476, "Full-finetune/Step Time": 6.747080659493804} {"Full-finetune/Learning Rate": 1.679618024135197e-05, "Full-finetune/Loss": 0.8174705505371094, "Full-finetune/Loss (Raw)": 0.3190113306045532, "Full-finetune/Step": 1477, "Full-finetune/Step Time": 6.7676538322120905} {"Full-finetune/Learning Rate": 1.679160909122011e-05, "Full-finetune/Loss": 0.8183833360671997, "Full-finetune/Loss (Raw)": 1.0218347311019897, "Full-finetune/Step": 1478, "Full-finetune/Step Time": 6.76180393435061} {"Full-finetune/Learning Rate": 1.6787035305397363e-05, "Full-finetune/Loss": 0.8171193599700928, "Full-finetune/Loss (Raw)": 0.758949875831604, "Full-finetune/Step": 1479, "Full-finetune/Step Time": 6.7753168772906065} {"Full-finetune/Learning Rate": 1.6782458885658716e-05, "Full-finetune/Loss": 0.8206404447555542, "Full-finetune/Loss (Raw)": 0.8843433856964111, "Full-finetune/Step": 1480, "Full-finetune/Step Time": 6.767650857567787} {"Full-finetune/Learning Rate": 1.6777879833780192e-05, "Full-finetune/Loss": 0.8210538625717163, "Full-finetune/Loss (Raw)": 0.909783661365509, "Full-finetune/Step": 1481, "Full-finetune/Step Time": 6.772255547344685} {"Full-finetune/Learning Rate": 1.677329815153883e-05, "Full-finetune/Loss": 0.8205655813217163, "Full-finetune/Loss (Raw)": 0.8202469944953918, "Full-finetune/Step": 1482, "Full-finetune/Step Time": 6.764022687450051} {"Full-finetune/Learning Rate": 1.6768713840712692e-05, "Full-finetune/Loss": 0.817416787147522, "Full-finetune/Loss (Raw)": 0.6107868552207947, "Full-finetune/Step": 1483, "Full-finetune/Step Time": 6.78837070427835} {"Full-finetune/Learning Rate": 1.6764126903080865e-05, "Full-finetune/Loss": 0.8166338801383972, "Full-finetune/Loss (Raw)": 0.905659556388855, "Full-finetune/Step": 1484, "Full-finetune/Step Time": 6.784713426604867} {"Full-finetune/Learning Rate": 1.675953734042344e-05, "Full-finetune/Loss": 0.8160011172294617, "Full-finetune/Loss (Raw)": 0.8120856881141663, "Full-finetune/Step": 1485, "Full-finetune/Step Time": 6.773185953497887} {"Full-finetune/Learning Rate": 1.6754945154521548e-05, "Full-finetune/Loss": 0.8113111257553101, "Full-finetune/Loss (Raw)": 0.3132706582546234, "Full-finetune/Step": 1486, "Full-finetune/Step Time": 6.79020113684237} {"Full-finetune/Learning Rate": 1.6750350347157315e-05, "Full-finetune/Loss": 0.8131953477859497, "Full-finetune/Loss (Raw)": 1.0266982316970825, "Full-finetune/Step": 1487, "Full-finetune/Step Time": 6.7858486864715815} {"Full-finetune/Learning Rate": 1.6745752920113907e-05, "Full-finetune/Loss": 0.8145189881324768, "Full-finetune/Loss (Raw)": 1.1155695915222168, "Full-finetune/Step": 1488, "Full-finetune/Step Time": 6.794774979352951} {"Full-finetune/Learning Rate": 1.6741152875175485e-05, "Full-finetune/Loss": 0.8118753433227539, "Full-finetune/Loss (Raw)": 0.6623662114143372, "Full-finetune/Step": 1489, "Full-finetune/Step Time": 6.796856492757797} {"Full-finetune/Learning Rate": 1.6736550214127247e-05, "Full-finetune/Loss": 0.8111858367919922, "Full-finetune/Loss (Raw)": 0.7882495522499084, "Full-finetune/Step": 1490, "Full-finetune/Step Time": 6.816696893423796} {"Full-finetune/Learning Rate": 1.673194493875539e-05, "Full-finetune/Loss": 0.8101434707641602, "Full-finetune/Loss (Raw)": 0.7191764116287231, "Full-finetune/Step": 1491, "Full-finetune/Step Time": 6.829417819157243} {"Full-finetune/Learning Rate": 1.672733705084713e-05, "Full-finetune/Loss": 0.8113888502120972, "Full-finetune/Loss (Raw)": 0.9511557221412659, "Full-finetune/Step": 1492, "Full-finetune/Step Time": 6.791700726374984} {"Full-finetune/Learning Rate": 1.67227265521907e-05, "Full-finetune/Loss": 0.8079121112823486, "Full-finetune/Loss (Raw)": 0.3386031687259674, "Full-finetune/Step": 1493, "Full-finetune/Step Time": 6.804855173453689} {"Full-finetune/Learning Rate": 1.671811344457535e-05, "Full-finetune/Loss": 0.8074058890342712, "Full-finetune/Loss (Raw)": 0.8334038257598877, "Full-finetune/Step": 1494, "Full-finetune/Step Time": 6.810368532314897} {"Full-finetune/Learning Rate": 1.671349772979133e-05, "Full-finetune/Loss": 0.8076549768447876, "Full-finetune/Loss (Raw)": 0.8478531837463379, "Full-finetune/Step": 1495, "Full-finetune/Step Time": 6.795815547928214} {"Full-finetune/Learning Rate": 1.6708879409629908e-05, "Full-finetune/Loss": 0.8080027103424072, "Full-finetune/Loss (Raw)": 0.9190087914466858, "Full-finetune/Step": 1496, "Full-finetune/Step Time": 6.796108407899737} {"Full-finetune/Learning Rate": 1.670425848588337e-05, "Full-finetune/Loss": 0.8063127994537354, "Full-finetune/Loss (Raw)": 0.7673404216766357, "Full-finetune/Step": 1497, "Full-finetune/Step Time": 6.79036465100944} {"Full-finetune/Learning Rate": 1.6699634960345005e-05, "Full-finetune/Loss": 0.8081135749816895, "Full-finetune/Loss (Raw)": 0.9849192500114441, "Full-finetune/Step": 1498, "Full-finetune/Step Time": 6.793448459357023} {"Full-finetune/Learning Rate": 1.6695008834809107e-05, "Full-finetune/Loss": 0.8045520186424255, "Full-finetune/Loss (Raw)": 0.44208571314811707, "Full-finetune/Step": 1499, "Full-finetune/Step Time": 6.804249485954642} {"Full-finetune/Learning Rate": 1.669038011107099e-05, "Full-finetune/Loss": 0.8043298721313477, "Full-finetune/Loss (Raw)": 0.3635156750679016, "Full-finetune/Step": 1500, "Full-finetune/Step Time": 6.804632538929582} {"Full-finetune/Learning Rate": 1.6685748790926975e-05, "Full-finetune/Loss": 0.8044619560241699, "Full-finetune/Loss (Raw)": 0.996866762638092, "Full-finetune/Step": 1501, "Full-finetune/Step Time": 6.83627556823194} {"Full-finetune/Learning Rate": 1.6681114876174377e-05, "Full-finetune/Loss": 0.806232750415802, "Full-finetune/Loss (Raw)": 0.9623157382011414, "Full-finetune/Step": 1502, "Full-finetune/Step Time": 6.829586572945118} {"Full-finetune/Learning Rate": 1.6676478368611536e-05, "Full-finetune/Loss": 0.8044339418411255, "Full-finetune/Loss (Raw)": 0.7245693802833557, "Full-finetune/Step": 1503, "Full-finetune/Step Time": 6.819762296974659} {"Full-finetune/Learning Rate": 1.667183927003779e-05, "Full-finetune/Loss": 0.8087678551673889, "Full-finetune/Loss (Raw)": 0.8811330199241638, "Full-finetune/Step": 1504, "Full-finetune/Step Time": 6.814338132739067} {"Full-finetune/Learning Rate": 1.6667197582253474e-05, "Full-finetune/Loss": 0.8090774416923523, "Full-finetune/Loss (Raw)": 0.8655507564544678, "Full-finetune/Step": 1505, "Full-finetune/Step Time": 6.798499645665288} {"Full-finetune/Learning Rate": 1.6662553307059947e-05, "Full-finetune/Loss": 0.8078298568725586, "Full-finetune/Loss (Raw)": 0.7868257164955139, "Full-finetune/Step": 1506, "Full-finetune/Step Time": 6.788076447322965} {"Full-finetune/Learning Rate": 1.6657906446259554e-05, "Full-finetune/Loss": 0.8105272054672241, "Full-finetune/Loss (Raw)": 0.9511379599571228, "Full-finetune/Step": 1507, "Full-finetune/Step Time": 6.76723557151854} {"Full-finetune/Learning Rate": 1.6653257001655652e-05, "Full-finetune/Loss": 0.8112124800682068, "Full-finetune/Loss (Raw)": 0.9445465803146362, "Full-finetune/Step": 1508, "Full-finetune/Step Time": 6.772039048373699} {"Full-finetune/Learning Rate": 1.664860497505261e-05, "Full-finetune/Loss": 0.8106608390808105, "Full-finetune/Loss (Raw)": 0.8614436388015747, "Full-finetune/Step": 1509, "Full-finetune/Step Time": 6.7660339046269655} {"Full-finetune/Learning Rate": 1.664395036825577e-05, "Full-finetune/Loss": 0.811286449432373, "Full-finetune/Loss (Raw)": 0.957065999507904, "Full-finetune/Step": 1510, "Full-finetune/Step Time": 6.7640769593417645} {"Full-finetune/Learning Rate": 1.6639293183071514e-05, "Full-finetune/Loss": 0.8128992319107056, "Full-finetune/Loss (Raw)": 0.9584020376205444, "Full-finetune/Step": 1511, "Full-finetune/Step Time": 6.763233922421932} {"Full-finetune/Learning Rate": 1.6634633421307187e-05, "Full-finetune/Loss": 0.8120577335357666, "Full-finetune/Loss (Raw)": 0.7514193058013916, "Full-finetune/Step": 1512, "Full-finetune/Step Time": 6.76693132892251} {"Full-finetune/Learning Rate": 1.6629971084771165e-05, "Full-finetune/Loss": 0.8124005794525146, "Full-finetune/Loss (Raw)": 0.99620521068573, "Full-finetune/Step": 1513, "Full-finetune/Step Time": 6.768590630963445} {"Full-finetune/Learning Rate": 1.66253061752728e-05, "Full-finetune/Loss": 0.8119750618934631, "Full-finetune/Loss (Raw)": 0.8012734651565552, "Full-finetune/Step": 1514, "Full-finetune/Step Time": 6.778223054483533} {"Full-finetune/Learning Rate": 1.662063869462246e-05, "Full-finetune/Loss": 0.8129821419715881, "Full-finetune/Loss (Raw)": 0.8722087740898132, "Full-finetune/Step": 1515, "Full-finetune/Step Time": 6.775405302643776} {"Full-finetune/Learning Rate": 1.6615968644631498e-05, "Full-finetune/Loss": 0.8126416206359863, "Full-finetune/Loss (Raw)": 0.8969324827194214, "Full-finetune/Step": 1516, "Full-finetune/Step Time": 6.7787905763834715} {"Full-finetune/Learning Rate": 1.661129602711227e-05, "Full-finetune/Loss": 0.811939001083374, "Full-finetune/Loss (Raw)": 0.7987203001976013, "Full-finetune/Step": 1517, "Full-finetune/Step Time": 6.779296109452844} {"Full-finetune/Learning Rate": 1.6606620843878128e-05, "Full-finetune/Loss": 0.811560869216919, "Full-finetune/Loss (Raw)": 0.9990731477737427, "Full-finetune/Step": 1518, "Full-finetune/Step Time": 6.772508192807436} {"Full-finetune/Learning Rate": 1.6601943096743426e-05, "Full-finetune/Loss": 0.8107185959815979, "Full-finetune/Loss (Raw)": 0.7168064117431641, "Full-finetune/Step": 1519, "Full-finetune/Step Time": 6.758031317964196} {"Full-finetune/Learning Rate": 1.6597262787523493e-05, "Full-finetune/Loss": 0.8076672554016113, "Full-finetune/Loss (Raw)": 0.41874492168426514, "Full-finetune/Step": 1520, "Full-finetune/Step Time": 6.740899778902531} {"Full-finetune/Learning Rate": 1.6592579918034678e-05, "Full-finetune/Loss": 0.8087188005447388, "Full-finetune/Loss (Raw)": 0.9094734787940979, "Full-finetune/Step": 1521, "Full-finetune/Step Time": 6.7053946647793055} {"Full-finetune/Learning Rate": 1.65878944900943e-05, "Full-finetune/Loss": 0.8071514368057251, "Full-finetune/Loss (Raw)": 0.7415274977684021, "Full-finetune/Step": 1522, "Full-finetune/Step Time": 6.71648683026433} {"Full-finetune/Learning Rate": 1.6583206505520695e-05, "Full-finetune/Loss": 0.8058682680130005, "Full-finetune/Loss (Raw)": 0.6665711402893066, "Full-finetune/Step": 1523, "Full-finetune/Step Time": 6.729836340993643} {"Full-finetune/Learning Rate": 1.6578515966133167e-05, "Full-finetune/Loss": 0.8066698908805847, "Full-finetune/Loss (Raw)": 0.716833233833313, "Full-finetune/Step": 1524, "Full-finetune/Step Time": 6.747002610936761} {"Full-finetune/Learning Rate": 1.6573822873752026e-05, "Full-finetune/Loss": 0.8053567409515381, "Full-finetune/Loss (Raw)": 0.7036985158920288, "Full-finetune/Step": 1525, "Full-finetune/Step Time": 6.73459099419415} {"Full-finetune/Learning Rate": 1.6569127230198575e-05, "Full-finetune/Loss": 0.8059607744216919, "Full-finetune/Loss (Raw)": 0.9480580687522888, "Full-finetune/Step": 1526, "Full-finetune/Step Time": 6.733804220333695} {"Full-finetune/Learning Rate": 1.6564429037295097e-05, "Full-finetune/Loss": 0.8078078627586365, "Full-finetune/Loss (Raw)": 0.8471897840499878, "Full-finetune/Step": 1527, "Full-finetune/Step Time": 6.73939798399806} {"Full-finetune/Learning Rate": 1.6559728296864868e-05, "Full-finetune/Loss": 0.8038145303726196, "Full-finetune/Loss (Raw)": 0.43219324946403503, "Full-finetune/Step": 1528, "Full-finetune/Step Time": 6.7501750234514475} {"Full-finetune/Learning Rate": 1.6555025010732156e-05, "Full-finetune/Loss": 0.8050029277801514, "Full-finetune/Loss (Raw)": 0.9711868762969971, "Full-finetune/Step": 1529, "Full-finetune/Step Time": 6.742664175108075} {"Full-finetune/Learning Rate": 1.6550319180722213e-05, "Full-finetune/Loss": 0.8065119981765747, "Full-finetune/Loss (Raw)": 0.5882647633552551, "Full-finetune/Step": 1530, "Full-finetune/Step Time": 6.7430888917297125} {"Full-finetune/Learning Rate": 1.6545610808661278e-05, "Full-finetune/Loss": 0.8072108626365662, "Full-finetune/Loss (Raw)": 0.9271942973136902, "Full-finetune/Step": 1531, "Full-finetune/Step Time": 6.745611634105444} {"Full-finetune/Learning Rate": 1.654089989637658e-05, "Full-finetune/Loss": 0.8070586919784546, "Full-finetune/Loss (Raw)": 0.8694681525230408, "Full-finetune/Step": 1532, "Full-finetune/Step Time": 6.7541293147951365} {"Full-finetune/Learning Rate": 1.6536186445696337e-05, "Full-finetune/Loss": 0.8068394660949707, "Full-finetune/Loss (Raw)": 0.6711425185203552, "Full-finetune/Step": 1533, "Full-finetune/Step Time": 6.757877780124545} {"Full-finetune/Learning Rate": 1.653147045844974e-05, "Full-finetune/Loss": 0.8056015372276306, "Full-finetune/Loss (Raw)": 0.6877558827400208, "Full-finetune/Step": 1534, "Full-finetune/Step Time": 6.765266807749867} {"Full-finetune/Learning Rate": 1.6526751936466974e-05, "Full-finetune/Loss": 0.8090461492538452, "Full-finetune/Loss (Raw)": 0.8211790919303894, "Full-finetune/Step": 1535, "Full-finetune/Step Time": 6.748895598575473} {"Full-finetune/Learning Rate": 1.6522030881579207e-05, "Full-finetune/Loss": 0.8095285296440125, "Full-finetune/Loss (Raw)": 0.7962818145751953, "Full-finetune/Step": 1536, "Full-finetune/Step Time": 6.752902774140239} {"Full-finetune/Learning Rate": 1.6517307295618582e-05, "Full-finetune/Loss": 0.8087361454963684, "Full-finetune/Loss (Raw)": 0.8228817582130432, "Full-finetune/Step": 1537, "Full-finetune/Step Time": 6.752509506419301} {"Full-finetune/Learning Rate": 1.6512581180418245e-05, "Full-finetune/Loss": 0.8062481880187988, "Full-finetune/Loss (Raw)": 0.4443746507167816, "Full-finetune/Step": 1538, "Full-finetune/Step Time": 6.740270620211959} {"Full-finetune/Learning Rate": 1.650785253781229e-05, "Full-finetune/Loss": 0.8062665462493896, "Full-finetune/Loss (Raw)": 0.8739175200462341, "Full-finetune/Step": 1539, "Full-finetune/Step Time": 6.742432111874223} {"Full-finetune/Learning Rate": 1.6503121369635822e-05, "Full-finetune/Loss": 0.8074227571487427, "Full-finetune/Loss (Raw)": 0.8198713064193726, "Full-finetune/Step": 1540, "Full-finetune/Step Time": 6.742165522649884} {"Full-finetune/Learning Rate": 1.6498387677724917e-05, "Full-finetune/Loss": 0.8075699806213379, "Full-finetune/Loss (Raw)": 0.9353756308555603, "Full-finetune/Step": 1541, "Full-finetune/Step Time": 6.74378190562129} {"Full-finetune/Learning Rate": 1.6493651463916622e-05, "Full-finetune/Loss": 0.8048031330108643, "Full-finetune/Loss (Raw)": 0.6679255962371826, "Full-finetune/Step": 1542, "Full-finetune/Step Time": 6.750224988907576} {"Full-finetune/Learning Rate": 1.6488912730048973e-05, "Full-finetune/Loss": 0.8041270971298218, "Full-finetune/Loss (Raw)": 0.7626203894615173, "Full-finetune/Step": 1543, "Full-finetune/Step Time": 6.746651094406843} {"Full-finetune/Learning Rate": 1.6484171477960978e-05, "Full-finetune/Loss": 0.8036507368087769, "Full-finetune/Loss (Raw)": 0.9080283641815186, "Full-finetune/Step": 1544, "Full-finetune/Step Time": 6.745001759380102} {"Full-finetune/Learning Rate": 1.6479427709492622e-05, "Full-finetune/Loss": 0.8034493923187256, "Full-finetune/Loss (Raw)": 0.8170924186706543, "Full-finetune/Step": 1545, "Full-finetune/Step Time": 6.737438166514039} {"Full-finetune/Learning Rate": 1.6474681426484878e-05, "Full-finetune/Loss": 0.8052508234977722, "Full-finetune/Loss (Raw)": 0.9510756731033325, "Full-finetune/Step": 1546, "Full-finetune/Step Time": 6.748011415824294} {"Full-finetune/Learning Rate": 1.6469932630779677e-05, "Full-finetune/Loss": 0.8062025904655457, "Full-finetune/Loss (Raw)": 0.8502110242843628, "Full-finetune/Step": 1547, "Full-finetune/Step Time": 6.752431467175484} {"Full-finetune/Learning Rate": 1.6465181324219937e-05, "Full-finetune/Loss": 0.8056372404098511, "Full-finetune/Loss (Raw)": 0.8418694734573364, "Full-finetune/Step": 1548, "Full-finetune/Step Time": 6.75165512226522} {"Full-finetune/Learning Rate": 1.6460427508649546e-05, "Full-finetune/Loss": 0.8024677038192749, "Full-finetune/Loss (Raw)": 0.4162440299987793, "Full-finetune/Step": 1549, "Full-finetune/Step Time": 6.769464934244752} {"Full-finetune/Learning Rate": 1.6455671185913367e-05, "Full-finetune/Loss": 0.8033535480499268, "Full-finetune/Loss (Raw)": 0.8891458511352539, "Full-finetune/Step": 1550, "Full-finetune/Step Time": 6.768491001799703} {"Full-finetune/Learning Rate": 1.645091235785724e-05, "Full-finetune/Loss": 0.8018426895141602, "Full-finetune/Loss (Raw)": 0.7546595931053162, "Full-finetune/Step": 1551, "Full-finetune/Step Time": 6.762954218313098} {"Full-finetune/Learning Rate": 1.644615102632797e-05, "Full-finetune/Loss": 0.8029125332832336, "Full-finetune/Loss (Raw)": 0.9228941798210144, "Full-finetune/Step": 1552, "Full-finetune/Step Time": 6.7638914454728365} {"Full-finetune/Learning Rate": 1.6441387193173337e-05, "Full-finetune/Loss": 0.8025015592575073, "Full-finetune/Loss (Raw)": 0.7881379723548889, "Full-finetune/Step": 1553, "Full-finetune/Step Time": 6.775436131283641} {"Full-finetune/Learning Rate": 1.6436620860242093e-05, "Full-finetune/Loss": 0.8023291826248169, "Full-finetune/Loss (Raw)": 0.7636072635650635, "Full-finetune/Step": 1554, "Full-finetune/Step Time": 6.766541289165616} {"Full-finetune/Learning Rate": 1.6431852029383955e-05, "Full-finetune/Loss": 0.8037528395652771, "Full-finetune/Loss (Raw)": 0.924914538860321, "Full-finetune/Step": 1555, "Full-finetune/Step Time": 6.769118586555123} {"Full-finetune/Learning Rate": 1.6427080702449616e-05, "Full-finetune/Loss": 0.8033574223518372, "Full-finetune/Loss (Raw)": 0.7075476050376892, "Full-finetune/Step": 1556, "Full-finetune/Step Time": 6.774922510609031} {"Full-finetune/Learning Rate": 1.6422306881290734e-05, "Full-finetune/Loss": 0.8040045499801636, "Full-finetune/Loss (Raw)": 0.8042434453964233, "Full-finetune/Step": 1557, "Full-finetune/Step Time": 6.774674961343408} {"Full-finetune/Learning Rate": 1.641753056775994e-05, "Full-finetune/Loss": 0.8041269779205322, "Full-finetune/Loss (Raw)": 0.8493586182594299, "Full-finetune/Step": 1558, "Full-finetune/Step Time": 6.765164624899626} {"Full-finetune/Learning Rate": 1.641275176371082e-05, "Full-finetune/Loss": 0.8066956400871277, "Full-finetune/Loss (Raw)": 1.0878016948699951, "Full-finetune/Step": 1559, "Full-finetune/Step Time": 6.75684616714716} {"Full-finetune/Learning Rate": 1.6407970470997943e-05, "Full-finetune/Loss": 0.8076692819595337, "Full-finetune/Loss (Raw)": 0.9058234095573425, "Full-finetune/Step": 1560, "Full-finetune/Step Time": 6.762440895661712} {"Full-finetune/Learning Rate": 1.6403186691476828e-05, "Full-finetune/Loss": 0.8070106506347656, "Full-finetune/Loss (Raw)": 0.8764896988868713, "Full-finetune/Step": 1561, "Full-finetune/Step Time": 6.749587723985314} {"Full-finetune/Learning Rate": 1.639840042700397e-05, "Full-finetune/Loss": 0.8109422922134399, "Full-finetune/Loss (Raw)": 0.8818965554237366, "Full-finetune/Step": 1562, "Full-finetune/Step Time": 6.741810563951731} {"Full-finetune/Learning Rate": 1.6393611679436828e-05, "Full-finetune/Loss": 0.8099410533905029, "Full-finetune/Loss (Raw)": 0.7830122113227844, "Full-finetune/Step": 1563, "Full-finetune/Step Time": 6.758237903937697} {"Full-finetune/Learning Rate": 1.6388820450633813e-05, "Full-finetune/Loss": 0.8113020658493042, "Full-finetune/Loss (Raw)": 0.7798011302947998, "Full-finetune/Step": 1564, "Full-finetune/Step Time": 6.7452891785651445} {"Full-finetune/Learning Rate": 1.6384026742454315e-05, "Full-finetune/Loss": 0.8100823760032654, "Full-finetune/Loss (Raw)": 0.7414989471435547, "Full-finetune/Step": 1565, "Full-finetune/Step Time": 6.744714306667447} {"Full-finetune/Learning Rate": 1.6379230556758676e-05, "Full-finetune/Loss": 0.8090273141860962, "Full-finetune/Loss (Raw)": 0.7672796249389648, "Full-finetune/Step": 1566, "Full-finetune/Step Time": 6.752469209954143} {"Full-finetune/Learning Rate": 1.63744318954082e-05, "Full-finetune/Loss": 0.8082230687141418, "Full-finetune/Loss (Raw)": 0.8476439714431763, "Full-finetune/Step": 1567, "Full-finetune/Step Time": 6.758869204670191} {"Full-finetune/Learning Rate": 1.636963076026516e-05, "Full-finetune/Loss": 0.8098961114883423, "Full-finetune/Loss (Raw)": 0.8379943370819092, "Full-finetune/Step": 1568, "Full-finetune/Step Time": 6.732065953314304} {"Full-finetune/Learning Rate": 1.6364827153192773e-05, "Full-finetune/Loss": 0.8086237907409668, "Full-finetune/Loss (Raw)": 0.7063856720924377, "Full-finetune/Step": 1569, "Full-finetune/Step Time": 6.743819050490856} {"Full-finetune/Learning Rate": 1.6360021076055234e-05, "Full-finetune/Loss": 0.8096632361412048, "Full-finetune/Loss (Raw)": 0.9621904492378235, "Full-finetune/Step": 1570, "Full-finetune/Step Time": 6.742423957213759} {"Full-finetune/Learning Rate": 1.6355212530717683e-05, "Full-finetune/Loss": 0.810053825378418, "Full-finetune/Loss (Raw)": 0.8931110501289368, "Full-finetune/Step": 1571, "Full-finetune/Step Time": 6.73828842677176} {"Full-finetune/Learning Rate": 1.6350401519046226e-05, "Full-finetune/Loss": 0.8111070394515991, "Full-finetune/Loss (Raw)": 0.9216011166572571, "Full-finetune/Step": 1572, "Full-finetune/Step Time": 6.73234823718667} {"Full-finetune/Learning Rate": 1.634558804290792e-05, "Full-finetune/Loss": 0.8108296394348145, "Full-finetune/Loss (Raw)": 0.8223961591720581, "Full-finetune/Step": 1573, "Full-finetune/Step Time": 6.748439682647586} {"Full-finetune/Learning Rate": 1.634077210417078e-05, "Full-finetune/Loss": 0.8108570575714111, "Full-finetune/Loss (Raw)": 0.7228041887283325, "Full-finetune/Step": 1574, "Full-finetune/Step Time": 6.732390254735947} {"Full-finetune/Learning Rate": 1.633595370470378e-05, "Full-finetune/Loss": 0.811316967010498, "Full-finetune/Loss (Raw)": 0.9693968892097473, "Full-finetune/Step": 1575, "Full-finetune/Step Time": 6.74162015132606} {"Full-finetune/Learning Rate": 1.6331132846376843e-05, "Full-finetune/Loss": 0.8109551668167114, "Full-finetune/Loss (Raw)": 0.87007737159729, "Full-finetune/Step": 1576, "Full-finetune/Step Time": 6.745184486731887} {"Full-finetune/Learning Rate": 1.632630953106086e-05, "Full-finetune/Loss": 0.8118997812271118, "Full-finetune/Loss (Raw)": 0.8644285798072815, "Full-finetune/Step": 1577, "Full-finetune/Step Time": 6.7358597833663225} {"Full-finetune/Learning Rate": 1.6321483760627658e-05, "Full-finetune/Loss": 0.8132619857788086, "Full-finetune/Loss (Raw)": 0.8925416469573975, "Full-finetune/Step": 1578, "Full-finetune/Step Time": 6.706173058599234} {"Full-finetune/Learning Rate": 1.6316655536950017e-05, "Full-finetune/Loss": 0.8127068877220154, "Full-finetune/Loss (Raw)": 0.7732105255126953, "Full-finetune/Step": 1579, "Full-finetune/Step Time": 6.711975486949086} {"Full-finetune/Learning Rate": 1.6311824861901693e-05, "Full-finetune/Loss": 0.8108001947402954, "Full-finetune/Loss (Raw)": 0.6869348883628845, "Full-finetune/Step": 1580, "Full-finetune/Step Time": 6.745927268639207} {"Full-finetune/Learning Rate": 1.6306991737357366e-05, "Full-finetune/Loss": 0.8105520606040955, "Full-finetune/Loss (Raw)": 0.7568461894989014, "Full-finetune/Step": 1581, "Full-finetune/Step Time": 6.741459436714649} {"Full-finetune/Learning Rate": 1.630215616519268e-05, "Full-finetune/Loss": 0.8088464736938477, "Full-finetune/Loss (Raw)": 0.8396249413490295, "Full-finetune/Step": 1582, "Full-finetune/Step Time": 6.741943567991257} {"Full-finetune/Learning Rate": 1.6297318147284223e-05, "Full-finetune/Loss": 0.8084327578544617, "Full-finetune/Loss (Raw)": 0.8850961327552795, "Full-finetune/Step": 1583, "Full-finetune/Step Time": 6.745045151561499} {"Full-finetune/Learning Rate": 1.6292477685509538e-05, "Full-finetune/Loss": 0.8086121678352356, "Full-finetune/Loss (Raw)": 0.8454082012176514, "Full-finetune/Step": 1584, "Full-finetune/Step Time": 6.746858399361372} {"Full-finetune/Learning Rate": 1.6287634781747113e-05, "Full-finetune/Loss": 0.8131081461906433, "Full-finetune/Loss (Raw)": 0.9527193903923035, "Full-finetune/Step": 1585, "Full-finetune/Step Time": 6.7290972881019115} {"Full-finetune/Learning Rate": 1.628278943787639e-05, "Full-finetune/Loss": 0.8146187663078308, "Full-finetune/Loss (Raw)": 1.0583053827285767, "Full-finetune/Step": 1586, "Full-finetune/Step Time": 6.769175257533789} {"Full-finetune/Learning Rate": 1.6277941655777744e-05, "Full-finetune/Loss": 0.8144354820251465, "Full-finetune/Loss (Raw)": 0.9504477977752686, "Full-finetune/Step": 1587, "Full-finetune/Step Time": 6.762488000094891} {"Full-finetune/Learning Rate": 1.6273091437332506e-05, "Full-finetune/Loss": 0.8131117820739746, "Full-finetune/Loss (Raw)": 0.8390188217163086, "Full-finetune/Step": 1588, "Full-finetune/Step Time": 6.773526286706328} {"Full-finetune/Learning Rate": 1.6268238784422954e-05, "Full-finetune/Loss": 0.8114892840385437, "Full-finetune/Loss (Raw)": 0.8435556888580322, "Full-finetune/Step": 1589, "Full-finetune/Step Time": 6.788400450721383} {"Full-finetune/Learning Rate": 1.6263383698932307e-05, "Full-finetune/Loss": 0.8117963671684265, "Full-finetune/Loss (Raw)": 0.9388488531112671, "Full-finetune/Step": 1590, "Full-finetune/Step Time": 6.793634528294206} {"Full-finetune/Learning Rate": 1.6258526182744727e-05, "Full-finetune/Loss": 0.8091691732406616, "Full-finetune/Loss (Raw)": 0.6271541714668274, "Full-finetune/Step": 1591, "Full-finetune/Step Time": 6.808096919208765} {"Full-finetune/Learning Rate": 1.6253666237745326e-05, "Full-finetune/Loss": 0.8082988262176514, "Full-finetune/Loss (Raw)": 0.6997517943382263, "Full-finetune/Step": 1592, "Full-finetune/Step Time": 6.820494892075658} {"Full-finetune/Learning Rate": 1.6248803865820152e-05, "Full-finetune/Loss": 0.8076047897338867, "Full-finetune/Loss (Raw)": 0.7735562920570374, "Full-finetune/Step": 1593, "Full-finetune/Step Time": 6.823495078831911} {"Full-finetune/Learning Rate": 1.624393906885619e-05, "Full-finetune/Loss": 0.8073129653930664, "Full-finetune/Loss (Raw)": 0.8309999704360962, "Full-finetune/Step": 1594, "Full-finetune/Step Time": 6.8287461791187525} {"Full-finetune/Learning Rate": 1.6239071848741385e-05, "Full-finetune/Loss": 0.8061820864677429, "Full-finetune/Loss (Raw)": 0.7041888236999512, "Full-finetune/Step": 1595, "Full-finetune/Step Time": 6.806720769032836} {"Full-finetune/Learning Rate": 1.6234202207364596e-05, "Full-finetune/Loss": 0.8056750297546387, "Full-finetune/Loss (Raw)": 0.8393676280975342, "Full-finetune/Step": 1596, "Full-finetune/Step Time": 6.811046205461025} {"Full-finetune/Learning Rate": 1.6229330146615646e-05, "Full-finetune/Loss": 0.808295488357544, "Full-finetune/Loss (Raw)": 0.72490394115448, "Full-finetune/Step": 1597, "Full-finetune/Step Time": 6.790822813287377} {"Full-finetune/Learning Rate": 1.6224455668385283e-05, "Full-finetune/Loss": 0.8083471059799194, "Full-finetune/Loss (Raw)": 0.8123667240142822, "Full-finetune/Step": 1598, "Full-finetune/Step Time": 6.80182196944952} {"Full-finetune/Learning Rate": 1.62195787745652e-05, "Full-finetune/Loss": 0.8047060966491699, "Full-finetune/Loss (Raw)": 0.3880433440208435, "Full-finetune/Step": 1599, "Full-finetune/Step Time": 6.820454932749271} {"Full-finetune/Learning Rate": 1.6214699467048017e-05, "Full-finetune/Loss": 0.8057130575180054, "Full-finetune/Loss (Raw)": 0.9148741960525513, "Full-finetune/Step": 1600, "Full-finetune/Step Time": 6.8252439219504595} {"Full-finetune/Learning Rate": 1.6209817747727304e-05, "Full-finetune/Loss": 0.804231584072113, "Full-finetune/Loss (Raw)": 0.4022423028945923, "Full-finetune/Step": 1601, "Full-finetune/Step Time": 6.837358381599188} {"Full-finetune/Learning Rate": 1.620493361849756e-05, "Full-finetune/Loss": 0.8037383556365967, "Full-finetune/Loss (Raw)": 0.9657278656959534, "Full-finetune/Step": 1602, "Full-finetune/Step Time": 6.847923494875431} {"Full-finetune/Learning Rate": 1.6200047081254223e-05, "Full-finetune/Loss": 0.8056817054748535, "Full-finetune/Loss (Raw)": 0.9363927245140076, "Full-finetune/Step": 1603, "Full-finetune/Step Time": 6.843421017751098} {"Full-finetune/Learning Rate": 1.6195158137893656e-05, "Full-finetune/Loss": 0.8046779632568359, "Full-finetune/Loss (Raw)": 0.8038413524627686, "Full-finetune/Step": 1604, "Full-finetune/Step Time": 6.8436274491250515} {"Full-finetune/Learning Rate": 1.6190266790313165e-05, "Full-finetune/Loss": 0.8074102401733398, "Full-finetune/Loss (Raw)": 0.6687481999397278, "Full-finetune/Step": 1605, "Full-finetune/Step Time": 6.834593329578638} {"Full-finetune/Learning Rate": 1.618537304041099e-05, "Full-finetune/Loss": 0.8043627738952637, "Full-finetune/Loss (Raw)": 0.6317569613456726, "Full-finetune/Step": 1606, "Full-finetune/Step Time": 6.863293945789337} {"Full-finetune/Learning Rate": 1.6180476890086297e-05, "Full-finetune/Loss": 0.8049236536026001, "Full-finetune/Loss (Raw)": 0.8307418823242188, "Full-finetune/Step": 1607, "Full-finetune/Step Time": 6.850832285359502} {"Full-finetune/Learning Rate": 1.6175578341239188e-05, "Full-finetune/Loss": 0.8011394739151001, "Full-finetune/Loss (Raw)": 0.3999689817428589, "Full-finetune/Step": 1608, "Full-finetune/Step Time": 6.858553620055318} {"Full-finetune/Learning Rate": 1.617067739577069e-05, "Full-finetune/Loss": 0.8009214401245117, "Full-finetune/Loss (Raw)": 0.881878137588501, "Full-finetune/Step": 1609, "Full-finetune/Step Time": 6.85491906106472} {"Full-finetune/Learning Rate": 1.616577405558277e-05, "Full-finetune/Loss": 0.8013551235198975, "Full-finetune/Loss (Raw)": 0.8757513165473938, "Full-finetune/Step": 1610, "Full-finetune/Step Time": 6.856707515195012} {"Full-finetune/Learning Rate": 1.616086832257831e-05, "Full-finetune/Loss": 0.8024442195892334, "Full-finetune/Loss (Raw)": 0.7501984238624573, "Full-finetune/Step": 1611, "Full-finetune/Step Time": 6.848939033225179} {"Full-finetune/Learning Rate": 1.615596019866114e-05, "Full-finetune/Loss": 0.8017815947532654, "Full-finetune/Loss (Raw)": 0.8208391070365906, "Full-finetune/Step": 1612, "Full-finetune/Step Time": 6.852433593943715} {"Full-finetune/Learning Rate": 1.6151049685736e-05, "Full-finetune/Loss": 0.8011480569839478, "Full-finetune/Loss (Raw)": 0.730995774269104, "Full-finetune/Step": 1613, "Full-finetune/Step Time": 6.863875467330217} {"Full-finetune/Learning Rate": 1.6146136785708564e-05, "Full-finetune/Loss": 0.8057147860527039, "Full-finetune/Loss (Raw)": 0.8978105187416077, "Full-finetune/Step": 1614, "Full-finetune/Step Time": 6.845954138785601} {"Full-finetune/Learning Rate": 1.6141221500485437e-05, "Full-finetune/Loss": 0.8051501512527466, "Full-finetune/Loss (Raw)": 0.9544179439544678, "Full-finetune/Step": 1615, "Full-finetune/Step Time": 6.8389454539865255} {"Full-finetune/Learning Rate": 1.6136303831974146e-05, "Full-finetune/Loss": 0.8030880689620972, "Full-finetune/Loss (Raw)": 0.8516255617141724, "Full-finetune/Step": 1616, "Full-finetune/Step Time": 6.827128944918513} {"Full-finetune/Learning Rate": 1.6131383782083135e-05, "Full-finetune/Loss": 0.80426025390625, "Full-finetune/Loss (Raw)": 0.8124091029167175, "Full-finetune/Step": 1617, "Full-finetune/Step Time": 6.824939347803593} {"Full-finetune/Learning Rate": 1.6126461352721784e-05, "Full-finetune/Loss": 0.8045802712440491, "Full-finetune/Loss (Raw)": 0.8292161822319031, "Full-finetune/Step": 1618, "Full-finetune/Step Time": 6.805246716365218} {"Full-finetune/Learning Rate": 1.612153654580039e-05, "Full-finetune/Loss": 0.8048317432403564, "Full-finetune/Loss (Raw)": 0.7513571977615356, "Full-finetune/Step": 1619, "Full-finetune/Step Time": 6.795707039535046} {"Full-finetune/Learning Rate": 1.6116609363230176e-05, "Full-finetune/Loss": 0.8059254884719849, "Full-finetune/Loss (Raw)": 1.0911508798599243, "Full-finetune/Step": 1620, "Full-finetune/Step Time": 6.798182539641857} {"Full-finetune/Learning Rate": 1.6111679806923287e-05, "Full-finetune/Loss": 0.8094000816345215, "Full-finetune/Loss (Raw)": 0.7833510041236877, "Full-finetune/Step": 1621, "Full-finetune/Step Time": 6.786960441619158} {"Full-finetune/Learning Rate": 1.6106747878792784e-05, "Full-finetune/Loss": 0.8104512691497803, "Full-finetune/Loss (Raw)": 0.9679588079452515, "Full-finetune/Step": 1622, "Full-finetune/Step Time": 6.7847733329981565} {"Full-finetune/Learning Rate": 1.610181358075265e-05, "Full-finetune/Loss": 0.8121278882026672, "Full-finetune/Loss (Raw)": 1.0624616146087646, "Full-finetune/Step": 1623, "Full-finetune/Step Time": 6.780182085931301} {"Full-finetune/Learning Rate": 1.60968769147178e-05, "Full-finetune/Loss": 0.8118383884429932, "Full-finetune/Loss (Raw)": 0.8819566369056702, "Full-finetune/Step": 1624, "Full-finetune/Step Time": 6.782315978780389} {"Full-finetune/Learning Rate": 1.6091937882604047e-05, "Full-finetune/Loss": 0.8127070069313049, "Full-finetune/Loss (Raw)": 0.8785200119018555, "Full-finetune/Step": 1625, "Full-finetune/Step Time": 6.782716602087021} {"Full-finetune/Learning Rate": 1.608699648632814e-05, "Full-finetune/Loss": 0.8120518922805786, "Full-finetune/Loss (Raw)": 0.9010629057884216, "Full-finetune/Step": 1626, "Full-finetune/Step Time": 6.7741165570914745} {"Full-finetune/Learning Rate": 1.6082052727807732e-05, "Full-finetune/Loss": 0.8159840106964111, "Full-finetune/Loss (Raw)": 0.9454078078269958, "Full-finetune/Step": 1627, "Full-finetune/Step Time": 6.754531981423497} {"Full-finetune/Learning Rate": 1.60771066089614e-05, "Full-finetune/Loss": 0.8195583820343018, "Full-finetune/Loss (Raw)": 0.8210272789001465, "Full-finetune/Step": 1628, "Full-finetune/Step Time": 6.735370151698589} {"Full-finetune/Learning Rate": 1.6072158131708643e-05, "Full-finetune/Loss": 0.8179537653923035, "Full-finetune/Loss (Raw)": 0.7914749979972839, "Full-finetune/Step": 1629, "Full-finetune/Step Time": 6.706513702869415} {"Full-finetune/Learning Rate": 1.6067207297969864e-05, "Full-finetune/Loss": 0.8172580003738403, "Full-finetune/Loss (Raw)": 0.8732632398605347, "Full-finetune/Step": 1630, "Full-finetune/Step Time": 6.704614773392677} {"Full-finetune/Learning Rate": 1.6062254109666383e-05, "Full-finetune/Loss": 0.818289041519165, "Full-finetune/Loss (Raw)": 0.8565414547920227, "Full-finetune/Step": 1631, "Full-finetune/Step Time": 6.708618752658367} {"Full-finetune/Learning Rate": 1.6057298568720436e-05, "Full-finetune/Loss": 0.8177827596664429, "Full-finetune/Loss (Raw)": 0.8163288831710815, "Full-finetune/Step": 1632, "Full-finetune/Step Time": 6.699631510302424} {"Full-finetune/Learning Rate": 1.6052340677055175e-05, "Full-finetune/Loss": 0.8172218203544617, "Full-finetune/Loss (Raw)": 0.7937493920326233, "Full-finetune/Step": 1633, "Full-finetune/Step Time": 6.701831253245473} {"Full-finetune/Learning Rate": 1.6047380436594663e-05, "Full-finetune/Loss": 0.8178753852844238, "Full-finetune/Loss (Raw)": 0.8704883456230164, "Full-finetune/Step": 1634, "Full-finetune/Step Time": 6.694856099784374} {"Full-finetune/Learning Rate": 1.6042417849263863e-05, "Full-finetune/Loss": 0.8166069984436035, "Full-finetune/Loss (Raw)": 0.7887760400772095, "Full-finetune/Step": 1635, "Full-finetune/Step Time": 6.717167446389794} {"Full-finetune/Learning Rate": 1.6037452916988665e-05, "Full-finetune/Loss": 0.8162517547607422, "Full-finetune/Loss (Raw)": 0.8990742564201355, "Full-finetune/Step": 1636, "Full-finetune/Step Time": 6.713808411732316} {"Full-finetune/Learning Rate": 1.6032485641695862e-05, "Full-finetune/Loss": 0.815747082233429, "Full-finetune/Loss (Raw)": 0.7968422174453735, "Full-finetune/Step": 1637, "Full-finetune/Step Time": 6.715618444606662} {"Full-finetune/Learning Rate": 1.602751602531316e-05, "Full-finetune/Loss": 0.8145390748977661, "Full-finetune/Loss (Raw)": 0.8024473190307617, "Full-finetune/Step": 1638, "Full-finetune/Step Time": 6.7183316219598055} {"Full-finetune/Learning Rate": 1.602254406976916e-05, "Full-finetune/Loss": 0.8128471970558167, "Full-finetune/Loss (Raw)": 0.7418414354324341, "Full-finetune/Step": 1639, "Full-finetune/Step Time": 6.715904578566551} {"Full-finetune/Learning Rate": 1.6017569776993388e-05, "Full-finetune/Loss": 0.8145260810852051, "Full-finetune/Loss (Raw)": 0.9663177132606506, "Full-finetune/Step": 1640, "Full-finetune/Step Time": 6.705848306417465} {"Full-finetune/Learning Rate": 1.601259314891627e-05, "Full-finetune/Loss": 0.8137940764427185, "Full-finetune/Loss (Raw)": 0.9025057554244995, "Full-finetune/Step": 1641, "Full-finetune/Step Time": 6.704480201005936} {"Full-finetune/Learning Rate": 1.6007614187469137e-05, "Full-finetune/Loss": 0.8141863346099854, "Full-finetune/Loss (Raw)": 0.8514770269393921, "Full-finetune/Step": 1642, "Full-finetune/Step Time": 6.6947189550846815} {"Full-finetune/Learning Rate": 1.6002632894584227e-05, "Full-finetune/Loss": 0.8150101900100708, "Full-finetune/Loss (Raw)": 0.9776703715324402, "Full-finetune/Step": 1643, "Full-finetune/Step Time": 6.696708543226123} {"Full-finetune/Learning Rate": 1.599764927219468e-05, "Full-finetune/Loss": 0.8142192363739014, "Full-finetune/Loss (Raw)": 0.7956841588020325, "Full-finetune/Step": 1644, "Full-finetune/Step Time": 6.731759425252676} {"Full-finetune/Learning Rate": 1.5992663322234548e-05, "Full-finetune/Loss": 0.8146265745162964, "Full-finetune/Loss (Raw)": 0.8508630990982056, "Full-finetune/Step": 1645, "Full-finetune/Step Time": 6.735056338831782} {"Full-finetune/Learning Rate": 1.5987675046638777e-05, "Full-finetune/Loss": 0.8140040636062622, "Full-finetune/Loss (Raw)": 0.9193987846374512, "Full-finetune/Step": 1646, "Full-finetune/Step Time": 6.760828256607056} {"Full-finetune/Learning Rate": 1.5982684447343214e-05, "Full-finetune/Loss": 0.8144870400428772, "Full-finetune/Loss (Raw)": 0.7786235809326172, "Full-finetune/Step": 1647, "Full-finetune/Step Time": 6.783239142969251} {"Full-finetune/Learning Rate": 1.5977691526284623e-05, "Full-finetune/Loss": 0.8185215592384338, "Full-finetune/Loss (Raw)": 0.9351610541343689, "Full-finetune/Step": 1648, "Full-finetune/Step Time": 6.7709806971251965} {"Full-finetune/Learning Rate": 1.5972696285400656e-05, "Full-finetune/Loss": 0.8178654313087463, "Full-finetune/Loss (Raw)": 0.825489342212677, "Full-finetune/Step": 1649, "Full-finetune/Step Time": 6.771582601591945} {"Full-finetune/Learning Rate": 1.596769872662987e-05, "Full-finetune/Loss": 0.8191073536872864, "Full-finetune/Loss (Raw)": 0.9004945158958435, "Full-finetune/Step": 1650, "Full-finetune/Step Time": 6.760431004688144} {"Full-finetune/Learning Rate": 1.5962698851911715e-05, "Full-finetune/Loss": 0.8196582794189453, "Full-finetune/Loss (Raw)": 0.7370875477790833, "Full-finetune/Step": 1651, "Full-finetune/Step Time": 6.762817559763789} {"Full-finetune/Learning Rate": 1.5957696663186547e-05, "Full-finetune/Loss": 0.820475697517395, "Full-finetune/Loss (Raw)": 0.8214625716209412, "Full-finetune/Step": 1652, "Full-finetune/Step Time": 6.747809436172247} {"Full-finetune/Learning Rate": 1.595269216239562e-05, "Full-finetune/Loss": 0.8214969038963318, "Full-finetune/Loss (Raw)": 0.8344118595123291, "Full-finetune/Step": 1653, "Full-finetune/Step Time": 6.740337762981653} {"Full-finetune/Learning Rate": 1.594768535148108e-05, "Full-finetune/Loss": 0.8211039304733276, "Full-finetune/Loss (Raw)": 0.8977671265602112, "Full-finetune/Step": 1654, "Full-finetune/Step Time": 6.753349535167217} {"Full-finetune/Learning Rate": 1.5942676232385973e-05, "Full-finetune/Loss": 0.8176324367523193, "Full-finetune/Loss (Raw)": 0.40284034609794617, "Full-finetune/Step": 1655, "Full-finetune/Step Time": 6.759631806984544} {"Full-finetune/Learning Rate": 1.5937664807054242e-05, "Full-finetune/Loss": 0.8211959600448608, "Full-finetune/Loss (Raw)": 0.888313889503479, "Full-finetune/Step": 1656, "Full-finetune/Step Time": 6.746385773643851} {"Full-finetune/Learning Rate": 1.593265107743072e-05, "Full-finetune/Loss": 0.8199533224105835, "Full-finetune/Loss (Raw)": 0.8121358752250671, "Full-finetune/Step": 1657, "Full-finetune/Step Time": 6.743936797603965} {"Full-finetune/Learning Rate": 1.5927635045461145e-05, "Full-finetune/Loss": 0.8222602605819702, "Full-finetune/Loss (Raw)": 0.8835456967353821, "Full-finetune/Step": 1658, "Full-finetune/Step Time": 6.7286385130137205} {"Full-finetune/Learning Rate": 1.5922616713092135e-05, "Full-finetune/Loss": 0.8211769461631775, "Full-finetune/Loss (Raw)": 0.788533091545105, "Full-finetune/Step": 1659, "Full-finetune/Step Time": 6.728448418900371} {"Full-finetune/Learning Rate": 1.5917596082271206e-05, "Full-finetune/Loss": 0.8213135004043579, "Full-finetune/Loss (Raw)": 0.8869398236274719, "Full-finetune/Step": 1660, "Full-finetune/Step Time": 6.726864390075207} {"Full-finetune/Learning Rate": 1.5912573154946768e-05, "Full-finetune/Loss": 0.8226227760314941, "Full-finetune/Loss (Raw)": 0.8387340903282166, "Full-finetune/Step": 1661, "Full-finetune/Step Time": 6.711777972057462} {"Full-finetune/Learning Rate": 1.5907547933068125e-05, "Full-finetune/Loss": 0.8246468305587769, "Full-finetune/Loss (Raw)": 0.9468385577201843, "Full-finetune/Step": 1662, "Full-finetune/Step Time": 6.711581727489829} {"Full-finetune/Learning Rate": 1.590252041858546e-05, "Full-finetune/Loss": 0.8228356838226318, "Full-finetune/Loss (Raw)": 0.5893499851226807, "Full-finetune/Step": 1663, "Full-finetune/Step Time": 6.721522878855467} {"Full-finetune/Learning Rate": 1.5897490613449864e-05, "Full-finetune/Loss": 0.8218699097633362, "Full-finetune/Loss (Raw)": 0.6726619601249695, "Full-finetune/Step": 1664, "Full-finetune/Step Time": 6.714566338807344} {"Full-finetune/Learning Rate": 1.5892458519613296e-05, "Full-finetune/Loss": 0.8231896758079529, "Full-finetune/Loss (Raw)": 0.991810142993927, "Full-finetune/Step": 1665, "Full-finetune/Step Time": 6.714656556025147} {"Full-finetune/Learning Rate": 1.588742413902862e-05, "Full-finetune/Loss": 0.8261699676513672, "Full-finetune/Loss (Raw)": 0.825849175453186, "Full-finetune/Step": 1666, "Full-finetune/Step Time": 6.692430458962917} {"Full-finetune/Learning Rate": 1.5882387473649572e-05, "Full-finetune/Loss": 0.826303243637085, "Full-finetune/Loss (Raw)": 0.8909786939620972, "Full-finetune/Step": 1667, "Full-finetune/Step Time": 6.688801420852542} {"Full-finetune/Learning Rate": 1.5877348525430793e-05, "Full-finetune/Loss": 0.8271620869636536, "Full-finetune/Loss (Raw)": 0.9298047423362732, "Full-finetune/Step": 1668, "Full-finetune/Step Time": 6.693878864869475} {"Full-finetune/Learning Rate": 1.5872307296327795e-05, "Full-finetune/Loss": 0.8266787528991699, "Full-finetune/Loss (Raw)": 0.873505175113678, "Full-finetune/Step": 1669, "Full-finetune/Step Time": 6.690244629979134} {"Full-finetune/Learning Rate": 1.5867263788296984e-05, "Full-finetune/Loss": 0.8287848234176636, "Full-finetune/Loss (Raw)": 0.9375104308128357, "Full-finetune/Step": 1670, "Full-finetune/Step Time": 6.678241029381752} {"Full-finetune/Learning Rate": 1.5862218003295647e-05, "Full-finetune/Loss": 0.8291471004486084, "Full-finetune/Loss (Raw)": 0.8089930415153503, "Full-finetune/Step": 1671, "Full-finetune/Step Time": 6.685987964272499} {"Full-finetune/Learning Rate": 1.5857169943281948e-05, "Full-finetune/Loss": 0.828647255897522, "Full-finetune/Loss (Raw)": 0.8440427780151367, "Full-finetune/Step": 1672, "Full-finetune/Step Time": 6.702249446883798} {"Full-finetune/Learning Rate": 1.585211961021495e-05, "Full-finetune/Loss": 0.8278093934059143, "Full-finetune/Loss (Raw)": 0.7098506093025208, "Full-finetune/Step": 1673, "Full-finetune/Step Time": 6.707636063918471} {"Full-finetune/Learning Rate": 1.5847067006054584e-05, "Full-finetune/Loss": 0.8248696327209473, "Full-finetune/Loss (Raw)": 0.5747836828231812, "Full-finetune/Step": 1674, "Full-finetune/Step Time": 6.702557496726513} {"Full-finetune/Learning Rate": 1.584201213276167e-05, "Full-finetune/Loss": 0.8235931396484375, "Full-finetune/Loss (Raw)": 0.6868242025375366, "Full-finetune/Step": 1675, "Full-finetune/Step Time": 6.706751378253102} {"Full-finetune/Learning Rate": 1.5836954992297897e-05, "Full-finetune/Loss": 0.8242826461791992, "Full-finetune/Loss (Raw)": 0.9301271438598633, "Full-finetune/Step": 1676, "Full-finetune/Step Time": 6.701804654672742} {"Full-finetune/Learning Rate": 1.583189558662585e-05, "Full-finetune/Loss": 0.828477144241333, "Full-finetune/Loss (Raw)": 0.953128457069397, "Full-finetune/Step": 1677, "Full-finetune/Step Time": 6.683704566210508} {"Full-finetune/Learning Rate": 1.5826833917708983e-05, "Full-finetune/Loss": 0.8276419043540955, "Full-finetune/Loss (Raw)": 0.7822412848472595, "Full-finetune/Step": 1678, "Full-finetune/Step Time": 6.680641295388341} {"Full-finetune/Learning Rate": 1.5821769987511634e-05, "Full-finetune/Loss": 0.824936032295227, "Full-finetune/Loss (Raw)": 0.40831315517425537, "Full-finetune/Step": 1679, "Full-finetune/Step Time": 6.699693636968732} {"Full-finetune/Learning Rate": 1.5816703797999015e-05, "Full-finetune/Loss": 0.8244127631187439, "Full-finetune/Loss (Raw)": 0.8559085726737976, "Full-finetune/Step": 1680, "Full-finetune/Step Time": 6.707585150375962} {"Full-finetune/Learning Rate": 1.5811635351137212e-05, "Full-finetune/Loss": 0.8250547647476196, "Full-finetune/Loss (Raw)": 0.8703178763389587, "Full-finetune/Step": 1681, "Full-finetune/Step Time": 6.689893260598183} {"Full-finetune/Learning Rate": 1.580656464889319e-05, "Full-finetune/Loss": 0.8271484375, "Full-finetune/Loss (Raw)": 1.031599998474121, "Full-finetune/Step": 1682, "Full-finetune/Step Time": 6.693078696727753} {"Full-finetune/Learning Rate": 1.5801491693234793e-05, "Full-finetune/Loss": 0.8253559470176697, "Full-finetune/Loss (Raw)": 0.6954692006111145, "Full-finetune/Step": 1683, "Full-finetune/Step Time": 6.700653973966837} {"Full-finetune/Learning Rate": 1.5796416486130735e-05, "Full-finetune/Loss": 0.8263282775878906, "Full-finetune/Loss (Raw)": 0.832002580165863, "Full-finetune/Step": 1684, "Full-finetune/Step Time": 6.692622393369675} {"Full-finetune/Learning Rate": 1.5791339029550604e-05, "Full-finetune/Loss": 0.8259559869766235, "Full-finetune/Loss (Raw)": 0.7565898299217224, "Full-finetune/Step": 1685, "Full-finetune/Step Time": 6.69625604338944} {"Full-finetune/Learning Rate": 1.578625932546486e-05, "Full-finetune/Loss": 0.8241549134254456, "Full-finetune/Loss (Raw)": 0.6188262104988098, "Full-finetune/Step": 1686, "Full-finetune/Step Time": 6.707638442516327} {"Full-finetune/Learning Rate": 1.578117737584484e-05, "Full-finetune/Loss": 0.8227689266204834, "Full-finetune/Loss (Raw)": 0.9103916883468628, "Full-finetune/Step": 1687, "Full-finetune/Step Time": 6.710717933252454} {"Full-finetune/Learning Rate": 1.577609318266275e-05, "Full-finetune/Loss": 0.8221538066864014, "Full-finetune/Loss (Raw)": 0.8270934224128723, "Full-finetune/Step": 1688, "Full-finetune/Step Time": 6.707516251131892} {"Full-finetune/Learning Rate": 1.5771006747891663e-05, "Full-finetune/Loss": 0.8178219795227051, "Full-finetune/Loss (Raw)": 0.3220140337944031, "Full-finetune/Step": 1689, "Full-finetune/Step Time": 6.722842760384083} {"Full-finetune/Learning Rate": 1.576591807350553e-05, "Full-finetune/Loss": 0.8183772563934326, "Full-finetune/Loss (Raw)": 0.9529694318771362, "Full-finetune/Step": 1690, "Full-finetune/Step Time": 6.714996602386236} {"Full-finetune/Learning Rate": 1.5760827161479158e-05, "Full-finetune/Loss": 0.8190795183181763, "Full-finetune/Loss (Raw)": 0.8729101419448853, "Full-finetune/Step": 1691, "Full-finetune/Step Time": 6.697454443201423} {"Full-finetune/Learning Rate": 1.5755734013788235e-05, "Full-finetune/Loss": 0.8187841176986694, "Full-finetune/Loss (Raw)": 0.7419857382774353, "Full-finetune/Step": 1692, "Full-finetune/Step Time": 6.718977378681302} {"Full-finetune/Learning Rate": 1.5750638632409315e-05, "Full-finetune/Loss": 0.8180919885635376, "Full-finetune/Loss (Raw)": 0.6529108881950378, "Full-finetune/Step": 1693, "Full-finetune/Step Time": 6.737329408526421} {"Full-finetune/Learning Rate": 1.574554101931981e-05, "Full-finetune/Loss": 0.8166921734809875, "Full-finetune/Loss (Raw)": 0.588096559047699, "Full-finetune/Step": 1694, "Full-finetune/Step Time": 6.739428097382188} {"Full-finetune/Learning Rate": 1.5740441176498015e-05, "Full-finetune/Loss": 0.8153998851776123, "Full-finetune/Loss (Raw)": 0.6822282671928406, "Full-finetune/Step": 1695, "Full-finetune/Step Time": 6.745118442922831} {"Full-finetune/Learning Rate": 1.5735339105923068e-05, "Full-finetune/Loss": 0.8154809474945068, "Full-finetune/Loss (Raw)": 0.8483704924583435, "Full-finetune/Step": 1696, "Full-finetune/Step Time": 6.743978206068277} {"Full-finetune/Learning Rate": 1.5730234809574986e-05, "Full-finetune/Loss": 0.8171697854995728, "Full-finetune/Loss (Raw)": 0.9225608110427856, "Full-finetune/Step": 1697, "Full-finetune/Step Time": 6.730689324438572} {"Full-finetune/Learning Rate": 1.5725128289434647e-05, "Full-finetune/Loss": 0.8161709308624268, "Full-finetune/Loss (Raw)": 0.8343349099159241, "Full-finetune/Step": 1698, "Full-finetune/Step Time": 6.732775278389454} {"Full-finetune/Learning Rate": 1.5720019547483795e-05, "Full-finetune/Loss": 0.8161949515342712, "Full-finetune/Loss (Raw)": 0.8961889743804932, "Full-finetune/Step": 1699, "Full-finetune/Step Time": 6.731504572555423} {"Full-finetune/Learning Rate": 1.571490858570503e-05, "Full-finetune/Loss": 0.8157839775085449, "Full-finetune/Loss (Raw)": 0.8689927458763123, "Full-finetune/Step": 1700, "Full-finetune/Step Time": 6.751348095014691} {"Full-finetune/Learning Rate": 1.5709795406081818e-05, "Full-finetune/Loss": 0.8168473243713379, "Full-finetune/Loss (Raw)": 0.9585050940513611, "Full-finetune/Step": 1701, "Full-finetune/Step Time": 6.7369963731616735} {"Full-finetune/Learning Rate": 1.570468001059848e-05, "Full-finetune/Loss": 0.8184428215026855, "Full-finetune/Loss (Raw)": 0.9270328879356384, "Full-finetune/Step": 1702, "Full-finetune/Step Time": 6.736690141260624} {"Full-finetune/Learning Rate": 1.5699562401240206e-05, "Full-finetune/Loss": 0.8160560131072998, "Full-finetune/Loss (Raw)": 0.6638795733451843, "Full-finetune/Step": 1703, "Full-finetune/Step Time": 6.744301091879606} {"Full-finetune/Learning Rate": 1.5694442579993042e-05, "Full-finetune/Loss": 0.8120617270469666, "Full-finetune/Loss (Raw)": 0.3588142693042755, "Full-finetune/Step": 1704, "Full-finetune/Step Time": 6.758530555292964} {"Full-finetune/Learning Rate": 1.5689320548843882e-05, "Full-finetune/Loss": 0.8119845390319824, "Full-finetune/Loss (Raw)": 0.8545520901679993, "Full-finetune/Step": 1705, "Full-finetune/Step Time": 6.773780835792422} {"Full-finetune/Learning Rate": 1.5684196309780495e-05, "Full-finetune/Loss": 0.812136709690094, "Full-finetune/Loss (Raw)": 0.9120153784751892, "Full-finetune/Step": 1706, "Full-finetune/Step Time": 6.765887854620814} {"Full-finetune/Learning Rate": 1.5679069864791493e-05, "Full-finetune/Loss": 0.8128618001937866, "Full-finetune/Loss (Raw)": 0.8660169839859009, "Full-finetune/Step": 1707, "Full-finetune/Step Time": 6.784067822620273} {"Full-finetune/Learning Rate": 1.567394121586635e-05, "Full-finetune/Loss": 0.8143072724342346, "Full-finetune/Loss (Raw)": 0.8719618320465088, "Full-finetune/Step": 1708, "Full-finetune/Step Time": 6.765002826228738} {"Full-finetune/Learning Rate": 1.56688103649954e-05, "Full-finetune/Loss": 0.8160675168037415, "Full-finetune/Loss (Raw)": 0.9821602702140808, "Full-finetune/Step": 1709, "Full-finetune/Step Time": 6.762662289664149} {"Full-finetune/Learning Rate": 1.566367731416982e-05, "Full-finetune/Loss": 0.816011905670166, "Full-finetune/Loss (Raw)": 0.8325003981590271, "Full-finetune/Step": 1710, "Full-finetune/Step Time": 6.772091928869486} {"Full-finetune/Learning Rate": 1.5658542065381644e-05, "Full-finetune/Loss": 0.8157261610031128, "Full-finetune/Loss (Raw)": 0.8485248684883118, "Full-finetune/Step": 1711, "Full-finetune/Step Time": 6.775890884920955} {"Full-finetune/Learning Rate": 1.565340462062377e-05, "Full-finetune/Loss": 0.8169148564338684, "Full-finetune/Loss (Raw)": 0.9975535869598389, "Full-finetune/Step": 1712, "Full-finetune/Step Time": 6.776879521086812} {"Full-finetune/Learning Rate": 1.5648264981889936e-05, "Full-finetune/Loss": 0.8163845539093018, "Full-finetune/Loss (Raw)": 0.8848451972007751, "Full-finetune/Step": 1713, "Full-finetune/Step Time": 6.776612335816026} {"Full-finetune/Learning Rate": 1.5643123151174736e-05, "Full-finetune/Loss": 0.8151493668556213, "Full-finetune/Loss (Raw)": 0.900200366973877, "Full-finetune/Step": 1714, "Full-finetune/Step Time": 6.743210861459374} {"Full-finetune/Learning Rate": 1.5637979130473612e-05, "Full-finetune/Loss": 0.8126943111419678, "Full-finetune/Loss (Raw)": 0.6362054347991943, "Full-finetune/Step": 1715, "Full-finetune/Step Time": 6.751963106915355} {"Full-finetune/Learning Rate": 1.563283292178286e-05, "Full-finetune/Loss": 0.8129646182060242, "Full-finetune/Loss (Raw)": 0.8736148476600647, "Full-finetune/Step": 1716, "Full-finetune/Step Time": 6.740349819883704} {"Full-finetune/Learning Rate": 1.5627684527099624e-05, "Full-finetune/Loss": 0.8132549524307251, "Full-finetune/Loss (Raw)": 0.8807249069213867, "Full-finetune/Step": 1717, "Full-finetune/Step Time": 6.741074360907078} {"Full-finetune/Learning Rate": 1.5622533948421895e-05, "Full-finetune/Loss": 0.8131532669067383, "Full-finetune/Loss (Raw)": 0.9258261919021606, "Full-finetune/Step": 1718, "Full-finetune/Step Time": 6.731953889131546} {"Full-finetune/Learning Rate": 1.5617381187748507e-05, "Full-finetune/Loss": 0.8145782351493835, "Full-finetune/Loss (Raw)": 0.8095495104789734, "Full-finetune/Step": 1719, "Full-finetune/Step Time": 6.719327161088586} {"Full-finetune/Learning Rate": 1.5612226247079155e-05, "Full-finetune/Loss": 0.8166651725769043, "Full-finetune/Loss (Raw)": 0.9668755531311035, "Full-finetune/Step": 1720, "Full-finetune/Step Time": 6.70798397436738} {"Full-finetune/Learning Rate": 1.5607069128414366e-05, "Full-finetune/Loss": 0.8177288770675659, "Full-finetune/Loss (Raw)": 0.9097123742103577, "Full-finetune/Step": 1721, "Full-finetune/Step Time": 6.709287455305457} {"Full-finetune/Learning Rate": 1.5601909833755514e-05, "Full-finetune/Loss": 0.817934513092041, "Full-finetune/Loss (Raw)": 0.8573261499404907, "Full-finetune/Step": 1722, "Full-finetune/Step Time": 6.700767625123262} {"Full-finetune/Learning Rate": 1.559674836510482e-05, "Full-finetune/Loss": 0.8189423084259033, "Full-finetune/Loss (Raw)": 0.8331863284111023, "Full-finetune/Step": 1723, "Full-finetune/Step Time": 6.716763189062476} {"Full-finetune/Learning Rate": 1.5591584724465363e-05, "Full-finetune/Loss": 0.8184952735900879, "Full-finetune/Loss (Raw)": 0.7821431756019592, "Full-finetune/Step": 1724, "Full-finetune/Step Time": 6.709733599796891} {"Full-finetune/Learning Rate": 1.5586418913841034e-05, "Full-finetune/Loss": 0.8193766474723816, "Full-finetune/Loss (Raw)": 0.837724506855011, "Full-finetune/Step": 1725, "Full-finetune/Step Time": 6.689925834536552} {"Full-finetune/Learning Rate": 1.558125093523659e-05, "Full-finetune/Loss": 0.8190529942512512, "Full-finetune/Loss (Raw)": 0.7709349393844604, "Full-finetune/Step": 1726, "Full-finetune/Step Time": 6.683506986126304} {"Full-finetune/Learning Rate": 1.5576080790657626e-05, "Full-finetune/Loss": 0.8226817846298218, "Full-finetune/Loss (Raw)": 0.8525307178497314, "Full-finetune/Step": 1727, "Full-finetune/Step Time": 6.682178374379873} {"Full-finetune/Learning Rate": 1.5570908482110566e-05, "Full-finetune/Loss": 0.8229801058769226, "Full-finetune/Loss (Raw)": 0.9530583024024963, "Full-finetune/Step": 1728, "Full-finetune/Step Time": 6.677688574418426} {"Full-finetune/Learning Rate": 1.5565734011602693e-05, "Full-finetune/Loss": 0.8266686201095581, "Full-finetune/Loss (Raw)": 0.8743672370910645, "Full-finetune/Step": 1729, "Full-finetune/Step Time": 6.662657987326384} {"Full-finetune/Learning Rate": 1.556055738114211e-05, "Full-finetune/Loss": 0.8250318169593811, "Full-finetune/Loss (Raw)": 0.7562249302864075, "Full-finetune/Step": 1730, "Full-finetune/Step Time": 6.653145521879196} {"Full-finetune/Learning Rate": 1.555537859273777e-05, "Full-finetune/Loss": 0.8257678747177124, "Full-finetune/Loss (Raw)": 1.0306050777435303, "Full-finetune/Step": 1731, "Full-finetune/Step Time": 6.650418201461434} {"Full-finetune/Learning Rate": 1.5550197648399453e-05, "Full-finetune/Loss": 0.826799750328064, "Full-finetune/Loss (Raw)": 0.9359254837036133, "Full-finetune/Step": 1732, "Full-finetune/Step Time": 6.664488537237048} {"Full-finetune/Learning Rate": 1.554501455013779e-05, "Full-finetune/Loss": 0.828391969203949, "Full-finetune/Loss (Raw)": 0.8725526332855225, "Full-finetune/Step": 1733, "Full-finetune/Step Time": 6.656658077612519} {"Full-finetune/Learning Rate": 1.5539829299964236e-05, "Full-finetune/Loss": 0.8304914236068726, "Full-finetune/Loss (Raw)": 0.9004829525947571, "Full-finetune/Step": 1734, "Full-finetune/Step Time": 6.640826029703021} {"Full-finetune/Learning Rate": 1.5534641899891084e-05, "Full-finetune/Loss": 0.8305341601371765, "Full-finetune/Loss (Raw)": 0.8362127542495728, "Full-finetune/Step": 1735, "Full-finetune/Step Time": 6.640766479074955} {"Full-finetune/Learning Rate": 1.5529452351931473e-05, "Full-finetune/Loss": 0.8341965675354004, "Full-finetune/Loss (Raw)": 0.8687544465065002, "Full-finetune/Step": 1736, "Full-finetune/Step Time": 6.623119492083788} {"Full-finetune/Learning Rate": 1.5524260658099352e-05, "Full-finetune/Loss": 0.8334851264953613, "Full-finetune/Loss (Raw)": 0.7908115983009338, "Full-finetune/Step": 1737, "Full-finetune/Step Time": 6.619839554652572} {"Full-finetune/Learning Rate": 1.551906682040952e-05, "Full-finetune/Loss": 0.8294757008552551, "Full-finetune/Loss (Raw)": 0.3625481426715851, "Full-finetune/Step": 1738, "Full-finetune/Step Time": 6.634458128362894} {"Full-finetune/Learning Rate": 1.5513870840877608e-05, "Full-finetune/Loss": 0.8301557898521423, "Full-finetune/Loss (Raw)": 0.837256669998169, "Full-finetune/Step": 1739, "Full-finetune/Step Time": 6.614242421463132} {"Full-finetune/Learning Rate": 1.5508672721520076e-05, "Full-finetune/Loss": 0.8297011852264404, "Full-finetune/Loss (Raw)": 0.7626436352729797, "Full-finetune/Step": 1740, "Full-finetune/Step Time": 6.628941360861063} {"Full-finetune/Learning Rate": 1.5503472464354207e-05, "Full-finetune/Loss": 0.832146942615509, "Full-finetune/Loss (Raw)": 1.044055700302124, "Full-finetune/Step": 1741, "Full-finetune/Step Time": 6.61970860697329} {"Full-finetune/Learning Rate": 1.5498270071398116e-05, "Full-finetune/Loss": 0.8304635882377625, "Full-finetune/Loss (Raw)": 0.6823378205299377, "Full-finetune/Step": 1742, "Full-finetune/Step Time": 6.627704199403524} {"Full-finetune/Learning Rate": 1.549306554467076e-05, "Full-finetune/Loss": 0.8296955227851868, "Full-finetune/Loss (Raw)": 0.856109619140625, "Full-finetune/Step": 1743, "Full-finetune/Step Time": 6.64304349757731} {"Full-finetune/Learning Rate": 1.5487858886191903e-05, "Full-finetune/Loss": 0.8295474052429199, "Full-finetune/Loss (Raw)": 0.8326679468154907, "Full-finetune/Step": 1744, "Full-finetune/Step Time": 6.643908394500613} {"Full-finetune/Learning Rate": 1.548265009798216e-05, "Full-finetune/Loss": 0.830083966255188, "Full-finetune/Loss (Raw)": 0.8810909986495972, "Full-finetune/Step": 1745, "Full-finetune/Step Time": 6.640709903091192} {"Full-finetune/Learning Rate": 1.547743918206295e-05, "Full-finetune/Loss": 0.8312848806381226, "Full-finetune/Loss (Raw)": 0.9829297661781311, "Full-finetune/Step": 1746, "Full-finetune/Step Time": 6.640101559460163} {"Full-finetune/Learning Rate": 1.5472226140456537e-05, "Full-finetune/Loss": 0.8289136290550232, "Full-finetune/Loss (Raw)": 0.44783729314804077, "Full-finetune/Step": 1747, "Full-finetune/Step Time": 6.656095584854484} {"Full-finetune/Learning Rate": 1.546701097518599e-05, "Full-finetune/Loss": 0.8256828784942627, "Full-finetune/Loss (Raw)": 0.6776122450828552, "Full-finetune/Step": 1748, "Full-finetune/Step Time": 6.668021423742175} {"Full-finetune/Learning Rate": 1.546179368827522e-05, "Full-finetune/Loss": 0.8264521360397339, "Full-finetune/Loss (Raw)": 0.8818143010139465, "Full-finetune/Step": 1749, "Full-finetune/Step Time": 6.666867190971971} {"Full-finetune/Learning Rate": 1.545657428174895e-05, "Full-finetune/Loss": 0.8264541029930115, "Full-finetune/Loss (Raw)": 0.9682141542434692, "Full-finetune/Step": 1750, "Full-finetune/Step Time": 6.684694526717067} {"Full-finetune/Learning Rate": 1.5451352757632734e-05, "Full-finetune/Loss": 0.8252111077308655, "Full-finetune/Loss (Raw)": 0.9033558368682861, "Full-finetune/Step": 1751, "Full-finetune/Step Time": 6.688432555645704} {"Full-finetune/Learning Rate": 1.544612911795294e-05, "Full-finetune/Loss": 0.8262182474136353, "Full-finetune/Loss (Raw)": 1.0108739137649536, "Full-finetune/Step": 1752, "Full-finetune/Step Time": 6.687939310446382} {"Full-finetune/Learning Rate": 1.5440903364736763e-05, "Full-finetune/Loss": 0.8261058330535889, "Full-finetune/Loss (Raw)": 0.8641231656074524, "Full-finetune/Step": 1753, "Full-finetune/Step Time": 6.685418574139476} {"Full-finetune/Learning Rate": 1.5435675500012212e-05, "Full-finetune/Loss": 0.8262591361999512, "Full-finetune/Loss (Raw)": 0.9206870198249817, "Full-finetune/Step": 1754, "Full-finetune/Step Time": 6.704876270145178} {"Full-finetune/Learning Rate": 1.5430445525808123e-05, "Full-finetune/Loss": 0.8256787061691284, "Full-finetune/Loss (Raw)": 0.8711175322532654, "Full-finetune/Step": 1755, "Full-finetune/Step Time": 6.711563423275948} {"Full-finetune/Learning Rate": 1.5425213444154152e-05, "Full-finetune/Loss": 0.8269820213317871, "Full-finetune/Loss (Raw)": 0.9878531098365784, "Full-finetune/Step": 1756, "Full-finetune/Step Time": 6.714691203087568} {"Full-finetune/Learning Rate": 1.5419979257080755e-05, "Full-finetune/Loss": 0.8236823081970215, "Full-finetune/Loss (Raw)": 0.3691045343875885, "Full-finetune/Step": 1757, "Full-finetune/Step Time": 6.732892131432891} {"Full-finetune/Learning Rate": 1.541474296661923e-05, "Full-finetune/Loss": 0.8227925300598145, "Full-finetune/Loss (Raw)": 0.7593801021575928, "Full-finetune/Step": 1758, "Full-finetune/Step Time": 6.738583104684949} {"Full-finetune/Learning Rate": 1.5409504574801667e-05, "Full-finetune/Loss": 0.8235195875167847, "Full-finetune/Loss (Raw)": 0.9496011137962341, "Full-finetune/Step": 1759, "Full-finetune/Step Time": 6.743578566238284} {"Full-finetune/Learning Rate": 1.5404264083660993e-05, "Full-finetune/Loss": 0.8247500061988831, "Full-finetune/Loss (Raw)": 0.9738264083862305, "Full-finetune/Step": 1760, "Full-finetune/Step Time": 6.752225033938885} {"Full-finetune/Learning Rate": 1.539902149523094e-05, "Full-finetune/Loss": 0.8244574666023254, "Full-finetune/Loss (Raw)": 0.7563026547431946, "Full-finetune/Step": 1761, "Full-finetune/Step Time": 6.767566591501236} {"Full-finetune/Learning Rate": 1.5393776811546047e-05, "Full-finetune/Loss": 0.8236474394798279, "Full-finetune/Loss (Raw)": 0.7668051719665527, "Full-finetune/Step": 1762, "Full-finetune/Step Time": 6.777854718267918} {"Full-finetune/Learning Rate": 1.538853003464168e-05, "Full-finetune/Loss": 0.8253058195114136, "Full-finetune/Loss (Raw)": 1.0010489225387573, "Full-finetune/Step": 1763, "Full-finetune/Step Time": 6.771976813673973} {"Full-finetune/Learning Rate": 1.538328116655401e-05, "Full-finetune/Loss": 0.8256460428237915, "Full-finetune/Loss (Raw)": 0.942623496055603, "Full-finetune/Step": 1764, "Full-finetune/Step Time": 6.791783686727285} {"Full-finetune/Learning Rate": 1.5378030209320018e-05, "Full-finetune/Loss": 0.8255940675735474, "Full-finetune/Loss (Raw)": 0.7901788949966431, "Full-finetune/Step": 1765, "Full-finetune/Step Time": 6.804167199879885} {"Full-finetune/Learning Rate": 1.53727771649775e-05, "Full-finetune/Loss": 0.8264835476875305, "Full-finetune/Loss (Raw)": 0.9163123965263367, "Full-finetune/Step": 1766, "Full-finetune/Step Time": 6.816191868856549} {"Full-finetune/Learning Rate": 1.5367522035565055e-05, "Full-finetune/Loss": 0.8265359401702881, "Full-finetune/Loss (Raw)": 0.7485436201095581, "Full-finetune/Step": 1767, "Full-finetune/Step Time": 6.816155105829239} {"Full-finetune/Learning Rate": 1.5362264823122107e-05, "Full-finetune/Loss": 0.8245948553085327, "Full-finetune/Loss (Raw)": 0.7178560495376587, "Full-finetune/Step": 1768, "Full-finetune/Step Time": 6.816480664536357} {"Full-finetune/Learning Rate": 1.5357005529688868e-05, "Full-finetune/Loss": 0.8230808973312378, "Full-finetune/Loss (Raw)": 0.7087157964706421, "Full-finetune/Step": 1769, "Full-finetune/Step Time": 6.830157097429037} {"Full-finetune/Learning Rate": 1.5351744157306366e-05, "Full-finetune/Loss": 0.8235416412353516, "Full-finetune/Loss (Raw)": 0.910455048084259, "Full-finetune/Step": 1770, "Full-finetune/Step Time": 6.830902958288789} {"Full-finetune/Learning Rate": 1.5346480708016445e-05, "Full-finetune/Loss": 0.8222572803497314, "Full-finetune/Loss (Raw)": 0.8132684230804443, "Full-finetune/Step": 1771, "Full-finetune/Step Time": 6.829938791692257} {"Full-finetune/Learning Rate": 1.5341215183861738e-05, "Full-finetune/Loss": 0.8237410187721252, "Full-finetune/Loss (Raw)": 0.9856051206588745, "Full-finetune/Step": 1772, "Full-finetune/Step Time": 6.79369748570025} {"Full-finetune/Learning Rate": 1.53359475868857e-05, "Full-finetune/Loss": 0.8232640027999878, "Full-finetune/Loss (Raw)": 0.7898086905479431, "Full-finetune/Step": 1773, "Full-finetune/Step Time": 6.800733113661408} {"Full-finetune/Learning Rate": 1.5330677919132577e-05, "Full-finetune/Loss": 0.821885347366333, "Full-finetune/Loss (Raw)": 0.742938756942749, "Full-finetune/Step": 1774, "Full-finetune/Step Time": 6.791250705718994} {"Full-finetune/Learning Rate": 1.5325406182647426e-05, "Full-finetune/Loss": 0.8232274055480957, "Full-finetune/Loss (Raw)": 0.9503974914550781, "Full-finetune/Step": 1775, "Full-finetune/Step Time": 6.769131734967232} {"Full-finetune/Learning Rate": 1.5320132379476107e-05, "Full-finetune/Loss": 0.8225948810577393, "Full-finetune/Loss (Raw)": 0.8541988134384155, "Full-finetune/Step": 1776, "Full-finetune/Step Time": 6.766299314796925} {"Full-finetune/Learning Rate": 1.531485651166528e-05, "Full-finetune/Loss": 0.822986364364624, "Full-finetune/Loss (Raw)": 0.8756042122840881, "Full-finetune/Step": 1777, "Full-finetune/Step Time": 6.760688291862607} {"Full-finetune/Learning Rate": 1.5309578581262404e-05, "Full-finetune/Loss": 0.8224623203277588, "Full-finetune/Loss (Raw)": 0.8334149718284607, "Full-finetune/Step": 1778, "Full-finetune/Step Time": 6.762091122567654} {"Full-finetune/Learning Rate": 1.530429859031574e-05, "Full-finetune/Loss": 0.8248535990715027, "Full-finetune/Loss (Raw)": 1.0431714057922363, "Full-finetune/Step": 1779, "Full-finetune/Step Time": 6.745994640514255} {"Full-finetune/Learning Rate": 1.5299016540874346e-05, "Full-finetune/Loss": 0.8248329758644104, "Full-finetune/Loss (Raw)": 0.8188223838806152, "Full-finetune/Step": 1780, "Full-finetune/Step Time": 6.73628837428987} {"Full-finetune/Learning Rate": 1.5293732434988095e-05, "Full-finetune/Loss": 0.8236094117164612, "Full-finetune/Loss (Raw)": 0.6777928471565247, "Full-finetune/Step": 1781, "Full-finetune/Step Time": 6.737191496416926} {"Full-finetune/Learning Rate": 1.5288446274707635e-05, "Full-finetune/Loss": 0.8233701586723328, "Full-finetune/Loss (Raw)": 0.8671443462371826, "Full-finetune/Step": 1782, "Full-finetune/Step Time": 6.724593169987202} {"Full-finetune/Learning Rate": 1.5283158062084423e-05, "Full-finetune/Loss": 0.8259266018867493, "Full-finetune/Loss (Raw)": 0.730061411857605, "Full-finetune/Step": 1783, "Full-finetune/Step Time": 6.706885289400816} {"Full-finetune/Learning Rate": 1.5277867799170708e-05, "Full-finetune/Loss": 0.826871395111084, "Full-finetune/Loss (Raw)": 1.009244680404663, "Full-finetune/Step": 1784, "Full-finetune/Step Time": 6.7098871525377035} {"Full-finetune/Learning Rate": 1.527257548801955e-05, "Full-finetune/Loss": 0.8253507614135742, "Full-finetune/Loss (Raw)": 0.6174997687339783, "Full-finetune/Step": 1785, "Full-finetune/Step Time": 6.722305068746209} {"Full-finetune/Learning Rate": 1.526728113068478e-05, "Full-finetune/Loss": 0.8250686526298523, "Full-finetune/Loss (Raw)": 0.8474363088607788, "Full-finetune/Step": 1786, "Full-finetune/Step Time": 6.7195024862885475} {"Full-finetune/Learning Rate": 1.526198472922104e-05, "Full-finetune/Loss": 0.8243978023529053, "Full-finetune/Loss (Raw)": 0.7026638388633728, "Full-finetune/Step": 1787, "Full-finetune/Step Time": 6.724312398582697} {"Full-finetune/Learning Rate": 1.5256686285683753e-05, "Full-finetune/Loss": 0.8236143589019775, "Full-finetune/Loss (Raw)": 0.7866615056991577, "Full-finetune/Step": 1788, "Full-finetune/Step Time": 6.718393303453922} {"Full-finetune/Learning Rate": 1.5251385802129149e-05, "Full-finetune/Loss": 0.8194957971572876, "Full-finetune/Loss (Raw)": 0.3115514814853668, "Full-finetune/Step": 1789, "Full-finetune/Step Time": 6.742091700434685} {"Full-finetune/Learning Rate": 1.5246083280614241e-05, "Full-finetune/Loss": 0.8191548585891724, "Full-finetune/Loss (Raw)": 0.9032037258148193, "Full-finetune/Step": 1790, "Full-finetune/Step Time": 6.745097581297159} {"Full-finetune/Learning Rate": 1.5240778723196833e-05, "Full-finetune/Loss": 0.8221844434738159, "Full-finetune/Loss (Raw)": 0.9771336913108826, "Full-finetune/Step": 1791, "Full-finetune/Step Time": 6.736662955954671} {"Full-finetune/Learning Rate": 1.523547213193552e-05, "Full-finetune/Loss": 0.8241166472434998, "Full-finetune/Loss (Raw)": 0.9199837446212769, "Full-finetune/Step": 1792, "Full-finetune/Step Time": 6.737173261120915} {"Full-finetune/Learning Rate": 1.5230163508889686e-05, "Full-finetune/Loss": 0.8219538927078247, "Full-finetune/Loss (Raw)": 0.714982807636261, "Full-finetune/Step": 1793, "Full-finetune/Step Time": 6.750566331669688} {"Full-finetune/Learning Rate": 1.5224852856119503e-05, "Full-finetune/Loss": 0.8228563070297241, "Full-finetune/Loss (Raw)": 0.9413571357727051, "Full-finetune/Step": 1794, "Full-finetune/Step Time": 6.755700236186385} {"Full-finetune/Learning Rate": 1.5219540175685938e-05, "Full-finetune/Loss": 0.8188621997833252, "Full-finetune/Loss (Raw)": 0.37973034381866455, "Full-finetune/Step": 1795, "Full-finetune/Step Time": 6.774124847725034} {"Full-finetune/Learning Rate": 1.5214225469650727e-05, "Full-finetune/Loss": 0.8181018233299255, "Full-finetune/Loss (Raw)": 0.8324791789054871, "Full-finetune/Step": 1796, "Full-finetune/Step Time": 6.767584940418601} {"Full-finetune/Learning Rate": 1.5208908740076416e-05, "Full-finetune/Loss": 0.8178747296333313, "Full-finetune/Loss (Raw)": 0.844433069229126, "Full-finetune/Step": 1797, "Full-finetune/Step Time": 6.766994491219521} {"Full-finetune/Learning Rate": 1.5203589989026318e-05, "Full-finetune/Loss": 0.8173475861549377, "Full-finetune/Loss (Raw)": 0.8700343370437622, "Full-finetune/Step": 1798, "Full-finetune/Step Time": 6.769562749192119} {"Full-finetune/Learning Rate": 1.5198269218564535e-05, "Full-finetune/Loss": 0.8173889517784119, "Full-finetune/Loss (Raw)": 0.814288854598999, "Full-finetune/Step": 1799, "Full-finetune/Step Time": 6.761455774307251} {"Full-finetune/Learning Rate": 1.5192946430755955e-05, "Full-finetune/Loss": 0.8171601891517639, "Full-finetune/Loss (Raw)": 0.8147624731063843, "Full-finetune/Step": 1800, "Full-finetune/Step Time": 6.75430597923696} {"Full-finetune/Learning Rate": 1.518762162766625e-05, "Full-finetune/Loss": 0.8186643123626709, "Full-finetune/Loss (Raw)": 0.9023793935775757, "Full-finetune/Step": 1801, "Full-finetune/Step Time": 6.758183965459466} {"Full-finetune/Learning Rate": 1.5182294811361871e-05, "Full-finetune/Loss": 0.8202864527702332, "Full-finetune/Loss (Raw)": 0.782418429851532, "Full-finetune/Step": 1802, "Full-finetune/Step Time": 6.760173885151744} {"Full-finetune/Learning Rate": 1.5176965983910053e-05, "Full-finetune/Loss": 0.8214547634124756, "Full-finetune/Loss (Raw)": 0.8363715410232544, "Full-finetune/Step": 1803, "Full-finetune/Step Time": 6.772204115986824} {"Full-finetune/Learning Rate": 1.5171635147378804e-05, "Full-finetune/Loss": 0.8211135864257812, "Full-finetune/Loss (Raw)": 0.8864592909812927, "Full-finetune/Step": 1804, "Full-finetune/Step Time": 6.775591541081667} {"Full-finetune/Learning Rate": 1.5166302303836928e-05, "Full-finetune/Loss": 0.8179017305374146, "Full-finetune/Loss (Raw)": 0.5419982671737671, "Full-finetune/Step": 1805, "Full-finetune/Step Time": 6.7937306240201} {"Full-finetune/Learning Rate": 1.5160967455353985e-05, "Full-finetune/Loss": 0.8176339864730835, "Full-finetune/Loss (Raw)": 0.7479737997055054, "Full-finetune/Step": 1806, "Full-finetune/Step Time": 6.805284928530455} {"Full-finetune/Learning Rate": 1.5155630604000338e-05, "Full-finetune/Loss": 0.8214173316955566, "Full-finetune/Loss (Raw)": 0.8925871849060059, "Full-finetune/Step": 1807, "Full-finetune/Step Time": 6.787236671894789} {"Full-finetune/Learning Rate": 1.5150291751847106e-05, "Full-finetune/Loss": 0.8210211992263794, "Full-finetune/Loss (Raw)": 0.8052027821540833, "Full-finetune/Step": 1808, "Full-finetune/Step Time": 6.775808617472649} {"Full-finetune/Learning Rate": 1.5144950900966198e-05, "Full-finetune/Loss": 0.8210645318031311, "Full-finetune/Loss (Raw)": 0.8758628964424133, "Full-finetune/Step": 1809, "Full-finetune/Step Time": 6.774405153468251} {"Full-finetune/Learning Rate": 1.5139608053430288e-05, "Full-finetune/Loss": 0.81926429271698, "Full-finetune/Loss (Raw)": 0.8011642694473267, "Full-finetune/Step": 1810, "Full-finetune/Step Time": 6.7756919134408236} {"Full-finetune/Learning Rate": 1.5134263211312837e-05, "Full-finetune/Loss": 0.8204374313354492, "Full-finetune/Loss (Raw)": 0.845628559589386, "Full-finetune/Step": 1811, "Full-finetune/Step Time": 6.769571494311094} {"Full-finetune/Learning Rate": 1.5128916376688071e-05, "Full-finetune/Loss": 0.8196540474891663, "Full-finetune/Loss (Raw)": 0.7317338585853577, "Full-finetune/Step": 1812, "Full-finetune/Step Time": 6.775700936093926} {"Full-finetune/Learning Rate": 1.5123567551630993e-05, "Full-finetune/Loss": 0.8201578855514526, "Full-finetune/Loss (Raw)": 0.821079671382904, "Full-finetune/Step": 1813, "Full-finetune/Step Time": 6.765465764328837} {"Full-finetune/Learning Rate": 1.511821673821738e-05, "Full-finetune/Loss": 0.8220927715301514, "Full-finetune/Loss (Raw)": 0.866487979888916, "Full-finetune/Step": 1814, "Full-finetune/Step Time": 6.7533400524407625} {"Full-finetune/Learning Rate": 1.5112863938523774e-05, "Full-finetune/Loss": 0.8213927745819092, "Full-finetune/Loss (Raw)": 0.8207976818084717, "Full-finetune/Step": 1815, "Full-finetune/Step Time": 6.7629041243344545} {"Full-finetune/Learning Rate": 1.5107509154627491e-05, "Full-finetune/Loss": 0.822263777256012, "Full-finetune/Loss (Raw)": 0.9385806322097778, "Full-finetune/Step": 1816, "Full-finetune/Step Time": 6.76948894187808} {"Full-finetune/Learning Rate": 1.5102152388606625e-05, "Full-finetune/Loss": 0.8263503313064575, "Full-finetune/Loss (Raw)": 0.8450963497161865, "Full-finetune/Step": 1817, "Full-finetune/Step Time": 6.752772590145469} {"Full-finetune/Learning Rate": 1.5096793642540027e-05, "Full-finetune/Loss": 0.8249643445014954, "Full-finetune/Loss (Raw)": 0.7755602598190308, "Full-finetune/Step": 1818, "Full-finetune/Step Time": 6.752046462148428} {"Full-finetune/Learning Rate": 1.5091432918507322e-05, "Full-finetune/Loss": 0.8252159953117371, "Full-finetune/Loss (Raw)": 0.9051228165626526, "Full-finetune/Step": 1819, "Full-finetune/Step Time": 6.7496172320097685} {"Full-finetune/Learning Rate": 1.5086070218588904e-05, "Full-finetune/Loss": 0.8252346515655518, "Full-finetune/Loss (Raw)": 0.7443774342536926, "Full-finetune/Step": 1820, "Full-finetune/Step Time": 6.727307237684727} {"Full-finetune/Learning Rate": 1.5080705544865932e-05, "Full-finetune/Loss": 0.8269270658493042, "Full-finetune/Loss (Raw)": 0.8695362210273743, "Full-finetune/Step": 1821, "Full-finetune/Step Time": 6.712767982855439} {"Full-finetune/Learning Rate": 1.5075338899420326e-05, "Full-finetune/Loss": 0.8307209014892578, "Full-finetune/Loss (Raw)": 1.0737078189849854, "Full-finetune/Step": 1822, "Full-finetune/Step Time": 6.703989375382662} {"Full-finetune/Learning Rate": 1.5069970284334785e-05, "Full-finetune/Loss": 0.8320659399032593, "Full-finetune/Loss (Raw)": 0.8543918132781982, "Full-finetune/Step": 1823, "Full-finetune/Step Time": 6.6959994323551655} {"Full-finetune/Learning Rate": 1.506459970169276e-05, "Full-finetune/Loss": 0.8304445743560791, "Full-finetune/Loss (Raw)": 0.640837550163269, "Full-finetune/Step": 1824, "Full-finetune/Step Time": 6.698934160172939} {"Full-finetune/Learning Rate": 1.5059227153578466e-05, "Full-finetune/Loss": 0.8298363089561462, "Full-finetune/Loss (Raw)": 0.844704806804657, "Full-finetune/Step": 1825, "Full-finetune/Step Time": 6.700637474656105} {"Full-finetune/Learning Rate": 1.505385264207689e-05, "Full-finetune/Loss": 0.8302221298217773, "Full-finetune/Loss (Raw)": 0.8837115168571472, "Full-finetune/Step": 1826, "Full-finetune/Step Time": 6.701239096000791} {"Full-finetune/Learning Rate": 1.5048476169273767e-05, "Full-finetune/Loss": 0.8295462131500244, "Full-finetune/Loss (Raw)": 0.8096765875816345, "Full-finetune/Step": 1827, "Full-finetune/Step Time": 6.7335805129259825} {"Full-finetune/Learning Rate": 1.5043097737255607e-05, "Full-finetune/Loss": 0.8295369744300842, "Full-finetune/Loss (Raw)": 0.8678081631660461, "Full-finetune/Step": 1828, "Full-finetune/Step Time": 6.721706198528409} {"Full-finetune/Learning Rate": 1.5037717348109677e-05, "Full-finetune/Loss": 0.8301119804382324, "Full-finetune/Loss (Raw)": 1.0321085453033447, "Full-finetune/Step": 1829, "Full-finetune/Step Time": 6.735766958445311} {"Full-finetune/Learning Rate": 1.5032335003923994e-05, "Full-finetune/Loss": 0.8304334878921509, "Full-finetune/Loss (Raw)": 0.9681885838508606, "Full-finetune/Step": 1830, "Full-finetune/Step Time": 6.7473305724561214} {"Full-finetune/Learning Rate": 1.5026950706787346e-05, "Full-finetune/Loss": 0.8317010998725891, "Full-finetune/Loss (Raw)": 0.8261305093765259, "Full-finetune/Step": 1831, "Full-finetune/Step Time": 6.73308714851737} {"Full-finetune/Learning Rate": 1.502156445878927e-05, "Full-finetune/Loss": 0.83650803565979, "Full-finetune/Loss (Raw)": 0.9741044640541077, "Full-finetune/Step": 1832, "Full-finetune/Step Time": 6.719799438491464} {"Full-finetune/Learning Rate": 1.5016176262020069e-05, "Full-finetune/Loss": 0.8355457782745361, "Full-finetune/Loss (Raw)": 0.7313796877861023, "Full-finetune/Step": 1833, "Full-finetune/Step Time": 6.726162992417812} {"Full-finetune/Learning Rate": 1.5010786118570791e-05, "Full-finetune/Loss": 0.8359943628311157, "Full-finetune/Loss (Raw)": 0.969438910484314, "Full-finetune/Step": 1834, "Full-finetune/Step Time": 6.724943766370416} {"Full-finetune/Learning Rate": 1.5005394030533246e-05, "Full-finetune/Loss": 0.8362340927124023, "Full-finetune/Loss (Raw)": 0.8966982960700989, "Full-finetune/Step": 1835, "Full-finetune/Step Time": 6.713106555864215} {"Full-finetune/Learning Rate": 1.5000000000000002e-05, "Full-finetune/Loss": 0.8363772630691528, "Full-finetune/Loss (Raw)": 0.890293300151825, "Full-finetune/Step": 1836, "Full-finetune/Step Time": 6.708632554858923} {"Full-finetune/Learning Rate": 1.4994604029064372e-05, "Full-finetune/Loss": 0.835159957408905, "Full-finetune/Loss (Raw)": 0.8263400793075562, "Full-finetune/Step": 1837, "Full-finetune/Step Time": 6.711332187056541} {"Full-finetune/Learning Rate": 1.4989206119820432e-05, "Full-finetune/Loss": 0.8352749347686768, "Full-finetune/Loss (Raw)": 0.8472154140472412, "Full-finetune/Step": 1838, "Full-finetune/Step Time": 6.704170169308782} {"Full-finetune/Learning Rate": 1.4983806274363005e-05, "Full-finetune/Loss": 0.835626482963562, "Full-finetune/Loss (Raw)": 0.8935219049453735, "Full-finetune/Step": 1839, "Full-finetune/Step Time": 6.7067653723061085} {"Full-finetune/Learning Rate": 1.497840449478766e-05, "Full-finetune/Loss": 0.8342349529266357, "Full-finetune/Loss (Raw)": 0.8194437026977539, "Full-finetune/Step": 1840, "Full-finetune/Step Time": 6.716203534975648} {"Full-finetune/Learning Rate": 1.4973000783190728e-05, "Full-finetune/Loss": 0.833361029624939, "Full-finetune/Loss (Raw)": 0.7729821801185608, "Full-finetune/Step": 1841, "Full-finetune/Step Time": 6.719247527420521} {"Full-finetune/Learning Rate": 1.4967595141669278e-05, "Full-finetune/Loss": 0.8337294459342957, "Full-finetune/Loss (Raw)": 0.9473552107810974, "Full-finetune/Step": 1842, "Full-finetune/Step Time": 6.717620620504022} {"Full-finetune/Learning Rate": 1.4962187572321143e-05, "Full-finetune/Loss": 0.8353805541992188, "Full-finetune/Loss (Raw)": 0.8475534319877625, "Full-finetune/Step": 1843, "Full-finetune/Step Time": 6.722899112850428} {"Full-finetune/Learning Rate": 1.4956778077244884e-05, "Full-finetune/Loss": 0.8347967267036438, "Full-finetune/Loss (Raw)": 0.7988784313201904, "Full-finetune/Step": 1844, "Full-finetune/Step Time": 6.733357585966587} {"Full-finetune/Learning Rate": 1.4951366658539832e-05, "Full-finetune/Loss": 0.8338088989257812, "Full-finetune/Loss (Raw)": 0.7542802095413208, "Full-finetune/Step": 1845, "Full-finetune/Step Time": 6.727683551609516} {"Full-finetune/Learning Rate": 1.4945953318306044e-05, "Full-finetune/Loss": 0.8340973854064941, "Full-finetune/Loss (Raw)": 0.9627562165260315, "Full-finetune/Step": 1846, "Full-finetune/Step Time": 6.729208057746291} {"Full-finetune/Learning Rate": 1.4940538058644332e-05, "Full-finetune/Loss": 0.8346229791641235, "Full-finetune/Loss (Raw)": 0.8768249154090881, "Full-finetune/Step": 1847, "Full-finetune/Step Time": 6.737106237560511} {"Full-finetune/Learning Rate": 1.4935120881656263e-05, "Full-finetune/Loss": 0.8338754177093506, "Full-finetune/Loss (Raw)": 0.8711930513381958, "Full-finetune/Step": 1848, "Full-finetune/Step Time": 6.729439295828342} {"Full-finetune/Learning Rate": 1.4929701789444122e-05, "Full-finetune/Loss": 0.8342793583869934, "Full-finetune/Loss (Raw)": 0.9614102244377136, "Full-finetune/Step": 1849, "Full-finetune/Step Time": 6.723287748172879} {"Full-finetune/Learning Rate": 1.4924280784110965e-05, "Full-finetune/Loss": 0.8347854614257812, "Full-finetune/Loss (Raw)": 0.9221155047416687, "Full-finetune/Step": 1850, "Full-finetune/Step Time": 6.729012219235301} {"Full-finetune/Learning Rate": 1.4918857867760571e-05, "Full-finetune/Loss": 0.8332688808441162, "Full-finetune/Loss (Raw)": 0.6390593647956848, "Full-finetune/Step": 1851, "Full-finetune/Step Time": 6.732390271499753} {"Full-finetune/Learning Rate": 1.4913433042497473e-05, "Full-finetune/Loss": 0.8340997695922852, "Full-finetune/Loss (Raw)": 0.8884981274604797, "Full-finetune/Step": 1852, "Full-finetune/Step Time": 6.7335499022156} {"Full-finetune/Learning Rate": 1.4908006310426942e-05, "Full-finetune/Loss": 0.8345578908920288, "Full-finetune/Loss (Raw)": 0.8963612914085388, "Full-finetune/Step": 1853, "Full-finetune/Step Time": 6.73379042558372} {"Full-finetune/Learning Rate": 1.4902577673654977e-05, "Full-finetune/Loss": 0.8354816436767578, "Full-finetune/Loss (Raw)": 0.8891801238059998, "Full-finetune/Step": 1854, "Full-finetune/Step Time": 6.727050160989165} {"Full-finetune/Learning Rate": 1.4897147134288334e-05, "Full-finetune/Loss": 0.8358107209205627, "Full-finetune/Loss (Raw)": 0.8946467638015747, "Full-finetune/Step": 1855, "Full-finetune/Step Time": 6.723448392003775} {"Full-finetune/Learning Rate": 1.4891714694434495e-05, "Full-finetune/Loss": 0.836143970489502, "Full-finetune/Loss (Raw)": 0.9957088828086853, "Full-finetune/Step": 1856, "Full-finetune/Step Time": 6.72311794012785} {"Full-finetune/Learning Rate": 1.4886280356201688e-05, "Full-finetune/Loss": 0.8352034091949463, "Full-finetune/Loss (Raw)": 0.7539805173873901, "Full-finetune/Step": 1857, "Full-finetune/Step Time": 6.758042044937611} {"Full-finetune/Learning Rate": 1.4880844121698872e-05, "Full-finetune/Loss": 0.8358309864997864, "Full-finetune/Loss (Raw)": 0.8365582227706909, "Full-finetune/Step": 1858, "Full-finetune/Step Time": 6.7661947682499886} {"Full-finetune/Learning Rate": 1.487540599303574e-05, "Full-finetune/Loss": 0.8353067636489868, "Full-finetune/Loss (Raw)": 0.9635041952133179, "Full-finetune/Step": 1859, "Full-finetune/Step Time": 6.7605394218117} {"Full-finetune/Learning Rate": 1.4869965972322728e-05, "Full-finetune/Loss": 0.8335415124893188, "Full-finetune/Loss (Raw)": 0.7099727988243103, "Full-finetune/Step": 1860, "Full-finetune/Step Time": 6.748717423528433} {"Full-finetune/Learning Rate": 1.4864524061670999e-05, "Full-finetune/Loss": 0.8329717516899109, "Full-finetune/Loss (Raw)": 0.7996202111244202, "Full-finetune/Step": 1861, "Full-finetune/Step Time": 6.748872188851237} {"Full-finetune/Learning Rate": 1.4859080263192456e-05, "Full-finetune/Loss": 0.8320180177688599, "Full-finetune/Loss (Raw)": 0.7784111499786377, "Full-finetune/Step": 1862, "Full-finetune/Step Time": 6.73319286480546} {"Full-finetune/Learning Rate": 1.4853634578999728e-05, "Full-finetune/Loss": 0.8313033580780029, "Full-finetune/Loss (Raw)": 0.7447304725646973, "Full-finetune/Step": 1863, "Full-finetune/Step Time": 6.732803910970688} {"Full-finetune/Learning Rate": 1.4848187011206177e-05, "Full-finetune/Loss": 0.8304851055145264, "Full-finetune/Loss (Raw)": 0.7640246152877808, "Full-finetune/Step": 1864, "Full-finetune/Step Time": 6.736880091950297} {"Full-finetune/Learning Rate": 1.4842737561925901e-05, "Full-finetune/Loss": 0.8314285278320312, "Full-finetune/Loss (Raw)": 0.9115665555000305, "Full-finetune/Step": 1865, "Full-finetune/Step Time": 6.7404830157756805} {"Full-finetune/Learning Rate": 1.4837286233273722e-05, "Full-finetune/Loss": 0.8359707593917847, "Full-finetune/Loss (Raw)": 0.9439525008201599, "Full-finetune/Step": 1866, "Full-finetune/Step Time": 6.735468650236726} {"Full-finetune/Learning Rate": 1.48318330273652e-05, "Full-finetune/Loss": 0.8368701338768005, "Full-finetune/Loss (Raw)": 0.9523768424987793, "Full-finetune/Step": 1867, "Full-finetune/Step Time": 6.740753078833222} {"Full-finetune/Learning Rate": 1.4826377946316611e-05, "Full-finetune/Loss": 0.8367772698402405, "Full-finetune/Loss (Raw)": 0.7507562041282654, "Full-finetune/Step": 1868, "Full-finetune/Step Time": 6.742469221353531} {"Full-finetune/Learning Rate": 1.4820920992244964e-05, "Full-finetune/Loss": 0.8359032273292542, "Full-finetune/Loss (Raw)": 0.9321831464767456, "Full-finetune/Step": 1869, "Full-finetune/Step Time": 6.751685900613666} {"Full-finetune/Learning Rate": 1.481546216726801e-05, "Full-finetune/Loss": 0.8365288972854614, "Full-finetune/Loss (Raw)": 0.7624227404594421, "Full-finetune/Step": 1870, "Full-finetune/Step Time": 6.755123544484377} {"Full-finetune/Learning Rate": 1.4810001473504194e-05, "Full-finetune/Loss": 0.8362025618553162, "Full-finetune/Loss (Raw)": 0.8143380880355835, "Full-finetune/Step": 1871, "Full-finetune/Step Time": 6.768270554021001} {"Full-finetune/Learning Rate": 1.4804538913072719e-05, "Full-finetune/Loss": 0.8350527286529541, "Full-finetune/Loss (Raw)": 0.6854837536811829, "Full-finetune/Step": 1872, "Full-finetune/Step Time": 6.772578740492463} {"Full-finetune/Learning Rate": 1.479907448809349e-05, "Full-finetune/Loss": 0.8310717940330505, "Full-finetune/Loss (Raw)": 0.3715358376502991, "Full-finetune/Step": 1873, "Full-finetune/Step Time": 6.783328870311379} {"Full-finetune/Learning Rate": 1.4793608200687146e-05, "Full-finetune/Loss": 0.8304510712623596, "Full-finetune/Loss (Raw)": 0.9034802913665771, "Full-finetune/Step": 1874, "Full-finetune/Step Time": 6.786140782758594} {"Full-finetune/Learning Rate": 1.478814005297505e-05, "Full-finetune/Loss": 0.8324196934700012, "Full-finetune/Loss (Raw)": 0.6998182535171509, "Full-finetune/Step": 1875, "Full-finetune/Step Time": 6.778874430805445} {"Full-finetune/Learning Rate": 1.4782670047079275e-05, "Full-finetune/Loss": 0.8340640664100647, "Full-finetune/Loss (Raw)": 0.8880878686904907, "Full-finetune/Step": 1876, "Full-finetune/Step Time": 6.7643592823296785} {"Full-finetune/Learning Rate": 1.477719818512263e-05, "Full-finetune/Loss": 0.8341696858406067, "Full-finetune/Loss (Raw)": 0.8953332901000977, "Full-finetune/Step": 1877, "Full-finetune/Step Time": 6.76889631152153} {"Full-finetune/Learning Rate": 1.4771724469228638e-05, "Full-finetune/Loss": 0.83183753490448, "Full-finetune/Loss (Raw)": 0.6697081327438354, "Full-finetune/Step": 1878, "Full-finetune/Step Time": 6.7608371917158365} {"Full-finetune/Learning Rate": 1.4766248901521539e-05, "Full-finetune/Loss": 0.83225017786026, "Full-finetune/Loss (Raw)": 0.9561688899993896, "Full-finetune/Step": 1879, "Full-finetune/Step Time": 6.752167731523514} {"Full-finetune/Learning Rate": 1.4760771484126296e-05, "Full-finetune/Loss": 0.8291765451431274, "Full-finetune/Loss (Raw)": 0.6174445748329163, "Full-finetune/Step": 1880, "Full-finetune/Step Time": 6.773028077557683} {"Full-finetune/Learning Rate": 1.4755292219168581e-05, "Full-finetune/Loss": 0.828291654586792, "Full-finetune/Loss (Raw)": 0.7508620619773865, "Full-finetune/Step": 1881, "Full-finetune/Step Time": 6.780164424329996} {"Full-finetune/Learning Rate": 1.4749811108774798e-05, "Full-finetune/Loss": 0.82799232006073, "Full-finetune/Loss (Raw)": 0.882368803024292, "Full-finetune/Step": 1882, "Full-finetune/Step Time": 6.765304306522012} {"Full-finetune/Learning Rate": 1.4744328155072058e-05, "Full-finetune/Loss": 0.8263639807701111, "Full-finetune/Loss (Raw)": 0.6626892685890198, "Full-finetune/Step": 1883, "Full-finetune/Step Time": 6.772708933800459} {"Full-finetune/Learning Rate": 1.4738843360188184e-05, "Full-finetune/Loss": 0.8214972615242004, "Full-finetune/Loss (Raw)": 0.36491259932518005, "Full-finetune/Step": 1884, "Full-finetune/Step Time": 6.788738165050745} {"Full-finetune/Learning Rate": 1.4733356726251722e-05, "Full-finetune/Loss": 0.8256070613861084, "Full-finetune/Loss (Raw)": 0.8951578736305237, "Full-finetune/Step": 1885, "Full-finetune/Step Time": 6.773572230711579} {"Full-finetune/Learning Rate": 1.4727868255391924e-05, "Full-finetune/Loss": 0.8275449275970459, "Full-finetune/Loss (Raw)": 1.0074325799942017, "Full-finetune/Step": 1886, "Full-finetune/Step Time": 6.776323929429054} {"Full-finetune/Learning Rate": 1.472237794973876e-05, "Full-finetune/Loss": 0.8257961273193359, "Full-finetune/Loss (Raw)": 0.7257553339004517, "Full-finetune/Step": 1887, "Full-finetune/Step Time": 6.771296152845025} {"Full-finetune/Learning Rate": 1.471688581142291e-05, "Full-finetune/Loss": 0.8235517740249634, "Full-finetune/Loss (Raw)": 0.686543345451355, "Full-finetune/Step": 1888, "Full-finetune/Step Time": 6.765224382281303} {"Full-finetune/Learning Rate": 1.4711391842575765e-05, "Full-finetune/Loss": 0.8239196538925171, "Full-finetune/Loss (Raw)": 0.8033948540687561, "Full-finetune/Step": 1889, "Full-finetune/Step Time": 6.770039839670062} {"Full-finetune/Learning Rate": 1.4705896045329432e-05, "Full-finetune/Loss": 0.824265718460083, "Full-finetune/Loss (Raw)": 0.8111037015914917, "Full-finetune/Step": 1890, "Full-finetune/Step Time": 6.762738790363073} {"Full-finetune/Learning Rate": 1.4700398421816714e-05, "Full-finetune/Loss": 0.8236554265022278, "Full-finetune/Loss (Raw)": 0.9229345321655273, "Full-finetune/Step": 1891, "Full-finetune/Step Time": 6.747314229607582} {"Full-finetune/Learning Rate": 1.469489897417114e-05, "Full-finetune/Loss": 0.8239385485649109, "Full-finetune/Loss (Raw)": 0.9788528084754944, "Full-finetune/Step": 1892, "Full-finetune/Step Time": 6.72799420915544} {"Full-finetune/Learning Rate": 1.468939770452693e-05, "Full-finetune/Loss": 0.8240045309066772, "Full-finetune/Loss (Raw)": 0.7986263036727905, "Full-finetune/Step": 1893, "Full-finetune/Step Time": 6.729729587212205} {"Full-finetune/Learning Rate": 1.4683894615019028e-05, "Full-finetune/Loss": 0.8226245045661926, "Full-finetune/Loss (Raw)": 0.7396732568740845, "Full-finetune/Step": 1894, "Full-finetune/Step Time": 6.710799742490053} {"Full-finetune/Learning Rate": 1.4678389707783072e-05, "Full-finetune/Loss": 0.8231800198554993, "Full-finetune/Loss (Raw)": 0.8196511268615723, "Full-finetune/Step": 1895, "Full-finetune/Step Time": 6.710946526378393} {"Full-finetune/Learning Rate": 1.4672882984955404e-05, "Full-finetune/Loss": 0.8242205381393433, "Full-finetune/Loss (Raw)": 0.8510369062423706, "Full-finetune/Step": 1896, "Full-finetune/Step Time": 6.713096670806408} {"Full-finetune/Learning Rate": 1.4667374448673081e-05, "Full-finetune/Loss": 0.826409101486206, "Full-finetune/Loss (Raw)": 0.9888507723808289, "Full-finetune/Step": 1897, "Full-finetune/Step Time": 6.7001685462892056} {"Full-finetune/Learning Rate": 1.4661864101073857e-05, "Full-finetune/Loss": 0.8260481953620911, "Full-finetune/Loss (Raw)": 0.8642653226852417, "Full-finetune/Step": 1898, "Full-finetune/Step Time": 6.704901972785592} {"Full-finetune/Learning Rate": 1.465635194429619e-05, "Full-finetune/Loss": 0.8252449631690979, "Full-finetune/Loss (Raw)": 0.7104561924934387, "Full-finetune/Step": 1899, "Full-finetune/Step Time": 6.723377091810107} {"Full-finetune/Learning Rate": 1.4650837980479246e-05, "Full-finetune/Loss": 0.8240368366241455, "Full-finetune/Loss (Raw)": 0.8309598565101624, "Full-finetune/Step": 1900, "Full-finetune/Step Time": 6.7253675907850266} {"Full-finetune/Learning Rate": 1.4645322211762876e-05, "Full-finetune/Loss": 0.8241525292396545, "Full-finetune/Loss (Raw)": 0.8046166896820068, "Full-finetune/Step": 1901, "Full-finetune/Step Time": 6.721900392323732} {"Full-finetune/Learning Rate": 1.4639804640287652e-05, "Full-finetune/Loss": 0.8257861137390137, "Full-finetune/Loss (Raw)": 0.9520452618598938, "Full-finetune/Step": 1902, "Full-finetune/Step Time": 6.713529501110315} {"Full-finetune/Learning Rate": 1.4634285268194828e-05, "Full-finetune/Loss": 0.8242436647415161, "Full-finetune/Loss (Raw)": 0.7529633641242981, "Full-finetune/Step": 1903, "Full-finetune/Step Time": 6.732279300689697} {"Full-finetune/Learning Rate": 1.462876409762637e-05, "Full-finetune/Loss": 0.824029266834259, "Full-finetune/Loss (Raw)": 0.8267503380775452, "Full-finetune/Step": 1904, "Full-finetune/Step Time": 6.738540975376964} {"Full-finetune/Learning Rate": 1.4623241130724937e-05, "Full-finetune/Loss": 0.8198207020759583, "Full-finetune/Loss (Raw)": 0.336907297372818, "Full-finetune/Step": 1905, "Full-finetune/Step Time": 6.759451264515519} {"Full-finetune/Learning Rate": 1.4617716369633876e-05, "Full-finetune/Loss": 0.8192321062088013, "Full-finetune/Loss (Raw)": 0.7580788731575012, "Full-finetune/Step": 1906, "Full-finetune/Step Time": 6.773579157888889} {"Full-finetune/Learning Rate": 1.461218981649725e-05, "Full-finetune/Loss": 0.8178392648696899, "Full-finetune/Loss (Raw)": 0.8648842573165894, "Full-finetune/Step": 1907, "Full-finetune/Step Time": 6.80005195364356} {"Full-finetune/Learning Rate": 1.4606661473459801e-05, "Full-finetune/Loss": 0.817704975605011, "Full-finetune/Loss (Raw)": 0.8016291856765747, "Full-finetune/Step": 1908, "Full-finetune/Step Time": 6.7989240773022175} {"Full-finetune/Learning Rate": 1.4601131342666973e-05, "Full-finetune/Loss": 0.8171291351318359, "Full-finetune/Loss (Raw)": 0.6040847301483154, "Full-finetune/Step": 1909, "Full-finetune/Step Time": 6.799901211634278} {"Full-finetune/Learning Rate": 1.45955994262649e-05, "Full-finetune/Loss": 0.8170967102050781, "Full-finetune/Loss (Raw)": 0.8630045056343079, "Full-finetune/Step": 1910, "Full-finetune/Step Time": 6.8021106366068125} {"Full-finetune/Learning Rate": 1.459006572640041e-05, "Full-finetune/Loss": 0.8185192942619324, "Full-finetune/Loss (Raw)": 0.9121439456939697, "Full-finetune/Step": 1911, "Full-finetune/Step Time": 6.808588983491063} {"Full-finetune/Learning Rate": 1.4584530245221025e-05, "Full-finetune/Loss": 0.8171274065971375, "Full-finetune/Loss (Raw)": 0.8310810327529907, "Full-finetune/Step": 1912, "Full-finetune/Step Time": 6.801351608708501} {"Full-finetune/Learning Rate": 1.4578992984874956e-05, "Full-finetune/Loss": 0.8189540505409241, "Full-finetune/Loss (Raw)": 0.8513129353523254, "Full-finetune/Step": 1913, "Full-finetune/Step Time": 6.788279268890619} {"Full-finetune/Learning Rate": 1.457345394751111e-05, "Full-finetune/Loss": 0.8201547265052795, "Full-finetune/Loss (Raw)": 1.001124620437622, "Full-finetune/Step": 1914, "Full-finetune/Step Time": 6.786090375855565} {"Full-finetune/Learning Rate": 1.4567913135279075e-05, "Full-finetune/Loss": 0.8219038248062134, "Full-finetune/Loss (Raw)": 0.9265488386154175, "Full-finetune/Step": 1915, "Full-finetune/Step Time": 6.808794345706701} {"Full-finetune/Learning Rate": 1.456237055032913e-05, "Full-finetune/Loss": 0.8222267031669617, "Full-finetune/Loss (Raw)": 0.8279909491539001, "Full-finetune/Step": 1916, "Full-finetune/Step Time": 6.809265261515975} {"Full-finetune/Learning Rate": 1.455682619481225e-05, "Full-finetune/Loss": 0.8257871866226196, "Full-finetune/Loss (Raw)": 0.7672910690307617, "Full-finetune/Step": 1917, "Full-finetune/Step Time": 6.785933854058385} {"Full-finetune/Learning Rate": 1.4551280070880089e-05, "Full-finetune/Loss": 0.825088620185852, "Full-finetune/Loss (Raw)": 0.8137915134429932, "Full-finetune/Step": 1918, "Full-finetune/Step Time": 6.7734333742409945} {"Full-finetune/Learning Rate": 1.4545732180684989e-05, "Full-finetune/Loss": 0.8241141438484192, "Full-finetune/Loss (Raw)": 0.8523992896080017, "Full-finetune/Step": 1919, "Full-finetune/Step Time": 6.770404512062669} {"Full-finetune/Learning Rate": 1.4540182526379976e-05, "Full-finetune/Loss": 0.8227376937866211, "Full-finetune/Loss (Raw)": 0.7437974214553833, "Full-finetune/Step": 1920, "Full-finetune/Step Time": 6.779652526602149} {"Full-finetune/Learning Rate": 1.4534631110118763e-05, "Full-finetune/Loss": 0.8246645927429199, "Full-finetune/Loss (Raw)": 0.961629331111908, "Full-finetune/Step": 1921, "Full-finetune/Step Time": 6.764711568132043} {"Full-finetune/Learning Rate": 1.4529077934055754e-05, "Full-finetune/Loss": 0.8198439478874207, "Full-finetune/Loss (Raw)": 0.3243117332458496, "Full-finetune/Step": 1922, "Full-finetune/Step Time": 6.7818731758743525} {"Full-finetune/Learning Rate": 1.4523523000346012e-05, "Full-finetune/Loss": 0.8235721588134766, "Full-finetune/Loss (Raw)": 0.8569449186325073, "Full-finetune/Step": 1923, "Full-finetune/Step Time": 6.7661779429763556} {"Full-finetune/Learning Rate": 1.4517966311145313e-05, "Full-finetune/Loss": 0.8240402936935425, "Full-finetune/Loss (Raw)": 0.8923894762992859, "Full-finetune/Step": 1924, "Full-finetune/Step Time": 6.7638098616153} {"Full-finetune/Learning Rate": 1.4512407868610094e-05, "Full-finetune/Loss": 0.8253346681594849, "Full-finetune/Loss (Raw)": 1.01012122631073, "Full-finetune/Step": 1925, "Full-finetune/Step Time": 6.760993082076311} {"Full-finetune/Learning Rate": 1.450684767489748e-05, "Full-finetune/Loss": 0.8212593793869019, "Full-finetune/Loss (Raw)": 0.3483901619911194, "Full-finetune/Step": 1926, "Full-finetune/Step Time": 6.774712266400456} {"Full-finetune/Learning Rate": 1.4501285732165271e-05, "Full-finetune/Loss": 0.8181053400039673, "Full-finetune/Loss (Raw)": 0.4105764627456665, "Full-finetune/Step": 1927, "Full-finetune/Step Time": 6.795370118692517} {"Full-finetune/Learning Rate": 1.4495722042571946e-05, "Full-finetune/Loss": 0.8191090822219849, "Full-finetune/Loss (Raw)": 0.9432427883148193, "Full-finetune/Step": 1928, "Full-finetune/Step Time": 6.796766925603151} {"Full-finetune/Learning Rate": 1.4490156608276674e-05, "Full-finetune/Loss": 0.8176164627075195, "Full-finetune/Loss (Raw)": 0.7113255262374878, "Full-finetune/Step": 1929, "Full-finetune/Step Time": 6.78893294557929} {"Full-finetune/Learning Rate": 1.4484589431439283e-05, "Full-finetune/Loss": 0.8167062401771545, "Full-finetune/Loss (Raw)": 0.6659058928489685, "Full-finetune/Step": 1930, "Full-finetune/Step Time": 6.774926036596298} {"Full-finetune/Learning Rate": 1.4479020514220285e-05, "Full-finetune/Loss": 0.8171156644821167, "Full-finetune/Loss (Raw)": 0.888778805732727, "Full-finetune/Step": 1931, "Full-finetune/Step Time": 6.7538346610963345} {"Full-finetune/Learning Rate": 1.4473449858780874e-05, "Full-finetune/Loss": 0.8151525259017944, "Full-finetune/Loss (Raw)": 0.6351805925369263, "Full-finetune/Step": 1932, "Full-finetune/Step Time": 6.767111221328378} {"Full-finetune/Learning Rate": 1.4467877467282906e-05, "Full-finetune/Loss": 0.817808985710144, "Full-finetune/Loss (Raw)": 0.8820237517356873, "Full-finetune/Step": 1933, "Full-finetune/Step Time": 6.750958181917667} {"Full-finetune/Learning Rate": 1.4462303341888925e-05, "Full-finetune/Loss": 0.8189523220062256, "Full-finetune/Loss (Raw)": 0.894315779209137, "Full-finetune/Step": 1934, "Full-finetune/Step Time": 6.7383693140000105} {"Full-finetune/Learning Rate": 1.4456727484762137e-05, "Full-finetune/Loss": 0.8193057179450989, "Full-finetune/Loss (Raw)": 0.9378262162208557, "Full-finetune/Step": 1935, "Full-finetune/Step Time": 6.745015600696206} {"Full-finetune/Learning Rate": 1.4451149898066418e-05, "Full-finetune/Loss": 0.8186733722686768, "Full-finetune/Loss (Raw)": 0.7242599129676819, "Full-finetune/Step": 1936, "Full-finetune/Step Time": 6.75192847289145} {"Full-finetune/Learning Rate": 1.4445570583966329e-05, "Full-finetune/Loss": 0.8184558153152466, "Full-finetune/Loss (Raw)": 0.8480208516120911, "Full-finetune/Step": 1937, "Full-finetune/Step Time": 6.751548724249005} {"Full-finetune/Learning Rate": 1.4439989544627085e-05, "Full-finetune/Loss": 0.8154789209365845, "Full-finetune/Loss (Raw)": 0.4201217591762543, "Full-finetune/Step": 1938, "Full-finetune/Step Time": 6.764023812487721} {"Full-finetune/Learning Rate": 1.4434406782214584e-05, "Full-finetune/Loss": 0.8154752850532532, "Full-finetune/Loss (Raw)": 0.845158576965332, "Full-finetune/Step": 1939, "Full-finetune/Step Time": 6.760801777243614} {"Full-finetune/Learning Rate": 1.4428822298895386e-05, "Full-finetune/Loss": 0.8134901523590088, "Full-finetune/Loss (Raw)": 0.477631539106369, "Full-finetune/Step": 1940, "Full-finetune/Step Time": 6.762347180396318} {"Full-finetune/Learning Rate": 1.442323609683672e-05, "Full-finetune/Loss": 0.8133160471916199, "Full-finetune/Loss (Raw)": 0.7987997531890869, "Full-finetune/Step": 1941, "Full-finetune/Step Time": 6.762432241812348} {"Full-finetune/Learning Rate": 1.4417648178206481e-05, "Full-finetune/Loss": 0.8142263889312744, "Full-finetune/Loss (Raw)": 0.98301762342453, "Full-finetune/Step": 1942, "Full-finetune/Step Time": 6.764077769592404} {"Full-finetune/Learning Rate": 1.4412058545173228e-05, "Full-finetune/Loss": 0.8151341676712036, "Full-finetune/Loss (Raw)": 0.9369881749153137, "Full-finetune/Step": 1943, "Full-finetune/Step Time": 6.7521172650158405} {"Full-finetune/Learning Rate": 1.4406467199906194e-05, "Full-finetune/Loss": 0.8145904541015625, "Full-finetune/Loss (Raw)": 0.8689817786216736, "Full-finetune/Step": 1944, "Full-finetune/Step Time": 6.746740855276585} {"Full-finetune/Learning Rate": 1.440087414457527e-05, "Full-finetune/Loss": 0.8110904693603516, "Full-finetune/Loss (Raw)": 0.3970976173877716, "Full-finetune/Step": 1945, "Full-finetune/Step Time": 6.763708675280213} {"Full-finetune/Learning Rate": 1.4395279381351008e-05, "Full-finetune/Loss": 0.8121459484100342, "Full-finetune/Loss (Raw)": 0.9106683135032654, "Full-finetune/Step": 1946, "Full-finetune/Step Time": 6.779061878100038} {"Full-finetune/Learning Rate": 1.438968291240463e-05, "Full-finetune/Loss": 0.8129712343215942, "Full-finetune/Loss (Raw)": 1.0107543468475342, "Full-finetune/Step": 1947, "Full-finetune/Step Time": 6.775934338569641} {"Full-finetune/Learning Rate": 1.4384084739908016e-05, "Full-finetune/Loss": 0.8153048753738403, "Full-finetune/Loss (Raw)": 1.0430859327316284, "Full-finetune/Step": 1948, "Full-finetune/Step Time": 6.776687296107411} {"Full-finetune/Learning Rate": 1.4378484866033705e-05, "Full-finetune/Loss": 0.8159602880477905, "Full-finetune/Loss (Raw)": 0.9534314274787903, "Full-finetune/Step": 1949, "Full-finetune/Step Time": 6.774713292717934} {"Full-finetune/Learning Rate": 1.43728832929549e-05, "Full-finetune/Loss": 0.8152863383293152, "Full-finetune/Loss (Raw)": 0.9874410629272461, "Full-finetune/Step": 1950, "Full-finetune/Step Time": 6.782191062346101} {"Full-finetune/Learning Rate": 1.4367280022845462e-05, "Full-finetune/Loss": 0.8160499334335327, "Full-finetune/Loss (Raw)": 0.9521325826644897, "Full-finetune/Step": 1951, "Full-finetune/Step Time": 6.785670578479767} {"Full-finetune/Learning Rate": 1.4361675057879913e-05, "Full-finetune/Loss": 0.8186928033828735, "Full-finetune/Loss (Raw)": 0.9791236519813538, "Full-finetune/Step": 1952, "Full-finetune/Step Time": 6.781062735244632} {"Full-finetune/Learning Rate": 1.4356068400233429e-05, "Full-finetune/Loss": 0.8166083097457886, "Full-finetune/Loss (Raw)": 0.577892541885376, "Full-finetune/Step": 1953, "Full-finetune/Step Time": 6.791919309645891} {"Full-finetune/Learning Rate": 1.435046005208184e-05, "Full-finetune/Loss": 0.8144708871841431, "Full-finetune/Loss (Raw)": 0.6101126670837402, "Full-finetune/Step": 1954, "Full-finetune/Step Time": 6.800472078844905} {"Full-finetune/Learning Rate": 1.4344850015601644e-05, "Full-finetune/Loss": 0.8138828873634338, "Full-finetune/Loss (Raw)": 0.7344160079956055, "Full-finetune/Step": 1955, "Full-finetune/Step Time": 6.785466169938445} {"Full-finetune/Learning Rate": 1.4339238292969981e-05, "Full-finetune/Loss": 0.8127708435058594, "Full-finetune/Loss (Raw)": 0.7254685163497925, "Full-finetune/Step": 1956, "Full-finetune/Step Time": 6.780076699331403} {"Full-finetune/Learning Rate": 1.4333624886364655e-05, "Full-finetune/Loss": 0.8117979764938354, "Full-finetune/Loss (Raw)": 0.9075824022293091, "Full-finetune/Step": 1957, "Full-finetune/Step Time": 6.763113919645548} {"Full-finetune/Learning Rate": 1.4328009797964115e-05, "Full-finetune/Loss": 0.8119058609008789, "Full-finetune/Loss (Raw)": 0.9819977879524231, "Full-finetune/Step": 1958, "Full-finetune/Step Time": 6.748905625194311} {"Full-finetune/Learning Rate": 1.432239302994747e-05, "Full-finetune/Loss": 0.8114497661590576, "Full-finetune/Loss (Raw)": 0.7677552103996277, "Full-finetune/Step": 1959, "Full-finetune/Step Time": 6.751027822494507} {"Full-finetune/Learning Rate": 1.431677458449448e-05, "Full-finetune/Loss": 0.8116058111190796, "Full-finetune/Loss (Raw)": 0.9940752983093262, "Full-finetune/Step": 1960, "Full-finetune/Step Time": 6.75491089746356} {"Full-finetune/Learning Rate": 1.431115446378555e-05, "Full-finetune/Loss": 0.8115172386169434, "Full-finetune/Loss (Raw)": 0.7200355529785156, "Full-finetune/Step": 1961, "Full-finetune/Step Time": 6.743251096457243} {"Full-finetune/Learning Rate": 1.4305532670001743e-05, "Full-finetune/Loss": 0.8072569370269775, "Full-finetune/Loss (Raw)": 0.4241202473640442, "Full-finetune/Step": 1962, "Full-finetune/Step Time": 6.7639247458428144} {"Full-finetune/Learning Rate": 1.4299909205324762e-05, "Full-finetune/Loss": 0.8034168481826782, "Full-finetune/Loss (Raw)": 0.4051656424999237, "Full-finetune/Step": 1963, "Full-finetune/Step Time": 6.76924485899508} {"Full-finetune/Learning Rate": 1.4294284071936968e-05, "Full-finetune/Loss": 0.8027637004852295, "Full-finetune/Loss (Raw)": 0.8066943287849426, "Full-finetune/Step": 1964, "Full-finetune/Step Time": 6.77662805095315} {"Full-finetune/Learning Rate": 1.4288657272021367e-05, "Full-finetune/Loss": 0.8011796474456787, "Full-finetune/Loss (Raw)": 0.6235878467559814, "Full-finetune/Step": 1965, "Full-finetune/Step Time": 6.774678887799382} {"Full-finetune/Learning Rate": 1.4283028807761603e-05, "Full-finetune/Loss": 0.8004313707351685, "Full-finetune/Loss (Raw)": 0.7514320611953735, "Full-finetune/Step": 1966, "Full-finetune/Step Time": 6.769777826964855} {"Full-finetune/Learning Rate": 1.4277398681341982e-05, "Full-finetune/Loss": 0.7994050979614258, "Full-finetune/Loss (Raw)": 0.7621544003486633, "Full-finetune/Step": 1967, "Full-finetune/Step Time": 6.75688480399549} {"Full-finetune/Learning Rate": 1.4271766894947441e-05, "Full-finetune/Loss": 0.8001468181610107, "Full-finetune/Loss (Raw)": 0.9143877625465393, "Full-finetune/Step": 1968, "Full-finetune/Step Time": 6.753507301211357} {"Full-finetune/Learning Rate": 1.4266133450763564e-05, "Full-finetune/Loss": 0.797673225402832, "Full-finetune/Loss (Raw)": 0.45636066794395447, "Full-finetune/Step": 1969, "Full-finetune/Step Time": 6.768162837252021} {"Full-finetune/Learning Rate": 1.426049835097659e-05, "Full-finetune/Loss": 0.796485424041748, "Full-finetune/Loss (Raw)": 0.7953160405158997, "Full-finetune/Step": 1970, "Full-finetune/Step Time": 6.764390956610441} {"Full-finetune/Learning Rate": 1.425486159777338e-05, "Full-finetune/Loss": 0.7931227087974548, "Full-finetune/Loss (Raw)": 0.41712746024131775, "Full-finetune/Step": 1971, "Full-finetune/Step Time": 6.765686150640249} {"Full-finetune/Learning Rate": 1.4249223193341456e-05, "Full-finetune/Loss": 0.7934486865997314, "Full-finetune/Loss (Raw)": 0.8405994176864624, "Full-finetune/Step": 1972, "Full-finetune/Step Time": 6.7554249484092} {"Full-finetune/Learning Rate": 1.4243583139868969e-05, "Full-finetune/Loss": 0.7936485409736633, "Full-finetune/Loss (Raw)": 0.7798683047294617, "Full-finetune/Step": 1973, "Full-finetune/Step Time": 6.780921323224902} {"Full-finetune/Learning Rate": 1.423794143954471e-05, "Full-finetune/Loss": 0.7937350273132324, "Full-finetune/Loss (Raw)": 0.9738218188285828, "Full-finetune/Step": 1974, "Full-finetune/Step Time": 6.779302842915058} {"Full-finetune/Learning Rate": 1.4232298094558125e-05, "Full-finetune/Loss": 0.7936421632766724, "Full-finetune/Loss (Raw)": 0.8649457693099976, "Full-finetune/Step": 1975, "Full-finetune/Step Time": 6.7687493823468685} {"Full-finetune/Learning Rate": 1.4226653107099273e-05, "Full-finetune/Loss": 0.7936307191848755, "Full-finetune/Loss (Raw)": 0.8697277903556824, "Full-finetune/Step": 1976, "Full-finetune/Step Time": 6.771150706335902} {"Full-finetune/Learning Rate": 1.422100647935887e-05, "Full-finetune/Loss": 0.793144941329956, "Full-finetune/Loss (Raw)": 0.8992297649383545, "Full-finetune/Step": 1977, "Full-finetune/Step Time": 6.778416309505701} {"Full-finetune/Learning Rate": 1.421535821352826e-05, "Full-finetune/Loss": 0.7935597896575928, "Full-finetune/Loss (Raw)": 0.9752150774002075, "Full-finetune/Step": 1978, "Full-finetune/Step Time": 6.770970471203327} {"Full-finetune/Learning Rate": 1.4209708311799424e-05, "Full-finetune/Loss": 0.7916845679283142, "Full-finetune/Loss (Raw)": 0.39902758598327637, "Full-finetune/Step": 1979, "Full-finetune/Step Time": 6.765044944360852} {"Full-finetune/Learning Rate": 1.4204056776364985e-05, "Full-finetune/Loss": 0.7899461984634399, "Full-finetune/Loss (Raw)": 0.6659858226776123, "Full-finetune/Step": 1980, "Full-finetune/Step Time": 6.778921319171786} {"Full-finetune/Learning Rate": 1.4198403609418184e-05, "Full-finetune/Loss": 0.7897242307662964, "Full-finetune/Loss (Raw)": 0.8679555654525757, "Full-finetune/Step": 1981, "Full-finetune/Step Time": 6.779313428327441} {"Full-finetune/Learning Rate": 1.419274881315291e-05, "Full-finetune/Loss": 0.7895225286483765, "Full-finetune/Loss (Raw)": 0.8633522987365723, "Full-finetune/Step": 1982, "Full-finetune/Step Time": 6.797768058255315} {"Full-finetune/Learning Rate": 1.418709238976368e-05, "Full-finetune/Loss": 0.7890152931213379, "Full-finetune/Loss (Raw)": 0.8297319412231445, "Full-finetune/Step": 1983, "Full-finetune/Step Time": 6.783662589266896} {"Full-finetune/Learning Rate": 1.4181434341445634e-05, "Full-finetune/Loss": 0.789089024066925, "Full-finetune/Loss (Raw)": 1.0051405429840088, "Full-finetune/Step": 1984, "Full-finetune/Step Time": 6.783377347514033} {"Full-finetune/Learning Rate": 1.4175774670394561e-05, "Full-finetune/Loss": 0.7906324863433838, "Full-finetune/Loss (Raw)": 0.9515411853790283, "Full-finetune/Step": 1985, "Full-finetune/Step Time": 6.744330072775483} {"Full-finetune/Learning Rate": 1.4170113378806862e-05, "Full-finetune/Loss": 0.7920675873756409, "Full-finetune/Loss (Raw)": 1.0202553272247314, "Full-finetune/Step": 1986, "Full-finetune/Step Time": 6.735383352264762} {"Full-finetune/Learning Rate": 1.4164450468879578e-05, "Full-finetune/Loss": 0.790804922580719, "Full-finetune/Loss (Raw)": 0.8018835186958313, "Full-finetune/Step": 1987, "Full-finetune/Step Time": 6.76096797734499} {"Full-finetune/Learning Rate": 1.4158785942810367e-05, "Full-finetune/Loss": 0.7915470600128174, "Full-finetune/Loss (Raw)": 0.8049646615982056, "Full-finetune/Step": 1988, "Full-finetune/Step Time": 6.759681211784482} {"Full-finetune/Learning Rate": 1.4153119802797526e-05, "Full-finetune/Loss": 0.7923422455787659, "Full-finetune/Loss (Raw)": 0.9014043211936951, "Full-finetune/Step": 1989, "Full-finetune/Step Time": 6.761026272550225} {"Full-finetune/Learning Rate": 1.4147452051039974e-05, "Full-finetune/Loss": 0.7932919263839722, "Full-finetune/Loss (Raw)": 0.8999719023704529, "Full-finetune/Step": 1990, "Full-finetune/Step Time": 6.760602589696646} {"Full-finetune/Learning Rate": 1.4141782689737252e-05, "Full-finetune/Loss": 0.7931782603263855, "Full-finetune/Loss (Raw)": 0.7301802039146423, "Full-finetune/Step": 1991, "Full-finetune/Step Time": 6.755639562383294} {"Full-finetune/Learning Rate": 1.413611172108953e-05, "Full-finetune/Loss": 0.7940815091133118, "Full-finetune/Loss (Raw)": 0.8796396851539612, "Full-finetune/Step": 1992, "Full-finetune/Step Time": 6.751193897798657} {"Full-finetune/Learning Rate": 1.41304391472976e-05, "Full-finetune/Loss": 0.7925336360931396, "Full-finetune/Loss (Raw)": 0.7134344577789307, "Full-finetune/Step": 1993, "Full-finetune/Step Time": 6.754102170467377} {"Full-finetune/Learning Rate": 1.4124764970562874e-05, "Full-finetune/Loss": 0.7902987003326416, "Full-finetune/Loss (Raw)": 0.6578815579414368, "Full-finetune/Step": 1994, "Full-finetune/Step Time": 6.75046481564641} {"Full-finetune/Learning Rate": 1.4119089193087396e-05, "Full-finetune/Loss": 0.7883546948432922, "Full-finetune/Loss (Raw)": 0.7035447359085083, "Full-finetune/Step": 1995, "Full-finetune/Step Time": 6.753365067765117} {"Full-finetune/Learning Rate": 1.4113411817073818e-05, "Full-finetune/Loss": 0.7883180379867554, "Full-finetune/Loss (Raw)": 0.7460620999336243, "Full-finetune/Step": 1996, "Full-finetune/Step Time": 6.740482479333878} {"Full-finetune/Learning Rate": 1.4107732844725424e-05, "Full-finetune/Loss": 0.7886183261871338, "Full-finetune/Loss (Raw)": 0.9706252813339233, "Full-finetune/Step": 1997, "Full-finetune/Step Time": 6.731071745976806} {"Full-finetune/Learning Rate": 1.4102052278246107e-05, "Full-finetune/Loss": 0.7889256477355957, "Full-finetune/Loss (Raw)": 0.8017556667327881, "Full-finetune/Step": 1998, "Full-finetune/Step Time": 6.720366716384888} {"Full-finetune/Learning Rate": 1.4096370119840389e-05, "Full-finetune/Loss": 0.7858682870864868, "Full-finetune/Loss (Raw)": 0.4229985475540161, "Full-finetune/Step": 1999, "Full-finetune/Step Time": 6.707542596384883} {"Full-finetune/Learning Rate": 1.4090686371713403e-05, "Full-finetune/Loss": 0.7857286930084229, "Full-finetune/Loss (Raw)": 0.6676161885261536, "Full-finetune/Step": 2000, "Full-finetune/Step Time": 6.732041712850332} {"Full-finetune/Learning Rate": 1.4085001036070896e-05, "Full-finetune/Loss": 0.789527177810669, "Full-finetune/Loss (Raw)": 0.8577374219894409, "Full-finetune/Step": 2001, "Full-finetune/Step Time": 6.719861283898354} {"Full-finetune/Learning Rate": 1.4079314115119242e-05, "Full-finetune/Loss": 0.7890448570251465, "Full-finetune/Loss (Raw)": 0.8417502045631409, "Full-finetune/Step": 2002, "Full-finetune/Step Time": 6.716050015762448} {"Full-finetune/Learning Rate": 1.4073625611065422e-05, "Full-finetune/Loss": 0.7896728515625, "Full-finetune/Loss (Raw)": 0.7802028059959412, "Full-finetune/Step": 2003, "Full-finetune/Step Time": 6.715081026777625} {"Full-finetune/Learning Rate": 1.4067935526117033e-05, "Full-finetune/Loss": 0.7890157103538513, "Full-finetune/Loss (Raw)": 0.8039692044258118, "Full-finetune/Step": 2004, "Full-finetune/Step Time": 6.7304685320705175} {"Full-finetune/Learning Rate": 1.4062243862482286e-05, "Full-finetune/Loss": 0.7881087064743042, "Full-finetune/Loss (Raw)": 0.7792374491691589, "Full-finetune/Step": 2005, "Full-finetune/Step Time": 6.72368318028748} {"Full-finetune/Learning Rate": 1.4056550622370004e-05, "Full-finetune/Loss": 0.7892087697982788, "Full-finetune/Loss (Raw)": 0.8105195760726929, "Full-finetune/Step": 2006, "Full-finetune/Step Time": 6.713432792574167} {"Full-finetune/Learning Rate": 1.4050855807989626e-05, "Full-finetune/Loss": 0.7900519371032715, "Full-finetune/Loss (Raw)": 1.0640959739685059, "Full-finetune/Step": 2007, "Full-finetune/Step Time": 6.713431518524885} {"Full-finetune/Learning Rate": 1.4045159421551191e-05, "Full-finetune/Loss": 0.7915849685668945, "Full-finetune/Loss (Raw)": 0.8136656880378723, "Full-finetune/Step": 2008, "Full-finetune/Step Time": 6.697690613567829} {"Full-finetune/Learning Rate": 1.4039461465265362e-05, "Full-finetune/Loss": 0.791832447052002, "Full-finetune/Loss (Raw)": 0.7825413942337036, "Full-finetune/Step": 2009, "Full-finetune/Step Time": 6.687405154109001} {"Full-finetune/Learning Rate": 1.4033761941343403e-05, "Full-finetune/Loss": 0.792179524898529, "Full-finetune/Loss (Raw)": 0.9267927408218384, "Full-finetune/Step": 2010, "Full-finetune/Step Time": 6.686900710687041} {"Full-finetune/Learning Rate": 1.4028060851997183e-05, "Full-finetune/Loss": 0.7950718402862549, "Full-finetune/Loss (Raw)": 1.0329062938690186, "Full-finetune/Step": 2011, "Full-finetune/Step Time": 6.680564938113093} {"Full-finetune/Learning Rate": 1.4022358199439193e-05, "Full-finetune/Loss": 0.7988129258155823, "Full-finetune/Loss (Raw)": 0.8437741994857788, "Full-finetune/Step": 2012, "Full-finetune/Step Time": 6.6611609030514956} {"Full-finetune/Learning Rate": 1.4016653985882512e-05, "Full-finetune/Loss": 0.7986915111541748, "Full-finetune/Loss (Raw)": 0.8796176910400391, "Full-finetune/Step": 2013, "Full-finetune/Step Time": 6.6756298542022705} {"Full-finetune/Learning Rate": 1.4010948213540838e-05, "Full-finetune/Loss": 0.796195387840271, "Full-finetune/Loss (Raw)": 0.6879255175590515, "Full-finetune/Step": 2014, "Full-finetune/Step Time": 6.683872424066067} {"Full-finetune/Learning Rate": 1.4005240884628473e-05, "Full-finetune/Loss": 0.7937108278274536, "Full-finetune/Loss (Raw)": 0.4077262878417969, "Full-finetune/Step": 2015, "Full-finetune/Step Time": 6.699669219553471} {"Full-finetune/Learning Rate": 1.3999532001360309e-05, "Full-finetune/Loss": 0.7915986776351929, "Full-finetune/Loss (Raw)": 0.41619160771369934, "Full-finetune/Step": 2016, "Full-finetune/Step Time": 6.714526139199734} {"Full-finetune/Learning Rate": 1.3993821565951868e-05, "Full-finetune/Loss": 0.7907534837722778, "Full-finetune/Loss (Raw)": 0.6952157616615295, "Full-finetune/Step": 2017, "Full-finetune/Step Time": 6.710994558408856} {"Full-finetune/Learning Rate": 1.3988109580619244e-05, "Full-finetune/Loss": 0.7913686037063599, "Full-finetune/Loss (Raw)": 0.8898319005966187, "Full-finetune/Step": 2018, "Full-finetune/Step Time": 6.71250656992197} {"Full-finetune/Learning Rate": 1.3982396047579156e-05, "Full-finetune/Loss": 0.7903303503990173, "Full-finetune/Loss (Raw)": 0.7900452613830566, "Full-finetune/Step": 2019, "Full-finetune/Step Time": 6.720335483551025} {"Full-finetune/Learning Rate": 1.3976680969048912e-05, "Full-finetune/Loss": 0.7895784974098206, "Full-finetune/Loss (Raw)": 0.8826091289520264, "Full-finetune/Step": 2020, "Full-finetune/Step Time": 6.723369816318154} {"Full-finetune/Learning Rate": 1.3970964347246419e-05, "Full-finetune/Loss": 0.789496660232544, "Full-finetune/Loss (Raw)": 0.7881596684455872, "Full-finetune/Step": 2021, "Full-finetune/Step Time": 6.729326892644167} {"Full-finetune/Learning Rate": 1.3965246184390193e-05, "Full-finetune/Loss": 0.7901982665061951, "Full-finetune/Loss (Raw)": 0.8294723629951477, "Full-finetune/Step": 2022, "Full-finetune/Step Time": 6.736960623413324} {"Full-finetune/Learning Rate": 1.3959526482699335e-05, "Full-finetune/Loss": 0.7887268662452698, "Full-finetune/Loss (Raw)": 0.6313111782073975, "Full-finetune/Step": 2023, "Full-finetune/Step Time": 6.760763876140118} {"Full-finetune/Learning Rate": 1.3953805244393554e-05, "Full-finetune/Loss": 0.7882537841796875, "Full-finetune/Loss (Raw)": 0.7904815673828125, "Full-finetune/Step": 2024, "Full-finetune/Step Time": 6.7604985460639} {"Full-finetune/Learning Rate": 1.394808247169315e-05, "Full-finetune/Loss": 0.7865697145462036, "Full-finetune/Loss (Raw)": 0.7732989192008972, "Full-finetune/Step": 2025, "Full-finetune/Step Time": 6.7716638296842575} {"Full-finetune/Learning Rate": 1.3942358166819016e-05, "Full-finetune/Loss": 0.7867944240570068, "Full-finetune/Loss (Raw)": 0.8930211663246155, "Full-finetune/Step": 2026, "Full-finetune/Step Time": 6.758482538163662} {"Full-finetune/Learning Rate": 1.3936632331992648e-05, "Full-finetune/Loss": 0.7875846028327942, "Full-finetune/Loss (Raw)": 0.8116023540496826, "Full-finetune/Step": 2027, "Full-finetune/Step Time": 6.738033125177026} {"Full-finetune/Learning Rate": 1.3930904969436125e-05, "Full-finetune/Loss": 0.7877602577209473, "Full-finetune/Loss (Raw)": 0.8534485697746277, "Full-finetune/Step": 2028, "Full-finetune/Step Time": 6.751636063680053} {"Full-finetune/Learning Rate": 1.3925176081372129e-05, "Full-finetune/Loss": 0.7885396480560303, "Full-finetune/Loss (Raw)": 0.9043686389923096, "Full-finetune/Step": 2029, "Full-finetune/Step Time": 6.744954964146018} {"Full-finetune/Learning Rate": 1.3919445670023933e-05, "Full-finetune/Loss": 0.7838611602783203, "Full-finetune/Loss (Raw)": 0.35319408774375916, "Full-finetune/Step": 2030, "Full-finetune/Step Time": 6.755562484264374} {"Full-finetune/Learning Rate": 1.3913713737615387e-05, "Full-finetune/Loss": 0.7842500805854797, "Full-finetune/Loss (Raw)": 0.8027511835098267, "Full-finetune/Step": 2031, "Full-finetune/Step Time": 6.73690371401608} {"Full-finetune/Learning Rate": 1.3907980286370952e-05, "Full-finetune/Loss": 0.7851607799530029, "Full-finetune/Loss (Raw)": 0.943320095539093, "Full-finetune/Step": 2032, "Full-finetune/Step Time": 6.732237625867128} {"Full-finetune/Learning Rate": 1.3902245318515661e-05, "Full-finetune/Loss": 0.7893078327178955, "Full-finetune/Loss (Raw)": 0.8677312135696411, "Full-finetune/Step": 2033, "Full-finetune/Step Time": 6.717493770644069} {"Full-finetune/Learning Rate": 1.389650883627515e-05, "Full-finetune/Loss": 0.7909859418869019, "Full-finetune/Loss (Raw)": 0.9728788733482361, "Full-finetune/Step": 2034, "Full-finetune/Step Time": 6.704770429059863} {"Full-finetune/Learning Rate": 1.3890770841875632e-05, "Full-finetune/Loss": 0.7893217206001282, "Full-finetune/Loss (Raw)": 0.6518635749816895, "Full-finetune/Step": 2035, "Full-finetune/Step Time": 6.703106110915542} {"Full-finetune/Learning Rate": 1.388503133754391e-05, "Full-finetune/Loss": 0.7898944616317749, "Full-finetune/Loss (Raw)": 0.8749374151229858, "Full-finetune/Step": 2036, "Full-finetune/Step Time": 6.715974433347583} {"Full-finetune/Learning Rate": 1.3879290325507378e-05, "Full-finetune/Loss": 0.7916630506515503, "Full-finetune/Loss (Raw)": 0.8304631114006042, "Full-finetune/Step": 2037, "Full-finetune/Step Time": 6.740989964455366} {"Full-finetune/Learning Rate": 1.3873547807994e-05, "Full-finetune/Loss": 0.7903043627738953, "Full-finetune/Loss (Raw)": 0.6890904903411865, "Full-finetune/Step": 2038, "Full-finetune/Step Time": 6.7432974353432655} {"Full-finetune/Learning Rate": 1.386780378723235e-05, "Full-finetune/Loss": 0.790648341178894, "Full-finetune/Loss (Raw)": 0.9561718702316284, "Full-finetune/Step": 2039, "Full-finetune/Step Time": 6.7371840719133615} {"Full-finetune/Learning Rate": 1.3862058265451561e-05, "Full-finetune/Loss": 0.7910124063491821, "Full-finetune/Loss (Raw)": 0.8776854276657104, "Full-finetune/Step": 2040, "Full-finetune/Step Time": 6.736649561673403} {"Full-finetune/Learning Rate": 1.385631124488136e-05, "Full-finetune/Loss": 0.7911763191223145, "Full-finetune/Loss (Raw)": 0.8722943663597107, "Full-finetune/Step": 2041, "Full-finetune/Step Time": 6.777453493326902} {"Full-finetune/Learning Rate": 1.385056272775205e-05, "Full-finetune/Loss": 0.7892086505889893, "Full-finetune/Loss (Raw)": 0.7492637038230896, "Full-finetune/Step": 2042, "Full-finetune/Step Time": 6.779188267886639} {"Full-finetune/Learning Rate": 1.3844812716294523e-05, "Full-finetune/Loss": 0.7887440919876099, "Full-finetune/Loss (Raw)": 0.86709064245224, "Full-finetune/Step": 2043, "Full-finetune/Step Time": 6.749327827244997} {"Full-finetune/Learning Rate": 1.3839061212740244e-05, "Full-finetune/Loss": 0.7891045212745667, "Full-finetune/Loss (Raw)": 0.874121367931366, "Full-finetune/Step": 2044, "Full-finetune/Step Time": 6.749507447704673} {"Full-finetune/Learning Rate": 1.383330821932126e-05, "Full-finetune/Loss": 0.7900559902191162, "Full-finetune/Loss (Raw)": 0.8890794515609741, "Full-finetune/Step": 2045, "Full-finetune/Step Time": 6.760559655725956} {"Full-finetune/Learning Rate": 1.3827553738270196e-05, "Full-finetune/Loss": 0.7911623120307922, "Full-finetune/Loss (Raw)": 0.9553938508033752, "Full-finetune/Step": 2046, "Full-finetune/Step Time": 6.760644918307662} {"Full-finetune/Learning Rate": 1.382179777182025e-05, "Full-finetune/Loss": 0.7917527556419373, "Full-finetune/Loss (Raw)": 0.9279823303222656, "Full-finetune/Step": 2047, "Full-finetune/Step Time": 6.7637748178094625} {"Full-finetune/Learning Rate": 1.3816040322205206e-05, "Full-finetune/Loss": 0.7916608452796936, "Full-finetune/Loss (Raw)": 0.73203444480896, "Full-finetune/Step": 2048, "Full-finetune/Step Time": 6.753264704719186} {"Full-finetune/Learning Rate": 1.3810281391659414e-05, "Full-finetune/Loss": 0.7884602546691895, "Full-finetune/Loss (Raw)": 0.5519532561302185, "Full-finetune/Step": 2049, "Full-finetune/Step Time": 6.767891779541969} {"Full-finetune/Learning Rate": 1.3804520982417802e-05, "Full-finetune/Loss": 0.7911559343338013, "Full-finetune/Loss (Raw)": 0.6693581938743591, "Full-finetune/Step": 2050, "Full-finetune/Step Time": 6.762473041191697} {"Full-finetune/Learning Rate": 1.3798759096715877e-05, "Full-finetune/Loss": 0.7921563982963562, "Full-finetune/Loss (Raw)": 0.9850058555603027, "Full-finetune/Step": 2051, "Full-finetune/Step Time": 6.765804070979357} {"Full-finetune/Learning Rate": 1.379299573678971e-05, "Full-finetune/Loss": 0.7925935983657837, "Full-finetune/Loss (Raw)": 0.9483487606048584, "Full-finetune/Step": 2052, "Full-finetune/Step Time": 6.777524869889021} {"Full-finetune/Learning Rate": 1.3787230904875947e-05, "Full-finetune/Loss": 0.7917325496673584, "Full-finetune/Loss (Raw)": 0.8999056220054626, "Full-finetune/Step": 2053, "Full-finetune/Step Time": 6.778938930481672} {"Full-finetune/Learning Rate": 1.3781464603211811e-05, "Full-finetune/Loss": 0.7965417504310608, "Full-finetune/Loss (Raw)": 0.9639636278152466, "Full-finetune/Step": 2054, "Full-finetune/Step Time": 6.763860022649169} {"Full-finetune/Learning Rate": 1.3775696834035088e-05, "Full-finetune/Loss": 0.8003129959106445, "Full-finetune/Loss (Raw)": 0.8933055400848389, "Full-finetune/Step": 2055, "Full-finetune/Step Time": 6.748817814514041} {"Full-finetune/Learning Rate": 1.3769927599584141e-05, "Full-finetune/Loss": 0.8000926971435547, "Full-finetune/Loss (Raw)": 0.9150416851043701, "Full-finetune/Step": 2056, "Full-finetune/Step Time": 6.736482525244355} {"Full-finetune/Learning Rate": 1.376415690209789e-05, "Full-finetune/Loss": 0.8004058003425598, "Full-finetune/Loss (Raw)": 0.7514024376869202, "Full-finetune/Step": 2057, "Full-finetune/Step Time": 6.744814299046993} {"Full-finetune/Learning Rate": 1.3758384743815838e-05, "Full-finetune/Loss": 0.8011363744735718, "Full-finetune/Loss (Raw)": 0.7594192028045654, "Full-finetune/Step": 2058, "Full-finetune/Step Time": 6.755828520283103} {"Full-finetune/Learning Rate": 1.3752611126978045e-05, "Full-finetune/Loss": 0.8002177476882935, "Full-finetune/Loss (Raw)": 0.7712030410766602, "Full-finetune/Step": 2059, "Full-finetune/Step Time": 6.756239768117666} {"Full-finetune/Learning Rate": 1.3746836053825133e-05, "Full-finetune/Loss": 0.8014723658561707, "Full-finetune/Loss (Raw)": 0.7957605719566345, "Full-finetune/Step": 2060, "Full-finetune/Step Time": 6.736372193321586} {"Full-finetune/Learning Rate": 1.3741059526598299e-05, "Full-finetune/Loss": 0.8018839359283447, "Full-finetune/Loss (Raw)": 0.9347055554389954, "Full-finetune/Step": 2061, "Full-finetune/Step Time": 6.735888948664069} {"Full-finetune/Learning Rate": 1.3735281547539304e-05, "Full-finetune/Loss": 0.8011034727096558, "Full-finetune/Loss (Raw)": 0.794411838054657, "Full-finetune/Step": 2062, "Full-finetune/Step Time": 6.735460570082068} {"Full-finetune/Learning Rate": 1.3729502118890463e-05, "Full-finetune/Loss": 0.8009695410728455, "Full-finetune/Loss (Raw)": 0.9206907153129578, "Full-finetune/Step": 2063, "Full-finetune/Step Time": 6.727401765063405} {"Full-finetune/Learning Rate": 1.3723721242894663e-05, "Full-finetune/Loss": 0.8010315895080566, "Full-finetune/Loss (Raw)": 0.7322027683258057, "Full-finetune/Step": 2064, "Full-finetune/Step Time": 6.7428627125918865} {"Full-finetune/Learning Rate": 1.3717938921795347e-05, "Full-finetune/Loss": 0.8008172512054443, "Full-finetune/Loss (Raw)": 0.8205799460411072, "Full-finetune/Step": 2065, "Full-finetune/Step Time": 6.761191578581929} {"Full-finetune/Learning Rate": 1.3712155157836522e-05, "Full-finetune/Loss": 0.8002433776855469, "Full-finetune/Loss (Raw)": 0.34666708111763, "Full-finetune/Step": 2066, "Full-finetune/Step Time": 6.761456372216344} {"Full-finetune/Learning Rate": 1.3706369953262754e-05, "Full-finetune/Loss": 0.8001688718795776, "Full-finetune/Loss (Raw)": 0.8356327414512634, "Full-finetune/Step": 2067, "Full-finetune/Step Time": 6.762885807082057} {"Full-finetune/Learning Rate": 1.3700583310319171e-05, "Full-finetune/Loss": 0.8030474185943604, "Full-finetune/Loss (Raw)": 0.8460759520530701, "Full-finetune/Step": 2068, "Full-finetune/Step Time": 6.751792969182134} {"Full-finetune/Learning Rate": 1.3694795231251453e-05, "Full-finetune/Loss": 0.8033349514007568, "Full-finetune/Loss (Raw)": 0.8356058597564697, "Full-finetune/Step": 2069, "Full-finetune/Step Time": 6.746705386787653} {"Full-finetune/Learning Rate": 1.368900571830584e-05, "Full-finetune/Loss": 0.7994030714035034, "Full-finetune/Loss (Raw)": 0.4797341227531433, "Full-finetune/Step": 2070, "Full-finetune/Step Time": 6.758565686643124} {"Full-finetune/Learning Rate": 1.3683214773729123e-05, "Full-finetune/Loss": 0.7990160584449768, "Full-finetune/Loss (Raw)": 0.8874524235725403, "Full-finetune/Step": 2071, "Full-finetune/Step Time": 6.764433324337006} {"Full-finetune/Learning Rate": 1.3677422399768667e-05, "Full-finetune/Loss": 0.7996810674667358, "Full-finetune/Loss (Raw)": 0.9540984034538269, "Full-finetune/Step": 2072, "Full-finetune/Step Time": 6.755110394209623} {"Full-finetune/Learning Rate": 1.3671628598672373e-05, "Full-finetune/Loss": 0.8022439479827881, "Full-finetune/Loss (Raw)": 0.7251471281051636, "Full-finetune/Step": 2073, "Full-finetune/Step Time": 6.747884200885892} {"Full-finetune/Learning Rate": 1.3665833372688702e-05, "Full-finetune/Loss": 0.8026307225227356, "Full-finetune/Loss (Raw)": 0.9601773619651794, "Full-finetune/Step": 2074, "Full-finetune/Step Time": 6.735855681821704} {"Full-finetune/Learning Rate": 1.366003672406667e-05, "Full-finetune/Loss": 0.8008331060409546, "Full-finetune/Loss (Raw)": 0.780663013458252, "Full-finetune/Step": 2075, "Full-finetune/Step Time": 6.7359009105712175} {"Full-finetune/Learning Rate": 1.3654238655055836e-05, "Full-finetune/Loss": 0.7980812788009644, "Full-finetune/Loss (Raw)": 0.690843939781189, "Full-finetune/Step": 2076, "Full-finetune/Step Time": 6.748285034671426} {"Full-finetune/Learning Rate": 1.364843916790633e-05, "Full-finetune/Loss": 0.7968716621398926, "Full-finetune/Loss (Raw)": 0.7986119985580444, "Full-finetune/Step": 2077, "Full-finetune/Step Time": 6.748125357553363} {"Full-finetune/Learning Rate": 1.3642638264868808e-05, "Full-finetune/Loss": 0.7951882481575012, "Full-finetune/Loss (Raw)": 0.7719573974609375, "Full-finetune/Step": 2078, "Full-finetune/Step Time": 6.737593172118068} {"Full-finetune/Learning Rate": 1.3636835948194494e-05, "Full-finetune/Loss": 0.7938277721405029, "Full-finetune/Loss (Raw)": 0.7779991626739502, "Full-finetune/Step": 2079, "Full-finetune/Step Time": 6.732559688389301} {"Full-finetune/Learning Rate": 1.3631032220135152e-05, "Full-finetune/Loss": 0.7917261719703674, "Full-finetune/Loss (Raw)": 0.7101133465766907, "Full-finetune/Step": 2080, "Full-finetune/Step Time": 6.740178123116493} {"Full-finetune/Learning Rate": 1.3625227082943093e-05, "Full-finetune/Loss": 0.7935227751731873, "Full-finetune/Loss (Raw)": 0.8078580498695374, "Full-finetune/Step": 2081, "Full-finetune/Step Time": 6.738650545477867} {"Full-finetune/Learning Rate": 1.361942053887118e-05, "Full-finetune/Loss": 0.7949396371841431, "Full-finetune/Loss (Raw)": 0.7914687395095825, "Full-finetune/Step": 2082, "Full-finetune/Step Time": 6.7249733824282885} {"Full-finetune/Learning Rate": 1.361361259017282e-05, "Full-finetune/Loss": 0.7962336540222168, "Full-finetune/Loss (Raw)": 0.9000499248504639, "Full-finetune/Step": 2083, "Full-finetune/Step Time": 6.7071597427129745} {"Full-finetune/Learning Rate": 1.3607803239101965e-05, "Full-finetune/Loss": 0.7983475923538208, "Full-finetune/Loss (Raw)": 0.9960529804229736, "Full-finetune/Step": 2084, "Full-finetune/Step Time": 6.712283620610833} {"Full-finetune/Learning Rate": 1.360199248791311e-05, "Full-finetune/Loss": 0.7937651872634888, "Full-finetune/Loss (Raw)": 0.32103949785232544, "Full-finetune/Step": 2085, "Full-finetune/Step Time": 6.728878188878298} {"Full-finetune/Learning Rate": 1.359618033886129e-05, "Full-finetune/Loss": 0.7930858135223389, "Full-finetune/Loss (Raw)": 0.8950335383415222, "Full-finetune/Step": 2086, "Full-finetune/Step Time": 6.734886351972818} {"Full-finetune/Learning Rate": 1.359036679420209e-05, "Full-finetune/Loss": 0.7947264313697815, "Full-finetune/Loss (Raw)": 0.9777578711509705, "Full-finetune/Step": 2087, "Full-finetune/Step Time": 6.723743710666895} {"Full-finetune/Learning Rate": 1.3584551856191635e-05, "Full-finetune/Loss": 0.794113039970398, "Full-finetune/Loss (Raw)": 0.9155637621879578, "Full-finetune/Step": 2088, "Full-finetune/Step Time": 6.714091680943966} {"Full-finetune/Learning Rate": 1.3578735527086585e-05, "Full-finetune/Loss": 0.7951722145080566, "Full-finetune/Loss (Raw)": 0.855606734752655, "Full-finetune/Step": 2089, "Full-finetune/Step Time": 6.708906648680568} {"Full-finetune/Learning Rate": 1.357291780914415e-05, "Full-finetune/Loss": 0.7939733266830444, "Full-finetune/Loss (Raw)": 0.2706626355648041, "Full-finetune/Step": 2090, "Full-finetune/Step Time": 6.708492843434215} {"Full-finetune/Learning Rate": 1.3567098704622064e-05, "Full-finetune/Loss": 0.7967480421066284, "Full-finetune/Loss (Raw)": 0.7603281736373901, "Full-finetune/Step": 2091, "Full-finetune/Step Time": 6.706047864630818} {"Full-finetune/Learning Rate": 1.3561278215778618e-05, "Full-finetune/Loss": 0.7972654700279236, "Full-finetune/Loss (Raw)": 0.8729228377342224, "Full-finetune/Step": 2092, "Full-finetune/Step Time": 6.691807197406888} {"Full-finetune/Learning Rate": 1.355545634487262e-05, "Full-finetune/Loss": 0.7995837330818176, "Full-finetune/Loss (Raw)": 0.9203265905380249, "Full-finetune/Step": 2093, "Full-finetune/Step Time": 6.690010452643037} {"Full-finetune/Learning Rate": 1.354963309416343e-05, "Full-finetune/Loss": 0.801311731338501, "Full-finetune/Loss (Raw)": 0.9726159572601318, "Full-finetune/Step": 2094, "Full-finetune/Step Time": 6.690601734444499} {"Full-finetune/Learning Rate": 1.3543808465910938e-05, "Full-finetune/Loss": 0.8020073175430298, "Full-finetune/Loss (Raw)": 0.8511894345283508, "Full-finetune/Step": 2095, "Full-finetune/Step Time": 6.695031281560659} {"Full-finetune/Learning Rate": 1.3537982462375566e-05, "Full-finetune/Loss": 0.8012972474098206, "Full-finetune/Loss (Raw)": 0.8234981298446655, "Full-finetune/Step": 2096, "Full-finetune/Step Time": 6.686213683336973} {"Full-finetune/Learning Rate": 1.3532155085818276e-05, "Full-finetune/Loss": 0.8046584129333496, "Full-finetune/Loss (Raw)": 0.8865948915481567, "Full-finetune/Step": 2097, "Full-finetune/Step Time": 6.667226558551192} {"Full-finetune/Learning Rate": 1.3526326338500553e-05, "Full-finetune/Loss": 0.804359495639801, "Full-finetune/Loss (Raw)": 0.7570531368255615, "Full-finetune/Step": 2098, "Full-finetune/Step Time": 6.674354815855622} {"Full-finetune/Learning Rate": 1.3520496222684424e-05, "Full-finetune/Loss": 0.8039113283157349, "Full-finetune/Loss (Raw)": 0.35976338386535645, "Full-finetune/Step": 2099, "Full-finetune/Step Time": 6.677491256967187} {"Full-finetune/Learning Rate": 1.3514664740632446e-05, "Full-finetune/Loss": 0.8024599552154541, "Full-finetune/Loss (Raw)": 0.6548143029212952, "Full-finetune/Step": 2100, "Full-finetune/Step Time": 6.682643871754408} {"Full-finetune/Learning Rate": 1.3508831894607696e-05, "Full-finetune/Loss": 0.8035010099411011, "Full-finetune/Loss (Raw)": 0.9131332635879517, "Full-finetune/Step": 2101, "Full-finetune/Step Time": 6.647970803081989} {"Full-finetune/Learning Rate": 1.3502997686873797e-05, "Full-finetune/Loss": 0.802247941493988, "Full-finetune/Loss (Raw)": 0.8134198188781738, "Full-finetune/Step": 2102, "Full-finetune/Step Time": 6.651471605524421} {"Full-finetune/Learning Rate": 1.3497162119694883e-05, "Full-finetune/Loss": 0.7984586954116821, "Full-finetune/Loss (Raw)": 0.37993207573890686, "Full-finetune/Step": 2103, "Full-finetune/Step Time": 6.670479213818908} {"Full-finetune/Learning Rate": 1.3491325195335627e-05, "Full-finetune/Loss": 0.7979602813720703, "Full-finetune/Loss (Raw)": 0.8059219121932983, "Full-finetune/Step": 2104, "Full-finetune/Step Time": 6.678339647129178} {"Full-finetune/Learning Rate": 1.3485486916061228e-05, "Full-finetune/Loss": 0.7975326776504517, "Full-finetune/Loss (Raw)": 0.844507098197937, "Full-finetune/Step": 2105, "Full-finetune/Step Time": 6.689100904390216} {"Full-finetune/Learning Rate": 1.3479647284137398e-05, "Full-finetune/Loss": 0.7966198921203613, "Full-finetune/Loss (Raw)": 0.8583706617355347, "Full-finetune/Step": 2106, "Full-finetune/Step Time": 6.688989190384746} {"Full-finetune/Learning Rate": 1.3473806301830398e-05, "Full-finetune/Loss": 0.8002711534500122, "Full-finetune/Loss (Raw)": 0.8663917183876038, "Full-finetune/Step": 2107, "Full-finetune/Step Time": 6.670403186231852} {"Full-finetune/Learning Rate": 1.346796397140699e-05, "Full-finetune/Loss": 0.8010891079902649, "Full-finetune/Loss (Raw)": 0.7706790566444397, "Full-finetune/Step": 2108, "Full-finetune/Step Time": 6.65742053091526} {"Full-finetune/Learning Rate": 1.3462120295134476e-05, "Full-finetune/Loss": 0.8013084530830383, "Full-finetune/Loss (Raw)": 0.8960338830947876, "Full-finetune/Step": 2109, "Full-finetune/Step Time": 6.65984502248466} {"Full-finetune/Learning Rate": 1.3456275275280661e-05, "Full-finetune/Loss": 0.8006971478462219, "Full-finetune/Loss (Raw)": 0.7851018309593201, "Full-finetune/Step": 2110, "Full-finetune/Step Time": 6.6566260270774364} {"Full-finetune/Learning Rate": 1.3450428914113892e-05, "Full-finetune/Loss": 0.8003301620483398, "Full-finetune/Loss (Raw)": 0.7827559113502502, "Full-finetune/Step": 2111, "Full-finetune/Step Time": 6.6724984012544155} {"Full-finetune/Learning Rate": 1.3444581213903026e-05, "Full-finetune/Loss": 0.7961689233779907, "Full-finetune/Loss (Raw)": 0.4725072979927063, "Full-finetune/Step": 2112, "Full-finetune/Step Time": 6.691362138837576} {"Full-finetune/Learning Rate": 1.3438732176917437e-05, "Full-finetune/Loss": 0.7952990531921387, "Full-finetune/Loss (Raw)": 0.8402056097984314, "Full-finetune/Step": 2113, "Full-finetune/Step Time": 6.691735876724124} {"Full-finetune/Learning Rate": 1.3432881805427032e-05, "Full-finetune/Loss": 0.7943816781044006, "Full-finetune/Loss (Raw)": 0.9028246998786926, "Full-finetune/Step": 2114, "Full-finetune/Step Time": 6.7087609600275755} {"Full-finetune/Learning Rate": 1.3427030101702215e-05, "Full-finetune/Loss": 0.7952233552932739, "Full-finetune/Loss (Raw)": 0.9096130728721619, "Full-finetune/Step": 2115, "Full-finetune/Step Time": 6.686041705310345} {"Full-finetune/Learning Rate": 1.3421177068013925e-05, "Full-finetune/Loss": 0.7921880483627319, "Full-finetune/Loss (Raw)": 0.4164460599422455, "Full-finetune/Step": 2116, "Full-finetune/Step Time": 6.70074013993144} {"Full-finetune/Learning Rate": 1.3415322706633606e-05, "Full-finetune/Loss": 0.790391206741333, "Full-finetune/Loss (Raw)": 0.6714197397232056, "Full-finetune/Step": 2117, "Full-finetune/Step Time": 6.718456890434027} {"Full-finetune/Learning Rate": 1.3409467019833223e-05, "Full-finetune/Loss": 0.7892801761627197, "Full-finetune/Loss (Raw)": 0.7577506899833679, "Full-finetune/Step": 2118, "Full-finetune/Step Time": 6.71586194075644} {"Full-finetune/Learning Rate": 1.3403610009885258e-05, "Full-finetune/Loss": 0.7894759178161621, "Full-finetune/Loss (Raw)": 0.7552439570426941, "Full-finetune/Step": 2119, "Full-finetune/Step Time": 6.739097837358713} {"Full-finetune/Learning Rate": 1.3397751679062692e-05, "Full-finetune/Loss": 0.7852858304977417, "Full-finetune/Loss (Raw)": 0.3432984948158264, "Full-finetune/Step": 2120, "Full-finetune/Step Time": 6.757581105455756} {"Full-finetune/Learning Rate": 1.3391892029639037e-05, "Full-finetune/Loss": 0.7863941192626953, "Full-finetune/Loss (Raw)": 0.8552977442741394, "Full-finetune/Step": 2121, "Full-finetune/Step Time": 6.750821374356747} {"Full-finetune/Learning Rate": 1.3386031063888309e-05, "Full-finetune/Loss": 0.7883042097091675, "Full-finetune/Loss (Raw)": 0.9023740291595459, "Full-finetune/Step": 2122, "Full-finetune/Step Time": 6.744541976600885} {"Full-finetune/Learning Rate": 1.3380168784085028e-05, "Full-finetune/Loss": 0.7897398471832275, "Full-finetune/Loss (Raw)": 0.8873093128204346, "Full-finetune/Step": 2123, "Full-finetune/Step Time": 6.736064160242677} {"Full-finetune/Learning Rate": 1.337430519250424e-05, "Full-finetune/Loss": 0.7906895875930786, "Full-finetune/Loss (Raw)": 0.8676242232322693, "Full-finetune/Step": 2124, "Full-finetune/Step Time": 6.732886491343379} {"Full-finetune/Learning Rate": 1.3368440291421475e-05, "Full-finetune/Loss": 0.7899119853973389, "Full-finetune/Loss (Raw)": 0.8710941672325134, "Full-finetune/Step": 2125, "Full-finetune/Step Time": 6.744826944544911} {"Full-finetune/Learning Rate": 1.3362574083112804e-05, "Full-finetune/Loss": 0.7908129692077637, "Full-finetune/Loss (Raw)": 0.9170753955841064, "Full-finetune/Step": 2126, "Full-finetune/Step Time": 6.741816898807883} {"Full-finetune/Learning Rate": 1.3356706569854778e-05, "Full-finetune/Loss": 0.7941908836364746, "Full-finetune/Loss (Raw)": 0.855379045009613, "Full-finetune/Step": 2127, "Full-finetune/Step Time": 6.724958699196577} {"Full-finetune/Learning Rate": 1.3350837753924466e-05, "Full-finetune/Loss": 0.7918124794960022, "Full-finetune/Loss (Raw)": 0.3631815016269684, "Full-finetune/Step": 2128, "Full-finetune/Step Time": 6.7193559519946575} {"Full-finetune/Learning Rate": 1.3344967637599445e-05, "Full-finetune/Loss": 0.7873836755752563, "Full-finetune/Loss (Raw)": 0.2908509373664856, "Full-finetune/Step": 2129, "Full-finetune/Step Time": 6.732439324259758} {"Full-finetune/Learning Rate": 1.3339096223157782e-05, "Full-finetune/Loss": 0.7876209020614624, "Full-finetune/Loss (Raw)": 0.8721132278442383, "Full-finetune/Step": 2130, "Full-finetune/Step Time": 6.732990963384509} {"Full-finetune/Learning Rate": 1.3333223512878068e-05, "Full-finetune/Loss": 0.7887225151062012, "Full-finetune/Loss (Raw)": 0.9212033748626709, "Full-finetune/Step": 2131, "Full-finetune/Step Time": 6.722693033516407} {"Full-finetune/Learning Rate": 1.3327349509039387e-05, "Full-finetune/Loss": 0.7887963652610779, "Full-finetune/Loss (Raw)": 0.8134300708770752, "Full-finetune/Step": 2132, "Full-finetune/Step Time": 6.715220615267754} {"Full-finetune/Learning Rate": 1.3321474213921317e-05, "Full-finetune/Loss": 0.7912160754203796, "Full-finetune/Loss (Raw)": 1.088956356048584, "Full-finetune/Step": 2133, "Full-finetune/Step Time": 6.7120011784136295} {"Full-finetune/Learning Rate": 1.3315597629803959e-05, "Full-finetune/Loss": 0.7907492518424988, "Full-finetune/Loss (Raw)": 0.7507691979408264, "Full-finetune/Step": 2134, "Full-finetune/Step Time": 6.722363201901317} {"Full-finetune/Learning Rate": 1.3309719758967885e-05, "Full-finetune/Loss": 0.7884477972984314, "Full-finetune/Loss (Raw)": 0.7695074677467346, "Full-finetune/Step": 2135, "Full-finetune/Step Time": 6.717699114233255} {"Full-finetune/Learning Rate": 1.3303840603694192e-05, "Full-finetune/Loss": 0.788709819316864, "Full-finetune/Loss (Raw)": 0.8472079634666443, "Full-finetune/Step": 2136, "Full-finetune/Step Time": 6.709818480536342} {"Full-finetune/Learning Rate": 1.3297960166264462e-05, "Full-finetune/Loss": 0.7859996557235718, "Full-finetune/Loss (Raw)": 0.4356333911418915, "Full-finetune/Step": 2137, "Full-finetune/Step Time": 6.727945253252983} {"Full-finetune/Learning Rate": 1.3292078448960778e-05, "Full-finetune/Loss": 0.7854785919189453, "Full-finetune/Loss (Raw)": 0.8600996732711792, "Full-finetune/Step": 2138, "Full-finetune/Step Time": 6.73658518306911} {"Full-finetune/Learning Rate": 1.3286195454065726e-05, "Full-finetune/Loss": 0.7832359075546265, "Full-finetune/Loss (Raw)": 0.7458397746086121, "Full-finetune/Step": 2139, "Full-finetune/Step Time": 6.750503156334162} {"Full-finetune/Learning Rate": 1.3280311183862371e-05, "Full-finetune/Loss": 0.78374183177948, "Full-finetune/Loss (Raw)": 0.90853351354599, "Full-finetune/Step": 2140, "Full-finetune/Step Time": 6.753620106726885} {"Full-finetune/Learning Rate": 1.3274425640634287e-05, "Full-finetune/Loss": 0.7844970226287842, "Full-finetune/Loss (Raw)": 0.9762882590293884, "Full-finetune/Step": 2141, "Full-finetune/Step Time": 6.739743761718273} {"Full-finetune/Learning Rate": 1.3268538826665544e-05, "Full-finetune/Loss": 0.7852792739868164, "Full-finetune/Loss (Raw)": 0.7880457639694214, "Full-finetune/Step": 2142, "Full-finetune/Step Time": 6.727924205362797} {"Full-finetune/Learning Rate": 1.326265074424069e-05, "Full-finetune/Loss": 0.7885066270828247, "Full-finetune/Loss (Raw)": 0.8208355903625488, "Full-finetune/Step": 2143, "Full-finetune/Step Time": 6.7250827476382256} {"Full-finetune/Learning Rate": 1.3256761395644787e-05, "Full-finetune/Loss": 0.7915253639221191, "Full-finetune/Loss (Raw)": 0.8025858998298645, "Full-finetune/Step": 2144, "Full-finetune/Step Time": 6.715838374570012} {"Full-finetune/Learning Rate": 1.3250870783163363e-05, "Full-finetune/Loss": 0.7902340888977051, "Full-finetune/Loss (Raw)": 0.5299321413040161, "Full-finetune/Step": 2145, "Full-finetune/Step Time": 6.714119194075465} {"Full-finetune/Learning Rate": 1.3244978909082457e-05, "Full-finetune/Loss": 0.7899369597434998, "Full-finetune/Loss (Raw)": 0.8518025279045105, "Full-finetune/Step": 2146, "Full-finetune/Step Time": 6.728087624534965} {"Full-finetune/Learning Rate": 1.3239085775688591e-05, "Full-finetune/Loss": 0.7896860837936401, "Full-finetune/Loss (Raw)": 0.757933497428894, "Full-finetune/Step": 2147, "Full-finetune/Step Time": 6.738772498443723} {"Full-finetune/Learning Rate": 1.3233191385268771e-05, "Full-finetune/Loss": 0.7904432415962219, "Full-finetune/Loss (Raw)": 0.9795274138450623, "Full-finetune/Step": 2148, "Full-finetune/Step Time": 6.734812518581748} {"Full-finetune/Learning Rate": 1.3227295740110502e-05, "Full-finetune/Loss": 0.7913361191749573, "Full-finetune/Loss (Raw)": 0.9024403095245361, "Full-finetune/Step": 2149, "Full-finetune/Step Time": 6.714003158733249} {"Full-finetune/Learning Rate": 1.3221398842501758e-05, "Full-finetune/Loss": 0.7911020517349243, "Full-finetune/Loss (Raw)": 0.7995172739028931, "Full-finetune/Step": 2150, "Full-finetune/Step Time": 6.710862470790744} {"Full-finetune/Learning Rate": 1.321550069473102e-05, "Full-finetune/Loss": 0.7934139966964722, "Full-finetune/Loss (Raw)": 0.9272342920303345, "Full-finetune/Step": 2151, "Full-finetune/Step Time": 6.6948283314704895} {"Full-finetune/Learning Rate": 1.3209601299087238e-05, "Full-finetune/Loss": 0.7930445671081543, "Full-finetune/Loss (Raw)": 0.7431982159614563, "Full-finetune/Step": 2152, "Full-finetune/Step Time": 6.699941860511899} {"Full-finetune/Learning Rate": 1.3203700657859855e-05, "Full-finetune/Loss": 0.7933059930801392, "Full-finetune/Loss (Raw)": 0.8067571520805359, "Full-finetune/Step": 2153, "Full-finetune/Step Time": 6.690079230815172} {"Full-finetune/Learning Rate": 1.3197798773338793e-05, "Full-finetune/Loss": 0.79175865650177, "Full-finetune/Loss (Raw)": 0.6949723362922668, "Full-finetune/Step": 2154, "Full-finetune/Step Time": 6.698749542236328} {"Full-finetune/Learning Rate": 1.3191895647814457e-05, "Full-finetune/Loss": 0.7917940020561218, "Full-finetune/Loss (Raw)": 0.8161221742630005, "Full-finetune/Step": 2155, "Full-finetune/Step Time": 6.7054436188191175} {"Full-finetune/Learning Rate": 1.318599128357774e-05, "Full-finetune/Loss": 0.7904247045516968, "Full-finetune/Loss (Raw)": 0.6781778335571289, "Full-finetune/Step": 2156, "Full-finetune/Step Time": 6.700313972309232} {"Full-finetune/Learning Rate": 1.3180085682920005e-05, "Full-finetune/Loss": 0.7892829179763794, "Full-finetune/Loss (Raw)": 0.7582147121429443, "Full-finetune/Step": 2157, "Full-finetune/Step Time": 6.709500135853887} {"Full-finetune/Learning Rate": 1.31741788481331e-05, "Full-finetune/Loss": 0.79361492395401, "Full-finetune/Loss (Raw)": 0.9076951146125793, "Full-finetune/Step": 2158, "Full-finetune/Step Time": 6.6915122754871845} {"Full-finetune/Learning Rate": 1.316827078150936e-05, "Full-finetune/Loss": 0.7940975427627563, "Full-finetune/Loss (Raw)": 0.8645260334014893, "Full-finetune/Step": 2159, "Full-finetune/Step Time": 6.705086104571819} {"Full-finetune/Learning Rate": 1.316236148534158e-05, "Full-finetune/Loss": 0.7931730151176453, "Full-finetune/Loss (Raw)": 0.8249778747558594, "Full-finetune/Step": 2160, "Full-finetune/Step Time": 6.7003477569669485} {"Full-finetune/Learning Rate": 1.3156450961923051e-05, "Full-finetune/Loss": 0.7923697233200073, "Full-finetune/Loss (Raw)": 0.7649093866348267, "Full-finetune/Step": 2161, "Full-finetune/Step Time": 6.694852456450462} {"Full-finetune/Learning Rate": 1.3150539213547528e-05, "Full-finetune/Loss": 0.7906745076179504, "Full-finetune/Loss (Raw)": 0.7558939456939697, "Full-finetune/Step": 2162, "Full-finetune/Step Time": 6.698476232588291} {"Full-finetune/Learning Rate": 1.3144626242509243e-05, "Full-finetune/Loss": 0.7916991710662842, "Full-finetune/Loss (Raw)": 0.7830195426940918, "Full-finetune/Step": 2163, "Full-finetune/Step Time": 6.683401444926858} {"Full-finetune/Learning Rate": 1.3138712051102908e-05, "Full-finetune/Loss": 0.7923731803894043, "Full-finetune/Loss (Raw)": 0.9612117409706116, "Full-finetune/Step": 2164, "Full-finetune/Step Time": 6.6696146838366985} {"Full-finetune/Learning Rate": 1.3132796641623703e-05, "Full-finetune/Loss": 0.7923955917358398, "Full-finetune/Loss (Raw)": 0.8333327770233154, "Full-finetune/Step": 2165, "Full-finetune/Step Time": 6.668983301147819} {"Full-finetune/Learning Rate": 1.3126880016367287e-05, "Full-finetune/Loss": 0.7940607666969299, "Full-finetune/Loss (Raw)": 0.9022344946861267, "Full-finetune/Step": 2166, "Full-finetune/Step Time": 6.669453704729676} {"Full-finetune/Learning Rate": 1.3120962177629783e-05, "Full-finetune/Loss": 0.7937231659889221, "Full-finetune/Loss (Raw)": 0.9129571318626404, "Full-finetune/Step": 2167, "Full-finetune/Step Time": 6.669054791331291} {"Full-finetune/Learning Rate": 1.311504312770779e-05, "Full-finetune/Loss": 0.7931454181671143, "Full-finetune/Loss (Raw)": 0.8037318587303162, "Full-finetune/Step": 2168, "Full-finetune/Step Time": 6.667771264910698} {"Full-finetune/Learning Rate": 1.3109122868898374e-05, "Full-finetune/Loss": 0.7931890487670898, "Full-finetune/Loss (Raw)": 0.8778847455978394, "Full-finetune/Step": 2169, "Full-finetune/Step Time": 6.6271869000047445} {"Full-finetune/Learning Rate": 1.310320140349907e-05, "Full-finetune/Loss": 0.7929642796516418, "Full-finetune/Loss (Raw)": 0.7204923033714294, "Full-finetune/Step": 2170, "Full-finetune/Step Time": 6.628467407077551} {"Full-finetune/Learning Rate": 1.3097278733807892e-05, "Full-finetune/Loss": 0.7918658256530762, "Full-finetune/Loss (Raw)": 0.7264899611473083, "Full-finetune/Step": 2171, "Full-finetune/Step Time": 6.6380000952631235} {"Full-finetune/Learning Rate": 1.3091354862123304e-05, "Full-finetune/Loss": 0.7919855713844299, "Full-finetune/Loss (Raw)": 0.8894432187080383, "Full-finetune/Step": 2172, "Full-finetune/Step Time": 6.637767881155014} {"Full-finetune/Learning Rate": 1.3085429790744248e-05, "Full-finetune/Loss": 0.7895412445068359, "Full-finetune/Loss (Raw)": 0.576204776763916, "Full-finetune/Step": 2173, "Full-finetune/Step Time": 6.642367737367749} {"Full-finetune/Learning Rate": 1.3079503521970127e-05, "Full-finetune/Loss": 0.7866960763931274, "Full-finetune/Loss (Raw)": 0.5912211537361145, "Full-finetune/Step": 2174, "Full-finetune/Step Time": 6.655047111213207} {"Full-finetune/Learning Rate": 1.3073576058100815e-05, "Full-finetune/Loss": 0.785367488861084, "Full-finetune/Loss (Raw)": 0.757914125919342, "Full-finetune/Step": 2175, "Full-finetune/Step Time": 6.671129792928696} {"Full-finetune/Learning Rate": 1.3067647401436636e-05, "Full-finetune/Loss": 0.7868406772613525, "Full-finetune/Loss (Raw)": 0.9206031560897827, "Full-finetune/Step": 2176, "Full-finetune/Step Time": 6.671019522473216} {"Full-finetune/Learning Rate": 1.3061717554278393e-05, "Full-finetune/Loss": 0.7896674871444702, "Full-finetune/Loss (Raw)": 0.913784384727478, "Full-finetune/Step": 2177, "Full-finetune/Step Time": 6.660380188375711} {"Full-finetune/Learning Rate": 1.3055786518927342e-05, "Full-finetune/Loss": 0.7900593280792236, "Full-finetune/Loss (Raw)": 0.7195128798484802, "Full-finetune/Step": 2178, "Full-finetune/Step Time": 6.658205710351467} {"Full-finetune/Learning Rate": 1.3049854297685206e-05, "Full-finetune/Loss": 0.788509726524353, "Full-finetune/Loss (Raw)": 0.7866640090942383, "Full-finetune/Step": 2179, "Full-finetune/Step Time": 6.667263586074114} {"Full-finetune/Learning Rate": 1.3043920892854155e-05, "Full-finetune/Loss": 0.7879328727722168, "Full-finetune/Loss (Raw)": 0.8745036125183105, "Full-finetune/Step": 2180, "Full-finetune/Step Time": 6.661122027784586} {"Full-finetune/Learning Rate": 1.3037986306736839e-05, "Full-finetune/Loss": 0.7880553603172302, "Full-finetune/Loss (Raw)": 0.915591299533844, "Full-finetune/Step": 2181, "Full-finetune/Step Time": 6.659674521535635} {"Full-finetune/Learning Rate": 1.3032050541636343e-05, "Full-finetune/Loss": 0.7878684997558594, "Full-finetune/Loss (Raw)": 0.9400380849838257, "Full-finetune/Step": 2182, "Full-finetune/Step Time": 6.663125034421682} {"Full-finetune/Learning Rate": 1.3026113599856229e-05, "Full-finetune/Loss": 0.7861585021018982, "Full-finetune/Loss (Raw)": 0.6744313836097717, "Full-finetune/Step": 2183, "Full-finetune/Step Time": 6.6787332240492105} {"Full-finetune/Learning Rate": 1.302017548370051e-05, "Full-finetune/Loss": 0.784310519695282, "Full-finetune/Loss (Raw)": 0.6784979104995728, "Full-finetune/Step": 2184, "Full-finetune/Step Time": 6.693417269736528} {"Full-finetune/Learning Rate": 1.3014236195473649e-05, "Full-finetune/Loss": 0.785210132598877, "Full-finetune/Loss (Raw)": 0.8665590882301331, "Full-finetune/Step": 2185, "Full-finetune/Step Time": 6.680038779973984} {"Full-finetune/Learning Rate": 1.3008295737480567e-05, "Full-finetune/Loss": 0.7856882810592651, "Full-finetune/Loss (Raw)": 0.820612907409668, "Full-finetune/Step": 2186, "Full-finetune/Step Time": 6.6658630929887295} {"Full-finetune/Learning Rate": 1.300235411202664e-05, "Full-finetune/Loss": 0.7840811014175415, "Full-finetune/Loss (Raw)": 0.5654899477958679, "Full-finetune/Step": 2187, "Full-finetune/Step Time": 6.699012650176883} {"Full-finetune/Learning Rate": 1.2996411321417699e-05, "Full-finetune/Loss": 0.7840791344642639, "Full-finetune/Loss (Raw)": 0.7955093383789062, "Full-finetune/Step": 2188, "Full-finetune/Step Time": 6.731830649077892} {"Full-finetune/Learning Rate": 1.2990467367960027e-05, "Full-finetune/Loss": 0.7820131182670593, "Full-finetune/Loss (Raw)": 0.6702530980110168, "Full-finetune/Step": 2189, "Full-finetune/Step Time": 6.738723872229457} {"Full-finetune/Learning Rate": 1.298452225396035e-05, "Full-finetune/Loss": 0.7810578346252441, "Full-finetune/Loss (Raw)": 0.6721410155296326, "Full-finetune/Step": 2190, "Full-finetune/Step Time": 6.7473650723695755} {"Full-finetune/Learning Rate": 1.2978575981725858e-05, "Full-finetune/Loss": 0.7800694704055786, "Full-finetune/Loss (Raw)": 0.7941722869873047, "Full-finetune/Step": 2191, "Full-finetune/Step Time": 6.762256875634193} {"Full-finetune/Learning Rate": 1.2972628553564178e-05, "Full-finetune/Loss": 0.7794561386108398, "Full-finetune/Loss (Raw)": 0.6537047624588013, "Full-finetune/Step": 2192, "Full-finetune/Step Time": 6.743188416585326} {"Full-finetune/Learning Rate": 1.2966679971783387e-05, "Full-finetune/Loss": 0.7801470756530762, "Full-finetune/Loss (Raw)": 0.909010112285614, "Full-finetune/Step": 2193, "Full-finetune/Step Time": 6.72828596830368} {"Full-finetune/Learning Rate": 1.2960730238692023e-05, "Full-finetune/Loss": 0.7839213609695435, "Full-finetune/Loss (Raw)": 0.8297745585441589, "Full-finetune/Step": 2194, "Full-finetune/Step Time": 6.723672347143292} {"Full-finetune/Learning Rate": 1.2954779356599056e-05, "Full-finetune/Loss": 0.7843834161758423, "Full-finetune/Loss (Raw)": 0.8947864770889282, "Full-finetune/Step": 2195, "Full-finetune/Step Time": 6.721125738695264} {"Full-finetune/Learning Rate": 1.2948827327813908e-05, "Full-finetune/Loss": 0.783552885055542, "Full-finetune/Loss (Raw)": 0.7397584319114685, "Full-finetune/Step": 2196, "Full-finetune/Step Time": 6.730334788560867} {"Full-finetune/Learning Rate": 1.2942874154646447e-05, "Full-finetune/Loss": 0.783423125743866, "Full-finetune/Loss (Raw)": 0.8190023303031921, "Full-finetune/Step": 2197, "Full-finetune/Step Time": 6.732957066968083} {"Full-finetune/Learning Rate": 1.2936919839406978e-05, "Full-finetune/Loss": 0.7856634855270386, "Full-finetune/Loss (Raw)": 0.7664914131164551, "Full-finetune/Step": 2198, "Full-finetune/Step Time": 6.7260967288166285} {"Full-finetune/Learning Rate": 1.2930964384406259e-05, "Full-finetune/Loss": 0.7847846746444702, "Full-finetune/Loss (Raw)": 0.774971067905426, "Full-finetune/Step": 2199, "Full-finetune/Step Time": 6.717162972316146} {"Full-finetune/Learning Rate": 1.2925007791955487e-05, "Full-finetune/Loss": 0.7830978035926819, "Full-finetune/Loss (Raw)": 0.7381824254989624, "Full-finetune/Step": 2200, "Full-finetune/Step Time": 6.7291866187006235} {"Full-finetune/Learning Rate": 1.2919050064366294e-05, "Full-finetune/Loss": 0.7803214192390442, "Full-finetune/Loss (Raw)": 0.3697664737701416, "Full-finetune/Step": 2201, "Full-finetune/Step Time": 6.734692607074976} {"Full-finetune/Learning Rate": 1.2913091203950764e-05, "Full-finetune/Loss": 0.7791328430175781, "Full-finetune/Loss (Raw)": 0.8080343008041382, "Full-finetune/Step": 2202, "Full-finetune/Step Time": 6.727909946814179} {"Full-finetune/Learning Rate": 1.2907131213021404e-05, "Full-finetune/Loss": 0.7792513370513916, "Full-finetune/Loss (Raw)": 0.7958422303199768, "Full-finetune/Step": 2203, "Full-finetune/Step Time": 6.72831848077476} {"Full-finetune/Learning Rate": 1.2901170093891182e-05, "Full-finetune/Loss": 0.7793290615081787, "Full-finetune/Loss (Raw)": 0.7007911801338196, "Full-finetune/Step": 2204, "Full-finetune/Step Time": 6.724958365783095} {"Full-finetune/Learning Rate": 1.2895207848873488e-05, "Full-finetune/Loss": 0.7765054106712341, "Full-finetune/Loss (Raw)": 0.4371773600578308, "Full-finetune/Step": 2205, "Full-finetune/Step Time": 6.7413893807679415} {"Full-finetune/Learning Rate": 1.288924448028215e-05, "Full-finetune/Loss": 0.775010347366333, "Full-finetune/Loss (Raw)": 0.5805890560150146, "Full-finetune/Step": 2206, "Full-finetune/Step Time": 6.791904129087925} {"Full-finetune/Learning Rate": 1.288327999043144e-05, "Full-finetune/Loss": 0.7756242752075195, "Full-finetune/Loss (Raw)": 0.856584906578064, "Full-finetune/Step": 2207, "Full-finetune/Step Time": 6.786265023052692} {"Full-finetune/Learning Rate": 1.2877314381636052e-05, "Full-finetune/Loss": 0.7760990858078003, "Full-finetune/Loss (Raw)": 0.7708925008773804, "Full-finetune/Step": 2208, "Full-finetune/Step Time": 6.790626730769873} {"Full-finetune/Learning Rate": 1.287134765621113e-05, "Full-finetune/Loss": 0.7759008407592773, "Full-finetune/Loss (Raw)": 0.7824769020080566, "Full-finetune/Step": 2209, "Full-finetune/Step Time": 6.780662976205349} {"Full-finetune/Learning Rate": 1.2865379816472244e-05, "Full-finetune/Loss": 0.7729195356369019, "Full-finetune/Loss (Raw)": 0.40986233949661255, "Full-finetune/Step": 2210, "Full-finetune/Step Time": 6.797475200146437} {"Full-finetune/Learning Rate": 1.2859410864735389e-05, "Full-finetune/Loss": 0.7731389999389648, "Full-finetune/Loss (Raw)": 0.9281439781188965, "Full-finetune/Step": 2211, "Full-finetune/Step Time": 6.822567755356431} {"Full-finetune/Learning Rate": 1.2853440803317003e-05, "Full-finetune/Loss": 0.7699094414710999, "Full-finetune/Loss (Raw)": 0.5826639533042908, "Full-finetune/Step": 2212, "Full-finetune/Step Time": 6.834298074245453} {"Full-finetune/Learning Rate": 1.2847469634533952e-05, "Full-finetune/Loss": 0.7729029059410095, "Full-finetune/Loss (Raw)": 0.7042072415351868, "Full-finetune/Step": 2213, "Full-finetune/Step Time": 6.8228790406137705} {"Full-finetune/Learning Rate": 1.2841497360703528e-05, "Full-finetune/Loss": 0.7706907987594604, "Full-finetune/Loss (Raw)": 0.611879825592041, "Full-finetune/Step": 2214, "Full-finetune/Step Time": 6.8272782396525145} {"Full-finetune/Learning Rate": 1.283552398414345e-05, "Full-finetune/Loss": 0.7699328064918518, "Full-finetune/Loss (Raw)": 0.8807385563850403, "Full-finetune/Step": 2215, "Full-finetune/Step Time": 6.827305503189564} {"Full-finetune/Learning Rate": 1.2829549507171877e-05, "Full-finetune/Loss": 0.7694774866104126, "Full-finetune/Loss (Raw)": 0.8572871088981628, "Full-finetune/Step": 2216, "Full-finetune/Step Time": 6.830294279381633} {"Full-finetune/Learning Rate": 1.2823573932107381e-05, "Full-finetune/Loss": 0.7701437473297119, "Full-finetune/Loss (Raw)": 0.9408769607543945, "Full-finetune/Step": 2217, "Full-finetune/Step Time": 6.830117551609874} {"Full-finetune/Learning Rate": 1.2817597261268965e-05, "Full-finetune/Loss": 0.7723220586776733, "Full-finetune/Loss (Raw)": 0.5494922399520874, "Full-finetune/Step": 2218, "Full-finetune/Step Time": 6.817151956260204} {"Full-finetune/Learning Rate": 1.2811619496976066e-05, "Full-finetune/Loss": 0.7718065977096558, "Full-finetune/Loss (Raw)": 0.694351315498352, "Full-finetune/Step": 2219, "Full-finetune/Step Time": 6.832942487671971} {"Full-finetune/Learning Rate": 1.2805640641548529e-05, "Full-finetune/Loss": 0.7717583179473877, "Full-finetune/Loss (Raw)": 0.866744875907898, "Full-finetune/Step": 2220, "Full-finetune/Step Time": 6.833693344146013} {"Full-finetune/Learning Rate": 1.2799660697306636e-05, "Full-finetune/Loss": 0.7700431942939758, "Full-finetune/Loss (Raw)": 0.7007866501808167, "Full-finetune/Step": 2221, "Full-finetune/Step Time": 6.849622290581465} {"Full-finetune/Learning Rate": 1.2793679666571084e-05, "Full-finetune/Loss": 0.769550085067749, "Full-finetune/Loss (Raw)": 0.9094941020011902, "Full-finetune/Step": 2222, "Full-finetune/Step Time": 6.858113799244165} {"Full-finetune/Learning Rate": 1.2787697551662994e-05, "Full-finetune/Loss": 0.7695691585540771, "Full-finetune/Loss (Raw)": 0.8536393642425537, "Full-finetune/Step": 2223, "Full-finetune/Step Time": 6.86293039098382} {"Full-finetune/Learning Rate": 1.2781714354903913e-05, "Full-finetune/Loss": 0.7696994543075562, "Full-finetune/Loss (Raw)": 0.8401773571968079, "Full-finetune/Step": 2224, "Full-finetune/Step Time": 6.864439653232694} {"Full-finetune/Learning Rate": 1.2775730078615798e-05, "Full-finetune/Loss": 0.7688488960266113, "Full-finetune/Loss (Raw)": 0.7777127027511597, "Full-finetune/Step": 2225, "Full-finetune/Step Time": 6.874923249706626} {"Full-finetune/Learning Rate": 1.276974472512103e-05, "Full-finetune/Loss": 0.7695472240447998, "Full-finetune/Loss (Raw)": 0.8464410305023193, "Full-finetune/Step": 2226, "Full-finetune/Step Time": 6.87531747110188} {"Full-finetune/Learning Rate": 1.276375829674241e-05, "Full-finetune/Loss": 0.7737898826599121, "Full-finetune/Loss (Raw)": 0.9028242230415344, "Full-finetune/Step": 2227, "Full-finetune/Step Time": 6.861449100077152} {"Full-finetune/Learning Rate": 1.275777079580315e-05, "Full-finetune/Loss": 0.7746203541755676, "Full-finetune/Loss (Raw)": 0.7611169219017029, "Full-finetune/Step": 2228, "Full-finetune/Step Time": 6.8580301739275455} {"Full-finetune/Learning Rate": 1.2751782224626891e-05, "Full-finetune/Loss": 0.7734869718551636, "Full-finetune/Loss (Raw)": 0.7680577635765076, "Full-finetune/Step": 2229, "Full-finetune/Step Time": 6.857568871229887} {"Full-finetune/Learning Rate": 1.274579258553767e-05, "Full-finetune/Loss": 0.773194432258606, "Full-finetune/Loss (Raw)": 0.7759755849838257, "Full-finetune/Step": 2230, "Full-finetune/Step Time": 6.858217265456915} {"Full-finetune/Learning Rate": 1.2739801880859954e-05, "Full-finetune/Loss": 0.7775921821594238, "Full-finetune/Loss (Raw)": 0.9428479075431824, "Full-finetune/Step": 2231, "Full-finetune/Step Time": 6.840709466487169} {"Full-finetune/Learning Rate": 1.2733810112918618e-05, "Full-finetune/Loss": 0.7788721323013306, "Full-finetune/Loss (Raw)": 0.9697498083114624, "Full-finetune/Step": 2232, "Full-finetune/Step Time": 6.841914879158139} {"Full-finetune/Learning Rate": 1.2727817284038953e-05, "Full-finetune/Loss": 0.7775515913963318, "Full-finetune/Loss (Raw)": 0.6754773259162903, "Full-finetune/Step": 2233, "Full-finetune/Step Time": 6.839596662670374} {"Full-finetune/Learning Rate": 1.2721823396546655e-05, "Full-finetune/Loss": 0.7780309915542603, "Full-finetune/Loss (Raw)": 0.9197342395782471, "Full-finetune/Step": 2234, "Full-finetune/Step Time": 6.842296112328768} {"Full-finetune/Learning Rate": 1.2715828452767833e-05, "Full-finetune/Loss": 0.7785497307777405, "Full-finetune/Loss (Raw)": 0.932793378829956, "Full-finetune/Step": 2235, "Full-finetune/Step Time": 6.853672096505761} {"Full-finetune/Learning Rate": 1.2709832455029017e-05, "Full-finetune/Loss": 0.7797582149505615, "Full-finetune/Loss (Raw)": 0.9253624677658081, "Full-finetune/Step": 2236, "Full-finetune/Step Time": 6.854475142434239} {"Full-finetune/Learning Rate": 1.2703835405657124e-05, "Full-finetune/Loss": 0.7786232233047485, "Full-finetune/Loss (Raw)": 0.7507637739181519, "Full-finetune/Step": 2237, "Full-finetune/Step Time": 6.856357967481017} {"Full-finetune/Learning Rate": 1.2697837306979499e-05, "Full-finetune/Loss": 0.7776142954826355, "Full-finetune/Loss (Raw)": 0.6559529900550842, "Full-finetune/Step": 2238, "Full-finetune/Step Time": 6.8603357914835215} {"Full-finetune/Learning Rate": 1.2691838161323888e-05, "Full-finetune/Loss": 0.7802435159683228, "Full-finetune/Loss (Raw)": 1.1192938089370728, "Full-finetune/Step": 2239, "Full-finetune/Step Time": 6.846712170168757} {"Full-finetune/Learning Rate": 1.2685837971018438e-05, "Full-finetune/Loss": 0.7842392921447754, "Full-finetune/Loss (Raw)": 0.9839651584625244, "Full-finetune/Step": 2240, "Full-finetune/Step Time": 6.830257585272193} {"Full-finetune/Learning Rate": 1.2679836738391714e-05, "Full-finetune/Loss": 0.7849920988082886, "Full-finetune/Loss (Raw)": 0.936564564704895, "Full-finetune/Step": 2241, "Full-finetune/Step Time": 6.837121069431305} {"Full-finetune/Learning Rate": 1.2673834465772662e-05, "Full-finetune/Loss": 0.7840496301651001, "Full-finetune/Loss (Raw)": 0.7821912169456482, "Full-finetune/Step": 2242, "Full-finetune/Step Time": 6.820494266226888} {"Full-finetune/Learning Rate": 1.2667831155490662e-05, "Full-finetune/Loss": 0.7832129001617432, "Full-finetune/Loss (Raw)": 0.8025102615356445, "Full-finetune/Step": 2243, "Full-finetune/Step Time": 6.822676634415984} {"Full-finetune/Learning Rate": 1.2661826809875474e-05, "Full-finetune/Loss": 0.7884493470191956, "Full-finetune/Loss (Raw)": 1.0867160558700562, "Full-finetune/Step": 2244, "Full-finetune/Step Time": 6.802242638543248} {"Full-finetune/Learning Rate": 1.265582143125727e-05, "Full-finetune/Loss": 0.7899347543716431, "Full-finetune/Loss (Raw)": 0.8615524768829346, "Full-finetune/Step": 2245, "Full-finetune/Step Time": 6.781528832390904} {"Full-finetune/Learning Rate": 1.264981502196662e-05, "Full-finetune/Loss": 0.7914953231811523, "Full-finetune/Loss (Raw)": 0.9575024247169495, "Full-finetune/Step": 2246, "Full-finetune/Step Time": 6.783965706825256} {"Full-finetune/Learning Rate": 1.2643807584334491e-05, "Full-finetune/Loss": 0.7912464141845703, "Full-finetune/Loss (Raw)": 0.7233823537826538, "Full-finetune/Step": 2247, "Full-finetune/Step Time": 6.772801794111729} {"Full-finetune/Learning Rate": 1.2637799120692253e-05, "Full-finetune/Loss": 0.7941433191299438, "Full-finetune/Loss (Raw)": 0.714099109172821, "Full-finetune/Step": 2248, "Full-finetune/Step Time": 6.751106947660446} {"Full-finetune/Learning Rate": 1.2631789633371678e-05, "Full-finetune/Loss": 0.7934521436691284, "Full-finetune/Loss (Raw)": 0.7668249607086182, "Full-finetune/Step": 2249, "Full-finetune/Step Time": 6.7503208965063095} {"Full-finetune/Learning Rate": 1.2625779124704925e-05, "Full-finetune/Loss": 0.7924699783325195, "Full-finetune/Loss (Raw)": 0.7766620516777039, "Full-finetune/Step": 2250, "Full-finetune/Step Time": 6.771934360265732} {"Full-finetune/Learning Rate": 1.2619767597024561e-05, "Full-finetune/Loss": 0.7923315763473511, "Full-finetune/Loss (Raw)": 0.869591474533081, "Full-finetune/Step": 2251, "Full-finetune/Step Time": 6.773549538105726} {"Full-finetune/Learning Rate": 1.2613755052663536e-05, "Full-finetune/Loss": 0.7881473302841187, "Full-finetune/Loss (Raw)": 0.332037091255188, "Full-finetune/Step": 2252, "Full-finetune/Step Time": 6.786802355200052} {"Full-finetune/Learning Rate": 1.2607741493955208e-05, "Full-finetune/Loss": 0.788338303565979, "Full-finetune/Loss (Raw)": 0.8955472707748413, "Full-finetune/Step": 2253, "Full-finetune/Step Time": 6.770484482869506} {"Full-finetune/Learning Rate": 1.2601726923233315e-05, "Full-finetune/Loss": 0.788528561592102, "Full-finetune/Loss (Raw)": 0.9414263963699341, "Full-finetune/Step": 2254, "Full-finetune/Step Time": 6.776389019563794} {"Full-finetune/Learning Rate": 1.2595711342831998e-05, "Full-finetune/Loss": 0.7884425520896912, "Full-finetune/Loss (Raw)": 0.8443711400032043, "Full-finetune/Step": 2255, "Full-finetune/Step Time": 6.7752866595983505} {"Full-finetune/Learning Rate": 1.2589694755085789e-05, "Full-finetune/Loss": 0.7916133403778076, "Full-finetune/Loss (Raw)": 0.76904296875, "Full-finetune/Step": 2256, "Full-finetune/Step Time": 6.753554338589311} {"Full-finetune/Learning Rate": 1.25836771623296e-05, "Full-finetune/Loss": 0.7957669496536255, "Full-finetune/Loss (Raw)": 0.8225130438804626, "Full-finetune/Step": 2257, "Full-finetune/Step Time": 6.737513240426779} {"Full-finetune/Learning Rate": 1.257765856689875e-05, "Full-finetune/Loss": 0.7939864993095398, "Full-finetune/Loss (Raw)": 0.6442158222198486, "Full-finetune/Step": 2258, "Full-finetune/Step Time": 6.735534206032753} {"Full-finetune/Learning Rate": 1.2571638971128933e-05, "Full-finetune/Loss": 0.7926769852638245, "Full-finetune/Loss (Raw)": 0.7535826563835144, "Full-finetune/Step": 2259, "Full-finetune/Step Time": 6.733408056199551} {"Full-finetune/Learning Rate": 1.2565618377356237e-05, "Full-finetune/Loss": 0.7924516201019287, "Full-finetune/Loss (Raw)": 0.7845883369445801, "Full-finetune/Step": 2260, "Full-finetune/Step Time": 6.7499411422759295} {"Full-finetune/Learning Rate": 1.2559596787917144e-05, "Full-finetune/Loss": 0.7908559441566467, "Full-finetune/Loss (Raw)": 0.8847019672393799, "Full-finetune/Step": 2261, "Full-finetune/Step Time": 6.751247750595212} {"Full-finetune/Learning Rate": 1.2553574205148502e-05, "Full-finetune/Loss": 0.7927261590957642, "Full-finetune/Loss (Raw)": 0.9901588559150696, "Full-finetune/Step": 2262, "Full-finetune/Step Time": 6.745532780885696} {"Full-finetune/Learning Rate": 1.2547550631387571e-05, "Full-finetune/Loss": 0.7929275035858154, "Full-finetune/Loss (Raw)": 0.795285701751709, "Full-finetune/Step": 2263, "Full-finetune/Step Time": 6.749586246907711} {"Full-finetune/Learning Rate": 1.2541526068971975e-05, "Full-finetune/Loss": 0.7931697368621826, "Full-finetune/Loss (Raw)": 0.8782134056091309, "Full-finetune/Step": 2264, "Full-finetune/Step Time": 6.753589976578951} {"Full-finetune/Learning Rate": 1.2535500520239727e-05, "Full-finetune/Loss": 0.7962199449539185, "Full-finetune/Loss (Raw)": 0.8260571360588074, "Full-finetune/Step": 2265, "Full-finetune/Step Time": 6.738288063555956} {"Full-finetune/Learning Rate": 1.2529473987529235e-05, "Full-finetune/Loss": 0.7937073111534119, "Full-finetune/Loss (Raw)": 0.5384888052940369, "Full-finetune/Step": 2266, "Full-finetune/Step Time": 6.742845140397549} {"Full-finetune/Learning Rate": 1.2523446473179267e-05, "Full-finetune/Loss": 0.7942350506782532, "Full-finetune/Loss (Raw)": 0.8133852481842041, "Full-finetune/Step": 2267, "Full-finetune/Step Time": 6.723670428618789} {"Full-finetune/Learning Rate": 1.2517417979528991e-05, "Full-finetune/Loss": 0.7930307388305664, "Full-finetune/Loss (Raw)": 0.7543799877166748, "Full-finetune/Step": 2268, "Full-finetune/Step Time": 6.72021285071969} {"Full-finetune/Learning Rate": 1.2511388508917946e-05, "Full-finetune/Loss": 0.7923848628997803, "Full-finetune/Loss (Raw)": 0.8936159014701843, "Full-finetune/Step": 2269, "Full-finetune/Step Time": 6.712672213092446} {"Full-finetune/Learning Rate": 1.250535806368605e-05, "Full-finetune/Loss": 0.7909142374992371, "Full-finetune/Loss (Raw)": 0.5998091697692871, "Full-finetune/Step": 2270, "Full-finetune/Step Time": 6.721437456086278} {"Full-finetune/Learning Rate": 1.2499326646173606e-05, "Full-finetune/Loss": 0.7914506196975708, "Full-finetune/Loss (Raw)": 0.8894936442375183, "Full-finetune/Step": 2271, "Full-finetune/Step Time": 6.702397629618645} {"Full-finetune/Learning Rate": 1.2493294258721282e-05, "Full-finetune/Loss": 0.7921578288078308, "Full-finetune/Loss (Raw)": 0.8931083679199219, "Full-finetune/Step": 2272, "Full-finetune/Step Time": 6.696394518017769} {"Full-finetune/Learning Rate": 1.2487260903670136e-05, "Full-finetune/Loss": 0.7942402958869934, "Full-finetune/Loss (Raw)": 0.796484649181366, "Full-finetune/Step": 2273, "Full-finetune/Step Time": 6.681875443086028} {"Full-finetune/Learning Rate": 1.2481226583361592e-05, "Full-finetune/Loss": 0.7935090065002441, "Full-finetune/Loss (Raw)": 0.7582001090049744, "Full-finetune/Step": 2274, "Full-finetune/Step Time": 6.692389523610473} {"Full-finetune/Learning Rate": 1.2475191300137452e-05, "Full-finetune/Loss": 0.7947865724563599, "Full-finetune/Loss (Raw)": 0.9214643239974976, "Full-finetune/Step": 2275, "Full-finetune/Step Time": 6.67408899590373} {"Full-finetune/Learning Rate": 1.2469155056339894e-05, "Full-finetune/Loss": 0.793773889541626, "Full-finetune/Loss (Raw)": 0.8499013781547546, "Full-finetune/Step": 2276, "Full-finetune/Step Time": 6.6766432374715805} {"Full-finetune/Learning Rate": 1.2463117854311463e-05, "Full-finetune/Loss": 0.792037844657898, "Full-finetune/Loss (Raw)": 0.6802186965942383, "Full-finetune/Step": 2277, "Full-finetune/Step Time": 6.681903440505266} {"Full-finetune/Learning Rate": 1.2457079696395082e-05, "Full-finetune/Loss": 0.7927533388137817, "Full-finetune/Loss (Raw)": 0.8911137580871582, "Full-finetune/Step": 2278, "Full-finetune/Step Time": 6.680008392781019} {"Full-finetune/Learning Rate": 1.2451040584934042e-05, "Full-finetune/Loss": 0.7922802567481995, "Full-finetune/Loss (Raw)": 0.8666727542877197, "Full-finetune/Step": 2279, "Full-finetune/Step Time": 6.6695119850337505} {"Full-finetune/Learning Rate": 1.2445000522272005e-05, "Full-finetune/Loss": 0.7930787801742554, "Full-finetune/Loss (Raw)": 0.8454134464263916, "Full-finetune/Step": 2280, "Full-finetune/Step Time": 6.664343222975731} {"Full-finetune/Learning Rate": 1.2438959510752997e-05, "Full-finetune/Loss": 0.7932069897651672, "Full-finetune/Loss (Raw)": 0.823158323764801, "Full-finetune/Step": 2281, "Full-finetune/Step Time": 6.665500994771719} {"Full-finetune/Learning Rate": 1.243291755272142e-05, "Full-finetune/Loss": 0.794061541557312, "Full-finetune/Loss (Raw)": 0.8043698072433472, "Full-finetune/Step": 2282, "Full-finetune/Step Time": 6.659233637154102} {"Full-finetune/Learning Rate": 1.2426874650522039e-05, "Full-finetune/Loss": 0.7951470017433167, "Full-finetune/Loss (Raw)": 0.9550526738166809, "Full-finetune/Step": 2283, "Full-finetune/Step Time": 6.646930918097496} {"Full-finetune/Learning Rate": 1.2420830806499988e-05, "Full-finetune/Loss": 0.7947927117347717, "Full-finetune/Loss (Raw)": 0.6328305006027222, "Full-finetune/Step": 2284, "Full-finetune/Step Time": 6.660917665809393} {"Full-finetune/Learning Rate": 1.2414786023000765e-05, "Full-finetune/Loss": 0.7949823141098022, "Full-finetune/Loss (Raw)": 0.782475471496582, "Full-finetune/Step": 2285, "Full-finetune/Step Time": 6.661735504865646} {"Full-finetune/Learning Rate": 1.240874030237023e-05, "Full-finetune/Loss": 0.793971061706543, "Full-finetune/Loss (Raw)": 0.7782658338546753, "Full-finetune/Step": 2286, "Full-finetune/Step Time": 6.670895216986537} {"Full-finetune/Learning Rate": 1.2402693646954607e-05, "Full-finetune/Loss": 0.7917934656143188, "Full-finetune/Loss (Raw)": 0.5857849717140198, "Full-finetune/Step": 2287, "Full-finetune/Step Time": 6.655953636392951} {"Full-finetune/Learning Rate": 1.2396646059100492e-05, "Full-finetune/Loss": 0.7925856113433838, "Full-finetune/Loss (Raw)": 0.9263777136802673, "Full-finetune/Step": 2288, "Full-finetune/Step Time": 6.659936310723424} {"Full-finetune/Learning Rate": 1.2390597541154832e-05, "Full-finetune/Loss": 0.7914798855781555, "Full-finetune/Loss (Raw)": 0.6233727335929871, "Full-finetune/Step": 2289, "Full-finetune/Step Time": 6.672485338523984} {"Full-finetune/Learning Rate": 1.238454809546494e-05, "Full-finetune/Loss": 0.7927809953689575, "Full-finetune/Loss (Raw)": 0.9224479794502258, "Full-finetune/Step": 2290, "Full-finetune/Step Time": 6.664853157475591} {"Full-finetune/Learning Rate": 1.2378497724378484e-05, "Full-finetune/Loss": 0.7927167415618896, "Full-finetune/Loss (Raw)": 0.7747830152511597, "Full-finetune/Step": 2291, "Full-finetune/Step Time": 6.676889907568693} {"Full-finetune/Learning Rate": 1.2372446430243497e-05, "Full-finetune/Loss": 0.791021466255188, "Full-finetune/Loss (Raw)": 0.7442156076431274, "Full-finetune/Step": 2292, "Full-finetune/Step Time": 6.684813071042299} {"Full-finetune/Learning Rate": 1.2366394215408369e-05, "Full-finetune/Loss": 0.7889412641525269, "Full-finetune/Loss (Raw)": 0.5670726299285889, "Full-finetune/Step": 2293, "Full-finetune/Step Time": 6.668564120307565} {"Full-finetune/Learning Rate": 1.2360341082221847e-05, "Full-finetune/Loss": 0.7886220812797546, "Full-finetune/Loss (Raw)": 0.8613799810409546, "Full-finetune/Step": 2294, "Full-finetune/Step Time": 6.665909070521593} {"Full-finetune/Learning Rate": 1.2354287033033032e-05, "Full-finetune/Loss": 0.78767991065979, "Full-finetune/Loss (Raw)": 0.7923503518104553, "Full-finetune/Step": 2295, "Full-finetune/Step Time": 6.693670891225338} {"Full-finetune/Learning Rate": 1.2348232070191383e-05, "Full-finetune/Loss": 0.7868354320526123, "Full-finetune/Loss (Raw)": 0.6956474184989929, "Full-finetune/Step": 2296, "Full-finetune/Step Time": 6.694122163578868} {"Full-finetune/Learning Rate": 1.2342176196046715e-05, "Full-finetune/Loss": 0.7862428426742554, "Full-finetune/Loss (Raw)": 0.8020254373550415, "Full-finetune/Step": 2297, "Full-finetune/Step Time": 6.695903860032558} {"Full-finetune/Learning Rate": 1.2336119412949193e-05, "Full-finetune/Loss": 0.7868771553039551, "Full-finetune/Loss (Raw)": 0.8016866445541382, "Full-finetune/Step": 2298, "Full-finetune/Step Time": 6.696962613612413} {"Full-finetune/Learning Rate": 1.2330061723249328e-05, "Full-finetune/Loss": 0.7880895137786865, "Full-finetune/Loss (Raw)": 0.8816717267036438, "Full-finetune/Step": 2299, "Full-finetune/Step Time": 6.694361314177513} {"Full-finetune/Learning Rate": 1.2324003129298006e-05, "Full-finetune/Loss": 0.7880926728248596, "Full-finetune/Loss (Raw)": 0.8898463249206543, "Full-finetune/Step": 2300, "Full-finetune/Step Time": 6.69956343062222} {"Full-finetune/Learning Rate": 1.2317943633446441e-05, "Full-finetune/Loss": 0.7895584106445312, "Full-finetune/Loss (Raw)": 0.7638163566589355, "Full-finetune/Step": 2301, "Full-finetune/Step Time": 6.685323411598802} {"Full-finetune/Learning Rate": 1.2311883238046204e-05, "Full-finetune/Loss": 0.79063880443573, "Full-finetune/Loss (Raw)": 0.729515790939331, "Full-finetune/Step": 2302, "Full-finetune/Step Time": 6.677855189889669} {"Full-finetune/Learning Rate": 1.2305821945449222e-05, "Full-finetune/Loss": 0.7913938164710999, "Full-finetune/Loss (Raw)": 0.8545615077018738, "Full-finetune/Step": 2303, "Full-finetune/Step Time": 6.658737178891897} {"Full-finetune/Learning Rate": 1.2299759758007755e-05, "Full-finetune/Loss": 0.7892090082168579, "Full-finetune/Loss (Raw)": 0.6409440040588379, "Full-finetune/Step": 2304, "Full-finetune/Step Time": 6.666113281622529} {"Full-finetune/Learning Rate": 1.2293696678074428e-05, "Full-finetune/Loss": 0.7879722118377686, "Full-finetune/Loss (Raw)": 0.7554765939712524, "Full-finetune/Step": 2305, "Full-finetune/Step Time": 6.661495417356491} {"Full-finetune/Learning Rate": 1.22876327080022e-05, "Full-finetune/Loss": 0.7897135019302368, "Full-finetune/Loss (Raw)": 0.9424043893814087, "Full-finetune/Step": 2306, "Full-finetune/Step Time": 6.653830800205469} {"Full-finetune/Learning Rate": 1.2281567850144383e-05, "Full-finetune/Loss": 0.7903292775154114, "Full-finetune/Loss (Raw)": 0.8654736876487732, "Full-finetune/Step": 2307, "Full-finetune/Step Time": 6.637552205473185} {"Full-finetune/Learning Rate": 1.2275502106854624e-05, "Full-finetune/Loss": 0.7875857353210449, "Full-finetune/Loss (Raw)": 0.5233272314071655, "Full-finetune/Step": 2308, "Full-finetune/Step Time": 6.639233831316233} {"Full-finetune/Learning Rate": 1.2269435480486923e-05, "Full-finetune/Loss": 0.7864696383476257, "Full-finetune/Loss (Raw)": 0.7727324962615967, "Full-finetune/Step": 2309, "Full-finetune/Step Time": 6.638797285035253} {"Full-finetune/Learning Rate": 1.226336797339562e-05, "Full-finetune/Loss": 0.7861059904098511, "Full-finetune/Loss (Raw)": 0.8934961557388306, "Full-finetune/Step": 2310, "Full-finetune/Step Time": 6.6323837246745825} {"Full-finetune/Learning Rate": 1.2257299587935394e-05, "Full-finetune/Loss": 0.7878561615943909, "Full-finetune/Loss (Raw)": 0.8984455466270447, "Full-finetune/Step": 2311, "Full-finetune/Step Time": 6.613299414515495} {"Full-finetune/Learning Rate": 1.2251230326461268e-05, "Full-finetune/Loss": 0.7879941463470459, "Full-finetune/Loss (Raw)": 0.6961659789085388, "Full-finetune/Step": 2312, "Full-finetune/Step Time": 6.621580453589559} {"Full-finetune/Learning Rate": 1.2245160191328605e-05, "Full-finetune/Loss": 0.7874287962913513, "Full-finetune/Loss (Raw)": 0.7941924333572388, "Full-finetune/Step": 2313, "Full-finetune/Step Time": 6.626131527125835} {"Full-finetune/Learning Rate": 1.2239089184893102e-05, "Full-finetune/Loss": 0.7869486808776855, "Full-finetune/Loss (Raw)": 0.7591569423675537, "Full-finetune/Step": 2314, "Full-finetune/Step Time": 6.644924795255065} {"Full-finetune/Learning Rate": 1.2233017309510804e-05, "Full-finetune/Loss": 0.7886106967926025, "Full-finetune/Loss (Raw)": 0.7782256007194519, "Full-finetune/Step": 2315, "Full-finetune/Step Time": 6.610745364800096} {"Full-finetune/Learning Rate": 1.2226944567538084e-05, "Full-finetune/Loss": 0.7895599007606506, "Full-finetune/Loss (Raw)": 0.9170105457305908, "Full-finetune/Step": 2316, "Full-finetune/Step Time": 6.58616959489882} {"Full-finetune/Learning Rate": 1.2220870961331654e-05, "Full-finetune/Loss": 0.7904190421104431, "Full-finetune/Loss (Raw)": 0.7802239060401917, "Full-finetune/Step": 2317, "Full-finetune/Step Time": 6.581762995570898} {"Full-finetune/Learning Rate": 1.2214796493248564e-05, "Full-finetune/Loss": 0.7908689975738525, "Full-finetune/Loss (Raw)": 0.7297304272651672, "Full-finetune/Step": 2318, "Full-finetune/Step Time": 6.570053830742836} {"Full-finetune/Learning Rate": 1.2208721165646193e-05, "Full-finetune/Loss": 0.7920496463775635, "Full-finetune/Loss (Raw)": 0.9453008770942688, "Full-finetune/Step": 2319, "Full-finetune/Step Time": 6.560387570410967} {"Full-finetune/Learning Rate": 1.2202644980882258e-05, "Full-finetune/Loss": 0.7931950092315674, "Full-finetune/Loss (Raw)": 0.8003075122833252, "Full-finetune/Step": 2320, "Full-finetune/Step Time": 6.558556132018566} {"Full-finetune/Learning Rate": 1.2196567941314815e-05, "Full-finetune/Loss": 0.7924847602844238, "Full-finetune/Loss (Raw)": 0.8181056976318359, "Full-finetune/Step": 2321, "Full-finetune/Step Time": 6.5643310546875} {"Full-finetune/Learning Rate": 1.2190490049302233e-05, "Full-finetune/Loss": 0.7925848960876465, "Full-finetune/Loss (Raw)": 0.8425875902175903, "Full-finetune/Step": 2322, "Full-finetune/Step Time": 6.553932290524244} {"Full-finetune/Learning Rate": 1.218441130720323e-05, "Full-finetune/Loss": 0.7910232543945312, "Full-finetune/Loss (Raw)": 0.6948944330215454, "Full-finetune/Step": 2323, "Full-finetune/Step Time": 6.566995024681091} {"Full-finetune/Learning Rate": 1.2178331717376848e-05, "Full-finetune/Loss": 0.791298508644104, "Full-finetune/Loss (Raw)": 0.7749975919723511, "Full-finetune/Step": 2324, "Full-finetune/Step Time": 6.556996436789632} {"Full-finetune/Learning Rate": 1.217225128218245e-05, "Full-finetune/Loss": 0.7916461229324341, "Full-finetune/Loss (Raw)": 0.8634950518608093, "Full-finetune/Step": 2325, "Full-finetune/Step Time": 6.56249862164259} {"Full-finetune/Learning Rate": 1.2166170003979742e-05, "Full-finetune/Loss": 0.7891570329666138, "Full-finetune/Loss (Raw)": 0.44788816571235657, "Full-finetune/Step": 2326, "Full-finetune/Step Time": 6.574587397277355} {"Full-finetune/Learning Rate": 1.2160087885128745e-05, "Full-finetune/Loss": 0.7896585464477539, "Full-finetune/Loss (Raw)": 0.8391619324684143, "Full-finetune/Step": 2327, "Full-finetune/Step Time": 6.591532187536359} {"Full-finetune/Learning Rate": 1.2154004927989815e-05, "Full-finetune/Loss": 0.7895814776420593, "Full-finetune/Loss (Raw)": 0.7283180356025696, "Full-finetune/Step": 2328, "Full-finetune/Step Time": 6.5855118576437235} {"Full-finetune/Learning Rate": 1.2147921134923624e-05, "Full-finetune/Loss": 0.7938515543937683, "Full-finetune/Loss (Raw)": 0.916338324546814, "Full-finetune/Step": 2329, "Full-finetune/Step Time": 6.570431353524327} {"Full-finetune/Learning Rate": 1.2141836508291175e-05, "Full-finetune/Loss": 0.7931303977966309, "Full-finetune/Loss (Raw)": 0.7157225012779236, "Full-finetune/Step": 2330, "Full-finetune/Step Time": 6.5816431269049644} {"Full-finetune/Learning Rate": 1.2135751050453793e-05, "Full-finetune/Loss": 0.7933363914489746, "Full-finetune/Loss (Raw)": 0.8222063779830933, "Full-finetune/Step": 2331, "Full-finetune/Step Time": 6.58963181450963} {"Full-finetune/Learning Rate": 1.2129664763773125e-05, "Full-finetune/Loss": 0.795726478099823, "Full-finetune/Loss (Raw)": 1.0067278146743774, "Full-finetune/Step": 2332, "Full-finetune/Step Time": 6.586576007306576} {"Full-finetune/Learning Rate": 1.212357765061114e-05, "Full-finetune/Loss": 0.7991061806678772, "Full-finetune/Loss (Raw)": 0.8697770833969116, "Full-finetune/Step": 2333, "Full-finetune/Step Time": 6.58623924292624} {"Full-finetune/Learning Rate": 1.211748971333013e-05, "Full-finetune/Loss": 0.8011133670806885, "Full-finetune/Loss (Raw)": 0.8375065326690674, "Full-finetune/Step": 2334, "Full-finetune/Step Time": 6.5497034173458815} {"Full-finetune/Learning Rate": 1.21114009542927e-05, "Full-finetune/Loss": 0.8008733987808228, "Full-finetune/Loss (Raw)": 0.8258754014968872, "Full-finetune/Step": 2335, "Full-finetune/Step Time": 6.552719371393323} {"Full-finetune/Learning Rate": 1.2105311375861786e-05, "Full-finetune/Loss": 0.8011599779129028, "Full-finetune/Loss (Raw)": 0.8075672388076782, "Full-finetune/Step": 2336, "Full-finetune/Step Time": 6.556544357910752} {"Full-finetune/Learning Rate": 1.2099220980400625e-05, "Full-finetune/Loss": 0.8023699522018433, "Full-finetune/Loss (Raw)": 0.9373591542243958, "Full-finetune/Step": 2337, "Full-finetune/Step Time": 6.557590154930949} {"Full-finetune/Learning Rate": 1.2093129770272787e-05, "Full-finetune/Loss": 0.8059079647064209, "Full-finetune/Loss (Raw)": 0.8627204895019531, "Full-finetune/Step": 2338, "Full-finetune/Step Time": 6.541176937520504} {"Full-finetune/Learning Rate": 1.208703774784215e-05, "Full-finetune/Loss": 0.8043354153633118, "Full-finetune/Loss (Raw)": 0.7268655896186829, "Full-finetune/Step": 2339, "Full-finetune/Step Time": 6.5361352507025} {"Full-finetune/Learning Rate": 1.2080944915472908e-05, "Full-finetune/Loss": 0.8057241439819336, "Full-finetune/Loss (Raw)": 0.7604183554649353, "Full-finetune/Step": 2340, "Full-finetune/Step Time": 6.52298523299396} {"Full-finetune/Learning Rate": 1.2074851275529575e-05, "Full-finetune/Loss": 0.8071601986885071, "Full-finetune/Loss (Raw)": 0.8880231380462646, "Full-finetune/Step": 2341, "Full-finetune/Step Time": 6.5500121135264635} {"Full-finetune/Learning Rate": 1.2068756830376966e-05, "Full-finetune/Loss": 0.8052588701248169, "Full-finetune/Loss (Raw)": 0.368510901927948, "Full-finetune/Step": 2342, "Full-finetune/Step Time": 6.5598580446094275} {"Full-finetune/Learning Rate": 1.2062661582380222e-05, "Full-finetune/Loss": 0.8055812120437622, "Full-finetune/Loss (Raw)": 0.9219998717308044, "Full-finetune/Step": 2343, "Full-finetune/Step Time": 6.579422656446695} {"Full-finetune/Learning Rate": 1.2056565533904787e-05, "Full-finetune/Loss": 0.8048770427703857, "Full-finetune/Loss (Raw)": 0.7671505212783813, "Full-finetune/Step": 2344, "Full-finetune/Step Time": 6.602783367037773} {"Full-finetune/Learning Rate": 1.2050468687316418e-05, "Full-finetune/Loss": 0.805579423904419, "Full-finetune/Loss (Raw)": 1.0307821035385132, "Full-finetune/Step": 2345, "Full-finetune/Step Time": 6.597956484183669} {"Full-finetune/Learning Rate": 1.2044371044981188e-05, "Full-finetune/Loss": 0.8075870275497437, "Full-finetune/Loss (Raw)": 0.8064649701118469, "Full-finetune/Step": 2346, "Full-finetune/Step Time": 6.594267273321748} {"Full-finetune/Learning Rate": 1.2038272609265464e-05, "Full-finetune/Loss": 0.8086973428726196, "Full-finetune/Loss (Raw)": 0.8364669680595398, "Full-finetune/Step": 2347, "Full-finetune/Step Time": 6.5645050425082445} {"Full-finetune/Learning Rate": 1.2032173382535936e-05, "Full-finetune/Loss": 0.8084648251533508, "Full-finetune/Loss (Raw)": 0.8369858264923096, "Full-finetune/Step": 2348, "Full-finetune/Step Time": 6.56659235432744} {"Full-finetune/Learning Rate": 1.2026073367159597e-05, "Full-finetune/Loss": 0.8090161681175232, "Full-finetune/Loss (Raw)": 0.7713562250137329, "Full-finetune/Step": 2349, "Full-finetune/Step Time": 6.558564400300384} {"Full-finetune/Learning Rate": 1.2019972565503733e-05, "Full-finetune/Loss": 0.8083648681640625, "Full-finetune/Loss (Raw)": 0.8261297941207886, "Full-finetune/Step": 2350, "Full-finetune/Step Time": 6.552580568939447} {"Full-finetune/Learning Rate": 1.2013870979935957e-05, "Full-finetune/Loss": 0.8089348077774048, "Full-finetune/Loss (Raw)": 0.9265961050987244, "Full-finetune/Step": 2351, "Full-finetune/Step Time": 6.550596235319972} {"Full-finetune/Learning Rate": 1.200776861282417e-05, "Full-finetune/Loss": 0.808956503868103, "Full-finetune/Loss (Raw)": 0.8429484367370605, "Full-finetune/Step": 2352, "Full-finetune/Step Time": 6.569135833531618} {"Full-finetune/Learning Rate": 1.2001665466536582e-05, "Full-finetune/Loss": 0.8095057606697083, "Full-finetune/Loss (Raw)": 0.8480209708213806, "Full-finetune/Step": 2353, "Full-finetune/Step Time": 6.564508564770222} {"Full-finetune/Learning Rate": 1.199556154344171e-05, "Full-finetune/Loss": 0.8095037937164307, "Full-finetune/Loss (Raw)": 0.8461897373199463, "Full-finetune/Step": 2354, "Full-finetune/Step Time": 6.556585785001516} {"Full-finetune/Learning Rate": 1.1989456845908356e-05, "Full-finetune/Loss": 0.8085731863975525, "Full-finetune/Loss (Raw)": 0.7837035655975342, "Full-finetune/Step": 2355, "Full-finetune/Step Time": 6.554259521886706} {"Full-finetune/Learning Rate": 1.1983351376305648e-05, "Full-finetune/Loss": 0.8079901337623596, "Full-finetune/Loss (Raw)": 0.6864867806434631, "Full-finetune/Step": 2356, "Full-finetune/Step Time": 6.556883467361331} {"Full-finetune/Learning Rate": 1.197724513700299e-05, "Full-finetune/Loss": 0.8083996772766113, "Full-finetune/Loss (Raw)": 0.8204774260520935, "Full-finetune/Step": 2357, "Full-finetune/Step Time": 6.560958582907915} {"Full-finetune/Learning Rate": 1.1971138130370097e-05, "Full-finetune/Loss": 0.8094255924224854, "Full-finetune/Loss (Raw)": 0.9072973728179932, "Full-finetune/Step": 2358, "Full-finetune/Step Time": 6.558255000039935} {"Full-finetune/Learning Rate": 1.1965030358776975e-05, "Full-finetune/Loss": 0.80439293384552, "Full-finetune/Loss (Raw)": 0.29866302013397217, "Full-finetune/Step": 2359, "Full-finetune/Step Time": 6.57517709210515} {"Full-finetune/Learning Rate": 1.1958921824593936e-05, "Full-finetune/Loss": 0.8037158250808716, "Full-finetune/Loss (Raw)": 0.8830781579017639, "Full-finetune/Step": 2360, "Full-finetune/Step Time": 6.565093353390694} {"Full-finetune/Learning Rate": 1.1952812530191586e-05, "Full-finetune/Loss": 0.8053772449493408, "Full-finetune/Loss (Raw)": 0.8881394267082214, "Full-finetune/Step": 2361, "Full-finetune/Step Time": 6.559794746339321} {"Full-finetune/Learning Rate": 1.1946702477940815e-05, "Full-finetune/Loss": 0.8046268820762634, "Full-finetune/Loss (Raw)": 0.823689877986908, "Full-finetune/Step": 2362, "Full-finetune/Step Time": 6.5532760582864285} {"Full-finetune/Learning Rate": 1.194059167021282e-05, "Full-finetune/Loss": 0.8015379905700684, "Full-finetune/Loss (Raw)": 0.5374186635017395, "Full-finetune/Step": 2363, "Full-finetune/Step Time": 6.559994529932737} {"Full-finetune/Learning Rate": 1.1934480109379083e-05, "Full-finetune/Loss": 0.8000915050506592, "Full-finetune/Loss (Raw)": 0.740206778049469, "Full-finetune/Step": 2364, "Full-finetune/Step Time": 6.557843118906021} {"Full-finetune/Learning Rate": 1.1928367797811384e-05, "Full-finetune/Loss": 0.8010425567626953, "Full-finetune/Loss (Raw)": 0.8724982738494873, "Full-finetune/Step": 2365, "Full-finetune/Step Time": 6.553156670182943} {"Full-finetune/Learning Rate": 1.1922254737881792e-05, "Full-finetune/Loss": 0.8003774881362915, "Full-finetune/Loss (Raw)": 0.570820152759552, "Full-finetune/Step": 2366, "Full-finetune/Step Time": 6.5501206163316965} {"Full-finetune/Learning Rate": 1.1916140931962661e-05, "Full-finetune/Loss": 0.7988764047622681, "Full-finetune/Loss (Raw)": 0.9271564483642578, "Full-finetune/Step": 2367, "Full-finetune/Step Time": 6.555099761113524} {"Full-finetune/Learning Rate": 1.191002638242665e-05, "Full-finetune/Loss": 0.7972928285598755, "Full-finetune/Loss (Raw)": 0.7812716364860535, "Full-finetune/Step": 2368, "Full-finetune/Step Time": 6.551911603659391} {"Full-finetune/Learning Rate": 1.1903911091646684e-05, "Full-finetune/Loss": 0.7974053621292114, "Full-finetune/Loss (Raw)": 0.9509745240211487, "Full-finetune/Step": 2369, "Full-finetune/Step Time": 6.558795453980565} {"Full-finetune/Learning Rate": 1.1897795061995998e-05, "Full-finetune/Loss": 0.7976522445678711, "Full-finetune/Loss (Raw)": 0.8137945532798767, "Full-finetune/Step": 2370, "Full-finetune/Step Time": 6.570758782327175} {"Full-finetune/Learning Rate": 1.18916782958481e-05, "Full-finetune/Loss": 0.7985891103744507, "Full-finetune/Loss (Raw)": 0.9224250316619873, "Full-finetune/Step": 2371, "Full-finetune/Step Time": 6.581382030621171} {"Full-finetune/Learning Rate": 1.1885560795576782e-05, "Full-finetune/Loss": 0.7960723638534546, "Full-finetune/Loss (Raw)": 0.7645719647407532, "Full-finetune/Step": 2372, "Full-finetune/Step Time": 6.583474263548851} {"Full-finetune/Learning Rate": 1.1879442563556141e-05, "Full-finetune/Loss": 0.7965596914291382, "Full-finetune/Loss (Raw)": 0.923922061920166, "Full-finetune/Step": 2373, "Full-finetune/Step Time": 6.6047033816576} {"Full-finetune/Learning Rate": 1.1873323602160526e-05, "Full-finetune/Loss": 0.7947613000869751, "Full-finetune/Loss (Raw)": 0.7273156046867371, "Full-finetune/Step": 2374, "Full-finetune/Step Time": 6.6054237093776464} {"Full-finetune/Learning Rate": 1.1867203913764599e-05, "Full-finetune/Loss": 0.7940874099731445, "Full-finetune/Loss (Raw)": 0.6371194124221802, "Full-finetune/Step": 2375, "Full-finetune/Step Time": 6.608021812513471} {"Full-finetune/Learning Rate": 1.1861083500743284e-05, "Full-finetune/Loss": 0.7956141233444214, "Full-finetune/Loss (Raw)": 0.9095249772071838, "Full-finetune/Step": 2376, "Full-finetune/Step Time": 6.612193539738655} {"Full-finetune/Learning Rate": 1.1854962365471798e-05, "Full-finetune/Loss": 0.7972244024276733, "Full-finetune/Loss (Raw)": 0.9729419350624084, "Full-finetune/Step": 2377, "Full-finetune/Step Time": 6.612057335674763} {"Full-finetune/Learning Rate": 1.1848840510325635e-05, "Full-finetune/Loss": 0.7982920408248901, "Full-finetune/Loss (Raw)": 0.9133153557777405, "Full-finetune/Step": 2378, "Full-finetune/Step Time": 6.590513000264764} {"Full-finetune/Learning Rate": 1.184271793768056e-05, "Full-finetune/Loss": 0.7982110381126404, "Full-finetune/Loss (Raw)": 0.8592296838760376, "Full-finetune/Step": 2379, "Full-finetune/Step Time": 6.606391109526157} {"Full-finetune/Learning Rate": 1.1836594649912636e-05, "Full-finetune/Loss": 0.8013027310371399, "Full-finetune/Loss (Raw)": 0.7277672290802002, "Full-finetune/Step": 2380, "Full-finetune/Step Time": 6.587072992697358} {"Full-finetune/Learning Rate": 1.1830470649398183e-05, "Full-finetune/Loss": 0.8007519245147705, "Full-finetune/Loss (Raw)": 0.8250447511672974, "Full-finetune/Step": 2381, "Full-finetune/Step Time": 6.589364057406783} {"Full-finetune/Learning Rate": 1.1824345938513806e-05, "Full-finetune/Loss": 0.7986198663711548, "Full-finetune/Loss (Raw)": 0.6685240268707275, "Full-finetune/Step": 2382, "Full-finetune/Step Time": 6.6002403032034636} {"Full-finetune/Learning Rate": 1.181822051963639e-05, "Full-finetune/Loss": 0.7964531183242798, "Full-finetune/Loss (Raw)": 0.5670233368873596, "Full-finetune/Step": 2383, "Full-finetune/Step Time": 6.612700937315822} {"Full-finetune/Learning Rate": 1.1812094395143086e-05, "Full-finetune/Loss": 0.7979475259780884, "Full-finetune/Loss (Raw)": 0.9603325128555298, "Full-finetune/Step": 2384, "Full-finetune/Step Time": 6.60980355925858} {"Full-finetune/Learning Rate": 1.1805967567411325e-05, "Full-finetune/Loss": 0.797818660736084, "Full-finetune/Loss (Raw)": 0.8060205578804016, "Full-finetune/Step": 2385, "Full-finetune/Step Time": 6.61479252204299} {"Full-finetune/Learning Rate": 1.1799840038818813e-05, "Full-finetune/Loss": 0.8004563450813293, "Full-finetune/Loss (Raw)": 0.9818393588066101, "Full-finetune/Step": 2386, "Full-finetune/Step Time": 6.621503418311477} {"Full-finetune/Learning Rate": 1.1793711811743515e-05, "Full-finetune/Loss": 0.8009451627731323, "Full-finetune/Loss (Raw)": 0.816142201423645, "Full-finetune/Step": 2387, "Full-finetune/Step Time": 6.6343904864043} {"Full-finetune/Learning Rate": 1.1787582888563689e-05, "Full-finetune/Loss": 0.8011685609817505, "Full-finetune/Loss (Raw)": 0.8131936192512512, "Full-finetune/Step": 2388, "Full-finetune/Step Time": 6.633328957483172} {"Full-finetune/Learning Rate": 1.1781453271657839e-05, "Full-finetune/Loss": 0.8017309904098511, "Full-finetune/Loss (Raw)": 0.9566872119903564, "Full-finetune/Step": 2389, "Full-finetune/Step Time": 6.632541364058852} {"Full-finetune/Learning Rate": 1.1775322963404757e-05, "Full-finetune/Loss": 0.8012277483940125, "Full-finetune/Loss (Raw)": 0.9257411956787109, "Full-finetune/Step": 2390, "Full-finetune/Step Time": 6.636525260284543} {"Full-finetune/Learning Rate": 1.176919196618349e-05, "Full-finetune/Loss": 0.8021628856658936, "Full-finetune/Loss (Raw)": 0.9149836301803589, "Full-finetune/Step": 2391, "Full-finetune/Step Time": 6.635963646695018} {"Full-finetune/Learning Rate": 1.1763060282373364e-05, "Full-finetune/Loss": 0.8020632266998291, "Full-finetune/Loss (Raw)": 0.8654661178588867, "Full-finetune/Step": 2392, "Full-finetune/Step Time": 6.635615553706884} {"Full-finetune/Learning Rate": 1.1756927914353963e-05, "Full-finetune/Loss": 0.8027963638305664, "Full-finetune/Loss (Raw)": 0.9198906421661377, "Full-finetune/Step": 2393, "Full-finetune/Step Time": 6.647970635443926} {"Full-finetune/Learning Rate": 1.175079486450514e-05, "Full-finetune/Loss": 0.8017425537109375, "Full-finetune/Loss (Raw)": 0.4036058783531189, "Full-finetune/Step": 2394, "Full-finetune/Step Time": 6.6580964755266905} {"Full-finetune/Learning Rate": 1.1744661135207009e-05, "Full-finetune/Loss": 0.8020994663238525, "Full-finetune/Loss (Raw)": 0.8590667843818665, "Full-finetune/Step": 2395, "Full-finetune/Step Time": 6.664677729830146} {"Full-finetune/Learning Rate": 1.1738526728839956e-05, "Full-finetune/Loss": 0.8024672865867615, "Full-finetune/Loss (Raw)": 0.801461398601532, "Full-finetune/Step": 2396, "Full-finetune/Step Time": 6.661901032552123} {"Full-finetune/Learning Rate": 1.1732391647784618e-05, "Full-finetune/Loss": 0.8021979928016663, "Full-finetune/Loss (Raw)": 0.859150230884552, "Full-finetune/Step": 2397, "Full-finetune/Step Time": 6.659885942935944} {"Full-finetune/Learning Rate": 1.1726255894421908e-05, "Full-finetune/Loss": 0.8054167032241821, "Full-finetune/Loss (Raw)": 1.0118016004562378, "Full-finetune/Step": 2398, "Full-finetune/Step Time": 6.651690425351262} {"Full-finetune/Learning Rate": 1.1720119471132985e-05, "Full-finetune/Loss": 0.8055967688560486, "Full-finetune/Loss (Raw)": 0.9125424027442932, "Full-finetune/Step": 2399, "Full-finetune/Step Time": 6.657836200669408} {"Full-finetune/Learning Rate": 1.1713982380299278e-05, "Full-finetune/Loss": 0.8043665885925293, "Full-finetune/Loss (Raw)": 0.7356420755386353, "Full-finetune/Step": 2400, "Full-finetune/Step Time": 6.650002034381032} {"Full-finetune/Learning Rate": 1.1707844624302471e-05, "Full-finetune/Loss": 0.8044861555099487, "Full-finetune/Loss (Raw)": 0.8117890954017639, "Full-finetune/Step": 2401, "Full-finetune/Step Time": 6.654438687488437} {"Full-finetune/Learning Rate": 1.170170620552451e-05, "Full-finetune/Loss": 0.8041032552719116, "Full-finetune/Loss (Raw)": 0.709191083908081, "Full-finetune/Step": 2402, "Full-finetune/Step Time": 6.639343032613397} {"Full-finetune/Learning Rate": 1.1695567126347593e-05, "Full-finetune/Loss": 0.8034680485725403, "Full-finetune/Loss (Raw)": 0.840156614780426, "Full-finetune/Step": 2403, "Full-finetune/Step Time": 6.647711638361216} {"Full-finetune/Learning Rate": 1.1689427389154176e-05, "Full-finetune/Loss": 0.8034834861755371, "Full-finetune/Loss (Raw)": 0.8518868684768677, "Full-finetune/Step": 2404, "Full-finetune/Step Time": 6.646106090396643} {"Full-finetune/Learning Rate": 1.1683286996326975e-05, "Full-finetune/Loss": 0.8043453693389893, "Full-finetune/Loss (Raw)": 0.7905285358428955, "Full-finetune/Step": 2405, "Full-finetune/Step Time": 6.641040554270148} {"Full-finetune/Learning Rate": 1.1677145950248954e-05, "Full-finetune/Loss": 0.8047037124633789, "Full-finetune/Loss (Raw)": 0.9369891285896301, "Full-finetune/Step": 2406, "Full-finetune/Step Time": 6.641899997368455} {"Full-finetune/Learning Rate": 1.167100425330333e-05, "Full-finetune/Loss": 0.80235356092453, "Full-finetune/Loss (Raw)": 0.5658528208732605, "Full-finetune/Step": 2407, "Full-finetune/Step Time": 6.663078939542174} {"Full-finetune/Learning Rate": 1.1664861907873583e-05, "Full-finetune/Loss": 0.8037323951721191, "Full-finetune/Loss (Raw)": 1.0219022035598755, "Full-finetune/Step": 2408, "Full-finetune/Step Time": 6.681790331378579} {"Full-finetune/Learning Rate": 1.165871891634343e-05, "Full-finetune/Loss": 0.8036782741546631, "Full-finetune/Loss (Raw)": 0.8162294626235962, "Full-finetune/Step": 2409, "Full-finetune/Step Time": 6.681742265820503} {"Full-finetune/Learning Rate": 1.165257528109685e-05, "Full-finetune/Loss": 0.8036761283874512, "Full-finetune/Loss (Raw)": 0.8041014075279236, "Full-finetune/Step": 2410, "Full-finetune/Step Time": 6.6905297841876745} {"Full-finetune/Learning Rate": 1.1646431004518068e-05, "Full-finetune/Loss": 0.8014793992042542, "Full-finetune/Loss (Raw)": 0.6738624572753906, "Full-finetune/Step": 2411, "Full-finetune/Step Time": 6.697167374193668} {"Full-finetune/Learning Rate": 1.1640286088991558e-05, "Full-finetune/Loss": 0.8024226427078247, "Full-finetune/Loss (Raw)": 0.7535669803619385, "Full-finetune/Step": 2412, "Full-finetune/Step Time": 6.675355143845081} {"Full-finetune/Learning Rate": 1.1634140536902039e-05, "Full-finetune/Loss": 0.8021615147590637, "Full-finetune/Loss (Raw)": 0.7490537762641907, "Full-finetune/Step": 2413, "Full-finetune/Step Time": 6.6645874716341496} {"Full-finetune/Learning Rate": 1.1627994350634482e-05, "Full-finetune/Loss": 0.8029594421386719, "Full-finetune/Loss (Raw)": 0.8804044723510742, "Full-finetune/Step": 2414, "Full-finetune/Step Time": 6.651116030290723} {"Full-finetune/Learning Rate": 1.1621847532574101e-05, "Full-finetune/Loss": 0.8048491477966309, "Full-finetune/Loss (Raw)": 0.8276606202125549, "Full-finetune/Step": 2415, "Full-finetune/Step Time": 6.651604287326336} {"Full-finetune/Learning Rate": 1.1615700085106357e-05, "Full-finetune/Loss": 0.8036811351776123, "Full-finetune/Loss (Raw)": 0.7768718600273132, "Full-finetune/Step": 2416, "Full-finetune/Step Time": 6.648950757458806} {"Full-finetune/Learning Rate": 1.1609552010616953e-05, "Full-finetune/Loss": 0.8046889305114746, "Full-finetune/Loss (Raw)": 0.7523704171180725, "Full-finetune/Step": 2417, "Full-finetune/Step Time": 6.647045021876693} {"Full-finetune/Learning Rate": 1.1603403311491841e-05, "Full-finetune/Loss": 0.8009880781173706, "Full-finetune/Loss (Raw)": 0.4487391710281372, "Full-finetune/Step": 2418, "Full-finetune/Step Time": 6.664250545203686} {"Full-finetune/Learning Rate": 1.1597253990117205e-05, "Full-finetune/Loss": 0.8009752035140991, "Full-finetune/Loss (Raw)": 0.7731378674507141, "Full-finetune/Step": 2419, "Full-finetune/Step Time": 6.672540370374918} {"Full-finetune/Learning Rate": 1.1591104048879479e-05, "Full-finetune/Loss": 0.802873969078064, "Full-finetune/Loss (Raw)": 0.9872578978538513, "Full-finetune/Step": 2420, "Full-finetune/Step Time": 6.661994446069002} {"Full-finetune/Learning Rate": 1.1584953490165334e-05, "Full-finetune/Loss": 0.803537130355835, "Full-finetune/Loss (Raw)": 0.6519597172737122, "Full-finetune/Step": 2421, "Full-finetune/Step Time": 6.671809189021587} {"Full-finetune/Learning Rate": 1.1578802316361685e-05, "Full-finetune/Loss": 0.8020026683807373, "Full-finetune/Loss (Raw)": 0.6649697422981262, "Full-finetune/Step": 2422, "Full-finetune/Step Time": 6.681511849164963} {"Full-finetune/Learning Rate": 1.157265052985568e-05, "Full-finetune/Loss": 0.8004732728004456, "Full-finetune/Loss (Raw)": 0.596584677696228, "Full-finetune/Step": 2423, "Full-finetune/Step Time": 6.664128055796027} {"Full-finetune/Learning Rate": 1.156649813303471e-05, "Full-finetune/Loss": 0.801982581615448, "Full-finetune/Loss (Raw)": 0.8888383507728577, "Full-finetune/Step": 2424, "Full-finetune/Step Time": 6.664343278855085} {"Full-finetune/Learning Rate": 1.1560345128286396e-05, "Full-finetune/Loss": 0.8011634945869446, "Full-finetune/Loss (Raw)": 0.6971827745437622, "Full-finetune/Step": 2425, "Full-finetune/Step Time": 6.685822376981378} {"Full-finetune/Learning Rate": 1.1554191517998598e-05, "Full-finetune/Loss": 0.7978849411010742, "Full-finetune/Loss (Raw)": 0.3820362091064453, "Full-finetune/Step": 2426, "Full-finetune/Step Time": 6.702064450830221} {"Full-finetune/Learning Rate": 1.1548037304559419e-05, "Full-finetune/Loss": 0.7975752353668213, "Full-finetune/Loss (Raw)": 0.8420235514640808, "Full-finetune/Step": 2427, "Full-finetune/Step Time": 6.694514114409685} {"Full-finetune/Learning Rate": 1.1541882490357183e-05, "Full-finetune/Loss": 0.7960972785949707, "Full-finetune/Loss (Raw)": 0.7006691694259644, "Full-finetune/Step": 2428, "Full-finetune/Step Time": 6.689208388328552} {"Full-finetune/Learning Rate": 1.1535727077780456e-05, "Full-finetune/Loss": 0.7965538501739502, "Full-finetune/Loss (Raw)": 0.8222599029541016, "Full-finetune/Step": 2429, "Full-finetune/Step Time": 6.701530616730452} {"Full-finetune/Learning Rate": 1.1529571069218033e-05, "Full-finetune/Loss": 0.7962768077850342, "Full-finetune/Loss (Raw)": 0.6940549612045288, "Full-finetune/Step": 2430, "Full-finetune/Step Time": 6.725684642791748} {"Full-finetune/Learning Rate": 1.152341446705894e-05, "Full-finetune/Loss": 0.7965801954269409, "Full-finetune/Loss (Raw)": 0.8933916091918945, "Full-finetune/Step": 2431, "Full-finetune/Step Time": 6.735977806150913} {"Full-finetune/Learning Rate": 1.1517257273692432e-05, "Full-finetune/Loss": 0.7978197932243347, "Full-finetune/Loss (Raw)": 0.7996103167533875, "Full-finetune/Step": 2432, "Full-finetune/Step Time": 6.7387638334184885} {"Full-finetune/Learning Rate": 1.1511099491508003e-05, "Full-finetune/Loss": 0.7977350950241089, "Full-finetune/Loss (Raw)": 0.7446393370628357, "Full-finetune/Step": 2433, "Full-finetune/Step Time": 6.738248609006405} {"Full-finetune/Learning Rate": 1.1504941122895361e-05, "Full-finetune/Loss": 0.7970179915428162, "Full-finetune/Loss (Raw)": 0.8506155610084534, "Full-finetune/Step": 2434, "Full-finetune/Step Time": 6.737429929897189} {"Full-finetune/Learning Rate": 1.1498782170244449e-05, "Full-finetune/Loss": 0.7965975403785706, "Full-finetune/Loss (Raw)": 0.811649739742279, "Full-finetune/Step": 2435, "Full-finetune/Step Time": 6.747039023786783} {"Full-finetune/Learning Rate": 1.1492622635945438e-05, "Full-finetune/Loss": 0.7993999719619751, "Full-finetune/Loss (Raw)": 0.8820415139198303, "Full-finetune/Step": 2436, "Full-finetune/Step Time": 6.741266764700413} {"Full-finetune/Learning Rate": 1.1486462522388724e-05, "Full-finetune/Loss": 0.7993333339691162, "Full-finetune/Loss (Raw)": 0.7642065286636353, "Full-finetune/Step": 2437, "Full-finetune/Step Time": 6.74726390093565} {"Full-finetune/Learning Rate": 1.1480301831964926e-05, "Full-finetune/Loss": 0.7995833158493042, "Full-finetune/Loss (Raw)": 0.925489604473114, "Full-finetune/Step": 2438, "Full-finetune/Step Time": 6.751360837370157} {"Full-finetune/Learning Rate": 1.1474140567064889e-05, "Full-finetune/Loss": 0.7996020913124084, "Full-finetune/Loss (Raw)": 0.9008498191833496, "Full-finetune/Step": 2439, "Full-finetune/Step Time": 6.752388536930084} {"Full-finetune/Learning Rate": 1.1467978730079677e-05, "Full-finetune/Loss": 0.8012627363204956, "Full-finetune/Loss (Raw)": 0.9087226986885071, "Full-finetune/Step": 2440, "Full-finetune/Step Time": 6.733066340908408} {"Full-finetune/Learning Rate": 1.1461816323400579e-05, "Full-finetune/Loss": 0.8022787570953369, "Full-finetune/Loss (Raw)": 0.9242477416992188, "Full-finetune/Step": 2441, "Full-finetune/Step Time": 6.729937283322215} {"Full-finetune/Learning Rate": 1.1455653349419105e-05, "Full-finetune/Loss": 0.802889347076416, "Full-finetune/Loss (Raw)": 0.8373119235038757, "Full-finetune/Step": 2442, "Full-finetune/Step Time": 6.71487545594573} {"Full-finetune/Learning Rate": 1.1449489810526985e-05, "Full-finetune/Loss": 0.8039986491203308, "Full-finetune/Loss (Raw)": 0.9202196002006531, "Full-finetune/Step": 2443, "Full-finetune/Step Time": 6.721622388809919} {"Full-finetune/Learning Rate": 1.144332570911617e-05, "Full-finetune/Loss": 0.8034666180610657, "Full-finetune/Loss (Raw)": 0.848906934261322, "Full-finetune/Step": 2444, "Full-finetune/Step Time": 6.718856159597635} {"Full-finetune/Learning Rate": 1.1437161047578826e-05, "Full-finetune/Loss": 0.8004889488220215, "Full-finetune/Loss (Raw)": 0.3990862965583801, "Full-finetune/Step": 2445, "Full-finetune/Step Time": 6.733507260680199} {"Full-finetune/Learning Rate": 1.1430995828307337e-05, "Full-finetune/Loss": 0.8016537427902222, "Full-finetune/Loss (Raw)": 0.8788216710090637, "Full-finetune/Step": 2446, "Full-finetune/Step Time": 6.736629897728562} {"Full-finetune/Learning Rate": 1.1424830053694303e-05, "Full-finetune/Loss": 0.8008547425270081, "Full-finetune/Loss (Raw)": 0.8430296182632446, "Full-finetune/Step": 2447, "Full-finetune/Step Time": 6.743146490305662} {"Full-finetune/Learning Rate": 1.1418663726132543e-05, "Full-finetune/Loss": 0.8017246723175049, "Full-finetune/Loss (Raw)": 0.9116612076759338, "Full-finetune/Step": 2448, "Full-finetune/Step Time": 6.744520319625735} {"Full-finetune/Learning Rate": 1.141249684801509e-05, "Full-finetune/Loss": 0.8009546995162964, "Full-finetune/Loss (Raw)": 0.7195496559143066, "Full-finetune/Step": 2449, "Full-finetune/Step Time": 6.737687343731523} {"Full-finetune/Learning Rate": 1.140632942173519e-05, "Full-finetune/Loss": 0.7982256412506104, "Full-finetune/Loss (Raw)": 0.49326926469802856, "Full-finetune/Step": 2450, "Full-finetune/Step Time": 6.753330597653985} {"Full-finetune/Learning Rate": 1.1400161449686293e-05, "Full-finetune/Loss": 0.7953620553016663, "Full-finetune/Loss (Raw)": 0.32834941148757935, "Full-finetune/Step": 2451, "Full-finetune/Step Time": 6.761004664003849} {"Full-finetune/Learning Rate": 1.1393992934262073e-05, "Full-finetune/Loss": 0.7953568696975708, "Full-finetune/Loss (Raw)": 0.7743340134620667, "Full-finetune/Step": 2452, "Full-finetune/Step Time": 6.764101352542639} {"Full-finetune/Learning Rate": 1.1387823877856413e-05, "Full-finetune/Loss": 0.7947891354560852, "Full-finetune/Loss (Raw)": 0.7908214330673218, "Full-finetune/Step": 2453, "Full-finetune/Step Time": 6.7826072089374065} {"Full-finetune/Learning Rate": 1.1381654282863399e-05, "Full-finetune/Loss": 0.7977192401885986, "Full-finetune/Loss (Raw)": 0.8229445219039917, "Full-finetune/Step": 2454, "Full-finetune/Step Time": 6.775321615859866} {"Full-finetune/Learning Rate": 1.1375484151677334e-05, "Full-finetune/Loss": 0.7979786992073059, "Full-finetune/Loss (Raw)": 0.8723763227462769, "Full-finetune/Step": 2455, "Full-finetune/Step Time": 6.761574620380998} {"Full-finetune/Learning Rate": 1.1369313486692723e-05, "Full-finetune/Loss": 0.7991248369216919, "Full-finetune/Loss (Raw)": 0.8750165700912476, "Full-finetune/Step": 2456, "Full-finetune/Step Time": 6.752361502498388} {"Full-finetune/Learning Rate": 1.136314229030428e-05, "Full-finetune/Loss": 0.7985837459564209, "Full-finetune/Loss (Raw)": 0.8470889925956726, "Full-finetune/Step": 2457, "Full-finetune/Step Time": 6.749607045203447} {"Full-finetune/Learning Rate": 1.1356970564906928e-05, "Full-finetune/Loss": 0.7992945909500122, "Full-finetune/Loss (Raw)": 0.8067057728767395, "Full-finetune/Step": 2458, "Full-finetune/Step Time": 6.756249258294702} {"Full-finetune/Learning Rate": 1.1350798312895795e-05, "Full-finetune/Loss": 0.7988828420639038, "Full-finetune/Loss (Raw)": 0.7695019245147705, "Full-finetune/Step": 2459, "Full-finetune/Step Time": 6.751019641757011} {"Full-finetune/Learning Rate": 1.1344625536666208e-05, "Full-finetune/Loss": 0.7969683408737183, "Full-finetune/Loss (Raw)": 0.7616773247718811, "Full-finetune/Step": 2460, "Full-finetune/Step Time": 6.744716487824917} {"Full-finetune/Learning Rate": 1.13384522386137e-05, "Full-finetune/Loss": 0.797529935836792, "Full-finetune/Loss (Raw)": 0.9416506886482239, "Full-finetune/Step": 2461, "Full-finetune/Step Time": 6.7424702905118465} {"Full-finetune/Learning Rate": 1.1332278421134007e-05, "Full-finetune/Loss": 0.7976641654968262, "Full-finetune/Loss (Raw)": 0.8546949028968811, "Full-finetune/Step": 2462, "Full-finetune/Step Time": 6.740134922787547} {"Full-finetune/Learning Rate": 1.1326104086623075e-05, "Full-finetune/Loss": 0.7972714900970459, "Full-finetune/Loss (Raw)": 0.7756047248840332, "Full-finetune/Step": 2463, "Full-finetune/Step Time": 6.7405542191118} {"Full-finetune/Learning Rate": 1.1319929237477033e-05, "Full-finetune/Loss": 0.7979362607002258, "Full-finetune/Loss (Raw)": 0.8926640152931213, "Full-finetune/Step": 2464, "Full-finetune/Step Time": 6.725478634238243} {"Full-finetune/Learning Rate": 1.1313753876092223e-05, "Full-finetune/Loss": 0.7970737218856812, "Full-finetune/Loss (Raw)": 0.8269508481025696, "Full-finetune/Step": 2465, "Full-finetune/Step Time": 6.727336769923568} {"Full-finetune/Learning Rate": 1.1307578004865188e-05, "Full-finetune/Loss": 0.7962844967842102, "Full-finetune/Loss (Raw)": 0.7617014050483704, "Full-finetune/Step": 2466, "Full-finetune/Step Time": 6.728618115186691} {"Full-finetune/Learning Rate": 1.130140162619265e-05, "Full-finetune/Loss": 0.7969546914100647, "Full-finetune/Loss (Raw)": 0.8126506209373474, "Full-finetune/Step": 2467, "Full-finetune/Step Time": 6.707967154681683} {"Full-finetune/Learning Rate": 1.1295224742471555e-05, "Full-finetune/Loss": 0.7982343435287476, "Full-finetune/Loss (Raw)": 0.9242181181907654, "Full-finetune/Step": 2468, "Full-finetune/Step Time": 6.71111673861742} {"Full-finetune/Learning Rate": 1.128904735609902e-05, "Full-finetune/Loss": 0.7982791662216187, "Full-finetune/Loss (Raw)": 0.8937535881996155, "Full-finetune/Step": 2469, "Full-finetune/Step Time": 6.6823407001793385} {"Full-finetune/Learning Rate": 1.1282869469472375e-05, "Full-finetune/Loss": 0.8012821078300476, "Full-finetune/Loss (Raw)": 0.7528886795043945, "Full-finetune/Step": 2470, "Full-finetune/Step Time": 6.67133067175746} {"Full-finetune/Learning Rate": 1.1276691084989135e-05, "Full-finetune/Loss": 0.8011589646339417, "Full-finetune/Loss (Raw)": 0.9062379598617554, "Full-finetune/Step": 2471, "Full-finetune/Step Time": 6.652212806046009} {"Full-finetune/Learning Rate": 1.1270512205047008e-05, "Full-finetune/Loss": 0.801679790019989, "Full-finetune/Loss (Raw)": 0.8338210582733154, "Full-finetune/Step": 2472, "Full-finetune/Step Time": 6.626096468418837} {"Full-finetune/Learning Rate": 1.1264332832043902e-05, "Full-finetune/Loss": 0.8004392385482788, "Full-finetune/Loss (Raw)": 0.8719844818115234, "Full-finetune/Step": 2473, "Full-finetune/Step Time": 6.632084913551807} {"Full-finetune/Learning Rate": 1.1258152968377907e-05, "Full-finetune/Loss": 0.8003652691841125, "Full-finetune/Loss (Raw)": 0.7970017790794373, "Full-finetune/Step": 2474, "Full-finetune/Step Time": 6.635380517691374} {"Full-finetune/Learning Rate": 1.125197261644731e-05, "Full-finetune/Loss": 0.8001554012298584, "Full-finetune/Loss (Raw)": 0.8096036911010742, "Full-finetune/Step": 2475, "Full-finetune/Step Time": 6.635013313964009} {"Full-finetune/Learning Rate": 1.1245791778650578e-05, "Full-finetune/Loss": 0.8003500699996948, "Full-finetune/Loss (Raw)": 0.8619083166122437, "Full-finetune/Step": 2476, "Full-finetune/Step Time": 6.632082311436534} {"Full-finetune/Learning Rate": 1.1239610457386381e-05, "Full-finetune/Loss": 0.8008005619049072, "Full-finetune/Loss (Raw)": 0.829006552696228, "Full-finetune/Step": 2477, "Full-finetune/Step Time": 6.629073474556208} {"Full-finetune/Learning Rate": 1.1233428655053572e-05, "Full-finetune/Loss": 0.8007239699363708, "Full-finetune/Loss (Raw)": 0.8163360953330994, "Full-finetune/Step": 2478, "Full-finetune/Step Time": 6.627605952322483} {"Full-finetune/Learning Rate": 1.1227246374051175e-05, "Full-finetune/Loss": 0.7991045713424683, "Full-finetune/Loss (Raw)": 0.7193167209625244, "Full-finetune/Step": 2479, "Full-finetune/Step Time": 6.626647170633078} {"Full-finetune/Learning Rate": 1.1221063616778425e-05, "Full-finetune/Loss": 0.7979898452758789, "Full-finetune/Loss (Raw)": 0.7002544403076172, "Full-finetune/Step": 2480, "Full-finetune/Step Time": 6.613957326859236} {"Full-finetune/Learning Rate": 1.1214880385634723e-05, "Full-finetune/Loss": 0.7943664789199829, "Full-finetune/Loss (Raw)": 0.38423314690589905, "Full-finetune/Step": 2481, "Full-finetune/Step Time": 6.627638511359692} {"Full-finetune/Learning Rate": 1.120869668301966e-05, "Full-finetune/Loss": 0.795742392539978, "Full-finetune/Loss (Raw)": 1.022301435470581, "Full-finetune/Step": 2482, "Full-finetune/Step Time": 6.625866048038006} {"Full-finetune/Learning Rate": 1.1202512511333017e-05, "Full-finetune/Loss": 0.7954979538917542, "Full-finetune/Loss (Raw)": 0.7524157166481018, "Full-finetune/Step": 2483, "Full-finetune/Step Time": 6.6301034316420555} {"Full-finetune/Learning Rate": 1.119632787297474e-05, "Full-finetune/Loss": 0.795811653137207, "Full-finetune/Loss (Raw)": 0.7266433238983154, "Full-finetune/Step": 2484, "Full-finetune/Step Time": 6.633494671434164} {"Full-finetune/Learning Rate": 1.1190142770344976e-05, "Full-finetune/Loss": 0.7959999442100525, "Full-finetune/Loss (Raw)": 0.8445776104927063, "Full-finetune/Step": 2485, "Full-finetune/Step Time": 6.634546663612127} {"Full-finetune/Learning Rate": 1.1183957205844037e-05, "Full-finetune/Loss": 0.7939722537994385, "Full-finetune/Loss (Raw)": 0.6477568745613098, "Full-finetune/Step": 2486, "Full-finetune/Step Time": 6.643293436616659} {"Full-finetune/Learning Rate": 1.1177771181872416e-05, "Full-finetune/Loss": 0.7941586971282959, "Full-finetune/Loss (Raw)": 0.3225262463092804, "Full-finetune/Step": 2487, "Full-finetune/Step Time": 6.641424056142569} {"Full-finetune/Learning Rate": 1.1171584700830803e-05, "Full-finetune/Loss": 0.7943537831306458, "Full-finetune/Loss (Raw)": 0.9080498218536377, "Full-finetune/Step": 2488, "Full-finetune/Step Time": 6.640841292217374} {"Full-finetune/Learning Rate": 1.1165397765120035e-05, "Full-finetune/Loss": 0.7941622138023376, "Full-finetune/Loss (Raw)": 0.8636196255683899, "Full-finetune/Step": 2489, "Full-finetune/Step Time": 6.657467111945152} {"Full-finetune/Learning Rate": 1.115921037714115e-05, "Full-finetune/Loss": 0.7902230620384216, "Full-finetune/Loss (Raw)": 0.3194769024848938, "Full-finetune/Step": 2490, "Full-finetune/Step Time": 6.679159471765161} {"Full-finetune/Learning Rate": 1.1153022539295345e-05, "Full-finetune/Loss": 0.7921469211578369, "Full-finetune/Loss (Raw)": 0.7836686968803406, "Full-finetune/Step": 2491, "Full-finetune/Step Time": 6.693628180772066} {"Full-finetune/Learning Rate": 1.1146834253984008e-05, "Full-finetune/Loss": 0.7919718027114868, "Full-finetune/Loss (Raw)": 0.7177945375442505, "Full-finetune/Step": 2492, "Full-finetune/Step Time": 6.694777576252818} {"Full-finetune/Learning Rate": 1.1140645523608684e-05, "Full-finetune/Loss": 0.7917468547821045, "Full-finetune/Loss (Raw)": 0.8437073826789856, "Full-finetune/Step": 2493, "Full-finetune/Step Time": 6.701586050912738} {"Full-finetune/Learning Rate": 1.1134456350571099e-05, "Full-finetune/Loss": 0.7945524454116821, "Full-finetune/Loss (Raw)": 0.929932713508606, "Full-finetune/Step": 2494, "Full-finetune/Step Time": 6.689788283780217} {"Full-finetune/Learning Rate": 1.112826673727316e-05, "Full-finetune/Loss": 0.7939871549606323, "Full-finetune/Loss (Raw)": 0.8548046946525574, "Full-finetune/Step": 2495, "Full-finetune/Step Time": 6.682420575991273} {"Full-finetune/Learning Rate": 1.112207668611692e-05, "Full-finetune/Loss": 0.7938183546066284, "Full-finetune/Loss (Raw)": 0.7596627473831177, "Full-finetune/Step": 2496, "Full-finetune/Step Time": 6.702415086328983} {"Full-finetune/Learning Rate": 1.1115886199504631e-05, "Full-finetune/Loss": 0.791703999042511, "Full-finetune/Loss (Raw)": 0.6803382635116577, "Full-finetune/Step": 2497, "Full-finetune/Step Time": 6.690019519999623} {"Full-finetune/Learning Rate": 1.110969527983869e-05, "Full-finetune/Loss": 0.7926520109176636, "Full-finetune/Loss (Raw)": 0.9351413249969482, "Full-finetune/Step": 2498, "Full-finetune/Step Time": 6.681281825527549} {"Full-finetune/Learning Rate": 1.1103503929521677e-05, "Full-finetune/Loss": 0.7893964052200317, "Full-finetune/Loss (Raw)": 0.5057010650634766, "Full-finetune/Step": 2499, "Full-finetune/Step Time": 6.681272925809026} {"Full-finetune/Learning Rate": 1.1097312150956336e-05, "Full-finetune/Loss": 0.7901525497436523, "Full-finetune/Loss (Raw)": 0.8613641262054443, "Full-finetune/Step": 2500, "Full-finetune/Step Time": 6.681096909567714} {"Full-finetune/Learning Rate": 1.1091119946545568e-05, "Full-finetune/Loss": 0.789920449256897, "Full-finetune/Loss (Raw)": 0.8942145109176636, "Full-finetune/Step": 2501, "Full-finetune/Step Time": 6.662362199276686} {"Full-finetune/Learning Rate": 1.1084927318692453e-05, "Full-finetune/Loss": 0.7910595536231995, "Full-finetune/Loss (Raw)": 0.8731211423873901, "Full-finetune/Step": 2502, "Full-finetune/Step Time": 6.662365570664406} {"Full-finetune/Learning Rate": 1.1078734269800227e-05, "Full-finetune/Loss": 0.7924778461456299, "Full-finetune/Loss (Raw)": 0.8186615109443665, "Full-finetune/Step": 2503, "Full-finetune/Step Time": 6.652069924399257} {"Full-finetune/Learning Rate": 1.1072540802272291e-05, "Full-finetune/Loss": 0.791928768157959, "Full-finetune/Loss (Raw)": 0.8392397165298462, "Full-finetune/Step": 2504, "Full-finetune/Step Time": 6.657307285815477} {"Full-finetune/Learning Rate": 1.1066346918512211e-05, "Full-finetune/Loss": 0.7908006906509399, "Full-finetune/Loss (Raw)": 0.8285489678382874, "Full-finetune/Step": 2505, "Full-finetune/Step Time": 6.656249629333615} {"Full-finetune/Learning Rate": 1.1060152620923707e-05, "Full-finetune/Loss": 0.7897911071777344, "Full-finetune/Loss (Raw)": 0.7840871810913086, "Full-finetune/Step": 2506, "Full-finetune/Step Time": 6.659812219440937} {"Full-finetune/Learning Rate": 1.105395791191067e-05, "Full-finetune/Loss": 0.7902393937110901, "Full-finetune/Loss (Raw)": 0.9166094064712524, "Full-finetune/Step": 2507, "Full-finetune/Step Time": 6.647831939160824} {"Full-finetune/Learning Rate": 1.1047762793877146e-05, "Full-finetune/Loss": 0.790266215801239, "Full-finetune/Loss (Raw)": 0.7311996221542358, "Full-finetune/Step": 2508, "Full-finetune/Step Time": 6.655236089602113} {"Full-finetune/Learning Rate": 1.1041567269227336e-05, "Full-finetune/Loss": 0.7900168895721436, "Full-finetune/Loss (Raw)": 0.7931290864944458, "Full-finetune/Step": 2509, "Full-finetune/Step Time": 6.658054685220122} {"Full-finetune/Learning Rate": 1.1035371340365606e-05, "Full-finetune/Loss": 0.7919495105743408, "Full-finetune/Loss (Raw)": 0.9158990979194641, "Full-finetune/Step": 2510, "Full-finetune/Step Time": 6.6425050515681505} {"Full-finetune/Learning Rate": 1.1029175009696468e-05, "Full-finetune/Loss": 0.7942907810211182, "Full-finetune/Loss (Raw)": 0.8667076826095581, "Full-finetune/Step": 2511, "Full-finetune/Step Time": 6.639530936256051} {"Full-finetune/Learning Rate": 1.1022978279624606e-05, "Full-finetune/Loss": 0.7922652363777161, "Full-finetune/Loss (Raw)": 0.7010650038719177, "Full-finetune/Step": 2512, "Full-finetune/Step Time": 6.64892815425992} {"Full-finetune/Learning Rate": 1.1016781152554846e-05, "Full-finetune/Loss": 0.791692852973938, "Full-finetune/Loss (Raw)": 0.7327619194984436, "Full-finetune/Step": 2513, "Full-finetune/Step Time": 6.648960039019585} {"Full-finetune/Learning Rate": 1.1010583630892166e-05, "Full-finetune/Loss": 0.7906999588012695, "Full-finetune/Loss (Raw)": 0.8547472953796387, "Full-finetune/Step": 2514, "Full-finetune/Step Time": 6.6547212563455105} {"Full-finetune/Learning Rate": 1.1004385717041716e-05, "Full-finetune/Loss": 0.7893173098564148, "Full-finetune/Loss (Raw)": 0.6391555666923523, "Full-finetune/Step": 2515, "Full-finetune/Step Time": 6.671585725620389} {"Full-finetune/Learning Rate": 1.0998187413408774e-05, "Full-finetune/Loss": 0.7891018390655518, "Full-finetune/Loss (Raw)": 0.7856131792068481, "Full-finetune/Step": 2516, "Full-finetune/Step Time": 6.650063026696444} {"Full-finetune/Learning Rate": 1.0991988722398786e-05, "Full-finetune/Loss": 0.7882359027862549, "Full-finetune/Loss (Raw)": 0.8458489179611206, "Full-finetune/Step": 2517, "Full-finetune/Step Time": 6.654113722965121} {"Full-finetune/Learning Rate": 1.0985789646417341e-05, "Full-finetune/Loss": 0.7867968678474426, "Full-finetune/Loss (Raw)": 0.7415456771850586, "Full-finetune/Step": 2518, "Full-finetune/Step Time": 6.660378599539399} {"Full-finetune/Learning Rate": 1.097959018787018e-05, "Full-finetune/Loss": 0.7857775092124939, "Full-finetune/Loss (Raw)": 0.7845055460929871, "Full-finetune/Step": 2519, "Full-finetune/Step Time": 6.665312338620424} {"Full-finetune/Learning Rate": 1.0973390349163195e-05, "Full-finetune/Loss": 0.7863309979438782, "Full-finetune/Loss (Raw)": 0.9363141059875488, "Full-finetune/Step": 2520, "Full-finetune/Step Time": 6.664939476177096} {"Full-finetune/Learning Rate": 1.0967190132702414e-05, "Full-finetune/Loss": 0.7823458909988403, "Full-finetune/Loss (Raw)": 0.4098040759563446, "Full-finetune/Step": 2521, "Full-finetune/Step Time": 6.667741542682052} {"Full-finetune/Learning Rate": 1.096098954089403e-05, "Full-finetune/Loss": 0.7857155799865723, "Full-finetune/Loss (Raw)": 0.834911584854126, "Full-finetune/Step": 2522, "Full-finetune/Step Time": 6.662115724757314} {"Full-finetune/Learning Rate": 1.0954788576144365e-05, "Full-finetune/Loss": 0.7837050557136536, "Full-finetune/Loss (Raw)": 0.6017270684242249, "Full-finetune/Step": 2523, "Full-finetune/Step Time": 6.661797625944018} {"Full-finetune/Learning Rate": 1.0948587240859893e-05, "Full-finetune/Loss": 0.7835423350334167, "Full-finetune/Loss (Raw)": 0.7806267142295837, "Full-finetune/Step": 2524, "Full-finetune/Step Time": 6.662017831578851} {"Full-finetune/Learning Rate": 1.0942385537447236e-05, "Full-finetune/Loss": 0.7815767526626587, "Full-finetune/Loss (Raw)": 0.6075663566589355, "Full-finetune/Step": 2525, "Full-finetune/Step Time": 6.675291473045945} {"Full-finetune/Learning Rate": 1.0936183468313147e-05, "Full-finetune/Loss": 0.780414879322052, "Full-finetune/Loss (Raw)": 0.863075315952301, "Full-finetune/Step": 2526, "Full-finetune/Step Time": 6.678295284509659} {"Full-finetune/Learning Rate": 1.0929981035864532e-05, "Full-finetune/Loss": 0.7797343730926514, "Full-finetune/Loss (Raw)": 0.8254397511482239, "Full-finetune/Step": 2527, "Full-finetune/Step Time": 6.687740679830313} {"Full-finetune/Learning Rate": 1.0923778242508437e-05, "Full-finetune/Loss": 0.7813080549240112, "Full-finetune/Loss (Raw)": 0.9370753169059753, "Full-finetune/Step": 2528, "Full-finetune/Step Time": 6.694672787562013} {"Full-finetune/Learning Rate": 1.0917575090652038e-05, "Full-finetune/Loss": 0.7771151065826416, "Full-finetune/Loss (Raw)": 0.2750939428806305, "Full-finetune/Step": 2529, "Full-finetune/Step Time": 6.709822837263346} {"Full-finetune/Learning Rate": 1.091137158270266e-05, "Full-finetune/Loss": 0.7783724665641785, "Full-finetune/Loss (Raw)": 0.8701255321502686, "Full-finetune/Step": 2530, "Full-finetune/Step Time": 6.703527146950364} {"Full-finetune/Learning Rate": 1.0905167721067762e-05, "Full-finetune/Loss": 0.7778249382972717, "Full-finetune/Loss (Raw)": 0.7700729370117188, "Full-finetune/Step": 2531, "Full-finetune/Step Time": 6.703002452850342} {"Full-finetune/Learning Rate": 1.0898963508154947e-05, "Full-finetune/Loss": 0.7773810625076294, "Full-finetune/Loss (Raw)": 0.7950660586357117, "Full-finetune/Step": 2532, "Full-finetune/Step Time": 6.720517521724105} {"Full-finetune/Learning Rate": 1.0892758946371943e-05, "Full-finetune/Loss": 0.7767124772071838, "Full-finetune/Loss (Raw)": 0.7049623727798462, "Full-finetune/Step": 2533, "Full-finetune/Step Time": 6.723716355860233} {"Full-finetune/Learning Rate": 1.0886554038126624e-05, "Full-finetune/Loss": 0.7724095582962036, "Full-finetune/Loss (Raw)": 0.38620299100875854, "Full-finetune/Step": 2534, "Full-finetune/Step Time": 6.740764124318957} {"Full-finetune/Learning Rate": 1.0880348785826991e-05, "Full-finetune/Loss": 0.7751016616821289, "Full-finetune/Loss (Raw)": 0.9104442000389099, "Full-finetune/Step": 2535, "Full-finetune/Step Time": 6.727584118023515} {"Full-finetune/Learning Rate": 1.087414319188118e-05, "Full-finetune/Loss": 0.7720091342926025, "Full-finetune/Loss (Raw)": 0.6260660290718079, "Full-finetune/Step": 2536, "Full-finetune/Step Time": 6.726360164582729} {"Full-finetune/Learning Rate": 1.0867937258697463e-05, "Full-finetune/Loss": 0.7714089155197144, "Full-finetune/Loss (Raw)": 0.7394076585769653, "Full-finetune/Step": 2537, "Full-finetune/Step Time": 6.7331572733819485} {"Full-finetune/Learning Rate": 1.086173098868424e-05, "Full-finetune/Loss": 0.7718791961669922, "Full-finetune/Loss (Raw)": 0.864288330078125, "Full-finetune/Step": 2538, "Full-finetune/Step Time": 6.737513216212392} {"Full-finetune/Learning Rate": 1.0855524384250048e-05, "Full-finetune/Loss": 0.7727974653244019, "Full-finetune/Loss (Raw)": 0.7914004325866699, "Full-finetune/Step": 2539, "Full-finetune/Step Time": 6.73743599653244} {"Full-finetune/Learning Rate": 1.0849317447803542e-05, "Full-finetune/Loss": 0.7733992338180542, "Full-finetune/Loss (Raw)": 0.8305865526199341, "Full-finetune/Step": 2540, "Full-finetune/Step Time": 6.7375997845083475} {"Full-finetune/Learning Rate": 1.0843110181753513e-05, "Full-finetune/Loss": 0.7725543975830078, "Full-finetune/Loss (Raw)": 0.6409223675727844, "Full-finetune/Step": 2541, "Full-finetune/Step Time": 6.754631387069821} {"Full-finetune/Learning Rate": 1.0836902588508888e-05, "Full-finetune/Loss": 0.77215576171875, "Full-finetune/Loss (Raw)": 0.8293734788894653, "Full-finetune/Step": 2542, "Full-finetune/Step Time": 6.759777678176761} {"Full-finetune/Learning Rate": 1.0830694670478706e-05, "Full-finetune/Loss": 0.7722835540771484, "Full-finetune/Loss (Raw)": 0.844017744064331, "Full-finetune/Step": 2543, "Full-finetune/Step Time": 6.754478054121137} {"Full-finetune/Learning Rate": 1.082448643007214e-05, "Full-finetune/Loss": 0.7712268829345703, "Full-finetune/Loss (Raw)": 0.6416224241256714, "Full-finetune/Step": 2544, "Full-finetune/Step Time": 6.764780111610889} {"Full-finetune/Learning Rate": 1.0818277869698488e-05, "Full-finetune/Loss": 0.771589994430542, "Full-finetune/Loss (Raw)": 0.7988569736480713, "Full-finetune/Step": 2545, "Full-finetune/Step Time": 6.759140798822045} {"Full-finetune/Learning Rate": 1.0812068991767171e-05, "Full-finetune/Loss": 0.7703036069869995, "Full-finetune/Loss (Raw)": 0.2840752899646759, "Full-finetune/Step": 2546, "Full-finetune/Step Time": 6.758654721081257} {"Full-finetune/Learning Rate": 1.0805859798687727e-05, "Full-finetune/Loss": 0.7704390287399292, "Full-finetune/Loss (Raw)": 0.7904734015464783, "Full-finetune/Step": 2547, "Full-finetune/Step Time": 6.730799751356244} {"Full-finetune/Learning Rate": 1.0799650292869835e-05, "Full-finetune/Loss": 0.7701886892318726, "Full-finetune/Loss (Raw)": 0.9552110433578491, "Full-finetune/Step": 2548, "Full-finetune/Step Time": 6.744586555287242} {"Full-finetune/Learning Rate": 1.0793440476723274e-05, "Full-finetune/Loss": 0.7676810026168823, "Full-finetune/Loss (Raw)": 0.3309797942638397, "Full-finetune/Step": 2549, "Full-finetune/Step Time": 6.742431027814746} {"Full-finetune/Learning Rate": 1.0787230352657954e-05, "Full-finetune/Loss": 0.7688183784484863, "Full-finetune/Loss (Raw)": 0.8105486035346985, "Full-finetune/Step": 2550, "Full-finetune/Step Time": 6.734138587489724} {"Full-finetune/Learning Rate": 1.0781019923083905e-05, "Full-finetune/Loss": 0.7710247039794922, "Full-finetune/Loss (Raw)": 0.8790007829666138, "Full-finetune/Step": 2551, "Full-finetune/Step Time": 6.7412433829158545} {"Full-finetune/Learning Rate": 1.077480919041127e-05, "Full-finetune/Loss": 0.772190511226654, "Full-finetune/Loss (Raw)": 1.03805410861969, "Full-finetune/Step": 2552, "Full-finetune/Step Time": 6.7426902167499065} {"Full-finetune/Learning Rate": 1.0768598157050313e-05, "Full-finetune/Loss": 0.7723996639251709, "Full-finetune/Loss (Raw)": 0.7239595651626587, "Full-finetune/Step": 2553, "Full-finetune/Step Time": 6.732672117650509} {"Full-finetune/Learning Rate": 1.0762386825411422e-05, "Full-finetune/Loss": 0.7759747505187988, "Full-finetune/Loss (Raw)": 0.8396434187889099, "Full-finetune/Step": 2554, "Full-finetune/Step Time": 6.743433533236384} {"Full-finetune/Learning Rate": 1.0756175197905086e-05, "Full-finetune/Loss": 0.772017240524292, "Full-finetune/Loss (Raw)": 0.3354589343070984, "Full-finetune/Step": 2555, "Full-finetune/Step Time": 6.759606862440705} {"Full-finetune/Learning Rate": 1.0749963276941919e-05, "Full-finetune/Loss": 0.7732051610946655, "Full-finetune/Loss (Raw)": 0.8527300357818604, "Full-finetune/Step": 2556, "Full-finetune/Step Time": 6.761376321315765} {"Full-finetune/Learning Rate": 1.0743751064932648e-05, "Full-finetune/Loss": 0.7722172141075134, "Full-finetune/Loss (Raw)": 0.695801317691803, "Full-finetune/Step": 2557, "Full-finetune/Step Time": 6.744977459311485} {"Full-finetune/Learning Rate": 1.0737538564288104e-05, "Full-finetune/Loss": 0.7730156183242798, "Full-finetune/Loss (Raw)": 0.7962526679039001, "Full-finetune/Step": 2558, "Full-finetune/Step Time": 6.7154744155704975} {"Full-finetune/Learning Rate": 1.0731325777419246e-05, "Full-finetune/Loss": 0.7717381715774536, "Full-finetune/Loss (Raw)": 0.7298805117607117, "Full-finetune/Step": 2559, "Full-finetune/Step Time": 6.703602870926261} {"Full-finetune/Learning Rate": 1.0725112706737133e-05, "Full-finetune/Loss": 0.7700234055519104, "Full-finetune/Loss (Raw)": 0.580114483833313, "Full-finetune/Step": 2560, "Full-finetune/Step Time": 6.712892457842827} {"Full-finetune/Learning Rate": 1.0718899354652931e-05, "Full-finetune/Loss": 0.7707140445709229, "Full-finetune/Loss (Raw)": 0.8330397605895996, "Full-finetune/Step": 2561, "Full-finetune/Step Time": 6.7164998687803745} {"Full-finetune/Learning Rate": 1.071268572357793e-05, "Full-finetune/Loss": 0.771066427230835, "Full-finetune/Loss (Raw)": 0.8957198858261108, "Full-finetune/Step": 2562, "Full-finetune/Step Time": 6.715733136981726} {"Full-finetune/Learning Rate": 1.0706471815923511e-05, "Full-finetune/Loss": 0.7680236101150513, "Full-finetune/Loss (Raw)": 0.4221721589565277, "Full-finetune/Step": 2563, "Full-finetune/Step Time": 6.713833441957831} {"Full-finetune/Learning Rate": 1.070025763410117e-05, "Full-finetune/Loss": 0.7677338123321533, "Full-finetune/Loss (Raw)": 0.8449522852897644, "Full-finetune/Step": 2564, "Full-finetune/Step Time": 6.716455297544599} {"Full-finetune/Learning Rate": 1.0694043180522515e-05, "Full-finetune/Loss": 0.7686870098114014, "Full-finetune/Loss (Raw)": 0.8862094283103943, "Full-finetune/Step": 2565, "Full-finetune/Step Time": 6.714157685637474} {"Full-finetune/Learning Rate": 1.0687828457599254e-05, "Full-finetune/Loss": 0.7679846286773682, "Full-finetune/Loss (Raw)": 0.8355873227119446, "Full-finetune/Step": 2566, "Full-finetune/Step Time": 6.718451561406255} {"Full-finetune/Learning Rate": 1.0681613467743195e-05, "Full-finetune/Loss": 0.7666246294975281, "Full-finetune/Loss (Raw)": 0.7267712950706482, "Full-finetune/Step": 2567, "Full-finetune/Step Time": 6.732591670006514} {"Full-finetune/Learning Rate": 1.0675398213366257e-05, "Full-finetune/Loss": 0.7652680277824402, "Full-finetune/Loss (Raw)": 0.7350753545761108, "Full-finetune/Step": 2568, "Full-finetune/Step Time": 6.736263018101454} {"Full-finetune/Learning Rate": 1.0669182696880459e-05, "Full-finetune/Loss": 0.7659525871276855, "Full-finetune/Loss (Raw)": 1.0118801593780518, "Full-finetune/Step": 2569, "Full-finetune/Step Time": 6.733190448954701} {"Full-finetune/Learning Rate": 1.0662966920697918e-05, "Full-finetune/Loss": 0.7656477689743042, "Full-finetune/Loss (Raw)": 0.7982792854309082, "Full-finetune/Step": 2570, "Full-finetune/Step Time": 6.737517470493913} {"Full-finetune/Learning Rate": 1.0656750887230862e-05, "Full-finetune/Loss": 0.7664245963096619, "Full-finetune/Loss (Raw)": 1.0196588039398193, "Full-finetune/Step": 2571, "Full-finetune/Step Time": 6.734648931771517} {"Full-finetune/Learning Rate": 1.0650534598891608e-05, "Full-finetune/Loss": 0.7662253379821777, "Full-finetune/Loss (Raw)": 0.8234082460403442, "Full-finetune/Step": 2572, "Full-finetune/Step Time": 6.731702970340848} {"Full-finetune/Learning Rate": 1.0644318058092577e-05, "Full-finetune/Loss": 0.7693585157394409, "Full-finetune/Loss (Raw)": 0.800125777721405, "Full-finetune/Step": 2573, "Full-finetune/Step Time": 6.721199901774526} {"Full-finetune/Learning Rate": 1.0638101267246283e-05, "Full-finetune/Loss": 0.7682751417160034, "Full-finetune/Loss (Raw)": 0.7401474118232727, "Full-finetune/Step": 2574, "Full-finetune/Step Time": 6.741645911708474} {"Full-finetune/Learning Rate": 1.063188422876535e-05, "Full-finetune/Loss": 0.7673951387405396, "Full-finetune/Loss (Raw)": 0.7303990125656128, "Full-finetune/Step": 2575, "Full-finetune/Step Time": 6.732866005972028} {"Full-finetune/Learning Rate": 1.0625666945062484e-05, "Full-finetune/Loss": 0.7665125131607056, "Full-finetune/Loss (Raw)": 0.798680305480957, "Full-finetune/Step": 2576, "Full-finetune/Step Time": 6.734629426151514} {"Full-finetune/Learning Rate": 1.061944941855049e-05, "Full-finetune/Loss": 0.7676236629486084, "Full-finetune/Loss (Raw)": 0.8617794513702393, "Full-finetune/Step": 2577, "Full-finetune/Step Time": 6.739250134676695} {"Full-finetune/Learning Rate": 1.061323165164227e-05, "Full-finetune/Loss": 0.7710217833518982, "Full-finetune/Loss (Raw)": 0.9282262325286865, "Full-finetune/Step": 2578, "Full-finetune/Step Time": 6.720807837322354} {"Full-finetune/Learning Rate": 1.0607013646750818e-05, "Full-finetune/Loss": 0.7756110429763794, "Full-finetune/Loss (Raw)": 0.9157667756080627, "Full-finetune/Step": 2579, "Full-finetune/Step Time": 6.701531711965799} {"Full-finetune/Learning Rate": 1.0600795406289221e-05, "Full-finetune/Loss": 0.7742435932159424, "Full-finetune/Loss (Raw)": 0.599308967590332, "Full-finetune/Step": 2580, "Full-finetune/Step Time": 6.709997270256281} {"Full-finetune/Learning Rate": 1.059457693267066e-05, "Full-finetune/Loss": 0.7752645611763, "Full-finetune/Loss (Raw)": 0.9215017557144165, "Full-finetune/Step": 2581, "Full-finetune/Step Time": 6.687028177082539} {"Full-finetune/Learning Rate": 1.0588358228308395e-05, "Full-finetune/Loss": 0.7749752402305603, "Full-finetune/Loss (Raw)": 0.7859111428260803, "Full-finetune/Step": 2582, "Full-finetune/Step Time": 6.6762561444193125} {"Full-finetune/Learning Rate": 1.0582139295615793e-05, "Full-finetune/Loss": 0.7737722396850586, "Full-finetune/Loss (Raw)": 0.7183966636657715, "Full-finetune/Step": 2583, "Full-finetune/Step Time": 6.684716975316405} {"Full-finetune/Learning Rate": 1.0575920137006288e-05, "Full-finetune/Loss": 0.7742575407028198, "Full-finetune/Loss (Raw)": 0.9371374249458313, "Full-finetune/Step": 2584, "Full-finetune/Step Time": 6.685055086389184} {"Full-finetune/Learning Rate": 1.056970075489343e-05, "Full-finetune/Loss": 0.772824227809906, "Full-finetune/Loss (Raw)": 0.6636245250701904, "Full-finetune/Step": 2585, "Full-finetune/Step Time": 6.688041200861335} {"Full-finetune/Learning Rate": 1.0563481151690827e-05, "Full-finetune/Loss": 0.7727673053741455, "Full-finetune/Loss (Raw)": 0.7994118928909302, "Full-finetune/Step": 2586, "Full-finetune/Step Time": 6.674085414037108} {"Full-finetune/Learning Rate": 1.055726132981219e-05, "Full-finetune/Loss": 0.7737193703651428, "Full-finetune/Loss (Raw)": 0.8913702964782715, "Full-finetune/Step": 2587, "Full-finetune/Step Time": 6.6748538594692945} {"Full-finetune/Learning Rate": 1.0551041291671311e-05, "Full-finetune/Loss": 0.7734034061431885, "Full-finetune/Loss (Raw)": 0.7212320566177368, "Full-finetune/Step": 2588, "Full-finetune/Step Time": 6.682917706668377} {"Full-finetune/Learning Rate": 1.0544821039682066e-05, "Full-finetune/Loss": 0.7727254629135132, "Full-finetune/Loss (Raw)": 0.8548746109008789, "Full-finetune/Step": 2589, "Full-finetune/Step Time": 6.68706519715488} {"Full-finetune/Learning Rate": 1.0538600576258413e-05, "Full-finetune/Loss": 0.7729676961898804, "Full-finetune/Loss (Raw)": 0.8857022523880005, "Full-finetune/Step": 2590, "Full-finetune/Step Time": 6.680282611399889} {"Full-finetune/Learning Rate": 1.0532379903814387e-05, "Full-finetune/Loss": 0.7725145816802979, "Full-finetune/Loss (Raw)": 0.7176022529602051, "Full-finetune/Step": 2591, "Full-finetune/Step Time": 6.71815299987793} {"Full-finetune/Learning Rate": 1.0526159024764118e-05, "Full-finetune/Loss": 0.7718783617019653, "Full-finetune/Loss (Raw)": 0.8112376928329468, "Full-finetune/Step": 2592, "Full-finetune/Step Time": 6.727951182052493} {"Full-finetune/Learning Rate": 1.0519937941521805e-05, "Full-finetune/Loss": 0.7708824872970581, "Full-finetune/Loss (Raw)": 0.6994755268096924, "Full-finetune/Step": 2593, "Full-finetune/Step Time": 6.7480293698608875} {"Full-finetune/Learning Rate": 1.0513716656501723e-05, "Full-finetune/Loss": 0.7703851461410522, "Full-finetune/Loss (Raw)": 0.6980387568473816, "Full-finetune/Step": 2594, "Full-finetune/Step Time": 6.759160304442048} {"Full-finetune/Learning Rate": 1.0507495172118242e-05, "Full-finetune/Loss": 0.7702311873435974, "Full-finetune/Loss (Raw)": 0.7929430603981018, "Full-finetune/Step": 2595, "Full-finetune/Step Time": 6.765655277296901} {"Full-finetune/Learning Rate": 1.0501273490785789e-05, "Full-finetune/Loss": 0.7686394453048706, "Full-finetune/Loss (Raw)": 0.7204813957214355, "Full-finetune/Step": 2596, "Full-finetune/Step Time": 6.7592104114592075} {"Full-finetune/Learning Rate": 1.049505161491888e-05, "Full-finetune/Loss": 0.7686208486557007, "Full-finetune/Loss (Raw)": 0.8913635015487671, "Full-finetune/Step": 2597, "Full-finetune/Step Time": 6.763068651780486} {"Full-finetune/Learning Rate": 1.0488829546932106e-05, "Full-finetune/Loss": 0.7672666907310486, "Full-finetune/Loss (Raw)": 0.579558789730072, "Full-finetune/Step": 2598, "Full-finetune/Step Time": 6.77420655079186} {"Full-finetune/Learning Rate": 1.0482607289240128e-05, "Full-finetune/Loss": 0.7668925523757935, "Full-finetune/Loss (Raw)": 0.8583533763885498, "Full-finetune/Step": 2599, "Full-finetune/Step Time": 6.7800280302762985} {"Full-finetune/Learning Rate": 1.0476384844257686e-05, "Full-finetune/Loss": 0.7665992975234985, "Full-finetune/Loss (Raw)": 0.7962793111801147, "Full-finetune/Step": 2600, "Full-finetune/Step Time": 6.785388141870499} {"Full-finetune/Learning Rate": 1.0470162214399582e-05, "Full-finetune/Loss": 0.7658884525299072, "Full-finetune/Loss (Raw)": 0.780994176864624, "Full-finetune/Step": 2601, "Full-finetune/Step Time": 6.792321313172579} {"Full-finetune/Learning Rate": 1.0463939402080706e-05, "Full-finetune/Loss": 0.7670341730117798, "Full-finetune/Loss (Raw)": 0.9436553716659546, "Full-finetune/Step": 2602, "Full-finetune/Step Time": 6.7916827872395515} {"Full-finetune/Learning Rate": 1.0457716409716002e-05, "Full-finetune/Loss": 0.7676581144332886, "Full-finetune/Loss (Raw)": 0.8894743323326111, "Full-finetune/Step": 2603, "Full-finetune/Step Time": 6.801174413412809} {"Full-finetune/Learning Rate": 1.0451493239720496e-05, "Full-finetune/Loss": 0.766937792301178, "Full-finetune/Loss (Raw)": 0.7697060704231262, "Full-finetune/Step": 2604, "Full-finetune/Step Time": 6.808194626122713} {"Full-finetune/Learning Rate": 1.0445269894509283e-05, "Full-finetune/Loss": 0.7661492228507996, "Full-finetune/Loss (Raw)": 0.7280668020248413, "Full-finetune/Step": 2605, "Full-finetune/Step Time": 6.806830367073417} {"Full-finetune/Learning Rate": 1.043904637649751e-05, "Full-finetune/Loss": 0.7652398347854614, "Full-finetune/Loss (Raw)": 0.699938178062439, "Full-finetune/Step": 2606, "Full-finetune/Step Time": 6.836279073730111} {"Full-finetune/Learning Rate": 1.0432822688100414e-05, "Full-finetune/Loss": 0.7660311460494995, "Full-finetune/Loss (Raw)": 0.8206034898757935, "Full-finetune/Step": 2607, "Full-finetune/Step Time": 6.85679030418396} {"Full-finetune/Learning Rate": 1.0426598831733283e-05, "Full-finetune/Loss": 0.7674362659454346, "Full-finetune/Loss (Raw)": 0.8801096677780151, "Full-finetune/Step": 2608, "Full-finetune/Step Time": 6.85116832703352} {"Full-finetune/Learning Rate": 1.0420374809811472e-05, "Full-finetune/Loss": 0.7714094519615173, "Full-finetune/Loss (Raw)": 0.8928032517433167, "Full-finetune/Step": 2609, "Full-finetune/Step Time": 6.838218158110976} {"Full-finetune/Learning Rate": 1.0414150624750408e-05, "Full-finetune/Loss": 0.7689740061759949, "Full-finetune/Loss (Raw)": 0.7105574607849121, "Full-finetune/Step": 2610, "Full-finetune/Step Time": 6.863984994590282} {"Full-finetune/Learning Rate": 1.0407926278965568e-05, "Full-finetune/Loss": 0.7685699462890625, "Full-finetune/Loss (Raw)": 0.7006969451904297, "Full-finetune/Step": 2611, "Full-finetune/Step Time": 6.865728348493576} {"Full-finetune/Learning Rate": 1.0401701774872508e-05, "Full-finetune/Loss": 0.769164502620697, "Full-finetune/Loss (Raw)": 0.8027512431144714, "Full-finetune/Step": 2612, "Full-finetune/Step Time": 6.854584980756044} {"Full-finetune/Learning Rate": 1.0395477114886827e-05, "Full-finetune/Loss": 0.7700666785240173, "Full-finetune/Loss (Raw)": 0.9600543975830078, "Full-finetune/Step": 2613, "Full-finetune/Step Time": 6.853582516312599} {"Full-finetune/Learning Rate": 1.0389252301424197e-05, "Full-finetune/Loss": 0.770961344242096, "Full-finetune/Loss (Raw)": 0.7622721195220947, "Full-finetune/Step": 2614, "Full-finetune/Step Time": 6.8472418785095215} {"Full-finetune/Learning Rate": 1.0383027336900356e-05, "Full-finetune/Loss": 0.7752711772918701, "Full-finetune/Loss (Raw)": 0.8741891384124756, "Full-finetune/Step": 2615, "Full-finetune/Step Time": 6.8399135787039995} {"Full-finetune/Learning Rate": 1.0376802223731077e-05, "Full-finetune/Loss": 0.7753057479858398, "Full-finetune/Loss (Raw)": 0.912476122379303, "Full-finetune/Step": 2616, "Full-finetune/Step Time": 6.842224488034844} {"Full-finetune/Learning Rate": 1.0370576964332217e-05, "Full-finetune/Loss": 0.771715521812439, "Full-finetune/Loss (Raw)": 0.4040660858154297, "Full-finetune/Step": 2617, "Full-finetune/Step Time": 6.837567338719964} {"Full-finetune/Learning Rate": 1.0364351561119667e-05, "Full-finetune/Loss": 0.7742710113525391, "Full-finetune/Loss (Raw)": 0.646577000617981, "Full-finetune/Step": 2618, "Full-finetune/Step Time": 6.827465059235692} {"Full-finetune/Learning Rate": 1.0358126016509392e-05, "Full-finetune/Loss": 0.7751649618148804, "Full-finetune/Loss (Raw)": 0.898097038269043, "Full-finetune/Step": 2619, "Full-finetune/Step Time": 6.79992401227355} {"Full-finetune/Learning Rate": 1.0351900332917401e-05, "Full-finetune/Loss": 0.7752600908279419, "Full-finetune/Loss (Raw)": 0.7299767732620239, "Full-finetune/Step": 2620, "Full-finetune/Step Time": 6.7959721852093935} {"Full-finetune/Learning Rate": 1.034567451275976e-05, "Full-finetune/Loss": 0.7746168375015259, "Full-finetune/Loss (Raw)": 0.7613616585731506, "Full-finetune/Step": 2621, "Full-finetune/Step Time": 6.79430915042758} {"Full-finetune/Learning Rate": 1.0339448558452594e-05, "Full-finetune/Loss": 0.7742562294006348, "Full-finetune/Loss (Raw)": 0.8837801814079285, "Full-finetune/Step": 2622, "Full-finetune/Step Time": 6.78994195535779} {"Full-finetune/Learning Rate": 1.0333222472412064e-05, "Full-finetune/Loss": 0.7735714912414551, "Full-finetune/Loss (Raw)": 0.7671545147895813, "Full-finetune/Step": 2623, "Full-finetune/Step Time": 6.811238572001457} {"Full-finetune/Learning Rate": 1.03269962570544e-05, "Full-finetune/Loss": 0.7741490602493286, "Full-finetune/Loss (Raw)": 0.8335912823677063, "Full-finetune/Step": 2624, "Full-finetune/Step Time": 6.8010748364031315} {"Full-finetune/Learning Rate": 1.0320769914795874e-05, "Full-finetune/Loss": 0.7735227346420288, "Full-finetune/Loss (Raw)": 0.6001778244972229, "Full-finetune/Step": 2625, "Full-finetune/Step Time": 6.815287543460727} {"Full-finetune/Learning Rate": 1.0314543448052806e-05, "Full-finetune/Loss": 0.7687145471572876, "Full-finetune/Loss (Raw)": 0.3196852505207062, "Full-finetune/Step": 2626, "Full-finetune/Step Time": 6.832146819680929} {"Full-finetune/Learning Rate": 1.0308316859241567e-05, "Full-finetune/Loss": 0.7708950638771057, "Full-finetune/Loss (Raw)": 0.784810483455658, "Full-finetune/Step": 2627, "Full-finetune/Step Time": 6.817185454070568} {"Full-finetune/Learning Rate": 1.0302090150778572e-05, "Full-finetune/Loss": 0.7710461020469666, "Full-finetune/Loss (Raw)": 0.880700945854187, "Full-finetune/Step": 2628, "Full-finetune/Step Time": 6.817536871880293} {"Full-finetune/Learning Rate": 1.029586332508029e-05, "Full-finetune/Loss": 0.7681963443756104, "Full-finetune/Loss (Raw)": 0.5294341444969177, "Full-finetune/Step": 2629, "Full-finetune/Step Time": 6.833398127928376} {"Full-finetune/Learning Rate": 1.0289636384563226e-05, "Full-finetune/Loss": 0.7667880654335022, "Full-finetune/Loss (Raw)": 0.6928674578666687, "Full-finetune/Step": 2630, "Full-finetune/Step Time": 6.834747243672609} {"Full-finetune/Learning Rate": 1.0283409331643936e-05, "Full-finetune/Loss": 0.7674325704574585, "Full-finetune/Loss (Raw)": 0.901157021522522, "Full-finetune/Step": 2631, "Full-finetune/Step Time": 6.832359174266458} {"Full-finetune/Learning Rate": 1.0277182168739019e-05, "Full-finetune/Loss": 0.7664445042610168, "Full-finetune/Loss (Raw)": 0.7127731442451477, "Full-finetune/Step": 2632, "Full-finetune/Step Time": 6.838005233556032} {"Full-finetune/Learning Rate": 1.0270954898265113e-05, "Full-finetune/Loss": 0.7666757702827454, "Full-finetune/Loss (Raw)": 0.8581487536430359, "Full-finetune/Step": 2633, "Full-finetune/Step Time": 6.8413810059428215} {"Full-finetune/Learning Rate": 1.0264727522638899e-05, "Full-finetune/Loss": 0.7667216062545776, "Full-finetune/Loss (Raw)": 0.7899475693702698, "Full-finetune/Step": 2634, "Full-finetune/Step Time": 6.843292279168963} {"Full-finetune/Learning Rate": 1.0258500044277106e-05, "Full-finetune/Loss": 0.7656717896461487, "Full-finetune/Loss (Raw)": 0.7822409272193909, "Full-finetune/Step": 2635, "Full-finetune/Step Time": 6.837448012083769} {"Full-finetune/Learning Rate": 1.025227246559649e-05, "Full-finetune/Loss": 0.7674615979194641, "Full-finetune/Loss (Raw)": 0.9602974653244019, "Full-finetune/Step": 2636, "Full-finetune/Step Time": 6.837818488478661} {"Full-finetune/Learning Rate": 1.0246044789013855e-05, "Full-finetune/Loss": 0.7672052383422852, "Full-finetune/Loss (Raw)": 0.76030433177948, "Full-finetune/Step": 2637, "Full-finetune/Step Time": 6.842412104830146} {"Full-finetune/Learning Rate": 1.023981701694604e-05, "Full-finetune/Loss": 0.7671172618865967, "Full-finetune/Loss (Raw)": 0.9046468138694763, "Full-finetune/Step": 2638, "Full-finetune/Step Time": 6.864403918385506} {"Full-finetune/Learning Rate": 1.0233589151809927e-05, "Full-finetune/Loss": 0.7663922309875488, "Full-finetune/Loss (Raw)": 0.7738997936248779, "Full-finetune/Step": 2639, "Full-finetune/Step Time": 6.854717554524541} {"Full-finetune/Learning Rate": 1.0227361196022421e-05, "Full-finetune/Loss": 0.7636801600456238, "Full-finetune/Loss (Raw)": 0.35392364859580994, "Full-finetune/Step": 2640, "Full-finetune/Step Time": 6.863238129764795} {"Full-finetune/Learning Rate": 1.0221133152000474e-05, "Full-finetune/Loss": 0.764217734336853, "Full-finetune/Loss (Raw)": 0.8015691041946411, "Full-finetune/Step": 2641, "Full-finetune/Step Time": 6.855945184826851} {"Full-finetune/Learning Rate": 1.021490502216107e-05, "Full-finetune/Loss": 0.7636158466339111, "Full-finetune/Loss (Raw)": 0.7776990532875061, "Full-finetune/Step": 2642, "Full-finetune/Step Time": 6.84663924202323} {"Full-finetune/Learning Rate": 1.020867680892122e-05, "Full-finetune/Loss": 0.7635520696640015, "Full-finetune/Loss (Raw)": 0.6309957504272461, "Full-finetune/Step": 2643, "Full-finetune/Step Time": 6.834295812994242} {"Full-finetune/Learning Rate": 1.0202448514697975e-05, "Full-finetune/Loss": 0.7625039219856262, "Full-finetune/Loss (Raw)": 0.6514578461647034, "Full-finetune/Step": 2644, "Full-finetune/Step Time": 6.867699254304171} {"Full-finetune/Learning Rate": 1.0196220141908417e-05, "Full-finetune/Loss": 0.7620618343353271, "Full-finetune/Loss (Raw)": 0.7892521619796753, "Full-finetune/Step": 2645, "Full-finetune/Step Time": 6.870861444622278} {"Full-finetune/Learning Rate": 1.018999169296965e-05, "Full-finetune/Loss": 0.7625502943992615, "Full-finetune/Loss (Raw)": 0.8040739297866821, "Full-finetune/Step": 2646, "Full-finetune/Step Time": 6.856988271698356} {"Full-finetune/Learning Rate": 1.0183763170298818e-05, "Full-finetune/Loss": 0.7621660828590393, "Full-finetune/Loss (Raw)": 0.7353276014328003, "Full-finetune/Step": 2647, "Full-finetune/Step Time": 6.862168539315462} {"Full-finetune/Learning Rate": 1.0177534576313084e-05, "Full-finetune/Loss": 0.7611745595932007, "Full-finetune/Loss (Raw)": 0.8094040155410767, "Full-finetune/Step": 2648, "Full-finetune/Step Time": 6.873395070433617} {"Full-finetune/Learning Rate": 1.0171305913429648e-05, "Full-finetune/Loss": 0.7635056972503662, "Full-finetune/Loss (Raw)": 0.7081759572029114, "Full-finetune/Step": 2649, "Full-finetune/Step Time": 6.867067191749811} {"Full-finetune/Learning Rate": 1.0165077184065732e-05, "Full-finetune/Loss": 0.762521505355835, "Full-finetune/Loss (Raw)": 0.7089391350746155, "Full-finetune/Step": 2650, "Full-finetune/Step Time": 6.8478468134999275} {"Full-finetune/Learning Rate": 1.0158848390638588e-05, "Full-finetune/Loss": 0.763950526714325, "Full-finetune/Loss (Raw)": 0.7846425175666809, "Full-finetune/Step": 2651, "Full-finetune/Step Time": 6.843517791479826} {"Full-finetune/Learning Rate": 1.015261953556548e-05, "Full-finetune/Loss": 0.764613151550293, "Full-finetune/Loss (Raw)": 0.8654447793960571, "Full-finetune/Step": 2652, "Full-finetune/Step Time": 6.849114490672946} {"Full-finetune/Learning Rate": 1.0146390621263709e-05, "Full-finetune/Loss": 0.7653051614761353, "Full-finetune/Loss (Raw)": 0.6961429119110107, "Full-finetune/Step": 2653, "Full-finetune/Step Time": 6.8600187841802835} {"Full-finetune/Learning Rate": 1.0140161650150598e-05, "Full-finetune/Loss": 0.7612985372543335, "Full-finetune/Loss (Raw)": 0.3502296209335327, "Full-finetune/Step": 2654, "Full-finetune/Step Time": 6.869617115706205} {"Full-finetune/Learning Rate": 1.0133932624643487e-05, "Full-finetune/Loss": 0.7611179947853088, "Full-finetune/Loss (Raw)": 0.8023290038108826, "Full-finetune/Step": 2655, "Full-finetune/Step Time": 6.869853645563126} {"Full-finetune/Learning Rate": 1.012770354715974e-05, "Full-finetune/Loss": 0.7600184082984924, "Full-finetune/Loss (Raw)": 0.7963308691978455, "Full-finetune/Step": 2656, "Full-finetune/Step Time": 6.866902399808168} {"Full-finetune/Learning Rate": 1.012147442011674e-05, "Full-finetune/Loss": 0.7624161243438721, "Full-finetune/Loss (Raw)": 0.5819924473762512, "Full-finetune/Step": 2657, "Full-finetune/Step Time": 6.86382170021534} {"Full-finetune/Learning Rate": 1.0115245245931883e-05, "Full-finetune/Loss": 0.7598940134048462, "Full-finetune/Loss (Raw)": 0.5473009943962097, "Full-finetune/Step": 2658, "Full-finetune/Step Time": 6.879627330228686} {"Full-finetune/Learning Rate": 1.0109016027022602e-05, "Full-finetune/Loss": 0.7604296207427979, "Full-finetune/Loss (Raw)": 0.8386254906654358, "Full-finetune/Step": 2659, "Full-finetune/Step Time": 6.880886793136597} {"Full-finetune/Learning Rate": 1.0102786765806329e-05, "Full-finetune/Loss": 0.7611097097396851, "Full-finetune/Loss (Raw)": 0.8821299076080322, "Full-finetune/Step": 2660, "Full-finetune/Step Time": 6.86445995606482} {"Full-finetune/Learning Rate": 1.0096557464700518e-05, "Full-finetune/Loss": 0.763545036315918, "Full-finetune/Loss (Raw)": 1.016670823097229, "Full-finetune/Step": 2661, "Full-finetune/Step Time": 6.8685801681131124} {"Full-finetune/Learning Rate": 1.0090328126122642e-05, "Full-finetune/Loss": 0.7673420310020447, "Full-finetune/Loss (Raw)": 0.8722241520881653, "Full-finetune/Step": 2662, "Full-finetune/Step Time": 6.847102200612426} {"Full-finetune/Learning Rate": 1.0084098752490179e-05, "Full-finetune/Loss": 0.7672069072723389, "Full-finetune/Loss (Raw)": 0.893145740032196, "Full-finetune/Step": 2663, "Full-finetune/Step Time": 6.8517572190612555} {"Full-finetune/Learning Rate": 1.0077869346220635e-05, "Full-finetune/Loss": 0.7657831907272339, "Full-finetune/Loss (Raw)": 0.4438328742980957, "Full-finetune/Step": 2664, "Full-finetune/Step Time": 6.861493520438671} {"Full-finetune/Learning Rate": 1.0071639909731519e-05, "Full-finetune/Loss": 0.7652100324630737, "Full-finetune/Loss (Raw)": 0.6660478115081787, "Full-finetune/Step": 2665, "Full-finetune/Step Time": 6.857006317004561} {"Full-finetune/Learning Rate": 1.0065410445440352e-05, "Full-finetune/Loss": 0.7651879191398621, "Full-finetune/Loss (Raw)": 0.8614534735679626, "Full-finetune/Step": 2666, "Full-finetune/Step Time": 6.852771149948239} {"Full-finetune/Learning Rate": 1.0059180955764666e-05, "Full-finetune/Loss": 0.7647102475166321, "Full-finetune/Loss (Raw)": 0.7302586436271667, "Full-finetune/Step": 2667, "Full-finetune/Step Time": 6.847218483686447} {"Full-finetune/Learning Rate": 1.005295144312201e-05, "Full-finetune/Loss": 0.7644255757331848, "Full-finetune/Loss (Raw)": 0.7941482663154602, "Full-finetune/Step": 2668, "Full-finetune/Step Time": 6.84408251196146} {"Full-finetune/Learning Rate": 1.0046721909929928e-05, "Full-finetune/Loss": 0.766595721244812, "Full-finetune/Loss (Raw)": 0.9186933636665344, "Full-finetune/Step": 2669, "Full-finetune/Step Time": 6.82742296718061} {"Full-finetune/Learning Rate": 1.0040492358605989e-05, "Full-finetune/Loss": 0.7666715979576111, "Full-finetune/Loss (Raw)": 0.8390932083129883, "Full-finetune/Step": 2670, "Full-finetune/Step Time": 6.8217754159122705} {"Full-finetune/Learning Rate": 1.0034262791567755e-05, "Full-finetune/Loss": 0.7650332450866699, "Full-finetune/Loss (Raw)": 0.6343058943748474, "Full-finetune/Step": 2671, "Full-finetune/Step Time": 6.828177822753787} {"Full-finetune/Learning Rate": 1.0028033211232802e-05, "Full-finetune/Loss": 0.7643967270851135, "Full-finetune/Loss (Raw)": 0.5601533055305481, "Full-finetune/Step": 2672, "Full-finetune/Step Time": 6.833081817254424} {"Full-finetune/Learning Rate": 1.0021803620018706e-05, "Full-finetune/Loss": 0.7644507884979248, "Full-finetune/Loss (Raw)": 0.8057750463485718, "Full-finetune/Step": 2673, "Full-finetune/Step Time": 6.834070660173893} {"Full-finetune/Learning Rate": 1.0015574020343048e-05, "Full-finetune/Loss": 0.767030656337738, "Full-finetune/Loss (Raw)": 0.6142957210540771, "Full-finetune/Step": 2674, "Full-finetune/Step Time": 6.827244624495506} {"Full-finetune/Learning Rate": 1.0009344414623418e-05, "Full-finetune/Loss": 0.7671637535095215, "Full-finetune/Loss (Raw)": 0.8075162172317505, "Full-finetune/Step": 2675, "Full-finetune/Step Time": 6.831250077113509} {"Full-finetune/Learning Rate": 1.0003114805277405e-05, "Full-finetune/Loss": 0.7657862901687622, "Full-finetune/Loss (Raw)": 0.7788869142532349, "Full-finetune/Step": 2676, "Full-finetune/Step Time": 6.823630170896649} {"Full-finetune/Learning Rate": 9.9968851947226e-06, "Full-finetune/Loss": 0.7700785398483276, "Full-finetune/Loss (Raw)": 0.8803870677947998, "Full-finetune/Step": 2677, "Full-finetune/Step Time": 6.8051560912281275} {"Full-finetune/Learning Rate": 9.990655585376583e-06, "Full-finetune/Loss": 0.7699146270751953, "Full-finetune/Loss (Raw)": 0.7895746231079102, "Full-finetune/Step": 2678, "Full-finetune/Step Time": 6.802724443376064} {"Full-finetune/Learning Rate": 9.984425979656955e-06, "Full-finetune/Loss": 0.7705851793289185, "Full-finetune/Loss (Raw)": 0.9648275375366211, "Full-finetune/Step": 2679, "Full-finetune/Step Time": 6.791092123836279} {"Full-finetune/Learning Rate": 9.978196379981297e-06, "Full-finetune/Loss": 0.7692999839782715, "Full-finetune/Loss (Raw)": 0.8735488057136536, "Full-finetune/Step": 2680, "Full-finetune/Step Time": 6.792267590761185} {"Full-finetune/Learning Rate": 9.971966788767198e-06, "Full-finetune/Loss": 0.7663202881813049, "Full-finetune/Loss (Raw)": 0.3425636291503906, "Full-finetune/Step": 2681, "Full-finetune/Step Time": 6.798036152496934} {"Full-finetune/Learning Rate": 9.96573720843225e-06, "Full-finetune/Loss": 0.7660672068595886, "Full-finetune/Loss (Raw)": 0.8072460293769836, "Full-finetune/Step": 2682, "Full-finetune/Step Time": 6.775809057056904} {"Full-finetune/Learning Rate": 9.959507641394015e-06, "Full-finetune/Loss": 0.7699631452560425, "Full-finetune/Loss (Raw)": 0.834137499332428, "Full-finetune/Step": 2683, "Full-finetune/Step Time": 6.756531335413456} {"Full-finetune/Learning Rate": 9.953278090070075e-06, "Full-finetune/Loss": 0.7694251537322998, "Full-finetune/Loss (Raw)": 0.7838708162307739, "Full-finetune/Step": 2684, "Full-finetune/Step Time": 6.755481602624059} {"Full-finetune/Learning Rate": 9.947048556877992e-06, "Full-finetune/Loss": 0.7706055045127869, "Full-finetune/Loss (Raw)": 0.8468818664550781, "Full-finetune/Step": 2685, "Full-finetune/Step Time": 6.759350933134556} {"Full-finetune/Learning Rate": 9.940819044235335e-06, "Full-finetune/Loss": 0.7700196504592896, "Full-finetune/Loss (Raw)": 0.7212646007537842, "Full-finetune/Step": 2686, "Full-finetune/Step Time": 6.761999309062958} {"Full-finetune/Learning Rate": 9.934589554559653e-06, "Full-finetune/Loss": 0.7704801559448242, "Full-finetune/Loss (Raw)": 0.7888271808624268, "Full-finetune/Step": 2687, "Full-finetune/Step Time": 6.773855904117227} {"Full-finetune/Learning Rate": 9.928360090268484e-06, "Full-finetune/Loss": 0.7726110219955444, "Full-finetune/Loss (Raw)": 0.8528632521629333, "Full-finetune/Step": 2688, "Full-finetune/Step Time": 6.755439192056656} {"Full-finetune/Learning Rate": 9.922130653779368e-06, "Full-finetune/Loss": 0.7734776735305786, "Full-finetune/Loss (Raw)": 0.9439674615859985, "Full-finetune/Step": 2689, "Full-finetune/Step Time": 6.752489861100912} {"Full-finetune/Learning Rate": 9.915901247509821e-06, "Full-finetune/Loss": 0.7694332599639893, "Full-finetune/Loss (Raw)": 0.37803328037261963, "Full-finetune/Step": 2690, "Full-finetune/Step Time": 6.770593980327249} {"Full-finetune/Learning Rate": 9.909671873877363e-06, "Full-finetune/Loss": 0.7730753421783447, "Full-finetune/Loss (Raw)": 0.8883655667304993, "Full-finetune/Step": 2691, "Full-finetune/Step Time": 6.764585234224796} {"Full-finetune/Learning Rate": 9.903442535299487e-06, "Full-finetune/Loss": 0.7725449204444885, "Full-finetune/Loss (Raw)": 0.7770545482635498, "Full-finetune/Step": 2692, "Full-finetune/Step Time": 6.7636344358325005} {"Full-finetune/Learning Rate": 9.897213234193673e-06, "Full-finetune/Loss": 0.7721781730651855, "Full-finetune/Loss (Raw)": 0.8392686247825623, "Full-finetune/Step": 2693, "Full-finetune/Step Time": 6.761192126199603} {"Full-finetune/Learning Rate": 9.890983972977401e-06, "Full-finetune/Loss": 0.7714354991912842, "Full-finetune/Loss (Raw)": 0.7405292391777039, "Full-finetune/Step": 2694, "Full-finetune/Step Time": 6.761429391801357} {"Full-finetune/Learning Rate": 9.884754754068116e-06, "Full-finetune/Loss": 0.7717941999435425, "Full-finetune/Loss (Raw)": 0.7726820111274719, "Full-finetune/Step": 2695, "Full-finetune/Step Time": 6.751229187473655} {"Full-finetune/Learning Rate": 9.878525579883264e-06, "Full-finetune/Loss": 0.7722872495651245, "Full-finetune/Loss (Raw)": 0.7981883883476257, "Full-finetune/Step": 2696, "Full-finetune/Step Time": 6.76232810690999} {"Full-finetune/Learning Rate": 9.872296452840266e-06, "Full-finetune/Loss": 0.7706658840179443, "Full-finetune/Loss (Raw)": 0.804341197013855, "Full-finetune/Step": 2697, "Full-finetune/Step Time": 6.765279712155461} {"Full-finetune/Learning Rate": 9.866067375356516e-06, "Full-finetune/Loss": 0.7700033187866211, "Full-finetune/Loss (Raw)": 0.713467538356781, "Full-finetune/Step": 2698, "Full-finetune/Step Time": 6.76070324704051} {"Full-finetune/Learning Rate": 9.859838349849404e-06, "Full-finetune/Loss": 0.7689483165740967, "Full-finetune/Loss (Raw)": 0.8846266865730286, "Full-finetune/Step": 2699, "Full-finetune/Step Time": 6.775559846311808} {"Full-finetune/Learning Rate": 9.853609378736293e-06, "Full-finetune/Loss": 0.768724799156189, "Full-finetune/Loss (Raw)": 0.7947912812232971, "Full-finetune/Step": 2700, "Full-finetune/Step Time": 6.775380447506905} {"Full-finetune/Learning Rate": 9.847380464434525e-06, "Full-finetune/Loss": 0.7685045599937439, "Full-finetune/Loss (Raw)": 0.7719345688819885, "Full-finetune/Step": 2701, "Full-finetune/Step Time": 6.797281069681048} {"Full-finetune/Learning Rate": 9.841151609361419e-06, "Full-finetune/Loss": 0.7693215608596802, "Full-finetune/Loss (Raw)": 0.8447297215461731, "Full-finetune/Step": 2702, "Full-finetune/Step Time": 6.786386443302035} {"Full-finetune/Learning Rate": 9.83492281593427e-06, "Full-finetune/Loss": 0.7690601348876953, "Full-finetune/Loss (Raw)": 0.6969314813613892, "Full-finetune/Step": 2703, "Full-finetune/Step Time": 6.803612507879734} {"Full-finetune/Learning Rate": 9.828694086570353e-06, "Full-finetune/Loss": 0.7682023048400879, "Full-finetune/Loss (Raw)": 0.6888810396194458, "Full-finetune/Step": 2704, "Full-finetune/Step Time": 6.815879745408893} {"Full-finetune/Learning Rate": 9.822465423686917e-06, "Full-finetune/Loss": 0.7674272656440735, "Full-finetune/Loss (Raw)": 0.7625717520713806, "Full-finetune/Step": 2705, "Full-finetune/Step Time": 6.815712520852685} {"Full-finetune/Learning Rate": 9.816236829701187e-06, "Full-finetune/Loss": 0.767042338848114, "Full-finetune/Loss (Raw)": 0.8789569735527039, "Full-finetune/Step": 2706, "Full-finetune/Step Time": 6.823642095550895} {"Full-finetune/Learning Rate": 9.810008307030356e-06, "Full-finetune/Loss": 0.7674670219421387, "Full-finetune/Loss (Raw)": 0.9701305031776428, "Full-finetune/Step": 2707, "Full-finetune/Step Time": 6.819722443819046} {"Full-finetune/Learning Rate": 9.803779858091586e-06, "Full-finetune/Loss": 0.7698173522949219, "Full-finetune/Loss (Raw)": 0.9001423120498657, "Full-finetune/Step": 2708, "Full-finetune/Step Time": 6.817320479080081} {"Full-finetune/Learning Rate": 9.797551485302027e-06, "Full-finetune/Loss": 0.7685993909835815, "Full-finetune/Loss (Raw)": 0.7656111121177673, "Full-finetune/Step": 2709, "Full-finetune/Step Time": 6.827109429985285} {"Full-finetune/Learning Rate": 9.791323191078781e-06, "Full-finetune/Loss": 0.7654263973236084, "Full-finetune/Loss (Raw)": 0.3797689974308014, "Full-finetune/Step": 2710, "Full-finetune/Step Time": 6.845682924613357} {"Full-finetune/Learning Rate": 9.785094977838935e-06, "Full-finetune/Loss": 0.7665648460388184, "Full-finetune/Loss (Raw)": 0.8641119003295898, "Full-finetune/Step": 2711, "Full-finetune/Step Time": 6.837581122294068} {"Full-finetune/Learning Rate": 9.77886684799953e-06, "Full-finetune/Loss": 0.7656402587890625, "Full-finetune/Loss (Raw)": 0.8187956213951111, "Full-finetune/Step": 2712, "Full-finetune/Step Time": 6.856306426227093} {"Full-finetune/Learning Rate": 9.77263880397758e-06, "Full-finetune/Loss": 0.7666113972663879, "Full-finetune/Loss (Raw)": 0.7879239916801453, "Full-finetune/Step": 2713, "Full-finetune/Step Time": 6.862722856923938} {"Full-finetune/Learning Rate": 9.766410848190076e-06, "Full-finetune/Loss": 0.7672406435012817, "Full-finetune/Loss (Raw)": 0.8799657821655273, "Full-finetune/Step": 2714, "Full-finetune/Step Time": 6.862856639549136} {"Full-finetune/Learning Rate": 9.76018298305396e-06, "Full-finetune/Loss": 0.7659111618995667, "Full-finetune/Loss (Raw)": 0.721192479133606, "Full-finetune/Step": 2715, "Full-finetune/Step Time": 6.8549363519996405} {"Full-finetune/Learning Rate": 9.75395521098615e-06, "Full-finetune/Loss": 0.7664353847503662, "Full-finetune/Loss (Raw)": 0.788330078125, "Full-finetune/Step": 2716, "Full-finetune/Step Time": 6.858743377029896} {"Full-finetune/Learning Rate": 9.747727534403515e-06, "Full-finetune/Loss": 0.7663837671279907, "Full-finetune/Loss (Raw)": 0.8482677340507507, "Full-finetune/Step": 2717, "Full-finetune/Step Time": 6.839883113279939} {"Full-finetune/Learning Rate": 9.741499955722897e-06, "Full-finetune/Loss": 0.7653465270996094, "Full-finetune/Loss (Raw)": 0.7529435157775879, "Full-finetune/Step": 2718, "Full-finetune/Step Time": 6.850445978343487} {"Full-finetune/Learning Rate": 9.735272477361103e-06, "Full-finetune/Loss": 0.7661263346672058, "Full-finetune/Loss (Raw)": 0.8174101114273071, "Full-finetune/Step": 2719, "Full-finetune/Step Time": 6.813843531534076} {"Full-finetune/Learning Rate": 9.729045101734889e-06, "Full-finetune/Loss": 0.7665046453475952, "Full-finetune/Loss (Raw)": 0.8596644401550293, "Full-finetune/Step": 2720, "Full-finetune/Step Time": 6.808383630588651} {"Full-finetune/Learning Rate": 9.722817831260985e-06, "Full-finetune/Loss": 0.7665364742279053, "Full-finetune/Loss (Raw)": 0.7035403847694397, "Full-finetune/Step": 2721, "Full-finetune/Step Time": 6.799122272059321} {"Full-finetune/Learning Rate": 9.716590668356067e-06, "Full-finetune/Loss": 0.7673208713531494, "Full-finetune/Loss (Raw)": 0.7984510064125061, "Full-finetune/Step": 2722, "Full-finetune/Step Time": 6.7901687398552895} {"Full-finetune/Learning Rate": 9.710363615436777e-06, "Full-finetune/Loss": 0.7668294906616211, "Full-finetune/Loss (Raw)": 0.7300393581390381, "Full-finetune/Step": 2723, "Full-finetune/Step Time": 6.7846351377666} {"Full-finetune/Learning Rate": 9.704136674919712e-06, "Full-finetune/Loss": 0.7675558924674988, "Full-finetune/Loss (Raw)": 0.8134600520133972, "Full-finetune/Step": 2724, "Full-finetune/Step Time": 6.787884896621108} {"Full-finetune/Learning Rate": 9.697909849221428e-06, "Full-finetune/Loss": 0.7668700218200684, "Full-finetune/Loss (Raw)": 0.8035793304443359, "Full-finetune/Step": 2725, "Full-finetune/Step Time": 6.780853753909469} {"Full-finetune/Learning Rate": 9.691683140758439e-06, "Full-finetune/Loss": 0.7686493396759033, "Full-finetune/Loss (Raw)": 0.80731201171875, "Full-finetune/Step": 2726, "Full-finetune/Step Time": 6.778668334707618} {"Full-finetune/Learning Rate": 9.685456551947199e-06, "Full-finetune/Loss": 0.7680476903915405, "Full-finetune/Loss (Raw)": 0.7813352346420288, "Full-finetune/Step": 2727, "Full-finetune/Step Time": 6.781250851228833} {"Full-finetune/Learning Rate": 9.67923008520413e-06, "Full-finetune/Loss": 0.7661663293838501, "Full-finetune/Loss (Raw)": 0.5554640293121338, "Full-finetune/Step": 2728, "Full-finetune/Step Time": 6.77938955463469} {"Full-finetune/Learning Rate": 9.673003742945602e-06, "Full-finetune/Loss": 0.763384222984314, "Full-finetune/Loss (Raw)": 0.42488813400268555, "Full-finetune/Step": 2729, "Full-finetune/Step Time": 6.787449890747666} {"Full-finetune/Learning Rate": 9.666777527587936e-06, "Full-finetune/Loss": 0.760851263999939, "Full-finetune/Loss (Raw)": 0.6194354891777039, "Full-finetune/Step": 2730, "Full-finetune/Step Time": 6.790276035666466} {"Full-finetune/Learning Rate": 9.660551441547411e-06, "Full-finetune/Loss": 0.7602411508560181, "Full-finetune/Loss (Raw)": 0.8113809823989868, "Full-finetune/Step": 2731, "Full-finetune/Step Time": 6.7773031033575535} {"Full-finetune/Learning Rate": 9.654325487240243e-06, "Full-finetune/Loss": 0.7606141567230225, "Full-finetune/Loss (Raw)": 0.8174554109573364, "Full-finetune/Step": 2732, "Full-finetune/Step Time": 6.772494953125715} {"Full-finetune/Learning Rate": 9.6480996670826e-06, "Full-finetune/Loss": 0.7616420984268188, "Full-finetune/Loss (Raw)": 0.8596370220184326, "Full-finetune/Step": 2733, "Full-finetune/Step Time": 6.77285703830421} {"Full-finetune/Learning Rate": 9.64187398349061e-06, "Full-finetune/Loss": 0.7621991634368896, "Full-finetune/Loss (Raw)": 0.7712511420249939, "Full-finetune/Step": 2734, "Full-finetune/Step Time": 6.749450808390975} {"Full-finetune/Learning Rate": 9.635648438880334e-06, "Full-finetune/Loss": 0.7625887393951416, "Full-finetune/Loss (Raw)": 0.8704628944396973, "Full-finetune/Step": 2735, "Full-finetune/Step Time": 6.727224297821522} {"Full-finetune/Learning Rate": 9.629423035667788e-06, "Full-finetune/Loss": 0.7593567967414856, "Full-finetune/Loss (Raw)": 0.4664199650287628, "Full-finetune/Step": 2736, "Full-finetune/Step Time": 6.742879925295711} {"Full-finetune/Learning Rate": 9.623197776268926e-06, "Full-finetune/Loss": 0.7566041946411133, "Full-finetune/Loss (Raw)": 0.5404621958732605, "Full-finetune/Step": 2737, "Full-finetune/Step Time": 6.748441407456994} {"Full-finetune/Learning Rate": 9.616972663099648e-06, "Full-finetune/Loss": 0.7577314376831055, "Full-finetune/Loss (Raw)": 0.8548538088798523, "Full-finetune/Step": 2738, "Full-finetune/Step Time": 6.728321589529514} {"Full-finetune/Learning Rate": 9.610747698575801e-06, "Full-finetune/Loss": 0.7591477036476135, "Full-finetune/Loss (Raw)": 0.881977915763855, "Full-finetune/Step": 2739, "Full-finetune/Step Time": 6.72079654596746} {"Full-finetune/Learning Rate": 9.60452288511318e-06, "Full-finetune/Loss": 0.7582523226737976, "Full-finetune/Loss (Raw)": 0.6881424784660339, "Full-finetune/Step": 2740, "Full-finetune/Step Time": 6.730437101796269} {"Full-finetune/Learning Rate": 9.598298225127499e-06, "Full-finetune/Loss": 0.757678747177124, "Full-finetune/Loss (Raw)": 0.8866332769393921, "Full-finetune/Step": 2741, "Full-finetune/Step Time": 6.73097406513989} {"Full-finetune/Learning Rate": 9.592073721034435e-06, "Full-finetune/Loss": 0.7577825784683228, "Full-finetune/Loss (Raw)": 0.7755693793296814, "Full-finetune/Step": 2742, "Full-finetune/Step Time": 6.736776180565357} {"Full-finetune/Learning Rate": 9.585849375249595e-06, "Full-finetune/Loss": 0.7579506635665894, "Full-finetune/Loss (Raw)": 0.8956976532936096, "Full-finetune/Step": 2743, "Full-finetune/Step Time": 6.728242361918092} {"Full-finetune/Learning Rate": 9.57962519018853e-06, "Full-finetune/Loss": 0.7573280334472656, "Full-finetune/Loss (Raw)": 0.832780659198761, "Full-finetune/Step": 2744, "Full-finetune/Step Time": 6.734120965003967} {"Full-finetune/Learning Rate": 9.573401168266722e-06, "Full-finetune/Loss": 0.7612577676773071, "Full-finetune/Loss (Raw)": 0.9070777297019958, "Full-finetune/Step": 2745, "Full-finetune/Step Time": 6.717017421498895} {"Full-finetune/Learning Rate": 9.567177311899589e-06, "Full-finetune/Loss": 0.7625598907470703, "Full-finetune/Loss (Raw)": 0.8132427334785461, "Full-finetune/Step": 2746, "Full-finetune/Step Time": 6.722540702670813} {"Full-finetune/Learning Rate": 9.560953623502493e-06, "Full-finetune/Loss": 0.761783242225647, "Full-finetune/Loss (Raw)": 0.798683226108551, "Full-finetune/Step": 2747, "Full-finetune/Step Time": 6.739131446927786} {"Full-finetune/Learning Rate": 9.554730105490722e-06, "Full-finetune/Loss": 0.7633962631225586, "Full-finetune/Loss (Raw)": 0.9364467263221741, "Full-finetune/Step": 2748, "Full-finetune/Step Time": 6.741223800927401} {"Full-finetune/Learning Rate": 9.548506760279506e-06, "Full-finetune/Loss": 0.7641544342041016, "Full-finetune/Loss (Raw)": 0.8584078550338745, "Full-finetune/Step": 2749, "Full-finetune/Step Time": 6.737973360344768} {"Full-finetune/Learning Rate": 9.542283590284003e-06, "Full-finetune/Loss": 0.7633664608001709, "Full-finetune/Loss (Raw)": 0.7829272150993347, "Full-finetune/Step": 2750, "Full-finetune/Step Time": 6.741832742467523} {"Full-finetune/Learning Rate": 9.5360605979193e-06, "Full-finetune/Loss": 0.7634596824645996, "Full-finetune/Loss (Raw)": 0.779083251953125, "Full-finetune/Step": 2751, "Full-finetune/Step Time": 6.730817589908838} {"Full-finetune/Learning Rate": 9.52983778560042e-06, "Full-finetune/Loss": 0.7634471654891968, "Full-finetune/Loss (Raw)": 0.8319844603538513, "Full-finetune/Step": 2752, "Full-finetune/Step Time": 6.730480536818504} {"Full-finetune/Learning Rate": 9.523615155742317e-06, "Full-finetune/Loss": 0.7660097479820251, "Full-finetune/Loss (Raw)": 0.9281879663467407, "Full-finetune/Step": 2753, "Full-finetune/Step Time": 6.716521434485912} {"Full-finetune/Learning Rate": 9.517392710759872e-06, "Full-finetune/Loss": 0.7697913646697998, "Full-finetune/Loss (Raw)": 0.8037369251251221, "Full-finetune/Step": 2754, "Full-finetune/Step Time": 6.696833783760667} {"Full-finetune/Learning Rate": 9.511170453067899e-06, "Full-finetune/Loss": 0.766505241394043, "Full-finetune/Loss (Raw)": 0.3641832768917084, "Full-finetune/Step": 2755, "Full-finetune/Step Time": 6.717159928753972} {"Full-finetune/Learning Rate": 9.504948385081121e-06, "Full-finetune/Loss": 0.7660876512527466, "Full-finetune/Loss (Raw)": 0.8272526860237122, "Full-finetune/Step": 2756, "Full-finetune/Step Time": 6.735102219507098} {"Full-finetune/Learning Rate": 9.498726509214216e-06, "Full-finetune/Loss": 0.7670965194702148, "Full-finetune/Loss (Raw)": 0.6585706472396851, "Full-finetune/Step": 2757, "Full-finetune/Step Time": 6.7513284888118505} {"Full-finetune/Learning Rate": 9.492504827881761e-06, "Full-finetune/Loss": 0.7685664892196655, "Full-finetune/Loss (Raw)": 0.8810256123542786, "Full-finetune/Step": 2758, "Full-finetune/Step Time": 6.75131668150425} {"Full-finetune/Learning Rate": 9.486283343498277e-06, "Full-finetune/Loss": 0.7671576738357544, "Full-finetune/Loss (Raw)": 0.7208183407783508, "Full-finetune/Step": 2759, "Full-finetune/Step Time": 6.7524734772741795} {"Full-finetune/Learning Rate": 9.480062058478201e-06, "Full-finetune/Loss": 0.7668434977531433, "Full-finetune/Loss (Raw)": 0.6725623607635498, "Full-finetune/Step": 2760, "Full-finetune/Step Time": 6.736387882381678} {"Full-finetune/Learning Rate": 9.473840975235883e-06, "Full-finetune/Loss": 0.7634941339492798, "Full-finetune/Loss (Raw)": 0.42942988872528076, "Full-finetune/Step": 2761, "Full-finetune/Step Time": 6.754828963428736} {"Full-finetune/Learning Rate": 9.467620096185616e-06, "Full-finetune/Loss": 0.7629958391189575, "Full-finetune/Loss (Raw)": 0.7261690497398376, "Full-finetune/Step": 2762, "Full-finetune/Step Time": 6.764850260689855} {"Full-finetune/Learning Rate": 9.46139942374159e-06, "Full-finetune/Loss": 0.7632591128349304, "Full-finetune/Loss (Raw)": 0.8159366846084595, "Full-finetune/Step": 2763, "Full-finetune/Step Time": 6.768596805632114} {"Full-finetune/Learning Rate": 9.455178960317936e-06, "Full-finetune/Loss": 0.7618275284767151, "Full-finetune/Loss (Raw)": 0.7770534753799438, "Full-finetune/Step": 2764, "Full-finetune/Step Time": 6.769480314105749} {"Full-finetune/Learning Rate": 9.448958708328694e-06, "Full-finetune/Loss": 0.7586793899536133, "Full-finetune/Loss (Raw)": 0.3573453426361084, "Full-finetune/Step": 2765, "Full-finetune/Step Time": 6.780347166582942} {"Full-finetune/Learning Rate": 9.442738670187814e-06, "Full-finetune/Loss": 0.7579530477523804, "Full-finetune/Loss (Raw)": 0.8116810917854309, "Full-finetune/Step": 2766, "Full-finetune/Step Time": 6.765887163579464} {"Full-finetune/Learning Rate": 9.436518848309178e-06, "Full-finetune/Loss": 0.7583757638931274, "Full-finetune/Loss (Raw)": 0.8279939889907837, "Full-finetune/Step": 2767, "Full-finetune/Step Time": 6.778943717479706} {"Full-finetune/Learning Rate": 9.430299245106574e-06, "Full-finetune/Loss": 0.7623234987258911, "Full-finetune/Loss (Raw)": 0.8592408895492554, "Full-finetune/Step": 2768, "Full-finetune/Step Time": 6.7623785473406315} {"Full-finetune/Learning Rate": 9.424079862993712e-06, "Full-finetune/Loss": 0.7627623081207275, "Full-finetune/Loss (Raw)": 0.8577396869659424, "Full-finetune/Step": 2769, "Full-finetune/Step Time": 6.769445892423391} {"Full-finetune/Learning Rate": 9.417860704384214e-06, "Full-finetune/Loss": 0.7607327699661255, "Full-finetune/Loss (Raw)": 0.5179170370101929, "Full-finetune/Step": 2770, "Full-finetune/Step Time": 6.784367008134723} {"Full-finetune/Learning Rate": 9.411641771691608e-06, "Full-finetune/Loss": 0.7626545429229736, "Full-finetune/Loss (Raw)": 0.876984715461731, "Full-finetune/Step": 2771, "Full-finetune/Step Time": 6.776959288865328} {"Full-finetune/Learning Rate": 9.405423067329344e-06, "Full-finetune/Loss": 0.7641952037811279, "Full-finetune/Loss (Raw)": 0.848652720451355, "Full-finetune/Step": 2772, "Full-finetune/Step Time": 6.744947383180261} {"Full-finetune/Learning Rate": 9.399204593710777e-06, "Full-finetune/Loss": 0.7650225758552551, "Full-finetune/Loss (Raw)": 0.8951627612113953, "Full-finetune/Step": 2773, "Full-finetune/Step Time": 6.747795393690467} {"Full-finetune/Learning Rate": 9.392986353249183e-06, "Full-finetune/Loss": 0.7613855600357056, "Full-finetune/Loss (Raw)": 0.33853042125701904, "Full-finetune/Step": 2774, "Full-finetune/Step Time": 6.764830682426691} {"Full-finetune/Learning Rate": 9.386768348357735e-06, "Full-finetune/Loss": 0.7624928951263428, "Full-finetune/Loss (Raw)": 0.8770667910575867, "Full-finetune/Step": 2775, "Full-finetune/Step Time": 6.755972541868687} {"Full-finetune/Learning Rate": 9.380550581449512e-06, "Full-finetune/Loss": 0.7619820237159729, "Full-finetune/Loss (Raw)": 0.7440145611763, "Full-finetune/Step": 2776, "Full-finetune/Step Time": 6.748731777071953} {"Full-finetune/Learning Rate": 9.37433305493752e-06, "Full-finetune/Loss": 0.7627606391906738, "Full-finetune/Loss (Raw)": 0.8078441023826599, "Full-finetune/Step": 2777, "Full-finetune/Step Time": 6.735924761742353} {"Full-finetune/Learning Rate": 9.368115771234651e-06, "Full-finetune/Loss": 0.7637699246406555, "Full-finetune/Loss (Raw)": 0.8381271958351135, "Full-finetune/Step": 2778, "Full-finetune/Step Time": 6.736763456836343} {"Full-finetune/Learning Rate": 9.361898732753715e-06, "Full-finetune/Loss": 0.7631993293762207, "Full-finetune/Loss (Raw)": 0.7116080522537231, "Full-finetune/Step": 2779, "Full-finetune/Step Time": 6.752701746299863} {"Full-finetune/Learning Rate": 9.355681941907428e-06, "Full-finetune/Loss": 0.7627836465835571, "Full-finetune/Loss (Raw)": 0.8122327327728271, "Full-finetune/Step": 2780, "Full-finetune/Step Time": 6.751969251781702} {"Full-finetune/Learning Rate": 9.349465401108393e-06, "Full-finetune/Loss": 0.7643511295318604, "Full-finetune/Loss (Raw)": 0.8967764973640442, "Full-finetune/Step": 2781, "Full-finetune/Step Time": 6.737658413127065} {"Full-finetune/Learning Rate": 9.34324911276914e-06, "Full-finetune/Loss": 0.7679932713508606, "Full-finetune/Loss (Raw)": 0.8164274096488953, "Full-finetune/Step": 2782, "Full-finetune/Step Time": 6.72163026407361} {"Full-finetune/Learning Rate": 9.337033079302082e-06, "Full-finetune/Loss": 0.7681533694267273, "Full-finetune/Loss (Raw)": 0.8228215575218201, "Full-finetune/Step": 2783, "Full-finetune/Step Time": 6.724289197474718} {"Full-finetune/Learning Rate": 9.330817303119545e-06, "Full-finetune/Loss": 0.7665855884552002, "Full-finetune/Loss (Raw)": 0.595652163028717, "Full-finetune/Step": 2784, "Full-finetune/Step Time": 6.736290585249662} {"Full-finetune/Learning Rate": 9.324601786633747e-06, "Full-finetune/Loss": 0.7681791186332703, "Full-finetune/Loss (Raw)": 0.7859656810760498, "Full-finetune/Step": 2785, "Full-finetune/Step Time": 6.726175643503666} {"Full-finetune/Learning Rate": 9.318386532256807e-06, "Full-finetune/Loss": 0.7705243825912476, "Full-finetune/Loss (Raw)": 0.8475020527839661, "Full-finetune/Step": 2786, "Full-finetune/Step Time": 6.703882463276386} {"Full-finetune/Learning Rate": 9.312171542400749e-06, "Full-finetune/Loss": 0.7699470520019531, "Full-finetune/Loss (Raw)": 0.7647238373756409, "Full-finetune/Step": 2787, "Full-finetune/Step Time": 6.693736096844077} {"Full-finetune/Learning Rate": 9.305956819477485e-06, "Full-finetune/Loss": 0.769831657409668, "Full-finetune/Loss (Raw)": 0.8673521876335144, "Full-finetune/Step": 2788, "Full-finetune/Step Time": 6.703894652426243} {"Full-finetune/Learning Rate": 9.299742365898831e-06, "Full-finetune/Loss": 0.7676897048950195, "Full-finetune/Loss (Raw)": 0.7425099611282349, "Full-finetune/Step": 2789, "Full-finetune/Step Time": 6.7162361815571785} {"Full-finetune/Learning Rate": 9.293528184076495e-06, "Full-finetune/Loss": 0.7671042680740356, "Full-finetune/Loss (Raw)": 0.7972844839096069, "Full-finetune/Step": 2790, "Full-finetune/Step Time": 6.720682566985488} {"Full-finetune/Learning Rate": 9.287314276422074e-06, "Full-finetune/Loss": 0.764782190322876, "Full-finetune/Loss (Raw)": 0.5959182381629944, "Full-finetune/Step": 2791, "Full-finetune/Step Time": 6.730711102485657} {"Full-finetune/Learning Rate": 9.28110064534707e-06, "Full-finetune/Loss": 0.7675203084945679, "Full-finetune/Loss (Raw)": 0.7943134903907776, "Full-finetune/Step": 2792, "Full-finetune/Step Time": 6.716653360053897} {"Full-finetune/Learning Rate": 9.274887293262869e-06, "Full-finetune/Loss": 0.7696269750595093, "Full-finetune/Loss (Raw)": 0.9357022643089294, "Full-finetune/Step": 2793, "Full-finetune/Step Time": 6.713141744956374} {"Full-finetune/Learning Rate": 9.268674222580757e-06, "Full-finetune/Loss": 0.7686945199966431, "Full-finetune/Loss (Raw)": 0.7421010136604309, "Full-finetune/Step": 2794, "Full-finetune/Step Time": 6.722391609102488} {"Full-finetune/Learning Rate": 9.262461435711898e-06, "Full-finetune/Loss": 0.7680047154426575, "Full-finetune/Loss (Raw)": 0.6419618725776672, "Full-finetune/Step": 2795, "Full-finetune/Step Time": 6.732959786430001} {"Full-finetune/Learning Rate": 9.256248935067356e-06, "Full-finetune/Loss": 0.767824649810791, "Full-finetune/Loss (Raw)": 0.7711002826690674, "Full-finetune/Step": 2796, "Full-finetune/Step Time": 6.749574035406113} {"Full-finetune/Learning Rate": 9.250036723058083e-06, "Full-finetune/Loss": 0.7652072906494141, "Full-finetune/Loss (Raw)": 0.5836660861968994, "Full-finetune/Step": 2797, "Full-finetune/Step Time": 6.762279646471143} {"Full-finetune/Learning Rate": 9.243824802094914e-06, "Full-finetune/Loss": 0.7616249322891235, "Full-finetune/Loss (Raw)": 0.3805523216724396, "Full-finetune/Step": 2798, "Full-finetune/Step Time": 6.785022361204028} {"Full-finetune/Learning Rate": 9.237613174588583e-06, "Full-finetune/Loss": 0.7618969678878784, "Full-finetune/Loss (Raw)": 0.6691336631774902, "Full-finetune/Step": 2799, "Full-finetune/Step Time": 6.784358581528068} {"Full-finetune/Learning Rate": 9.231401842949688e-06, "Full-finetune/Loss": 0.7623288631439209, "Full-finetune/Loss (Raw)": 0.615431010723114, "Full-finetune/Step": 2800, "Full-finetune/Step Time": 6.7789510153234005} {"Full-finetune/Learning Rate": 9.225190809588731e-06, "Full-finetune/Loss": 0.7629504799842834, "Full-finetune/Loss (Raw)": 0.8853389620780945, "Full-finetune/Step": 2801, "Full-finetune/Step Time": 6.792137250304222} {"Full-finetune/Learning Rate": 9.218980076916099e-06, "Full-finetune/Loss": 0.7647896409034729, "Full-finetune/Loss (Raw)": 0.8497109413146973, "Full-finetune/Step": 2802, "Full-finetune/Step Time": 6.791590875014663} {"Full-finetune/Learning Rate": 9.212769647342046e-06, "Full-finetune/Loss": 0.7641562223434448, "Full-finetune/Loss (Raw)": 0.7264434695243835, "Full-finetune/Step": 2803, "Full-finetune/Step Time": 6.7903107069432735} {"Full-finetune/Learning Rate": 9.206559523276731e-06, "Full-finetune/Loss": 0.7645899057388306, "Full-finetune/Loss (Raw)": 0.8343856930732727, "Full-finetune/Step": 2804, "Full-finetune/Step Time": 6.785099817439914} {"Full-finetune/Learning Rate": 9.200349707130168e-06, "Full-finetune/Loss": 0.7652264833450317, "Full-finetune/Loss (Raw)": 0.9618697166442871, "Full-finetune/Step": 2805, "Full-finetune/Step Time": 6.783777859061956} {"Full-finetune/Learning Rate": 9.194140201312274e-06, "Full-finetune/Loss": 0.7648545503616333, "Full-finetune/Loss (Raw)": 0.741973876953125, "Full-finetune/Step": 2806, "Full-finetune/Step Time": 6.790355607867241} {"Full-finetune/Learning Rate": 9.187931008232832e-06, "Full-finetune/Loss": 0.7634362578392029, "Full-finetune/Loss (Raw)": 0.7832857370376587, "Full-finetune/Step": 2807, "Full-finetune/Step Time": 6.787688547745347} {"Full-finetune/Learning Rate": 9.181722130301512e-06, "Full-finetune/Loss": 0.7613961696624756, "Full-finetune/Loss (Raw)": 0.6124240756034851, "Full-finetune/Step": 2808, "Full-finetune/Step Time": 6.79900505207479} {"Full-finetune/Learning Rate": 9.175513569927865e-06, "Full-finetune/Loss": 0.761460542678833, "Full-finetune/Loss (Raw)": 0.3508017063140869, "Full-finetune/Step": 2809, "Full-finetune/Step Time": 6.799843840301037} {"Full-finetune/Learning Rate": 9.169305329521296e-06, "Full-finetune/Loss": 0.7605305314064026, "Full-finetune/Loss (Raw)": 0.6881993412971497, "Full-finetune/Step": 2810, "Full-finetune/Step Time": 6.7985868994146585} {"Full-finetune/Learning Rate": 9.163097411491115e-06, "Full-finetune/Loss": 0.7605422139167786, "Full-finetune/Loss (Raw)": 0.8356348872184753, "Full-finetune/Step": 2811, "Full-finetune/Step Time": 6.801783308386803} {"Full-finetune/Learning Rate": 9.156889818246485e-06, "Full-finetune/Loss": 0.7604987621307373, "Full-finetune/Loss (Raw)": 0.7783032655715942, "Full-finetune/Step": 2812, "Full-finetune/Step Time": 6.800444481894374} {"Full-finetune/Learning Rate": 9.150682552196461e-06, "Full-finetune/Loss": 0.7607067823410034, "Full-finetune/Loss (Raw)": 0.8735059499740601, "Full-finetune/Step": 2813, "Full-finetune/Step Time": 6.79907245375216} {"Full-finetune/Learning Rate": 9.144475615749957e-06, "Full-finetune/Loss": 0.7578656077384949, "Full-finetune/Loss (Raw)": 0.3576015830039978, "Full-finetune/Step": 2814, "Full-finetune/Step Time": 6.8144593723118305} {"Full-finetune/Learning Rate": 9.138269011315761e-06, "Full-finetune/Loss": 0.7588862180709839, "Full-finetune/Loss (Raw)": 0.9194666743278503, "Full-finetune/Step": 2815, "Full-finetune/Step Time": 6.807449785992503} {"Full-finetune/Learning Rate": 9.13206274130254e-06, "Full-finetune/Loss": 0.7585833668708801, "Full-finetune/Loss (Raw)": 0.8140989542007446, "Full-finetune/Step": 2816, "Full-finetune/Step Time": 6.804526671767235} {"Full-finetune/Learning Rate": 9.125856808118822e-06, "Full-finetune/Loss": 0.7573602199554443, "Full-finetune/Loss (Raw)": 0.787402331829071, "Full-finetune/Step": 2817, "Full-finetune/Step Time": 6.81172663718462} {"Full-finetune/Learning Rate": 9.119651214173012e-06, "Full-finetune/Loss": 0.7605551481246948, "Full-finetune/Loss (Raw)": 0.7869833707809448, "Full-finetune/Step": 2818, "Full-finetune/Step Time": 6.798436658456922} {"Full-finetune/Learning Rate": 9.11344596187338e-06, "Full-finetune/Loss": 0.7609373331069946, "Full-finetune/Loss (Raw)": 0.9372866153717041, "Full-finetune/Step": 2819, "Full-finetune/Step Time": 6.798915537074208} {"Full-finetune/Learning Rate": 9.107241053628058e-06, "Full-finetune/Loss": 0.7620043754577637, "Full-finetune/Loss (Raw)": 0.9136375188827515, "Full-finetune/Step": 2820, "Full-finetune/Step Time": 6.7981254030019045} {"Full-finetune/Learning Rate": 9.101036491845056e-06, "Full-finetune/Loss": 0.7604188919067383, "Full-finetune/Loss (Raw)": 0.6363196969032288, "Full-finetune/Step": 2821, "Full-finetune/Step Time": 6.810210907831788} {"Full-finetune/Learning Rate": 9.094832278932238e-06, "Full-finetune/Loss": 0.7611827254295349, "Full-finetune/Loss (Raw)": 0.838303804397583, "Full-finetune/Step": 2822, "Full-finetune/Step Time": 6.805678902193904} {"Full-finetune/Learning Rate": 9.088628417297346e-06, "Full-finetune/Loss": 0.7583527565002441, "Full-finetune/Loss (Raw)": 0.4104416072368622, "Full-finetune/Step": 2823, "Full-finetune/Step Time": 6.816162995994091} {"Full-finetune/Learning Rate": 9.082424909347969e-06, "Full-finetune/Loss": 0.7568458318710327, "Full-finetune/Loss (Raw)": 0.6053063869476318, "Full-finetune/Step": 2824, "Full-finetune/Step Time": 6.803958043456078} {"Full-finetune/Learning Rate": 9.076221757491568e-06, "Full-finetune/Loss": 0.7560635805130005, "Full-finetune/Loss (Raw)": 0.7042155861854553, "Full-finetune/Step": 2825, "Full-finetune/Step Time": 6.8134179431945086} {"Full-finetune/Learning Rate": 9.070018964135471e-06, "Full-finetune/Loss": 0.755750834941864, "Full-finetune/Loss (Raw)": 0.6734318733215332, "Full-finetune/Step": 2826, "Full-finetune/Step Time": 6.814985658973455} {"Full-finetune/Learning Rate": 9.063816531686855e-06, "Full-finetune/Loss": 0.7552330493927002, "Full-finetune/Loss (Raw)": 0.8183548450469971, "Full-finetune/Step": 2827, "Full-finetune/Step Time": 6.797027330845594} {"Full-finetune/Learning Rate": 9.057614462552769e-06, "Full-finetune/Loss": 0.7564375400543213, "Full-finetune/Loss (Raw)": 0.9489638805389404, "Full-finetune/Step": 2828, "Full-finetune/Step Time": 6.796747179701924} {"Full-finetune/Learning Rate": 9.05141275914011e-06, "Full-finetune/Loss": 0.7569669485092163, "Full-finetune/Loss (Raw)": 0.8396981358528137, "Full-finetune/Step": 2829, "Full-finetune/Step Time": 6.778568781912327} {"Full-finetune/Learning Rate": 9.045211423855637e-06, "Full-finetune/Loss": 0.757332980632782, "Full-finetune/Loss (Raw)": 0.8915844559669495, "Full-finetune/Step": 2830, "Full-finetune/Step Time": 6.770833743736148} {"Full-finetune/Learning Rate": 9.039010459105974e-06, "Full-finetune/Loss": 0.7598408460617065, "Full-finetune/Loss (Raw)": 1.017932415008545, "Full-finetune/Step": 2831, "Full-finetune/Step Time": 6.754045020788908} {"Full-finetune/Learning Rate": 9.032809867297586e-06, "Full-finetune/Loss": 0.7618024945259094, "Full-finetune/Loss (Raw)": 0.9399772882461548, "Full-finetune/Step": 2832, "Full-finetune/Step Time": 6.759124545380473} {"Full-finetune/Learning Rate": 9.02660965083681e-06, "Full-finetune/Loss": 0.7620981335639954, "Full-finetune/Loss (Raw)": 0.8004100918769836, "Full-finetune/Step": 2833, "Full-finetune/Step Time": 6.766410533338785} {"Full-finetune/Learning Rate": 9.020409812129824e-06, "Full-finetune/Loss": 0.7580794095993042, "Full-finetune/Loss (Raw)": 0.36456480622291565, "Full-finetune/Step": 2834, "Full-finetune/Step Time": 6.775859989225864} {"Full-finetune/Learning Rate": 9.014210353582662e-06, "Full-finetune/Loss": 0.7570033073425293, "Full-finetune/Loss (Raw)": 0.8323854804039001, "Full-finetune/Step": 2835, "Full-finetune/Step Time": 6.77935847081244} {"Full-finetune/Learning Rate": 9.008011277601217e-06, "Full-finetune/Loss": 0.7548806667327881, "Full-finetune/Loss (Raw)": 0.6284418106079102, "Full-finetune/Step": 2836, "Full-finetune/Step Time": 6.772228999063373} {"Full-finetune/Learning Rate": 9.001812586591228e-06, "Full-finetune/Loss": 0.7555116415023804, "Full-finetune/Loss (Raw)": 0.8463861346244812, "Full-finetune/Step": 2837, "Full-finetune/Step Time": 6.759341958910227} {"Full-finetune/Learning Rate": 8.995614282958289e-06, "Full-finetune/Loss": 0.7580209374427795, "Full-finetune/Loss (Raw)": 0.700954258441925, "Full-finetune/Step": 2838, "Full-finetune/Step Time": 6.737252399325371} {"Full-finetune/Learning Rate": 8.989416369107837e-06, "Full-finetune/Loss": 0.7567486763000488, "Full-finetune/Loss (Raw)": 0.7012578248977661, "Full-finetune/Step": 2839, "Full-finetune/Step Time": 6.740882996469736} {"Full-finetune/Learning Rate": 8.983218847445157e-06, "Full-finetune/Loss": 0.7566007375717163, "Full-finetune/Loss (Raw)": 0.7998656630516052, "Full-finetune/Step": 2840, "Full-finetune/Step Time": 6.724171394482255} {"Full-finetune/Learning Rate": 8.977021720375395e-06, "Full-finetune/Loss": 0.7577977776527405, "Full-finetune/Loss (Raw)": 0.94114089012146, "Full-finetune/Step": 2841, "Full-finetune/Step Time": 6.7188468258827925} {"Full-finetune/Learning Rate": 8.970824990303532e-06, "Full-finetune/Loss": 0.757325291633606, "Full-finetune/Loss (Raw)": 0.8194852471351624, "Full-finetune/Step": 2842, "Full-finetune/Step Time": 6.720086749643087} {"Full-finetune/Learning Rate": 8.9646286596344e-06, "Full-finetune/Loss": 0.7569615840911865, "Full-finetune/Loss (Raw)": 0.674643874168396, "Full-finetune/Step": 2843, "Full-finetune/Step Time": 6.72515613399446} {"Full-finetune/Learning Rate": 8.958432730772667e-06, "Full-finetune/Loss": 0.7574392557144165, "Full-finetune/Loss (Raw)": 0.8494681715965271, "Full-finetune/Step": 2844, "Full-finetune/Step Time": 6.714585807174444} {"Full-finetune/Learning Rate": 8.952237206122855e-06, "Full-finetune/Loss": 0.7571576833724976, "Full-finetune/Loss (Raw)": 0.8122336268424988, "Full-finetune/Step": 2845, "Full-finetune/Step Time": 6.725779354572296} {"Full-finetune/Learning Rate": 8.946042088089331e-06, "Full-finetune/Loss": 0.757112443447113, "Full-finetune/Loss (Raw)": 0.7471479177474976, "Full-finetune/Step": 2846, "Full-finetune/Step Time": 6.712595848366618} {"Full-finetune/Learning Rate": 8.939847379076293e-06, "Full-finetune/Loss": 0.7554175853729248, "Full-finetune/Loss (Raw)": 0.6004665493965149, "Full-finetune/Step": 2847, "Full-finetune/Step Time": 6.724542802199721} {"Full-finetune/Learning Rate": 8.933653081487794e-06, "Full-finetune/Loss": 0.7554093599319458, "Full-finetune/Loss (Raw)": 0.8586068153381348, "Full-finetune/Step": 2848, "Full-finetune/Step Time": 6.731206685304642} {"Full-finetune/Learning Rate": 8.927459197727712e-06, "Full-finetune/Loss": 0.7558004856109619, "Full-finetune/Loss (Raw)": 0.7536103129386902, "Full-finetune/Step": 2849, "Full-finetune/Step Time": 6.727695196866989} {"Full-finetune/Learning Rate": 8.921265730199775e-06, "Full-finetune/Loss": 0.7557886242866516, "Full-finetune/Loss (Raw)": 0.7969295978546143, "Full-finetune/Step": 2850, "Full-finetune/Step Time": 6.722445139661431} {"Full-finetune/Learning Rate": 8.91507268130755e-06, "Full-finetune/Loss": 0.7567011117935181, "Full-finetune/Loss (Raw)": 0.8468343019485474, "Full-finetune/Step": 2851, "Full-finetune/Step Time": 6.719250597059727} {"Full-finetune/Learning Rate": 8.908880053454432e-06, "Full-finetune/Loss": 0.7568148374557495, "Full-finetune/Loss (Raw)": 0.828019917011261, "Full-finetune/Step": 2852, "Full-finetune/Step Time": 6.71967988461256} {"Full-finetune/Learning Rate": 8.90268784904367e-06, "Full-finetune/Loss": 0.7564519643783569, "Full-finetune/Loss (Raw)": 0.757138729095459, "Full-finetune/Step": 2853, "Full-finetune/Step Time": 6.721030840650201} {"Full-finetune/Learning Rate": 8.896496070478326e-06, "Full-finetune/Loss": 0.7569757699966431, "Full-finetune/Loss (Raw)": 0.8743531107902527, "Full-finetune/Step": 2854, "Full-finetune/Step Time": 6.714190512895584} {"Full-finetune/Learning Rate": 8.890304720161311e-06, "Full-finetune/Loss": 0.7569800615310669, "Full-finetune/Loss (Raw)": 0.7818887829780579, "Full-finetune/Step": 2855, "Full-finetune/Step Time": 6.718850934877992} {"Full-finetune/Learning Rate": 8.884113800495372e-06, "Full-finetune/Loss": 0.7586413025856018, "Full-finetune/Loss (Raw)": 0.7680987119674683, "Full-finetune/Step": 2856, "Full-finetune/Step Time": 6.732876636087894} {"Full-finetune/Learning Rate": 8.87792331388308e-06, "Full-finetune/Loss": 0.761860728263855, "Full-finetune/Loss (Raw)": 0.8369755744934082, "Full-finetune/Step": 2857, "Full-finetune/Step Time": 6.718758603557944} {"Full-finetune/Learning Rate": 8.871733262726846e-06, "Full-finetune/Loss": 0.7626045942306519, "Full-finetune/Loss (Raw)": 0.7146444916725159, "Full-finetune/Step": 2858, "Full-finetune/Step Time": 6.710503526031971} {"Full-finetune/Learning Rate": 8.865543649428903e-06, "Full-finetune/Loss": 0.76263427734375, "Full-finetune/Loss (Raw)": 0.8151801228523254, "Full-finetune/Step": 2859, "Full-finetune/Step Time": 6.717171832919121} {"Full-finetune/Learning Rate": 8.859354476391319e-06, "Full-finetune/Loss": 0.761573076248169, "Full-finetune/Loss (Raw)": 0.6816263198852539, "Full-finetune/Step": 2860, "Full-finetune/Step Time": 6.726994015276432} {"Full-finetune/Learning Rate": 8.853165746015997e-06, "Full-finetune/Loss": 0.7624026536941528, "Full-finetune/Loss (Raw)": 0.9658241271972656, "Full-finetune/Step": 2861, "Full-finetune/Step Time": 6.737756362184882} {"Full-finetune/Learning Rate": 8.846977460704656e-06, "Full-finetune/Loss": 0.7623789310455322, "Full-finetune/Loss (Raw)": 0.7682144641876221, "Full-finetune/Step": 2862, "Full-finetune/Step Time": 6.753432517871261} {"Full-finetune/Learning Rate": 8.840789622858856e-06, "Full-finetune/Loss": 0.7616990804672241, "Full-finetune/Loss (Raw)": 0.7834464907646179, "Full-finetune/Step": 2863, "Full-finetune/Step Time": 6.750836139544845} {"Full-finetune/Learning Rate": 8.834602234879968e-06, "Full-finetune/Loss": 0.7635968923568726, "Full-finetune/Loss (Raw)": 0.709340512752533, "Full-finetune/Step": 2864, "Full-finetune/Step Time": 6.73138253390789} {"Full-finetune/Learning Rate": 8.8284152991692e-06, "Full-finetune/Loss": 0.7662267684936523, "Full-finetune/Loss (Raw)": 0.8770820498466492, "Full-finetune/Step": 2865, "Full-finetune/Step Time": 6.721709828823805} {"Full-finetune/Learning Rate": 8.822228818127584e-06, "Full-finetune/Loss": 0.7626429200172424, "Full-finetune/Loss (Raw)": 0.3961203992366791, "Full-finetune/Step": 2866, "Full-finetune/Step Time": 6.735444277524948} {"Full-finetune/Learning Rate": 8.81604279415597e-06, "Full-finetune/Loss": 0.762325644493103, "Full-finetune/Loss (Raw)": 0.8413649797439575, "Full-finetune/Step": 2867, "Full-finetune/Step Time": 6.739245232194662} {"Full-finetune/Learning Rate": 8.809857229655029e-06, "Full-finetune/Loss": 0.7642199397087097, "Full-finetune/Loss (Raw)": 0.930615246295929, "Full-finetune/Step": 2868, "Full-finetune/Step Time": 6.733357980847359} {"Full-finetune/Learning Rate": 8.803672127025264e-06, "Full-finetune/Loss": 0.7620168924331665, "Full-finetune/Loss (Raw)": 0.6046423316001892, "Full-finetune/Step": 2869, "Full-finetune/Step Time": 6.763688558712602} {"Full-finetune/Learning Rate": 8.797487488666986e-06, "Full-finetune/Loss": 0.7589120864868164, "Full-finetune/Loss (Raw)": 0.37815526127815247, "Full-finetune/Step": 2870, "Full-finetune/Step Time": 6.774615500122309} {"Full-finetune/Learning Rate": 8.791303316980342e-06, "Full-finetune/Loss": 0.7593104839324951, "Full-finetune/Loss (Raw)": 0.9466869831085205, "Full-finetune/Step": 2871, "Full-finetune/Step Time": 6.775100465863943} {"Full-finetune/Learning Rate": 8.785119614365282e-06, "Full-finetune/Loss": 0.7600720524787903, "Full-finetune/Loss (Raw)": 0.9302672147750854, "Full-finetune/Step": 2872, "Full-finetune/Step Time": 6.764499790966511} {"Full-finetune/Learning Rate": 8.77893638322158e-06, "Full-finetune/Loss": 0.7595624923706055, "Full-finetune/Loss (Raw)": 0.841850996017456, "Full-finetune/Step": 2873, "Full-finetune/Step Time": 6.76592168956995} {"Full-finetune/Learning Rate": 8.772753625948828e-06, "Full-finetune/Loss": 0.7560546398162842, "Full-finetune/Loss (Raw)": 0.36423611640930176, "Full-finetune/Step": 2874, "Full-finetune/Step Time": 6.770318454131484} {"Full-finetune/Learning Rate": 8.766571344946433e-06, "Full-finetune/Loss": 0.7566827535629272, "Full-finetune/Loss (Raw)": 0.8790837526321411, "Full-finetune/Step": 2875, "Full-finetune/Step Time": 6.751256283372641} {"Full-finetune/Learning Rate": 8.760389542613619e-06, "Full-finetune/Loss": 0.7556951642036438, "Full-finetune/Loss (Raw)": 0.8100324869155884, "Full-finetune/Step": 2876, "Full-finetune/Step Time": 6.760555244982243} {"Full-finetune/Learning Rate": 8.754208221349427e-06, "Full-finetune/Loss": 0.7551730275154114, "Full-finetune/Loss (Raw)": 0.7915760278701782, "Full-finetune/Step": 2877, "Full-finetune/Step Time": 6.762698709964752} {"Full-finetune/Learning Rate": 8.748027383552696e-06, "Full-finetune/Loss": 0.7546952366828918, "Full-finetune/Loss (Raw)": 0.7217738032341003, "Full-finetune/Step": 2878, "Full-finetune/Step Time": 6.77849986590445} {"Full-finetune/Learning Rate": 8.741847031622096e-06, "Full-finetune/Loss": 0.7542762756347656, "Full-finetune/Loss (Raw)": 0.7254584431648254, "Full-finetune/Step": 2879, "Full-finetune/Step Time": 6.779352147132158} {"Full-finetune/Learning Rate": 8.7356671679561e-06, "Full-finetune/Loss": 0.7549030780792236, "Full-finetune/Loss (Raw)": 0.9122114181518555, "Full-finetune/Step": 2880, "Full-finetune/Step Time": 6.770642532035708} {"Full-finetune/Learning Rate": 8.729487794952992e-06, "Full-finetune/Loss": 0.7541334629058838, "Full-finetune/Loss (Raw)": 0.8296738266944885, "Full-finetune/Step": 2881, "Full-finetune/Step Time": 6.775797370821238} {"Full-finetune/Learning Rate": 8.72330891501087e-06, "Full-finetune/Loss": 0.7547643184661865, "Full-finetune/Loss (Raw)": 0.8844878673553467, "Full-finetune/Step": 2882, "Full-finetune/Step Time": 6.776427678763866} {"Full-finetune/Learning Rate": 8.717130530527626e-06, "Full-finetune/Loss": 0.759105920791626, "Full-finetune/Loss (Raw)": 0.9199130535125732, "Full-finetune/Step": 2883, "Full-finetune/Step Time": 6.761012315750122} {"Full-finetune/Learning Rate": 8.710952643900983e-06, "Full-finetune/Loss": 0.758077085018158, "Full-finetune/Loss (Raw)": 0.6955611109733582, "Full-finetune/Step": 2884, "Full-finetune/Step Time": 6.764991531148553} {"Full-finetune/Learning Rate": 8.704775257528446e-06, "Full-finetune/Loss": 0.7597655057907104, "Full-finetune/Loss (Raw)": 0.8746919631958008, "Full-finetune/Step": 2885, "Full-finetune/Step Time": 6.73253646120429} {"Full-finetune/Learning Rate": 8.69859837380735e-06, "Full-finetune/Loss": 0.7573330402374268, "Full-finetune/Loss (Raw)": 0.5696653723716736, "Full-finetune/Step": 2886, "Full-finetune/Step Time": 6.754935996606946} {"Full-finetune/Learning Rate": 8.692421995134819e-06, "Full-finetune/Loss": 0.7569748163223267, "Full-finetune/Loss (Raw)": 0.674969494342804, "Full-finetune/Step": 2887, "Full-finetune/Step Time": 6.758829984813929} {"Full-finetune/Learning Rate": 8.686246123907778e-06, "Full-finetune/Loss": 0.7594555616378784, "Full-finetune/Loss (Raw)": 0.9900926947593689, "Full-finetune/Step": 2888, "Full-finetune/Step Time": 6.766798585653305} {"Full-finetune/Learning Rate": 8.680070762522969e-06, "Full-finetune/Loss": 0.7587656378746033, "Full-finetune/Loss (Raw)": 0.341122567653656, "Full-finetune/Step": 2889, "Full-finetune/Step Time": 6.764825960621238} {"Full-finetune/Learning Rate": 8.673895913376928e-06, "Full-finetune/Loss": 0.7589738368988037, "Full-finetune/Loss (Raw)": 0.7528219223022461, "Full-finetune/Step": 2890, "Full-finetune/Step Time": 6.764426622539759} {"Full-finetune/Learning Rate": 8.667721578865993e-06, "Full-finetune/Loss": 0.7586124539375305, "Full-finetune/Loss (Raw)": 0.7696805596351624, "Full-finetune/Step": 2891, "Full-finetune/Step Time": 6.76331364735961} {"Full-finetune/Learning Rate": 8.661547761386307e-06, "Full-finetune/Loss": 0.760248064994812, "Full-finetune/Loss (Raw)": 0.9863989353179932, "Full-finetune/Step": 2892, "Full-finetune/Step Time": 6.771637445315719} {"Full-finetune/Learning Rate": 8.655374463333797e-06, "Full-finetune/Loss": 0.7635348439216614, "Full-finetune/Loss (Raw)": 0.7780593633651733, "Full-finetune/Step": 2893, "Full-finetune/Step Time": 6.753025397658348} {"Full-finetune/Learning Rate": 8.649201687104209e-06, "Full-finetune/Loss": 0.76424241065979, "Full-finetune/Loss (Raw)": 0.9022514224052429, "Full-finetune/Step": 2894, "Full-finetune/Step Time": 6.750264963135123} {"Full-finetune/Learning Rate": 8.643029435093072e-06, "Full-finetune/Loss": 0.7642450928688049, "Full-finetune/Loss (Raw)": 0.8283393979072571, "Full-finetune/Step": 2895, "Full-finetune/Step Time": 6.737013481557369} {"Full-finetune/Learning Rate": 8.63685770969572e-06, "Full-finetune/Loss": 0.7643337249755859, "Full-finetune/Loss (Raw)": 0.870581328868866, "Full-finetune/Step": 2896, "Full-finetune/Step Time": 6.739616272971034} {"Full-finetune/Learning Rate": 8.630686513307282e-06, "Full-finetune/Loss": 0.7644178867340088, "Full-finetune/Loss (Raw)": 0.868513822555542, "Full-finetune/Step": 2897, "Full-finetune/Step Time": 6.733158668503165} {"Full-finetune/Learning Rate": 8.62451584832267e-06, "Full-finetune/Loss": 0.7667961716651917, "Full-finetune/Loss (Raw)": 0.8223376274108887, "Full-finetune/Step": 2898, "Full-finetune/Step Time": 6.717774260789156} {"Full-finetune/Learning Rate": 8.618345717136604e-06, "Full-finetune/Loss": 0.7661935687065125, "Full-finetune/Loss (Raw)": 0.7998559474945068, "Full-finetune/Step": 2899, "Full-finetune/Step Time": 6.717212326824665} {"Full-finetune/Learning Rate": 8.612176122143589e-06, "Full-finetune/Loss": 0.7647798657417297, "Full-finetune/Loss (Raw)": 0.6676947474479675, "Full-finetune/Step": 2900, "Full-finetune/Step Time": 6.721502870321274} {"Full-finetune/Learning Rate": 8.606007065737928e-06, "Full-finetune/Loss": 0.7630089521408081, "Full-finetune/Loss (Raw)": 0.6684815883636475, "Full-finetune/Step": 2901, "Full-finetune/Step Time": 6.719863122329116} {"Full-finetune/Learning Rate": 8.599838550313714e-06, "Full-finetune/Loss": 0.76650470495224, "Full-finetune/Loss (Raw)": 0.7859897613525391, "Full-finetune/Step": 2902, "Full-finetune/Step Time": 6.724581243470311} {"Full-finetune/Learning Rate": 8.593670578264814e-06, "Full-finetune/Loss": 0.7661482095718384, "Full-finetune/Loss (Raw)": 0.8314346671104431, "Full-finetune/Step": 2903, "Full-finetune/Step Time": 6.726560853421688} {"Full-finetune/Learning Rate": 8.587503151984911e-06, "Full-finetune/Loss": 0.766514241695404, "Full-finetune/Loss (Raw)": 0.7908673882484436, "Full-finetune/Step": 2904, "Full-finetune/Step Time": 6.723606713116169} {"Full-finetune/Learning Rate": 8.581336273867455e-06, "Full-finetune/Loss": 0.7665130496025085, "Full-finetune/Loss (Raw)": 0.8076924681663513, "Full-finetune/Step": 2905, "Full-finetune/Step Time": 6.725560894235969} {"Full-finetune/Learning Rate": 8.575169946305699e-06, "Full-finetune/Loss": 0.7668648958206177, "Full-finetune/Loss (Raw)": 0.8831698894500732, "Full-finetune/Step": 2906, "Full-finetune/Step Time": 6.726919298991561} {"Full-finetune/Learning Rate": 8.569004171692667e-06, "Full-finetune/Loss": 0.7675352692604065, "Full-finetune/Loss (Raw)": 0.7974116206169128, "Full-finetune/Step": 2907, "Full-finetune/Step Time": 6.7081074099987745} {"Full-finetune/Learning Rate": 8.562838952421177e-06, "Full-finetune/Loss": 0.7679750919342041, "Full-finetune/Loss (Raw)": 0.8685266375541687, "Full-finetune/Step": 2908, "Full-finetune/Step Time": 6.710909266024828} {"Full-finetune/Learning Rate": 8.556674290883833e-06, "Full-finetune/Loss": 0.768735408782959, "Full-finetune/Loss (Raw)": 0.9940990209579468, "Full-finetune/Step": 2909, "Full-finetune/Step Time": 6.704460406675935} {"Full-finetune/Learning Rate": 8.550510189473016e-06, "Full-finetune/Loss": 0.7695345878601074, "Full-finetune/Loss (Raw)": 0.9187219738960266, "Full-finetune/Step": 2910, "Full-finetune/Step Time": 6.706360127776861} {"Full-finetune/Learning Rate": 8.544346650580899e-06, "Full-finetune/Loss": 0.7705353498458862, "Full-finetune/Loss (Raw)": 0.950921356678009, "Full-finetune/Step": 2911, "Full-finetune/Step Time": 6.6916117910295725} {"Full-finetune/Learning Rate": 8.538183676599426e-06, "Full-finetune/Loss": 0.7724307775497437, "Full-finetune/Loss (Raw)": 0.8382619023323059, "Full-finetune/Step": 2912, "Full-finetune/Step Time": 6.684196721762419} {"Full-finetune/Learning Rate": 8.532021269920326e-06, "Full-finetune/Loss": 0.7728248238563538, "Full-finetune/Loss (Raw)": 0.8364052176475525, "Full-finetune/Step": 2913, "Full-finetune/Step Time": 6.6761712692677975} {"Full-finetune/Learning Rate": 8.525859432935114e-06, "Full-finetune/Loss": 0.7709529399871826, "Full-finetune/Loss (Raw)": 0.6079071164131165, "Full-finetune/Step": 2914, "Full-finetune/Step Time": 6.701934520155191} {"Full-finetune/Learning Rate": 8.519698168035074e-06, "Full-finetune/Loss": 0.7693582773208618, "Full-finetune/Loss (Raw)": 0.5606043934822083, "Full-finetune/Step": 2915, "Full-finetune/Step Time": 6.713115243241191} {"Full-finetune/Learning Rate": 8.513537477611278e-06, "Full-finetune/Loss": 0.7693228721618652, "Full-finetune/Loss (Raw)": 0.8628146052360535, "Full-finetune/Step": 2916, "Full-finetune/Step Time": 6.700535951182246} {"Full-finetune/Learning Rate": 8.507377364054566e-06, "Full-finetune/Loss": 0.7707074880599976, "Full-finetune/Loss (Raw)": 0.9197455048561096, "Full-finetune/Step": 2917, "Full-finetune/Step Time": 6.684250188991427} {"Full-finetune/Learning Rate": 8.501217829755554e-06, "Full-finetune/Loss": 0.7700283527374268, "Full-finetune/Loss (Raw)": 0.7103591561317444, "Full-finetune/Step": 2918, "Full-finetune/Step Time": 6.6990401819348335} {"Full-finetune/Learning Rate": 8.495058877104642e-06, "Full-finetune/Loss": 0.7718932628631592, "Full-finetune/Loss (Raw)": 0.8346192836761475, "Full-finetune/Step": 2919, "Full-finetune/Step Time": 6.677443427965045} {"Full-finetune/Learning Rate": 8.488900508491999e-06, "Full-finetune/Loss": 0.768445611000061, "Full-finetune/Loss (Raw)": 0.35301950573921204, "Full-finetune/Step": 2920, "Full-finetune/Step Time": 6.681021098047495} {"Full-finetune/Learning Rate": 8.48274272630757e-06, "Full-finetune/Loss": 0.7671964168548584, "Full-finetune/Loss (Raw)": 0.7757996320724487, "Full-finetune/Step": 2921, "Full-finetune/Step Time": 6.699011446908116} {"Full-finetune/Learning Rate": 8.476585532941064e-06, "Full-finetune/Loss": 0.7672152519226074, "Full-finetune/Loss (Raw)": 0.7445200681686401, "Full-finetune/Step": 2922, "Full-finetune/Step Time": 6.691608253866434} {"Full-finetune/Learning Rate": 8.470428930781969e-06, "Full-finetune/Loss": 0.7677110433578491, "Full-finetune/Loss (Raw)": 0.705413281917572, "Full-finetune/Step": 2923, "Full-finetune/Step Time": 6.712497660890222} {"Full-finetune/Learning Rate": 8.464272922219547e-06, "Full-finetune/Loss": 0.7685731649398804, "Full-finetune/Loss (Raw)": 0.8814582824707031, "Full-finetune/Step": 2924, "Full-finetune/Step Time": 6.69557861238718} {"Full-finetune/Learning Rate": 8.458117509642817e-06, "Full-finetune/Loss": 0.7710670232772827, "Full-finetune/Loss (Raw)": 0.9028748869895935, "Full-finetune/Step": 2925, "Full-finetune/Step Time": 6.685575786978006} {"Full-finetune/Learning Rate": 8.451962695440586e-06, "Full-finetune/Loss": 0.7743411660194397, "Full-finetune/Loss (Raw)": 0.7996482849121094, "Full-finetune/Step": 2926, "Full-finetune/Step Time": 6.693858824670315} {"Full-finetune/Learning Rate": 8.445808482001405e-06, "Full-finetune/Loss": 0.7733521461486816, "Full-finetune/Loss (Raw)": 0.5425321459770203, "Full-finetune/Step": 2927, "Full-finetune/Step Time": 6.701302213594317} {"Full-finetune/Learning Rate": 8.439654871713608e-06, "Full-finetune/Loss": 0.7747315168380737, "Full-finetune/Loss (Raw)": 0.7919945120811462, "Full-finetune/Step": 2928, "Full-finetune/Step Time": 6.7101645059883595} {"Full-finetune/Learning Rate": 8.433501866965294e-06, "Full-finetune/Loss": 0.7744442820549011, "Full-finetune/Loss (Raw)": 0.8485722541809082, "Full-finetune/Step": 2929, "Full-finetune/Step Time": 6.69051062874496} {"Full-finetune/Learning Rate": 8.42734947014432e-06, "Full-finetune/Loss": 0.7746382355690002, "Full-finetune/Loss (Raw)": 0.8745393753051758, "Full-finetune/Step": 2930, "Full-finetune/Step Time": 6.683732563629746} {"Full-finetune/Learning Rate": 8.42119768363832e-06, "Full-finetune/Loss": 0.7758084535598755, "Full-finetune/Loss (Raw)": 0.8762328028678894, "Full-finetune/Step": 2931, "Full-finetune/Step Time": 6.6865724958479404} {"Full-finetune/Learning Rate": 8.415046509834668e-06, "Full-finetune/Loss": 0.7760946750640869, "Full-finetune/Loss (Raw)": 0.8710190653800964, "Full-finetune/Step": 2932, "Full-finetune/Step Time": 6.689765045419335} {"Full-finetune/Learning Rate": 8.408895951120526e-06, "Full-finetune/Loss": 0.7749407291412354, "Full-finetune/Loss (Raw)": 0.8141602277755737, "Full-finetune/Step": 2933, "Full-finetune/Step Time": 6.700336303561926} {"Full-finetune/Learning Rate": 8.402746009882798e-06, "Full-finetune/Loss": 0.7752948999404907, "Full-finetune/Loss (Raw)": 0.7873131036758423, "Full-finetune/Step": 2934, "Full-finetune/Step Time": 6.704155847430229} {"Full-finetune/Learning Rate": 8.39659668850816e-06, "Full-finetune/Loss": 0.7754607200622559, "Full-finetune/Loss (Raw)": 0.8045114874839783, "Full-finetune/Step": 2935, "Full-finetune/Step Time": 6.709290025755763} {"Full-finetune/Learning Rate": 8.39044798938305e-06, "Full-finetune/Loss": 0.7773828506469727, "Full-finetune/Loss (Raw)": 0.8584490418434143, "Full-finetune/Step": 2936, "Full-finetune/Step Time": 6.713928131386638} {"Full-finetune/Learning Rate": 8.384299914893645e-06, "Full-finetune/Loss": 0.7824428081512451, "Full-finetune/Loss (Raw)": 0.9984850883483887, "Full-finetune/Step": 2937, "Full-finetune/Step Time": 6.697281870990992} {"Full-finetune/Learning Rate": 8.378152467425902e-06, "Full-finetune/Loss": 0.7837123274803162, "Full-finetune/Loss (Raw)": 0.8506947159767151, "Full-finetune/Step": 2938, "Full-finetune/Step Time": 6.691397611051798} {"Full-finetune/Learning Rate": 8.37200564936552e-06, "Full-finetune/Loss": 0.7836993932723999, "Full-finetune/Loss (Raw)": 0.8339774012565613, "Full-finetune/Step": 2939, "Full-finetune/Step Time": 6.690077533945441} {"Full-finetune/Learning Rate": 8.365859463097963e-06, "Full-finetune/Loss": 0.7828526496887207, "Full-finetune/Loss (Raw)": 0.6699220538139343, "Full-finetune/Step": 2940, "Full-finetune/Step Time": 6.692086772993207} {"Full-finetune/Learning Rate": 8.359713911008447e-06, "Full-finetune/Loss": 0.7800713777542114, "Full-finetune/Loss (Raw)": 0.517509937286377, "Full-finetune/Step": 2941, "Full-finetune/Step Time": 6.709900166839361} {"Full-finetune/Learning Rate": 8.353568995481934e-06, "Full-finetune/Loss": 0.7847528457641602, "Full-finetune/Loss (Raw)": 0.9568220973014832, "Full-finetune/Step": 2942, "Full-finetune/Step Time": 6.6977983471006155} {"Full-finetune/Learning Rate": 8.347424718903152e-06, "Full-finetune/Loss": 0.7834398746490479, "Full-finetune/Loss (Raw)": 0.7514104247093201, "Full-finetune/Step": 2943, "Full-finetune/Step Time": 6.701501686125994} {"Full-finetune/Learning Rate": 8.341281083656571e-06, "Full-finetune/Loss": 0.7834441661834717, "Full-finetune/Loss (Raw)": 0.814645528793335, "Full-finetune/Step": 2944, "Full-finetune/Step Time": 6.7191398944705725} {"Full-finetune/Learning Rate": 8.33513809212642e-06, "Full-finetune/Loss": 0.7831364870071411, "Full-finetune/Loss (Raw)": 0.7480171918869019, "Full-finetune/Step": 2945, "Full-finetune/Step Time": 6.727666389197111} {"Full-finetune/Learning Rate": 8.328995746696674e-06, "Full-finetune/Loss": 0.7832025289535522, "Full-finetune/Loss (Raw)": 0.7954340577125549, "Full-finetune/Step": 2946, "Full-finetune/Step Time": 6.719516774639487} {"Full-finetune/Learning Rate": 8.32285404975105e-06, "Full-finetune/Loss": 0.7818547487258911, "Full-finetune/Loss (Raw)": 0.7647801041603088, "Full-finetune/Step": 2947, "Full-finetune/Step Time": 6.737144744023681} {"Full-finetune/Learning Rate": 8.316713003673028e-06, "Full-finetune/Loss": 0.7801480293273926, "Full-finetune/Loss (Raw)": 0.6951702833175659, "Full-finetune/Step": 2948, "Full-finetune/Step Time": 6.747740568593144} {"Full-finetune/Learning Rate": 8.310572610845824e-06, "Full-finetune/Loss": 0.781845211982727, "Full-finetune/Loss (Raw)": 0.8535587787628174, "Full-finetune/Step": 2949, "Full-finetune/Step Time": 6.7388491947203875} {"Full-finetune/Learning Rate": 8.304432873652412e-06, "Full-finetune/Loss": 0.7819047570228577, "Full-finetune/Loss (Raw)": 0.8459308743476868, "Full-finetune/Step": 2950, "Full-finetune/Step Time": 6.738502932712436} {"Full-finetune/Learning Rate": 8.298293794475495e-06, "Full-finetune/Loss": 0.7811582684516907, "Full-finetune/Loss (Raw)": 0.314890056848526, "Full-finetune/Step": 2951, "Full-finetune/Step Time": 6.738681595772505} {"Full-finetune/Learning Rate": 8.29215537569753e-06, "Full-finetune/Loss": 0.783100962638855, "Full-finetune/Loss (Raw)": 0.8539666533470154, "Full-finetune/Step": 2952, "Full-finetune/Step Time": 6.737489279359579} {"Full-finetune/Learning Rate": 8.286017619700725e-06, "Full-finetune/Loss": 0.7846787571907043, "Full-finetune/Loss (Raw)": 0.9061766862869263, "Full-finetune/Step": 2953, "Full-finetune/Step Time": 6.725159572437406} {"Full-finetune/Learning Rate": 8.279880528867017e-06, "Full-finetune/Loss": 0.7850710153579712, "Full-finetune/Loss (Raw)": 0.7236442565917969, "Full-finetune/Step": 2954, "Full-finetune/Step Time": 6.731545330956578} {"Full-finetune/Learning Rate": 8.273744105578097e-06, "Full-finetune/Loss": 0.7864438891410828, "Full-finetune/Loss (Raw)": 0.9940797686576843, "Full-finetune/Step": 2955, "Full-finetune/Step Time": 6.739401565864682} {"Full-finetune/Learning Rate": 8.267608352215385e-06, "Full-finetune/Loss": 0.7853526473045349, "Full-finetune/Loss (Raw)": 0.8092877864837646, "Full-finetune/Step": 2956, "Full-finetune/Step Time": 6.73567040450871} {"Full-finetune/Learning Rate": 8.261473271160046e-06, "Full-finetune/Loss": 0.7848187685012817, "Full-finetune/Loss (Raw)": 0.7713587880134583, "Full-finetune/Step": 2957, "Full-finetune/Step Time": 6.725626336410642} {"Full-finetune/Learning Rate": 8.255338864792993e-06, "Full-finetune/Loss": 0.784144401550293, "Full-finetune/Loss (Raw)": 0.8052698373794556, "Full-finetune/Step": 2958, "Full-finetune/Step Time": 6.729136602953076} {"Full-finetune/Learning Rate": 8.249205135494862e-06, "Full-finetune/Loss": 0.7787593007087708, "Full-finetune/Loss (Raw)": 0.3286368250846863, "Full-finetune/Step": 2959, "Full-finetune/Step Time": 6.743886983022094} {"Full-finetune/Learning Rate": 8.24307208564604e-06, "Full-finetune/Loss": 0.7771701216697693, "Full-finetune/Loss (Raw)": 0.7365589141845703, "Full-finetune/Step": 2960, "Full-finetune/Step Time": 6.733705842867494} {"Full-finetune/Learning Rate": 8.236939717626639e-06, "Full-finetune/Loss": 0.7768746614456177, "Full-finetune/Loss (Raw)": 0.7625868916511536, "Full-finetune/Step": 2961, "Full-finetune/Step Time": 6.718884751200676} {"Full-finetune/Learning Rate": 8.23080803381651e-06, "Full-finetune/Loss": 0.7819104194641113, "Full-finetune/Loss (Raw)": 1.0091464519500732, "Full-finetune/Step": 2962, "Full-finetune/Step Time": 6.710216272622347} {"Full-finetune/Learning Rate": 8.224677036595245e-06, "Full-finetune/Loss": 0.7820421457290649, "Full-finetune/Loss (Raw)": 0.8492490649223328, "Full-finetune/Step": 2963, "Full-finetune/Step Time": 6.719758417457342} {"Full-finetune/Learning Rate": 8.218546728342161e-06, "Full-finetune/Loss": 0.7839509844779968, "Full-finetune/Loss (Raw)": 0.8727738857269287, "Full-finetune/Step": 2964, "Full-finetune/Step Time": 6.722567664459348} {"Full-finetune/Learning Rate": 8.212417111436315e-06, "Full-finetune/Loss": 0.7843534350395203, "Full-finetune/Loss (Raw)": 0.8978963494300842, "Full-finetune/Step": 2965, "Full-finetune/Step Time": 6.71966446377337} {"Full-finetune/Learning Rate": 8.206288188256486e-06, "Full-finetune/Loss": 0.7845145463943481, "Full-finetune/Loss (Raw)": 0.7215738296508789, "Full-finetune/Step": 2966, "Full-finetune/Step Time": 6.727967815473676} {"Full-finetune/Learning Rate": 8.20015996118119e-06, "Full-finetune/Loss": 0.7843824625015259, "Full-finetune/Loss (Raw)": 0.6843542456626892, "Full-finetune/Step": 2967, "Full-finetune/Step Time": 6.74007817171514} {"Full-finetune/Learning Rate": 8.194032432588677e-06, "Full-finetune/Loss": 0.7848262786865234, "Full-finetune/Loss (Raw)": 0.8566769361495972, "Full-finetune/Step": 2968, "Full-finetune/Step Time": 6.748502820730209} {"Full-finetune/Learning Rate": 8.187905604856916e-06, "Full-finetune/Loss": 0.783722996711731, "Full-finetune/Loss (Raw)": 0.799919605255127, "Full-finetune/Step": 2969, "Full-finetune/Step Time": 6.7451951783150434} {"Full-finetune/Learning Rate": 8.181779480363614e-06, "Full-finetune/Loss": 0.7835129499435425, "Full-finetune/Loss (Raw)": 0.7925935387611389, "Full-finetune/Step": 2970, "Full-finetune/Step Time": 6.7571881711483} {"Full-finetune/Learning Rate": 8.175654061486197e-06, "Full-finetune/Loss": 0.7844811677932739, "Full-finetune/Loss (Raw)": 0.7985811829566956, "Full-finetune/Step": 2971, "Full-finetune/Step Time": 6.764342054724693} {"Full-finetune/Learning Rate": 8.16952935060182e-06, "Full-finetune/Loss": 0.7841312885284424, "Full-finetune/Loss (Raw)": 0.8046870231628418, "Full-finetune/Step": 2972, "Full-finetune/Step Time": 6.760415306314826} {"Full-finetune/Learning Rate": 8.163405350087366e-06, "Full-finetune/Loss": 0.7833386063575745, "Full-finetune/Loss (Raw)": 0.7107653617858887, "Full-finetune/Step": 2973, "Full-finetune/Step Time": 6.764279989525676} {"Full-finetune/Learning Rate": 8.157282062319437e-06, "Full-finetune/Loss": 0.7840371131896973, "Full-finetune/Loss (Raw)": 0.836553692817688, "Full-finetune/Step": 2974, "Full-finetune/Step Time": 6.766363749280572} {"Full-finetune/Learning Rate": 8.15115948967437e-06, "Full-finetune/Loss": 0.782476544380188, "Full-finetune/Loss (Raw)": 0.40072518587112427, "Full-finetune/Step": 2975, "Full-finetune/Step Time": 6.769802987575531} {"Full-finetune/Learning Rate": 8.145037634528205e-06, "Full-finetune/Loss": 0.781490683555603, "Full-finetune/Loss (Raw)": 0.7324036955833435, "Full-finetune/Step": 2976, "Full-finetune/Step Time": 6.765484359115362} {"Full-finetune/Learning Rate": 8.138916499256718e-06, "Full-finetune/Loss": 0.7816315293312073, "Full-finetune/Loss (Raw)": 0.7716419100761414, "Full-finetune/Step": 2977, "Full-finetune/Step Time": 6.762952694669366} {"Full-finetune/Learning Rate": 8.132796086235405e-06, "Full-finetune/Loss": 0.7825751304626465, "Full-finetune/Loss (Raw)": 0.9177120327949524, "Full-finetune/Step": 2978, "Full-finetune/Step Time": 6.764354510232806} {"Full-finetune/Learning Rate": 8.126676397839475e-06, "Full-finetune/Loss": 0.7824392914772034, "Full-finetune/Loss (Raw)": 0.8294452428817749, "Full-finetune/Step": 2979, "Full-finetune/Step Time": 6.783569356426597} {"Full-finetune/Learning Rate": 8.120557436443866e-06, "Full-finetune/Loss": 0.7818282842636108, "Full-finetune/Loss (Raw)": 0.7498135566711426, "Full-finetune/Step": 2980, "Full-finetune/Step Time": 6.785866003483534} {"Full-finetune/Learning Rate": 8.11443920442322e-06, "Full-finetune/Loss": 0.7815345525741577, "Full-finetune/Loss (Raw)": 0.7195392847061157, "Full-finetune/Step": 2981, "Full-finetune/Step Time": 6.811984710395336} {"Full-finetune/Learning Rate": 8.108321704151903e-06, "Full-finetune/Loss": 0.7806481122970581, "Full-finetune/Loss (Raw)": 0.7608922123908997, "Full-finetune/Step": 2982, "Full-finetune/Step Time": 6.805167818441987} {"Full-finetune/Learning Rate": 8.102204938004004e-06, "Full-finetune/Loss": 0.7819050550460815, "Full-finetune/Loss (Raw)": 0.9427780508995056, "Full-finetune/Step": 2983, "Full-finetune/Step Time": 6.795473709702492} {"Full-finetune/Learning Rate": 8.096088908353316e-06, "Full-finetune/Loss": 0.7819044589996338, "Full-finetune/Loss (Raw)": 0.768014669418335, "Full-finetune/Step": 2984, "Full-finetune/Step Time": 6.7949472945183516} {"Full-finetune/Learning Rate": 8.089973617573355e-06, "Full-finetune/Loss": 0.781853199005127, "Full-finetune/Loss (Raw)": 0.830418586730957, "Full-finetune/Step": 2985, "Full-finetune/Step Time": 6.793445475399494} {"Full-finetune/Learning Rate": 8.08385906803734e-06, "Full-finetune/Loss": 0.778901219367981, "Full-finetune/Loss (Raw)": 0.33678996562957764, "Full-finetune/Step": 2986, "Full-finetune/Step Time": 6.813177002593875} {"Full-finetune/Learning Rate": 8.07774526211821e-06, "Full-finetune/Loss": 0.778668999671936, "Full-finetune/Loss (Raw)": 0.7854591012001038, "Full-finetune/Step": 2987, "Full-finetune/Step Time": 6.827288961037993} {"Full-finetune/Learning Rate": 8.071632202188618e-06, "Full-finetune/Loss": 0.7801591753959656, "Full-finetune/Loss (Raw)": 0.8723698258399963, "Full-finetune/Step": 2988, "Full-finetune/Step Time": 6.825621835887432} {"Full-finetune/Learning Rate": 8.065519890620922e-06, "Full-finetune/Loss": 0.7775169014930725, "Full-finetune/Loss (Raw)": 0.6276124715805054, "Full-finetune/Step": 2989, "Full-finetune/Step Time": 6.818381929770112} {"Full-finetune/Learning Rate": 8.059408329787184e-06, "Full-finetune/Loss": 0.7787601947784424, "Full-finetune/Loss (Raw)": 0.927353024482727, "Full-finetune/Step": 2990, "Full-finetune/Step Time": 6.795083701610565} {"Full-finetune/Learning Rate": 8.053297522059189e-06, "Full-finetune/Loss": 0.7794104814529419, "Full-finetune/Loss (Raw)": 0.8666906952857971, "Full-finetune/Step": 2991, "Full-finetune/Step Time": 6.809765297919512} {"Full-finetune/Learning Rate": 8.047187469808418e-06, "Full-finetune/Loss": 0.7792297005653381, "Full-finetune/Loss (Raw)": 0.6861927509307861, "Full-finetune/Step": 2992, "Full-finetune/Step Time": 6.820576870813966} {"Full-finetune/Learning Rate": 8.041078175406064e-06, "Full-finetune/Loss": 0.7795437574386597, "Full-finetune/Loss (Raw)": 0.9172874689102173, "Full-finetune/Step": 2993, "Full-finetune/Step Time": 6.816668210551143} {"Full-finetune/Learning Rate": 8.03496964122303e-06, "Full-finetune/Loss": 0.7836548089981079, "Full-finetune/Loss (Raw)": 0.9223347902297974, "Full-finetune/Step": 2994, "Full-finetune/Step Time": 6.795191396027803} {"Full-finetune/Learning Rate": 8.02886186962991e-06, "Full-finetune/Loss": 0.7835811376571655, "Full-finetune/Loss (Raw)": 0.8319247961044312, "Full-finetune/Step": 2995, "Full-finetune/Step Time": 6.797027420252562} {"Full-finetune/Learning Rate": 8.022754862997016e-06, "Full-finetune/Loss": 0.7820794582366943, "Full-finetune/Loss (Raw)": 0.7384010553359985, "Full-finetune/Step": 2996, "Full-finetune/Step Time": 6.811844175681472} {"Full-finetune/Learning Rate": 8.016648623694354e-06, "Full-finetune/Loss": 0.7831251621246338, "Full-finetune/Loss (Raw)": 0.7384960055351257, "Full-finetune/Step": 2997, "Full-finetune/Step Time": 6.78182614967227} {"Full-finetune/Learning Rate": 8.010543154091644e-06, "Full-finetune/Loss": 0.7859318256378174, "Full-finetune/Loss (Raw)": 0.7374141812324524, "Full-finetune/Step": 2998, "Full-finetune/Step Time": 6.772376665845513} {"Full-finetune/Learning Rate": 8.004438456558297e-06, "Full-finetune/Loss": 0.7849473357200623, "Full-finetune/Loss (Raw)": 0.8206676244735718, "Full-finetune/Step": 2999, "Full-finetune/Step Time": 6.776818066835403} {"Full-finetune/Learning Rate": 7.99833453346342e-06, "Full-finetune/Loss": 0.7842656970024109, "Full-finetune/Loss (Raw)": 0.8430125713348389, "Full-finetune/Step": 3000, "Full-finetune/Step Time": 6.782175797969103} {"Full-finetune/Learning Rate": 7.992231387175832e-06, "Full-finetune/Loss": 0.7835696339607239, "Full-finetune/Loss (Raw)": 0.7527615427970886, "Full-finetune/Step": 3001, "Full-finetune/Step Time": 6.785676138475537} {"Full-finetune/Learning Rate": 7.986129020064044e-06, "Full-finetune/Loss": 0.7876412272453308, "Full-finetune/Loss (Raw)": 0.8853971362113953, "Full-finetune/Step": 3002, "Full-finetune/Step Time": 6.781668359413743} {"Full-finetune/Learning Rate": 7.980027434496267e-06, "Full-finetune/Loss": 0.7870538830757141, "Full-finetune/Loss (Raw)": 0.8039080500602722, "Full-finetune/Step": 3003, "Full-finetune/Step Time": 6.786095652729273} {"Full-finetune/Learning Rate": 7.97392663284041e-06, "Full-finetune/Loss": 0.7876351475715637, "Full-finetune/Loss (Raw)": 0.884429931640625, "Full-finetune/Step": 3004, "Full-finetune/Step Time": 6.777816636487842} {"Full-finetune/Learning Rate": 7.967826617464065e-06, "Full-finetune/Loss": 0.7860696315765381, "Full-finetune/Loss (Raw)": 0.5911905765533447, "Full-finetune/Step": 3005, "Full-finetune/Step Time": 6.7959524765610695} {"Full-finetune/Learning Rate": 7.961727390734538e-06, "Full-finetune/Loss": 0.7873395681381226, "Full-finetune/Loss (Raw)": 0.8843282461166382, "Full-finetune/Step": 3006, "Full-finetune/Step Time": 6.778199892491102} {"Full-finetune/Learning Rate": 7.955628955018813e-06, "Full-finetune/Loss": 0.7867809534072876, "Full-finetune/Loss (Raw)": 0.6539566516876221, "Full-finetune/Step": 3007, "Full-finetune/Step Time": 6.7695091515779495} {"Full-finetune/Learning Rate": 7.949531312683582e-06, "Full-finetune/Loss": 0.7864560484886169, "Full-finetune/Loss (Raw)": 0.8706153631210327, "Full-finetune/Step": 3008, "Full-finetune/Step Time": 6.770546108484268} {"Full-finetune/Learning Rate": 7.943434466095216e-06, "Full-finetune/Loss": 0.7872999906539917, "Full-finetune/Loss (Raw)": 0.9377099871635437, "Full-finetune/Step": 3009, "Full-finetune/Step Time": 6.763148734346032} {"Full-finetune/Learning Rate": 7.937338417619781e-06, "Full-finetune/Loss": 0.7864865064620972, "Full-finetune/Loss (Raw)": 0.7803537845611572, "Full-finetune/Step": 3010, "Full-finetune/Step Time": 6.7642746195197105} {"Full-finetune/Learning Rate": 7.931243169623037e-06, "Full-finetune/Loss": 0.7862527370452881, "Full-finetune/Loss (Raw)": 0.8899906873703003, "Full-finetune/Step": 3011, "Full-finetune/Step Time": 6.761752467602491} {"Full-finetune/Learning Rate": 7.925148724470428e-06, "Full-finetune/Loss": 0.7875706553459167, "Full-finetune/Loss (Raw)": 0.8642548322677612, "Full-finetune/Step": 3012, "Full-finetune/Step Time": 6.734699109569192} {"Full-finetune/Learning Rate": 7.919055084527094e-06, "Full-finetune/Loss": 0.787257969379425, "Full-finetune/Loss (Raw)": 0.8346646428108215, "Full-finetune/Step": 3013, "Full-finetune/Step Time": 6.731502028182149} {"Full-finetune/Learning Rate": 7.912962252157854e-06, "Full-finetune/Loss": 0.7881388664245605, "Full-finetune/Loss (Raw)": 0.6824271082878113, "Full-finetune/Step": 3014, "Full-finetune/Step Time": 6.715520352125168} {"Full-finetune/Learning Rate": 7.906870229727216e-06, "Full-finetune/Loss": 0.7875890731811523, "Full-finetune/Loss (Raw)": 0.6045950651168823, "Full-finetune/Step": 3015, "Full-finetune/Step Time": 6.725199108943343} {"Full-finetune/Learning Rate": 7.900779019599378e-06, "Full-finetune/Loss": 0.7858281135559082, "Full-finetune/Loss (Raw)": 0.7646881937980652, "Full-finetune/Step": 3016, "Full-finetune/Step Time": 6.728605410084128} {"Full-finetune/Learning Rate": 7.894688624138217e-06, "Full-finetune/Loss": 0.7886204719543457, "Full-finetune/Loss (Raw)": 0.6985505819320679, "Full-finetune/Step": 3017, "Full-finetune/Step Time": 6.716435380280018} {"Full-finetune/Learning Rate": 7.8885990457073e-06, "Full-finetune/Loss": 0.7884439826011658, "Full-finetune/Loss (Raw)": 0.7302234768867493, "Full-finetune/Step": 3018, "Full-finetune/Step Time": 6.699767867103219} {"Full-finetune/Learning Rate": 7.882510286669875e-06, "Full-finetune/Loss": 0.7886493802070618, "Full-finetune/Loss (Raw)": 0.7959737181663513, "Full-finetune/Step": 3019, "Full-finetune/Step Time": 6.701542526483536} {"Full-finetune/Learning Rate": 7.876422349388861e-06, "Full-finetune/Loss": 0.7869650721549988, "Full-finetune/Loss (Raw)": 0.7708072066307068, "Full-finetune/Step": 3020, "Full-finetune/Step Time": 6.683774350211024} {"Full-finetune/Learning Rate": 7.870335236226878e-06, "Full-finetune/Loss": 0.7870510816574097, "Full-finetune/Loss (Raw)": 0.7890699505805969, "Full-finetune/Step": 3021, "Full-finetune/Step Time": 6.6876070741564035} {"Full-finetune/Learning Rate": 7.864248949546209e-06, "Full-finetune/Loss": 0.7871420383453369, "Full-finetune/Loss (Raw)": 0.9138904809951782, "Full-finetune/Step": 3022, "Full-finetune/Step Time": 6.688747715204954} {"Full-finetune/Learning Rate": 7.858163491708827e-06, "Full-finetune/Loss": 0.7858700752258301, "Full-finetune/Loss (Raw)": 0.6655230522155762, "Full-finetune/Step": 3023, "Full-finetune/Step Time": 6.692938299849629} {"Full-finetune/Learning Rate": 7.85207886507638e-06, "Full-finetune/Loss": 0.7854057550430298, "Full-finetune/Loss (Raw)": 0.8111470341682434, "Full-finetune/Step": 3024, "Full-finetune/Step Time": 6.697249257937074} {"Full-finetune/Learning Rate": 7.845995072010188e-06, "Full-finetune/Loss": 0.7853783965110779, "Full-finetune/Loss (Raw)": 0.8650211691856384, "Full-finetune/Step": 3025, "Full-finetune/Step Time": 6.697661666199565} {"Full-finetune/Learning Rate": 7.839912114871256e-06, "Full-finetune/Loss": 0.7852559685707092, "Full-finetune/Loss (Raw)": 0.8066631555557251, "Full-finetune/Step": 3026, "Full-finetune/Step Time": 6.702304719015956} {"Full-finetune/Learning Rate": 7.83382999602026e-06, "Full-finetune/Loss": 0.7860747575759888, "Full-finetune/Loss (Raw)": 0.9046570062637329, "Full-finetune/Step": 3027, "Full-finetune/Step Time": 6.702725602313876} {"Full-finetune/Learning Rate": 7.827748717817552e-06, "Full-finetune/Loss": 0.7871820330619812, "Full-finetune/Loss (Raw)": 0.8094335198402405, "Full-finetune/Step": 3028, "Full-finetune/Step Time": 6.705821949988604} {"Full-finetune/Learning Rate": 7.821668282623159e-06, "Full-finetune/Loss": 0.7898445129394531, "Full-finetune/Loss (Raw)": 1.0092802047729492, "Full-finetune/Step": 3029, "Full-finetune/Step Time": 6.69730612821877} {"Full-finetune/Learning Rate": 7.815588692796771e-06, "Full-finetune/Loss": 0.7897546887397766, "Full-finetune/Loss (Raw)": 0.7744889259338379, "Full-finetune/Step": 3030, "Full-finetune/Step Time": 6.690176609903574} {"Full-finetune/Learning Rate": 7.809509950697769e-06, "Full-finetune/Loss": 0.7898298501968384, "Full-finetune/Loss (Raw)": 0.8410516977310181, "Full-finetune/Step": 3031, "Full-finetune/Step Time": 6.687738044187427} {"Full-finetune/Learning Rate": 7.803432058685187e-06, "Full-finetune/Loss": 0.7900247573852539, "Full-finetune/Loss (Raw)": 0.8158144354820251, "Full-finetune/Step": 3032, "Full-finetune/Step Time": 6.6933761313557625} {"Full-finetune/Learning Rate": 7.797355019117743e-06, "Full-finetune/Loss": 0.7903996109962463, "Full-finetune/Loss (Raw)": 0.855681836605072, "Full-finetune/Step": 3033, "Full-finetune/Step Time": 6.691319962963462} {"Full-finetune/Learning Rate": 7.79127883435381e-06, "Full-finetune/Loss": 0.7907529473304749, "Full-finetune/Loss (Raw)": 0.9283953905105591, "Full-finetune/Step": 3034, "Full-finetune/Step Time": 6.698138322681189} {"Full-finetune/Learning Rate": 7.78520350675144e-06, "Full-finetune/Loss": 0.7909884452819824, "Full-finetune/Loss (Raw)": 0.8275582790374756, "Full-finetune/Step": 3035, "Full-finetune/Step Time": 6.701148599386215} {"Full-finetune/Learning Rate": 7.779129038668347e-06, "Full-finetune/Loss": 0.7902989387512207, "Full-finetune/Loss (Raw)": 0.7802661061286926, "Full-finetune/Step": 3036, "Full-finetune/Step Time": 6.7144420221447945} {"Full-finetune/Learning Rate": 7.773055432461916e-06, "Full-finetune/Loss": 0.7898907661437988, "Full-finetune/Loss (Raw)": 0.9418484568595886, "Full-finetune/Step": 3037, "Full-finetune/Step Time": 6.71826926805079} {"Full-finetune/Learning Rate": 7.766982690489198e-06, "Full-finetune/Loss": 0.7896057367324829, "Full-finetune/Loss (Raw)": 0.8822425007820129, "Full-finetune/Step": 3038, "Full-finetune/Step Time": 6.716991543769836} {"Full-finetune/Learning Rate": 7.7609108151069e-06, "Full-finetune/Loss": 0.7878739237785339, "Full-finetune/Loss (Raw)": 0.7292494177818298, "Full-finetune/Step": 3039, "Full-finetune/Step Time": 6.718753229826689} {"Full-finetune/Learning Rate": 7.754839808671397e-06, "Full-finetune/Loss": 0.7878648042678833, "Full-finetune/Loss (Raw)": 0.8370927572250366, "Full-finetune/Step": 3040, "Full-finetune/Step Time": 6.711819134652615} {"Full-finetune/Learning Rate": 7.748769673538734e-06, "Full-finetune/Loss": 0.7870808243751526, "Full-finetune/Loss (Raw)": 0.7360556721687317, "Full-finetune/Step": 3041, "Full-finetune/Step Time": 6.7186075411736965} {"Full-finetune/Learning Rate": 7.742700412064608e-06, "Full-finetune/Loss": 0.7887081503868103, "Full-finetune/Loss (Raw)": 0.8162078261375427, "Full-finetune/Step": 3042, "Full-finetune/Step Time": 6.695306170731783} {"Full-finetune/Learning Rate": 7.736632026604386e-06, "Full-finetune/Loss": 0.7867924571037292, "Full-finetune/Loss (Raw)": 0.3153946101665497, "Full-finetune/Step": 3043, "Full-finetune/Step Time": 6.704064320772886} {"Full-finetune/Learning Rate": 7.730564519513082e-06, "Full-finetune/Loss": 0.7852740287780762, "Full-finetune/Loss (Raw)": 0.6684547066688538, "Full-finetune/Step": 3044, "Full-finetune/Step Time": 6.724768904969096} {"Full-finetune/Learning Rate": 7.724497893145379e-06, "Full-finetune/Loss": 0.7825378775596619, "Full-finetune/Loss (Raw)": 0.5695199966430664, "Full-finetune/Step": 3045, "Full-finetune/Step Time": 6.731751423329115} {"Full-finetune/Learning Rate": 7.718432149855622e-06, "Full-finetune/Loss": 0.7847071886062622, "Full-finetune/Loss (Raw)": 0.9880269765853882, "Full-finetune/Step": 3046, "Full-finetune/Step Time": 6.716379899531603} {"Full-finetune/Learning Rate": 7.7123672919978e-06, "Full-finetune/Loss": 0.7845954298973083, "Full-finetune/Loss (Raw)": 0.8203189969062805, "Full-finetune/Step": 3047, "Full-finetune/Step Time": 6.723229918628931} {"Full-finetune/Learning Rate": 7.706303321925577e-06, "Full-finetune/Loss": 0.789289116859436, "Full-finetune/Loss (Raw)": 0.953809916973114, "Full-finetune/Step": 3048, "Full-finetune/Step Time": 6.705701354891062} {"Full-finetune/Learning Rate": 7.700240241992247e-06, "Full-finetune/Loss": 0.7893201112747192, "Full-finetune/Loss (Raw)": 0.7797676920890808, "Full-finetune/Step": 3049, "Full-finetune/Step Time": 6.6889674216508865} {"Full-finetune/Learning Rate": 7.694178054550782e-06, "Full-finetune/Loss": 0.7908820509910583, "Full-finetune/Loss (Raw)": 0.9444459676742554, "Full-finetune/Step": 3050, "Full-finetune/Step Time": 6.691149173304439} {"Full-finetune/Learning Rate": 7.688116761953797e-06, "Full-finetune/Loss": 0.7925719022750854, "Full-finetune/Loss (Raw)": 0.9217116832733154, "Full-finetune/Step": 3051, "Full-finetune/Step Time": 6.6712088864296675} {"Full-finetune/Learning Rate": 7.682056366553559e-06, "Full-finetune/Loss": 0.7917492389678955, "Full-finetune/Loss (Raw)": 0.7761610746383667, "Full-finetune/Step": 3052, "Full-finetune/Step Time": 6.674709986895323} {"Full-finetune/Learning Rate": 7.675996870701999e-06, "Full-finetune/Loss": 0.79173344373703, "Full-finetune/Loss (Raw)": 0.9008514285087585, "Full-finetune/Step": 3053, "Full-finetune/Step Time": 6.687461396679282} {"Full-finetune/Learning Rate": 7.669938276750673e-06, "Full-finetune/Loss": 0.793063759803772, "Full-finetune/Loss (Raw)": 0.9699265360832214, "Full-finetune/Step": 3054, "Full-finetune/Step Time": 6.662942569702864} {"Full-finetune/Learning Rate": 7.663880587050812e-06, "Full-finetune/Loss": 0.7952067255973816, "Full-finetune/Loss (Raw)": 0.8168361186981201, "Full-finetune/Step": 3055, "Full-finetune/Step Time": 6.654992043972015} {"Full-finetune/Learning Rate": 7.657823803953288e-06, "Full-finetune/Loss": 0.7957085967063904, "Full-finetune/Loss (Raw)": 0.8562361598014832, "Full-finetune/Step": 3056, "Full-finetune/Step Time": 6.633949929848313} {"Full-finetune/Learning Rate": 7.651767929808615e-06, "Full-finetune/Loss": 0.7946265339851379, "Full-finetune/Loss (Raw)": 0.7100638151168823, "Full-finetune/Step": 3057, "Full-finetune/Step Time": 6.645843235775828} {"Full-finetune/Learning Rate": 7.645712966966973e-06, "Full-finetune/Loss": 0.7944875359535217, "Full-finetune/Loss (Raw)": 0.856749415397644, "Full-finetune/Step": 3058, "Full-finetune/Step Time": 6.643260706216097} {"Full-finetune/Learning Rate": 7.639658917778156e-06, "Full-finetune/Loss": 0.7943451404571533, "Full-finetune/Loss (Raw)": 0.8580033779144287, "Full-finetune/Step": 3059, "Full-finetune/Step Time": 6.639348039403558} {"Full-finetune/Learning Rate": 7.633605784591633e-06, "Full-finetune/Loss": 0.7945986390113831, "Full-finetune/Loss (Raw)": 0.9034727811813354, "Full-finetune/Step": 3060, "Full-finetune/Step Time": 6.644863339141011} {"Full-finetune/Learning Rate": 7.627553569756505e-06, "Full-finetune/Loss": 0.7952920198440552, "Full-finetune/Loss (Raw)": 0.9029080867767334, "Full-finetune/Step": 3061, "Full-finetune/Step Time": 6.6364349238574505} {"Full-finetune/Learning Rate": 7.621502275621518e-06, "Full-finetune/Loss": 0.7948265075683594, "Full-finetune/Loss (Raw)": 0.7277314066886902, "Full-finetune/Step": 3062, "Full-finetune/Step Time": 6.654512289911509} {"Full-finetune/Learning Rate": 7.615451904535066e-06, "Full-finetune/Loss": 0.7947596311569214, "Full-finetune/Loss (Raw)": 0.7959514856338501, "Full-finetune/Step": 3063, "Full-finetune/Step Time": 6.645882476121187} {"Full-finetune/Learning Rate": 7.609402458845171e-06, "Full-finetune/Loss": 0.7947235107421875, "Full-finetune/Loss (Raw)": 0.8538284301757812, "Full-finetune/Step": 3064, "Full-finetune/Step Time": 6.629070673137903} {"Full-finetune/Learning Rate": 7.60335394089951e-06, "Full-finetune/Loss": 0.7918730974197388, "Full-finetune/Loss (Raw)": 0.6336286067962646, "Full-finetune/Step": 3065, "Full-finetune/Step Time": 6.645449420437217} {"Full-finetune/Learning Rate": 7.597306353045393e-06, "Full-finetune/Loss": 0.7893839478492737, "Full-finetune/Loss (Raw)": 0.5320808291435242, "Full-finetune/Step": 3066, "Full-finetune/Step Time": 6.678818020969629} {"Full-finetune/Learning Rate": 7.5912596976297735e-06, "Full-finetune/Loss": 0.7893007397651672, "Full-finetune/Loss (Raw)": 0.8233250379562378, "Full-finetune/Step": 3067, "Full-finetune/Step Time": 6.67959145270288} {"Full-finetune/Learning Rate": 7.58521397699924e-06, "Full-finetune/Loss": 0.7899150848388672, "Full-finetune/Loss (Raw)": 0.7485650181770325, "Full-finetune/Step": 3068, "Full-finetune/Step Time": 6.691598115488887} {"Full-finetune/Learning Rate": 7.579169193500014e-06, "Full-finetune/Loss": 0.7911641001701355, "Full-finetune/Loss (Raw)": 0.6773799061775208, "Full-finetune/Step": 3069, "Full-finetune/Step Time": 6.683004532009363} {"Full-finetune/Learning Rate": 7.573125349477964e-06, "Full-finetune/Loss": 0.7900575399398804, "Full-finetune/Loss (Raw)": 0.8151823282241821, "Full-finetune/Step": 3070, "Full-finetune/Step Time": 6.681176032871008} {"Full-finetune/Learning Rate": 7.567082447278582e-06, "Full-finetune/Loss": 0.7917768955230713, "Full-finetune/Loss (Raw)": 0.9714871048927307, "Full-finetune/Step": 3071, "Full-finetune/Step Time": 6.672941243276} {"Full-finetune/Learning Rate": 7.561040489247008e-06, "Full-finetune/Loss": 0.7916005849838257, "Full-finetune/Loss (Raw)": 0.7920814752578735, "Full-finetune/Step": 3072, "Full-finetune/Step Time": 6.664866715669632} {"Full-finetune/Learning Rate": 7.554999477728001e-06, "Full-finetune/Loss": 0.7921651601791382, "Full-finetune/Loss (Raw)": 0.8202842473983765, "Full-finetune/Step": 3073, "Full-finetune/Step Time": 6.655134974047542} {"Full-finetune/Learning Rate": 7.548959415065961e-06, "Full-finetune/Loss": 0.7921826243400574, "Full-finetune/Loss (Raw)": 0.7976625561714172, "Full-finetune/Step": 3074, "Full-finetune/Step Time": 6.657264253124595} {"Full-finetune/Learning Rate": 7.54292030360492e-06, "Full-finetune/Loss": 0.7885643243789673, "Full-finetune/Loss (Raw)": 0.3016433119773865, "Full-finetune/Step": 3075, "Full-finetune/Step Time": 6.6590287797153} {"Full-finetune/Learning Rate": 7.5368821456885375e-06, "Full-finetune/Loss": 0.7884818911552429, "Full-finetune/Loss (Raw)": 0.6846169233322144, "Full-finetune/Step": 3076, "Full-finetune/Step Time": 6.647328736260533} {"Full-finetune/Learning Rate": 7.530844943660111e-06, "Full-finetune/Loss": 0.7884730100631714, "Full-finetune/Loss (Raw)": 0.8524203896522522, "Full-finetune/Step": 3077, "Full-finetune/Step Time": 6.647888882085681} {"Full-finetune/Learning Rate": 7.524808699862552e-06, "Full-finetune/Loss": 0.7872856259346008, "Full-finetune/Loss (Raw)": 0.6939480304718018, "Full-finetune/Step": 3078, "Full-finetune/Step Time": 6.644967684522271} {"Full-finetune/Learning Rate": 7.518773416638412e-06, "Full-finetune/Loss": 0.789677619934082, "Full-finetune/Loss (Raw)": 0.6210634708404541, "Full-finetune/Step": 3079, "Full-finetune/Step Time": 6.650726418942213} {"Full-finetune/Learning Rate": 7.512739096329868e-06, "Full-finetune/Loss": 0.7884558439254761, "Full-finetune/Loss (Raw)": 0.6975847482681274, "Full-finetune/Step": 3080, "Full-finetune/Step Time": 6.653728079050779} {"Full-finetune/Learning Rate": 7.50670574127872e-06, "Full-finetune/Loss": 0.7870818972587585, "Full-finetune/Loss (Raw)": 0.7303088903427124, "Full-finetune/Step": 3081, "Full-finetune/Step Time": 6.658194011077285} {"Full-finetune/Learning Rate": 7.500673353826399e-06, "Full-finetune/Loss": 0.7878566980361938, "Full-finetune/Loss (Raw)": 0.822814404964447, "Full-finetune/Step": 3082, "Full-finetune/Step Time": 6.652687327936292} {"Full-finetune/Learning Rate": 7.4946419363139535e-06, "Full-finetune/Loss": 0.7867381572723389, "Full-finetune/Loss (Raw)": 0.8509073257446289, "Full-finetune/Step": 3083, "Full-finetune/Step Time": 6.658204657956958} {"Full-finetune/Learning Rate": 7.488611491082056e-06, "Full-finetune/Loss": 0.786383330821991, "Full-finetune/Loss (Raw)": 0.763871431350708, "Full-finetune/Step": 3084, "Full-finetune/Step Time": 6.670710291713476} {"Full-finetune/Learning Rate": 7.482582020471012e-06, "Full-finetune/Loss": 0.7880473136901855, "Full-finetune/Loss (Raw)": 0.9843502640724182, "Full-finetune/Step": 3085, "Full-finetune/Step Time": 6.66763518191874} {"Full-finetune/Learning Rate": 7.476553526820733e-06, "Full-finetune/Loss": 0.7882716655731201, "Full-finetune/Loss (Raw)": 0.8339816927909851, "Full-finetune/Step": 3086, "Full-finetune/Step Time": 6.667219625785947} {"Full-finetune/Learning Rate": 7.4705260124707705e-06, "Full-finetune/Loss": 0.7919231653213501, "Full-finetune/Loss (Raw)": 0.7960333824157715, "Full-finetune/Step": 3087, "Full-finetune/Step Time": 6.6544507425278425} {"Full-finetune/Learning Rate": 7.464499479760275e-06, "Full-finetune/Loss": 0.7904897332191467, "Full-finetune/Loss (Raw)": 0.5530803799629211, "Full-finetune/Step": 3088, "Full-finetune/Step Time": 6.650833696126938} {"Full-finetune/Learning Rate": 7.458473931028029e-06, "Full-finetune/Loss": 0.7915743589401245, "Full-finetune/Loss (Raw)": 0.9014233350753784, "Full-finetune/Step": 3089, "Full-finetune/Step Time": 6.655304277315736} {"Full-finetune/Learning Rate": 7.452449368612433e-06, "Full-finetune/Loss": 0.7894039154052734, "Full-finetune/Loss (Raw)": 0.7313256859779358, "Full-finetune/Step": 3090, "Full-finetune/Step Time": 6.647240981459618} {"Full-finetune/Learning Rate": 7.446425794851498e-06, "Full-finetune/Loss": 0.7883797883987427, "Full-finetune/Loss (Raw)": 0.7181593775749207, "Full-finetune/Step": 3091, "Full-finetune/Step Time": 6.6516021359711885} {"Full-finetune/Learning Rate": 7.440403212082863e-06, "Full-finetune/Loss": 0.7839359641075134, "Full-finetune/Loss (Raw)": 0.30396515130996704, "Full-finetune/Step": 3092, "Full-finetune/Step Time": 6.666835891082883} {"Full-finetune/Learning Rate": 7.434381622643766e-06, "Full-finetune/Loss": 0.7833806276321411, "Full-finetune/Loss (Raw)": 0.8268123269081116, "Full-finetune/Step": 3093, "Full-finetune/Step Time": 6.674990413710475} {"Full-finetune/Learning Rate": 7.42836102887107e-06, "Full-finetune/Loss": 0.7844409942626953, "Full-finetune/Loss (Raw)": 0.8573004603385925, "Full-finetune/Step": 3094, "Full-finetune/Step Time": 6.669912768527865} {"Full-finetune/Learning Rate": 7.422341433101254e-06, "Full-finetune/Loss": 0.7851806879043579, "Full-finetune/Loss (Raw)": 0.7790355682373047, "Full-finetune/Step": 3095, "Full-finetune/Step Time": 6.651960192248225} {"Full-finetune/Learning Rate": 7.4163228376704e-06, "Full-finetune/Loss": 0.7849252820014954, "Full-finetune/Loss (Raw)": 0.8239883184432983, "Full-finetune/Step": 3096, "Full-finetune/Step Time": 6.657153896987438} {"Full-finetune/Learning Rate": 7.410305244914216e-06, "Full-finetune/Loss": 0.7811273336410522, "Full-finetune/Loss (Raw)": 0.31377893686294556, "Full-finetune/Step": 3097, "Full-finetune/Step Time": 6.676682388409972} {"Full-finetune/Learning Rate": 7.404288657168006e-06, "Full-finetune/Loss": 0.77940833568573, "Full-finetune/Loss (Raw)": 0.5725613832473755, "Full-finetune/Step": 3098, "Full-finetune/Step Time": 6.666312135756016} {"Full-finetune/Learning Rate": 7.398273076766687e-06, "Full-finetune/Loss": 0.7797417640686035, "Full-finetune/Loss (Raw)": 0.8412575721740723, "Full-finetune/Step": 3099, "Full-finetune/Step Time": 6.669704116880894} {"Full-finetune/Learning Rate": 7.392258506044796e-06, "Full-finetune/Loss": 0.781347393989563, "Full-finetune/Loss (Raw)": 1.0102094411849976, "Full-finetune/Step": 3100, "Full-finetune/Step Time": 6.671425469219685} {"Full-finetune/Learning Rate": 7.386244947336463e-06, "Full-finetune/Loss": 0.7801427841186523, "Full-finetune/Loss (Raw)": 0.5565769672393799, "Full-finetune/Step": 3101, "Full-finetune/Step Time": 6.673283876851201} {"Full-finetune/Learning Rate": 7.380232402975443e-06, "Full-finetune/Loss": 0.780644953250885, "Full-finetune/Loss (Raw)": 0.9008309245109558, "Full-finetune/Step": 3102, "Full-finetune/Step Time": 6.68073064647615} {"Full-finetune/Learning Rate": 7.3742208752950776e-06, "Full-finetune/Loss": 0.7847815155982971, "Full-finetune/Loss (Raw)": 0.9302065372467041, "Full-finetune/Step": 3103, "Full-finetune/Step Time": 6.675460884347558} {"Full-finetune/Learning Rate": 7.3682103666283245e-06, "Full-finetune/Loss": 0.7857986688613892, "Full-finetune/Loss (Raw)": 0.8625994920730591, "Full-finetune/Step": 3104, "Full-finetune/Step Time": 6.668991420418024} {"Full-finetune/Learning Rate": 7.362200879307748e-06, "Full-finetune/Loss": 0.7864382863044739, "Full-finetune/Loss (Raw)": 0.8535138368606567, "Full-finetune/Step": 3105, "Full-finetune/Step Time": 6.659341674298048} {"Full-finetune/Learning Rate": 7.3561924156655105e-06, "Full-finetune/Loss": 0.7848914861679077, "Full-finetune/Loss (Raw)": 0.7197208404541016, "Full-finetune/Step": 3106, "Full-finetune/Step Time": 6.669523807242513} {"Full-finetune/Learning Rate": 7.350184978033386e-06, "Full-finetune/Loss": 0.785929262638092, "Full-finetune/Loss (Raw)": 0.9622791409492493, "Full-finetune/Step": 3107, "Full-finetune/Step Time": 6.653686719015241} {"Full-finetune/Learning Rate": 7.344178568742734e-06, "Full-finetune/Loss": 0.7859441041946411, "Full-finetune/Loss (Raw)": 0.7517094016075134, "Full-finetune/Step": 3108, "Full-finetune/Step Time": 6.657291302457452} {"Full-finetune/Learning Rate": 7.338173190124526e-06, "Full-finetune/Loss": 0.7862420082092285, "Full-finetune/Loss (Raw)": 0.757673978805542, "Full-finetune/Step": 3109, "Full-finetune/Step Time": 6.632739832624793} {"Full-finetune/Learning Rate": 7.3321688445093395e-06, "Full-finetune/Loss": 0.7860019207000732, "Full-finetune/Loss (Raw)": 0.7301614880561829, "Full-finetune/Step": 3110, "Full-finetune/Step Time": 6.6292138528078794} {"Full-finetune/Learning Rate": 7.326165534227341e-06, "Full-finetune/Loss": 0.7846073508262634, "Full-finetune/Loss (Raw)": 0.7642728686332703, "Full-finetune/Step": 3111, "Full-finetune/Step Time": 6.63862500153482} {"Full-finetune/Learning Rate": 7.320163261608293e-06, "Full-finetune/Loss": 0.7855209708213806, "Full-finetune/Loss (Raw)": 0.8849574327468872, "Full-finetune/Step": 3112, "Full-finetune/Step Time": 6.623354895040393} {"Full-finetune/Learning Rate": 7.314162028981564e-06, "Full-finetune/Loss": 0.7851614952087402, "Full-finetune/Loss (Raw)": 0.7844104766845703, "Full-finetune/Step": 3113, "Full-finetune/Step Time": 6.636491613462567} {"Full-finetune/Learning Rate": 7.308161838676113e-06, "Full-finetune/Loss": 0.7893019318580627, "Full-finetune/Loss (Raw)": 0.8667664527893066, "Full-finetune/Step": 3114, "Full-finetune/Step Time": 6.62420280277729} {"Full-finetune/Learning Rate": 7.302162693020503e-06, "Full-finetune/Loss": 0.7892530560493469, "Full-finetune/Loss (Raw)": 0.7791994214057922, "Full-finetune/Step": 3115, "Full-finetune/Step Time": 6.605008324608207} {"Full-finetune/Learning Rate": 7.2961645943428804e-06, "Full-finetune/Loss": 0.7868791818618774, "Full-finetune/Loss (Raw)": 0.5685086846351624, "Full-finetune/Step": 3116, "Full-finetune/Step Time": 6.630813643336296} {"Full-finetune/Learning Rate": 7.290167544970989e-06, "Full-finetune/Loss": 0.788648784160614, "Full-finetune/Loss (Raw)": 0.8541297316551208, "Full-finetune/Step": 3117, "Full-finetune/Step Time": 6.626406027004123} {"Full-finetune/Learning Rate": 7.284171547232168e-06, "Full-finetune/Loss": 0.7851698994636536, "Full-finetune/Loss (Raw)": 0.48205286264419556, "Full-finetune/Step": 3118, "Full-finetune/Step Time": 6.646065089851618} {"Full-finetune/Learning Rate": 7.278176603453347e-06, "Full-finetune/Loss": 0.7847365736961365, "Full-finetune/Loss (Raw)": 0.8112232089042664, "Full-finetune/Step": 3119, "Full-finetune/Step Time": 6.637456443160772} {"Full-finetune/Learning Rate": 7.272182715961049e-06, "Full-finetune/Loss": 0.7859344482421875, "Full-finetune/Loss (Raw)": 0.8395182490348816, "Full-finetune/Step": 3120, "Full-finetune/Step Time": 6.631743686273694} {"Full-finetune/Learning Rate": 7.266189887081385e-06, "Full-finetune/Loss": 0.7835267782211304, "Full-finetune/Loss (Raw)": 0.609113335609436, "Full-finetune/Step": 3121, "Full-finetune/Step Time": 6.642730353400111} {"Full-finetune/Learning Rate": 7.2601981191400484e-06, "Full-finetune/Loss": 0.7823169827461243, "Full-finetune/Loss (Raw)": 0.7674792408943176, "Full-finetune/Step": 3122, "Full-finetune/Step Time": 6.644669007509947} {"Full-finetune/Learning Rate": 7.254207414462333e-06, "Full-finetune/Loss": 0.7819764614105225, "Full-finetune/Loss (Raw)": 0.7883354425430298, "Full-finetune/Step": 3123, "Full-finetune/Step Time": 6.640046006068587} {"Full-finetune/Learning Rate": 7.248217775373112e-06, "Full-finetune/Loss": 0.7829529643058777, "Full-finetune/Loss (Raw)": 0.8633938431739807, "Full-finetune/Step": 3124, "Full-finetune/Step Time": 6.623836375772953} {"Full-finetune/Learning Rate": 7.242229204196849e-06, "Full-finetune/Loss": 0.7822816371917725, "Full-finetune/Loss (Raw)": 0.6525728106498718, "Full-finetune/Step": 3125, "Full-finetune/Step Time": 6.623944139108062} {"Full-finetune/Learning Rate": 7.236241703257594e-06, "Full-finetune/Loss": 0.783282995223999, "Full-finetune/Loss (Raw)": 0.8655770421028137, "Full-finetune/Step": 3126, "Full-finetune/Step Time": 6.615060847252607} {"Full-finetune/Learning Rate": 7.230255274878972e-06, "Full-finetune/Loss": 0.7842846512794495, "Full-finetune/Loss (Raw)": 0.9488826394081116, "Full-finetune/Step": 3127, "Full-finetune/Step Time": 6.612211776897311} {"Full-finetune/Learning Rate": 7.224269921384206e-06, "Full-finetune/Loss": 0.7816982865333557, "Full-finetune/Loss (Raw)": 0.5119560360908508, "Full-finetune/Step": 3128, "Full-finetune/Step Time": 6.647412059828639} {"Full-finetune/Learning Rate": 7.218285645096088e-06, "Full-finetune/Loss": 0.7818107008934021, "Full-finetune/Loss (Raw)": 0.7671557664871216, "Full-finetune/Step": 3129, "Full-finetune/Step Time": 6.643748465925455} {"Full-finetune/Learning Rate": 7.2123024483370056e-06, "Full-finetune/Loss": 0.7805068492889404, "Full-finetune/Loss (Raw)": 0.7185033559799194, "Full-finetune/Step": 3130, "Full-finetune/Step Time": 6.647082641720772} {"Full-finetune/Learning Rate": 7.20632033342892e-06, "Full-finetune/Loss": 0.7800531387329102, "Full-finetune/Loss (Raw)": 0.7458277940750122, "Full-finetune/Step": 3131, "Full-finetune/Step Time": 6.650151047855616} {"Full-finetune/Learning Rate": 7.200339302693367e-06, "Full-finetune/Loss": 0.7763874530792236, "Full-finetune/Loss (Raw)": 0.41522452235221863, "Full-finetune/Step": 3132, "Full-finetune/Step Time": 6.667605614289641} {"Full-finetune/Learning Rate": 7.194359358451475e-06, "Full-finetune/Loss": 0.7781820893287659, "Full-finetune/Loss (Raw)": 0.8209081292152405, "Full-finetune/Step": 3133, "Full-finetune/Step Time": 6.650115337222815} {"Full-finetune/Learning Rate": 7.188380503023936e-06, "Full-finetune/Loss": 0.776511549949646, "Full-finetune/Loss (Raw)": 0.6704961657524109, "Full-finetune/Step": 3134, "Full-finetune/Step Time": 6.66477451659739} {"Full-finetune/Learning Rate": 7.182402738731036e-06, "Full-finetune/Loss": 0.7781845331192017, "Full-finetune/Loss (Raw)": 0.8680986166000366, "Full-finetune/Step": 3135, "Full-finetune/Step Time": 6.66822168789804} {"Full-finetune/Learning Rate": 7.176426067892623e-06, "Full-finetune/Loss": 0.7777970433235168, "Full-finetune/Loss (Raw)": 0.821014940738678, "Full-finetune/Step": 3136, "Full-finetune/Step Time": 6.669592656195164} {"Full-finetune/Learning Rate": 7.170450492828126e-06, "Full-finetune/Loss": 0.7760388851165771, "Full-finetune/Loss (Raw)": 0.712670087814331, "Full-finetune/Step": 3137, "Full-finetune/Step Time": 6.686070127412677} {"Full-finetune/Learning Rate": 7.164476015856552e-06, "Full-finetune/Loss": 0.7750325202941895, "Full-finetune/Loss (Raw)": 0.6515418291091919, "Full-finetune/Step": 3138, "Full-finetune/Step Time": 6.693729482591152} {"Full-finetune/Learning Rate": 7.158502639296475e-06, "Full-finetune/Loss": 0.7742941975593567, "Full-finetune/Loss (Raw)": 0.7954815626144409, "Full-finetune/Step": 3139, "Full-finetune/Step Time": 6.694315703585744} {"Full-finetune/Learning Rate": 7.152530365466051e-06, "Full-finetune/Loss": 0.7736052870750427, "Full-finetune/Loss (Raw)": 0.7760729789733887, "Full-finetune/Step": 3140, "Full-finetune/Step Time": 6.706064531579614} {"Full-finetune/Learning Rate": 7.146559196683001e-06, "Full-finetune/Loss": 0.7732251286506653, "Full-finetune/Loss (Raw)": 0.7860055565834045, "Full-finetune/Step": 3141, "Full-finetune/Step Time": 6.707203157246113} {"Full-finetune/Learning Rate": 7.140589135264614e-06, "Full-finetune/Loss": 0.7736470103263855, "Full-finetune/Loss (Raw)": 0.7364261746406555, "Full-finetune/Step": 3142, "Full-finetune/Step Time": 6.7116327583789825} {"Full-finetune/Learning Rate": 7.13462018352776e-06, "Full-finetune/Loss": 0.772993803024292, "Full-finetune/Loss (Raw)": 0.5209845304489136, "Full-finetune/Step": 3143, "Full-finetune/Step Time": 6.710669092833996} {"Full-finetune/Learning Rate": 7.12865234378887e-06, "Full-finetune/Loss": 0.7742574214935303, "Full-finetune/Loss (Raw)": 0.9264302253723145, "Full-finetune/Step": 3144, "Full-finetune/Step Time": 6.702485861256719} {"Full-finetune/Learning Rate": 7.122685618363948e-06, "Full-finetune/Loss": 0.7758495807647705, "Full-finetune/Loss (Raw)": 0.9023434519767761, "Full-finetune/Step": 3145, "Full-finetune/Step Time": 6.697520684450865} {"Full-finetune/Learning Rate": 7.116720009568565e-06, "Full-finetune/Loss": 0.7768615484237671, "Full-finetune/Loss (Raw)": 0.8597617745399475, "Full-finetune/Step": 3146, "Full-finetune/Step Time": 6.698655661195517} {"Full-finetune/Learning Rate": 7.110755519717851e-06, "Full-finetune/Loss": 0.7768438458442688, "Full-finetune/Loss (Raw)": 0.7937074303627014, "Full-finetune/Step": 3147, "Full-finetune/Step Time": 6.705954631790519} {"Full-finetune/Learning Rate": 7.104792151126515e-06, "Full-finetune/Loss": 0.7781667709350586, "Full-finetune/Loss (Raw)": 0.9401419162750244, "Full-finetune/Step": 3148, "Full-finetune/Step Time": 6.726248864084482} {"Full-finetune/Learning Rate": 7.098829906108819e-06, "Full-finetune/Loss": 0.7753376960754395, "Full-finetune/Loss (Raw)": 0.42695146799087524, "Full-finetune/Step": 3149, "Full-finetune/Step Time": 6.741151675581932} {"Full-finetune/Learning Rate": 7.092868786978596e-06, "Full-finetune/Loss": 0.7742263078689575, "Full-finetune/Loss (Raw)": 0.7716342806816101, "Full-finetune/Step": 3150, "Full-finetune/Step Time": 6.73874225653708} {"Full-finetune/Learning Rate": 7.086908796049243e-06, "Full-finetune/Loss": 0.7755202054977417, "Full-finetune/Loss (Raw)": 0.831138014793396, "Full-finetune/Step": 3151, "Full-finetune/Step Time": 6.75054301135242} {"Full-finetune/Learning Rate": 7.08094993563371e-06, "Full-finetune/Loss": 0.775374174118042, "Full-finetune/Loss (Raw)": 0.7924538850784302, "Full-finetune/Step": 3152, "Full-finetune/Step Time": 6.745957428589463} {"Full-finetune/Learning Rate": 7.074992208044518e-06, "Full-finetune/Loss": 0.7751536965370178, "Full-finetune/Loss (Raw)": 0.836800754070282, "Full-finetune/Step": 3153, "Full-finetune/Step Time": 6.745798472315073} {"Full-finetune/Learning Rate": 7.069035615593743e-06, "Full-finetune/Loss": 0.775026261806488, "Full-finetune/Loss (Raw)": 0.7903500199317932, "Full-finetune/Step": 3154, "Full-finetune/Step Time": 6.737465962767601} {"Full-finetune/Learning Rate": 7.063080160593026e-06, "Full-finetune/Loss": 0.7747349143028259, "Full-finetune/Loss (Raw)": 0.867363691329956, "Full-finetune/Step": 3155, "Full-finetune/Step Time": 6.73075015284121} {"Full-finetune/Learning Rate": 7.0571258453535585e-06, "Full-finetune/Loss": 0.7743148803710938, "Full-finetune/Loss (Raw)": 0.7556631565093994, "Full-finetune/Step": 3156, "Full-finetune/Step Time": 6.725349368527532} {"Full-finetune/Learning Rate": 7.051172672186093e-06, "Full-finetune/Loss": 0.7737342715263367, "Full-finetune/Loss (Raw)": 0.9349648356437683, "Full-finetune/Step": 3157, "Full-finetune/Step Time": 6.73170268908143} {"Full-finetune/Learning Rate": 7.045220643400946e-06, "Full-finetune/Loss": 0.7699424028396606, "Full-finetune/Loss (Raw)": 0.28913578391075134, "Full-finetune/Step": 3158, "Full-finetune/Step Time": 6.73621928319335} {"Full-finetune/Learning Rate": 7.039269761307977e-06, "Full-finetune/Loss": 0.7690437436103821, "Full-finetune/Loss (Raw)": 0.7260245084762573, "Full-finetune/Step": 3159, "Full-finetune/Step Time": 6.740825366228819} {"Full-finetune/Learning Rate": 7.0333200282166146e-06, "Full-finetune/Loss": 0.7690612077713013, "Full-finetune/Loss (Raw)": 0.8180402517318726, "Full-finetune/Step": 3160, "Full-finetune/Step Time": 6.731048649176955} {"Full-finetune/Learning Rate": 7.027371446435829e-06, "Full-finetune/Loss": 0.7689938545227051, "Full-finetune/Loss (Raw)": 0.8470659255981445, "Full-finetune/Step": 3161, "Full-finetune/Step Time": 6.760465748608112} {"Full-finetune/Learning Rate": 7.0214240182741455e-06, "Full-finetune/Loss": 0.7649730443954468, "Full-finetune/Loss (Raw)": 0.41373616456985474, "Full-finetune/Step": 3162, "Full-finetune/Step Time": 6.779410917311907} {"Full-finetune/Learning Rate": 7.015477746039652e-06, "Full-finetune/Loss": 0.7657065391540527, "Full-finetune/Loss (Raw)": 0.9214378595352173, "Full-finetune/Step": 3163, "Full-finetune/Step Time": 6.800766730681062} {"Full-finetune/Learning Rate": 7.009532632039975e-06, "Full-finetune/Loss": 0.7661116123199463, "Full-finetune/Loss (Raw)": 0.8321182727813721, "Full-finetune/Step": 3164, "Full-finetune/Step Time": 6.78211653418839} {"Full-finetune/Learning Rate": 7.003588678582303e-06, "Full-finetune/Loss": 0.7628662586212158, "Full-finetune/Loss (Raw)": 0.526448667049408, "Full-finetune/Step": 3165, "Full-finetune/Step Time": 6.78312655352056} {"Full-finetune/Learning Rate": 6.9976458879733635e-06, "Full-finetune/Loss": 0.7620809674263, "Full-finetune/Loss (Raw)": 0.7817227244377136, "Full-finetune/Step": 3166, "Full-finetune/Step Time": 6.783617600798607} {"Full-finetune/Learning Rate": 6.991704262519436e-06, "Full-finetune/Loss": 0.7626745104789734, "Full-finetune/Loss (Raw)": 0.8052230477333069, "Full-finetune/Step": 3167, "Full-finetune/Step Time": 6.780620127916336} {"Full-finetune/Learning Rate": 6.985763804526354e-06, "Full-finetune/Loss": 0.763470470905304, "Full-finetune/Loss (Raw)": 0.9389737844467163, "Full-finetune/Step": 3168, "Full-finetune/Step Time": 6.786332318559289} {"Full-finetune/Learning Rate": 6.979824516299491e-06, "Full-finetune/Loss": 0.7602279186248779, "Full-finetune/Loss (Raw)": 0.32101425528526306, "Full-finetune/Step": 3169, "Full-finetune/Step Time": 6.799328912049532} {"Full-finetune/Learning Rate": 6.973886400143772e-06, "Full-finetune/Loss": 0.7590265274047852, "Full-finetune/Loss (Raw)": 0.6624215841293335, "Full-finetune/Step": 3170, "Full-finetune/Step Time": 6.80242177657783} {"Full-finetune/Learning Rate": 6.9679494583636595e-06, "Full-finetune/Loss": 0.7623686194419861, "Full-finetune/Loss (Raw)": 0.7431821227073669, "Full-finetune/Step": 3171, "Full-finetune/Step Time": 6.7863249611109495} {"Full-finetune/Learning Rate": 6.9620136932631655e-06, "Full-finetune/Loss": 0.7630053758621216, "Full-finetune/Loss (Raw)": 0.7499654293060303, "Full-finetune/Step": 3172, "Full-finetune/Step Time": 6.767547337338328} {"Full-finetune/Learning Rate": 6.956079107145845e-06, "Full-finetune/Loss": 0.7637906074523926, "Full-finetune/Loss (Raw)": 0.6700243353843689, "Full-finetune/Step": 3173, "Full-finetune/Step Time": 6.778670990839601} {"Full-finetune/Learning Rate": 6.950145702314795e-06, "Full-finetune/Loss": 0.7636895775794983, "Full-finetune/Loss (Raw)": 0.9750977158546448, "Full-finetune/Step": 3174, "Full-finetune/Step Time": 6.784387532621622} {"Full-finetune/Learning Rate": 6.9442134810726615e-06, "Full-finetune/Loss": 0.7633623480796814, "Full-finetune/Loss (Raw)": 0.7784384489059448, "Full-finetune/Step": 3175, "Full-finetune/Step Time": 6.77997181750834} {"Full-finetune/Learning Rate": 6.938282445721609e-06, "Full-finetune/Loss": 0.7634021639823914, "Full-finetune/Loss (Raw)": 0.9589018821716309, "Full-finetune/Step": 3176, "Full-finetune/Step Time": 6.782691592350602} {"Full-finetune/Learning Rate": 6.932352598563365e-06, "Full-finetune/Loss": 0.7632544040679932, "Full-finetune/Loss (Raw)": 0.7608562707901001, "Full-finetune/Step": 3177, "Full-finetune/Step Time": 6.785208001732826} {"Full-finetune/Learning Rate": 6.92642394189919e-06, "Full-finetune/Loss": 0.7622426748275757, "Full-finetune/Loss (Raw)": 0.8149423003196716, "Full-finetune/Step": 3178, "Full-finetune/Step Time": 6.780417388305068} {"Full-finetune/Learning Rate": 6.920496478029873e-06, "Full-finetune/Loss": 0.7622872591018677, "Full-finetune/Loss (Raw)": 0.9274187684059143, "Full-finetune/Step": 3179, "Full-finetune/Step Time": 6.785887842997909} {"Full-finetune/Learning Rate": 6.914570209255758e-06, "Full-finetune/Loss": 0.7620169520378113, "Full-finetune/Loss (Raw)": 0.7415619492530823, "Full-finetune/Step": 3180, "Full-finetune/Step Time": 6.78308585844934} {"Full-finetune/Learning Rate": 6.908645137876699e-06, "Full-finetune/Loss": 0.761568546295166, "Full-finetune/Loss (Raw)": 0.8434548377990723, "Full-finetune/Step": 3181, "Full-finetune/Step Time": 6.768743289634585} {"Full-finetune/Learning Rate": 6.902721266192112e-06, "Full-finetune/Loss": 0.7604938745498657, "Full-finetune/Loss (Raw)": 0.8323675394058228, "Full-finetune/Step": 3182, "Full-finetune/Step Time": 6.772018173709512} {"Full-finetune/Learning Rate": 6.896798596500929e-06, "Full-finetune/Loss": 0.7595221996307373, "Full-finetune/Loss (Raw)": 0.6924690008163452, "Full-finetune/Step": 3183, "Full-finetune/Step Time": 6.771073758602142} {"Full-finetune/Learning Rate": 6.8908771311016276e-06, "Full-finetune/Loss": 0.7587910294532776, "Full-finetune/Loss (Raw)": 0.7626424431800842, "Full-finetune/Step": 3184, "Full-finetune/Step Time": 6.78214792907238} {"Full-finetune/Learning Rate": 6.884956872292216e-06, "Full-finetune/Loss": 0.7600554823875427, "Full-finetune/Loss (Raw)": 0.8719111084938049, "Full-finetune/Step": 3185, "Full-finetune/Step Time": 6.777022508904338} {"Full-finetune/Learning Rate": 6.87903782237022e-06, "Full-finetune/Loss": 0.7594223022460938, "Full-finetune/Loss (Raw)": 0.7757007479667664, "Full-finetune/Step": 3186, "Full-finetune/Step Time": 6.780490363016725} {"Full-finetune/Learning Rate": 6.8731199836327155e-06, "Full-finetune/Loss": 0.7595114707946777, "Full-finetune/Loss (Raw)": 0.8694202303886414, "Full-finetune/Step": 3187, "Full-finetune/Step Time": 6.779756898060441} {"Full-finetune/Learning Rate": 6.867203358376295e-06, "Full-finetune/Loss": 0.7579715847969055, "Full-finetune/Loss (Raw)": 0.706367015838623, "Full-finetune/Step": 3188, "Full-finetune/Step Time": 6.7876789048314095} {"Full-finetune/Learning Rate": 6.861287948897091e-06, "Full-finetune/Loss": 0.7567532658576965, "Full-finetune/Loss (Raw)": 0.7469689249992371, "Full-finetune/Step": 3189, "Full-finetune/Step Time": 6.7855569291859865} {"Full-finetune/Learning Rate": 6.855373757490761e-06, "Full-finetune/Loss": 0.7576366662979126, "Full-finetune/Loss (Raw)": 0.8408029079437256, "Full-finetune/Step": 3190, "Full-finetune/Step Time": 6.761349970474839} {"Full-finetune/Learning Rate": 6.849460786452476e-06, "Full-finetune/Loss": 0.7579714059829712, "Full-finetune/Loss (Raw)": 0.8388035893440247, "Full-finetune/Step": 3191, "Full-finetune/Step Time": 6.786857469007373} {"Full-finetune/Learning Rate": 6.843549038076951e-06, "Full-finetune/Loss": 0.7576309442520142, "Full-finetune/Loss (Raw)": 0.8102392554283142, "Full-finetune/Step": 3192, "Full-finetune/Step Time": 6.792648021131754} {"Full-finetune/Learning Rate": 6.83763851465842e-06, "Full-finetune/Loss": 0.7588726282119751, "Full-finetune/Loss (Raw)": 0.7925640344619751, "Full-finetune/Step": 3193, "Full-finetune/Step Time": 6.776747651398182} {"Full-finetune/Learning Rate": 6.831729218490645e-06, "Full-finetune/Loss": 0.7605043053627014, "Full-finetune/Loss (Raw)": 0.7409359216690063, "Full-finetune/Step": 3194, "Full-finetune/Step Time": 6.750572016462684} {"Full-finetune/Learning Rate": 6.825821151866903e-06, "Full-finetune/Loss": 0.7606147527694702, "Full-finetune/Loss (Raw)": 0.8374664187431335, "Full-finetune/Step": 3195, "Full-finetune/Step Time": 6.756807800382376} {"Full-finetune/Learning Rate": 6.819914317079999e-06, "Full-finetune/Loss": 0.7610176801681519, "Full-finetune/Loss (Raw)": 0.8001367449760437, "Full-finetune/Step": 3196, "Full-finetune/Step Time": 6.747352844104171} {"Full-finetune/Learning Rate": 6.814008716422265e-06, "Full-finetune/Loss": 0.7625411748886108, "Full-finetune/Loss (Raw)": 0.8723958730697632, "Full-finetune/Step": 3197, "Full-finetune/Step Time": 6.73904499784112} {"Full-finetune/Learning Rate": 6.808104352185544e-06, "Full-finetune/Loss": 0.761770486831665, "Full-finetune/Loss (Raw)": 0.7165291905403137, "Full-finetune/Step": 3198, "Full-finetune/Step Time": 6.747085452079773} {"Full-finetune/Learning Rate": 6.8022012266612116e-06, "Full-finetune/Loss": 0.7605668306350708, "Full-finetune/Loss (Raw)": 0.8174187541007996, "Full-finetune/Step": 3199, "Full-finetune/Step Time": 6.747134068980813} {"Full-finetune/Learning Rate": 6.79629934214015e-06, "Full-finetune/Loss": 0.7623095512390137, "Full-finetune/Loss (Raw)": 1.0151509046554565, "Full-finetune/Step": 3200, "Full-finetune/Step Time": 6.737281857058406} {"Full-finetune/Learning Rate": 6.790398700912764e-06, "Full-finetune/Loss": 0.7606545686721802, "Full-finetune/Loss (Raw)": 0.6084396243095398, "Full-finetune/Step": 3201, "Full-finetune/Step Time": 6.73924127407372} {"Full-finetune/Learning Rate": 6.784499305268983e-06, "Full-finetune/Loss": 0.7604433298110962, "Full-finetune/Loss (Raw)": 0.7706220746040344, "Full-finetune/Step": 3202, "Full-finetune/Step Time": 6.736566131934524} {"Full-finetune/Learning Rate": 6.778601157498241e-06, "Full-finetune/Loss": 0.7642102241516113, "Full-finetune/Loss (Raw)": 0.7838107347488403, "Full-finetune/Step": 3203, "Full-finetune/Step Time": 6.719252901151776} {"Full-finetune/Learning Rate": 6.772704259889503e-06, "Full-finetune/Loss": 0.7648767232894897, "Full-finetune/Loss (Raw)": 0.7699339389801025, "Full-finetune/Step": 3204, "Full-finetune/Step Time": 6.725094605237246} {"Full-finetune/Learning Rate": 6.76680861473123e-06, "Full-finetune/Loss": 0.7639914751052856, "Full-finetune/Loss (Raw)": 0.739111602306366, "Full-finetune/Step": 3205, "Full-finetune/Step Time": 6.720814865082502} {"Full-finetune/Learning Rate": 6.7609142243114115e-06, "Full-finetune/Loss": 0.7638174295425415, "Full-finetune/Loss (Raw)": 0.6716622114181519, "Full-finetune/Step": 3206, "Full-finetune/Step Time": 6.722389021888375} {"Full-finetune/Learning Rate": 6.755021090917545e-06, "Full-finetune/Loss": 0.7649546265602112, "Full-finetune/Loss (Raw)": 0.7666314244270325, "Full-finetune/Step": 3207, "Full-finetune/Step Time": 6.7220762334764} {"Full-finetune/Learning Rate": 6.7491292168366386e-06, "Full-finetune/Loss": 0.765888512134552, "Full-finetune/Loss (Raw)": 0.8171159029006958, "Full-finetune/Step": 3208, "Full-finetune/Step Time": 6.715314542874694} {"Full-finetune/Learning Rate": 6.743238604355219e-06, "Full-finetune/Loss": 0.766227662563324, "Full-finetune/Loss (Raw)": 0.7737196087837219, "Full-finetune/Step": 3209, "Full-finetune/Step Time": 6.718386923894286} {"Full-finetune/Learning Rate": 6.737349255759312e-06, "Full-finetune/Loss": 0.7654210329055786, "Full-finetune/Loss (Raw)": 0.7195634841918945, "Full-finetune/Step": 3210, "Full-finetune/Step Time": 6.727108808234334} {"Full-finetune/Learning Rate": 6.731461173334459e-06, "Full-finetune/Loss": 0.7649999856948853, "Full-finetune/Loss (Raw)": 0.7970199584960938, "Full-finetune/Step": 3211, "Full-finetune/Step Time": 6.72061537578702} {"Full-finetune/Learning Rate": 6.725574359365714e-06, "Full-finetune/Loss": 0.7621673345565796, "Full-finetune/Loss (Raw)": 0.40128225088119507, "Full-finetune/Step": 3212, "Full-finetune/Step Time": 6.729560486972332} {"Full-finetune/Learning Rate": 6.7196888161376305e-06, "Full-finetune/Loss": 0.7610838413238525, "Full-finetune/Loss (Raw)": 0.8456692695617676, "Full-finetune/Step": 3213, "Full-finetune/Step Time": 6.740383578464389} {"Full-finetune/Learning Rate": 6.71380454593428e-06, "Full-finetune/Loss": 0.7600506544113159, "Full-finetune/Loss (Raw)": 0.7017310261726379, "Full-finetune/Step": 3214, "Full-finetune/Step Time": 6.7519885916262865} {"Full-finetune/Learning Rate": 6.707921551039224e-06, "Full-finetune/Loss": 0.7589415311813354, "Full-finetune/Loss (Raw)": 0.6540650725364685, "Full-finetune/Step": 3215, "Full-finetune/Step Time": 6.760475968942046} {"Full-finetune/Learning Rate": 6.70203983373554e-06, "Full-finetune/Loss": 0.7597070932388306, "Full-finetune/Loss (Raw)": 0.651081919670105, "Full-finetune/Step": 3216, "Full-finetune/Step Time": 6.791068632155657} {"Full-finetune/Learning Rate": 6.69615939630581e-06, "Full-finetune/Loss": 0.759014904499054, "Full-finetune/Loss (Raw)": 0.8128146529197693, "Full-finetune/Step": 3217, "Full-finetune/Step Time": 6.782705707475543} {"Full-finetune/Learning Rate": 6.690280241032117e-06, "Full-finetune/Loss": 0.7597205638885498, "Full-finetune/Loss (Raw)": 0.8216561675071716, "Full-finetune/Step": 3218, "Full-finetune/Step Time": 6.780589820817113} {"Full-finetune/Learning Rate": 6.684402370196048e-06, "Full-finetune/Loss": 0.7594219446182251, "Full-finetune/Loss (Raw)": 0.6799288988113403, "Full-finetune/Step": 3219, "Full-finetune/Step Time": 6.764291243627667} {"Full-finetune/Learning Rate": 6.678525786078684e-06, "Full-finetune/Loss": 0.7644694447517395, "Full-finetune/Loss (Raw)": 0.9500492215156555, "Full-finetune/Step": 3220, "Full-finetune/Step Time": 6.74479428678751} {"Full-finetune/Learning Rate": 6.672650490960617e-06, "Full-finetune/Loss": 0.7638422250747681, "Full-finetune/Loss (Raw)": 0.7465294599533081, "Full-finetune/Step": 3221, "Full-finetune/Step Time": 6.747610619291663} {"Full-finetune/Learning Rate": 6.666776487121934e-06, "Full-finetune/Loss": 0.7597728371620178, "Full-finetune/Loss (Raw)": 0.3364168405532837, "Full-finetune/Step": 3222, "Full-finetune/Step Time": 6.7624350767582655} {"Full-finetune/Learning Rate": 6.660903776842218e-06, "Full-finetune/Loss": 0.7596895694732666, "Full-finetune/Loss (Raw)": 0.7683790922164917, "Full-finetune/Step": 3223, "Full-finetune/Step Time": 6.768268762156367} {"Full-finetune/Learning Rate": 6.655032362400562e-06, "Full-finetune/Loss": 0.7594407200813293, "Full-finetune/Loss (Raw)": 0.7921338677406311, "Full-finetune/Step": 3224, "Full-finetune/Step Time": 6.759758232161403} {"Full-finetune/Learning Rate": 6.649162246075538e-06, "Full-finetune/Loss": 0.7631604671478271, "Full-finetune/Loss (Raw)": 0.7899054884910583, "Full-finetune/Step": 3225, "Full-finetune/Step Time": 6.744817182421684} {"Full-finetune/Learning Rate": 6.6432934301452235e-06, "Full-finetune/Loss": 0.7648231983184814, "Full-finetune/Loss (Raw)": 0.7853935360908508, "Full-finetune/Step": 3226, "Full-finetune/Step Time": 6.751943968236446} {"Full-finetune/Learning Rate": 6.637425916887198e-06, "Full-finetune/Loss": 0.7636616230010986, "Full-finetune/Loss (Raw)": 0.6925784349441528, "Full-finetune/Step": 3227, "Full-finetune/Step Time": 6.752124415710568} {"Full-finetune/Learning Rate": 6.631559708578524e-06, "Full-finetune/Loss": 0.7617956399917603, "Full-finetune/Loss (Raw)": 0.7713550329208374, "Full-finetune/Step": 3228, "Full-finetune/Step Time": 6.750158032402396} {"Full-finetune/Learning Rate": 6.625694807495766e-06, "Full-finetune/Loss": 0.764390230178833, "Full-finetune/Loss (Raw)": 0.8886829018592834, "Full-finetune/Step": 3229, "Full-finetune/Step Time": 6.734725210815668} {"Full-finetune/Learning Rate": 6.619831215914974e-06, "Full-finetune/Loss": 0.7629538178443909, "Full-finetune/Loss (Raw)": 0.7169761657714844, "Full-finetune/Step": 3230, "Full-finetune/Step Time": 6.735576890408993} {"Full-finetune/Learning Rate": 6.6139689361116945e-06, "Full-finetune/Loss": 0.7618751525878906, "Full-finetune/Loss (Raw)": 0.7921411991119385, "Full-finetune/Step": 3231, "Full-finetune/Step Time": 6.748615212738514} {"Full-finetune/Learning Rate": 6.608107970360964e-06, "Full-finetune/Loss": 0.7624834775924683, "Full-finetune/Loss (Raw)": 0.9404639005661011, "Full-finetune/Step": 3232, "Full-finetune/Step Time": 6.750313842669129} {"Full-finetune/Learning Rate": 6.602248320937308e-06, "Full-finetune/Loss": 0.763678789138794, "Full-finetune/Loss (Raw)": 1.0065116882324219, "Full-finetune/Step": 3233, "Full-finetune/Step Time": 6.754162544384599} {"Full-finetune/Learning Rate": 6.596389990114748e-06, "Full-finetune/Loss": 0.7635538578033447, "Full-finetune/Loss (Raw)": 0.7037295699119568, "Full-finetune/Step": 3234, "Full-finetune/Step Time": 6.764284087345004} {"Full-finetune/Learning Rate": 6.5905329801667796e-06, "Full-finetune/Loss": 0.7632676959037781, "Full-finetune/Loss (Raw)": 0.925652265548706, "Full-finetune/Step": 3235, "Full-finetune/Step Time": 6.769055875018239} {"Full-finetune/Learning Rate": 6.584677293366396e-06, "Full-finetune/Loss": 0.7623739838600159, "Full-finetune/Loss (Raw)": 0.6373139023780823, "Full-finetune/Step": 3236, "Full-finetune/Step Time": 6.763468232005835} {"Full-finetune/Learning Rate": 6.578822931986078e-06, "Full-finetune/Loss": 0.763275682926178, "Full-finetune/Loss (Raw)": 0.8730849623680115, "Full-finetune/Step": 3237, "Full-finetune/Step Time": 6.761633213609457} {"Full-finetune/Learning Rate": 6.57296989829779e-06, "Full-finetune/Loss": 0.7607199549674988, "Full-finetune/Loss (Raw)": 0.40303170680999756, "Full-finetune/Step": 3238, "Full-finetune/Step Time": 6.781636621803045} {"Full-finetune/Learning Rate": 6.567118194572972e-06, "Full-finetune/Loss": 0.7623252868652344, "Full-finetune/Loss (Raw)": 0.9697590470314026, "Full-finetune/Step": 3239, "Full-finetune/Step Time": 6.7783878687769175} {"Full-finetune/Learning Rate": 6.5612678230825646e-06, "Full-finetune/Loss": 0.7627227306365967, "Full-finetune/Loss (Raw)": 0.9358248710632324, "Full-finetune/Step": 3240, "Full-finetune/Step Time": 6.779156971722841} {"Full-finetune/Learning Rate": 6.555418786096976e-06, "Full-finetune/Loss": 0.763584315776825, "Full-finetune/Loss (Raw)": 0.89469313621521, "Full-finetune/Step": 3241, "Full-finetune/Step Time": 6.769002368673682} {"Full-finetune/Learning Rate": 6.549571085886111e-06, "Full-finetune/Loss": 0.7640923857688904, "Full-finetune/Loss (Raw)": 0.9318058490753174, "Full-finetune/Step": 3242, "Full-finetune/Step Time": 6.7659674268215895} {"Full-finetune/Learning Rate": 6.5437247247193445e-06, "Full-finetune/Loss": 0.7645494937896729, "Full-finetune/Loss (Raw)": 0.8377025127410889, "Full-finetune/Step": 3243, "Full-finetune/Step Time": 6.764717441052198} {"Full-finetune/Learning Rate": 6.537879704865531e-06, "Full-finetune/Loss": 0.7667088508605957, "Full-finetune/Loss (Raw)": 0.8449141979217529, "Full-finetune/Step": 3244, "Full-finetune/Step Time": 6.72997316904366} {"Full-finetune/Learning Rate": 6.532036028593012e-06, "Full-finetune/Loss": 0.7669218182563782, "Full-finetune/Loss (Raw)": 0.8813817501068115, "Full-finetune/Step": 3245, "Full-finetune/Step Time": 6.736266674473882} {"Full-finetune/Learning Rate": 6.526193698169603e-06, "Full-finetune/Loss": 0.7684778571128845, "Full-finetune/Loss (Raw)": 0.6812301874160767, "Full-finetune/Step": 3246, "Full-finetune/Step Time": 6.720797993242741} {"Full-finetune/Learning Rate": 6.520352715862601e-06, "Full-finetune/Loss": 0.7679354548454285, "Full-finetune/Loss (Raw)": 0.7417967319488525, "Full-finetune/Step": 3247, "Full-finetune/Step Time": 6.718874394893646} {"Full-finetune/Learning Rate": 6.514513083938779e-06, "Full-finetune/Loss": 0.768764317035675, "Full-finetune/Loss (Raw)": 0.9456082582473755, "Full-finetune/Step": 3248, "Full-finetune/Step Time": 6.723624315112829} {"Full-finetune/Learning Rate": 6.508674804664374e-06, "Full-finetune/Loss": 0.771417498588562, "Full-finetune/Loss (Raw)": 0.9487226009368896, "Full-finetune/Step": 3249, "Full-finetune/Step Time": 6.712868286296725} {"Full-finetune/Learning Rate": 6.50283788030512e-06, "Full-finetune/Loss": 0.7713265419006348, "Full-finetune/Loss (Raw)": 0.7558329105377197, "Full-finetune/Step": 3250, "Full-finetune/Step Time": 6.7323589865118265} {"Full-finetune/Learning Rate": 6.497002313126204e-06, "Full-finetune/Loss": 0.7703033089637756, "Full-finetune/Loss (Raw)": 0.6573648452758789, "Full-finetune/Step": 3251, "Full-finetune/Step Time": 6.759486023336649} {"Full-finetune/Learning Rate": 6.491168105392303e-06, "Full-finetune/Loss": 0.7704744338989258, "Full-finetune/Loss (Raw)": 0.8852940201759338, "Full-finetune/Step": 3252, "Full-finetune/Step Time": 6.760985516011715} {"Full-finetune/Learning Rate": 6.485335259367558e-06, "Full-finetune/Loss": 0.7718802690505981, "Full-finetune/Loss (Raw)": 0.8325269222259521, "Full-finetune/Step": 3253, "Full-finetune/Step Time": 6.76323795504868} {"Full-finetune/Learning Rate": 6.4795037773155765e-06, "Full-finetune/Loss": 0.7719190120697021, "Full-finetune/Loss (Raw)": 0.8705286383628845, "Full-finetune/Step": 3254, "Full-finetune/Step Time": 6.763052858412266} {"Full-finetune/Learning Rate": 6.473673661499451e-06, "Full-finetune/Loss": 0.7713279724121094, "Full-finetune/Loss (Raw)": 0.8732350468635559, "Full-finetune/Step": 3255, "Full-finetune/Step Time": 6.773626871407032} {"Full-finetune/Learning Rate": 6.467844914181727e-06, "Full-finetune/Loss": 0.774124264717102, "Full-finetune/Loss (Raw)": 0.8698766827583313, "Full-finetune/Step": 3256, "Full-finetune/Step Time": 6.740112539380789} {"Full-finetune/Learning Rate": 6.462017537624437e-06, "Full-finetune/Loss": 0.772998034954071, "Full-finetune/Loss (Raw)": 0.6230020523071289, "Full-finetune/Step": 3257, "Full-finetune/Step Time": 6.750181224197149} {"Full-finetune/Learning Rate": 6.4561915340890665e-06, "Full-finetune/Loss": 0.7752820253372192, "Full-finetune/Loss (Raw)": 1.0108535289764404, "Full-finetune/Step": 3258, "Full-finetune/Step Time": 6.739221161231399} {"Full-finetune/Learning Rate": 6.450366905836573e-06, "Full-finetune/Loss": 0.7749963998794556, "Full-finetune/Loss (Raw)": 0.7092683911323547, "Full-finetune/Step": 3259, "Full-finetune/Step Time": 6.741863174363971} {"Full-finetune/Learning Rate": 6.444543655127383e-06, "Full-finetune/Loss": 0.7781621217727661, "Full-finetune/Loss (Raw)": 0.8204318881034851, "Full-finetune/Step": 3260, "Full-finetune/Step Time": 6.742234472185373} {"Full-finetune/Learning Rate": 6.438721784221385e-06, "Full-finetune/Loss": 0.7776020765304565, "Full-finetune/Loss (Raw)": 0.7492309212684631, "Full-finetune/Step": 3261, "Full-finetune/Step Time": 6.741895450279117} {"Full-finetune/Learning Rate": 6.432901295377936e-06, "Full-finetune/Loss": 0.7795310020446777, "Full-finetune/Loss (Raw)": 0.9173963665962219, "Full-finetune/Step": 3262, "Full-finetune/Step Time": 6.728427955880761} {"Full-finetune/Learning Rate": 6.427082190855855e-06, "Full-finetune/Loss": 0.7800878286361694, "Full-finetune/Loss (Raw)": 0.9393777847290039, "Full-finetune/Step": 3263, "Full-finetune/Step Time": 6.726226607337594} {"Full-finetune/Learning Rate": 6.4212644729134155e-06, "Full-finetune/Loss": 0.7792614698410034, "Full-finetune/Loss (Raw)": 0.7152301073074341, "Full-finetune/Step": 3264, "Full-finetune/Step Time": 6.728099409490824} {"Full-finetune/Learning Rate": 6.415448143808367e-06, "Full-finetune/Loss": 0.7811552882194519, "Full-finetune/Loss (Raw)": 0.9550792574882507, "Full-finetune/Step": 3265, "Full-finetune/Step Time": 6.722145805135369} {"Full-finetune/Learning Rate": 6.409633205797911e-06, "Full-finetune/Loss": 0.7822614908218384, "Full-finetune/Loss (Raw)": 0.7931324243545532, "Full-finetune/Step": 3266, "Full-finetune/Step Time": 6.7268605418503284} {"Full-finetune/Learning Rate": 6.403819661138712e-06, "Full-finetune/Loss": 0.7816799283027649, "Full-finetune/Loss (Raw)": 0.7210443615913391, "Full-finetune/Step": 3267, "Full-finetune/Step Time": 6.74079879000783} {"Full-finetune/Learning Rate": 6.398007512086897e-06, "Full-finetune/Loss": 0.78256756067276, "Full-finetune/Loss (Raw)": 0.8896908164024353, "Full-finetune/Step": 3268, "Full-finetune/Step Time": 6.733920317143202} {"Full-finetune/Learning Rate": 6.3921967608980375e-06, "Full-finetune/Loss": 0.7832950353622437, "Full-finetune/Loss (Raw)": 0.8791278600692749, "Full-finetune/Step": 3269, "Full-finetune/Step Time": 6.737765474244952} {"Full-finetune/Learning Rate": 6.386387409827182e-06, "Full-finetune/Loss": 0.7833241820335388, "Full-finetune/Loss (Raw)": 0.7401536703109741, "Full-finetune/Step": 3270, "Full-finetune/Step Time": 6.737706299871206} {"Full-finetune/Learning Rate": 6.38057946112882e-06, "Full-finetune/Loss": 0.7853280305862427, "Full-finetune/Loss (Raw)": 0.7774760723114014, "Full-finetune/Step": 3271, "Full-finetune/Step Time": 6.722070246934891} {"Full-finetune/Learning Rate": 6.374772917056908e-06, "Full-finetune/Loss": 0.7840679883956909, "Full-finetune/Loss (Raw)": 0.7651498913764954, "Full-finetune/Step": 3272, "Full-finetune/Step Time": 6.750507045537233} {"Full-finetune/Learning Rate": 6.368967779864854e-06, "Full-finetune/Loss": 0.7827509045600891, "Full-finetune/Loss (Raw)": 0.7337554097175598, "Full-finetune/Step": 3273, "Full-finetune/Step Time": 6.756080796942115} {"Full-finetune/Learning Rate": 6.3631640518055085e-06, "Full-finetune/Loss": 0.7785717248916626, "Full-finetune/Loss (Raw)": 0.3248192071914673, "Full-finetune/Step": 3274, "Full-finetune/Step Time": 6.773098360747099} {"Full-finetune/Learning Rate": 6.3573617351311945e-06, "Full-finetune/Loss": 0.7779955267906189, "Full-finetune/Loss (Raw)": 0.7199591398239136, "Full-finetune/Step": 3275, "Full-finetune/Step Time": 6.772705452516675} {"Full-finetune/Learning Rate": 6.351560832093673e-06, "Full-finetune/Loss": 0.7771620750427246, "Full-finetune/Loss (Raw)": 0.8334540128707886, "Full-finetune/Step": 3276, "Full-finetune/Step Time": 6.756861265748739} {"Full-finetune/Learning Rate": 6.3457613449441654e-06, "Full-finetune/Loss": 0.7802557945251465, "Full-finetune/Loss (Raw)": 0.8229543566703796, "Full-finetune/Step": 3277, "Full-finetune/Step Time": 6.7377398405224085} {"Full-finetune/Learning Rate": 6.3399632759333364e-06, "Full-finetune/Loss": 0.7804296016693115, "Full-finetune/Loss (Raw)": 0.7938787937164307, "Full-finetune/Step": 3278, "Full-finetune/Step Time": 6.739969165995717} {"Full-finetune/Learning Rate": 6.3341666273113e-06, "Full-finetune/Loss": 0.7804291844367981, "Full-finetune/Loss (Raw)": 0.8310897350311279, "Full-finetune/Step": 3279, "Full-finetune/Step Time": 6.736128905788064} {"Full-finetune/Learning Rate": 6.328371401327629e-06, "Full-finetune/Loss": 0.7814962863922119, "Full-finetune/Loss (Raw)": 0.9290328025817871, "Full-finetune/Step": 3280, "Full-finetune/Step Time": 6.735498066991568} {"Full-finetune/Learning Rate": 6.322577600231332e-06, "Full-finetune/Loss": 0.7826164364814758, "Full-finetune/Loss (Raw)": 0.9801878929138184, "Full-finetune/Step": 3281, "Full-finetune/Step Time": 6.7374298721551895} {"Full-finetune/Learning Rate": 6.316785226270878e-06, "Full-finetune/Loss": 0.7825030088424683, "Full-finetune/Loss (Raw)": 0.7758257985115051, "Full-finetune/Step": 3282, "Full-finetune/Step Time": 6.738971706479788} {"Full-finetune/Learning Rate": 6.310994281694167e-06, "Full-finetune/Loss": 0.7786507606506348, "Full-finetune/Loss (Raw)": 0.37428006529808044, "Full-finetune/Step": 3283, "Full-finetune/Step Time": 6.755764376372099} {"Full-finetune/Learning Rate": 6.305204768748551e-06, "Full-finetune/Loss": 0.7786288261413574, "Full-finetune/Loss (Raw)": 0.7528535723686218, "Full-finetune/Step": 3284, "Full-finetune/Step Time": 6.771067516878247} {"Full-finetune/Learning Rate": 6.299416689680831e-06, "Full-finetune/Loss": 0.7779884338378906, "Full-finetune/Loss (Raw)": 0.852998673915863, "Full-finetune/Step": 3285, "Full-finetune/Step Time": 6.764479553326964} {"Full-finetune/Learning Rate": 6.293630046737245e-06, "Full-finetune/Loss": 0.7819663286209106, "Full-finetune/Loss (Raw)": 0.798301100730896, "Full-finetune/Step": 3286, "Full-finetune/Step Time": 6.75227689743042} {"Full-finetune/Learning Rate": 6.28784484216348e-06, "Full-finetune/Loss": 0.7827877998352051, "Full-finetune/Loss (Raw)": 0.8311777710914612, "Full-finetune/Step": 3287, "Full-finetune/Step Time": 6.757832793518901} {"Full-finetune/Learning Rate": 6.282061078204657e-06, "Full-finetune/Loss": 0.7835582494735718, "Full-finetune/Loss (Raw)": 0.9166600108146667, "Full-finetune/Step": 3288, "Full-finetune/Step Time": 6.7630102802067995} {"Full-finetune/Learning Rate": 6.276278757105341e-06, "Full-finetune/Loss": 0.7827930450439453, "Full-finetune/Loss (Raw)": 0.7491214871406555, "Full-finetune/Step": 3289, "Full-finetune/Step Time": 6.733758609741926} {"Full-finetune/Learning Rate": 6.270497881109541e-06, "Full-finetune/Loss": 0.7861913442611694, "Full-finetune/Loss (Raw)": 0.848713755607605, "Full-finetune/Step": 3290, "Full-finetune/Step Time": 6.704859439283609} {"Full-finetune/Learning Rate": 6.264718452460699e-06, "Full-finetune/Loss": 0.784292459487915, "Full-finetune/Loss (Raw)": 0.6783834099769592, "Full-finetune/Step": 3291, "Full-finetune/Step Time": 6.689745381474495} {"Full-finetune/Learning Rate": 6.258940473401705e-06, "Full-finetune/Loss": 0.7837047576904297, "Full-finetune/Loss (Raw)": 0.7568894028663635, "Full-finetune/Step": 3292, "Full-finetune/Step Time": 6.689988570287824} {"Full-finetune/Learning Rate": 6.253163946174872e-06, "Full-finetune/Loss": 0.7859815359115601, "Full-finetune/Loss (Raw)": 0.8178824186325073, "Full-finetune/Step": 3293, "Full-finetune/Step Time": 6.686756396666169} {"Full-finetune/Learning Rate": 6.247388873021959e-06, "Full-finetune/Loss": 0.7854670882225037, "Full-finetune/Loss (Raw)": 0.7158679962158203, "Full-finetune/Step": 3294, "Full-finetune/Step Time": 6.7130304127931595} {"Full-finetune/Learning Rate": 6.2416152561841635e-06, "Full-finetune/Loss": 0.7853336334228516, "Full-finetune/Loss (Raw)": 0.788144588470459, "Full-finetune/Step": 3295, "Full-finetune/Step Time": 6.716397682204843} {"Full-finetune/Learning Rate": 6.2358430979021075e-06, "Full-finetune/Loss": 0.7845170497894287, "Full-finetune/Loss (Raw)": 0.834449827671051, "Full-finetune/Step": 3296, "Full-finetune/Step Time": 6.717368740588427} {"Full-finetune/Learning Rate": 6.230072400415863e-06, "Full-finetune/Loss": 0.787293553352356, "Full-finetune/Loss (Raw)": 0.6764055490493774, "Full-finetune/Step": 3297, "Full-finetune/Step Time": 6.720759965479374} {"Full-finetune/Learning Rate": 6.224303165964913e-06, "Full-finetune/Loss": 0.7891489267349243, "Full-finetune/Loss (Raw)": 0.8999064564704895, "Full-finetune/Step": 3298, "Full-finetune/Step Time": 6.71772420220077} {"Full-finetune/Learning Rate": 6.21853539678819e-06, "Full-finetune/Loss": 0.786423921585083, "Full-finetune/Loss (Raw)": 0.39438286423683167, "Full-finetune/Step": 3299, "Full-finetune/Step Time": 6.734395857900381} {"Full-finetune/Learning Rate": 6.212769095124055e-06, "Full-finetune/Loss": 0.78733229637146, "Full-finetune/Loss (Raw)": 0.8662366271018982, "Full-finetune/Step": 3300, "Full-finetune/Step Time": 6.737246951088309} {"Full-finetune/Learning Rate": 6.207004263210292e-06, "Full-finetune/Loss": 0.7873219847679138, "Full-finetune/Loss (Raw)": 0.6687055826187134, "Full-finetune/Step": 3301, "Full-finetune/Step Time": 6.7261983919888735} {"Full-finetune/Learning Rate": 6.2012409032841294e-06, "Full-finetune/Loss": 0.7867785692214966, "Full-finetune/Loss (Raw)": 0.9055349826812744, "Full-finetune/Step": 3302, "Full-finetune/Step Time": 6.723866144195199} {"Full-finetune/Learning Rate": 6.1954790175822e-06, "Full-finetune/Loss": 0.7868010997772217, "Full-finetune/Loss (Raw)": 0.7813258767127991, "Full-finetune/Step": 3303, "Full-finetune/Step Time": 6.722361532971263} {"Full-finetune/Learning Rate": 6.189718608340587e-06, "Full-finetune/Loss": 0.7849764823913574, "Full-finetune/Loss (Raw)": 0.7253557443618774, "Full-finetune/Step": 3304, "Full-finetune/Step Time": 6.731363398954272} {"Full-finetune/Learning Rate": 6.183959677794795e-06, "Full-finetune/Loss": 0.7857948541641235, "Full-finetune/Loss (Raw)": 0.8656037449836731, "Full-finetune/Step": 3305, "Full-finetune/Step Time": 6.728768588975072} {"Full-finetune/Learning Rate": 6.1782022281797485e-06, "Full-finetune/Loss": 0.7850156426429749, "Full-finetune/Loss (Raw)": 0.7152082920074463, "Full-finetune/Step": 3306, "Full-finetune/Step Time": 6.720760948956013} {"Full-finetune/Learning Rate": 6.17244626172981e-06, "Full-finetune/Loss": 0.7851570844650269, "Full-finetune/Loss (Raw)": 0.9455158114433289, "Full-finetune/Step": 3307, "Full-finetune/Step Time": 6.710917130112648} {"Full-finetune/Learning Rate": 6.166691780678743e-06, "Full-finetune/Loss": 0.7858213186264038, "Full-finetune/Loss (Raw)": 0.8265839219093323, "Full-finetune/Step": 3308, "Full-finetune/Step Time": 6.716271795332432} {"Full-finetune/Learning Rate": 6.1609387872597595e-06, "Full-finetune/Loss": 0.7860800623893738, "Full-finetune/Loss (Raw)": 0.8765760660171509, "Full-finetune/Step": 3309, "Full-finetune/Step Time": 6.7259704396128654} {"Full-finetune/Learning Rate": 6.155187283705479e-06, "Full-finetune/Loss": 0.7853946685791016, "Full-finetune/Loss (Raw)": 0.7446406483650208, "Full-finetune/Step": 3310, "Full-finetune/Step Time": 6.722897270694375} {"Full-finetune/Learning Rate": 6.14943727224795e-06, "Full-finetune/Loss": 0.7855216860771179, "Full-finetune/Loss (Raw)": 0.7087278366088867, "Full-finetune/Step": 3311, "Full-finetune/Step Time": 6.73033525981009} {"Full-finetune/Learning Rate": 6.1436887551186466e-06, "Full-finetune/Loss": 0.7856600284576416, "Full-finetune/Loss (Raw)": 0.7803453207015991, "Full-finetune/Step": 3312, "Full-finetune/Step Time": 6.723726436495781} {"Full-finetune/Learning Rate": 6.137941734548443e-06, "Full-finetune/Loss": 0.7822109460830688, "Full-finetune/Loss (Raw)": 0.430438756942749, "Full-finetune/Step": 3313, "Full-finetune/Step Time": 6.735162692144513} {"Full-finetune/Learning Rate": 6.132196212767653e-06, "Full-finetune/Loss": 0.7818841338157654, "Full-finetune/Loss (Raw)": 0.7338581681251526, "Full-finetune/Step": 3314, "Full-finetune/Step Time": 6.733138531446457} {"Full-finetune/Learning Rate": 6.126452192005999e-06, "Full-finetune/Loss": 0.7827706336975098, "Full-finetune/Loss (Raw)": 0.9828975200653076, "Full-finetune/Step": 3315, "Full-finetune/Step Time": 6.7320668417960405} {"Full-finetune/Learning Rate": 6.120709674492627e-06, "Full-finetune/Loss": 0.7831182479858398, "Full-finetune/Loss (Raw)": 0.7508605122566223, "Full-finetune/Step": 3316, "Full-finetune/Step Time": 6.723766461014748} {"Full-finetune/Learning Rate": 6.114968662456093e-06, "Full-finetune/Loss": 0.7831273674964905, "Full-finetune/Loss (Raw)": 0.7481388449668884, "Full-finetune/Step": 3317, "Full-finetune/Step Time": 6.723179368302226} {"Full-finetune/Learning Rate": 6.10922915812437e-06, "Full-finetune/Loss": 0.7792361974716187, "Full-finetune/Loss (Raw)": 0.34272563457489014, "Full-finetune/Step": 3318, "Full-finetune/Step Time": 6.735122302547097} {"Full-finetune/Learning Rate": 6.103491163724853e-06, "Full-finetune/Loss": 0.7796649932861328, "Full-finetune/Loss (Raw)": 0.8936905860900879, "Full-finetune/Step": 3319, "Full-finetune/Step Time": 6.716648010537028} {"Full-finetune/Learning Rate": 6.09775468148434e-06, "Full-finetune/Loss": 0.7793259620666504, "Full-finetune/Loss (Raw)": 0.7668406367301941, "Full-finetune/Step": 3320, "Full-finetune/Step Time": 6.707929879426956} {"Full-finetune/Learning Rate": 6.092019713629053e-06, "Full-finetune/Loss": 0.7786598205566406, "Full-finetune/Loss (Raw)": 0.7073032855987549, "Full-finetune/Step": 3321, "Full-finetune/Step Time": 6.707198219373822} {"Full-finetune/Learning Rate": 6.086286262384617e-06, "Full-finetune/Loss": 0.7788304090499878, "Full-finetune/Loss (Raw)": 0.7627712488174438, "Full-finetune/Step": 3322, "Full-finetune/Step Time": 6.698759159073234} {"Full-finetune/Learning Rate": 6.080554329976071e-06, "Full-finetune/Loss": 0.7793418169021606, "Full-finetune/Loss (Raw)": 0.90293288230896, "Full-finetune/Step": 3323, "Full-finetune/Step Time": 6.725700415670872} {"Full-finetune/Learning Rate": 6.074823918627872e-06, "Full-finetune/Loss": 0.7790459990501404, "Full-finetune/Loss (Raw)": 0.7622654438018799, "Full-finetune/Step": 3324, "Full-finetune/Step Time": 6.7475737147033215} {"Full-finetune/Learning Rate": 6.069095030563877e-06, "Full-finetune/Loss": 0.7775788307189941, "Full-finetune/Loss (Raw)": 0.6846007704734802, "Full-finetune/Step": 3325, "Full-finetune/Step Time": 6.746904203668237} {"Full-finetune/Learning Rate": 6.063367668007356e-06, "Full-finetune/Loss": 0.778812050819397, "Full-finetune/Loss (Raw)": 0.8743746876716614, "Full-finetune/Step": 3326, "Full-finetune/Step Time": 6.738063715398312} {"Full-finetune/Learning Rate": 6.057641833180987e-06, "Full-finetune/Loss": 0.7769085168838501, "Full-finetune/Loss (Raw)": 0.5737811923027039, "Full-finetune/Step": 3327, "Full-finetune/Step Time": 6.7494173515588045} {"Full-finetune/Learning Rate": 6.0519175283068534e-06, "Full-finetune/Loss": 0.7762137055397034, "Full-finetune/Loss (Raw)": 0.9262023568153381, "Full-finetune/Step": 3328, "Full-finetune/Step Time": 6.752252858132124} {"Full-finetune/Learning Rate": 6.046194755606448e-06, "Full-finetune/Loss": 0.7777895331382751, "Full-finetune/Loss (Raw)": 0.8101498484611511, "Full-finetune/Step": 3329, "Full-finetune/Step Time": 6.754234140738845} {"Full-finetune/Learning Rate": 6.040473517300665e-06, "Full-finetune/Loss": 0.777205228805542, "Full-finetune/Loss (Raw)": 0.6958245635032654, "Full-finetune/Step": 3330, "Full-finetune/Step Time": 6.766353024169803} {"Full-finetune/Learning Rate": 6.0347538156098104e-06, "Full-finetune/Loss": 0.7774025201797485, "Full-finetune/Loss (Raw)": 0.8090744614601135, "Full-finetune/Step": 3331, "Full-finetune/Step Time": 6.778281960636377} {"Full-finetune/Learning Rate": 6.0290356527535836e-06, "Full-finetune/Loss": 0.7775623202323914, "Full-finetune/Loss (Raw)": 0.790384829044342, "Full-finetune/Step": 3332, "Full-finetune/Step Time": 6.770660731941462} {"Full-finetune/Learning Rate": 6.023319030951091e-06, "Full-finetune/Loss": 0.778179407119751, "Full-finetune/Loss (Raw)": 0.8180946707725525, "Full-finetune/Step": 3333, "Full-finetune/Step Time": 6.769871350377798} {"Full-finetune/Learning Rate": 6.017603952420846e-06, "Full-finetune/Loss": 0.7793124914169312, "Full-finetune/Loss (Raw)": 0.8166991472244263, "Full-finetune/Step": 3334, "Full-finetune/Step Time": 6.769153993576765} {"Full-finetune/Learning Rate": 6.011890419380756e-06, "Full-finetune/Loss": 0.7789636850357056, "Full-finetune/Loss (Raw)": 0.7219854593276978, "Full-finetune/Step": 3335, "Full-finetune/Step Time": 6.746677653864026} {"Full-finetune/Learning Rate": 6.0061784340481366e-06, "Full-finetune/Loss": 0.7785377502441406, "Full-finetune/Loss (Raw)": 0.7625989317893982, "Full-finetune/Step": 3336, "Full-finetune/Step Time": 6.745254136621952} {"Full-finetune/Learning Rate": 6.000467998639692e-06, "Full-finetune/Loss": 0.7778239250183105, "Full-finetune/Loss (Raw)": 0.6823421120643616, "Full-finetune/Step": 3337, "Full-finetune/Step Time": 6.746837355196476} {"Full-finetune/Learning Rate": 5.994759115371531e-06, "Full-finetune/Loss": 0.7784551978111267, "Full-finetune/Loss (Raw)": 0.8003696799278259, "Full-finetune/Step": 3338, "Full-finetune/Step Time": 6.739214219152927} {"Full-finetune/Learning Rate": 5.989051786459163e-06, "Full-finetune/Loss": 0.7788019180297852, "Full-finetune/Loss (Raw)": 0.8413978219032288, "Full-finetune/Step": 3339, "Full-finetune/Step Time": 6.731569377705455} {"Full-finetune/Learning Rate": 5.98334601411749e-06, "Full-finetune/Loss": 0.7819388508796692, "Full-finetune/Loss (Raw)": 0.8028140068054199, "Full-finetune/Step": 3340, "Full-finetune/Step Time": 6.715888297185302} {"Full-finetune/Learning Rate": 5.977641800560811e-06, "Full-finetune/Loss": 0.7828361988067627, "Full-finetune/Loss (Raw)": 0.9605286121368408, "Full-finetune/Step": 3341, "Full-finetune/Step Time": 6.709449093788862} {"Full-finetune/Learning Rate": 5.971939148002819e-06, "Full-finetune/Loss": 0.7841963768005371, "Full-finetune/Loss (Raw)": 0.8758307099342346, "Full-finetune/Step": 3342, "Full-finetune/Step Time": 6.689772306010127} {"Full-finetune/Learning Rate": 5.966238058656601e-06, "Full-finetune/Loss": 0.7856837511062622, "Full-finetune/Loss (Raw)": 0.8444568514823914, "Full-finetune/Step": 3343, "Full-finetune/Step Time": 6.677200498059392} {"Full-finetune/Learning Rate": 5.960538534734641e-06, "Full-finetune/Loss": 0.7854653596878052, "Full-finetune/Loss (Raw)": 0.6231265664100647, "Full-finetune/Step": 3344, "Full-finetune/Step Time": 6.646287959069014} {"Full-finetune/Learning Rate": 5.9548405784488095e-06, "Full-finetune/Loss": 0.7819303870201111, "Full-finetune/Loss (Raw)": 0.360333651304245, "Full-finetune/Step": 3345, "Full-finetune/Step Time": 6.667435023933649} {"Full-finetune/Learning Rate": 5.949144192010378e-06, "Full-finetune/Loss": 0.782507598400116, "Full-finetune/Loss (Raw)": 0.8955385684967041, "Full-finetune/Step": 3346, "Full-finetune/Step Time": 6.66668601706624} {"Full-finetune/Learning Rate": 5.94344937763e-06, "Full-finetune/Loss": 0.7825540900230408, "Full-finetune/Loss (Raw)": 0.6858869194984436, "Full-finetune/Step": 3347, "Full-finetune/Step Time": 6.666943587362766} {"Full-finetune/Learning Rate": 5.9377561375177164e-06, "Full-finetune/Loss": 0.7817904353141785, "Full-finetune/Loss (Raw)": 0.8522974252700806, "Full-finetune/Step": 3348, "Full-finetune/Step Time": 6.6813713777810335} {"Full-finetune/Learning Rate": 5.9320644738829706e-06, "Full-finetune/Loss": 0.7825971841812134, "Full-finetune/Loss (Raw)": 0.8497845530509949, "Full-finetune/Step": 3349, "Full-finetune/Step Time": 6.677632484585047} {"Full-finetune/Learning Rate": 5.926374388934579e-06, "Full-finetune/Loss": 0.7877890467643738, "Full-finetune/Loss (Raw)": 1.000981092453003, "Full-finetune/Step": 3350, "Full-finetune/Step Time": 6.667447037994862} {"Full-finetune/Learning Rate": 5.920685884880762e-06, "Full-finetune/Loss": 0.7876068353652954, "Full-finetune/Loss (Raw)": 0.7450613379478455, "Full-finetune/Step": 3351, "Full-finetune/Step Time": 6.671907600015402} {"Full-finetune/Learning Rate": 5.914998963929109e-06, "Full-finetune/Loss": 0.7880504131317139, "Full-finetune/Loss (Raw)": 0.8489102125167847, "Full-finetune/Step": 3352, "Full-finetune/Step Time": 6.6753417737782} {"Full-finetune/Learning Rate": 5.9093136282866014e-06, "Full-finetune/Loss": 0.7890157103538513, "Full-finetune/Loss (Raw)": 0.9134607315063477, "Full-finetune/Step": 3353, "Full-finetune/Step Time": 6.6897856667637825} {"Full-finetune/Learning Rate": 5.903629880159614e-06, "Full-finetune/Loss": 0.7898149490356445, "Full-finetune/Loss (Raw)": 0.8876988291740417, "Full-finetune/Step": 3354, "Full-finetune/Step Time": 6.6774886306375265} {"Full-finetune/Learning Rate": 5.897947721753894e-06, "Full-finetune/Loss": 0.790911853313446, "Full-finetune/Loss (Raw)": 0.8329816460609436, "Full-finetune/Step": 3355, "Full-finetune/Step Time": 6.6695753168314695} {"Full-finetune/Learning Rate": 5.89226715527458e-06, "Full-finetune/Loss": 0.7917623519897461, "Full-finetune/Loss (Raw)": 0.8802149295806885, "Full-finetune/Step": 3356, "Full-finetune/Step Time": 6.670458355918527} {"Full-finetune/Learning Rate": 5.886588182926185e-06, "Full-finetune/Loss": 0.7911043167114258, "Full-finetune/Loss (Raw)": 0.8044592142105103, "Full-finetune/Step": 3357, "Full-finetune/Step Time": 6.670147048309445} {"Full-finetune/Learning Rate": 5.880910806912607e-06, "Full-finetune/Loss": 0.7913063764572144, "Full-finetune/Loss (Raw)": 0.7428300976753235, "Full-finetune/Step": 3358, "Full-finetune/Step Time": 6.664600824937224} {"Full-finetune/Learning Rate": 5.8752350294371274e-06, "Full-finetune/Loss": 0.7928191423416138, "Full-finetune/Loss (Raw)": 0.9857817888259888, "Full-finetune/Step": 3359, "Full-finetune/Step Time": 6.637870091944933} {"Full-finetune/Learning Rate": 5.8695608527024075e-06, "Full-finetune/Loss": 0.792359471321106, "Full-finetune/Loss (Raw)": 0.8816269040107727, "Full-finetune/Step": 3360, "Full-finetune/Step Time": 6.638211393728852} {"Full-finetune/Learning Rate": 5.863888278910475e-06, "Full-finetune/Loss": 0.7910017967224121, "Full-finetune/Loss (Raw)": 0.8327337503433228, "Full-finetune/Step": 3361, "Full-finetune/Step Time": 6.650390952825546} {"Full-finetune/Learning Rate": 5.8582173102627526e-06, "Full-finetune/Loss": 0.7910764813423157, "Full-finetune/Loss (Raw)": 0.7132857441902161, "Full-finetune/Step": 3362, "Full-finetune/Step Time": 6.6330993212759495} {"Full-finetune/Learning Rate": 5.852547948960029e-06, "Full-finetune/Loss": 0.7907898426055908, "Full-finetune/Loss (Raw)": 0.8889625668525696, "Full-finetune/Step": 3363, "Full-finetune/Step Time": 6.632326470687985} {"Full-finetune/Learning Rate": 5.8468801972024755e-06, "Full-finetune/Loss": 0.7926806211471558, "Full-finetune/Loss (Raw)": 0.8793284296989441, "Full-finetune/Step": 3364, "Full-finetune/Step Time": 6.625272074714303} {"Full-finetune/Learning Rate": 5.841214057189638e-06, "Full-finetune/Loss": 0.7918137907981873, "Full-finetune/Loss (Raw)": 0.7621358633041382, "Full-finetune/Step": 3365, "Full-finetune/Step Time": 6.639185881242156} {"Full-finetune/Learning Rate": 5.8355495311204266e-06, "Full-finetune/Loss": 0.793927788734436, "Full-finetune/Loss (Raw)": 0.6736167073249817, "Full-finetune/Step": 3366, "Full-finetune/Step Time": 6.632191698998213} {"Full-finetune/Learning Rate": 5.829886621193141e-06, "Full-finetune/Loss": 0.7918728590011597, "Full-finetune/Loss (Raw)": 0.7067327499389648, "Full-finetune/Step": 3367, "Full-finetune/Step Time": 6.6274845488369465} {"Full-finetune/Learning Rate": 5.824225329605442e-06, "Full-finetune/Loss": 0.7880316972732544, "Full-finetune/Loss (Raw)": 0.4441594183444977, "Full-finetune/Step": 3368, "Full-finetune/Step Time": 6.643736215308309} {"Full-finetune/Learning Rate": 5.818565658554364e-06, "Full-finetune/Loss": 0.7875732183456421, "Full-finetune/Loss (Raw)": 0.8360080122947693, "Full-finetune/Step": 3369, "Full-finetune/Step Time": 6.642666952684522} {"Full-finetune/Learning Rate": 5.8129076102363266e-06, "Full-finetune/Loss": 0.7870449423789978, "Full-finetune/Loss (Raw)": 0.8641881942749023, "Full-finetune/Step": 3370, "Full-finetune/Step Time": 6.6400357112288475} {"Full-finetune/Learning Rate": 5.807251186847095e-06, "Full-finetune/Loss": 0.7848662734031677, "Full-finetune/Loss (Raw)": 0.5588309168815613, "Full-finetune/Step": 3371, "Full-finetune/Step Time": 6.661945017054677} {"Full-finetune/Learning Rate": 5.801596390581817e-06, "Full-finetune/Loss": 0.7840461730957031, "Full-finetune/Loss (Raw)": 0.739946722984314, "Full-finetune/Step": 3372, "Full-finetune/Step Time": 6.672693280503154} {"Full-finetune/Learning Rate": 5.795943223635018e-06, "Full-finetune/Loss": 0.7837324142456055, "Full-finetune/Loss (Raw)": 0.8412212133407593, "Full-finetune/Step": 3373, "Full-finetune/Step Time": 6.676909759640694} {"Full-finetune/Learning Rate": 5.7902916882005755e-06, "Full-finetune/Loss": 0.7847588062286377, "Full-finetune/Loss (Raw)": 0.8125916719436646, "Full-finetune/Step": 3374, "Full-finetune/Step Time": 6.6800232250243425} {"Full-finetune/Learning Rate": 5.784641786471742e-06, "Full-finetune/Loss": 0.7871368527412415, "Full-finetune/Loss (Raw)": 1.0461915731430054, "Full-finetune/Step": 3375, "Full-finetune/Step Time": 6.6873776745051146} {"Full-finetune/Learning Rate": 5.778993520641132e-06, "Full-finetune/Loss": 0.7870060801506042, "Full-finetune/Loss (Raw)": 0.9288752675056458, "Full-finetune/Step": 3376, "Full-finetune/Step Time": 6.679686903953552} {"Full-finetune/Learning Rate": 5.7733468929007294e-06, "Full-finetune/Loss": 0.7845607995986938, "Full-finetune/Loss (Raw)": 0.6357234716415405, "Full-finetune/Step": 3377, "Full-finetune/Step Time": 6.689688324928284} {"Full-finetune/Learning Rate": 5.7677019054418805e-06, "Full-finetune/Loss": 0.7856536507606506, "Full-finetune/Loss (Raw)": 0.895717442035675, "Full-finetune/Step": 3378, "Full-finetune/Step Time": 6.694651361554861} {"Full-finetune/Learning Rate": 5.762058560455287e-06, "Full-finetune/Loss": 0.7853163480758667, "Full-finetune/Loss (Raw)": 0.614189863204956, "Full-finetune/Step": 3379, "Full-finetune/Step Time": 6.6689733024686575} {"Full-finetune/Learning Rate": 5.756416860131036e-06, "Full-finetune/Loss": 0.7838926315307617, "Full-finetune/Loss (Raw)": 0.7030613422393799, "Full-finetune/Step": 3380, "Full-finetune/Step Time": 6.664606431499124} {"Full-finetune/Learning Rate": 5.750776806658549e-06, "Full-finetune/Loss": 0.7832397818565369, "Full-finetune/Loss (Raw)": 0.7489619851112366, "Full-finetune/Step": 3381, "Full-finetune/Step Time": 6.659699192270637} {"Full-finetune/Learning Rate": 5.745138402226621e-06, "Full-finetune/Loss": 0.7833285331726074, "Full-finetune/Loss (Raw)": 0.8818806409835815, "Full-finetune/Step": 3382, "Full-finetune/Step Time": 6.660773178562522} {"Full-finetune/Learning Rate": 5.739501649023413e-06, "Full-finetune/Loss": 0.7826509475708008, "Full-finetune/Loss (Raw)": 0.7865077257156372, "Full-finetune/Step": 3383, "Full-finetune/Step Time": 6.658173410221934} {"Full-finetune/Learning Rate": 5.7338665492364365e-06, "Full-finetune/Loss": 0.7827873229980469, "Full-finetune/Loss (Raw)": 0.887344241142273, "Full-finetune/Step": 3384, "Full-finetune/Step Time": 6.653068332001567} {"Full-finetune/Learning Rate": 5.728233105052563e-06, "Full-finetune/Loss": 0.7830116748809814, "Full-finetune/Loss (Raw)": 0.6517096757888794, "Full-finetune/Step": 3385, "Full-finetune/Step Time": 6.641925126314163} {"Full-finetune/Learning Rate": 5.722601318658021e-06, "Full-finetune/Loss": 0.7812755107879639, "Full-finetune/Loss (Raw)": 0.7886253595352173, "Full-finetune/Step": 3386, "Full-finetune/Step Time": 6.632577503100038} {"Full-finetune/Learning Rate": 5.7169711922383984e-06, "Full-finetune/Loss": 0.7813201546669006, "Full-finetune/Loss (Raw)": 0.7149817943572998, "Full-finetune/Step": 3387, "Full-finetune/Step Time": 6.620687063783407} {"Full-finetune/Learning Rate": 5.711342727978637e-06, "Full-finetune/Loss": 0.7808976173400879, "Full-finetune/Loss (Raw)": 0.7663537263870239, "Full-finetune/Step": 3388, "Full-finetune/Step Time": 6.601996732875705} {"Full-finetune/Learning Rate": 5.7057159280630316e-06, "Full-finetune/Loss": 0.7814376354217529, "Full-finetune/Loss (Raw)": 0.8183497786521912, "Full-finetune/Step": 3389, "Full-finetune/Step Time": 6.60601133108139} {"Full-finetune/Learning Rate": 5.700090794675241e-06, "Full-finetune/Loss": 0.7807773351669312, "Full-finetune/Loss (Raw)": 0.8328700661659241, "Full-finetune/Step": 3390, "Full-finetune/Step Time": 6.6093388963490725} {"Full-finetune/Learning Rate": 5.694467329998264e-06, "Full-finetune/Loss": 0.780961275100708, "Full-finetune/Loss (Raw)": 0.9629249572753906, "Full-finetune/Step": 3391, "Full-finetune/Step Time": 6.604645328596234} {"Full-finetune/Learning Rate": 5.68884553621445e-06, "Full-finetune/Loss": 0.7812793254852295, "Full-finetune/Loss (Raw)": 0.7559467554092407, "Full-finetune/Step": 3392, "Full-finetune/Step Time": 6.603177530691028} {"Full-finetune/Learning Rate": 5.683225415505521e-06, "Full-finetune/Loss": 0.7808248996734619, "Full-finetune/Loss (Raw)": 0.89690762758255, "Full-finetune/Step": 3393, "Full-finetune/Step Time": 6.588441390544176} {"Full-finetune/Learning Rate": 5.67760697005253e-06, "Full-finetune/Loss": 0.7804097533226013, "Full-finetune/Loss (Raw)": 0.7399923801422119, "Full-finetune/Step": 3394, "Full-finetune/Step Time": 6.597303248941898} {"Full-finetune/Learning Rate": 5.671990202035886e-06, "Full-finetune/Loss": 0.7816885113716125, "Full-finetune/Loss (Raw)": 0.8847301006317139, "Full-finetune/Step": 3395, "Full-finetune/Step Time": 6.580435456708074} {"Full-finetune/Learning Rate": 5.666375113635347e-06, "Full-finetune/Loss": 0.7799391746520996, "Full-finetune/Loss (Raw)": 0.6657758355140686, "Full-finetune/Step": 3396, "Full-finetune/Step Time": 6.579829728230834} {"Full-finetune/Learning Rate": 5.660761707030022e-06, "Full-finetune/Loss": 0.7790262699127197, "Full-finetune/Loss (Raw)": 0.762270987033844, "Full-finetune/Step": 3397, "Full-finetune/Step Time": 6.576920228078961} {"Full-finetune/Learning Rate": 5.655149984398359e-06, "Full-finetune/Loss": 0.7796502113342285, "Full-finetune/Loss (Raw)": 0.8200222849845886, "Full-finetune/Step": 3398, "Full-finetune/Step Time": 6.575363345444202} {"Full-finetune/Learning Rate": 5.649539947918159e-06, "Full-finetune/Loss": 0.7795863747596741, "Full-finetune/Loss (Raw)": 0.7693026065826416, "Full-finetune/Step": 3399, "Full-finetune/Step Time": 6.594960242509842} {"Full-finetune/Learning Rate": 5.643931599766577e-06, "Full-finetune/Loss": 0.7796302437782288, "Full-finetune/Loss (Raw)": 0.7707681059837341, "Full-finetune/Step": 3400, "Full-finetune/Step Time": 6.567662326619029} {"Full-finetune/Learning Rate": 5.638324942120091e-06, "Full-finetune/Loss": 0.7795140743255615, "Full-finetune/Loss (Raw)": 0.7188820838928223, "Full-finetune/Step": 3401, "Full-finetune/Step Time": 6.559616554528475} {"Full-finetune/Learning Rate": 5.632719977154537e-06, "Full-finetune/Loss": 0.7830754518508911, "Full-finetune/Loss (Raw)": 0.7806788086891174, "Full-finetune/Step": 3402, "Full-finetune/Step Time": 6.5417449455708265} {"Full-finetune/Learning Rate": 5.6271167070451016e-06, "Full-finetune/Loss": 0.7832871675491333, "Full-finetune/Loss (Raw)": 0.747061550617218, "Full-finetune/Step": 3403, "Full-finetune/Step Time": 6.5345166977494955} {"Full-finetune/Learning Rate": 5.621515133966301e-06, "Full-finetune/Loss": 0.7827012538909912, "Full-finetune/Loss (Raw)": 0.7584566473960876, "Full-finetune/Step": 3404, "Full-finetune/Step Time": 6.531687363982201} {"Full-finetune/Learning Rate": 5.615915260091987e-06, "Full-finetune/Loss": 0.7808048725128174, "Full-finetune/Loss (Raw)": 0.5802109837532043, "Full-finetune/Step": 3405, "Full-finetune/Step Time": 6.570953143760562} {"Full-finetune/Learning Rate": 5.610317087595372e-06, "Full-finetune/Loss": 0.7812771797180176, "Full-finetune/Loss (Raw)": 0.8543359637260437, "Full-finetune/Step": 3406, "Full-finetune/Step Time": 6.567206505686045} {"Full-finetune/Learning Rate": 5.604720618648993e-06, "Full-finetune/Loss": 0.7812418937683105, "Full-finetune/Loss (Raw)": 0.826568067073822, "Full-finetune/Step": 3407, "Full-finetune/Step Time": 6.555506559088826} {"Full-finetune/Learning Rate": 5.599125855424734e-06, "Full-finetune/Loss": 0.7805944681167603, "Full-finetune/Loss (Raw)": 0.8461728096008301, "Full-finetune/Step": 3408, "Full-finetune/Step Time": 6.555442562326789} {"Full-finetune/Learning Rate": 5.593532800093808e-06, "Full-finetune/Loss": 0.7794376611709595, "Full-finetune/Loss (Raw)": 0.832111120223999, "Full-finetune/Step": 3409, "Full-finetune/Step Time": 6.553197691217065} {"Full-finetune/Learning Rate": 5.5879414548267754e-06, "Full-finetune/Loss": 0.7788301110267639, "Full-finetune/Loss (Raw)": 0.6980586051940918, "Full-finetune/Step": 3410, "Full-finetune/Step Time": 6.556469457224011} {"Full-finetune/Learning Rate": 5.582351821793525e-06, "Full-finetune/Loss": 0.7832949757575989, "Full-finetune/Loss (Raw)": 0.9457889199256897, "Full-finetune/Step": 3411, "Full-finetune/Step Time": 6.538990056142211} {"Full-finetune/Learning Rate": 5.576763903163281e-06, "Full-finetune/Loss": 0.7839521169662476, "Full-finetune/Loss (Raw)": 0.8369607925415039, "Full-finetune/Step": 3412, "Full-finetune/Step Time": 6.5234490521252155} {"Full-finetune/Learning Rate": 5.571177701104614e-06, "Full-finetune/Loss": 0.7844966053962708, "Full-finetune/Loss (Raw)": 0.9226974844932556, "Full-finetune/Step": 3413, "Full-finetune/Step Time": 6.531191585585475} {"Full-finetune/Learning Rate": 5.56559321778542e-06, "Full-finetune/Loss": 0.7854311466217041, "Full-finetune/Loss (Raw)": 0.9179232716560364, "Full-finetune/Step": 3414, "Full-finetune/Step Time": 6.528691524639726} {"Full-finetune/Learning Rate": 5.560010455372915e-06, "Full-finetune/Loss": 0.7840756177902222, "Full-finetune/Loss (Raw)": 0.6576681137084961, "Full-finetune/Step": 3415, "Full-finetune/Step Time": 6.526291437447071} {"Full-finetune/Learning Rate": 5.5544294160336735e-06, "Full-finetune/Loss": 0.7802925109863281, "Full-finetune/Loss (Raw)": 0.43241918087005615, "Full-finetune/Step": 3416, "Full-finetune/Step Time": 6.541691733524203} {"Full-finetune/Learning Rate": 5.5488501019335824e-06, "Full-finetune/Loss": 0.7800270318984985, "Full-finetune/Loss (Raw)": 0.715147852897644, "Full-finetune/Step": 3417, "Full-finetune/Step Time": 6.5496388375759125} {"Full-finetune/Learning Rate": 5.5432725152378675e-06, "Full-finetune/Loss": 0.7798172235488892, "Full-finetune/Loss (Raw)": 0.8218523859977722, "Full-finetune/Step": 3418, "Full-finetune/Step Time": 6.561194596812129} {"Full-finetune/Learning Rate": 5.537696658111077e-06, "Full-finetune/Loss": 0.7809090614318848, "Full-finetune/Loss (Raw)": 0.8181442618370056, "Full-finetune/Step": 3419, "Full-finetune/Step Time": 6.553378758952022} {"Full-finetune/Learning Rate": 5.532122532717095e-06, "Full-finetune/Loss": 0.7790800333023071, "Full-finetune/Loss (Raw)": 0.5227681398391724, "Full-finetune/Step": 3420, "Full-finetune/Step Time": 6.578799081966281} {"Full-finetune/Learning Rate": 5.52655014121913e-06, "Full-finetune/Loss": 0.7794040441513062, "Full-finetune/Loss (Raw)": 0.8593570590019226, "Full-finetune/Step": 3421, "Full-finetune/Step Time": 6.58091045729816} {"Full-finetune/Learning Rate": 5.5209794857797135e-06, "Full-finetune/Loss": 0.7790420651435852, "Full-finetune/Loss (Raw)": 0.6695350408554077, "Full-finetune/Step": 3422, "Full-finetune/Step Time": 6.57415902800858} {"Full-finetune/Learning Rate": 5.51541056856072e-06, "Full-finetune/Loss": 0.7788878083229065, "Full-finetune/Loss (Raw)": 0.7684012651443481, "Full-finetune/Step": 3423, "Full-finetune/Step Time": 6.572680598124862} {"Full-finetune/Learning Rate": 5.509843391723332e-06, "Full-finetune/Loss": 0.7751755714416504, "Full-finetune/Loss (Raw)": 0.3592769503593445, "Full-finetune/Step": 3424, "Full-finetune/Step Time": 6.584687160328031} {"Full-finetune/Learning Rate": 5.504277957428053e-06, "Full-finetune/Loss": 0.7758747935295105, "Full-finetune/Loss (Raw)": 0.765912652015686, "Full-finetune/Step": 3425, "Full-finetune/Step Time": 6.573238598182797} {"Full-finetune/Learning Rate": 5.498714267834731e-06, "Full-finetune/Loss": 0.7739516496658325, "Full-finetune/Loss (Raw)": 0.6537418961524963, "Full-finetune/Step": 3426, "Full-finetune/Step Time": 6.574177028611302} {"Full-finetune/Learning Rate": 5.493152325102523e-06, "Full-finetune/Loss": 0.7774567008018494, "Full-finetune/Loss (Raw)": 0.8430326581001282, "Full-finetune/Step": 3427, "Full-finetune/Step Time": 6.5610585659742355} {"Full-finetune/Learning Rate": 5.487592131389907e-06, "Full-finetune/Loss": 0.7765512466430664, "Full-finetune/Loss (Raw)": 0.7503271698951721, "Full-finetune/Step": 3428, "Full-finetune/Step Time": 6.567335274070501} {"Full-finetune/Learning Rate": 5.482033688854689e-06, "Full-finetune/Loss": 0.7779567241668701, "Full-finetune/Loss (Raw)": 0.8486117720603943, "Full-finetune/Step": 3429, "Full-finetune/Step Time": 6.564101783558726} {"Full-finetune/Learning Rate": 5.47647699965399e-06, "Full-finetune/Loss": 0.7758962512016296, "Full-finetune/Loss (Raw)": 0.6417986154556274, "Full-finetune/Step": 3430, "Full-finetune/Step Time": 6.558290760964155} {"Full-finetune/Learning Rate": 5.470922065944253e-06, "Full-finetune/Loss": 0.7770318984985352, "Full-finetune/Loss (Raw)": 0.9266951680183411, "Full-finetune/Step": 3431, "Full-finetune/Step Time": 6.562914686277509} {"Full-finetune/Learning Rate": 5.465368889881235e-06, "Full-finetune/Loss": 0.7764268517494202, "Full-finetune/Loss (Raw)": 0.6479002237319946, "Full-finetune/Step": 3432, "Full-finetune/Step Time": 6.575011612847447} {"Full-finetune/Learning Rate": 5.459817473620024e-06, "Full-finetune/Loss": 0.7758225202560425, "Full-finetune/Loss (Raw)": 0.7882454991340637, "Full-finetune/Step": 3433, "Full-finetune/Step Time": 6.59194471500814} {"Full-finetune/Learning Rate": 5.454267819315016e-06, "Full-finetune/Loss": 0.7755783796310425, "Full-finetune/Loss (Raw)": 0.683968722820282, "Full-finetune/Step": 3434, "Full-finetune/Step Time": 6.626610057428479} {"Full-finetune/Learning Rate": 5.448719929119916e-06, "Full-finetune/Loss": 0.7739091515541077, "Full-finetune/Loss (Raw)": 0.7318502068519592, "Full-finetune/Step": 3435, "Full-finetune/Step Time": 6.639454551041126} {"Full-finetune/Learning Rate": 5.44317380518775e-06, "Full-finetune/Loss": 0.7729480862617493, "Full-finetune/Loss (Raw)": 0.7035651206970215, "Full-finetune/Step": 3436, "Full-finetune/Step Time": 6.633539045229554} {"Full-finetune/Learning Rate": 5.43762944967087e-06, "Full-finetune/Loss": 0.7723343372344971, "Full-finetune/Loss (Raw)": 0.798020601272583, "Full-finetune/Step": 3437, "Full-finetune/Step Time": 6.622722806409001} {"Full-finetune/Learning Rate": 5.432086864720929e-06, "Full-finetune/Loss": 0.7719689607620239, "Full-finetune/Loss (Raw)": 0.6978741884231567, "Full-finetune/Step": 3438, "Full-finetune/Step Time": 6.632403377443552} {"Full-finetune/Learning Rate": 5.426546052488892e-06, "Full-finetune/Loss": 0.772770881652832, "Full-finetune/Loss (Raw)": 0.8113725185394287, "Full-finetune/Step": 3439, "Full-finetune/Step Time": 6.638190321624279} {"Full-finetune/Learning Rate": 5.421007015125046e-06, "Full-finetune/Loss": 0.7722389698028564, "Full-finetune/Loss (Raw)": 0.7122519016265869, "Full-finetune/Step": 3440, "Full-finetune/Step Time": 6.640121202915907} {"Full-finetune/Learning Rate": 5.415469754778979e-06, "Full-finetune/Loss": 0.7737387418746948, "Full-finetune/Loss (Raw)": 0.62242192029953, "Full-finetune/Step": 3441, "Full-finetune/Step Time": 6.644616512581706} {"Full-finetune/Learning Rate": 5.409934273599594e-06, "Full-finetune/Loss": 0.7737083435058594, "Full-finetune/Loss (Raw)": 0.7299603223800659, "Full-finetune/Step": 3442, "Full-finetune/Step Time": 6.650618935003877} {"Full-finetune/Learning Rate": 5.404400573735105e-06, "Full-finetune/Loss": 0.7715499401092529, "Full-finetune/Loss (Raw)": 0.7066214084625244, "Full-finetune/Step": 3443, "Full-finetune/Step Time": 6.659974716603756} {"Full-finetune/Learning Rate": 5.3988686573330316e-06, "Full-finetune/Loss": 0.7720843553543091, "Full-finetune/Loss (Raw)": 0.8192605972290039, "Full-finetune/Step": 3444, "Full-finetune/Step Time": 6.653490858152509} {"Full-finetune/Learning Rate": 5.393338526540202e-06, "Full-finetune/Loss": 0.7707411050796509, "Full-finetune/Loss (Raw)": 0.5762043595314026, "Full-finetune/Step": 3445, "Full-finetune/Step Time": 6.671967271715403} {"Full-finetune/Learning Rate": 5.387810183502749e-06, "Full-finetune/Loss": 0.7738832235336304, "Full-finetune/Loss (Raw)": 0.7449252009391785, "Full-finetune/Step": 3446, "Full-finetune/Step Time": 6.663898115977645} {"Full-finetune/Learning Rate": 5.382283630366123e-06, "Full-finetune/Loss": 0.7739137411117554, "Full-finetune/Loss (Raw)": 0.897596001625061, "Full-finetune/Step": 3447, "Full-finetune/Step Time": 6.65816661156714} {"Full-finetune/Learning Rate": 5.37675886927507e-06, "Full-finetune/Loss": 0.7736634016036987, "Full-finetune/Loss (Raw)": 0.7347889542579651, "Full-finetune/Step": 3448, "Full-finetune/Step Time": 6.6638513300567865} {"Full-finetune/Learning Rate": 5.371235902373632e-06, "Full-finetune/Loss": 0.7751492857933044, "Full-finetune/Loss (Raw)": 0.8975018262863159, "Full-finetune/Step": 3449, "Full-finetune/Step Time": 6.660388007760048} {"Full-finetune/Learning Rate": 5.365714731805174e-06, "Full-finetune/Loss": 0.7752164602279663, "Full-finetune/Loss (Raw)": 0.7713738083839417, "Full-finetune/Step": 3450, "Full-finetune/Step Time": 6.681194482371211} {"Full-finetune/Learning Rate": 5.360195359712353e-06, "Full-finetune/Loss": 0.7742554545402527, "Full-finetune/Loss (Raw)": 0.7799215912818909, "Full-finetune/Step": 3451, "Full-finetune/Step Time": 6.644434817135334} {"Full-finetune/Learning Rate": 5.354677788237126e-06, "Full-finetune/Loss": 0.7751296162605286, "Full-finetune/Loss (Raw)": 0.8741554617881775, "Full-finetune/Step": 3452, "Full-finetune/Step Time": 6.619277121499181} {"Full-finetune/Learning Rate": 5.349162019520759e-06, "Full-finetune/Loss": 0.7759745121002197, "Full-finetune/Loss (Raw)": 0.7927504181861877, "Full-finetune/Step": 3453, "Full-finetune/Step Time": 6.617847494781017} {"Full-finetune/Learning Rate": 5.343648055703812e-06, "Full-finetune/Loss": 0.7750028371810913, "Full-finetune/Loss (Raw)": 0.7499936819076538, "Full-finetune/Step": 3454, "Full-finetune/Step Time": 6.636305006220937} {"Full-finetune/Learning Rate": 5.338135898926147e-06, "Full-finetune/Loss": 0.7731986045837402, "Full-finetune/Loss (Raw)": 0.34284836053848267, "Full-finetune/Step": 3455, "Full-finetune/Step Time": 6.64465151540935} {"Full-finetune/Learning Rate": 5.33262555132692e-06, "Full-finetune/Loss": 0.772691011428833, "Full-finetune/Loss (Raw)": 0.8612234592437744, "Full-finetune/Step": 3456, "Full-finetune/Step Time": 6.642961032688618} {"Full-finetune/Learning Rate": 5.327117015044599e-06, "Full-finetune/Loss": 0.7730093002319336, "Full-finetune/Loss (Raw)": 0.8508954048156738, "Full-finetune/Step": 3457, "Full-finetune/Step Time": 6.634192392230034} {"Full-finetune/Learning Rate": 5.321610292216937e-06, "Full-finetune/Loss": 0.7739236354827881, "Full-finetune/Loss (Raw)": 0.8128558993339539, "Full-finetune/Step": 3458, "Full-finetune/Step Time": 6.643150458112359} {"Full-finetune/Learning Rate": 5.316105384980974e-06, "Full-finetune/Loss": 0.7739429473876953, "Full-finetune/Loss (Raw)": 0.8115407824516296, "Full-finetune/Step": 3459, "Full-finetune/Step Time": 6.6340127270668745} {"Full-finetune/Learning Rate": 5.3106022954730704e-06, "Full-finetune/Loss": 0.7747448682785034, "Full-finetune/Loss (Raw)": 0.8930315971374512, "Full-finetune/Step": 3460, "Full-finetune/Step Time": 6.636074062436819} {"Full-finetune/Learning Rate": 5.3051010258288624e-06, "Full-finetune/Loss": 0.7710564136505127, "Full-finetune/Loss (Raw)": 0.34597083926200867, "Full-finetune/Step": 3461, "Full-finetune/Step Time": 6.654298035427928} {"Full-finetune/Learning Rate": 5.299601578183289e-06, "Full-finetune/Loss": 0.7709078788757324, "Full-finetune/Loss (Raw)": 0.7976961731910706, "Full-finetune/Step": 3462, "Full-finetune/Step Time": 6.652631217613816} {"Full-finetune/Learning Rate": 5.294103954670573e-06, "Full-finetune/Loss": 0.7712231278419495, "Full-finetune/Loss (Raw)": 0.7623317837715149, "Full-finetune/Step": 3463, "Full-finetune/Step Time": 6.655834745615721} {"Full-finetune/Learning Rate": 5.288608157424238e-06, "Full-finetune/Loss": 0.7723056077957153, "Full-finetune/Loss (Raw)": 0.9011552929878235, "Full-finetune/Step": 3464, "Full-finetune/Step Time": 6.656326923519373} {"Full-finetune/Learning Rate": 5.283114188577094e-06, "Full-finetune/Loss": 0.7726401090621948, "Full-finetune/Loss (Raw)": 0.7251603007316589, "Full-finetune/Step": 3465, "Full-finetune/Step Time": 6.6524353697896} {"Full-finetune/Learning Rate": 5.277622050261242e-06, "Full-finetune/Loss": 0.771693766117096, "Full-finetune/Loss (Raw)": 0.6792414784431458, "Full-finetune/Step": 3466, "Full-finetune/Step Time": 6.65157312899828} {"Full-finetune/Learning Rate": 5.272131744608078e-06, "Full-finetune/Loss": 0.7702020406723022, "Full-finetune/Loss (Raw)": 0.6504579186439514, "Full-finetune/Step": 3467, "Full-finetune/Step Time": 6.65070678666234} {"Full-finetune/Learning Rate": 5.2666432737482844e-06, "Full-finetune/Loss": 0.7689626216888428, "Full-finetune/Loss (Raw)": 0.6441641449928284, "Full-finetune/Step": 3468, "Full-finetune/Step Time": 6.659159662202001} {"Full-finetune/Learning Rate": 5.261156639811817e-06, "Full-finetune/Loss": 0.7669690847396851, "Full-finetune/Loss (Raw)": 0.7053518295288086, "Full-finetune/Step": 3469, "Full-finetune/Step Time": 6.6679572351276875} {"Full-finetune/Learning Rate": 5.255671844927944e-06, "Full-finetune/Loss": 0.7661389112472534, "Full-finetune/Loss (Raw)": 0.7695721983909607, "Full-finetune/Step": 3470, "Full-finetune/Step Time": 6.676040066406131} {"Full-finetune/Learning Rate": 5.250188891225203e-06, "Full-finetune/Loss": 0.766289472579956, "Full-finetune/Loss (Raw)": 0.8637343645095825, "Full-finetune/Step": 3471, "Full-finetune/Step Time": 6.682092444971204} {"Full-finetune/Learning Rate": 5.24470778083142e-06, "Full-finetune/Loss": 0.7671939730644226, "Full-finetune/Loss (Raw)": 0.7388960719108582, "Full-finetune/Step": 3472, "Full-finetune/Step Time": 6.683562338352203} {"Full-finetune/Learning Rate": 5.2392285158737085e-06, "Full-finetune/Loss": 0.7710477113723755, "Full-finetune/Loss (Raw)": 0.8536192178726196, "Full-finetune/Step": 3473, "Full-finetune/Step Time": 6.686075011268258} {"Full-finetune/Learning Rate": 5.233751098478465e-06, "Full-finetune/Loss": 0.7692318558692932, "Full-finetune/Loss (Raw)": 0.6631056666374207, "Full-finetune/Step": 3474, "Full-finetune/Step Time": 6.691004231572151} {"Full-finetune/Learning Rate": 5.228275530771366e-06, "Full-finetune/Loss": 0.7701634168624878, "Full-finetune/Loss (Raw)": 0.8051283955574036, "Full-finetune/Step": 3475, "Full-finetune/Step Time": 6.698870979249477} {"Full-finetune/Learning Rate": 5.22280181487737e-06, "Full-finetune/Loss": 0.7683970928192139, "Full-finetune/Loss (Raw)": 0.6262004375457764, "Full-finetune/Step": 3476, "Full-finetune/Step Time": 6.6934839598834515} {"Full-finetune/Learning Rate": 5.217329952920727e-06, "Full-finetune/Loss": 0.7674708962440491, "Full-finetune/Loss (Raw)": 0.7312336564064026, "Full-finetune/Step": 3477, "Full-finetune/Step Time": 6.711977718397975} {"Full-finetune/Learning Rate": 5.2118599470249575e-06, "Full-finetune/Loss": 0.7669754028320312, "Full-finetune/Loss (Raw)": 0.9375664591789246, "Full-finetune/Step": 3478, "Full-finetune/Step Time": 6.70516268350184} {"Full-finetune/Learning Rate": 5.206391799312855e-06, "Full-finetune/Loss": 0.769128680229187, "Full-finetune/Loss (Raw)": 1.0206754207611084, "Full-finetune/Step": 3479, "Full-finetune/Step Time": 6.694054294377565} {"Full-finetune/Learning Rate": 5.200925511906513e-06, "Full-finetune/Loss": 0.7690623998641968, "Full-finetune/Loss (Raw)": 0.8404272794723511, "Full-finetune/Step": 3480, "Full-finetune/Step Time": 6.6858404986560345} {"Full-finetune/Learning Rate": 5.195461086927284e-06, "Full-finetune/Loss": 0.7685815095901489, "Full-finetune/Loss (Raw)": 0.8519085645675659, "Full-finetune/Step": 3481, "Full-finetune/Step Time": 6.664602579548955} {"Full-finetune/Learning Rate": 5.189998526495808e-06, "Full-finetune/Loss": 0.7675307989120483, "Full-finetune/Loss (Raw)": 0.7532009482383728, "Full-finetune/Step": 3482, "Full-finetune/Step Time": 6.66544939763844} {"Full-finetune/Learning Rate": 5.184537832731996e-06, "Full-finetune/Loss": 0.7667220830917358, "Full-finetune/Loss (Raw)": 0.729465901851654, "Full-finetune/Step": 3483, "Full-finetune/Step Time": 6.700823422521353} {"Full-finetune/Learning Rate": 5.1790790077550366e-06, "Full-finetune/Loss": 0.7658786773681641, "Full-finetune/Loss (Raw)": 0.7722586393356323, "Full-finetune/Step": 3484, "Full-finetune/Step Time": 6.7184605449438095} {"Full-finetune/Learning Rate": 5.173622053683395e-06, "Full-finetune/Loss": 0.7668725848197937, "Full-finetune/Loss (Raw)": 0.9316831827163696, "Full-finetune/Step": 3485, "Full-finetune/Step Time": 6.716250428929925} {"Full-finetune/Learning Rate": 5.168166972634803e-06, "Full-finetune/Loss": 0.7672691941261292, "Full-finetune/Loss (Raw)": 0.7935969233512878, "Full-finetune/Step": 3486, "Full-finetune/Step Time": 6.711014745756984} {"Full-finetune/Learning Rate": 5.1627137667262815e-06, "Full-finetune/Loss": 0.7651605010032654, "Full-finetune/Loss (Raw)": 0.7158676981925964, "Full-finetune/Step": 3487, "Full-finetune/Step Time": 6.710592612624168} {"Full-finetune/Learning Rate": 5.1572624380741044e-06, "Full-finetune/Loss": 0.7636129260063171, "Full-finetune/Loss (Raw)": 0.6835323572158813, "Full-finetune/Step": 3488, "Full-finetune/Step Time": 6.721124308183789} {"Full-finetune/Learning Rate": 5.151812988793825e-06, "Full-finetune/Loss": 0.7639256715774536, "Full-finetune/Loss (Raw)": 0.8727746605873108, "Full-finetune/Step": 3489, "Full-finetune/Step Time": 6.709384258836508} {"Full-finetune/Learning Rate": 5.146365421000275e-06, "Full-finetune/Loss": 0.7639032602310181, "Full-finetune/Loss (Raw)": 0.7104171514511108, "Full-finetune/Step": 3490, "Full-finetune/Step Time": 6.718667592853308} {"Full-finetune/Learning Rate": 5.140919736807546e-06, "Full-finetune/Loss": 0.764089822769165, "Full-finetune/Loss (Raw)": 0.9128384590148926, "Full-finetune/Step": 3491, "Full-finetune/Step Time": 6.714499434456229} {"Full-finetune/Learning Rate": 5.135475938329002e-06, "Full-finetune/Loss": 0.7634812593460083, "Full-finetune/Loss (Raw)": 0.801430881023407, "Full-finetune/Step": 3492, "Full-finetune/Step Time": 6.724315317347646} {"Full-finetune/Learning Rate": 5.1300340276772756e-06, "Full-finetune/Loss": 0.7612817287445068, "Full-finetune/Loss (Raw)": 0.4805958867073059, "Full-finetune/Step": 3493, "Full-finetune/Step Time": 6.716454165056348} {"Full-finetune/Learning Rate": 5.124594006964263e-06, "Full-finetune/Loss": 0.7633235454559326, "Full-finetune/Loss (Raw)": 0.9349666833877563, "Full-finetune/Step": 3494, "Full-finetune/Step Time": 6.704565566033125} {"Full-finetune/Learning Rate": 5.1191558783011315e-06, "Full-finetune/Loss": 0.7644274234771729, "Full-finetune/Loss (Raw)": 0.8480334281921387, "Full-finetune/Step": 3495, "Full-finetune/Step Time": 6.704213064163923} {"Full-finetune/Learning Rate": 5.113719643798312e-06, "Full-finetune/Loss": 0.7677168846130371, "Full-finetune/Loss (Raw)": 0.8652141690254211, "Full-finetune/Step": 3496, "Full-finetune/Step Time": 6.694866733625531} {"Full-finetune/Learning Rate": 5.108285305565508e-06, "Full-finetune/Loss": 0.766098141670227, "Full-finetune/Loss (Raw)": 0.6288089752197266, "Full-finetune/Step": 3497, "Full-finetune/Step Time": 6.70255421474576} {"Full-finetune/Learning Rate": 5.1028528657116715e-06, "Full-finetune/Loss": 0.7661845088005066, "Full-finetune/Loss (Raw)": 0.8752370476722717, "Full-finetune/Step": 3498, "Full-finetune/Step Time": 6.706398257985711} {"Full-finetune/Learning Rate": 5.097422326345024e-06, "Full-finetune/Loss": 0.767928957939148, "Full-finetune/Loss (Raw)": 0.7821218967437744, "Full-finetune/Step": 3499, "Full-finetune/Step Time": 6.692671438679099} {"Full-finetune/Learning Rate": 5.091993689573062e-06, "Full-finetune/Loss": 0.7691220045089722, "Full-finetune/Loss (Raw)": 0.8926610350608826, "Full-finetune/Step": 3500, "Full-finetune/Step Time": 6.682623257860541} {"Full-finetune/Learning Rate": 5.086566957502527e-06, "Full-finetune/Loss": 0.7692621946334839, "Full-finetune/Loss (Raw)": 0.8591609597206116, "Full-finetune/Step": 3501, "Full-finetune/Step Time": 6.668919438496232} {"Full-finetune/Learning Rate": 5.081142132239429e-06, "Full-finetune/Loss": 0.7702517509460449, "Full-finetune/Loss (Raw)": 0.9392552971839905, "Full-finetune/Step": 3502, "Full-finetune/Step Time": 6.676299799233675} {"Full-finetune/Learning Rate": 5.075719215889039e-06, "Full-finetune/Loss": 0.7688825130462646, "Full-finetune/Loss (Raw)": 0.8709323406219482, "Full-finetune/Step": 3503, "Full-finetune/Step Time": 6.666257670149207} {"Full-finetune/Learning Rate": 5.070298210555881e-06, "Full-finetune/Loss": 0.7675356864929199, "Full-finetune/Loss (Raw)": 0.7564777731895447, "Full-finetune/Step": 3504, "Full-finetune/Step Time": 6.667281934991479} {"Full-finetune/Learning Rate": 5.064879118343744e-06, "Full-finetune/Loss": 0.7685633897781372, "Full-finetune/Loss (Raw)": 0.7672726511955261, "Full-finetune/Step": 3505, "Full-finetune/Step Time": 6.664325164631009} {"Full-finetune/Learning Rate": 5.059461941355665e-06, "Full-finetune/Loss": 0.7672091722488403, "Full-finetune/Loss (Raw)": 0.7223829627037048, "Full-finetune/Step": 3506, "Full-finetune/Step Time": 6.650952562689781} {"Full-finetune/Learning Rate": 5.05404668169396e-06, "Full-finetune/Loss": 0.7670596837997437, "Full-finetune/Loss (Raw)": 0.5950468182563782, "Full-finetune/Step": 3507, "Full-finetune/Step Time": 6.665778623893857} {"Full-finetune/Learning Rate": 5.048633341460174e-06, "Full-finetune/Loss": 0.7677392363548279, "Full-finetune/Loss (Raw)": 0.7900431156158447, "Full-finetune/Step": 3508, "Full-finetune/Step Time": 6.66720730997622} {"Full-finetune/Learning Rate": 5.043221922755115e-06, "Full-finetune/Loss": 0.7677751779556274, "Full-finetune/Loss (Raw)": 0.7535661458969116, "Full-finetune/Step": 3509, "Full-finetune/Step Time": 6.690330961719155} {"Full-finetune/Learning Rate": 5.03781242767886e-06, "Full-finetune/Loss": 0.7665584683418274, "Full-finetune/Loss (Raw)": 0.7261371612548828, "Full-finetune/Step": 3510, "Full-finetune/Step Time": 6.689179942011833} {"Full-finetune/Learning Rate": 5.032404858330721e-06, "Full-finetune/Loss": 0.7659351825714111, "Full-finetune/Loss (Raw)": 0.7067317366600037, "Full-finetune/Step": 3511, "Full-finetune/Step Time": 6.686070676892996} {"Full-finetune/Learning Rate": 5.026999216809275e-06, "Full-finetune/Loss": 0.764991044998169, "Full-finetune/Loss (Raw)": 0.7664976119995117, "Full-finetune/Step": 3512, "Full-finetune/Step Time": 6.68425583653152} {"Full-finetune/Learning Rate": 5.021595505212342e-06, "Full-finetune/Loss": 0.7664666175842285, "Full-finetune/Loss (Raw)": 0.8405822515487671, "Full-finetune/Step": 3513, "Full-finetune/Step Time": 6.683747682720423} {"Full-finetune/Learning Rate": 5.016193725637e-06, "Full-finetune/Loss": 0.7660117149353027, "Full-finetune/Loss (Raw)": 0.7303928732872009, "Full-finetune/Step": 3514, "Full-finetune/Step Time": 6.689426155760884} {"Full-finetune/Learning Rate": 5.0107938801795695e-06, "Full-finetune/Loss": 0.7666353583335876, "Full-finetune/Loss (Raw)": 0.7948146462440491, "Full-finetune/Step": 3515, "Full-finetune/Step Time": 6.696206975728273} {"Full-finetune/Learning Rate": 5.005395970935627e-06, "Full-finetune/Loss": 0.7667932510375977, "Full-finetune/Loss (Raw)": 0.7865551114082336, "Full-finetune/Step": 3516, "Full-finetune/Step Time": 6.696755671873689} {"Full-finetune/Learning Rate": 5.000000000000003e-06, "Full-finetune/Loss": 0.7669516801834106, "Full-finetune/Loss (Raw)": 0.8386393785476685, "Full-finetune/Step": 3517, "Full-finetune/Step Time": 6.693789195269346} {"Full-finetune/Learning Rate": 4.994605969466758e-06, "Full-finetune/Loss": 0.7654024362564087, "Full-finetune/Loss (Raw)": 0.6345570087432861, "Full-finetune/Step": 3518, "Full-finetune/Step Time": 6.718965124338865} {"Full-finetune/Learning Rate": 4.989213881429211e-06, "Full-finetune/Loss": 0.7631059885025024, "Full-finetune/Loss (Raw)": 0.6689768433570862, "Full-finetune/Step": 3519, "Full-finetune/Step Time": 6.742328388616443} {"Full-finetune/Learning Rate": 4.983823737979933e-06, "Full-finetune/Loss": 0.7630402445793152, "Full-finetune/Loss (Raw)": 0.7475368976593018, "Full-finetune/Step": 3520, "Full-finetune/Step Time": 6.7437021266669035} {"Full-finetune/Learning Rate": 4.97843554121073e-06, "Full-finetune/Loss": 0.7629089951515198, "Full-finetune/Loss (Raw)": 0.8801080584526062, "Full-finetune/Step": 3521, "Full-finetune/Step Time": 6.744360627606511} {"Full-finetune/Learning Rate": 4.973049293212655e-06, "Full-finetune/Loss": 0.7620887756347656, "Full-finetune/Loss (Raw)": 0.635008692741394, "Full-finetune/Step": 3522, "Full-finetune/Step Time": 6.740091348066926} {"Full-finetune/Learning Rate": 4.967664996076007e-06, "Full-finetune/Loss": 0.7601259350776672, "Full-finetune/Loss (Raw)": 0.6334829926490784, "Full-finetune/Step": 3523, "Full-finetune/Step Time": 6.764508740976453} {"Full-finetune/Learning Rate": 4.962282651890325e-06, "Full-finetune/Loss": 0.7601107954978943, "Full-finetune/Loss (Raw)": 0.6638346314430237, "Full-finetune/Step": 3524, "Full-finetune/Step Time": 6.773127494379878} {"Full-finetune/Learning Rate": 4.9569022627443944e-06, "Full-finetune/Loss": 0.7617782354354858, "Full-finetune/Loss (Raw)": 0.9757049083709717, "Full-finetune/Step": 3525, "Full-finetune/Step Time": 6.776473479345441} {"Full-finetune/Learning Rate": 4.951523830726236e-06, "Full-finetune/Loss": 0.7605738043785095, "Full-finetune/Loss (Raw)": 0.6658563017845154, "Full-finetune/Step": 3526, "Full-finetune/Step Time": 6.773658620193601} {"Full-finetune/Learning Rate": 4.946147357923116e-06, "Full-finetune/Loss": 0.7607976198196411, "Full-finetune/Loss (Raw)": 0.7979502081871033, "Full-finetune/Step": 3527, "Full-finetune/Step Time": 6.751072954386473} {"Full-finetune/Learning Rate": 4.940772846421538e-06, "Full-finetune/Loss": 0.7608131766319275, "Full-finetune/Loss (Raw)": 0.7727593779563904, "Full-finetune/Step": 3528, "Full-finetune/Step Time": 6.750819506123662} {"Full-finetune/Learning Rate": 4.935400298307241e-06, "Full-finetune/Loss": 0.7601031064987183, "Full-finetune/Loss (Raw)": 0.6279918551445007, "Full-finetune/Step": 3529, "Full-finetune/Step Time": 6.7602637223899364} {"Full-finetune/Learning Rate": 4.930029715665215e-06, "Full-finetune/Loss": 0.7598732709884644, "Full-finetune/Loss (Raw)": 0.751259446144104, "Full-finetune/Step": 3530, "Full-finetune/Step Time": 6.762205183506012} {"Full-finetune/Learning Rate": 4.9246611005796776e-06, "Full-finetune/Loss": 0.7606899738311768, "Full-finetune/Loss (Raw)": 0.8516057729721069, "Full-finetune/Step": 3531, "Full-finetune/Step Time": 6.7616344932466745} {"Full-finetune/Learning Rate": 4.919294455134071e-06, "Full-finetune/Loss": 0.7598967552185059, "Full-finetune/Loss (Raw)": 0.6569156646728516, "Full-finetune/Step": 3532, "Full-finetune/Step Time": 6.773614859208465} {"Full-finetune/Learning Rate": 4.913929781411099e-06, "Full-finetune/Loss": 0.7579118609428406, "Full-finetune/Loss (Raw)": 0.3261526823043823, "Full-finetune/Step": 3533, "Full-finetune/Step Time": 6.7531620394438505} {"Full-finetune/Learning Rate": 4.90856708149268e-06, "Full-finetune/Loss": 0.7577081322669983, "Full-finetune/Loss (Raw)": 0.828255832195282, "Full-finetune/Step": 3534, "Full-finetune/Step Time": 6.762104477733374} {"Full-finetune/Learning Rate": 4.903206357459976e-06, "Full-finetune/Loss": 0.7539092898368835, "Full-finetune/Loss (Raw)": 0.3403165638446808, "Full-finetune/Step": 3535, "Full-finetune/Step Time": 6.779358804225922} {"Full-finetune/Learning Rate": 4.897847611393378e-06, "Full-finetune/Loss": 0.75336754322052, "Full-finetune/Loss (Raw)": 0.7768311500549316, "Full-finetune/Step": 3536, "Full-finetune/Step Time": 6.777934409677982} {"Full-finetune/Learning Rate": 4.892490845372511e-06, "Full-finetune/Loss": 0.7543244361877441, "Full-finetune/Loss (Raw)": 0.9545949101448059, "Full-finetune/Step": 3537, "Full-finetune/Step Time": 6.786092976108193} {"Full-finetune/Learning Rate": 4.8871360614762295e-06, "Full-finetune/Loss": 0.7559378743171692, "Full-finetune/Loss (Raw)": 0.9045761823654175, "Full-finetune/Step": 3538, "Full-finetune/Step Time": 6.788671771064401} {"Full-finetune/Learning Rate": 4.881783261782621e-06, "Full-finetune/Loss": 0.7554595470428467, "Full-finetune/Loss (Raw)": 0.8845630884170532, "Full-finetune/Step": 3539, "Full-finetune/Step Time": 6.792262384667993} {"Full-finetune/Learning Rate": 4.876432448369006e-06, "Full-finetune/Loss": 0.755279541015625, "Full-finetune/Loss (Raw)": 0.8139135241508484, "Full-finetune/Step": 3540, "Full-finetune/Step Time": 6.795716801658273} {"Full-finetune/Learning Rate": 4.871083623311933e-06, "Full-finetune/Loss": 0.7544594407081604, "Full-finetune/Loss (Raw)": 0.817732036113739, "Full-finetune/Step": 3541, "Full-finetune/Step Time": 6.785760400816798} {"Full-finetune/Learning Rate": 4.865736788687164e-06, "Full-finetune/Loss": 0.7527256011962891, "Full-finetune/Loss (Raw)": 0.6959912776947021, "Full-finetune/Step": 3542, "Full-finetune/Step Time": 6.796376351267099} {"Full-finetune/Learning Rate": 4.860391946569714e-06, "Full-finetune/Loss": 0.7536097168922424, "Full-finetune/Loss (Raw)": 0.770837128162384, "Full-finetune/Step": 3543, "Full-finetune/Step Time": 6.8090248461812735} {"Full-finetune/Learning Rate": 4.855049099033806e-06, "Full-finetune/Loss": 0.756816029548645, "Full-finetune/Loss (Raw)": 0.8428221344947815, "Full-finetune/Step": 3544, "Full-finetune/Step Time": 6.790596237406135} {"Full-finetune/Learning Rate": 4.849708248152897e-06, "Full-finetune/Loss": 0.7545404434204102, "Full-finetune/Loss (Raw)": 0.423871248960495, "Full-finetune/Step": 3545, "Full-finetune/Step Time": 6.801956677809358} {"Full-finetune/Learning Rate": 4.8443693959996655e-06, "Full-finetune/Loss": 0.7538472414016724, "Full-finetune/Loss (Raw)": 0.7331313490867615, "Full-finetune/Step": 3546, "Full-finetune/Step Time": 6.799811763688922} {"Full-finetune/Learning Rate": 4.839032544646016e-06, "Full-finetune/Loss": 0.7545102834701538, "Full-finetune/Loss (Raw)": 0.9030094742774963, "Full-finetune/Step": 3547, "Full-finetune/Step Time": 6.796675577759743} {"Full-finetune/Learning Rate": 4.833697696163077e-06, "Full-finetune/Loss": 0.7563427686691284, "Full-finetune/Loss (Raw)": 0.7573205232620239, "Full-finetune/Step": 3548, "Full-finetune/Step Time": 6.784391297027469} {"Full-finetune/Learning Rate": 4.828364852621195e-06, "Full-finetune/Loss": 0.7568143606185913, "Full-finetune/Loss (Raw)": 0.9197285175323486, "Full-finetune/Step": 3549, "Full-finetune/Step Time": 6.793883612379432} {"Full-finetune/Learning Rate": 4.823034016089949e-06, "Full-finetune/Loss": 0.7543008327484131, "Full-finetune/Loss (Raw)": 0.3477990925312042, "Full-finetune/Step": 3550, "Full-finetune/Step Time": 6.791057106107473} {"Full-finetune/Learning Rate": 4.8177051886381345e-06, "Full-finetune/Loss": 0.7536352872848511, "Full-finetune/Loss (Raw)": 0.6832090020179749, "Full-finetune/Step": 3551, "Full-finetune/Step Time": 6.8212086874991655} {"Full-finetune/Learning Rate": 4.812378372333751e-06, "Full-finetune/Loss": 0.7572841048240662, "Full-finetune/Loss (Raw)": 0.8263310194015503, "Full-finetune/Step": 3552, "Full-finetune/Step Time": 6.800809059292078} {"Full-finetune/Learning Rate": 4.807053569244047e-06, "Full-finetune/Loss": 0.7561629414558411, "Full-finetune/Loss (Raw)": 0.6224035620689392, "Full-finetune/Step": 3553, "Full-finetune/Step Time": 6.806800520047545} {"Full-finetune/Learning Rate": 4.801730781435469e-06, "Full-finetune/Loss": 0.7582157254219055, "Full-finetune/Loss (Raw)": 0.9164968729019165, "Full-finetune/Step": 3554, "Full-finetune/Step Time": 6.809568205848336} {"Full-finetune/Learning Rate": 4.796410010973686e-06, "Full-finetune/Loss": 0.7588694095611572, "Full-finetune/Loss (Raw)": 0.9266992211341858, "Full-finetune/Step": 3555, "Full-finetune/Step Time": 6.806947760283947} {"Full-finetune/Learning Rate": 4.791091259923587e-06, "Full-finetune/Loss": 0.7557053565979004, "Full-finetune/Loss (Raw)": 0.3453374207019806, "Full-finetune/Step": 3556, "Full-finetune/Step Time": 6.8159107603132725} {"Full-finetune/Learning Rate": 4.785774530349275e-06, "Full-finetune/Loss": 0.7567466497421265, "Full-finetune/Loss (Raw)": 0.9818927645683289, "Full-finetune/Step": 3557, "Full-finetune/Step Time": 6.815665502101183} {"Full-finetune/Learning Rate": 4.7804598243140664e-06, "Full-finetune/Loss": 0.7579858303070068, "Full-finetune/Loss (Raw)": 0.8004170656204224, "Full-finetune/Step": 3558, "Full-finetune/Step Time": 6.816832505166531} {"Full-finetune/Learning Rate": 4.775147143880496e-06, "Full-finetune/Loss": 0.7573566436767578, "Full-finetune/Loss (Raw)": 0.846156656742096, "Full-finetune/Step": 3559, "Full-finetune/Step Time": 6.815703433007002} {"Full-finetune/Learning Rate": 4.769836491110314e-06, "Full-finetune/Loss": 0.7592658400535583, "Full-finetune/Loss (Raw)": 0.8922803401947021, "Full-finetune/Step": 3560, "Full-finetune/Step Time": 6.800527108833194} {"Full-finetune/Learning Rate": 4.764527868064484e-06, "Full-finetune/Loss": 0.7605754137039185, "Full-finetune/Loss (Raw)": 0.9558688402175903, "Full-finetune/Step": 3561, "Full-finetune/Step Time": 6.784011529758573} {"Full-finetune/Learning Rate": 4.759221276803172e-06, "Full-finetune/Loss": 0.7601158022880554, "Full-finetune/Loss (Raw)": 0.6251353025436401, "Full-finetune/Step": 3562, "Full-finetune/Step Time": 6.749365493655205} {"Full-finetune/Learning Rate": 4.75391671938576e-06, "Full-finetune/Loss": 0.7597124576568604, "Full-finetune/Loss (Raw)": 0.6802254915237427, "Full-finetune/Step": 3563, "Full-finetune/Step Time": 6.735415484756231} {"Full-finetune/Learning Rate": 4.748614197870852e-06, "Full-finetune/Loss": 0.7602376937866211, "Full-finetune/Loss (Raw)": 0.7707908153533936, "Full-finetune/Step": 3564, "Full-finetune/Step Time": 6.748793128877878} {"Full-finetune/Learning Rate": 4.743313714316249e-06, "Full-finetune/Loss": 0.7599132061004639, "Full-finetune/Loss (Raw)": 0.7564957141876221, "Full-finetune/Step": 3565, "Full-finetune/Step Time": 6.751783469691873} {"Full-finetune/Learning Rate": 4.738015270778965e-06, "Full-finetune/Loss": 0.7613972425460815, "Full-finetune/Loss (Raw)": 0.8878176212310791, "Full-finetune/Step": 3566, "Full-finetune/Step Time": 6.739440929144621} {"Full-finetune/Learning Rate": 4.7327188693152235e-06, "Full-finetune/Loss": 0.7612642049789429, "Full-finetune/Loss (Raw)": 0.794349193572998, "Full-finetune/Step": 3567, "Full-finetune/Step Time": 6.743422262370586} {"Full-finetune/Learning Rate": 4.727424511980453e-06, "Full-finetune/Loss": 0.7620004415512085, "Full-finetune/Loss (Raw)": 0.8065003156661987, "Full-finetune/Step": 3568, "Full-finetune/Step Time": 6.748382311314344} {"Full-finetune/Learning Rate": 4.722132200829291e-06, "Full-finetune/Loss": 0.7630205750465393, "Full-finetune/Loss (Raw)": 0.7529943585395813, "Full-finetune/Step": 3569, "Full-finetune/Step Time": 6.736304933205247} {"Full-finetune/Learning Rate": 4.716841937915582e-06, "Full-finetune/Loss": 0.7632272243499756, "Full-finetune/Loss (Raw)": 0.756409764289856, "Full-finetune/Step": 3570, "Full-finetune/Step Time": 6.730565892532468} {"Full-finetune/Learning Rate": 4.71155372529237e-06, "Full-finetune/Loss": 0.7640494704246521, "Full-finetune/Loss (Raw)": 0.8118607997894287, "Full-finetune/Step": 3571, "Full-finetune/Step Time": 6.723574684932828} {"Full-finetune/Learning Rate": 4.706267565011909e-06, "Full-finetune/Loss": 0.7645970582962036, "Full-finetune/Loss (Raw)": 0.8893557190895081, "Full-finetune/Step": 3572, "Full-finetune/Step Time": 6.724609142169356} {"Full-finetune/Learning Rate": 4.700983459125652e-06, "Full-finetune/Loss": 0.7669473886489868, "Full-finetune/Loss (Raw)": 0.8770468235015869, "Full-finetune/Step": 3573, "Full-finetune/Step Time": 6.706943068653345} {"Full-finetune/Learning Rate": 4.695701409684264e-06, "Full-finetune/Loss": 0.7681565284729004, "Full-finetune/Loss (Raw)": 0.8996939659118652, "Full-finetune/Step": 3574, "Full-finetune/Step Time": 6.705223923549056} {"Full-finetune/Learning Rate": 4.690421418737604e-06, "Full-finetune/Loss": 0.7672362327575684, "Full-finetune/Loss (Raw)": 0.7797989249229431, "Full-finetune/Step": 3575, "Full-finetune/Step Time": 6.706997986882925} {"Full-finetune/Learning Rate": 4.685143488334723e-06, "Full-finetune/Loss": 0.7684702277183533, "Full-finetune/Loss (Raw)": 0.892745316028595, "Full-finetune/Step": 3576, "Full-finetune/Step Time": 6.705281274393201} {"Full-finetune/Learning Rate": 4.679867620523895e-06, "Full-finetune/Loss": 0.7676565647125244, "Full-finetune/Loss (Raw)": 0.7933471202850342, "Full-finetune/Step": 3577, "Full-finetune/Step Time": 6.723486555740237} {"Full-finetune/Learning Rate": 4.674593817352575e-06, "Full-finetune/Loss": 0.768446683883667, "Full-finetune/Loss (Raw)": 0.8725119829177856, "Full-finetune/Step": 3578, "Full-finetune/Step Time": 6.7081295400857925} {"Full-finetune/Learning Rate": 4.669322080867427e-06, "Full-finetune/Loss": 0.7688049674034119, "Full-finetune/Loss (Raw)": 0.8257836103439331, "Full-finetune/Step": 3579, "Full-finetune/Step Time": 6.708947293460369} {"Full-finetune/Learning Rate": 4.664052413114305e-06, "Full-finetune/Loss": 0.7685196995735168, "Full-finetune/Loss (Raw)": 0.837635338306427, "Full-finetune/Step": 3580, "Full-finetune/Step Time": 6.716400757431984} {"Full-finetune/Learning Rate": 4.658784816138266e-06, "Full-finetune/Loss": 0.7691255807876587, "Full-finetune/Loss (Raw)": 0.870310366153717, "Full-finetune/Step": 3581, "Full-finetune/Step Time": 6.726949954405427} {"Full-finetune/Learning Rate": 4.653519291983561e-06, "Full-finetune/Loss": 0.7687187194824219, "Full-finetune/Loss (Raw)": 0.6979113221168518, "Full-finetune/Step": 3582, "Full-finetune/Step Time": 6.7183418199419975} {"Full-finetune/Learning Rate": 4.648255842693635e-06, "Full-finetune/Loss": 0.7718079090118408, "Full-finetune/Loss (Raw)": 0.7382655739784241, "Full-finetune/Step": 3583, "Full-finetune/Step Time": 6.701491825282574} {"Full-finetune/Learning Rate": 4.642994470311136e-06, "Full-finetune/Loss": 0.7713887691497803, "Full-finetune/Loss (Raw)": 0.807577908039093, "Full-finetune/Step": 3584, "Full-finetune/Step Time": 6.702370248734951} {"Full-finetune/Learning Rate": 4.6377351768779e-06, "Full-finetune/Loss": 0.7721621990203857, "Full-finetune/Loss (Raw)": 0.9498925805091858, "Full-finetune/Step": 3585, "Full-finetune/Step Time": 6.7026709001511335} {"Full-finetune/Learning Rate": 4.632477964434944e-06, "Full-finetune/Loss": 0.7710107564926147, "Full-finetune/Loss (Raw)": 0.6654664874076843, "Full-finetune/Step": 3586, "Full-finetune/Step Time": 6.700318088755012} {"Full-finetune/Learning Rate": 4.6272228350225026e-06, "Full-finetune/Loss": 0.7703790068626404, "Full-finetune/Loss (Raw)": 0.7306778430938721, "Full-finetune/Step": 3587, "Full-finetune/Step Time": 6.693758960813284} {"Full-finetune/Learning Rate": 4.621969790679984e-06, "Full-finetune/Loss": 0.7685055732727051, "Full-finetune/Loss (Raw)": 0.653234601020813, "Full-finetune/Step": 3588, "Full-finetune/Step Time": 6.695253815501928} {"Full-finetune/Learning Rate": 4.6167188334459936e-06, "Full-finetune/Loss": 0.7717618942260742, "Full-finetune/Loss (Raw)": 0.7627809047698975, "Full-finetune/Step": 3589, "Full-finetune/Step Time": 6.686029523611069} {"Full-finetune/Learning Rate": 4.611469965358323e-06, "Full-finetune/Loss": 0.7722093462944031, "Full-finetune/Loss (Raw)": 0.854968786239624, "Full-finetune/Step": 3590, "Full-finetune/Step Time": 6.688904756680131} {"Full-finetune/Learning Rate": 4.606223188453956e-06, "Full-finetune/Loss": 0.7719353437423706, "Full-finetune/Loss (Raw)": 0.7272666692733765, "Full-finetune/Step": 3591, "Full-finetune/Step Time": 6.6872469168156385} {"Full-finetune/Learning Rate": 4.6009785047690666e-06, "Full-finetune/Loss": 0.7709641456604004, "Full-finetune/Loss (Raw)": 0.7768320441246033, "Full-finetune/Step": 3592, "Full-finetune/Step Time": 6.6940727308392525} {"Full-finetune/Learning Rate": 4.595735916339008e-06, "Full-finetune/Loss": 0.7724132537841797, "Full-finetune/Loss (Raw)": 0.9106431603431702, "Full-finetune/Step": 3593, "Full-finetune/Step Time": 6.6903379168361425} {"Full-finetune/Learning Rate": 4.590495425198335e-06, "Full-finetune/Loss": 0.7712651491165161, "Full-finetune/Loss (Raw)": 0.5322893857955933, "Full-finetune/Step": 3594, "Full-finetune/Step Time": 6.6969678439199924} {"Full-finetune/Learning Rate": 4.585257033380779e-06, "Full-finetune/Loss": 0.7726630568504333, "Full-finetune/Loss (Raw)": 0.8293911218643188, "Full-finetune/Step": 3595, "Full-finetune/Step Time": 6.7081778682768345} {"Full-finetune/Learning Rate": 4.580020742919247e-06, "Full-finetune/Loss": 0.7751421332359314, "Full-finetune/Loss (Raw)": 0.9614834189414978, "Full-finetune/Step": 3596, "Full-finetune/Step Time": 6.694525999948382} {"Full-finetune/Learning Rate": 4.5747865558458524e-06, "Full-finetune/Loss": 0.7715861797332764, "Full-finetune/Loss (Raw)": 0.25018438696861267, "Full-finetune/Step": 3597, "Full-finetune/Step Time": 6.701463708654046} {"Full-finetune/Learning Rate": 4.569554474191878e-06, "Full-finetune/Loss": 0.7713624238967896, "Full-finetune/Loss (Raw)": 0.7409363985061646, "Full-finetune/Step": 3598, "Full-finetune/Step Time": 6.7163349483162165} {"Full-finetune/Learning Rate": 4.56432449998779e-06, "Full-finetune/Loss": 0.7699781656265259, "Full-finetune/Loss (Raw)": 0.6865502595901489, "Full-finetune/Step": 3599, "Full-finetune/Step Time": 6.708592230454087} {"Full-finetune/Learning Rate": 4.559096635263242e-06, "Full-finetune/Loss": 0.7676526308059692, "Full-finetune/Loss (Raw)": 0.4412221312522888, "Full-finetune/Step": 3600, "Full-finetune/Step Time": 6.728533582761884} {"Full-finetune/Learning Rate": 4.553870882047064e-06, "Full-finetune/Loss": 0.7674974203109741, "Full-finetune/Loss (Raw)": 0.8337535262107849, "Full-finetune/Step": 3601, "Full-finetune/Step Time": 6.712484234943986} {"Full-finetune/Learning Rate": 4.548647242367271e-06, "Full-finetune/Loss": 0.7684324979782104, "Full-finetune/Loss (Raw)": 0.7827955484390259, "Full-finetune/Step": 3602, "Full-finetune/Step Time": 6.707379212602973} {"Full-finetune/Learning Rate": 4.543425718251051e-06, "Full-finetune/Loss": 0.7683547735214233, "Full-finetune/Loss (Raw)": 0.7951821088790894, "Full-finetune/Step": 3603, "Full-finetune/Step Time": 6.705991206690669} {"Full-finetune/Learning Rate": 4.538206311724782e-06, "Full-finetune/Loss": 0.7698255777359009, "Full-finetune/Loss (Raw)": 0.8144571185112, "Full-finetune/Step": 3604, "Full-finetune/Step Time": 6.702442081645131} {"Full-finetune/Learning Rate": 4.532989024814015e-06, "Full-finetune/Loss": 0.7690983414649963, "Full-finetune/Loss (Raw)": 0.638154923915863, "Full-finetune/Step": 3605, "Full-finetune/Step Time": 6.688990829512477} {"Full-finetune/Learning Rate": 4.527773859543466e-06, "Full-finetune/Loss": 0.7674498558044434, "Full-finetune/Loss (Raw)": 0.726563036441803, "Full-finetune/Step": 3606, "Full-finetune/Step Time": 6.700389685109258} {"Full-finetune/Learning Rate": 4.52256081793705e-06, "Full-finetune/Loss": 0.7657870650291443, "Full-finetune/Loss (Raw)": 0.8078389763832092, "Full-finetune/Step": 3607, "Full-finetune/Step Time": 6.714530171826482} {"Full-finetune/Learning Rate": 4.517349902017841e-06, "Full-finetune/Loss": 0.7642524242401123, "Full-finetune/Loss (Raw)": 0.6439917683601379, "Full-finetune/Step": 3608, "Full-finetune/Step Time": 6.741308951750398} {"Full-finetune/Learning Rate": 4.512141113808097e-06, "Full-finetune/Loss": 0.7633492946624756, "Full-finetune/Loss (Raw)": 0.7363067269325256, "Full-finetune/Step": 3609, "Full-finetune/Step Time": 6.741920651867986} {"Full-finetune/Learning Rate": 4.5069344553292436e-06, "Full-finetune/Loss": 0.7627081871032715, "Full-finetune/Loss (Raw)": 0.6711386442184448, "Full-finetune/Step": 3610, "Full-finetune/Step Time": 6.74117636308074} {"Full-finetune/Learning Rate": 4.501729928601887e-06, "Full-finetune/Loss": 0.7596423625946045, "Full-finetune/Loss (Raw)": 0.33703964948654175, "Full-finetune/Step": 3611, "Full-finetune/Step Time": 6.7194708827883005} {"Full-finetune/Learning Rate": 4.4965275356458e-06, "Full-finetune/Loss": 0.759633481502533, "Full-finetune/Loss (Raw)": 0.7711215615272522, "Full-finetune/Step": 3612, "Full-finetune/Step Time": 6.717809297144413} {"Full-finetune/Learning Rate": 4.491327278479924e-06, "Full-finetune/Loss": 0.7577468156814575, "Full-finetune/Loss (Raw)": 0.6901918053627014, "Full-finetune/Step": 3613, "Full-finetune/Step Time": 6.7191034108400345} {"Full-finetune/Learning Rate": 4.4861291591223934e-06, "Full-finetune/Loss": 0.757713794708252, "Full-finetune/Loss (Raw)": 0.7893717288970947, "Full-finetune/Step": 3614, "Full-finetune/Step Time": 6.72209014557302} {"Full-finetune/Learning Rate": 4.4809331795904835e-06, "Full-finetune/Loss": 0.7586321830749512, "Full-finetune/Loss (Raw)": 0.8334217667579651, "Full-finetune/Step": 3615, "Full-finetune/Step Time": 6.727288478985429} {"Full-finetune/Learning Rate": 4.47573934190065e-06, "Full-finetune/Loss": 0.7586314082145691, "Full-finetune/Loss (Raw)": 0.6834294199943542, "Full-finetune/Step": 3616, "Full-finetune/Step Time": 6.720697546377778} {"Full-finetune/Learning Rate": 4.470547648068532e-06, "Full-finetune/Loss": 0.7573649883270264, "Full-finetune/Loss (Raw)": 0.7106797099113464, "Full-finetune/Step": 3617, "Full-finetune/Step Time": 6.728517226874828} {"Full-finetune/Learning Rate": 4.465358100108916e-06, "Full-finetune/Loss": 0.757046103477478, "Full-finetune/Loss (Raw)": 0.6695952415466309, "Full-finetune/Step": 3618, "Full-finetune/Step Time": 6.7256541680544615} {"Full-finetune/Learning Rate": 4.460170700035769e-06, "Full-finetune/Loss": 0.755870521068573, "Full-finetune/Loss (Raw)": 0.7623617649078369, "Full-finetune/Step": 3619, "Full-finetune/Step Time": 6.724005335941911} {"Full-finetune/Learning Rate": 4.454985449862215e-06, "Full-finetune/Loss": 0.7557535767555237, "Full-finetune/Loss (Raw)": 0.7864611148834229, "Full-finetune/Step": 3620, "Full-finetune/Step Time": 6.724468121305108} {"Full-finetune/Learning Rate": 4.449802351600551e-06, "Full-finetune/Loss": 0.7577456831932068, "Full-finetune/Loss (Raw)": 0.7355867028236389, "Full-finetune/Step": 3621, "Full-finetune/Step Time": 6.719112277030945} {"Full-finetune/Learning Rate": 4.444621407262237e-06, "Full-finetune/Loss": 0.7565144300460815, "Full-finetune/Loss (Raw)": 0.7773601412773132, "Full-finetune/Step": 3622, "Full-finetune/Step Time": 6.71756811439991} {"Full-finetune/Learning Rate": 4.439442618857891e-06, "Full-finetune/Loss": 0.7563450932502747, "Full-finetune/Loss (Raw)": 0.8263647556304932, "Full-finetune/Step": 3623, "Full-finetune/Step Time": 6.7302716840058565} {"Full-finetune/Learning Rate": 4.434265988397312e-06, "Full-finetune/Loss": 0.7565406560897827, "Full-finetune/Loss (Raw)": 0.890249490737915, "Full-finetune/Step": 3624, "Full-finetune/Step Time": 6.723923269659281} {"Full-finetune/Learning Rate": 4.4290915178894355e-06, "Full-finetune/Loss": 0.7585266828536987, "Full-finetune/Loss (Raw)": 0.8830204606056213, "Full-finetune/Step": 3625, "Full-finetune/Step Time": 6.715639486908913} {"Full-finetune/Learning Rate": 4.423919209342376e-06, "Full-finetune/Loss": 0.7568548321723938, "Full-finetune/Loss (Raw)": 0.6612371206283569, "Full-finetune/Step": 3626, "Full-finetune/Step Time": 6.71199569106102} {"Full-finetune/Learning Rate": 4.41874906476341e-06, "Full-finetune/Loss": 0.7573995590209961, "Full-finetune/Loss (Raw)": 0.8518504500389099, "Full-finetune/Step": 3627, "Full-finetune/Step Time": 6.725974153727293} {"Full-finetune/Learning Rate": 4.413581086158969e-06, "Full-finetune/Loss": 0.7566097378730774, "Full-finetune/Loss (Raw)": 0.7915576696395874, "Full-finetune/Step": 3628, "Full-finetune/Step Time": 6.731927514076233} {"Full-finetune/Learning Rate": 4.408415275534641e-06, "Full-finetune/Loss": 0.7558083534240723, "Full-finetune/Loss (Raw)": 0.7565916180610657, "Full-finetune/Step": 3629, "Full-finetune/Step Time": 6.738323803991079} {"Full-finetune/Learning Rate": 4.4032516348951795e-06, "Full-finetune/Loss": 0.7539582848548889, "Full-finetune/Loss (Raw)": 0.702444314956665, "Full-finetune/Step": 3630, "Full-finetune/Step Time": 6.727344358339906} {"Full-finetune/Learning Rate": 4.3980901662444905e-06, "Full-finetune/Loss": 0.7520744800567627, "Full-finetune/Loss (Raw)": 0.6298074722290039, "Full-finetune/Step": 3631, "Full-finetune/Step Time": 6.741127310320735} {"Full-finetune/Learning Rate": 4.39293087158564e-06, "Full-finetune/Loss": 0.7530430555343628, "Full-finetune/Loss (Raw)": 0.8804505467414856, "Full-finetune/Step": 3632, "Full-finetune/Step Time": 6.744049057364464} {"Full-finetune/Learning Rate": 4.387773752920845e-06, "Full-finetune/Loss": 0.7539056539535522, "Full-finetune/Loss (Raw)": 0.8776820302009583, "Full-finetune/Step": 3633, "Full-finetune/Step Time": 6.73826502263546} {"Full-finetune/Learning Rate": 4.382618812251495e-06, "Full-finetune/Loss": 0.754264235496521, "Full-finetune/Loss (Raw)": 0.768281877040863, "Full-finetune/Step": 3634, "Full-finetune/Step Time": 6.731132799759507} {"Full-finetune/Learning Rate": 4.377466051578111e-06, "Full-finetune/Loss": 0.7547813057899475, "Full-finetune/Loss (Raw)": 0.6612342000007629, "Full-finetune/Step": 3635, "Full-finetune/Step Time": 6.71545997262001} {"Full-finetune/Learning Rate": 4.372315472900377e-06, "Full-finetune/Loss": 0.755003809928894, "Full-finetune/Loss (Raw)": 0.818527340888977, "Full-finetune/Step": 3636, "Full-finetune/Step Time": 6.715012978762388} {"Full-finetune/Learning Rate": 4.367167078217141e-06, "Full-finetune/Loss": 0.7557792663574219, "Full-finetune/Loss (Raw)": 0.8528187274932861, "Full-finetune/Step": 3637, "Full-finetune/Step Time": 6.691829597577453} {"Full-finetune/Learning Rate": 4.36202086952639e-06, "Full-finetune/Loss": 0.7565961480140686, "Full-finetune/Loss (Raw)": 0.8306998610496521, "Full-finetune/Step": 3638, "Full-finetune/Step Time": 6.690119776874781} {"Full-finetune/Learning Rate": 4.356876848825266e-06, "Full-finetune/Loss": 0.7574544548988342, "Full-finetune/Loss (Raw)": 0.816593587398529, "Full-finetune/Step": 3639, "Full-finetune/Step Time": 6.679939182475209} {"Full-finetune/Learning Rate": 4.351735018110066e-06, "Full-finetune/Loss": 0.7538613080978394, "Full-finetune/Loss (Raw)": 0.30657485127449036, "Full-finetune/Step": 3640, "Full-finetune/Step Time": 6.69871155358851} {"Full-finetune/Learning Rate": 4.346595379376232e-06, "Full-finetune/Loss": 0.7529455423355103, "Full-finetune/Loss (Raw)": 0.7233631014823914, "Full-finetune/Step": 3641, "Full-finetune/Step Time": 6.704024573788047} {"Full-finetune/Learning Rate": 4.341457934618357e-06, "Full-finetune/Loss": 0.7535114288330078, "Full-finetune/Loss (Raw)": 0.8028232455253601, "Full-finetune/Step": 3642, "Full-finetune/Step Time": 6.707383017987013} {"Full-finetune/Learning Rate": 4.336322685830181e-06, "Full-finetune/Loss": 0.7539201974868774, "Full-finetune/Loss (Raw)": 0.847140908241272, "Full-finetune/Step": 3643, "Full-finetune/Step Time": 6.701658571138978} {"Full-finetune/Learning Rate": 4.331189635004604e-06, "Full-finetune/Loss": 0.7538942694664001, "Full-finetune/Loss (Raw)": 0.7832383513450623, "Full-finetune/Step": 3644, "Full-finetune/Step Time": 6.700503332540393} {"Full-finetune/Learning Rate": 4.326058784133652e-06, "Full-finetune/Loss": 0.7532809972763062, "Full-finetune/Loss (Raw)": 0.7601381540298462, "Full-finetune/Step": 3645, "Full-finetune/Step Time": 6.700232304632664} {"Full-finetune/Learning Rate": 4.3209301352085075e-06, "Full-finetune/Loss": 0.7548812627792358, "Full-finetune/Loss (Raw)": 0.8393934369087219, "Full-finetune/Step": 3646, "Full-finetune/Step Time": 6.686763135716319} {"Full-finetune/Learning Rate": 4.315803690219507e-06, "Full-finetune/Loss": 0.7561866641044617, "Full-finetune/Loss (Raw)": 0.8360697627067566, "Full-finetune/Step": 3647, "Full-finetune/Step Time": 6.666055474430323} {"Full-finetune/Learning Rate": 4.310679451156122e-06, "Full-finetune/Loss": 0.7573332786560059, "Full-finetune/Loss (Raw)": 0.8943049311637878, "Full-finetune/Step": 3648, "Full-finetune/Step Time": 6.668879376724362} {"Full-finetune/Learning Rate": 4.305557420006962e-06, "Full-finetune/Loss": 0.7573988437652588, "Full-finetune/Loss (Raw)": 0.8884944915771484, "Full-finetune/Step": 3649, "Full-finetune/Step Time": 6.6740117240697145} {"Full-finetune/Learning Rate": 4.300437598759794e-06, "Full-finetune/Loss": 0.7598344087600708, "Full-finetune/Loss (Raw)": 0.9467641115188599, "Full-finetune/Step": 3650, "Full-finetune/Step Time": 6.6551167238503695} {"Full-finetune/Learning Rate": 4.295319989401522e-06, "Full-finetune/Loss": 0.7599937915802002, "Full-finetune/Loss (Raw)": 0.6538808345794678, "Full-finetune/Step": 3651, "Full-finetune/Step Time": 6.645846432074904} {"Full-finetune/Learning Rate": 4.2902045939181855e-06, "Full-finetune/Loss": 0.7611314058303833, "Full-finetune/Loss (Raw)": 0.809455394744873, "Full-finetune/Step": 3652, "Full-finetune/Step Time": 6.639894148334861} {"Full-finetune/Learning Rate": 4.285091414294973e-06, "Full-finetune/Loss": 0.7604924440383911, "Full-finetune/Loss (Raw)": 0.8939146399497986, "Full-finetune/Step": 3653, "Full-finetune/Step Time": 6.637996010482311} {"Full-finetune/Learning Rate": 4.2799804525162085e-06, "Full-finetune/Loss": 0.7611359357833862, "Full-finetune/Loss (Raw)": 0.7482269406318665, "Full-finetune/Step": 3654, "Full-finetune/Step Time": 6.648504422977567} {"Full-finetune/Learning Rate": 4.274871710565356e-06, "Full-finetune/Loss": 0.7622480988502502, "Full-finetune/Loss (Raw)": 0.940303385257721, "Full-finetune/Step": 3655, "Full-finetune/Step Time": 6.653482012450695} {"Full-finetune/Learning Rate": 4.269765190425015e-06, "Full-finetune/Loss": 0.7616189122200012, "Full-finetune/Loss (Raw)": 0.6922253370285034, "Full-finetune/Step": 3656, "Full-finetune/Step Time": 6.658603327348828} {"Full-finetune/Learning Rate": 4.264660894076934e-06, "Full-finetune/Loss": 0.763106644153595, "Full-finetune/Loss (Raw)": 0.8184213638305664, "Full-finetune/Step": 3657, "Full-finetune/Step Time": 6.654439073055983} {"Full-finetune/Learning Rate": 4.25955882350199e-06, "Full-finetune/Loss": 0.763109564781189, "Full-finetune/Loss (Raw)": 0.7516286969184875, "Full-finetune/Step": 3658, "Full-finetune/Step Time": 6.670255409553647} {"Full-finetune/Learning Rate": 4.254458980680188e-06, "Full-finetune/Loss": 0.7627341151237488, "Full-finetune/Loss (Raw)": 0.803548276424408, "Full-finetune/Step": 3659, "Full-finetune/Step Time": 6.674530116841197} {"Full-finetune/Learning Rate": 4.2493613675906865e-06, "Full-finetune/Loss": 0.764731764793396, "Full-finetune/Loss (Raw)": 0.9126213192939758, "Full-finetune/Step": 3660, "Full-finetune/Step Time": 6.662616968154907} {"Full-finetune/Learning Rate": 4.244265986211766e-06, "Full-finetune/Loss": 0.768518328666687, "Full-finetune/Loss (Raw)": 0.8108310103416443, "Full-finetune/Step": 3661, "Full-finetune/Step Time": 6.665995201095939} {"Full-finetune/Learning Rate": 4.239172838520846e-06, "Full-finetune/Loss": 0.7691782712936401, "Full-finetune/Loss (Raw)": 0.9127325415611267, "Full-finetune/Step": 3662, "Full-finetune/Step Time": 6.655507490038872} {"Full-finetune/Learning Rate": 4.2340819264944755e-06, "Full-finetune/Loss": 0.7742743492126465, "Full-finetune/Loss (Raw)": 0.9926149845123291, "Full-finetune/Step": 3663, "Full-finetune/Step Time": 6.636273261159658} {"Full-finetune/Learning Rate": 4.22899325210834e-06, "Full-finetune/Loss": 0.7738572955131531, "Full-finetune/Loss (Raw)": 0.7234398722648621, "Full-finetune/Step": 3664, "Full-finetune/Step Time": 6.6472746450454} {"Full-finetune/Learning Rate": 4.223906817337253e-06, "Full-finetune/Loss": 0.7728455066680908, "Full-finetune/Loss (Raw)": 0.825086236000061, "Full-finetune/Step": 3665, "Full-finetune/Step Time": 6.638497170060873} {"Full-finetune/Learning Rate": 4.218822624155159e-06, "Full-finetune/Loss": 0.7720516920089722, "Full-finetune/Loss (Raw)": 0.8029634952545166, "Full-finetune/Step": 3666, "Full-finetune/Step Time": 6.633273670449853} {"Full-finetune/Learning Rate": 4.213740674535141e-06, "Full-finetune/Loss": 0.7684376239776611, "Full-finetune/Loss (Raw)": 0.42196881771087646, "Full-finetune/Step": 3667, "Full-finetune/Step Time": 6.647939009591937} {"Full-finetune/Learning Rate": 4.2086609704494015e-06, "Full-finetune/Loss": 0.7678479552268982, "Full-finetune/Loss (Raw)": 0.7384399175643921, "Full-finetune/Step": 3668, "Full-finetune/Step Time": 6.639506643638015} {"Full-finetune/Learning Rate": 4.203583513869267e-06, "Full-finetune/Loss": 0.7672128677368164, "Full-finetune/Loss (Raw)": 0.7364407181739807, "Full-finetune/Step": 3669, "Full-finetune/Step Time": 6.64472066052258} {"Full-finetune/Learning Rate": 4.198508306765209e-06, "Full-finetune/Loss": 0.7681298851966858, "Full-finetune/Loss (Raw)": 0.8133675456047058, "Full-finetune/Step": 3670, "Full-finetune/Step Time": 6.6292734779417515} {"Full-finetune/Learning Rate": 4.193435351106811e-06, "Full-finetune/Loss": 0.769080638885498, "Full-finetune/Loss (Raw)": 0.8925305604934692, "Full-finetune/Step": 3671, "Full-finetune/Step Time": 6.6107806377112865} {"Full-finetune/Learning Rate": 4.188364648862791e-06, "Full-finetune/Loss": 0.7680577635765076, "Full-finetune/Loss (Raw)": 0.7118948101997375, "Full-finetune/Step": 3672, "Full-finetune/Step Time": 6.6183756329119205} {"Full-finetune/Learning Rate": 4.183296202000988e-06, "Full-finetune/Loss": 0.7689209580421448, "Full-finetune/Loss (Raw)": 0.5343599319458008, "Full-finetune/Step": 3673, "Full-finetune/Step Time": 6.6045201029628515} {"Full-finetune/Learning Rate": 4.178230012488368e-06, "Full-finetune/Loss": 0.7693348526954651, "Full-finetune/Loss (Raw)": 0.7861112952232361, "Full-finetune/Step": 3674, "Full-finetune/Step Time": 6.594532826915383} {"Full-finetune/Learning Rate": 4.173166082291019e-06, "Full-finetune/Loss": 0.7652465105056763, "Full-finetune/Loss (Raw)": 0.3796992301940918, "Full-finetune/Step": 3675, "Full-finetune/Step Time": 6.616549037396908} {"Full-finetune/Learning Rate": 4.16810441337415e-06, "Full-finetune/Loss": 0.7656666040420532, "Full-finetune/Loss (Raw)": 0.8110954761505127, "Full-finetune/Step": 3676, "Full-finetune/Step Time": 6.6212389543652534} {"Full-finetune/Learning Rate": 4.163045007702105e-06, "Full-finetune/Loss": 0.7655971050262451, "Full-finetune/Loss (Raw)": 0.910828709602356, "Full-finetune/Step": 3677, "Full-finetune/Step Time": 6.61040379293263} {"Full-finetune/Learning Rate": 4.157987867238338e-06, "Full-finetune/Loss": 0.767181932926178, "Full-finetune/Loss (Raw)": 0.5506625771522522, "Full-finetune/Step": 3678, "Full-finetune/Step Time": 6.593902254477143} {"Full-finetune/Learning Rate": 4.152932993945418e-06, "Full-finetune/Loss": 0.7676703929901123, "Full-finetune/Loss (Raw)": 0.7457289099693298, "Full-finetune/Step": 3679, "Full-finetune/Step Time": 6.579567741602659} {"Full-finetune/Learning Rate": 4.147880389785051e-06, "Full-finetune/Loss": 0.7679964900016785, "Full-finetune/Loss (Raw)": 0.8680697679519653, "Full-finetune/Step": 3680, "Full-finetune/Step Time": 6.606146141886711} {"Full-finetune/Learning Rate": 4.142830056718052e-06, "Full-finetune/Loss": 0.7697180509567261, "Full-finetune/Loss (Raw)": 0.8427673578262329, "Full-finetune/Step": 3681, "Full-finetune/Step Time": 6.587544221431017} {"Full-finetune/Learning Rate": 4.137781996704356e-06, "Full-finetune/Loss": 0.7706058025360107, "Full-finetune/Loss (Raw)": 1.0301275253295898, "Full-finetune/Step": 3682, "Full-finetune/Step Time": 6.584828808903694} {"Full-finetune/Learning Rate": 4.1327362117030174e-06, "Full-finetune/Loss": 0.7696797847747803, "Full-finetune/Loss (Raw)": 0.8081753253936768, "Full-finetune/Step": 3683, "Full-finetune/Step Time": 6.584696389734745} {"Full-finetune/Learning Rate": 4.127692703672207e-06, "Full-finetune/Loss": 0.7712118029594421, "Full-finetune/Loss (Raw)": 0.5414304733276367, "Full-finetune/Step": 3684, "Full-finetune/Step Time": 6.585834685713053} {"Full-finetune/Learning Rate": 4.1226514745692115e-06, "Full-finetune/Loss": 0.7695232629776001, "Full-finetune/Loss (Raw)": 0.7657593488693237, "Full-finetune/Step": 3685, "Full-finetune/Step Time": 6.580639332532883} {"Full-finetune/Learning Rate": 4.117612526350429e-06, "Full-finetune/Loss": 0.7697632312774658, "Full-finetune/Loss (Raw)": 0.8311342597007751, "Full-finetune/Step": 3686, "Full-finetune/Step Time": 6.58004935644567} {"Full-finetune/Learning Rate": 4.112575860971384e-06, "Full-finetune/Loss": 0.7696394920349121, "Full-finetune/Loss (Raw)": 0.830312192440033, "Full-finetune/Step": 3687, "Full-finetune/Step Time": 6.576785897836089} {"Full-finetune/Learning Rate": 4.107541480386708e-06, "Full-finetune/Loss": 0.7675808668136597, "Full-finetune/Loss (Raw)": 0.6287811994552612, "Full-finetune/Step": 3688, "Full-finetune/Step Time": 6.573778798803687} {"Full-finetune/Learning Rate": 4.102509386550141e-06, "Full-finetune/Loss": 0.7669435739517212, "Full-finetune/Loss (Raw)": 0.8742973804473877, "Full-finetune/Step": 3689, "Full-finetune/Step Time": 6.572693655267358} {"Full-finetune/Learning Rate": 4.0974795814145384e-06, "Full-finetune/Loss": 0.768708348274231, "Full-finetune/Loss (Raw)": 0.8510173559188843, "Full-finetune/Step": 3690, "Full-finetune/Step Time": 6.57534658908844} {"Full-finetune/Learning Rate": 4.092452066931877e-06, "Full-finetune/Loss": 0.7658066749572754, "Full-finetune/Loss (Raw)": 0.30881422758102417, "Full-finetune/Step": 3691, "Full-finetune/Step Time": 6.5912044532597065} {"Full-finetune/Learning Rate": 4.087426845053236e-06, "Full-finetune/Loss": 0.766594409942627, "Full-finetune/Loss (Raw)": 0.871622622013092, "Full-finetune/Step": 3692, "Full-finetune/Step Time": 6.575316097587347} {"Full-finetune/Learning Rate": 4.082403917728796e-06, "Full-finetune/Loss": 0.767014741897583, "Full-finetune/Loss (Raw)": 0.8103030323982239, "Full-finetune/Step": 3693, "Full-finetune/Step Time": 6.574230752885342} {"Full-finetune/Learning Rate": 4.077383286907869e-06, "Full-finetune/Loss": 0.7651975750923157, "Full-finetune/Loss (Raw)": 0.6552149653434753, "Full-finetune/Step": 3694, "Full-finetune/Step Time": 6.5971615463495255} {"Full-finetune/Learning Rate": 4.072364954538858e-06, "Full-finetune/Loss": 0.766956090927124, "Full-finetune/Loss (Raw)": 1.0194432735443115, "Full-finetune/Step": 3695, "Full-finetune/Step Time": 6.576056757941842} {"Full-finetune/Learning Rate": 4.067348922569281e-06, "Full-finetune/Loss": 0.7666818499565125, "Full-finetune/Loss (Raw)": 0.7713935375213623, "Full-finetune/Step": 3696, "Full-finetune/Step Time": 6.569391703233123} {"Full-finetune/Learning Rate": 4.062335192945762e-06, "Full-finetune/Loss": 0.764799952507019, "Full-finetune/Loss (Raw)": 0.512114942073822, "Full-finetune/Step": 3697, "Full-finetune/Step Time": 6.5839513055980206} {"Full-finetune/Learning Rate": 4.057323767614032e-06, "Full-finetune/Loss": 0.7653431296348572, "Full-finetune/Loss (Raw)": 0.8259373307228088, "Full-finetune/Step": 3698, "Full-finetune/Step Time": 6.589454321190715} {"Full-finetune/Learning Rate": 4.052314648518925e-06, "Full-finetune/Loss": 0.765189528465271, "Full-finetune/Loss (Raw)": 0.7921925783157349, "Full-finetune/Step": 3699, "Full-finetune/Step Time": 6.58700742572546} {"Full-finetune/Learning Rate": 4.047307837604383e-06, "Full-finetune/Loss": 0.7646355032920837, "Full-finetune/Loss (Raw)": 0.818443238735199, "Full-finetune/Step": 3700, "Full-finetune/Step Time": 6.587543208152056} {"Full-finetune/Learning Rate": 4.0423033368134545e-06, "Full-finetune/Loss": 0.7627740502357483, "Full-finetune/Loss (Raw)": 0.6387811303138733, "Full-finetune/Step": 3701, "Full-finetune/Step Time": 6.5943149253726006} {"Full-finetune/Learning Rate": 4.037301148088292e-06, "Full-finetune/Loss": 0.7622156143188477, "Full-finetune/Loss (Raw)": 0.8282167315483093, "Full-finetune/Step": 3702, "Full-finetune/Step Time": 6.583317490294576} {"Full-finetune/Learning Rate": 4.032301273370132e-06, "Full-finetune/Loss": 0.762836217880249, "Full-finetune/Loss (Raw)": 0.8592308163642883, "Full-finetune/Step": 3703, "Full-finetune/Step Time": 6.579489886760712} {"Full-finetune/Learning Rate": 4.027303714599345e-06, "Full-finetune/Loss": 0.7621111869812012, "Full-finetune/Loss (Raw)": 0.799945592880249, "Full-finetune/Step": 3704, "Full-finetune/Step Time": 6.583535837009549} {"Full-finetune/Learning Rate": 4.022308473715379e-06, "Full-finetune/Loss": 0.760662317276001, "Full-finetune/Loss (Raw)": 0.6078850030899048, "Full-finetune/Step": 3705, "Full-finetune/Step Time": 6.565493252128363} {"Full-finetune/Learning Rate": 4.017315552656788e-06, "Full-finetune/Loss": 0.7598615884780884, "Full-finetune/Loss (Raw)": 0.7700223326683044, "Full-finetune/Step": 3706, "Full-finetune/Step Time": 6.567622490227222} {"Full-finetune/Learning Rate": 4.01232495336123e-06, "Full-finetune/Loss": 0.7590460777282715, "Full-finetune/Loss (Raw)": 0.7214004397392273, "Full-finetune/Step": 3707, "Full-finetune/Step Time": 6.591980716213584} {"Full-finetune/Learning Rate": 4.0073366777654574e-06, "Full-finetune/Loss": 0.757427453994751, "Full-finetune/Loss (Raw)": 0.6304499506950378, "Full-finetune/Step": 3708, "Full-finetune/Step Time": 6.602462342008948} {"Full-finetune/Learning Rate": 4.002350727805324e-06, "Full-finetune/Loss": 0.7578525543212891, "Full-finetune/Loss (Raw)": 0.9247216582298279, "Full-finetune/Step": 3709, "Full-finetune/Step Time": 6.592384437099099} {"Full-finetune/Learning Rate": 3.997367105415775e-06, "Full-finetune/Loss": 0.7596713304519653, "Full-finetune/Loss (Raw)": 0.9307200312614441, "Full-finetune/Step": 3710, "Full-finetune/Step Time": 6.579561326652765} {"Full-finetune/Learning Rate": 3.992385812530864e-06, "Full-finetune/Loss": 0.7589429020881653, "Full-finetune/Loss (Raw)": 0.6450259685516357, "Full-finetune/Step": 3711, "Full-finetune/Step Time": 6.590278388932347} {"Full-finetune/Learning Rate": 3.987406851083735e-06, "Full-finetune/Loss": 0.759636402130127, "Full-finetune/Loss (Raw)": 0.8963465690612793, "Full-finetune/Step": 3712, "Full-finetune/Step Time": 6.588929817080498} {"Full-finetune/Learning Rate": 3.982430223006614e-06, "Full-finetune/Loss": 0.7544616460800171, "Full-finetune/Loss (Raw)": 0.28752070665359497, "Full-finetune/Step": 3713, "Full-finetune/Step Time": 6.6164906937628984} {"Full-finetune/Learning Rate": 3.977455930230842e-06, "Full-finetune/Loss": 0.7555025815963745, "Full-finetune/Loss (Raw)": 0.7987133264541626, "Full-finetune/Step": 3714, "Full-finetune/Step Time": 6.611533939838409} {"Full-finetune/Learning Rate": 3.972483974686845e-06, "Full-finetune/Loss": 0.7562402486801147, "Full-finetune/Loss (Raw)": 0.8250904679298401, "Full-finetune/Step": 3715, "Full-finetune/Step Time": 6.615149386227131} {"Full-finetune/Learning Rate": 3.967514358304139e-06, "Full-finetune/Loss": 0.7564854621887207, "Full-finetune/Loss (Raw)": 0.684619665145874, "Full-finetune/Step": 3716, "Full-finetune/Step Time": 6.637233667075634} {"Full-finetune/Learning Rate": 3.962547083011338e-06, "Full-finetune/Loss": 0.7557451128959656, "Full-finetune/Loss (Raw)": 0.6680194139480591, "Full-finetune/Step": 3717, "Full-finetune/Step Time": 6.656839903444052} {"Full-finetune/Learning Rate": 3.957582150736141e-06, "Full-finetune/Loss": 0.7538133263587952, "Full-finetune/Loss (Raw)": 0.6077014207839966, "Full-finetune/Step": 3718, "Full-finetune/Step Time": 6.660413045436144} {"Full-finetune/Learning Rate": 3.952619563405343e-06, "Full-finetune/Loss": 0.752939760684967, "Full-finetune/Loss (Raw)": 0.6154512763023376, "Full-finetune/Step": 3719, "Full-finetune/Step Time": 6.670715706422925} {"Full-finetune/Learning Rate": 3.9476593229448245e-06, "Full-finetune/Loss": 0.7533836960792542, "Full-finetune/Loss (Raw)": 0.8336566090583801, "Full-finetune/Step": 3720, "Full-finetune/Step Time": 6.673371126875281} {"Full-finetune/Learning Rate": 3.942701431279564e-06, "Full-finetune/Loss": 0.7533255815505981, "Full-finetune/Loss (Raw)": 0.9032024145126343, "Full-finetune/Step": 3721, "Full-finetune/Step Time": 6.674069628119469} {"Full-finetune/Learning Rate": 3.937745890333623e-06, "Full-finetune/Loss": 0.7556754946708679, "Full-finetune/Loss (Raw)": 0.8330825567245483, "Full-finetune/Step": 3722, "Full-finetune/Step Time": 6.671096859499812} {"Full-finetune/Learning Rate": 3.932792702030139e-06, "Full-finetune/Loss": 0.7565706372261047, "Full-finetune/Loss (Raw)": 0.9439706802368164, "Full-finetune/Step": 3723, "Full-finetune/Step Time": 6.667940527200699} {"Full-finetune/Learning Rate": 3.92784186829136e-06, "Full-finetune/Loss": 0.7549940347671509, "Full-finetune/Loss (Raw)": 0.7596734166145325, "Full-finetune/Step": 3724, "Full-finetune/Step Time": 6.66799576766789} {"Full-finetune/Learning Rate": 3.9228933910386014e-06, "Full-finetune/Loss": 0.759948194026947, "Full-finetune/Loss (Raw)": 0.8843196034431458, "Full-finetune/Step": 3725, "Full-finetune/Step Time": 6.653440598398447} {"Full-finetune/Learning Rate": 3.917947272192273e-06, "Full-finetune/Loss": 0.757172703742981, "Full-finetune/Loss (Raw)": 0.38566744327545166, "Full-finetune/Step": 3726, "Full-finetune/Step Time": 6.650879783555865} {"Full-finetune/Learning Rate": 3.913003513671866e-06, "Full-finetune/Loss": 0.7574840784072876, "Full-finetune/Loss (Raw)": 0.7264160513877869, "Full-finetune/Step": 3727, "Full-finetune/Step Time": 6.656208632513881} {"Full-finetune/Learning Rate": 3.908062117395956e-06, "Full-finetune/Loss": 0.7581533789634705, "Full-finetune/Loss (Raw)": 0.5268876552581787, "Full-finetune/Step": 3728, "Full-finetune/Step Time": 6.636324537917972} {"Full-finetune/Learning Rate": 3.903123085282204e-06, "Full-finetune/Loss": 0.7580385208129883, "Full-finetune/Loss (Raw)": 0.8190472722053528, "Full-finetune/Step": 3729, "Full-finetune/Step Time": 6.643456162884831} {"Full-finetune/Learning Rate": 3.898186419247347e-06, "Full-finetune/Loss": 0.7585874795913696, "Full-finetune/Loss (Raw)": 0.8530631065368652, "Full-finetune/Step": 3730, "Full-finetune/Step Time": 6.6720843110233545} {"Full-finetune/Learning Rate": 3.893252121207221e-06, "Full-finetune/Loss": 0.7589049935340881, "Full-finetune/Loss (Raw)": 0.83582603931427, "Full-finetune/Step": 3731, "Full-finetune/Step Time": 6.6714501697570086} {"Full-finetune/Learning Rate": 3.888320193076719e-06, "Full-finetune/Loss": 0.7591227293014526, "Full-finetune/Loss (Raw)": 0.8423237800598145, "Full-finetune/Step": 3732, "Full-finetune/Step Time": 6.669805796816945} {"Full-finetune/Learning Rate": 3.8833906367698245e-06, "Full-finetune/Loss": 0.7601422071456909, "Full-finetune/Loss (Raw)": 0.7686453461647034, "Full-finetune/Step": 3733, "Full-finetune/Step Time": 6.676656814292073} {"Full-finetune/Learning Rate": 3.878463454199613e-06, "Full-finetune/Loss": 0.7614394426345825, "Full-finetune/Loss (Raw)": 0.8926178216934204, "Full-finetune/Step": 3734, "Full-finetune/Step Time": 6.680674118921161} {"Full-finetune/Learning Rate": 3.87353864727822e-06, "Full-finetune/Loss": 0.7600491046905518, "Full-finetune/Loss (Raw)": 0.6298783421516418, "Full-finetune/Step": 3735, "Full-finetune/Step Time": 6.6705616209656} {"Full-finetune/Learning Rate": 3.8686162179168695e-06, "Full-finetune/Loss": 0.7628370523452759, "Full-finetune/Loss (Raw)": 1.0008347034454346, "Full-finetune/Step": 3736, "Full-finetune/Step Time": 6.652121154591441} {"Full-finetune/Learning Rate": 3.863696168025859e-06, "Full-finetune/Loss": 0.7629661560058594, "Full-finetune/Loss (Raw)": 0.7528457641601562, "Full-finetune/Step": 3737, "Full-finetune/Step Time": 6.6798047088086605} {"Full-finetune/Learning Rate": 3.858778499514565e-06, "Full-finetune/Loss": 0.7631985545158386, "Full-finetune/Loss (Raw)": 0.700882613658905, "Full-finetune/Step": 3738, "Full-finetune/Step Time": 6.688615348190069} {"Full-finetune/Learning Rate": 3.853863214291439e-06, "Full-finetune/Loss": 0.7661586403846741, "Full-finetune/Loss (Raw)": 0.715933620929718, "Full-finetune/Step": 3739, "Full-finetune/Step Time": 6.670781377702951} {"Full-finetune/Learning Rate": 3.848950314264001e-06, "Full-finetune/Loss": 0.766063928604126, "Full-finetune/Loss (Raw)": 0.7589898109436035, "Full-finetune/Step": 3740, "Full-finetune/Step Time": 6.652594856917858} {"Full-finetune/Learning Rate": 3.8440398013388655e-06, "Full-finetune/Loss": 0.7677181363105774, "Full-finetune/Loss (Raw)": 0.9019385576248169, "Full-finetune/Step": 3741, "Full-finetune/Step Time": 6.651466645300388} {"Full-finetune/Learning Rate": 3.839131677421694e-06, "Full-finetune/Loss": 0.7679529786109924, "Full-finetune/Loss (Raw)": 0.8194290995597839, "Full-finetune/Step": 3742, "Full-finetune/Step Time": 6.647547269240022} {"Full-finetune/Learning Rate": 3.834225944417235e-06, "Full-finetune/Loss": 0.7678588628768921, "Full-finetune/Loss (Raw)": 0.8213709592819214, "Full-finetune/Step": 3743, "Full-finetune/Step Time": 6.679072979837656} {"Full-finetune/Learning Rate": 3.829322604229313e-06, "Full-finetune/Loss": 0.7653818726539612, "Full-finetune/Loss (Raw)": 0.3663761019706726, "Full-finetune/Step": 3744, "Full-finetune/Step Time": 6.693637182936072} {"Full-finetune/Learning Rate": 3.824421658760815e-06, "Full-finetune/Loss": 0.7623994946479797, "Full-finetune/Loss (Raw)": 0.3289356529712677, "Full-finetune/Step": 3745, "Full-finetune/Step Time": 6.70061718672514} {"Full-finetune/Learning Rate": 3.819523109913706e-06, "Full-finetune/Loss": 0.7625866532325745, "Full-finetune/Loss (Raw)": 0.6935524940490723, "Full-finetune/Step": 3746, "Full-finetune/Step Time": 6.692586524412036} {"Full-finetune/Learning Rate": 3.814626959589012e-06, "Full-finetune/Loss": 0.7613413333892822, "Full-finetune/Loss (Raw)": 0.6029642820358276, "Full-finetune/Step": 3747, "Full-finetune/Step Time": 6.7125443164259195} {"Full-finetune/Learning Rate": 3.8097332096868368e-06, "Full-finetune/Loss": 0.7616804838180542, "Full-finetune/Loss (Raw)": 0.8298649787902832, "Full-finetune/Step": 3748, "Full-finetune/Step Time": 6.712677758187056} {"Full-finetune/Learning Rate": 3.8048418621063475e-06, "Full-finetune/Loss": 0.7607451677322388, "Full-finetune/Loss (Raw)": 0.6158691048622131, "Full-finetune/Step": 3749, "Full-finetune/Step Time": 6.718963209539652} {"Full-finetune/Learning Rate": 3.7999529187457775e-06, "Full-finetune/Loss": 0.7615720629692078, "Full-finetune/Loss (Raw)": 0.8832055926322937, "Full-finetune/Step": 3750, "Full-finetune/Step Time": 6.72387282922864} {"Full-finetune/Learning Rate": 3.7950663815024424e-06, "Full-finetune/Loss": 0.7626740336418152, "Full-finetune/Loss (Raw)": 0.9674195647239685, "Full-finetune/Step": 3751, "Full-finetune/Step Time": 6.718758340924978} {"Full-finetune/Learning Rate": 3.790182252272698e-06, "Full-finetune/Loss": 0.7618540525436401, "Full-finetune/Loss (Raw)": 0.7852811217308044, "Full-finetune/Step": 3752, "Full-finetune/Step Time": 6.722342565655708} {"Full-finetune/Learning Rate": 3.785300532951983e-06, "Full-finetune/Loss": 0.7604081630706787, "Full-finetune/Loss (Raw)": 0.6979541182518005, "Full-finetune/Step": 3753, "Full-finetune/Step Time": 6.7364777233451605} {"Full-finetune/Learning Rate": 3.780421225434803e-06, "Full-finetune/Loss": 0.7613646984100342, "Full-finetune/Loss (Raw)": 0.7836723923683167, "Full-finetune/Step": 3754, "Full-finetune/Step Time": 6.74044656008482} {"Full-finetune/Learning Rate": 3.7755443316147177e-06, "Full-finetune/Loss": 0.7607314586639404, "Full-finetune/Loss (Raw)": 0.7707983255386353, "Full-finetune/Step": 3755, "Full-finetune/Step Time": 6.7278800923377275} {"Full-finetune/Learning Rate": 3.7706698533843556e-06, "Full-finetune/Loss": 0.7592304944992065, "Full-finetune/Loss (Raw)": 0.5994364023208618, "Full-finetune/Step": 3756, "Full-finetune/Step Time": 6.729399802163243} {"Full-finetune/Learning Rate": 3.765797792635406e-06, "Full-finetune/Loss": 0.7604497671127319, "Full-finetune/Loss (Raw)": 0.912656307220459, "Full-finetune/Step": 3757, "Full-finetune/Step Time": 6.732831122353673} {"Full-finetune/Learning Rate": 3.7609281512586203e-06, "Full-finetune/Loss": 0.7601945400238037, "Full-finetune/Loss (Raw)": 0.669770359992981, "Full-finetune/Step": 3758, "Full-finetune/Step Time": 6.746187385171652} {"Full-finetune/Learning Rate": 3.756060931143811e-06, "Full-finetune/Loss": 0.7618958353996277, "Full-finetune/Loss (Raw)": 0.847576916217804, "Full-finetune/Step": 3759, "Full-finetune/Step Time": 6.731225857511163} {"Full-finetune/Learning Rate": 3.7511961341798487e-06, "Full-finetune/Loss": 0.7606046199798584, "Full-finetune/Loss (Raw)": 0.7151691317558289, "Full-finetune/Step": 3760, "Full-finetune/Step Time": 6.736238298937678} {"Full-finetune/Learning Rate": 3.746333762254677e-06, "Full-finetune/Loss": 0.7587409615516663, "Full-finetune/Loss (Raw)": 0.63913893699646, "Full-finetune/Step": 3761, "Full-finetune/Step Time": 6.747530184686184} {"Full-finetune/Learning Rate": 3.7414738172552744e-06, "Full-finetune/Loss": 0.7587980031967163, "Full-finetune/Loss (Raw)": 0.7755803465843201, "Full-finetune/Step": 3762, "Full-finetune/Step Time": 6.754007352516055} {"Full-finetune/Learning Rate": 3.7366163010676937e-06, "Full-finetune/Loss": 0.7606204152107239, "Full-finetune/Loss (Raw)": 0.8945096135139465, "Full-finetune/Step": 3763, "Full-finetune/Step Time": 6.754391252994537} {"Full-finetune/Learning Rate": 3.7317612155770467e-06, "Full-finetune/Loss": 0.7598539590835571, "Full-finetune/Loss (Raw)": 0.7204143404960632, "Full-finetune/Step": 3764, "Full-finetune/Step Time": 6.754231369122863} {"Full-finetune/Learning Rate": 3.726908562667496e-06, "Full-finetune/Loss": 0.7580671310424805, "Full-finetune/Loss (Raw)": 0.6241078972816467, "Full-finetune/Step": 3765, "Full-finetune/Step Time": 6.762191804125905} {"Full-finetune/Learning Rate": 3.72205834422226e-06, "Full-finetune/Loss": 0.7585352063179016, "Full-finetune/Loss (Raw)": 0.8906142115592957, "Full-finetune/Step": 3766, "Full-finetune/Step Time": 6.765782251954079} {"Full-finetune/Learning Rate": 3.717210562123613e-06, "Full-finetune/Loss": 0.7586264610290527, "Full-finetune/Loss (Raw)": 0.8282732963562012, "Full-finetune/Step": 3767, "Full-finetune/Step Time": 6.768854603171349} {"Full-finetune/Learning Rate": 3.712365218252887e-06, "Full-finetune/Loss": 0.7615804076194763, "Full-finetune/Loss (Raw)": 0.6846787929534912, "Full-finetune/Step": 3768, "Full-finetune/Step Time": 6.760162644088268} {"Full-finetune/Learning Rate": 3.7075223144904638e-06, "Full-finetune/Loss": 0.7578179836273193, "Full-finetune/Loss (Raw)": 0.24177689850330353, "Full-finetune/Step": 3769, "Full-finetune/Step Time": 6.784881941974163} {"Full-finetune/Learning Rate": 3.702681852715777e-06, "Full-finetune/Loss": 0.7589454054832458, "Full-finetune/Loss (Raw)": 0.9471279978752136, "Full-finetune/Step": 3770, "Full-finetune/Step Time": 6.777720449492335} {"Full-finetune/Learning Rate": 3.6978438348073253e-06, "Full-finetune/Loss": 0.7596460580825806, "Full-finetune/Loss (Raw)": 0.9368299245834351, "Full-finetune/Step": 3771, "Full-finetune/Step Time": 6.7789196614176035} {"Full-finetune/Learning Rate": 3.6930082626426377e-06, "Full-finetune/Loss": 0.7601600885391235, "Full-finetune/Loss (Raw)": 0.849032461643219, "Full-finetune/Step": 3772, "Full-finetune/Step Time": 6.784955473616719} {"Full-finetune/Learning Rate": 3.6881751380983066e-06, "Full-finetune/Loss": 0.7569251656532288, "Full-finetune/Loss (Raw)": 0.34606921672821045, "Full-finetune/Step": 3773, "Full-finetune/Step Time": 6.800643350929022} {"Full-finetune/Learning Rate": 3.6833444630499804e-06, "Full-finetune/Loss": 0.7571007609367371, "Full-finetune/Loss (Raw)": 0.8618665337562561, "Full-finetune/Step": 3774, "Full-finetune/Step Time": 6.803143898025155} {"Full-finetune/Learning Rate": 3.6785162393723482e-06, "Full-finetune/Loss": 0.7574202418327332, "Full-finetune/Loss (Raw)": 0.8769636154174805, "Full-finetune/Step": 3775, "Full-finetune/Step Time": 6.807794090360403} {"Full-finetune/Learning Rate": 3.6736904689391417e-06, "Full-finetune/Loss": 0.7571932077407837, "Full-finetune/Loss (Raw)": 0.86524498462677, "Full-finetune/Step": 3776, "Full-finetune/Step Time": 6.809917040169239} {"Full-finetune/Learning Rate": 3.6688671536231557e-06, "Full-finetune/Loss": 0.7523635625839233, "Full-finetune/Loss (Raw)": 0.2703058123588562, "Full-finetune/Step": 3777, "Full-finetune/Step Time": 6.827532907947898} {"Full-finetune/Learning Rate": 3.6640462952962217e-06, "Full-finetune/Loss": 0.7514247298240662, "Full-finetune/Loss (Raw)": 0.8265835046768188, "Full-finetune/Step": 3778, "Full-finetune/Step Time": 6.837059237062931} {"Full-finetune/Learning Rate": 3.6592278958292225e-06, "Full-finetune/Loss": 0.7514817118644714, "Full-finetune/Loss (Raw)": 0.6611822843551636, "Full-finetune/Step": 3779, "Full-finetune/Step Time": 6.835111474618316} {"Full-finetune/Learning Rate": 3.6544119570920845e-06, "Full-finetune/Loss": 0.7509026527404785, "Full-finetune/Loss (Raw)": 0.7353361248970032, "Full-finetune/Step": 3780, "Full-finetune/Step Time": 6.845501454547048} {"Full-finetune/Learning Rate": 3.6495984809537775e-06, "Full-finetune/Loss": 0.750503420829773, "Full-finetune/Loss (Raw)": 0.8428062796592712, "Full-finetune/Step": 3781, "Full-finetune/Step Time": 6.841489167883992} {"Full-finetune/Learning Rate": 3.6447874692823195e-06, "Full-finetune/Loss": 0.750530481338501, "Full-finetune/Loss (Raw)": 0.7516882419586182, "Full-finetune/Step": 3782, "Full-finetune/Step Time": 6.824345475062728} {"Full-finetune/Learning Rate": 3.639978923944766e-06, "Full-finetune/Loss": 0.7485423684120178, "Full-finetune/Loss (Raw)": 0.6858259439468384, "Full-finetune/Step": 3783, "Full-finetune/Step Time": 6.8251671474426985} {"Full-finetune/Learning Rate": 3.6351728468072265e-06, "Full-finetune/Loss": 0.748702883720398, "Full-finetune/Loss (Raw)": 0.7127747535705566, "Full-finetune/Step": 3784, "Full-finetune/Step Time": 6.829483276233077} {"Full-finetune/Learning Rate": 3.6303692397348455e-06, "Full-finetune/Loss": 0.748525857925415, "Full-finetune/Loss (Raw)": 0.795759916305542, "Full-finetune/Step": 3785, "Full-finetune/Step Time": 6.827044393867254} {"Full-finetune/Learning Rate": 3.6255681045918e-06, "Full-finetune/Loss": 0.7498912215232849, "Full-finetune/Loss (Raw)": 0.9263955354690552, "Full-finetune/Step": 3786, "Full-finetune/Step Time": 6.809754868969321} {"Full-finetune/Learning Rate": 3.6207694432413255e-06, "Full-finetune/Loss": 0.7501251697540283, "Full-finetune/Loss (Raw)": 0.8335002660751343, "Full-finetune/Step": 3787, "Full-finetune/Step Time": 6.806697227060795} {"Full-finetune/Learning Rate": 3.6159732575456862e-06, "Full-finetune/Loss": 0.7488769292831421, "Full-finetune/Loss (Raw)": 0.7528375387191772, "Full-finetune/Step": 3788, "Full-finetune/Step Time": 6.8212299812585115} {"Full-finetune/Learning Rate": 3.6111795493661885e-06, "Full-finetune/Loss": 0.7475138306617737, "Full-finetune/Loss (Raw)": 0.6363612413406372, "Full-finetune/Step": 3789, "Full-finetune/Step Time": 6.817336658015847} {"Full-finetune/Learning Rate": 3.606388320563177e-06, "Full-finetune/Loss": 0.7471193075180054, "Full-finetune/Loss (Raw)": 0.8622344732284546, "Full-finetune/Step": 3790, "Full-finetune/Step Time": 6.8191496301442385} {"Full-finetune/Learning Rate": 3.6015995729960316e-06, "Full-finetune/Loss": 0.745771050453186, "Full-finetune/Loss (Raw)": 0.8200292587280273, "Full-finetune/Step": 3791, "Full-finetune/Step Time": 6.8183120135217905} {"Full-finetune/Learning Rate": 3.596813308523176e-06, "Full-finetune/Loss": 0.7427642345428467, "Full-finetune/Loss (Raw)": 0.3385743200778961, "Full-finetune/Step": 3792, "Full-finetune/Step Time": 6.823376068845391} {"Full-finetune/Learning Rate": 3.592029529002059e-06, "Full-finetune/Loss": 0.7425262928009033, "Full-finetune/Loss (Raw)": 0.7946276068687439, "Full-finetune/Step": 3793, "Full-finetune/Step Time": 6.842838799580932} {"Full-finetune/Learning Rate": 3.5872482362891802e-06, "Full-finetune/Loss": 0.742536187171936, "Full-finetune/Loss (Raw)": 0.8042265772819519, "Full-finetune/Step": 3794, "Full-finetune/Step Time": 6.841762583702803} {"Full-finetune/Learning Rate": 3.5824694322400653e-06, "Full-finetune/Loss": 0.7445188760757446, "Full-finetune/Loss (Raw)": 0.6757651567459106, "Full-finetune/Step": 3795, "Full-finetune/Step Time": 6.847468502819538} {"Full-finetune/Learning Rate": 3.5776931187092667e-06, "Full-finetune/Loss": 0.7452707290649414, "Full-finetune/Loss (Raw)": 0.8346697092056274, "Full-finetune/Step": 3796, "Full-finetune/Step Time": 6.8508455362170935} {"Full-finetune/Learning Rate": 3.5729192975503856e-06, "Full-finetune/Loss": 0.7442010641098022, "Full-finetune/Loss (Raw)": 0.5995232462882996, "Full-finetune/Step": 3797, "Full-finetune/Step Time": 6.863400984555483} {"Full-finetune/Learning Rate": 3.5681479706160483e-06, "Full-finetune/Loss": 0.7435958385467529, "Full-finetune/Loss (Raw)": 0.7359007596969604, "Full-finetune/Step": 3798, "Full-finetune/Step Time": 6.859867284074426} {"Full-finetune/Learning Rate": 3.563379139757911e-06, "Full-finetune/Loss": 0.7411459684371948, "Full-finetune/Loss (Raw)": 0.5789443850517273, "Full-finetune/Step": 3799, "Full-finetune/Step Time": 6.868776509538293} {"Full-finetune/Learning Rate": 3.558612806826666e-06, "Full-finetune/Loss": 0.7421835660934448, "Full-finetune/Loss (Raw)": 0.8447052836418152, "Full-finetune/Step": 3800, "Full-finetune/Step Time": 6.864461787045002} {"Full-finetune/Learning Rate": 3.553848973672033e-06, "Full-finetune/Loss": 0.7440558075904846, "Full-finetune/Loss (Raw)": 0.7740123271942139, "Full-finetune/Step": 3801, "Full-finetune/Step Time": 6.863580577075481} {"Full-finetune/Learning Rate": 3.549087642142762e-06, "Full-finetune/Loss": 0.7448334693908691, "Full-finetune/Loss (Raw)": 0.8856484889984131, "Full-finetune/Step": 3802, "Full-finetune/Step Time": 6.86422842182219} {"Full-finetune/Learning Rate": 3.544328814086632e-06, "Full-finetune/Loss": 0.748886227607727, "Full-finetune/Loss (Raw)": 0.8984572887420654, "Full-finetune/Step": 3803, "Full-finetune/Step Time": 6.842100705951452} {"Full-finetune/Learning Rate": 3.5395724913504546e-06, "Full-finetune/Loss": 0.7490208745002747, "Full-finetune/Loss (Raw)": 0.8283262252807617, "Full-finetune/Step": 3804, "Full-finetune/Step Time": 6.8242153618484735} {"Full-finetune/Learning Rate": 3.5348186757800683e-06, "Full-finetune/Loss": 0.7465077638626099, "Full-finetune/Loss (Raw)": 0.5891507267951965, "Full-finetune/Step": 3805, "Full-finetune/Step Time": 6.839792121201754} {"Full-finetune/Learning Rate": 3.530067369220326e-06, "Full-finetune/Loss": 0.7482726573944092, "Full-finetune/Loss (Raw)": 0.7765722870826721, "Full-finetune/Step": 3806, "Full-finetune/Step Time": 6.85120945610106} {"Full-finetune/Learning Rate": 3.525318573515125e-06, "Full-finetune/Loss": 0.748154878616333, "Full-finetune/Loss (Raw)": 0.7306509613990784, "Full-finetune/Step": 3807, "Full-finetune/Step Time": 6.842698007822037} {"Full-finetune/Learning Rate": 3.520572290507378e-06, "Full-finetune/Loss": 0.746117353439331, "Full-finetune/Loss (Raw)": 0.607258677482605, "Full-finetune/Step": 3808, "Full-finetune/Step Time": 6.827786384150386} {"Full-finetune/Learning Rate": 3.5158285220390256e-06, "Full-finetune/Loss": 0.7453441619873047, "Full-finetune/Loss (Raw)": 0.7438013553619385, "Full-finetune/Step": 3809, "Full-finetune/Step Time": 6.841149155050516} {"Full-finetune/Learning Rate": 3.5110872699510313e-06, "Full-finetune/Loss": 0.7439554929733276, "Full-finetune/Loss (Raw)": 0.8523778319358826, "Full-finetune/Step": 3810, "Full-finetune/Step Time": 6.837501775473356} {"Full-finetune/Learning Rate": 3.5063485360833816e-06, "Full-finetune/Loss": 0.7431583404541016, "Full-finetune/Loss (Raw)": 0.7061397433280945, "Full-finetune/Step": 3811, "Full-finetune/Step Time": 6.8424664456397295} {"Full-finetune/Learning Rate": 3.5016123222750863e-06, "Full-finetune/Loss": 0.7444344162940979, "Full-finetune/Loss (Raw)": 0.704770028591156, "Full-finetune/Step": 3812, "Full-finetune/Step Time": 6.836391843855381} {"Full-finetune/Learning Rate": 3.4968786303641764e-06, "Full-finetune/Loss": 0.7445201873779297, "Full-finetune/Loss (Raw)": 0.7767391800880432, "Full-finetune/Step": 3813, "Full-finetune/Step Time": 6.8333597630262375} {"Full-finetune/Learning Rate": 3.492147462187714e-06, "Full-finetune/Loss": 0.7437056303024292, "Full-finetune/Loss (Raw)": 0.7268726825714111, "Full-finetune/Step": 3814, "Full-finetune/Step Time": 6.831395752727985} {"Full-finetune/Learning Rate": 3.4874188195817626e-06, "Full-finetune/Loss": 0.7435673475265503, "Full-finetune/Loss (Raw)": 0.8126152753829956, "Full-finetune/Step": 3815, "Full-finetune/Step Time": 6.849634973332286} {"Full-finetune/Learning Rate": 3.4826927043814197e-06, "Full-finetune/Loss": 0.7450281977653503, "Full-finetune/Loss (Raw)": 0.8157662153244019, "Full-finetune/Step": 3816, "Full-finetune/Step Time": 6.845427945256233} {"Full-finetune/Learning Rate": 3.4779691184207954e-06, "Full-finetune/Loss": 0.7438745498657227, "Full-finetune/Loss (Raw)": 0.7266342639923096, "Full-finetune/Step": 3817, "Full-finetune/Step Time": 6.842054909095168} {"Full-finetune/Learning Rate": 3.473248063533028e-06, "Full-finetune/Loss": 0.7431197166442871, "Full-finetune/Loss (Raw)": 0.7543940544128418, "Full-finetune/Step": 3818, "Full-finetune/Step Time": 6.835749847814441} {"Full-finetune/Learning Rate": 3.4685295415502663e-06, "Full-finetune/Loss": 0.7465344071388245, "Full-finetune/Loss (Raw)": 0.7458951473236084, "Full-finetune/Step": 3819, "Full-finetune/Step Time": 6.8207822404801846} {"Full-finetune/Learning Rate": 3.4638135543036656e-06, "Full-finetune/Loss": 0.7448227405548096, "Full-finetune/Loss (Raw)": 0.6525353193283081, "Full-finetune/Step": 3820, "Full-finetune/Step Time": 6.820304466411471} {"Full-finetune/Learning Rate": 3.4591001036234206e-06, "Full-finetune/Loss": 0.7437174916267395, "Full-finetune/Loss (Raw)": 0.6688227653503418, "Full-finetune/Step": 3821, "Full-finetune/Step Time": 6.825307635590434} {"Full-finetune/Learning Rate": 3.4543891913387253e-06, "Full-finetune/Loss": 0.7440937757492065, "Full-finetune/Loss (Raw)": 0.7033825516700745, "Full-finetune/Step": 3822, "Full-finetune/Step Time": 6.805894428864121} {"Full-finetune/Learning Rate": 3.449680819277792e-06, "Full-finetune/Loss": 0.7431718111038208, "Full-finetune/Loss (Raw)": 0.9014285206794739, "Full-finetune/Step": 3823, "Full-finetune/Step Time": 6.819167386740446} {"Full-finetune/Learning Rate": 3.4449749892678486e-06, "Full-finetune/Loss": 0.7436201572418213, "Full-finetune/Loss (Raw)": 0.8287860155105591, "Full-finetune/Step": 3824, "Full-finetune/Step Time": 6.8175393883138895} {"Full-finetune/Learning Rate": 3.4402717031351363e-06, "Full-finetune/Loss": 0.7446398735046387, "Full-finetune/Loss (Raw)": 0.6426317095756531, "Full-finetune/Step": 3825, "Full-finetune/Step Time": 6.796871036291122} {"Full-finetune/Learning Rate": 3.4355709627049073e-06, "Full-finetune/Loss": 0.7435325384140015, "Full-finetune/Loss (Raw)": 0.6841986775398254, "Full-finetune/Step": 3826, "Full-finetune/Step Time": 6.793413547798991} {"Full-finetune/Learning Rate": 3.4308727698014243e-06, "Full-finetune/Loss": 0.7433583736419678, "Full-finetune/Loss (Raw)": 0.7698972821235657, "Full-finetune/Step": 3827, "Full-finetune/Step Time": 6.819810764864087} {"Full-finetune/Learning Rate": 3.4261771262479724e-06, "Full-finetune/Loss": 0.7434484362602234, "Full-finetune/Loss (Raw)": 0.8299762010574341, "Full-finetune/Step": 3828, "Full-finetune/Step Time": 6.82407096400857} {"Full-finetune/Learning Rate": 3.4214840338668376e-06, "Full-finetune/Loss": 0.7406545281410217, "Full-finetune/Loss (Raw)": 0.28115856647491455, "Full-finetune/Step": 3829, "Full-finetune/Step Time": 6.836094422265887} {"Full-finetune/Learning Rate": 3.416793494479308e-06, "Full-finetune/Loss": 0.7403804063796997, "Full-finetune/Loss (Raw)": 0.7931291460990906, "Full-finetune/Step": 3830, "Full-finetune/Step Time": 6.8468121737241745} {"Full-finetune/Learning Rate": 3.412105509905701e-06, "Full-finetune/Loss": 0.7394386529922485, "Full-finetune/Loss (Raw)": 0.7386954426765442, "Full-finetune/Step": 3831, "Full-finetune/Step Time": 6.854047879576683} {"Full-finetune/Learning Rate": 3.407420081965327e-06, "Full-finetune/Loss": 0.7407914400100708, "Full-finetune/Loss (Raw)": 0.9730969071388245, "Full-finetune/Step": 3832, "Full-finetune/Step Time": 6.851884456351399} {"Full-finetune/Learning Rate": 3.402737212476509e-06, "Full-finetune/Loss": 0.7426931858062744, "Full-finetune/Loss (Raw)": 0.8513000011444092, "Full-finetune/Step": 3833, "Full-finetune/Step Time": 6.873539598658681} {"Full-finetune/Learning Rate": 3.398056903256579e-06, "Full-finetune/Loss": 0.7435833811759949, "Full-finetune/Loss (Raw)": 0.8839757442474365, "Full-finetune/Step": 3834, "Full-finetune/Step Time": 6.866760129109025} {"Full-finetune/Learning Rate": 3.393379156121873e-06, "Full-finetune/Loss": 0.7446610331535339, "Full-finetune/Loss (Raw)": 0.8593367338180542, "Full-finetune/Step": 3835, "Full-finetune/Step Time": 6.858355268836021} {"Full-finetune/Learning Rate": 3.3887039728877336e-06, "Full-finetune/Loss": 0.7454132437705994, "Full-finetune/Loss (Raw)": 0.7267385125160217, "Full-finetune/Step": 3836, "Full-finetune/Step Time": 6.836975950747728} {"Full-finetune/Learning Rate": 3.3840313553685034e-06, "Full-finetune/Loss": 0.7440042495727539, "Full-finetune/Loss (Raw)": 0.7443627119064331, "Full-finetune/Step": 3837, "Full-finetune/Step Time": 6.8418571036309} {"Full-finetune/Learning Rate": 3.379361305377542e-06, "Full-finetune/Loss": 0.7423580884933472, "Full-finetune/Loss (Raw)": 0.720012903213501, "Full-finetune/Step": 3838, "Full-finetune/Step Time": 6.839526819065213} {"Full-finetune/Learning Rate": 3.374693824727204e-06, "Full-finetune/Loss": 0.7435664534568787, "Full-finetune/Loss (Raw)": 0.7997024059295654, "Full-finetune/Step": 3839, "Full-finetune/Step Time": 6.853846000507474} {"Full-finetune/Learning Rate": 3.3700289152288377e-06, "Full-finetune/Loss": 0.7423087358474731, "Full-finetune/Loss (Raw)": 0.7353582978248596, "Full-finetune/Step": 3840, "Full-finetune/Step Time": 6.854308804497123} {"Full-finetune/Learning Rate": 3.3653665786928138e-06, "Full-finetune/Loss": 0.7474685907363892, "Full-finetune/Loss (Raw)": 0.9479761719703674, "Full-finetune/Step": 3841, "Full-finetune/Step Time": 6.825680194422603} {"Full-finetune/Learning Rate": 3.360706816928491e-06, "Full-finetune/Loss": 0.7471596002578735, "Full-finetune/Loss (Raw)": 0.7591733932495117, "Full-finetune/Step": 3842, "Full-finetune/Step Time": 6.815297763794661} {"Full-finetune/Learning Rate": 3.3560496317442305e-06, "Full-finetune/Loss": 0.7453658580780029, "Full-finetune/Loss (Raw)": 0.5954853892326355, "Full-finetune/Step": 3843, "Full-finetune/Step Time": 6.822838172316551} {"Full-finetune/Learning Rate": 3.351395024947395e-06, "Full-finetune/Loss": 0.7459726929664612, "Full-finetune/Loss (Raw)": 0.76229327917099, "Full-finetune/Step": 3844, "Full-finetune/Step Time": 6.805097743868828} {"Full-finetune/Learning Rate": 3.3467429983443477e-06, "Full-finetune/Loss": 0.7467070817947388, "Full-finetune/Loss (Raw)": 0.7620216012001038, "Full-finetune/Step": 3845, "Full-finetune/Step Time": 6.777540383860469} {"Full-finetune/Learning Rate": 3.34209355374045e-06, "Full-finetune/Loss": 0.7468631267547607, "Full-finetune/Loss (Raw)": 0.6276789903640747, "Full-finetune/Step": 3846, "Full-finetune/Step Time": 6.773241689428687} {"Full-finetune/Learning Rate": 3.337446692940055e-06, "Full-finetune/Loss": 0.7489468455314636, "Full-finetune/Loss (Raw)": 0.8821654915809631, "Full-finetune/Step": 3847, "Full-finetune/Step Time": 6.771189304068685} {"Full-finetune/Learning Rate": 3.332802417746527e-06, "Full-finetune/Loss": 0.7450457811355591, "Full-finetune/Loss (Raw)": 0.3343135714530945, "Full-finetune/Step": 3848, "Full-finetune/Step Time": 6.778971202671528} {"Full-finetune/Learning Rate": 3.3281607299622175e-06, "Full-finetune/Loss": 0.7449513077735901, "Full-finetune/Loss (Raw)": 0.8911106586456299, "Full-finetune/Step": 3849, "Full-finetune/Step Time": 6.794639404863119} {"Full-finetune/Learning Rate": 3.3235216313884657e-06, "Full-finetune/Loss": 0.7422010898590088, "Full-finetune/Loss (Raw)": 0.48106256127357483, "Full-finetune/Step": 3850, "Full-finetune/Step Time": 6.801818374544382} {"Full-finetune/Learning Rate": 3.3188851238256246e-06, "Full-finetune/Loss": 0.7398576736450195, "Full-finetune/Loss (Raw)": 0.6440063714981079, "Full-finetune/Step": 3851, "Full-finetune/Step Time": 6.794140161946416} {"Full-finetune/Learning Rate": 3.3142512090730283e-06, "Full-finetune/Loss": 0.7409270405769348, "Full-finetune/Loss (Raw)": 0.8965529799461365, "Full-finetune/Step": 3852, "Full-finetune/Step Time": 6.806122263893485} {"Full-finetune/Learning Rate": 3.309619888929011e-06, "Full-finetune/Loss": 0.737613320350647, "Full-finetune/Loss (Raw)": 0.46016815304756165, "Full-finetune/Step": 3853, "Full-finetune/Step Time": 6.818387746810913} {"Full-finetune/Learning Rate": 3.304991165190895e-06, "Full-finetune/Loss": 0.7408841848373413, "Full-finetune/Loss (Raw)": 0.8043407201766968, "Full-finetune/Step": 3854, "Full-finetune/Step Time": 6.816647427156568} {"Full-finetune/Learning Rate": 3.300365039655e-06, "Full-finetune/Loss": 0.739125669002533, "Full-finetune/Loss (Raw)": 0.5013250112533569, "Full-finetune/Step": 3855, "Full-finetune/Step Time": 6.825232634320855} {"Full-finetune/Learning Rate": 3.2957415141166327e-06, "Full-finetune/Loss": 0.7419850826263428, "Full-finetune/Loss (Raw)": 0.8928927779197693, "Full-finetune/Step": 3856, "Full-finetune/Step Time": 6.8177146948874} {"Full-finetune/Learning Rate": 3.2911205903700916e-06, "Full-finetune/Loss": 0.7384200096130371, "Full-finetune/Loss (Raw)": 0.3627122938632965, "Full-finetune/Step": 3857, "Full-finetune/Step Time": 6.824183948338032} {"Full-finetune/Learning Rate": 3.2865022702086756e-06, "Full-finetune/Loss": 0.7343762516975403, "Full-finetune/Loss (Raw)": 0.3354698717594147, "Full-finetune/Step": 3858, "Full-finetune/Step Time": 6.816859195008874} {"Full-finetune/Learning Rate": 3.281886555424656e-06, "Full-finetune/Loss": 0.7353066802024841, "Full-finetune/Loss (Raw)": 0.9549108147621155, "Full-finetune/Step": 3859, "Full-finetune/Step Time": 6.810138156637549} {"Full-finetune/Learning Rate": 3.277273447809299e-06, "Full-finetune/Loss": 0.73509681224823, "Full-finetune/Loss (Raw)": 0.8154638409614563, "Full-finetune/Step": 3860, "Full-finetune/Step Time": 6.803794395178556} {"Full-finetune/Learning Rate": 3.2726629491528714e-06, "Full-finetune/Loss": 0.7347864508628845, "Full-finetune/Loss (Raw)": 0.7289206981658936, "Full-finetune/Step": 3861, "Full-finetune/Step Time": 6.788128644227982} {"Full-finetune/Learning Rate": 3.268055061244614e-06, "Full-finetune/Loss": 0.7342092990875244, "Full-finetune/Loss (Raw)": 0.8187404274940491, "Full-finetune/Step": 3862, "Full-finetune/Step Time": 6.775943044573069} {"Full-finetune/Learning Rate": 3.2634497858727553e-06, "Full-finetune/Loss": 0.7343918085098267, "Full-finetune/Loss (Raw)": 0.6532419919967651, "Full-finetune/Step": 3863, "Full-finetune/Step Time": 6.7839932311326265} {"Full-finetune/Learning Rate": 3.2588471248245157e-06, "Full-finetune/Loss": 0.732915997505188, "Full-finetune/Loss (Raw)": 0.8119320869445801, "Full-finetune/Step": 3864, "Full-finetune/Step Time": 6.775490991771221} {"Full-finetune/Learning Rate": 3.254247079886097e-06, "Full-finetune/Loss": 0.7325894236564636, "Full-finetune/Loss (Raw)": 0.7110417485237122, "Full-finetune/Step": 3865, "Full-finetune/Step Time": 6.782412242144346} {"Full-finetune/Learning Rate": 3.2496496528426867e-06, "Full-finetune/Loss": 0.7342932224273682, "Full-finetune/Loss (Raw)": 0.9189695715904236, "Full-finetune/Step": 3866, "Full-finetune/Step Time": 6.773449689149857} {"Full-finetune/Learning Rate": 3.245054845478454e-06, "Full-finetune/Loss": 0.7363519668579102, "Full-finetune/Loss (Raw)": 0.9794519543647766, "Full-finetune/Step": 3867, "Full-finetune/Step Time": 6.773548807948828} {"Full-finetune/Learning Rate": 3.240462659576562e-06, "Full-finetune/Loss": 0.7369889616966248, "Full-finetune/Loss (Raw)": 0.8405231833457947, "Full-finetune/Step": 3868, "Full-finetune/Step Time": 6.776769902557135} {"Full-finetune/Learning Rate": 3.23587309691914e-06, "Full-finetune/Loss": 0.7355479001998901, "Full-finetune/Loss (Raw)": 0.7174869775772095, "Full-finetune/Step": 3869, "Full-finetune/Step Time": 6.7790090180933475} {"Full-finetune/Learning Rate": 3.231286159287308e-06, "Full-finetune/Loss": 0.7343610525131226, "Full-finetune/Loss (Raw)": 0.6675189733505249, "Full-finetune/Step": 3870, "Full-finetune/Step Time": 6.802369402721524} {"Full-finetune/Learning Rate": 3.2267018484611725e-06, "Full-finetune/Loss": 0.7341228723526001, "Full-finetune/Loss (Raw)": 0.7908751368522644, "Full-finetune/Step": 3871, "Full-finetune/Step Time": 6.766269661486149} {"Full-finetune/Learning Rate": 3.222120166219812e-06, "Full-finetune/Loss": 0.7342636585235596, "Full-finetune/Loss (Raw)": 0.38439759612083435, "Full-finetune/Step": 3872, "Full-finetune/Step Time": 6.781906144693494} {"Full-finetune/Learning Rate": 3.217541114341288e-06, "Full-finetune/Loss": 0.7367639541625977, "Full-finetune/Loss (Raw)": 0.648975133895874, "Full-finetune/Step": 3873, "Full-finetune/Step Time": 6.771225532516837} {"Full-finetune/Learning Rate": 3.2129646946026407e-06, "Full-finetune/Loss": 0.7373796701431274, "Full-finetune/Loss (Raw)": 0.7723686099052429, "Full-finetune/Step": 3874, "Full-finetune/Step Time": 6.769402073696256} {"Full-finetune/Learning Rate": 3.2083909087798905e-06, "Full-finetune/Loss": 0.7386363744735718, "Full-finetune/Loss (Raw)": 0.763810396194458, "Full-finetune/Step": 3875, "Full-finetune/Step Time": 6.750854080542922} {"Full-finetune/Learning Rate": 3.2038197586480336e-06, "Full-finetune/Loss": 0.7384322881698608, "Full-finetune/Loss (Raw)": 0.8037459850311279, "Full-finetune/Step": 3876, "Full-finetune/Step Time": 6.738097902387381} {"Full-finetune/Learning Rate": 3.199251245981041e-06, "Full-finetune/Loss": 0.7390326261520386, "Full-finetune/Loss (Raw)": 0.6927155256271362, "Full-finetune/Step": 3877, "Full-finetune/Step Time": 6.744763502851129} {"Full-finetune/Learning Rate": 3.1946853725518724e-06, "Full-finetune/Loss": 0.7358521223068237, "Full-finetune/Loss (Raw)": 0.47610166668891907, "Full-finetune/Step": 3878, "Full-finetune/Step Time": 6.771216005086899} {"Full-finetune/Learning Rate": 3.1901221401324446e-06, "Full-finetune/Loss": 0.7335929274559021, "Full-finetune/Loss (Raw)": 0.6782390475273132, "Full-finetune/Step": 3879, "Full-finetune/Step Time": 6.796865232288837} {"Full-finetune/Learning Rate": 3.18556155049366e-06, "Full-finetune/Loss": 0.7337298393249512, "Full-finetune/Loss (Raw)": 0.8027999997138977, "Full-finetune/Step": 3880, "Full-finetune/Step Time": 6.793013561517} {"Full-finetune/Learning Rate": 3.1810036054054005e-06, "Full-finetune/Loss": 0.7347302436828613, "Full-finetune/Loss (Raw)": 0.8260160684585571, "Full-finetune/Step": 3881, "Full-finetune/Step Time": 6.7842723950743675} {"Full-finetune/Learning Rate": 3.1764483066365117e-06, "Full-finetune/Loss": 0.7349585294723511, "Full-finetune/Loss (Raw)": 0.8128917217254639, "Full-finetune/Step": 3882, "Full-finetune/Step Time": 6.787327155470848} {"Full-finetune/Learning Rate": 3.1718956559548154e-06, "Full-finetune/Loss": 0.7358195781707764, "Full-finetune/Loss (Raw)": 0.8810067772865295, "Full-finetune/Step": 3883, "Full-finetune/Step Time": 6.780143596231937} {"Full-finetune/Learning Rate": 3.1673456551271086e-06, "Full-finetune/Loss": 0.7367019653320312, "Full-finetune/Loss (Raw)": 0.7123806476593018, "Full-finetune/Step": 3884, "Full-finetune/Step Time": 6.773107968270779} {"Full-finetune/Learning Rate": 3.162798305919157e-06, "Full-finetune/Loss": 0.7336997985839844, "Full-finetune/Loss (Raw)": 0.5283803939819336, "Full-finetune/Step": 3885, "Full-finetune/Step Time": 6.778142085298896} {"Full-finetune/Learning Rate": 3.1582536100956973e-06, "Full-finetune/Loss": 0.7337721586227417, "Full-finetune/Loss (Raw)": 0.6790314316749573, "Full-finetune/Step": 3886, "Full-finetune/Step Time": 6.765564287081361} {"Full-finetune/Learning Rate": 3.1537115694204345e-06, "Full-finetune/Loss": 0.7326791882514954, "Full-finetune/Loss (Raw)": 0.7076809406280518, "Full-finetune/Step": 3887, "Full-finetune/Step Time": 6.766093524172902} {"Full-finetune/Learning Rate": 3.1491721856560555e-06, "Full-finetune/Loss": 0.7331706285476685, "Full-finetune/Loss (Raw)": 0.7780719995498657, "Full-finetune/Step": 3888, "Full-finetune/Step Time": 6.75844039209187} {"Full-finetune/Learning Rate": 3.144635460564197e-06, "Full-finetune/Loss": 0.7342481017112732, "Full-finetune/Loss (Raw)": 0.7770541310310364, "Full-finetune/Step": 3889, "Full-finetune/Step Time": 6.749239034950733} {"Full-finetune/Learning Rate": 3.1401013959054726e-06, "Full-finetune/Loss": 0.7342096567153931, "Full-finetune/Loss (Raw)": 0.7706623077392578, "Full-finetune/Step": 3890, "Full-finetune/Step Time": 6.740645702928305} {"Full-finetune/Learning Rate": 3.1355699934394724e-06, "Full-finetune/Loss": 0.7324711680412292, "Full-finetune/Loss (Raw)": 0.6719832420349121, "Full-finetune/Step": 3891, "Full-finetune/Step Time": 6.755410555750132} {"Full-finetune/Learning Rate": 3.1310412549247403e-06, "Full-finetune/Loss": 0.7330608367919922, "Full-finetune/Loss (Raw)": 0.7958879470825195, "Full-finetune/Step": 3892, "Full-finetune/Step Time": 6.758360544219613} {"Full-finetune/Learning Rate": 3.1265151821187933e-06, "Full-finetune/Loss": 0.7311093211174011, "Full-finetune/Loss (Raw)": 0.37431657314300537, "Full-finetune/Step": 3893, "Full-finetune/Step Time": 6.776087887585163} {"Full-finetune/Learning Rate": 3.121991776778109e-06, "Full-finetune/Loss": 0.7304911613464355, "Full-finetune/Loss (Raw)": 0.8114885687828064, "Full-finetune/Step": 3894, "Full-finetune/Step Time": 6.774886712431908} {"Full-finetune/Learning Rate": 3.1174710406581355e-06, "Full-finetune/Loss": 0.7302170395851135, "Full-finetune/Loss (Raw)": 0.7931885719299316, "Full-finetune/Step": 3895, "Full-finetune/Step Time": 6.772943323478103} {"Full-finetune/Learning Rate": 3.1129529755132815e-06, "Full-finetune/Loss": 0.730985701084137, "Full-finetune/Loss (Raw)": 0.7830605506896973, "Full-finetune/Step": 3896, "Full-finetune/Step Time": 6.761087339371443} {"Full-finetune/Learning Rate": 3.108437583096918e-06, "Full-finetune/Loss": 0.7349746227264404, "Full-finetune/Loss (Raw)": 0.7523607611656189, "Full-finetune/Step": 3897, "Full-finetune/Step Time": 6.731119362637401} {"Full-finetune/Learning Rate": 3.103924865161383e-06, "Full-finetune/Loss": 0.7332248687744141, "Full-finetune/Loss (Raw)": 0.7231692671775818, "Full-finetune/Step": 3898, "Full-finetune/Step Time": 6.745177956297994} {"Full-finetune/Learning Rate": 3.0994148234579734e-06, "Full-finetune/Loss": 0.7329686284065247, "Full-finetune/Loss (Raw)": 0.9040242433547974, "Full-finetune/Step": 3899, "Full-finetune/Step Time": 6.748763186857104} {"Full-finetune/Learning Rate": 3.0949074597369444e-06, "Full-finetune/Loss": 0.7322679758071899, "Full-finetune/Loss (Raw)": 0.7593528628349304, "Full-finetune/Step": 3900, "Full-finetune/Step Time": 6.757550366222858} {"Full-finetune/Learning Rate": 3.090402775747523e-06, "Full-finetune/Loss": 0.7361838221549988, "Full-finetune/Loss (Raw)": 0.8472952246665955, "Full-finetune/Step": 3901, "Full-finetune/Step Time": 6.739269195124507} {"Full-finetune/Learning Rate": 3.0859007732378897e-06, "Full-finetune/Loss": 0.7367006540298462, "Full-finetune/Loss (Raw)": 0.9280245304107666, "Full-finetune/Step": 3902, "Full-finetune/Step Time": 6.722668757662177} {"Full-finetune/Learning Rate": 3.0814014539551727e-06, "Full-finetune/Loss": 0.7366921901702881, "Full-finetune/Loss (Raw)": 0.8758754134178162, "Full-finetune/Step": 3903, "Full-finetune/Step Time": 6.716612242162228} {"Full-finetune/Learning Rate": 3.076904819645481e-06, "Full-finetune/Loss": 0.7359195351600647, "Full-finetune/Loss (Raw)": 0.7663403749465942, "Full-finetune/Step": 3904, "Full-finetune/Step Time": 6.719650795683265} {"Full-finetune/Learning Rate": 3.072410872053867e-06, "Full-finetune/Loss": 0.7386256456375122, "Full-finetune/Loss (Raw)": 0.6166973114013672, "Full-finetune/Step": 3905, "Full-finetune/Step Time": 6.724911902099848} {"Full-finetune/Learning Rate": 3.067919612924343e-06, "Full-finetune/Loss": 0.7380834221839905, "Full-finetune/Loss (Raw)": 0.7571778297424316, "Full-finetune/Step": 3906, "Full-finetune/Step Time": 6.754684388637543} {"Full-finetune/Learning Rate": 3.0634310439998803e-06, "Full-finetune/Loss": 0.7410117983818054, "Full-finetune/Loss (Raw)": 1.0360134840011597, "Full-finetune/Step": 3907, "Full-finetune/Step Time": 6.7515344973653555} {"Full-finetune/Learning Rate": 3.0589451670224037e-06, "Full-finetune/Loss": 0.739600658416748, "Full-finetune/Loss (Raw)": 0.5547041296958923, "Full-finetune/Step": 3908, "Full-finetune/Step Time": 6.747719369828701} {"Full-finetune/Learning Rate": 3.054461983732795e-06, "Full-finetune/Loss": 0.7381457090377808, "Full-finetune/Loss (Raw)": 0.656582236289978, "Full-finetune/Step": 3909, "Full-finetune/Step Time": 6.763505270704627} {"Full-finetune/Learning Rate": 3.0499814958708883e-06, "Full-finetune/Loss": 0.7388918995857239, "Full-finetune/Loss (Raw)": 0.8471987247467041, "Full-finetune/Step": 3910, "Full-finetune/Step Time": 6.763136081397533} {"Full-finetune/Learning Rate": 3.045503705175478e-06, "Full-finetune/Loss": 0.7407888770103455, "Full-finetune/Loss (Raw)": 0.9286389350891113, "Full-finetune/Step": 3911, "Full-finetune/Step Time": 6.759763857349753} {"Full-finetune/Learning Rate": 3.041028613384307e-06, "Full-finetune/Loss": 0.7405020594596863, "Full-finetune/Loss (Raw)": 0.6760566234588623, "Full-finetune/Step": 3912, "Full-finetune/Step Time": 6.763155430555344} {"Full-finetune/Learning Rate": 3.036556222234063e-06, "Full-finetune/Loss": 0.7403523921966553, "Full-finetune/Loss (Raw)": 0.776604950428009, "Full-finetune/Step": 3913, "Full-finetune/Step Time": 6.7671555783599615} {"Full-finetune/Learning Rate": 3.032086533460402e-06, "Full-finetune/Loss": 0.7388899326324463, "Full-finetune/Loss (Raw)": 0.7391982078552246, "Full-finetune/Step": 3914, "Full-finetune/Step Time": 6.769324576482177} {"Full-finetune/Learning Rate": 3.027619548797921e-06, "Full-finetune/Loss": 0.7388415336608887, "Full-finetune/Loss (Raw)": 0.8273101449012756, "Full-finetune/Step": 3915, "Full-finetune/Step Time": 6.766228802502155} {"Full-finetune/Learning Rate": 3.023155269980169e-06, "Full-finetune/Loss": 0.7394940853118896, "Full-finetune/Loss (Raw)": 0.8363606929779053, "Full-finetune/Step": 3916, "Full-finetune/Step Time": 6.757077168673277} {"Full-finetune/Learning Rate": 3.0186936987396442e-06, "Full-finetune/Loss": 0.7402350902557373, "Full-finetune/Loss (Raw)": 0.7312131524085999, "Full-finetune/Step": 3917, "Full-finetune/Step Time": 6.757979957386851} {"Full-finetune/Learning Rate": 3.014234836807798e-06, "Full-finetune/Loss": 0.7403661608695984, "Full-finetune/Loss (Raw)": 0.8790132403373718, "Full-finetune/Step": 3918, "Full-finetune/Step Time": 6.756933705881238} {"Full-finetune/Learning Rate": 3.0097786859150247e-06, "Full-finetune/Loss": 0.739786684513092, "Full-finetune/Loss (Raw)": 0.7458563446998596, "Full-finetune/Step": 3919, "Full-finetune/Step Time": 6.76792417652905} {"Full-finetune/Learning Rate": 3.005325247790668e-06, "Full-finetune/Loss": 0.7433369159698486, "Full-finetune/Loss (Raw)": 0.7929986119270325, "Full-finetune/Step": 3920, "Full-finetune/Step Time": 6.784758752211928} {"Full-finetune/Learning Rate": 3.0008745241630266e-06, "Full-finetune/Loss": 0.7417165040969849, "Full-finetune/Loss (Raw)": 0.587213933467865, "Full-finetune/Step": 3921, "Full-finetune/Step Time": 6.792485518381} {"Full-finetune/Learning Rate": 2.9964265167593386e-06, "Full-finetune/Loss": 0.7421727180480957, "Full-finetune/Loss (Raw)": 0.8626202940940857, "Full-finetune/Step": 3922, "Full-finetune/Step Time": 6.795534994453192} {"Full-finetune/Learning Rate": 2.9919812273057815e-06, "Full-finetune/Loss": 0.7409096956253052, "Full-finetune/Loss (Raw)": 0.5141062140464783, "Full-finetune/Step": 3923, "Full-finetune/Step Time": 6.784780757501721} {"Full-finetune/Learning Rate": 2.987538657527491e-06, "Full-finetune/Loss": 0.7409010529518127, "Full-finetune/Loss (Raw)": 0.8335609436035156, "Full-finetune/Step": 3924, "Full-finetune/Step Time": 6.789408760145307} {"Full-finetune/Learning Rate": 2.9830988091485426e-06, "Full-finetune/Loss": 0.742684006690979, "Full-finetune/Loss (Raw)": 0.8277461528778076, "Full-finetune/Step": 3925, "Full-finetune/Step Time": 6.775249268859625} {"Full-finetune/Learning Rate": 2.9786616838919515e-06, "Full-finetune/Loss": 0.7423710227012634, "Full-finetune/Loss (Raw)": 0.6958342790603638, "Full-finetune/Step": 3926, "Full-finetune/Step Time": 6.780870897695422} {"Full-finetune/Learning Rate": 2.9742272834796813e-06, "Full-finetune/Loss": 0.743599534034729, "Full-finetune/Loss (Raw)": 0.736190676689148, "Full-finetune/Step": 3927, "Full-finetune/Step Time": 6.788492316380143} {"Full-finetune/Learning Rate": 2.969795609632636e-06, "Full-finetune/Loss": 0.7436984777450562, "Full-finetune/Loss (Raw)": 0.8573708534240723, "Full-finetune/Step": 3928, "Full-finetune/Step Time": 6.788969399407506} {"Full-finetune/Learning Rate": 2.9653666640706614e-06, "Full-finetune/Loss": 0.7441298365592957, "Full-finetune/Loss (Raw)": 0.829228401184082, "Full-finetune/Step": 3929, "Full-finetune/Step Time": 6.7864710837602615} {"Full-finetune/Learning Rate": 2.9609404485125414e-06, "Full-finetune/Loss": 0.7426538467407227, "Full-finetune/Loss (Raw)": 0.6967222094535828, "Full-finetune/Step": 3930, "Full-finetune/Step Time": 6.79224949143827} {"Full-finetune/Learning Rate": 2.9565169646760108e-06, "Full-finetune/Loss": 0.7426068782806396, "Full-finetune/Loss (Raw)": 0.8924415707588196, "Full-finetune/Step": 3931, "Full-finetune/Step Time": 6.7935510743409395} {"Full-finetune/Learning Rate": 2.9520962142777367e-06, "Full-finetune/Loss": 0.741534948348999, "Full-finetune/Loss (Raw)": 0.6911236047744751, "Full-finetune/Step": 3932, "Full-finetune/Step Time": 6.80863512866199} {"Full-finetune/Learning Rate": 2.947678199033317e-06, "Full-finetune/Loss": 0.7410316467285156, "Full-finetune/Loss (Raw)": 0.5247241854667664, "Full-finetune/Step": 3933, "Full-finetune/Step Time": 6.8173014130443335} {"Full-finetune/Learning Rate": 2.943262920657306e-06, "Full-finetune/Loss": 0.740481972694397, "Full-finetune/Loss (Raw)": 0.7062164545059204, "Full-finetune/Step": 3934, "Full-finetune/Step Time": 6.804490072652698} {"Full-finetune/Learning Rate": 2.938850380863184e-06, "Full-finetune/Loss": 0.7406735420227051, "Full-finetune/Loss (Raw)": 0.7551676630973816, "Full-finetune/Step": 3935, "Full-finetune/Step Time": 6.8077486250549555} {"Full-finetune/Learning Rate": 2.9344405813633724e-06, "Full-finetune/Loss": 0.7428659796714783, "Full-finetune/Loss (Raw)": 0.887893795967102, "Full-finetune/Step": 3936, "Full-finetune/Step Time": 6.8118392284959555} {"Full-finetune/Learning Rate": 2.930033523869228e-06, "Full-finetune/Loss": 0.7434003353118896, "Full-finetune/Loss (Raw)": 0.8121968507766724, "Full-finetune/Step": 3937, "Full-finetune/Step Time": 6.807008353993297} {"Full-finetune/Learning Rate": 2.925629210091043e-06, "Full-finetune/Loss": 0.7426252365112305, "Full-finetune/Loss (Raw)": 0.7531691193580627, "Full-finetune/Step": 3938, "Full-finetune/Step Time": 6.807079944759607} {"Full-finetune/Learning Rate": 2.9212276417380468e-06, "Full-finetune/Loss": 0.7435769438743591, "Full-finetune/Loss (Raw)": 0.8279586434364319, "Full-finetune/Step": 3939, "Full-finetune/Step Time": 6.808947836980224} {"Full-finetune/Learning Rate": 2.9168288205183983e-06, "Full-finetune/Loss": 0.7444667220115662, "Full-finetune/Loss (Raw)": 0.8186609148979187, "Full-finetune/Step": 3940, "Full-finetune/Step Time": 6.797261433675885} {"Full-finetune/Learning Rate": 2.9124327481392023e-06, "Full-finetune/Loss": 0.7450212240219116, "Full-finetune/Loss (Raw)": 0.8477098345756531, "Full-finetune/Step": 3941, "Full-finetune/Step Time": 6.802039785310626} {"Full-finetune/Learning Rate": 2.9080394263064815e-06, "Full-finetune/Loss": 0.7460961937904358, "Full-finetune/Loss (Raw)": 0.8644744157791138, "Full-finetune/Step": 3942, "Full-finetune/Step Time": 6.808510400354862} {"Full-finetune/Learning Rate": 2.9036488567252006e-06, "Full-finetune/Loss": 0.7445553541183472, "Full-finetune/Loss (Raw)": 0.6153876781463623, "Full-finetune/Step": 3943, "Full-finetune/Step Time": 6.814728630706668} {"Full-finetune/Learning Rate": 2.8992610410992506e-06, "Full-finetune/Loss": 0.7447205781936646, "Full-finetune/Loss (Raw)": 0.8369113802909851, "Full-finetune/Step": 3944, "Full-finetune/Step Time": 6.8150888960808516} {"Full-finetune/Learning Rate": 2.894875981131463e-06, "Full-finetune/Loss": 0.7465439438819885, "Full-finetune/Loss (Raw)": 0.9600272178649902, "Full-finetune/Step": 3945, "Full-finetune/Step Time": 6.837725343182683} {"Full-finetune/Learning Rate": 2.890493678523595e-06, "Full-finetune/Loss": 0.7466863393783569, "Full-finetune/Loss (Raw)": 0.7726234197616577, "Full-finetune/Step": 3946, "Full-finetune/Step Time": 6.851974509656429} {"Full-finetune/Learning Rate": 2.8861141349763224e-06, "Full-finetune/Loss": 0.7470607757568359, "Full-finetune/Loss (Raw)": 0.7938227653503418, "Full-finetune/Step": 3947, "Full-finetune/Step Time": 6.858862057328224} {"Full-finetune/Learning Rate": 2.881737352189271e-06, "Full-finetune/Loss": 0.74481600522995, "Full-finetune/Loss (Raw)": 0.36520272493362427, "Full-finetune/Step": 3948, "Full-finetune/Step Time": 6.8782737124711275} {"Full-finetune/Learning Rate": 2.8773633318609796e-06, "Full-finetune/Loss": 0.7463815212249756, "Full-finetune/Loss (Raw)": 0.8692131638526917, "Full-finetune/Step": 3949, "Full-finetune/Step Time": 6.876857565715909} {"Full-finetune/Learning Rate": 2.872992075688922e-06, "Full-finetune/Loss": 0.7477165460586548, "Full-finetune/Loss (Raw)": 0.8742601275444031, "Full-finetune/Step": 3950, "Full-finetune/Step Time": 6.873140346258879} {"Full-finetune/Learning Rate": 2.868623585369497e-06, "Full-finetune/Loss": 0.7475576996803284, "Full-finetune/Loss (Raw)": 0.8810975551605225, "Full-finetune/Step": 3951, "Full-finetune/Step Time": 6.86744948849082} {"Full-finetune/Learning Rate": 2.864257862598029e-06, "Full-finetune/Loss": 0.7463235855102539, "Full-finetune/Loss (Raw)": 0.6708211302757263, "Full-finetune/Step": 3952, "Full-finetune/Step Time": 6.876219756901264} {"Full-finetune/Learning Rate": 2.8598949090687723e-06, "Full-finetune/Loss": 0.7472018003463745, "Full-finetune/Loss (Raw)": 0.7550420165061951, "Full-finetune/Step": 3953, "Full-finetune/Step Time": 6.8747643660753965} {"Full-finetune/Learning Rate": 2.8555347264748988e-06, "Full-finetune/Loss": 0.7475360035896301, "Full-finetune/Loss (Raw)": 0.7269794344902039, "Full-finetune/Step": 3954, "Full-finetune/Step Time": 6.869559481739998} {"Full-finetune/Learning Rate": 2.8511773165085164e-06, "Full-finetune/Loss": 0.7466608285903931, "Full-finetune/Loss (Raw)": 0.6578668355941772, "Full-finetune/Step": 3955, "Full-finetune/Step Time": 6.838591421023011} {"Full-finetune/Learning Rate": 2.8468226808606525e-06, "Full-finetune/Loss": 0.7471635341644287, "Full-finetune/Loss (Raw)": 0.8943274617195129, "Full-finetune/Step": 3956, "Full-finetune/Step Time": 6.838695645332336} {"Full-finetune/Learning Rate": 2.842470821221246e-06, "Full-finetune/Loss": 0.7513124942779541, "Full-finetune/Loss (Raw)": 0.8122273087501526, "Full-finetune/Step": 3957, "Full-finetune/Step Time": 6.831950027495623} {"Full-finetune/Learning Rate": 2.8381217392791784e-06, "Full-finetune/Loss": 0.750751793384552, "Full-finetune/Loss (Raw)": 0.7213617563247681, "Full-finetune/Step": 3958, "Full-finetune/Step Time": 6.827614979818463} {"Full-finetune/Learning Rate": 2.833775436722238e-06, "Full-finetune/Loss": 0.7499464154243469, "Full-finetune/Loss (Raw)": 0.6356021761894226, "Full-finetune/Step": 3959, "Full-finetune/Step Time": 6.840303480625153} {"Full-finetune/Learning Rate": 2.829431915237144e-06, "Full-finetune/Loss": 0.7475016713142395, "Full-finetune/Loss (Raw)": 0.6601688861846924, "Full-finetune/Step": 3960, "Full-finetune/Step Time": 6.854101710021496} {"Full-finetune/Learning Rate": 2.825091176509529e-06, "Full-finetune/Loss": 0.7470616698265076, "Full-finetune/Loss (Raw)": 0.794979989528656, "Full-finetune/Step": 3961, "Full-finetune/Step Time": 6.848349668085575} {"Full-finetune/Learning Rate": 2.8207532222239497e-06, "Full-finetune/Loss": 0.7451800107955933, "Full-finetune/Loss (Raw)": 0.6431309580802917, "Full-finetune/Step": 3962, "Full-finetune/Step Time": 6.856440087780356} {"Full-finetune/Learning Rate": 2.816418054063883e-06, "Full-finetune/Loss": 0.7450138330459595, "Full-finetune/Loss (Raw)": 0.8380577564239502, "Full-finetune/Step": 3963, "Full-finetune/Step Time": 6.85147220082581} {"Full-finetune/Learning Rate": 2.8120856737117187e-06, "Full-finetune/Loss": 0.7456417083740234, "Full-finetune/Loss (Raw)": 0.80710768699646, "Full-finetune/Step": 3964, "Full-finetune/Step Time": 6.870118428021669} {"Full-finetune/Learning Rate": 2.807756082848775e-06, "Full-finetune/Loss": 0.7458896040916443, "Full-finetune/Loss (Raw)": 0.7760940790176392, "Full-finetune/Step": 3965, "Full-finetune/Step Time": 6.881505636498332} {"Full-finetune/Learning Rate": 2.803429283155281e-06, "Full-finetune/Loss": 0.7433077096939087, "Full-finetune/Loss (Raw)": 0.38953983783721924, "Full-finetune/Step": 3966, "Full-finetune/Step Time": 6.903012219816446} {"Full-finetune/Learning Rate": 2.7991052763103767e-06, "Full-finetune/Loss": 0.7432386875152588, "Full-finetune/Loss (Raw)": 0.7908569574356079, "Full-finetune/Step": 3967, "Full-finetune/Step Time": 6.877407124266028} {"Full-finetune/Learning Rate": 2.7947840639921308e-06, "Full-finetune/Loss": 0.7438405752182007, "Full-finetune/Loss (Raw)": 0.8123964667320251, "Full-finetune/Step": 3968, "Full-finetune/Step Time": 6.875348154455423} {"Full-finetune/Learning Rate": 2.7904656478775196e-06, "Full-finetune/Loss": 0.7425565719604492, "Full-finetune/Loss (Raw)": 0.7836245894432068, "Full-finetune/Step": 3969, "Full-finetune/Step Time": 6.899367557838559} {"Full-finetune/Learning Rate": 2.7861500296424372e-06, "Full-finetune/Loss": 0.7432684302330017, "Full-finetune/Loss (Raw)": 0.8502973914146423, "Full-finetune/Step": 3970, "Full-finetune/Step Time": 6.900230372324586} {"Full-finetune/Learning Rate": 2.781837210961689e-06, "Full-finetune/Loss": 0.7457019686698914, "Full-finetune/Loss (Raw)": 0.9069746732711792, "Full-finetune/Step": 3971, "Full-finetune/Step Time": 6.891823813319206} {"Full-finetune/Learning Rate": 2.7775271935089964e-06, "Full-finetune/Loss": 0.7456641793251038, "Full-finetune/Loss (Raw)": 0.7574566602706909, "Full-finetune/Step": 3972, "Full-finetune/Step Time": 6.888229724019766} {"Full-finetune/Learning Rate": 2.7732199789569937e-06, "Full-finetune/Loss": 0.7458004951477051, "Full-finetune/Loss (Raw)": 0.779469907283783, "Full-finetune/Step": 3973, "Full-finetune/Step Time": 6.885837730020285} {"Full-finetune/Learning Rate": 2.768915568977222e-06, "Full-finetune/Loss": 0.7465872168540955, "Full-finetune/Loss (Raw)": 0.7283868193626404, "Full-finetune/Step": 3974, "Full-finetune/Step Time": 6.888761797919869} {"Full-finetune/Learning Rate": 2.7646139652401473e-06, "Full-finetune/Loss": 0.7468031644821167, "Full-finetune/Loss (Raw)": 0.9098038077354431, "Full-finetune/Step": 3975, "Full-finetune/Step Time": 6.876870010048151} {"Full-finetune/Learning Rate": 2.7603151694151354e-06, "Full-finetune/Loss": 0.749897837638855, "Full-finetune/Loss (Raw)": 0.730427086353302, "Full-finetune/Step": 3976, "Full-finetune/Step Time": 6.870288219302893} {"Full-finetune/Learning Rate": 2.7560191831704584e-06, "Full-finetune/Loss": 0.7476598024368286, "Full-finetune/Loss (Raw)": 0.6046463251113892, "Full-finetune/Step": 3977, "Full-finetune/Step Time": 6.863386791199446} {"Full-finetune/Learning Rate": 2.751726008173312e-06, "Full-finetune/Loss": 0.7475370168685913, "Full-finetune/Loss (Raw)": 0.4653395414352417, "Full-finetune/Step": 3978, "Full-finetune/Step Time": 6.854305939748883} {"Full-finetune/Learning Rate": 2.747435646089791e-06, "Full-finetune/Loss": 0.7488937377929688, "Full-finetune/Loss (Raw)": 0.8176755309104919, "Full-finetune/Step": 3979, "Full-finetune/Step Time": 6.857624989002943} {"Full-finetune/Learning Rate": 2.7431480985849023e-06, "Full-finetune/Loss": 0.7463901042938232, "Full-finetune/Loss (Raw)": 0.5760836005210876, "Full-finetune/Step": 3980, "Full-finetune/Step Time": 6.848749555647373} {"Full-finetune/Learning Rate": 2.7388633673225586e-06, "Full-finetune/Loss": 0.7495434880256653, "Full-finetune/Loss (Raw)": 0.8638012409210205, "Full-finetune/Step": 3981, "Full-finetune/Step Time": 6.832339443266392} {"Full-finetune/Learning Rate": 2.734581453965579e-06, "Full-finetune/Loss": 0.7494332194328308, "Full-finetune/Loss (Raw)": 0.7902264595031738, "Full-finetune/Step": 3982, "Full-finetune/Step Time": 6.820713993161917} {"Full-finetune/Learning Rate": 2.730302360175693e-06, "Full-finetune/Loss": 0.7516164779663086, "Full-finetune/Loss (Raw)": 0.7807827591896057, "Full-finetune/Step": 3983, "Full-finetune/Step Time": 6.8044646717607975} {"Full-finetune/Learning Rate": 2.726026087613528e-06, "Full-finetune/Loss": 0.7473223209381104, "Full-finetune/Loss (Raw)": 0.34324169158935547, "Full-finetune/Step": 3984, "Full-finetune/Step Time": 6.822299106046557} {"Full-finetune/Learning Rate": 2.7217526379386327e-06, "Full-finetune/Loss": 0.7510203123092651, "Full-finetune/Loss (Raw)": 0.8360507488250732, "Full-finetune/Step": 3985, "Full-finetune/Step Time": 6.808558978140354} {"Full-finetune/Learning Rate": 2.7174820128094393e-06, "Full-finetune/Loss": 0.755818247795105, "Full-finetune/Loss (Raw)": 0.9496119022369385, "Full-finetune/Step": 3986, "Full-finetune/Step Time": 6.787829630076885} {"Full-finetune/Learning Rate": 2.713214213883295e-06, "Full-finetune/Loss": 0.7538819313049316, "Full-finetune/Loss (Raw)": 0.7070522904396057, "Full-finetune/Step": 3987, "Full-finetune/Step Time": 6.794565927237272} {"Full-finetune/Learning Rate": 2.7089492428164556e-06, "Full-finetune/Loss": 0.7539570331573486, "Full-finetune/Loss (Raw)": 0.8250784873962402, "Full-finetune/Step": 3988, "Full-finetune/Step Time": 6.797709088772535} {"Full-finetune/Learning Rate": 2.7046871012640687e-06, "Full-finetune/Loss": 0.7550504207611084, "Full-finetune/Loss (Raw)": 0.8688770532608032, "Full-finetune/Step": 3989, "Full-finetune/Step Time": 6.7973280511796474} {"Full-finetune/Learning Rate": 2.7004277908801877e-06, "Full-finetune/Loss": 0.7548969984054565, "Full-finetune/Loss (Raw)": 0.7991024255752563, "Full-finetune/Step": 3990, "Full-finetune/Step Time": 6.802236933261156} {"Full-finetune/Learning Rate": 2.6961713133177705e-06, "Full-finetune/Loss": 0.7564758062362671, "Full-finetune/Loss (Raw)": 0.8553286790847778, "Full-finetune/Step": 3991, "Full-finetune/Step Time": 6.7953552436083555} {"Full-finetune/Learning Rate": 2.6919176702286696e-06, "Full-finetune/Loss": 0.7551750540733337, "Full-finetune/Loss (Raw)": 0.6454380750656128, "Full-finetune/Step": 3992, "Full-finetune/Step Time": 6.812994325533509} {"Full-finetune/Learning Rate": 2.687666863263644e-06, "Full-finetune/Loss": 0.7560358047485352, "Full-finetune/Loss (Raw)": 0.8212212920188904, "Full-finetune/Step": 3993, "Full-finetune/Step Time": 6.780468080192804} {"Full-finetune/Learning Rate": 2.683418894072345e-06, "Full-finetune/Loss": 0.7540938854217529, "Full-finetune/Loss (Raw)": 0.6704018115997314, "Full-finetune/Step": 3994, "Full-finetune/Step Time": 6.793339470401406} {"Full-finetune/Learning Rate": 2.6791737643033333e-06, "Full-finetune/Loss": 0.7521970868110657, "Full-finetune/Loss (Raw)": 0.7366625666618347, "Full-finetune/Step": 3995, "Full-finetune/Step Time": 6.804478386417031} {"Full-finetune/Learning Rate": 2.6749314756040547e-06, "Full-finetune/Loss": 0.751101016998291, "Full-finetune/Loss (Raw)": 0.700228750705719, "Full-finetune/Step": 3996, "Full-finetune/Step Time": 6.803073592483997} {"Full-finetune/Learning Rate": 2.6706920296208573e-06, "Full-finetune/Loss": 0.750946044921875, "Full-finetune/Loss (Raw)": 0.697645902633667, "Full-finetune/Step": 3997, "Full-finetune/Step Time": 6.801078809425235} {"Full-finetune/Learning Rate": 2.666455427998994e-06, "Full-finetune/Loss": 0.7525708079338074, "Full-finetune/Loss (Raw)": 0.8754826784133911, "Full-finetune/Step": 3998, "Full-finetune/Step Time": 6.786131050437689} {"Full-finetune/Learning Rate": 2.6622216723826035e-06, "Full-finetune/Loss": 0.7523475885391235, "Full-finetune/Loss (Raw)": 0.7623109221458435, "Full-finetune/Step": 3999, "Full-finetune/Step Time": 6.801654931157827} {"Full-finetune/Learning Rate": 2.6579907644147227e-06, "Full-finetune/Loss": 0.7553997039794922, "Full-finetune/Loss (Raw)": 0.7750639915466309, "Full-finetune/Step": 4000, "Full-finetune/Step Time": 6.7728407476097345} {"Full-finetune/Learning Rate": 2.653762705737287e-06, "Full-finetune/Loss": 0.7559970021247864, "Full-finetune/Loss (Raw)": 0.7254292964935303, "Full-finetune/Step": 4001, "Full-finetune/Step Time": 6.784372037276626} {"Full-finetune/Learning Rate": 2.6495374979911215e-06, "Full-finetune/Loss": 0.7552857398986816, "Full-finetune/Loss (Raw)": 0.6813257336616516, "Full-finetune/Step": 4002, "Full-finetune/Step Time": 6.794834032654762} {"Full-finetune/Learning Rate": 2.6453151428159472e-06, "Full-finetune/Loss": 0.755101203918457, "Full-finetune/Loss (Raw)": 0.7401897311210632, "Full-finetune/Step": 4003, "Full-finetune/Step Time": 6.8070490546524525} {"Full-finetune/Learning Rate": 2.641095641850375e-06, "Full-finetune/Loss": 0.7553145885467529, "Full-finetune/Loss (Raw)": 0.8310580849647522, "Full-finetune/Step": 4004, "Full-finetune/Step Time": 6.8115954995155334} {"Full-finetune/Learning Rate": 2.6368789967319196e-06, "Full-finetune/Loss": 0.7560250759124756, "Full-finetune/Loss (Raw)": 0.7836585640907288, "Full-finetune/Step": 4005, "Full-finetune/Step Time": 6.797025872394443} {"Full-finetune/Learning Rate": 2.6326652090969697e-06, "Full-finetune/Loss": 0.7597165107727051, "Full-finetune/Loss (Raw)": 0.9486121535301208, "Full-finetune/Step": 4006, "Full-finetune/Step Time": 6.766526579856873} {"Full-finetune/Learning Rate": 2.6284542805808144e-06, "Full-finetune/Loss": 0.757238507270813, "Full-finetune/Loss (Raw)": 0.3610471189022064, "Full-finetune/Step": 4007, "Full-finetune/Step Time": 6.748312434181571} {"Full-finetune/Learning Rate": 2.624246212817638e-06, "Full-finetune/Loss": 0.7561154961585999, "Full-finetune/Loss (Raw)": 0.6590635776519775, "Full-finetune/Step": 4008, "Full-finetune/Step Time": 6.748494315892458} {"Full-finetune/Learning Rate": 2.620041007440508e-06, "Full-finetune/Loss": 0.755872905254364, "Full-finetune/Loss (Raw)": 0.7949612736701965, "Full-finetune/Step": 4009, "Full-finetune/Step Time": 6.7449373081326485} {"Full-finetune/Learning Rate": 2.6158386660813807e-06, "Full-finetune/Loss": 0.7548087239265442, "Full-finetune/Loss (Raw)": 0.6766736507415771, "Full-finetune/Step": 4010, "Full-finetune/Step Time": 6.753127535805106} {"Full-finetune/Learning Rate": 2.6116391903711035e-06, "Full-finetune/Loss": 0.7543694972991943, "Full-finetune/Loss (Raw)": 0.8247919678688049, "Full-finetune/Step": 4011, "Full-finetune/Step Time": 6.759800085797906} {"Full-finetune/Learning Rate": 2.607442581939411e-06, "Full-finetune/Loss": 0.7531104683876038, "Full-finetune/Loss (Raw)": 0.5512214303016663, "Full-finetune/Step": 4012, "Full-finetune/Step Time": 6.777630373835564} {"Full-finetune/Learning Rate": 2.603248842414925e-06, "Full-finetune/Loss": 0.7546032071113586, "Full-finetune/Loss (Raw)": 0.719452440738678, "Full-finetune/Step": 4013, "Full-finetune/Step Time": 6.772389797493815} {"Full-finetune/Learning Rate": 2.599057973425151e-06, "Full-finetune/Loss": 0.7548478841781616, "Full-finetune/Loss (Raw)": 0.7103491425514221, "Full-finetune/Step": 4014, "Full-finetune/Step Time": 6.797424955293536} {"Full-finetune/Learning Rate": 2.5948699765964915e-06, "Full-finetune/Loss": 0.7554048299789429, "Full-finetune/Loss (Raw)": 0.7789683938026428, "Full-finetune/Step": 4015, "Full-finetune/Step Time": 6.810111276805401} {"Full-finetune/Learning Rate": 2.5906848535542196e-06, "Full-finetune/Loss": 0.7555721998214722, "Full-finetune/Loss (Raw)": 0.7994967103004456, "Full-finetune/Step": 4016, "Full-finetune/Step Time": 6.833544077351689} {"Full-finetune/Learning Rate": 2.5865026059224974e-06, "Full-finetune/Loss": 0.7559301853179932, "Full-finetune/Loss (Raw)": 0.8228755593299866, "Full-finetune/Step": 4017, "Full-finetune/Step Time": 6.8356166034936905} {"Full-finetune/Learning Rate": 2.58232323532438e-06, "Full-finetune/Loss": 0.752480149269104, "Full-finetune/Loss (Raw)": 0.3290553092956543, "Full-finetune/Step": 4018, "Full-finetune/Step Time": 6.854933649301529} {"Full-finetune/Learning Rate": 2.5781467433817973e-06, "Full-finetune/Loss": 0.7545149326324463, "Full-finetune/Loss (Raw)": 0.9324424862861633, "Full-finetune/Step": 4019, "Full-finetune/Step Time": 6.837220335379243} {"Full-finetune/Learning Rate": 2.573973131715565e-06, "Full-finetune/Loss": 0.7551907300949097, "Full-finetune/Loss (Raw)": 0.8823874592781067, "Full-finetune/Step": 4020, "Full-finetune/Step Time": 6.846288681030273} {"Full-finetune/Learning Rate": 2.569802401945378e-06, "Full-finetune/Loss": 0.7584684491157532, "Full-finetune/Loss (Raw)": 0.7938637733459473, "Full-finetune/Step": 4021, "Full-finetune/Step Time": 6.829151453450322} {"Full-finetune/Learning Rate": 2.565634555689818e-06, "Full-finetune/Loss": 0.757972002029419, "Full-finetune/Loss (Raw)": 0.7479463815689087, "Full-finetune/Step": 4022, "Full-finetune/Step Time": 6.831831755116582} {"Full-finetune/Learning Rate": 2.5614695945663415e-06, "Full-finetune/Loss": 0.7581427097320557, "Full-finetune/Loss (Raw)": 0.8150379061698914, "Full-finetune/Step": 4023, "Full-finetune/Step Time": 6.835349950939417} {"Full-finetune/Learning Rate": 2.557307520191291e-06, "Full-finetune/Loss": 0.7588092088699341, "Full-finetune/Loss (Raw)": 0.8683679103851318, "Full-finetune/Step": 4024, "Full-finetune/Step Time": 6.840844701975584} {"Full-finetune/Learning Rate": 2.5531483341798856e-06, "Full-finetune/Loss": 0.7585951089859009, "Full-finetune/Loss (Raw)": 0.724961519241333, "Full-finetune/Step": 4025, "Full-finetune/Step Time": 6.841158600524068} {"Full-finetune/Learning Rate": 2.548992038146224e-06, "Full-finetune/Loss": 0.7595442533493042, "Full-finetune/Loss (Raw)": 0.8446543216705322, "Full-finetune/Step": 4026, "Full-finetune/Step Time": 6.829252464696765} {"Full-finetune/Learning Rate": 2.5448386337032817e-06, "Full-finetune/Loss": 0.7593992948532104, "Full-finetune/Loss (Raw)": 0.8854748010635376, "Full-finetune/Step": 4027, "Full-finetune/Step Time": 6.824766984209418} {"Full-finetune/Learning Rate": 2.5406881224629175e-06, "Full-finetune/Loss": 0.759147047996521, "Full-finetune/Loss (Raw)": 0.7270527482032776, "Full-finetune/Step": 4028, "Full-finetune/Step Time": 6.810861283913255} {"Full-finetune/Learning Rate": 2.5365405060358663e-06, "Full-finetune/Loss": 0.7591259479522705, "Full-finetune/Loss (Raw)": 0.8446063995361328, "Full-finetune/Step": 4029, "Full-finetune/Step Time": 6.811092497780919} {"Full-finetune/Learning Rate": 2.532395786031726e-06, "Full-finetune/Loss": 0.7549695372581482, "Full-finetune/Loss (Raw)": 0.3959990441799164, "Full-finetune/Step": 4030, "Full-finetune/Step Time": 6.826137159019709} {"Full-finetune/Learning Rate": 2.528253964058992e-06, "Full-finetune/Loss": 0.7539402842521667, "Full-finetune/Loss (Raw)": 0.7441355586051941, "Full-finetune/Step": 4031, "Full-finetune/Step Time": 6.828490871936083} {"Full-finetune/Learning Rate": 2.5241150417250216e-06, "Full-finetune/Loss": 0.7537572383880615, "Full-finetune/Loss (Raw)": 0.7429062128067017, "Full-finetune/Step": 4032, "Full-finetune/Step Time": 6.822432404384017} {"Full-finetune/Learning Rate": 2.5199790206360486e-06, "Full-finetune/Loss": 0.7555542588233948, "Full-finetune/Loss (Raw)": 0.8467186093330383, "Full-finetune/Step": 4033, "Full-finetune/Step Time": 6.800230843946338} {"Full-finetune/Learning Rate": 2.5158459023971826e-06, "Full-finetune/Loss": 0.755679726600647, "Full-finetune/Loss (Raw)": 0.7732314467430115, "Full-finetune/Step": 4034, "Full-finetune/Step Time": 6.761883364990354} {"Full-finetune/Learning Rate": 2.511715688612405e-06, "Full-finetune/Loss": 0.7526164650917053, "Full-finetune/Loss (Raw)": 0.6439269185066223, "Full-finetune/Step": 4035, "Full-finetune/Step Time": 6.757460841909051} {"Full-finetune/Learning Rate": 2.5075883808845736e-06, "Full-finetune/Loss": 0.756015419960022, "Full-finetune/Loss (Raw)": 0.9897623062133789, "Full-finetune/Step": 4036, "Full-finetune/Step Time": 6.744962116703391} {"Full-finetune/Learning Rate": 2.5034639808154115e-06, "Full-finetune/Loss": 0.756299614906311, "Full-finetune/Loss (Raw)": 0.6929632425308228, "Full-finetune/Step": 4037, "Full-finetune/Step Time": 6.741358237341046} {"Full-finetune/Learning Rate": 2.4993424900055228e-06, "Full-finetune/Loss": 0.7525179386138916, "Full-finetune/Loss (Raw)": 0.3631438612937927, "Full-finetune/Step": 4038, "Full-finetune/Step Time": 6.757057528942823} {"Full-finetune/Learning Rate": 2.4952239100543808e-06, "Full-finetune/Loss": 0.7511535882949829, "Full-finetune/Loss (Raw)": 0.7540018558502197, "Full-finetune/Step": 4039, "Full-finetune/Step Time": 6.778405075892806} {"Full-finetune/Learning Rate": 2.491108242560313e-06, "Full-finetune/Loss": 0.7525351643562317, "Full-finetune/Loss (Raw)": 0.8528892993927002, "Full-finetune/Step": 4040, "Full-finetune/Step Time": 6.76797790825367} {"Full-finetune/Learning Rate": 2.486995489120542e-06, "Full-finetune/Loss": 0.7519216537475586, "Full-finetune/Loss (Raw)": 0.6980805397033691, "Full-finetune/Step": 4041, "Full-finetune/Step Time": 6.761804010719061} {"Full-finetune/Learning Rate": 2.4828856513311405e-06, "Full-finetune/Loss": 0.7521873116493225, "Full-finetune/Loss (Raw)": 0.7732052803039551, "Full-finetune/Step": 4042, "Full-finetune/Step Time": 6.77552399225533} {"Full-finetune/Learning Rate": 2.4787787307870605e-06, "Full-finetune/Loss": 0.7522810101509094, "Full-finetune/Loss (Raw)": 0.8393036723136902, "Full-finetune/Step": 4043, "Full-finetune/Step Time": 6.777923736721277} {"Full-finetune/Learning Rate": 2.474674729082115e-06, "Full-finetune/Loss": 0.7518186569213867, "Full-finetune/Loss (Raw)": 0.7771779298782349, "Full-finetune/Step": 4044, "Full-finetune/Step Time": 6.769812293350697} {"Full-finetune/Learning Rate": 2.470573647808987e-06, "Full-finetune/Loss": 0.7484457492828369, "Full-finetune/Loss (Raw)": 0.2994804084300995, "Full-finetune/Step": 4045, "Full-finetune/Step Time": 6.76819977350533} {"Full-finetune/Learning Rate": 2.466475488559227e-06, "Full-finetune/Loss": 0.7470979690551758, "Full-finetune/Loss (Raw)": 0.7064935564994812, "Full-finetune/Step": 4046, "Full-finetune/Step Time": 6.766292456537485} {"Full-finetune/Learning Rate": 2.4623802529232476e-06, "Full-finetune/Loss": 0.747033953666687, "Full-finetune/Loss (Raw)": 0.7376654744148254, "Full-finetune/Step": 4047, "Full-finetune/Step Time": 6.768315222114325} {"Full-finetune/Learning Rate": 2.458287942490335e-06, "Full-finetune/Loss": 0.7437553405761719, "Full-finetune/Loss (Raw)": 0.3733338415622711, "Full-finetune/Step": 4048, "Full-finetune/Step Time": 6.751971807330847} {"Full-finetune/Learning Rate": 2.454198558848635e-06, "Full-finetune/Loss": 0.7442470192909241, "Full-finetune/Loss (Raw)": 0.6501547694206238, "Full-finetune/Step": 4049, "Full-finetune/Step Time": 6.7475734781473875} {"Full-finetune/Learning Rate": 2.4501121035851494e-06, "Full-finetune/Loss": 0.743324875831604, "Full-finetune/Loss (Raw)": 0.744583785533905, "Full-finetune/Step": 4050, "Full-finetune/Step Time": 6.746651453897357} {"Full-finetune/Learning Rate": 2.4460285782857606e-06, "Full-finetune/Loss": 0.7448856234550476, "Full-finetune/Loss (Raw)": 0.713882327079773, "Full-finetune/Step": 4051, "Full-finetune/Step Time": 6.737965494394302} {"Full-finetune/Learning Rate": 2.441947984535201e-06, "Full-finetune/Loss": 0.7441139221191406, "Full-finetune/Loss (Raw)": 0.7347872853279114, "Full-finetune/Step": 4052, "Full-finetune/Step Time": 6.757553247734904} {"Full-finetune/Learning Rate": 2.4378703239170707e-06, "Full-finetune/Loss": 0.7405924797058105, "Full-finetune/Loss (Raw)": 0.3770010471343994, "Full-finetune/Step": 4053, "Full-finetune/Step Time": 6.774004558101296} {"Full-finetune/Learning Rate": 2.4337955980138293e-06, "Full-finetune/Loss": 0.7400813698768616, "Full-finetune/Loss (Raw)": 0.6304018497467041, "Full-finetune/Step": 4054, "Full-finetune/Step Time": 6.783670188859105} {"Full-finetune/Learning Rate": 2.4297238084067988e-06, "Full-finetune/Loss": 0.7404504418373108, "Full-finetune/Loss (Raw)": 0.7834385633468628, "Full-finetune/Step": 4055, "Full-finetune/Step Time": 6.7659259252250195} {"Full-finetune/Learning Rate": 2.425654956676161e-06, "Full-finetune/Loss": 0.7381357550621033, "Full-finetune/Loss (Raw)": 0.5610897541046143, "Full-finetune/Step": 4056, "Full-finetune/Step Time": 6.781694313511252} {"Full-finetune/Learning Rate": 2.421589044400955e-06, "Full-finetune/Loss": 0.7385244369506836, "Full-finetune/Loss (Raw)": 0.8789827227592468, "Full-finetune/Step": 4057, "Full-finetune/Step Time": 6.778320694342256} {"Full-finetune/Learning Rate": 2.4175260731590878e-06, "Full-finetune/Loss": 0.7388039231300354, "Full-finetune/Loss (Raw)": 0.7324941754341125, "Full-finetune/Step": 4058, "Full-finetune/Step Time": 6.7701081689447165} {"Full-finetune/Learning Rate": 2.413466044527318e-06, "Full-finetune/Loss": 0.7372639179229736, "Full-finetune/Loss (Raw)": 0.6953249573707581, "Full-finetune/Step": 4059, "Full-finetune/Step Time": 6.784155813977122} {"Full-finetune/Learning Rate": 2.4094089600812576e-06, "Full-finetune/Loss": 0.7353201508522034, "Full-finetune/Loss (Raw)": 0.4423159956932068, "Full-finetune/Step": 4060, "Full-finetune/Step Time": 6.786480838432908} {"Full-finetune/Learning Rate": 2.405354821395388e-06, "Full-finetune/Loss": 0.7357236742973328, "Full-finetune/Loss (Raw)": 0.5763729214668274, "Full-finetune/Step": 4061, "Full-finetune/Step Time": 6.779241289943457} {"Full-finetune/Learning Rate": 2.4013036300430403e-06, "Full-finetune/Loss": 0.7365493774414062, "Full-finetune/Loss (Raw)": 0.8119075894355774, "Full-finetune/Step": 4062, "Full-finetune/Step Time": 6.782573625445366} {"Full-finetune/Learning Rate": 2.3972553875964035e-06, "Full-finetune/Loss": 0.7334485054016113, "Full-finetune/Loss (Raw)": 0.35825666785240173, "Full-finetune/Step": 4063, "Full-finetune/Step Time": 6.789090817794204} {"Full-finetune/Learning Rate": 2.393210095626515e-06, "Full-finetune/Loss": 0.7295534610748291, "Full-finetune/Loss (Raw)": 0.3893250524997711, "Full-finetune/Step": 4064, "Full-finetune/Step Time": 6.797796383500099} {"Full-finetune/Learning Rate": 2.38916775570328e-06, "Full-finetune/Loss": 0.7281216979026794, "Full-finetune/Loss (Raw)": 0.6289366483688354, "Full-finetune/Step": 4065, "Full-finetune/Step Time": 6.8047552444040775} {"Full-finetune/Learning Rate": 2.3851283693954506e-06, "Full-finetune/Loss": 0.7284080386161804, "Full-finetune/Loss (Raw)": 0.7898218035697937, "Full-finetune/Step": 4066, "Full-finetune/Step Time": 6.8095897026360035} {"Full-finetune/Learning Rate": 2.3810919382706297e-06, "Full-finetune/Loss": 0.7285995483398438, "Full-finetune/Loss (Raw)": 0.8524640798568726, "Full-finetune/Step": 4067, "Full-finetune/Step Time": 6.808555889874697} {"Full-finetune/Learning Rate": 2.3770584638952874e-06, "Full-finetune/Loss": 0.7285487651824951, "Full-finetune/Loss (Raw)": 0.8121708631515503, "Full-finetune/Step": 4068, "Full-finetune/Step Time": 6.8107303362339735} {"Full-finetune/Learning Rate": 2.3730279478347275e-06, "Full-finetune/Loss": 0.7285611629486084, "Full-finetune/Loss (Raw)": 0.8492990136146545, "Full-finetune/Step": 4069, "Full-finetune/Step Time": 6.823286879807711} {"Full-finetune/Learning Rate": 2.3690003916531157e-06, "Full-finetune/Loss": 0.7276654839515686, "Full-finetune/Loss (Raw)": 0.7498177289962769, "Full-finetune/Step": 4070, "Full-finetune/Step Time": 6.818598657846451} {"Full-finetune/Learning Rate": 2.3649757969134667e-06, "Full-finetune/Loss": 0.7285716533660889, "Full-finetune/Loss (Raw)": 0.731381356716156, "Full-finetune/Step": 4071, "Full-finetune/Step Time": 6.825219161808491} {"Full-finetune/Learning Rate": 2.3609541651776537e-06, "Full-finetune/Loss": 0.7271859645843506, "Full-finetune/Loss (Raw)": 0.6595502495765686, "Full-finetune/Step": 4072, "Full-finetune/Step Time": 6.8445443622767925} {"Full-finetune/Learning Rate": 2.356935498006391e-06, "Full-finetune/Loss": 0.7255880832672119, "Full-finetune/Loss (Raw)": 0.7554905414581299, "Full-finetune/Step": 4073, "Full-finetune/Step Time": 6.838262030854821} {"Full-finetune/Learning Rate": 2.3529197969592375e-06, "Full-finetune/Loss": 0.7265272736549377, "Full-finetune/Loss (Raw)": 0.8928424715995789, "Full-finetune/Step": 4074, "Full-finetune/Step Time": 6.825850686058402} {"Full-finetune/Learning Rate": 2.3489070635946175e-06, "Full-finetune/Loss": 0.7259663939476013, "Full-finetune/Loss (Raw)": 0.722025990486145, "Full-finetune/Step": 4075, "Full-finetune/Step Time": 6.825592337176204} {"Full-finetune/Learning Rate": 2.3448972994697906e-06, "Full-finetune/Loss": 0.7276480197906494, "Full-finetune/Loss (Raw)": 0.5804519057273865, "Full-finetune/Step": 4076, "Full-finetune/Step Time": 6.8272212743759155} {"Full-finetune/Learning Rate": 2.340890506140869e-06, "Full-finetune/Loss": 0.727000892162323, "Full-finetune/Loss (Raw)": 0.7863813638687134, "Full-finetune/Step": 4077, "Full-finetune/Step Time": 6.8392083793878555} {"Full-finetune/Learning Rate": 2.3368866851628102e-06, "Full-finetune/Loss": 0.7258440256118774, "Full-finetune/Loss (Raw)": 0.7261860370635986, "Full-finetune/Step": 4078, "Full-finetune/Step Time": 6.851678501814604} {"Full-finetune/Learning Rate": 2.3328858380894194e-06, "Full-finetune/Loss": 0.725632905960083, "Full-finetune/Loss (Raw)": 0.8540735840797424, "Full-finetune/Step": 4079, "Full-finetune/Step Time": 6.862186558544636} {"Full-finetune/Learning Rate": 2.328887966473348e-06, "Full-finetune/Loss": 0.7256911993026733, "Full-finetune/Loss (Raw)": 0.6782746911048889, "Full-finetune/Step": 4080, "Full-finetune/Step Time": 6.862711941823363} {"Full-finetune/Learning Rate": 2.324893071866087e-06, "Full-finetune/Loss": 0.7252834439277649, "Full-finetune/Loss (Raw)": 0.7028542757034302, "Full-finetune/Step": 4081, "Full-finetune/Step Time": 6.883395284414291} {"Full-finetune/Learning Rate": 2.320901155817983e-06, "Full-finetune/Loss": 0.7252768278121948, "Full-finetune/Loss (Raw)": 0.7261390089988708, "Full-finetune/Step": 4082, "Full-finetune/Step Time": 6.890561057254672} {"Full-finetune/Learning Rate": 2.316912219878222e-06, "Full-finetune/Loss": 0.7267638444900513, "Full-finetune/Loss (Raw)": 0.8481894731521606, "Full-finetune/Step": 4083, "Full-finetune/Step Time": 6.892772885039449} {"Full-finetune/Learning Rate": 2.312926265594825e-06, "Full-finetune/Loss": 0.7234556674957275, "Full-finetune/Loss (Raw)": 0.47089287638664246, "Full-finetune/Step": 4084, "Full-finetune/Step Time": 6.905712651088834} {"Full-finetune/Learning Rate": 2.308943294514667e-06, "Full-finetune/Loss": 0.7218242883682251, "Full-finetune/Loss (Raw)": 0.6034073829650879, "Full-finetune/Step": 4085, "Full-finetune/Step Time": 6.907756684347987} {"Full-finetune/Learning Rate": 2.3049633081834634e-06, "Full-finetune/Loss": 0.7215089797973633, "Full-finetune/Loss (Raw)": 0.6809962391853333, "Full-finetune/Step": 4086, "Full-finetune/Step Time": 6.9142617247998714} {"Full-finetune/Learning Rate": 2.3009863081457663e-06, "Full-finetune/Loss": 0.7233445048332214, "Full-finetune/Loss (Raw)": 0.8705562949180603, "Full-finetune/Step": 4087, "Full-finetune/Step Time": 6.902316652238369} {"Full-finetune/Learning Rate": 2.2970122959449716e-06, "Full-finetune/Loss": 0.7245219945907593, "Full-finetune/Loss (Raw)": 0.8108875155448914, "Full-finetune/Step": 4088, "Full-finetune/Step Time": 6.88765923678875} {"Full-finetune/Learning Rate": 2.2930412731233174e-06, "Full-finetune/Loss": 0.7244415283203125, "Full-finetune/Loss (Raw)": 0.7846800088882446, "Full-finetune/Step": 4089, "Full-finetune/Step Time": 6.878297653049231} {"Full-finetune/Learning Rate": 2.2890732412218795e-06, "Full-finetune/Loss": 0.7255966067314148, "Full-finetune/Loss (Raw)": 0.7909821271896362, "Full-finetune/Step": 4090, "Full-finetune/Step Time": 6.870634315535426} {"Full-finetune/Learning Rate": 2.2851082017805704e-06, "Full-finetune/Loss": 0.7249293327331543, "Full-finetune/Loss (Raw)": 0.7526381015777588, "Full-finetune/Step": 4091, "Full-finetune/Step Time": 6.883164552971721} {"Full-finetune/Learning Rate": 2.2811461563381497e-06, "Full-finetune/Loss": 0.7243344783782959, "Full-finetune/Loss (Raw)": 0.7309814691543579, "Full-finetune/Step": 4092, "Full-finetune/Step Time": 6.875183217227459} {"Full-finetune/Learning Rate": 2.2771871064322117e-06, "Full-finetune/Loss": 0.7251163125038147, "Full-finetune/Loss (Raw)": 0.876162052154541, "Full-finetune/Step": 4093, "Full-finetune/Step Time": 6.861418420448899} {"Full-finetune/Learning Rate": 2.2732310535991775e-06, "Full-finetune/Loss": 0.7296408414840698, "Full-finetune/Loss (Raw)": 0.9686821699142456, "Full-finetune/Step": 4094, "Full-finetune/Step Time": 6.845695065334439} {"Full-finetune/Learning Rate": 2.2692779993743207e-06, "Full-finetune/Loss": 0.7288565635681152, "Full-finetune/Loss (Raw)": 0.690467357635498, "Full-finetune/Step": 4095, "Full-finetune/Step Time": 6.852596005424857} {"Full-finetune/Learning Rate": 2.265327945291742e-06, "Full-finetune/Loss": 0.7295925617218018, "Full-finetune/Loss (Raw)": 0.9066073894500732, "Full-finetune/Step": 4096, "Full-finetune/Step Time": 6.85783083178103} {"Full-finetune/Learning Rate": 2.2613808928843827e-06, "Full-finetune/Loss": 0.7300258278846741, "Full-finetune/Loss (Raw)": 0.8390798568725586, "Full-finetune/Step": 4097, "Full-finetune/Step Time": 6.844356559216976} {"Full-finetune/Learning Rate": 2.2574368436840144e-06, "Full-finetune/Loss": 0.7280170917510986, "Full-finetune/Loss (Raw)": 0.5931783318519592, "Full-finetune/Step": 4098, "Full-finetune/Step Time": 6.85305504873395} {"Full-finetune/Learning Rate": 2.2534957992212457e-06, "Full-finetune/Loss": 0.7271454334259033, "Full-finetune/Loss (Raw)": 0.795400083065033, "Full-finetune/Step": 4099, "Full-finetune/Step Time": 6.857166849076748} {"Full-finetune/Learning Rate": 2.2495577610255203e-06, "Full-finetune/Loss": 0.7284053564071655, "Full-finetune/Loss (Raw)": 0.918735146522522, "Full-finetune/Step": 4100, "Full-finetune/Step Time": 6.858164645731449} {"Full-finetune/Learning Rate": 2.24562273062511e-06, "Full-finetune/Loss": 0.7282413244247437, "Full-finetune/Loss (Raw)": 0.7584719657897949, "Full-finetune/Step": 4101, "Full-finetune/Step Time": 6.87917690910399} {"Full-finetune/Learning Rate": 2.2416907095471296e-06, "Full-finetune/Loss": 0.7284945845603943, "Full-finetune/Loss (Raw)": 0.7607977986335754, "Full-finetune/Step": 4102, "Full-finetune/Step Time": 6.8766296450048685} {"Full-finetune/Learning Rate": 2.237761699317519e-06, "Full-finetune/Loss": 0.7275699377059937, "Full-finetune/Loss (Raw)": 0.791448712348938, "Full-finetune/Step": 4103, "Full-finetune/Step Time": 6.879575364291668} {"Full-finetune/Learning Rate": 2.2338357014610425e-06, "Full-finetune/Loss": 0.7282416820526123, "Full-finetune/Loss (Raw)": 0.8164118528366089, "Full-finetune/Step": 4104, "Full-finetune/Step Time": 6.868341038003564} {"Full-finetune/Learning Rate": 2.2299127175013113e-06, "Full-finetune/Loss": 0.7285537719726562, "Full-finetune/Loss (Raw)": 0.6445990800857544, "Full-finetune/Step": 4105, "Full-finetune/Step Time": 6.863279787823558} {"Full-finetune/Learning Rate": 2.2259927489607568e-06, "Full-finetune/Loss": 0.7306366562843323, "Full-finetune/Loss (Raw)": 0.7319449782371521, "Full-finetune/Step": 4106, "Full-finetune/Step Time": 6.862577185034752} {"Full-finetune/Learning Rate": 2.222075797360641e-06, "Full-finetune/Loss": 0.7299172878265381, "Full-finetune/Loss (Raw)": 0.7255923748016357, "Full-finetune/Step": 4107, "Full-finetune/Step Time": 6.865477066487074} {"Full-finetune/Learning Rate": 2.2181618642210577e-06, "Full-finetune/Loss": 0.7318215370178223, "Full-finetune/Loss (Raw)": 0.819825291633606, "Full-finetune/Step": 4108, "Full-finetune/Step Time": 6.864443350583315} {"Full-finetune/Learning Rate": 2.2142509510609277e-06, "Full-finetune/Loss": 0.7307772636413574, "Full-finetune/Loss (Raw)": 0.730140209197998, "Full-finetune/Step": 4109, "Full-finetune/Step Time": 6.872840540483594} {"Full-finetune/Learning Rate": 2.210343059397999e-06, "Full-finetune/Loss": 0.7305670976638794, "Full-finetune/Loss (Raw)": 0.7633194327354431, "Full-finetune/Step": 4110, "Full-finetune/Step Time": 6.8800750728696585} {"Full-finetune/Learning Rate": 2.206438190748846e-06, "Full-finetune/Loss": 0.731675922870636, "Full-finetune/Loss (Raw)": 0.9227184653282166, "Full-finetune/Step": 4111, "Full-finetune/Step Time": 6.898751083761454} {"Full-finetune/Learning Rate": 2.2025363466288796e-06, "Full-finetune/Loss": 0.7335978746414185, "Full-finetune/Loss (Raw)": 0.5892505049705505, "Full-finetune/Step": 4112, "Full-finetune/Step Time": 6.893998330458999} {"Full-finetune/Learning Rate": 2.1986375285523233e-06, "Full-finetune/Loss": 0.731987714767456, "Full-finetune/Loss (Raw)": 0.6299495697021484, "Full-finetune/Step": 4113, "Full-finetune/Step Time": 6.891872853040695} {"Full-finetune/Learning Rate": 2.194741738032229e-06, "Full-finetune/Loss": 0.7308835983276367, "Full-finetune/Loss (Raw)": 0.8082804679870605, "Full-finetune/Step": 4114, "Full-finetune/Step Time": 6.894761081784964} {"Full-finetune/Learning Rate": 2.1908489765804853e-06, "Full-finetune/Loss": 0.731698751449585, "Full-finetune/Loss (Raw)": 0.8113967180252075, "Full-finetune/Step": 4115, "Full-finetune/Step Time": 6.890365751460195} {"Full-finetune/Learning Rate": 2.1869592457077924e-06, "Full-finetune/Loss": 0.729305624961853, "Full-finetune/Loss (Raw)": 0.5187556147575378, "Full-finetune/Step": 4116, "Full-finetune/Step Time": 6.906050654128194} {"Full-finetune/Learning Rate": 2.183072546923679e-06, "Full-finetune/Loss": 0.7273691892623901, "Full-finetune/Loss (Raw)": 0.6210160255432129, "Full-finetune/Step": 4117, "Full-finetune/Step Time": 6.915316503494978} {"Full-finetune/Learning Rate": 2.179188881736498e-06, "Full-finetune/Loss": 0.7285211682319641, "Full-finetune/Loss (Raw)": 0.9465550184249878, "Full-finetune/Step": 4118, "Full-finetune/Step Time": 6.907562553882599} {"Full-finetune/Learning Rate": 2.1753082516534242e-06, "Full-finetune/Loss": 0.7288817167282104, "Full-finetune/Loss (Raw)": 0.9014818668365479, "Full-finetune/Step": 4119, "Full-finetune/Step Time": 6.902567101642489} {"Full-finetune/Learning Rate": 2.1714306581804533e-06, "Full-finetune/Loss": 0.7291266918182373, "Full-finetune/Loss (Raw)": 0.6767914295196533, "Full-finetune/Step": 4120, "Full-finetune/Step Time": 6.8982496820390224} {"Full-finetune/Learning Rate": 2.1675561028224012e-06, "Full-finetune/Loss": 0.7289750576019287, "Full-finetune/Loss (Raw)": 0.8018116354942322, "Full-finetune/Step": 4121, "Full-finetune/Step Time": 6.904035400599241} {"Full-finetune/Learning Rate": 2.163684587082915e-06, "Full-finetune/Loss": 0.729176938533783, "Full-finetune/Loss (Raw)": 0.6962478160858154, "Full-finetune/Step": 4122, "Full-finetune/Step Time": 6.913316175341606} {"Full-finetune/Learning Rate": 2.159816112464447e-06, "Full-finetune/Loss": 0.7306722402572632, "Full-finetune/Loss (Raw)": 0.9280573725700378, "Full-finetune/Step": 4123, "Full-finetune/Step Time": 6.904035670682788} {"Full-finetune/Learning Rate": 2.1559506804682762e-06, "Full-finetune/Loss": 0.7295549511909485, "Full-finetune/Loss (Raw)": 0.5572156310081482, "Full-finetune/Step": 4124, "Full-finetune/Step Time": 6.917073231190443} {"Full-finetune/Learning Rate": 2.1520882925945074e-06, "Full-finetune/Loss": 0.7303701043128967, "Full-finetune/Loss (Raw)": 0.8019877672195435, "Full-finetune/Step": 4125, "Full-finetune/Step Time": 6.920867267996073} {"Full-finetune/Learning Rate": 2.1482289503420537e-06, "Full-finetune/Loss": 0.7300207614898682, "Full-finetune/Loss (Raw)": 0.8307681083679199, "Full-finetune/Step": 4126, "Full-finetune/Step Time": 6.914884310215712} {"Full-finetune/Learning Rate": 2.144372655208653e-06, "Full-finetune/Loss": 0.7303645014762878, "Full-finetune/Loss (Raw)": 0.8063114881515503, "Full-finetune/Step": 4127, "Full-finetune/Step Time": 6.895692070946097} {"Full-finetune/Learning Rate": 2.1405194086908555e-06, "Full-finetune/Loss": 0.7301149368286133, "Full-finetune/Loss (Raw)": 0.7431159615516663, "Full-finetune/Step": 4128, "Full-finetune/Step Time": 6.891073713079095} {"Full-finetune/Learning Rate": 2.1366692122840317e-06, "Full-finetune/Loss": 0.7306535243988037, "Full-finetune/Loss (Raw)": 0.7943639159202576, "Full-finetune/Step": 4129, "Full-finetune/Step Time": 6.871962798759341} {"Full-finetune/Learning Rate": 2.1328220674823686e-06, "Full-finetune/Loss": 0.7313472032546997, "Full-finetune/Loss (Raw)": 0.7701172232627869, "Full-finetune/Step": 4130, "Full-finetune/Step Time": 6.862934809178114} {"Full-finetune/Learning Rate": 2.1289779757788655e-06, "Full-finetune/Loss": 0.7316998243331909, "Full-finetune/Loss (Raw)": 0.7853318452835083, "Full-finetune/Step": 4131, "Full-finetune/Step Time": 6.856546411290765} {"Full-finetune/Learning Rate": 2.1251369386653454e-06, "Full-finetune/Loss": 0.7314370274543762, "Full-finetune/Loss (Raw)": 0.7974153161048889, "Full-finetune/Step": 4132, "Full-finetune/Step Time": 6.857367159798741} {"Full-finetune/Learning Rate": 2.121298957632434e-06, "Full-finetune/Loss": 0.7308647632598877, "Full-finetune/Loss (Raw)": 0.7104097604751587, "Full-finetune/Step": 4133, "Full-finetune/Step Time": 6.865473626181483} {"Full-finetune/Learning Rate": 2.117464034169575e-06, "Full-finetune/Loss": 0.7291887998580933, "Full-finetune/Loss (Raw)": 0.7340871691703796, "Full-finetune/Step": 4134, "Full-finetune/Step Time": 6.881705982610583} {"Full-finetune/Learning Rate": 2.113632169765033e-06, "Full-finetune/Loss": 0.7317559719085693, "Full-finetune/Loss (Raw)": 0.6896505355834961, "Full-finetune/Step": 4135, "Full-finetune/Step Time": 6.876822290942073} {"Full-finetune/Learning Rate": 2.1098033659058792e-06, "Full-finetune/Loss": 0.7323641777038574, "Full-finetune/Loss (Raw)": 0.7369127869606018, "Full-finetune/Step": 4136, "Full-finetune/Step Time": 6.87517823278904} {"Full-finetune/Learning Rate": 2.105977624077995e-06, "Full-finetune/Loss": 0.7314996719360352, "Full-finetune/Loss (Raw)": 0.684299647808075, "Full-finetune/Step": 4137, "Full-finetune/Step Time": 6.873076885938644} {"Full-finetune/Learning Rate": 2.102154945766076e-06, "Full-finetune/Loss": 0.7329067587852478, "Full-finetune/Loss (Raw)": 0.8567818999290466, "Full-finetune/Step": 4138, "Full-finetune/Step Time": 6.858723422512412} {"Full-finetune/Learning Rate": 2.0983353324536316e-06, "Full-finetune/Loss": 0.7319371700286865, "Full-finetune/Loss (Raw)": 0.7006859183311462, "Full-finetune/Step": 4139, "Full-finetune/Step Time": 6.848536258563399} {"Full-finetune/Learning Rate": 2.0945187856229778e-06, "Full-finetune/Loss": 0.7332219481468201, "Full-finetune/Loss (Raw)": 0.7156735062599182, "Full-finetune/Step": 4140, "Full-finetune/Step Time": 6.848765341565013} {"Full-finetune/Learning Rate": 2.0907053067552384e-06, "Full-finetune/Loss": 0.730638861656189, "Full-finetune/Loss (Raw)": 0.38882097601890564, "Full-finetune/Step": 4141, "Full-finetune/Step Time": 6.856407416984439} {"Full-finetune/Learning Rate": 2.0868948973303593e-06, "Full-finetune/Loss": 0.7305843830108643, "Full-finetune/Loss (Raw)": 0.7033681273460388, "Full-finetune/Step": 4142, "Full-finetune/Step Time": 6.834935596212745} {"Full-finetune/Learning Rate": 2.0830875588270773e-06, "Full-finetune/Loss": 0.7300938963890076, "Full-finetune/Loss (Raw)": 0.7161920070648193, "Full-finetune/Step": 4143, "Full-finetune/Step Time": 6.820460876449943} {"Full-finetune/Learning Rate": 2.0792832927229465e-06, "Full-finetune/Loss": 0.7260951995849609, "Full-finetune/Loss (Raw)": 0.28766506910324097, "Full-finetune/Step": 4144, "Full-finetune/Step Time": 6.812595259398222} {"Full-finetune/Learning Rate": 2.0754821004943337e-06, "Full-finetune/Loss": 0.7247062921524048, "Full-finetune/Loss (Raw)": 0.6450906991958618, "Full-finetune/Step": 4145, "Full-finetune/Step Time": 6.822512907907367} {"Full-finetune/Learning Rate": 2.071683983616406e-06, "Full-finetune/Loss": 0.7286059856414795, "Full-finetune/Loss (Raw)": 0.8282166719436646, "Full-finetune/Step": 4146, "Full-finetune/Step Time": 6.800309918820858} {"Full-finetune/Learning Rate": 2.0678889435631333e-06, "Full-finetune/Loss": 0.7273655533790588, "Full-finetune/Loss (Raw)": 0.7736639976501465, "Full-finetune/Step": 4147, "Full-finetune/Step Time": 6.804652808234096} {"Full-finetune/Learning Rate": 2.064096981807302e-06, "Full-finetune/Loss": 0.7269895076751709, "Full-finetune/Loss (Raw)": 0.8342552185058594, "Full-finetune/Step": 4148, "Full-finetune/Step Time": 6.794784748926759} {"Full-finetune/Learning Rate": 2.060308099820496e-06, "Full-finetune/Loss": 0.7267310619354248, "Full-finetune/Loss (Raw)": 0.7607804536819458, "Full-finetune/Step": 4149, "Full-finetune/Step Time": 6.784495107829571} {"Full-finetune/Learning Rate": 2.056522299073108e-06, "Full-finetune/Loss": 0.7257696986198425, "Full-finetune/Loss (Raw)": 0.624896228313446, "Full-finetune/Step": 4150, "Full-finetune/Step Time": 6.809180106967688} {"Full-finetune/Learning Rate": 2.052739581034332e-06, "Full-finetune/Loss": 0.725437343120575, "Full-finetune/Loss (Raw)": 0.7724969983100891, "Full-finetune/Step": 4151, "Full-finetune/Step Time": 6.813967816531658} {"Full-finetune/Learning Rate": 2.0489599471721678e-06, "Full-finetune/Loss": 0.7245632410049438, "Full-finetune/Loss (Raw)": 0.7564862966537476, "Full-finetune/Step": 4152, "Full-finetune/Step Time": 6.81908699683845} {"Full-finetune/Learning Rate": 2.0451833989534165e-06, "Full-finetune/Loss": 0.7244757413864136, "Full-finetune/Loss (Raw)": 0.7137603759765625, "Full-finetune/Step": 4153, "Full-finetune/Step Time": 6.816367641091347} {"Full-finetune/Learning Rate": 2.0414099378436812e-06, "Full-finetune/Loss": 0.7241070866584778, "Full-finetune/Loss (Raw)": 0.7974625825881958, "Full-finetune/Step": 4154, "Full-finetune/Step Time": 6.820846490561962} {"Full-finetune/Learning Rate": 2.0376395653073735e-06, "Full-finetune/Loss": 0.7228502035140991, "Full-finetune/Loss (Raw)": 0.7245937585830688, "Full-finetune/Step": 4155, "Full-finetune/Step Time": 6.823677049949765} {"Full-finetune/Learning Rate": 2.0338722828077007e-06, "Full-finetune/Loss": 0.7224752902984619, "Full-finetune/Loss (Raw)": 0.6790642738342285, "Full-finetune/Step": 4156, "Full-finetune/Step Time": 6.822810761630535} {"Full-finetune/Learning Rate": 2.030108091806664e-06, "Full-finetune/Loss": 0.7227426767349243, "Full-finetune/Loss (Raw)": 0.8788272142410278, "Full-finetune/Step": 4157, "Full-finetune/Step Time": 6.8245675172656775} {"Full-finetune/Learning Rate": 2.0263469937650813e-06, "Full-finetune/Loss": 0.7256645560264587, "Full-finetune/Loss (Raw)": 0.7700009346008301, "Full-finetune/Step": 4158, "Full-finetune/Step Time": 6.816282223910093} {"Full-finetune/Learning Rate": 2.0225889901425565e-06, "Full-finetune/Loss": 0.7264349460601807, "Full-finetune/Loss (Raw)": 0.8427486419677734, "Full-finetune/Step": 4159, "Full-finetune/Step Time": 6.814698217436671} {"Full-finetune/Learning Rate": 2.0188340823975005e-06, "Full-finetune/Loss": 0.7275972962379456, "Full-finetune/Loss (Raw)": 0.891692042350769, "Full-finetune/Step": 4160, "Full-finetune/Step Time": 6.808526556938887} {"Full-finetune/Learning Rate": 2.0150822719871167e-06, "Full-finetune/Loss": 0.7271963357925415, "Full-finetune/Loss (Raw)": 0.795390784740448, "Full-finetune/Step": 4161, "Full-finetune/Step Time": 6.807385379448533} {"Full-finetune/Learning Rate": 2.011333560367411e-06, "Full-finetune/Loss": 0.7282423377037048, "Full-finetune/Loss (Raw)": 0.9071227312088013, "Full-finetune/Step": 4162, "Full-finetune/Step Time": 6.809153474867344} {"Full-finetune/Learning Rate": 2.007587948993185e-06, "Full-finetune/Loss": 0.7280395030975342, "Full-finetune/Loss (Raw)": 0.6179618239402771, "Full-finetune/Step": 4163, "Full-finetune/Step Time": 6.820469107478857} {"Full-finetune/Learning Rate": 2.003845439318035e-06, "Full-finetune/Loss": 0.7239785194396973, "Full-finetune/Loss (Raw)": 0.4699600338935852, "Full-finetune/Step": 4164, "Full-finetune/Step Time": 6.832105511799455} {"Full-finetune/Learning Rate": 2.000106032794361e-06, "Full-finetune/Loss": 0.7255215644836426, "Full-finetune/Loss (Raw)": 0.8904623985290527, "Full-finetune/Step": 4165, "Full-finetune/Step Time": 6.826601134613156} {"Full-finetune/Learning Rate": 1.996369730873353e-06, "Full-finetune/Loss": 0.729523777961731, "Full-finetune/Loss (Raw)": 0.875438928604126, "Full-finetune/Step": 4166, "Full-finetune/Step Time": 6.819691205397248} {"Full-finetune/Learning Rate": 1.99263653500499e-06, "Full-finetune/Loss": 0.7292563915252686, "Full-finetune/Loss (Raw)": 0.7197740077972412, "Full-finetune/Step": 4167, "Full-finetune/Step Time": 6.802878065034747} {"Full-finetune/Learning Rate": 1.9889064466380593e-06, "Full-finetune/Loss": 0.7273944616317749, "Full-finetune/Loss (Raw)": 0.6145654320716858, "Full-finetune/Step": 4168, "Full-finetune/Step Time": 6.800710517913103} {"Full-finetune/Learning Rate": 1.985179467220133e-06, "Full-finetune/Loss": 0.7247926592826843, "Full-finetune/Loss (Raw)": 0.3650466501712799, "Full-finetune/Step": 4169, "Full-finetune/Step Time": 6.819008802995086} {"Full-finetune/Learning Rate": 1.9814555981975803e-06, "Full-finetune/Loss": 0.7237456440925598, "Full-finetune/Loss (Raw)": 0.6391817331314087, "Full-finetune/Step": 4170, "Full-finetune/Step Time": 6.812135262414813} {"Full-finetune/Learning Rate": 1.9777348410155606e-06, "Full-finetune/Loss": 0.7235175371170044, "Full-finetune/Loss (Raw)": 0.8101040720939636, "Full-finetune/Step": 4171, "Full-finetune/Step Time": 6.8148188944906} {"Full-finetune/Learning Rate": 1.9740171971180278e-06, "Full-finetune/Loss": 0.7233150005340576, "Full-finetune/Loss (Raw)": 0.7512573003768921, "Full-finetune/Step": 4172, "Full-finetune/Step Time": 6.83436038531363} {"Full-finetune/Learning Rate": 1.9703026679477253e-06, "Full-finetune/Loss": 0.726719856262207, "Full-finetune/Loss (Raw)": 0.7352999448776245, "Full-finetune/Step": 4173, "Full-finetune/Step Time": 6.821936018764973} {"Full-finetune/Learning Rate": 1.9665912549461898e-06, "Full-finetune/Loss": 0.724004864692688, "Full-finetune/Loss (Raw)": 0.35897761583328247, "Full-finetune/Step": 4174, "Full-finetune/Step Time": 6.839380085468292} {"Full-finetune/Learning Rate": 1.962882959553751e-06, "Full-finetune/Loss": 0.7237308025360107, "Full-finetune/Loss (Raw)": 0.702588677406311, "Full-finetune/Step": 4175, "Full-finetune/Step Time": 6.829959366470575} {"Full-finetune/Learning Rate": 1.959177783209527e-06, "Full-finetune/Loss": 0.7266230583190918, "Full-finetune/Loss (Raw)": 0.7435382008552551, "Full-finetune/Step": 4176, "Full-finetune/Step Time": 6.82635760307312} {"Full-finetune/Learning Rate": 1.955475727351417e-06, "Full-finetune/Loss": 0.7277653217315674, "Full-finetune/Loss (Raw)": 0.7963669300079346, "Full-finetune/Step": 4177, "Full-finetune/Step Time": 6.833817010745406} {"Full-finetune/Learning Rate": 1.9517767934161224e-06, "Full-finetune/Loss": 0.7285799384117126, "Full-finetune/Loss (Raw)": 0.8488505482673645, "Full-finetune/Step": 4178, "Full-finetune/Step Time": 6.8352677300572395} {"Full-finetune/Learning Rate": 1.9480809828391266e-06, "Full-finetune/Loss": 0.7287148833274841, "Full-finetune/Loss (Raw)": 0.7311571836471558, "Full-finetune/Step": 4179, "Full-finetune/Step Time": 6.848207568749785} {"Full-finetune/Learning Rate": 1.9443882970547003e-06, "Full-finetune/Loss": 0.7295264005661011, "Full-finetune/Loss (Raw)": 0.8386563658714294, "Full-finetune/Step": 4180, "Full-finetune/Step Time": 6.838951168581843} {"Full-finetune/Learning Rate": 1.940698737495904e-06, "Full-finetune/Loss": 0.7325376868247986, "Full-finetune/Loss (Raw)": 0.7624513506889343, "Full-finetune/Step": 4181, "Full-finetune/Step Time": 6.824660511687398} {"Full-finetune/Learning Rate": 1.937012305594583e-06, "Full-finetune/Loss": 0.7318453192710876, "Full-finetune/Loss (Raw)": 0.5417810082435608, "Full-finetune/Step": 4182, "Full-finetune/Step Time": 6.823407532647252} {"Full-finetune/Learning Rate": 1.93332900278137e-06, "Full-finetune/Loss": 0.7322268486022949, "Full-finetune/Loss (Raw)": 0.832272469997406, "Full-finetune/Step": 4183, "Full-finetune/Step Time": 6.8225389160215855} {"Full-finetune/Learning Rate": 1.9296488304856785e-06, "Full-finetune/Loss": 0.7337242364883423, "Full-finetune/Loss (Raw)": 0.7527574896812439, "Full-finetune/Step": 4184, "Full-finetune/Step Time": 6.800288015976548} {"Full-finetune/Learning Rate": 1.92597179013572e-06, "Full-finetune/Loss": 0.7335511445999146, "Full-finetune/Loss (Raw)": 0.856825053691864, "Full-finetune/Step": 4185, "Full-finetune/Step Time": 6.803418766707182} {"Full-finetune/Learning Rate": 1.9222978831584793e-06, "Full-finetune/Loss": 0.7346819639205933, "Full-finetune/Loss (Raw)": 0.8772326111793518, "Full-finetune/Step": 4186, "Full-finetune/Step Time": 6.8172565922141075} {"Full-finetune/Learning Rate": 1.9186271109797215e-06, "Full-finetune/Loss": 0.7347689867019653, "Full-finetune/Loss (Raw)": 0.7064674496650696, "Full-finetune/Step": 4187, "Full-finetune/Step Time": 6.804320054128766} {"Full-finetune/Learning Rate": 1.9149594750240087e-06, "Full-finetune/Loss": 0.7358506917953491, "Full-finetune/Loss (Raw)": 0.5807788968086243, "Full-finetune/Step": 4188, "Full-finetune/Step Time": 6.794480500742793} {"Full-finetune/Learning Rate": 1.911294976714676e-06, "Full-finetune/Loss": 0.73721843957901, "Full-finetune/Loss (Raw)": 0.7514437437057495, "Full-finetune/Step": 4189, "Full-finetune/Step Time": 6.776752606034279} {"Full-finetune/Learning Rate": 1.9076336174738476e-06, "Full-finetune/Loss": 0.7350221872329712, "Full-finetune/Loss (Raw)": 0.5307815074920654, "Full-finetune/Step": 4190, "Full-finetune/Step Time": 6.776142681017518} {"Full-finetune/Learning Rate": 1.9039753987224164e-06, "Full-finetune/Loss": 0.7381821870803833, "Full-finetune/Loss (Raw)": 0.7627412676811218, "Full-finetune/Step": 4191, "Full-finetune/Step Time": 6.763480691239238} {"Full-finetune/Learning Rate": 1.9003203218800748e-06, "Full-finetune/Loss": 0.7422730922698975, "Full-finetune/Loss (Raw)": 0.9129606485366821, "Full-finetune/Step": 4192, "Full-finetune/Step Time": 6.741769151762128} {"Full-finetune/Learning Rate": 1.8966683883652826e-06, "Full-finetune/Loss": 0.7397667169570923, "Full-finetune/Loss (Raw)": 0.3081168532371521, "Full-finetune/Step": 4193, "Full-finetune/Step Time": 6.747950132936239} {"Full-finetune/Learning Rate": 1.8930195995952827e-06, "Full-finetune/Loss": 0.739976167678833, "Full-finetune/Loss (Raw)": 0.8166270852088928, "Full-finetune/Step": 4194, "Full-finetune/Step Time": 6.759222155436873} {"Full-finetune/Learning Rate": 1.8893739569861058e-06, "Full-finetune/Loss": 0.7399762868881226, "Full-finetune/Loss (Raw)": 0.8524911999702454, "Full-finetune/Step": 4195, "Full-finetune/Step Time": 6.760723892599344} {"Full-finetune/Learning Rate": 1.8857314619525469e-06, "Full-finetune/Loss": 0.7393478155136108, "Full-finetune/Loss (Raw)": 0.7317160964012146, "Full-finetune/Step": 4196, "Full-finetune/Step Time": 6.75675175152719} {"Full-finetune/Learning Rate": 1.8820921159081896e-06, "Full-finetune/Loss": 0.7361732721328735, "Full-finetune/Loss (Raw)": 0.442963570356369, "Full-finetune/Step": 4197, "Full-finetune/Step Time": 6.765201602131128} {"Full-finetune/Learning Rate": 1.8784559202653907e-06, "Full-finetune/Loss": 0.7370253801345825, "Full-finetune/Loss (Raw)": 0.8588899970054626, "Full-finetune/Step": 4198, "Full-finetune/Step Time": 6.764186592772603} {"Full-finetune/Learning Rate": 1.8748228764352915e-06, "Full-finetune/Loss": 0.7385242581367493, "Full-finetune/Loss (Raw)": 0.9232399463653564, "Full-finetune/Step": 4199, "Full-finetune/Step Time": 6.734538527205586} {"Full-finetune/Learning Rate": 1.8711929858278066e-06, "Full-finetune/Loss": 0.7397053837776184, "Full-finetune/Loss (Raw)": 0.8107284903526306, "Full-finetune/Step": 4200, "Full-finetune/Step Time": 6.730151267722249} {"Full-finetune/Learning Rate": 1.8675662498516167e-06, "Full-finetune/Loss": 0.7392397522926331, "Full-finetune/Loss (Raw)": 0.6958926916122437, "Full-finetune/Step": 4201, "Full-finetune/Step Time": 6.72980328835547} {"Full-finetune/Learning Rate": 1.863942669914196e-06, "Full-finetune/Loss": 0.7383760213851929, "Full-finetune/Loss (Raw)": 0.7822771072387695, "Full-finetune/Step": 4202, "Full-finetune/Step Time": 6.732547551393509} {"Full-finetune/Learning Rate": 1.8603222474217819e-06, "Full-finetune/Loss": 0.7383415102958679, "Full-finetune/Loss (Raw)": 0.7176202535629272, "Full-finetune/Step": 4203, "Full-finetune/Step Time": 6.723024647682905} {"Full-finetune/Learning Rate": 1.8567049837793894e-06, "Full-finetune/Loss": 0.7393919825553894, "Full-finetune/Loss (Raw)": 0.7149079442024231, "Full-finetune/Step": 4204, "Full-finetune/Step Time": 6.703304147347808} {"Full-finetune/Learning Rate": 1.8530908803908088e-06, "Full-finetune/Loss": 0.7394925355911255, "Full-finetune/Loss (Raw)": 0.7992587089538574, "Full-finetune/Step": 4205, "Full-finetune/Step Time": 6.685046806931496} {"Full-finetune/Learning Rate": 1.8494799386586015e-06, "Full-finetune/Loss": 0.739251971244812, "Full-finetune/Loss (Raw)": 0.6953818202018738, "Full-finetune/Step": 4206, "Full-finetune/Step Time": 6.674573764204979} {"Full-finetune/Learning Rate": 1.8458721599841056e-06, "Full-finetune/Loss": 0.7388103008270264, "Full-finetune/Loss (Raw)": 0.7975485324859619, "Full-finetune/Step": 4207, "Full-finetune/Step Time": 6.667576616629958} {"Full-finetune/Learning Rate": 1.8422675457674254e-06, "Full-finetune/Loss": 0.7395321130752563, "Full-finetune/Loss (Raw)": 0.7706560492515564, "Full-finetune/Step": 4208, "Full-finetune/Step Time": 6.657156527042389} {"Full-finetune/Learning Rate": 1.8386660974074466e-06, "Full-finetune/Loss": 0.7413054704666138, "Full-finetune/Loss (Raw)": 0.9298518300056458, "Full-finetune/Step": 4209, "Full-finetune/Step Time": 6.637158740311861} {"Full-finetune/Learning Rate": 1.8350678163018221e-06, "Full-finetune/Loss": 0.741645336151123, "Full-finetune/Loss (Raw)": 0.7696438431739807, "Full-finetune/Step": 4210, "Full-finetune/Step Time": 6.632524840533733} {"Full-finetune/Learning Rate": 1.831472703846966e-06, "Full-finetune/Loss": 0.7414746880531311, "Full-finetune/Loss (Raw)": 0.826340913772583, "Full-finetune/Step": 4211, "Full-finetune/Step Time": 6.63819064386189} {"Full-finetune/Learning Rate": 1.827880761438079e-06, "Full-finetune/Loss": 0.7429555654525757, "Full-finetune/Loss (Raw)": 0.66044020652771, "Full-finetune/Step": 4212, "Full-finetune/Step Time": 6.619106089696288} {"Full-finetune/Learning Rate": 1.8242919904691214e-06, "Full-finetune/Loss": 0.7430702447891235, "Full-finetune/Loss (Raw)": 0.6181005835533142, "Full-finetune/Step": 4213, "Full-finetune/Step Time": 6.616434762254357} {"Full-finetune/Learning Rate": 1.820706392332824e-06, "Full-finetune/Loss": 0.7438111305236816, "Full-finetune/Loss (Raw)": 0.7758170962333679, "Full-finetune/Step": 4214, "Full-finetune/Step Time": 6.607434364035726} {"Full-finetune/Learning Rate": 1.8171239684206898e-06, "Full-finetune/Loss": 0.7424437999725342, "Full-finetune/Loss (Raw)": 0.6955406665802002, "Full-finetune/Step": 4215, "Full-finetune/Step Time": 6.633381336927414} {"Full-finetune/Learning Rate": 1.8135447201229861e-06, "Full-finetune/Loss": 0.7423669099807739, "Full-finetune/Loss (Raw)": 0.8010455369949341, "Full-finetune/Step": 4216, "Full-finetune/Step Time": 6.628280317410827} {"Full-finetune/Learning Rate": 1.8099686488287483e-06, "Full-finetune/Loss": 0.7426351308822632, "Full-finetune/Loss (Raw)": 0.8190168142318726, "Full-finetune/Step": 4217, "Full-finetune/Step Time": 6.62530929222703} {"Full-finetune/Learning Rate": 1.8063957559257794e-06, "Full-finetune/Loss": 0.7424249649047852, "Full-finetune/Loss (Raw)": 0.7640798091888428, "Full-finetune/Step": 4218, "Full-finetune/Step Time": 6.625591687858105} {"Full-finetune/Learning Rate": 1.8028260428006538e-06, "Full-finetune/Loss": 0.7409976720809937, "Full-finetune/Loss (Raw)": 0.5699429512023926, "Full-finetune/Step": 4219, "Full-finetune/Step Time": 6.608434967696667} {"Full-finetune/Learning Rate": 1.7992595108387068e-06, "Full-finetune/Loss": 0.7416286468505859, "Full-finetune/Loss (Raw)": 0.8117469549179077, "Full-finetune/Step": 4220, "Full-finetune/Step Time": 6.597770405933261} {"Full-finetune/Learning Rate": 1.795696161424032e-06, "Full-finetune/Loss": 0.7417627573013306, "Full-finetune/Loss (Raw)": 0.8933281302452087, "Full-finetune/Step": 4221, "Full-finetune/Step Time": 6.599339691922069} {"Full-finetune/Learning Rate": 1.7921359959395046e-06, "Full-finetune/Loss": 0.7406226396560669, "Full-finetune/Loss (Raw)": 0.8227450251579285, "Full-finetune/Step": 4222, "Full-finetune/Step Time": 6.603514859452844} {"Full-finetune/Learning Rate": 1.7885790157667526e-06, "Full-finetune/Loss": 0.7419428825378418, "Full-finetune/Loss (Raw)": 0.8594588041305542, "Full-finetune/Step": 4223, "Full-finetune/Step Time": 6.594269186258316} {"Full-finetune/Learning Rate": 1.7850252222861697e-06, "Full-finetune/Loss": 0.7404384613037109, "Full-finetune/Loss (Raw)": 0.7140458822250366, "Full-finetune/Step": 4224, "Full-finetune/Step Time": 6.599275551736355} {"Full-finetune/Learning Rate": 1.7814746168769138e-06, "Full-finetune/Loss": 0.7392216920852661, "Full-finetune/Loss (Raw)": 0.6833310127258301, "Full-finetune/Step": 4225, "Full-finetune/Step Time": 6.591924166306853} {"Full-finetune/Learning Rate": 1.7779272009169068e-06, "Full-finetune/Loss": 0.7414170503616333, "Full-finetune/Loss (Raw)": 0.8741887807846069, "Full-finetune/Step": 4226, "Full-finetune/Step Time": 6.586531117558479} {"Full-finetune/Learning Rate": 1.77438297578283e-06, "Full-finetune/Loss": 0.7428171634674072, "Full-finetune/Loss (Raw)": 0.9746111631393433, "Full-finetune/Step": 4227, "Full-finetune/Step Time": 6.596284838393331} {"Full-finetune/Learning Rate": 1.7708419428501256e-06, "Full-finetune/Loss": 0.7417453527450562, "Full-finetune/Loss (Raw)": 0.7815365791320801, "Full-finetune/Step": 4228, "Full-finetune/Step Time": 6.598056513816118} {"Full-finetune/Learning Rate": 1.7673041034930083e-06, "Full-finetune/Loss": 0.7424103021621704, "Full-finetune/Loss (Raw)": 0.8435971140861511, "Full-finetune/Step": 4229, "Full-finetune/Step Time": 6.600734557956457} {"Full-finetune/Learning Rate": 1.7637694590844357e-06, "Full-finetune/Loss": 0.7385631799697876, "Full-finetune/Loss (Raw)": 0.26836735010147095, "Full-finetune/Step": 4230, "Full-finetune/Step Time": 6.612588666379452} {"Full-finetune/Learning Rate": 1.7602380109961348e-06, "Full-finetune/Loss": 0.7367063164710999, "Full-finetune/Loss (Raw)": 0.5537608861923218, "Full-finetune/Step": 4231, "Full-finetune/Step Time": 6.651305081322789} {"Full-finetune/Learning Rate": 1.7567097605985972e-06, "Full-finetune/Loss": 0.7364020347595215, "Full-finetune/Loss (Raw)": 0.7774587273597717, "Full-finetune/Step": 4232, "Full-finetune/Step Time": 6.659305987879634} {"Full-finetune/Learning Rate": 1.7531847092610644e-06, "Full-finetune/Loss": 0.7362785935401917, "Full-finetune/Loss (Raw)": 0.6288033723831177, "Full-finetune/Step": 4233, "Full-finetune/Step Time": 6.664907434955239} {"Full-finetune/Learning Rate": 1.7496628583515395e-06, "Full-finetune/Loss": 0.7379887104034424, "Full-finetune/Loss (Raw)": 0.9508450627326965, "Full-finetune/Step": 4234, "Full-finetune/Step Time": 6.667927676811814} {"Full-finetune/Learning Rate": 1.7461442092367864e-06, "Full-finetune/Loss": 0.7386952638626099, "Full-finetune/Loss (Raw)": 0.8160256147384644, "Full-finetune/Step": 4235, "Full-finetune/Step Time": 6.666113944724202} {"Full-finetune/Learning Rate": 1.7426287632823213e-06, "Full-finetune/Loss": 0.7371798753738403, "Full-finetune/Loss (Raw)": 0.6258610486984253, "Full-finetune/Step": 4236, "Full-finetune/Step Time": 6.674694735556841} {"Full-finetune/Learning Rate": 1.7391165218524232e-06, "Full-finetune/Loss": 0.7385032773017883, "Full-finetune/Loss (Raw)": 0.8995374441146851, "Full-finetune/Step": 4237, "Full-finetune/Step Time": 6.666686717420816} {"Full-finetune/Learning Rate": 1.735607486310119e-06, "Full-finetune/Loss": 0.7392181158065796, "Full-finetune/Loss (Raw)": 0.8548161387443542, "Full-finetune/Step": 4238, "Full-finetune/Step Time": 6.655179901048541} {"Full-finetune/Learning Rate": 1.7321016580172068e-06, "Full-finetune/Loss": 0.7384084463119507, "Full-finetune/Loss (Raw)": 0.8190776109695435, "Full-finetune/Step": 4239, "Full-finetune/Step Time": 6.639005001634359} {"Full-finetune/Learning Rate": 1.728599038334221e-06, "Full-finetune/Loss": 0.739769458770752, "Full-finetune/Loss (Raw)": 0.7634634971618652, "Full-finetune/Step": 4240, "Full-finetune/Step Time": 6.6282127015292645} {"Full-finetune/Learning Rate": 1.7250996286204602e-06, "Full-finetune/Loss": 0.7409816384315491, "Full-finetune/Loss (Raw)": 0.7851078510284424, "Full-finetune/Step": 4241, "Full-finetune/Step Time": 6.6289735566824675} {"Full-finetune/Learning Rate": 1.7216034302339824e-06, "Full-finetune/Loss": 0.741289496421814, "Full-finetune/Loss (Raw)": 0.8476859331130981, "Full-finetune/Step": 4242, "Full-finetune/Step Time": 6.626691857352853} {"Full-finetune/Learning Rate": 1.718110444531591e-06, "Full-finetune/Loss": 0.7406450510025024, "Full-finetune/Loss (Raw)": 0.728912353515625, "Full-finetune/Step": 4243, "Full-finetune/Step Time": 6.647411024197936} {"Full-finetune/Learning Rate": 1.7146206728688465e-06, "Full-finetune/Loss": 0.7429360151290894, "Full-finetune/Loss (Raw)": 0.811993420124054, "Full-finetune/Step": 4244, "Full-finetune/Step Time": 6.631375698372722} {"Full-finetune/Learning Rate": 1.7111341166000594e-06, "Full-finetune/Loss": 0.7442965507507324, "Full-finetune/Loss (Raw)": 0.7951657176017761, "Full-finetune/Step": 4245, "Full-finetune/Step Time": 6.6342668402940035} {"Full-finetune/Learning Rate": 1.7076507770782957e-06, "Full-finetune/Loss": 0.7425166368484497, "Full-finetune/Loss (Raw)": 0.7187279462814331, "Full-finetune/Step": 4246, "Full-finetune/Step Time": 6.669476293027401} {"Full-finetune/Learning Rate": 1.7041706556553717e-06, "Full-finetune/Loss": 0.7419635057449341, "Full-finetune/Loss (Raw)": 0.8306833505630493, "Full-finetune/Step": 4247, "Full-finetune/Step Time": 6.671090882271528} {"Full-finetune/Learning Rate": 1.7006937536818503e-06, "Full-finetune/Loss": 0.7389557957649231, "Full-finetune/Loss (Raw)": 0.291795015335083, "Full-finetune/Step": 4248, "Full-finetune/Step Time": 6.675999203696847} {"Full-finetune/Learning Rate": 1.6972200725070587e-06, "Full-finetune/Loss": 0.7392460107803345, "Full-finetune/Loss (Raw)": 0.8389639854431152, "Full-finetune/Step": 4249, "Full-finetune/Step Time": 6.6705456636846066} {"Full-finetune/Learning Rate": 1.6937496134790554e-06, "Full-finetune/Loss": 0.7383263111114502, "Full-finetune/Loss (Raw)": 0.5785248875617981, "Full-finetune/Step": 4250, "Full-finetune/Step Time": 6.656887700781226} {"Full-finetune/Learning Rate": 1.6902823779446586e-06, "Full-finetune/Loss": 0.736362636089325, "Full-finetune/Loss (Raw)": 0.6767096519470215, "Full-finetune/Step": 4251, "Full-finetune/Step Time": 6.656541503965855} {"Full-finetune/Learning Rate": 1.686818367249441e-06, "Full-finetune/Loss": 0.7396445274353027, "Full-finetune/Loss (Raw)": 0.9772961735725403, "Full-finetune/Step": 4252, "Full-finetune/Step Time": 6.6582393273711205} {"Full-finetune/Learning Rate": 1.6833575827377135e-06, "Full-finetune/Loss": 0.7375829219818115, "Full-finetune/Loss (Raw)": 0.5381023287773132, "Full-finetune/Step": 4253, "Full-finetune/Step Time": 6.678311081603169} {"Full-finetune/Learning Rate": 1.6799000257525388e-06, "Full-finetune/Loss": 0.7347635626792908, "Full-finetune/Loss (Raw)": 0.46988606452941895, "Full-finetune/Step": 4254, "Full-finetune/Step Time": 6.688553394749761} {"Full-finetune/Learning Rate": 1.6764456976357279e-06, "Full-finetune/Loss": 0.733776330947876, "Full-finetune/Loss (Raw)": 0.6799531579017639, "Full-finetune/Step": 4255, "Full-finetune/Step Time": 6.694724595174193} {"Full-finetune/Learning Rate": 1.6729945997278386e-06, "Full-finetune/Loss": 0.7341387271881104, "Full-finetune/Loss (Raw)": 0.7894997596740723, "Full-finetune/Step": 4256, "Full-finetune/Step Time": 6.7105446718633175} {"Full-finetune/Learning Rate": 1.6695467333681748e-06, "Full-finetune/Loss": 0.7343485355377197, "Full-finetune/Loss (Raw)": 0.8212189674377441, "Full-finetune/Step": 4257, "Full-finetune/Step Time": 6.715206816792488} {"Full-finetune/Learning Rate": 1.6661020998947818e-06, "Full-finetune/Loss": 0.73314368724823, "Full-finetune/Loss (Raw)": 0.6158875823020935, "Full-finetune/Step": 4258, "Full-finetune/Step Time": 6.725667366757989} {"Full-finetune/Learning Rate": 1.6626607006444628e-06, "Full-finetune/Loss": 0.7339732646942139, "Full-finetune/Loss (Raw)": 0.8915286064147949, "Full-finetune/Step": 4259, "Full-finetune/Step Time": 6.729414205998182} {"Full-finetune/Learning Rate": 1.6592225369527515e-06, "Full-finetune/Loss": 0.73374342918396, "Full-finetune/Loss (Raw)": 0.7679874300956726, "Full-finetune/Step": 4260, "Full-finetune/Step Time": 6.72612351924181} {"Full-finetune/Learning Rate": 1.6557876101539305e-06, "Full-finetune/Loss": 0.7344425320625305, "Full-finetune/Loss (Raw)": 0.7999052405357361, "Full-finetune/Step": 4261, "Full-finetune/Step Time": 6.7200284991413355} {"Full-finetune/Learning Rate": 1.6523559215810336e-06, "Full-finetune/Loss": 0.7345072031021118, "Full-finetune/Loss (Raw)": 0.7423601150512695, "Full-finetune/Step": 4262, "Full-finetune/Step Time": 6.710237588733435} {"Full-finetune/Learning Rate": 1.64892747256583e-06, "Full-finetune/Loss": 0.7364410161972046, "Full-finetune/Loss (Raw)": 0.9371839165687561, "Full-finetune/Step": 4263, "Full-finetune/Step Time": 6.69670370221138} {"Full-finetune/Learning Rate": 1.645502264438833e-06, "Full-finetune/Loss": 0.7355133295059204, "Full-finetune/Loss (Raw)": 0.6181652545928955, "Full-finetune/Step": 4264, "Full-finetune/Step Time": 6.70822493173182} {"Full-finetune/Learning Rate": 1.6420802985292995e-06, "Full-finetune/Loss": 0.7354779839515686, "Full-finetune/Loss (Raw)": 0.6797753572463989, "Full-finetune/Step": 4265, "Full-finetune/Step Time": 6.715403692796826} {"Full-finetune/Learning Rate": 1.638661576165227e-06, "Full-finetune/Loss": 0.7358771562576294, "Full-finetune/Loss (Raw)": 0.9078741073608398, "Full-finetune/Step": 4266, "Full-finetune/Step Time": 6.714941198006272} {"Full-finetune/Learning Rate": 1.6352460986733564e-06, "Full-finetune/Loss": 0.735367476940155, "Full-finetune/Loss (Raw)": 0.6354472041130066, "Full-finetune/Step": 4267, "Full-finetune/Step Time": 6.7286947052925825} {"Full-finetune/Learning Rate": 1.6318338673791655e-06, "Full-finetune/Loss": 0.7352979183197021, "Full-finetune/Loss (Raw)": 0.7067708373069763, "Full-finetune/Step": 4268, "Full-finetune/Step Time": 6.7091507110744715} {"Full-finetune/Learning Rate": 1.6284248836068772e-06, "Full-finetune/Loss": 0.7365221977233887, "Full-finetune/Loss (Raw)": 0.5455320477485657, "Full-finetune/Step": 4269, "Full-finetune/Step Time": 6.727060118690133} {"Full-finetune/Learning Rate": 1.6250191486794497e-06, "Full-finetune/Loss": 0.7370685338973999, "Full-finetune/Loss (Raw)": 0.7732890248298645, "Full-finetune/Step": 4270, "Full-finetune/Step Time": 6.725976679474115} {"Full-finetune/Learning Rate": 1.6216166639185804e-06, "Full-finetune/Loss": 0.7340610027313232, "Full-finetune/Loss (Raw)": 0.33123162388801575, "Full-finetune/Step": 4271, "Full-finetune/Step Time": 6.744574004784226} {"Full-finetune/Learning Rate": 1.618217430644713e-06, "Full-finetune/Loss": 0.7382524013519287, "Full-finetune/Loss (Raw)": 0.8241639733314514, "Full-finetune/Step": 4272, "Full-finetune/Step Time": 6.729519121348858} {"Full-finetune/Learning Rate": 1.6148214501770221e-06, "Full-finetune/Loss": 0.7385855913162231, "Full-finetune/Loss (Raw)": 0.687739908695221, "Full-finetune/Step": 4273, "Full-finetune/Step Time": 6.716493336483836} {"Full-finetune/Learning Rate": 1.6114287238334148e-06, "Full-finetune/Loss": 0.7379523515701294, "Full-finetune/Loss (Raw)": 0.747158408164978, "Full-finetune/Step": 4274, "Full-finetune/Step Time": 6.722267704084516} {"Full-finetune/Learning Rate": 1.6080392529305489e-06, "Full-finetune/Loss": 0.7366299033164978, "Full-finetune/Loss (Raw)": 0.6043959259986877, "Full-finetune/Step": 4275, "Full-finetune/Step Time": 6.733201598748565} {"Full-finetune/Learning Rate": 1.60465303878381e-06, "Full-finetune/Loss": 0.7365280389785767, "Full-finetune/Loss (Raw)": 0.821218729019165, "Full-finetune/Step": 4276, "Full-finetune/Step Time": 6.734546149149537} {"Full-finetune/Learning Rate": 1.6012700827073213e-06, "Full-finetune/Loss": 0.7360458374023438, "Full-finetune/Loss (Raw)": 0.6990619897842407, "Full-finetune/Step": 4277, "Full-finetune/Step Time": 6.756548311561346} {"Full-finetune/Learning Rate": 1.5978903860139416e-06, "Full-finetune/Loss": 0.7378131747245789, "Full-finetune/Loss (Raw)": 0.8511109352111816, "Full-finetune/Step": 4278, "Full-finetune/Step Time": 6.730580378323793} {"Full-finetune/Learning Rate": 1.5945139500152662e-06, "Full-finetune/Loss": 0.7372129559516907, "Full-finetune/Loss (Raw)": 0.6956680417060852, "Full-finetune/Step": 4279, "Full-finetune/Step Time": 6.728380249813199} {"Full-finetune/Learning Rate": 1.5911407760216235e-06, "Full-finetune/Loss": 0.7377787232398987, "Full-finetune/Loss (Raw)": 0.8289085030555725, "Full-finetune/Step": 4280, "Full-finetune/Step Time": 6.734034081920981} {"Full-finetune/Learning Rate": 1.5877708653420742e-06, "Full-finetune/Loss": 0.7393138408660889, "Full-finetune/Loss (Raw)": 0.9102469086647034, "Full-finetune/Step": 4281, "Full-finetune/Step Time": 6.73976799659431} {"Full-finetune/Learning Rate": 1.58440421928442e-06, "Full-finetune/Loss": 0.739609956741333, "Full-finetune/Loss (Raw)": 0.835365355014801, "Full-finetune/Step": 4282, "Full-finetune/Step Time": 6.736961690708995} {"Full-finetune/Learning Rate": 1.581040839155189e-06, "Full-finetune/Loss": 0.739920437335968, "Full-finetune/Loss (Raw)": 0.7643409967422485, "Full-finetune/Step": 4283, "Full-finetune/Step Time": 6.741495696827769} {"Full-finetune/Learning Rate": 1.5776807262596371e-06, "Full-finetune/Loss": 0.7408580780029297, "Full-finetune/Loss (Raw)": 0.7990840077400208, "Full-finetune/Step": 4284, "Full-finetune/Step Time": 6.739799654111266} {"Full-finetune/Learning Rate": 1.574323881901767e-06, "Full-finetune/Loss": 0.7386753559112549, "Full-finetune/Loss (Raw)": 0.5994451642036438, "Full-finetune/Step": 4285, "Full-finetune/Step Time": 6.783985607326031} {"Full-finetune/Learning Rate": 1.5709703073843008e-06, "Full-finetune/Loss": 0.7385329604148865, "Full-finetune/Loss (Raw)": 0.7517671585083008, "Full-finetune/Step": 4286, "Full-finetune/Step Time": 6.775309573858976} {"Full-finetune/Learning Rate": 1.5676200040086941e-06, "Full-finetune/Loss": 0.7383041381835938, "Full-finetune/Loss (Raw)": 0.8134651184082031, "Full-finetune/Step": 4287, "Full-finetune/Step Time": 6.788118980824947} {"Full-finetune/Learning Rate": 1.564272973075136e-06, "Full-finetune/Loss": 0.7382692694664001, "Full-finetune/Loss (Raw)": 0.8872266411781311, "Full-finetune/Step": 4288, "Full-finetune/Step Time": 6.801883563399315} {"Full-finetune/Learning Rate": 1.560929215882544e-06, "Full-finetune/Loss": 0.7377904653549194, "Full-finetune/Loss (Raw)": 0.7341007590293884, "Full-finetune/Step": 4289, "Full-finetune/Step Time": 6.802578218281269} {"Full-finetune/Learning Rate": 1.5575887337285644e-06, "Full-finetune/Loss": 0.7375158071517944, "Full-finetune/Loss (Raw)": 0.8719648122787476, "Full-finetune/Step": 4290, "Full-finetune/Step Time": 6.804219914600253} {"Full-finetune/Learning Rate": 1.5542515279095704e-06, "Full-finetune/Loss": 0.7395927309989929, "Full-finetune/Loss (Raw)": 0.8838043808937073, "Full-finetune/Step": 4291, "Full-finetune/Step Time": 6.788971295580268} {"Full-finetune/Learning Rate": 1.5509175997206705e-06, "Full-finetune/Loss": 0.7424743175506592, "Full-finetune/Loss (Raw)": 0.8388068675994873, "Full-finetune/Step": 4292, "Full-finetune/Step Time": 6.7848669197410345} {"Full-finetune/Learning Rate": 1.5475869504556996e-06, "Full-finetune/Loss": 0.7418781518936157, "Full-finetune/Loss (Raw)": 0.8141558170318604, "Full-finetune/Step": 4293, "Full-finetune/Step Time": 6.791936539113522} {"Full-finetune/Learning Rate": 1.5442595814072081e-06, "Full-finetune/Loss": 0.7409614324569702, "Full-finetune/Loss (Raw)": 0.7580975890159607, "Full-finetune/Step": 4294, "Full-finetune/Step Time": 6.803697543218732} {"Full-finetune/Learning Rate": 1.540935493866491e-06, "Full-finetune/Loss": 0.7377208471298218, "Full-finetune/Loss (Raw)": 0.30498236417770386, "Full-finetune/Step": 4295, "Full-finetune/Step Time": 6.8195089139044285} {"Full-finetune/Learning Rate": 1.53761468912356e-06, "Full-finetune/Loss": 0.7400710582733154, "Full-finetune/Loss (Raw)": 0.9153894782066345, "Full-finetune/Step": 4296, "Full-finetune/Step Time": 6.818497641012073} {"Full-finetune/Learning Rate": 1.5342971684671537e-06, "Full-finetune/Loss": 0.7437820434570312, "Full-finetune/Loss (Raw)": 0.8400510549545288, "Full-finetune/Step": 4297, "Full-finetune/Step Time": 6.804168000817299} {"Full-finetune/Learning Rate": 1.530982933184737e-06, "Full-finetune/Loss": 0.7433987855911255, "Full-finetune/Loss (Raw)": 0.5901195406913757, "Full-finetune/Step": 4298, "Full-finetune/Step Time": 6.797793265432119} {"Full-finetune/Learning Rate": 1.527671984562501e-06, "Full-finetune/Loss": 0.7433589696884155, "Full-finetune/Loss (Raw)": 0.8050141930580139, "Full-finetune/Step": 4299, "Full-finetune/Step Time": 6.8039423525333405} {"Full-finetune/Learning Rate": 1.524364323885359e-06, "Full-finetune/Loss": 0.7444012761116028, "Full-finetune/Loss (Raw)": 0.8846748471260071, "Full-finetune/Step": 4300, "Full-finetune/Step Time": 6.7894705273211} {"Full-finetune/Learning Rate": 1.5210599524369484e-06, "Full-finetune/Loss": 0.7415099143981934, "Full-finetune/Loss (Raw)": 0.3651975989341736, "Full-finetune/Step": 4301, "Full-finetune/Step Time": 6.801968617364764} {"Full-finetune/Learning Rate": 1.517758871499635e-06, "Full-finetune/Loss": 0.7456189393997192, "Full-finetune/Loss (Raw)": 0.8849334120750427, "Full-finetune/Step": 4302, "Full-finetune/Step Time": 6.789531596004963} {"Full-finetune/Learning Rate": 1.5144610823545035e-06, "Full-finetune/Loss": 0.7472097277641296, "Full-finetune/Loss (Raw)": 0.9062147736549377, "Full-finetune/Step": 4303, "Full-finetune/Step Time": 6.7882288340479136} {"Full-finetune/Learning Rate": 1.511166586281355e-06, "Full-finetune/Loss": 0.744012176990509, "Full-finetune/Loss (Raw)": 0.334254652261734, "Full-finetune/Step": 4304, "Full-finetune/Step Time": 6.792042654007673} {"Full-finetune/Learning Rate": 1.507875384558727e-06, "Full-finetune/Loss": 0.7427997589111328, "Full-finetune/Loss (Raw)": 0.6411738991737366, "Full-finetune/Step": 4305, "Full-finetune/Step Time": 6.775419680401683} {"Full-finetune/Learning Rate": 1.504587478463867e-06, "Full-finetune/Loss": 0.7417568564414978, "Full-finetune/Loss (Raw)": 0.7153574824333191, "Full-finetune/Step": 4306, "Full-finetune/Step Time": 6.781187683343887} {"Full-finetune/Learning Rate": 1.5013028692727483e-06, "Full-finetune/Loss": 0.740516185760498, "Full-finetune/Loss (Raw)": 0.5723505020141602, "Full-finetune/Step": 4307, "Full-finetune/Step Time": 6.77056691609323} {"Full-finetune/Learning Rate": 1.4980215582600621e-06, "Full-finetune/Loss": 0.7392607927322388, "Full-finetune/Loss (Raw)": 0.6779703497886658, "Full-finetune/Step": 4308, "Full-finetune/Step Time": 6.759846301749349} {"Full-finetune/Learning Rate": 1.4947435466992222e-06, "Full-finetune/Loss": 0.7390167117118835, "Full-finetune/Loss (Raw)": 0.7312076091766357, "Full-finetune/Step": 4309, "Full-finetune/Step Time": 6.763865815475583} {"Full-finetune/Learning Rate": 1.4914688358623608e-06, "Full-finetune/Loss": 0.7375253438949585, "Full-finetune/Loss (Raw)": 0.3508831262588501, "Full-finetune/Step": 4310, "Full-finetune/Step Time": 6.771331813186407} {"Full-finetune/Learning Rate": 1.488197427020327e-06, "Full-finetune/Loss": 0.7371753454208374, "Full-finetune/Loss (Raw)": 0.7874709963798523, "Full-finetune/Step": 4311, "Full-finetune/Step Time": 6.779108824208379} {"Full-finetune/Learning Rate": 1.4849293214426975e-06, "Full-finetune/Loss": 0.7379621267318726, "Full-finetune/Loss (Raw)": 0.8534692525863647, "Full-finetune/Step": 4312, "Full-finetune/Step Time": 6.786480190232396} {"Full-finetune/Learning Rate": 1.481664520397753e-06, "Full-finetune/Loss": 0.7381910085678101, "Full-finetune/Loss (Raw)": 0.8861207365989685, "Full-finetune/Step": 4313, "Full-finetune/Step Time": 6.806497288867831} {"Full-finetune/Learning Rate": 1.4784030251524994e-06, "Full-finetune/Loss": 0.7379772067070007, "Full-finetune/Loss (Raw)": 0.8498618006706238, "Full-finetune/Step": 4314, "Full-finetune/Step Time": 6.795734103769064} {"Full-finetune/Learning Rate": 1.4751448369726652e-06, "Full-finetune/Loss": 0.7383530139923096, "Full-finetune/Loss (Raw)": 0.7545790672302246, "Full-finetune/Step": 4315, "Full-finetune/Step Time": 6.793433660641313} {"Full-finetune/Learning Rate": 1.4718899571226842e-06, "Full-finetune/Loss": 0.739063560962677, "Full-finetune/Loss (Raw)": 0.6717280149459839, "Full-finetune/Step": 4316, "Full-finetune/Step Time": 6.806474193930626} {"Full-finetune/Learning Rate": 1.468638386865716e-06, "Full-finetune/Loss": 0.7392077445983887, "Full-finetune/Loss (Raw)": 0.7698920369148254, "Full-finetune/Step": 4317, "Full-finetune/Step Time": 6.804435273632407} {"Full-finetune/Learning Rate": 1.465390127463625e-06, "Full-finetune/Loss": 0.7397711277008057, "Full-finetune/Loss (Raw)": 0.6028989553451538, "Full-finetune/Step": 4318, "Full-finetune/Step Time": 6.808242868632078} {"Full-finetune/Learning Rate": 1.4621451801770038e-06, "Full-finetune/Loss": 0.7382853031158447, "Full-finetune/Loss (Raw)": 0.572553813457489, "Full-finetune/Step": 4319, "Full-finetune/Step Time": 6.812676278874278} {"Full-finetune/Learning Rate": 1.4589035462651502e-06, "Full-finetune/Loss": 0.736896812915802, "Full-finetune/Loss (Raw)": 0.735235869884491, "Full-finetune/Step": 4320, "Full-finetune/Step Time": 6.816585928201675} {"Full-finetune/Learning Rate": 1.4556652269860772e-06, "Full-finetune/Loss": 0.739044725894928, "Full-finetune/Loss (Raw)": 0.5830488801002502, "Full-finetune/Step": 4321, "Full-finetune/Step Time": 6.811060084030032} {"Full-finetune/Learning Rate": 1.4524302235965203e-06, "Full-finetune/Loss": 0.7395015358924866, "Full-finetune/Loss (Raw)": 0.8750964403152466, "Full-finetune/Step": 4322, "Full-finetune/Step Time": 6.8020436353981495} {"Full-finetune/Learning Rate": 1.4491985373519146e-06, "Full-finetune/Loss": 0.7373465299606323, "Full-finetune/Loss (Raw)": 0.57664954662323, "Full-finetune/Step": 4323, "Full-finetune/Step Time": 6.797910751774907} {"Full-finetune/Learning Rate": 1.4459701695064155e-06, "Full-finetune/Loss": 0.7375049591064453, "Full-finetune/Loss (Raw)": 0.7520022392272949, "Full-finetune/Step": 4324, "Full-finetune/Step Time": 6.7978488728404045} {"Full-finetune/Learning Rate": 1.4427451213128873e-06, "Full-finetune/Loss": 0.7381170988082886, "Full-finetune/Loss (Raw)": 0.5213092565536499, "Full-finetune/Step": 4325, "Full-finetune/Step Time": 6.7873421385884285} {"Full-finetune/Learning Rate": 1.4395233940229147e-06, "Full-finetune/Loss": 0.736372709274292, "Full-finetune/Loss (Raw)": 0.6356084942817688, "Full-finetune/Step": 4326, "Full-finetune/Step Time": 6.7953194957226515} {"Full-finetune/Learning Rate": 1.4363049888867853e-06, "Full-finetune/Loss": 0.733077347278595, "Full-finetune/Loss (Raw)": 0.5014362931251526, "Full-finetune/Step": 4327, "Full-finetune/Step Time": 6.811450954526663} {"Full-finetune/Learning Rate": 1.4330899071534942e-06, "Full-finetune/Loss": 0.7324817180633545, "Full-finetune/Loss (Raw)": 0.7344925999641418, "Full-finetune/Step": 4328, "Full-finetune/Step Time": 6.795942638069391} {"Full-finetune/Learning Rate": 1.4298781500707558e-06, "Full-finetune/Loss": 0.7332051992416382, "Full-finetune/Loss (Raw)": 0.78848797082901, "Full-finetune/Step": 4329, "Full-finetune/Step Time": 6.785803975537419} {"Full-finetune/Learning Rate": 1.426669718884991e-06, "Full-finetune/Loss": 0.7336858510971069, "Full-finetune/Loss (Raw)": 0.8438082933425903, "Full-finetune/Step": 4330, "Full-finetune/Step Time": 6.785372111946344} {"Full-finetune/Learning Rate": 1.4234646148413277e-06, "Full-finetune/Loss": 0.7345743179321289, "Full-finetune/Loss (Raw)": 0.8313421607017517, "Full-finetune/Step": 4331, "Full-finetune/Step Time": 6.785537043586373} {"Full-finetune/Learning Rate": 1.420262839183605e-06, "Full-finetune/Loss": 0.735183596611023, "Full-finetune/Loss (Raw)": 0.7928977012634277, "Full-finetune/Step": 4332, "Full-finetune/Step Time": 6.7887564562261105} {"Full-finetune/Learning Rate": 1.4170643931543693e-06, "Full-finetune/Loss": 0.7354402542114258, "Full-finetune/Loss (Raw)": 0.8321113586425781, "Full-finetune/Step": 4333, "Full-finetune/Step Time": 6.790585709735751} {"Full-finetune/Learning Rate": 1.4138692779948749e-06, "Full-finetune/Loss": 0.7368700504302979, "Full-finetune/Loss (Raw)": 0.8783979415893555, "Full-finetune/Step": 4334, "Full-finetune/Step Time": 6.798908729106188} {"Full-finetune/Learning Rate": 1.4106774949450808e-06, "Full-finetune/Loss": 0.7368097305297852, "Full-finetune/Loss (Raw)": 0.7898217439651489, "Full-finetune/Step": 4335, "Full-finetune/Step Time": 6.795548304915428} {"Full-finetune/Learning Rate": 1.4074890452436607e-06, "Full-finetune/Loss": 0.7377341985702515, "Full-finetune/Loss (Raw)": 0.8889926671981812, "Full-finetune/Step": 4336, "Full-finetune/Step Time": 6.8109367694705725} {"Full-finetune/Learning Rate": 1.4043039301279904e-06, "Full-finetune/Loss": 0.7345553040504456, "Full-finetune/Loss (Raw)": 0.522951066493988, "Full-finetune/Step": 4337, "Full-finetune/Step Time": 6.8325180020183325} {"Full-finetune/Learning Rate": 1.4011221508341433e-06, "Full-finetune/Loss": 0.7346798181533813, "Full-finetune/Loss (Raw)": 0.7855819463729858, "Full-finetune/Step": 4338, "Full-finetune/Step Time": 6.832575948908925} {"Full-finetune/Learning Rate": 1.3979437085969127e-06, "Full-finetune/Loss": 0.7340197563171387, "Full-finetune/Loss (Raw)": 0.7418549656867981, "Full-finetune/Step": 4339, "Full-finetune/Step Time": 6.83225629106164} {"Full-finetune/Learning Rate": 1.3947686046497878e-06, "Full-finetune/Loss": 0.7343413829803467, "Full-finetune/Loss (Raw)": 0.7015997171401978, "Full-finetune/Step": 4340, "Full-finetune/Step Time": 6.84674665145576} {"Full-finetune/Learning Rate": 1.3915968402249647e-06, "Full-finetune/Loss": 0.7365772724151611, "Full-finetune/Loss (Raw)": 0.9042975902557373, "Full-finetune/Step": 4341, "Full-finetune/Step Time": 6.838153298944235} {"Full-finetune/Learning Rate": 1.3884284165533423e-06, "Full-finetune/Loss": 0.7359274625778198, "Full-finetune/Loss (Raw)": 0.6926412582397461, "Full-finetune/Step": 4342, "Full-finetune/Step Time": 6.877821588888764} {"Full-finetune/Learning Rate": 1.3852633348645262e-06, "Full-finetune/Loss": 0.7358103394508362, "Full-finetune/Loss (Raw)": 0.680553138256073, "Full-finetune/Step": 4343, "Full-finetune/Step Time": 6.851664112880826} {"Full-finetune/Learning Rate": 1.38210159638682e-06, "Full-finetune/Loss": 0.7357130646705627, "Full-finetune/Loss (Raw)": 0.788589358329773, "Full-finetune/Step": 4344, "Full-finetune/Step Time": 6.8717367593199015} {"Full-finetune/Learning Rate": 1.3789432023472315e-06, "Full-finetune/Loss": 0.735056459903717, "Full-finetune/Loss (Raw)": 0.7349720597267151, "Full-finetune/Step": 4345, "Full-finetune/Step Time": 6.870254931971431} {"Full-finetune/Learning Rate": 1.3757881539714768e-06, "Full-finetune/Loss": 0.7351891994476318, "Full-finetune/Loss (Raw)": 0.7810707092285156, "Full-finetune/Step": 4346, "Full-finetune/Step Time": 6.873233921825886} {"Full-finetune/Learning Rate": 1.3726364524839687e-06, "Full-finetune/Loss": 0.7375243902206421, "Full-finetune/Loss (Raw)": 0.8688543438911438, "Full-finetune/Step": 4347, "Full-finetune/Step Time": 6.86569857224822} {"Full-finetune/Learning Rate": 1.369488099107814e-06, "Full-finetune/Loss": 0.7374837398529053, "Full-finetune/Loss (Raw)": 0.8065391778945923, "Full-finetune/Step": 4348, "Full-finetune/Step Time": 6.87147044017911} {"Full-finetune/Learning Rate": 1.3663430950648326e-06, "Full-finetune/Loss": 0.737289547920227, "Full-finetune/Loss (Raw)": 0.8684712648391724, "Full-finetune/Step": 4349, "Full-finetune/Step Time": 6.868690149858594} {"Full-finetune/Learning Rate": 1.363201441575539e-06, "Full-finetune/Loss": 0.735026478767395, "Full-finetune/Loss (Raw)": 0.5330778360366821, "Full-finetune/Step": 4350, "Full-finetune/Step Time": 6.879678085446358} {"Full-finetune/Learning Rate": 1.360063139859147e-06, "Full-finetune/Loss": 0.7346363663673401, "Full-finetune/Loss (Raw)": 0.8095188140869141, "Full-finetune/Step": 4351, "Full-finetune/Step Time": 6.879976373165846} {"Full-finetune/Learning Rate": 1.3569281911335686e-06, "Full-finetune/Loss": 0.7352570295333862, "Full-finetune/Loss (Raw)": 0.7934920191764832, "Full-finetune/Step": 4352, "Full-finetune/Step Time": 6.871592078357935} {"Full-finetune/Learning Rate": 1.3537965966154177e-06, "Full-finetune/Loss": 0.7363739609718323, "Full-finetune/Loss (Raw)": 0.8262945413589478, "Full-finetune/Step": 4353, "Full-finetune/Step Time": 6.868962498381734} {"Full-finetune/Learning Rate": 1.3506683575200053e-06, "Full-finetune/Loss": 0.7351813316345215, "Full-finetune/Loss (Raw)": 0.7215351462364197, "Full-finetune/Step": 4354, "Full-finetune/Step Time": 6.866794189438224} {"Full-finetune/Learning Rate": 1.347543475061337e-06, "Full-finetune/Loss": 0.7338255643844604, "Full-finetune/Loss (Raw)": 0.8010768294334412, "Full-finetune/Step": 4355, "Full-finetune/Step Time": 6.8527439292520285} {"Full-finetune/Learning Rate": 1.344421950452126e-06, "Full-finetune/Loss": 0.7340076565742493, "Full-finetune/Loss (Raw)": 0.8048387765884399, "Full-finetune/Step": 4356, "Full-finetune/Step Time": 6.850765747949481} {"Full-finetune/Learning Rate": 1.341303784903767e-06, "Full-finetune/Loss": 0.7342016696929932, "Full-finetune/Loss (Raw)": 0.8684309720993042, "Full-finetune/Step": 4357, "Full-finetune/Step Time": 6.8306408785283566} {"Full-finetune/Learning Rate": 1.3381889796263602e-06, "Full-finetune/Loss": 0.7383724451065063, "Full-finetune/Loss (Raw)": 0.8022301197052002, "Full-finetune/Step": 4358, "Full-finetune/Step Time": 6.823573522269726} {"Full-finetune/Learning Rate": 1.335077535828706e-06, "Full-finetune/Loss": 0.7400503754615784, "Full-finetune/Loss (Raw)": 0.7685355544090271, "Full-finetune/Step": 4359, "Full-finetune/Step Time": 6.783803913742304} {"Full-finetune/Learning Rate": 1.3319694547182916e-06, "Full-finetune/Loss": 0.7394804954528809, "Full-finetune/Loss (Raw)": 0.7045117616653442, "Full-finetune/Step": 4360, "Full-finetune/Step Time": 6.787860553711653} {"Full-finetune/Learning Rate": 1.328864737501302e-06, "Full-finetune/Loss": 0.7422471046447754, "Full-finetune/Loss (Raw)": 0.982934296131134, "Full-finetune/Step": 4361, "Full-finetune/Step Time": 6.796750467270613} {"Full-finetune/Learning Rate": 1.325763385382618e-06, "Full-finetune/Loss": 0.740778923034668, "Full-finetune/Loss (Raw)": 0.762914776802063, "Full-finetune/Step": 4362, "Full-finetune/Step Time": 6.799104832112789} {"Full-finetune/Learning Rate": 1.3226653995658144e-06, "Full-finetune/Loss": 0.7402088046073914, "Full-finetune/Loss (Raw)": 0.7430537939071655, "Full-finetune/Step": 4363, "Full-finetune/Step Time": 6.79598649777472} {"Full-finetune/Learning Rate": 1.319570781253159e-06, "Full-finetune/Loss": 0.7407584190368652, "Full-finetune/Loss (Raw)": 0.6962092518806458, "Full-finetune/Step": 4364, "Full-finetune/Step Time": 6.817484952509403} {"Full-finetune/Learning Rate": 1.3164795316456091e-06, "Full-finetune/Loss": 0.7390134334564209, "Full-finetune/Loss (Raw)": 0.6761851906776428, "Full-finetune/Step": 4365, "Full-finetune/Step Time": 6.831816490739584} {"Full-finetune/Learning Rate": 1.313391651942827e-06, "Full-finetune/Loss": 0.73890620470047, "Full-finetune/Loss (Raw)": 0.8410869240760803, "Full-finetune/Step": 4366, "Full-finetune/Step Time": 6.830179676413536} {"Full-finetune/Learning Rate": 1.3103071433431501e-06, "Full-finetune/Loss": 0.7381386160850525, "Full-finetune/Loss (Raw)": 0.7208265066146851, "Full-finetune/Step": 4367, "Full-finetune/Step Time": 6.835862765088677} {"Full-finetune/Learning Rate": 1.307226007043617e-06, "Full-finetune/Loss": 0.7385133504867554, "Full-finetune/Loss (Raw)": 0.8114213347434998, "Full-finetune/Step": 4368, "Full-finetune/Step Time": 6.8312969990074635} {"Full-finetune/Learning Rate": 1.30414824423996e-06, "Full-finetune/Loss": 0.738139271736145, "Full-finetune/Loss (Raw)": 0.7372267842292786, "Full-finetune/Step": 4369, "Full-finetune/Step Time": 6.837103679776192} {"Full-finetune/Learning Rate": 1.301073856126598e-06, "Full-finetune/Loss": 0.7381950616836548, "Full-finetune/Loss (Raw)": 0.8548370003700256, "Full-finetune/Step": 4370, "Full-finetune/Step Time": 6.839148985221982} {"Full-finetune/Learning Rate": 1.2980028438966396e-06, "Full-finetune/Loss": 0.7393519878387451, "Full-finetune/Loss (Raw)": 0.8769931793212891, "Full-finetune/Step": 4371, "Full-finetune/Step Time": 6.81689783371985} {"Full-finetune/Learning Rate": 1.2949352087418864e-06, "Full-finetune/Loss": 0.7348943948745728, "Full-finetune/Loss (Raw)": 0.24142462015151978, "Full-finetune/Step": 4372, "Full-finetune/Step Time": 6.834924604743719} {"Full-finetune/Learning Rate": 1.291870951852826e-06, "Full-finetune/Loss": 0.7341851592063904, "Full-finetune/Loss (Raw)": 0.7043811082839966, "Full-finetune/Step": 4373, "Full-finetune/Step Time": 6.8324041571468115} {"Full-finetune/Learning Rate": 1.288810074418637e-06, "Full-finetune/Loss": 0.7345896363258362, "Full-finetune/Loss (Raw)": 0.7704960703849792, "Full-finetune/Step": 4374, "Full-finetune/Step Time": 6.797402793541551} {"Full-finetune/Learning Rate": 1.2857525776271852e-06, "Full-finetune/Loss": 0.733845055103302, "Full-finetune/Loss (Raw)": 0.7353788614273071, "Full-finetune/Step": 4375, "Full-finetune/Step Time": 6.805294696241617} {"Full-finetune/Learning Rate": 1.2826984626650297e-06, "Full-finetune/Loss": 0.7373244762420654, "Full-finetune/Loss (Raw)": 0.737164318561554, "Full-finetune/Step": 4376, "Full-finetune/Step Time": 6.801158260554075} {"Full-finetune/Learning Rate": 1.2796477307174082e-06, "Full-finetune/Loss": 0.7368263006210327, "Full-finetune/Loss (Raw)": 0.7751938104629517, "Full-finetune/Step": 4377, "Full-finetune/Step Time": 6.800237558782101} {"Full-finetune/Learning Rate": 1.2766003829682504e-06, "Full-finetune/Loss": 0.7379708290100098, "Full-finetune/Loss (Raw)": 0.7250308394432068, "Full-finetune/Step": 4378, "Full-finetune/Step Time": 6.8090120777487755} {"Full-finetune/Learning Rate": 1.273556420600175e-06, "Full-finetune/Loss": 0.7374677062034607, "Full-finetune/Loss (Raw)": 0.612305223941803, "Full-finetune/Step": 4379, "Full-finetune/Step Time": 6.818669080734253} {"Full-finetune/Learning Rate": 1.2705158447944843e-06, "Full-finetune/Loss": 0.736666202545166, "Full-finetune/Loss (Raw)": 0.874701738357544, "Full-finetune/Step": 4380, "Full-finetune/Step Time": 6.810613000765443} {"Full-finetune/Learning Rate": 1.2674786567311648e-06, "Full-finetune/Loss": 0.738910973072052, "Full-finetune/Loss (Raw)": 0.8254373073577881, "Full-finetune/Step": 4381, "Full-finetune/Step Time": 6.796522423624992} {"Full-finetune/Learning Rate": 1.2644448575888913e-06, "Full-finetune/Loss": 0.7420971393585205, "Full-finetune/Loss (Raw)": 0.8777194619178772, "Full-finetune/Step": 4382, "Full-finetune/Step Time": 6.789923597127199} {"Full-finetune/Learning Rate": 1.2614144485450208e-06, "Full-finetune/Loss": 0.741487443447113, "Full-finetune/Loss (Raw)": 0.6019076704978943, "Full-finetune/Step": 4383, "Full-finetune/Step Time": 6.803675862029195} {"Full-finetune/Learning Rate": 1.2583874307755961e-06, "Full-finetune/Loss": 0.7413451671600342, "Full-finetune/Loss (Raw)": 0.7712970972061157, "Full-finetune/Step": 4384, "Full-finetune/Step Time": 6.80075859092176} {"Full-finetune/Learning Rate": 1.2553638054553418e-06, "Full-finetune/Loss": 0.740874707698822, "Full-finetune/Loss (Raw)": 0.7609912753105164, "Full-finetune/Step": 4385, "Full-finetune/Step Time": 6.792798252776265} {"Full-finetune/Learning Rate": 1.252343573757674e-06, "Full-finetune/Loss": 0.7409900426864624, "Full-finetune/Loss (Raw)": 0.6306536793708801, "Full-finetune/Step": 4386, "Full-finetune/Step Time": 6.7999503538012505} {"Full-finetune/Learning Rate": 1.2493267368546802e-06, "Full-finetune/Loss": 0.7413554191589355, "Full-finetune/Loss (Raw)": 0.9382848739624023, "Full-finetune/Step": 4387, "Full-finetune/Step Time": 6.787519773468375} {"Full-finetune/Learning Rate": 1.2463132959171343e-06, "Full-finetune/Loss": 0.7416678667068481, "Full-finetune/Loss (Raw)": 0.8079844117164612, "Full-finetune/Step": 4388, "Full-finetune/Step Time": 6.7926422487944365} {"Full-finetune/Learning Rate": 1.2433032521145006e-06, "Full-finetune/Loss": 0.7415720224380493, "Full-finetune/Loss (Raw)": 0.7876430749893188, "Full-finetune/Step": 4389, "Full-finetune/Step Time": 6.790583770722151} {"Full-finetune/Learning Rate": 1.2402966066149135e-06, "Full-finetune/Loss": 0.7417814135551453, "Full-finetune/Loss (Raw)": 0.7691578269004822, "Full-finetune/Step": 4390, "Full-finetune/Step Time": 6.789304103702307} {"Full-finetune/Learning Rate": 1.2372933605851944e-06, "Full-finetune/Loss": 0.7404212355613708, "Full-finetune/Loss (Raw)": 0.7630860805511475, "Full-finetune/Step": 4391, "Full-finetune/Step Time": 6.797322951257229} {"Full-finetune/Learning Rate": 1.2342935151908453e-06, "Full-finetune/Loss": 0.7383708953857422, "Full-finetune/Loss (Raw)": 0.35572388768196106, "Full-finetune/Step": 4392, "Full-finetune/Step Time": 6.8063660860061646} {"Full-finetune/Learning Rate": 1.231297071596046e-06, "Full-finetune/Loss": 0.7379516363143921, "Full-finetune/Loss (Raw)": 0.6261109709739685, "Full-finetune/Step": 4393, "Full-finetune/Step Time": 6.815490135923028} {"Full-finetune/Learning Rate": 1.2283040309636596e-06, "Full-finetune/Loss": 0.7365185022354126, "Full-finetune/Loss (Raw)": 0.7244341373443604, "Full-finetune/Step": 4394, "Full-finetune/Step Time": 6.824776753783226} {"Full-finetune/Learning Rate": 1.2253143944552258e-06, "Full-finetune/Loss": 0.7375891208648682, "Full-finetune/Loss (Raw)": 0.7724766731262207, "Full-finetune/Step": 4395, "Full-finetune/Step Time": 6.814714025706053} {"Full-finetune/Learning Rate": 1.2223281632309636e-06, "Full-finetune/Loss": 0.7385143041610718, "Full-finetune/Loss (Raw)": 0.8251928091049194, "Full-finetune/Step": 4396, "Full-finetune/Step Time": 6.822870509698987} {"Full-finetune/Learning Rate": 1.2193453384497722e-06, "Full-finetune/Loss": 0.7375246286392212, "Full-finetune/Loss (Raw)": 0.4188568592071533, "Full-finetune/Step": 4397, "Full-finetune/Step Time": 6.817956365644932} {"Full-finetune/Learning Rate": 1.2163659212692246e-06, "Full-finetune/Loss": 0.7374135255813599, "Full-finetune/Loss (Raw)": 0.759071409702301, "Full-finetune/Step": 4398, "Full-finetune/Step Time": 6.821452837437391} {"Full-finetune/Learning Rate": 1.2133899128455796e-06, "Full-finetune/Loss": 0.7403084635734558, "Full-finetune/Loss (Raw)": 0.7017801403999329, "Full-finetune/Step": 4399, "Full-finetune/Step Time": 6.805845944210887} {"Full-finetune/Learning Rate": 1.2104173143337672e-06, "Full-finetune/Loss": 0.7398104071617126, "Full-finetune/Loss (Raw)": 0.7604154348373413, "Full-finetune/Step": 4400, "Full-finetune/Step Time": 6.822266183793545} {"Full-finetune/Learning Rate": 1.2074481268873894e-06, "Full-finetune/Loss": 0.7370388507843018, "Full-finetune/Loss (Raw)": 0.3329784870147705, "Full-finetune/Step": 4401, "Full-finetune/Step Time": 6.840669695287943} {"Full-finetune/Learning Rate": 1.2044823516587345e-06, "Full-finetune/Loss": 0.7361670136451721, "Full-finetune/Loss (Raw)": 0.635566234588623, "Full-finetune/Step": 4402, "Full-finetune/Step Time": 6.859958853572607} {"Full-finetune/Learning Rate": 1.2015199897987616e-06, "Full-finetune/Loss": 0.7378336191177368, "Full-finetune/Loss (Raw)": 0.8177136182785034, "Full-finetune/Step": 4403, "Full-finetune/Step Time": 6.844122109934688} {"Full-finetune/Learning Rate": 1.198561042457105e-06, "Full-finetune/Loss": 0.7387081980705261, "Full-finetune/Loss (Raw)": 0.9331706166267395, "Full-finetune/Step": 4404, "Full-finetune/Step Time": 6.845434645190835} {"Full-finetune/Learning Rate": 1.195605510782073e-06, "Full-finetune/Loss": 0.7389518022537231, "Full-finetune/Loss (Raw)": 0.7302478551864624, "Full-finetune/Step": 4405, "Full-finetune/Step Time": 6.835135025903583} {"Full-finetune/Learning Rate": 1.192653395920652e-06, "Full-finetune/Loss": 0.737180769443512, "Full-finetune/Loss (Raw)": 0.6244135499000549, "Full-finetune/Step": 4406, "Full-finetune/Step Time": 6.832521595060825} {"Full-finetune/Learning Rate": 1.189704699018499e-06, "Full-finetune/Loss": 0.736827552318573, "Full-finetune/Loss (Raw)": 0.6504614949226379, "Full-finetune/Step": 4407, "Full-finetune/Step Time": 6.8303939290344715} {"Full-finetune/Learning Rate": 1.186759421219943e-06, "Full-finetune/Loss": 0.7376327514648438, "Full-finetune/Loss (Raw)": 0.9319689273834229, "Full-finetune/Step": 4408, "Full-finetune/Step Time": 6.817926742136478} {"Full-finetune/Learning Rate": 1.1838175636679938e-06, "Full-finetune/Loss": 0.7356991767883301, "Full-finetune/Loss (Raw)": 0.6627531051635742, "Full-finetune/Step": 4409, "Full-finetune/Step Time": 6.81673707999289} {"Full-finetune/Learning Rate": 1.1808791275043297e-06, "Full-finetune/Loss": 0.7353671789169312, "Full-finetune/Loss (Raw)": 0.7928707599639893, "Full-finetune/Step": 4410, "Full-finetune/Step Time": 6.814036227762699} {"Full-finetune/Learning Rate": 1.1779441138692926e-06, "Full-finetune/Loss": 0.7353333234786987, "Full-finetune/Loss (Raw)": 0.7600021958351135, "Full-finetune/Step": 4411, "Full-finetune/Step Time": 6.8076732363551855} {"Full-finetune/Learning Rate": 1.1750125239019118e-06, "Full-finetune/Loss": 0.7347931861877441, "Full-finetune/Loss (Raw)": 0.7299487590789795, "Full-finetune/Step": 4412, "Full-finetune/Step Time": 6.80958417057991} {"Full-finetune/Learning Rate": 1.1720843587398756e-06, "Full-finetune/Loss": 0.7377381324768066, "Full-finetune/Loss (Raw)": 0.9763900637626648, "Full-finetune/Step": 4413, "Full-finetune/Step Time": 6.762351484969258} {"Full-finetune/Learning Rate": 1.1691596195195498e-06, "Full-finetune/Loss": 0.737561047077179, "Full-finetune/Loss (Raw)": 0.7291065454483032, "Full-finetune/Step": 4414, "Full-finetune/Step Time": 6.762150164693594} {"Full-finetune/Learning Rate": 1.1662383073759686e-06, "Full-finetune/Loss": 0.7378813624382019, "Full-finetune/Loss (Raw)": 0.8544637560844421, "Full-finetune/Step": 4415, "Full-finetune/Step Time": 6.757548252120614} {"Full-finetune/Learning Rate": 1.1633204234428352e-06, "Full-finetune/Loss": 0.7361166477203369, "Full-finetune/Loss (Raw)": 0.6613454222679138, "Full-finetune/Step": 4416, "Full-finetune/Step Time": 6.759530771523714} {"Full-finetune/Learning Rate": 1.1604059688525238e-06, "Full-finetune/Loss": 0.7372636795043945, "Full-finetune/Loss (Raw)": 0.8809221982955933, "Full-finetune/Step": 4417, "Full-finetune/Step Time": 6.760881565511227} {"Full-finetune/Learning Rate": 1.157494944736075e-06, "Full-finetune/Loss": 0.7365401983261108, "Full-finetune/Loss (Raw)": 0.7793567180633545, "Full-finetune/Step": 4418, "Full-finetune/Step Time": 6.760842602699995} {"Full-finetune/Learning Rate": 1.1545873522232055e-06, "Full-finetune/Loss": 0.7362042665481567, "Full-finetune/Loss (Raw)": 0.840806245803833, "Full-finetune/Step": 4419, "Full-finetune/Step Time": 6.777221025899053} {"Full-finetune/Learning Rate": 1.1516831924422945e-06, "Full-finetune/Loss": 0.734958291053772, "Full-finetune/Loss (Raw)": 0.679324209690094, "Full-finetune/Step": 4420, "Full-finetune/Step Time": 6.789076860994101} {"Full-finetune/Learning Rate": 1.1487824665203828e-06, "Full-finetune/Loss": 0.7344499826431274, "Full-finetune/Loss (Raw)": 0.749085545539856, "Full-finetune/Step": 4421, "Full-finetune/Step Time": 6.786854155361652} {"Full-finetune/Learning Rate": 1.1458851755831934e-06, "Full-finetune/Loss": 0.7344873547554016, "Full-finetune/Loss (Raw)": 0.7628858089447021, "Full-finetune/Step": 4422, "Full-finetune/Step Time": 6.779735013842583} {"Full-finetune/Learning Rate": 1.1429913207551046e-06, "Full-finetune/Loss": 0.7368109822273254, "Full-finetune/Loss (Raw)": 0.602404773235321, "Full-finetune/Step": 4423, "Full-finetune/Step Time": 6.766147630289197} {"Full-finetune/Learning Rate": 1.1401009031591658e-06, "Full-finetune/Loss": 0.7322750091552734, "Full-finetune/Loss (Raw)": 0.33478933572769165, "Full-finetune/Step": 4424, "Full-finetune/Step Time": 6.7865485064685345} {"Full-finetune/Learning Rate": 1.137213923917092e-06, "Full-finetune/Loss": 0.7304450869560242, "Full-finetune/Loss (Raw)": 0.6058176159858704, "Full-finetune/Step": 4425, "Full-finetune/Step Time": 6.803704997524619} {"Full-finetune/Learning Rate": 1.1343303841492626e-06, "Full-finetune/Loss": 0.7322957515716553, "Full-finetune/Loss (Raw)": 0.8270049691200256, "Full-finetune/Step": 4426, "Full-finetune/Step Time": 6.799772758036852} {"Full-finetune/Learning Rate": 1.1314502849747244e-06, "Full-finetune/Loss": 0.7307083606719971, "Full-finetune/Loss (Raw)": 0.6018270254135132, "Full-finetune/Step": 4427, "Full-finetune/Step Time": 6.809132816269994} {"Full-finetune/Learning Rate": 1.1285736275111836e-06, "Full-finetune/Loss": 0.7300378680229187, "Full-finetune/Loss (Raw)": 0.798855185508728, "Full-finetune/Step": 4428, "Full-finetune/Step Time": 6.814517984166741} {"Full-finetune/Learning Rate": 1.1257004128750205e-06, "Full-finetune/Loss": 0.7332320213317871, "Full-finetune/Loss (Raw)": 0.774050772190094, "Full-finetune/Step": 4429, "Full-finetune/Step Time": 6.798125050961971} {"Full-finetune/Learning Rate": 1.122830642181273e-06, "Full-finetune/Loss": 0.7309255003929138, "Full-finetune/Loss (Raw)": 0.5896955132484436, "Full-finetune/Step": 4430, "Full-finetune/Step Time": 6.834215650334954} {"Full-finetune/Learning Rate": 1.119964316543637e-06, "Full-finetune/Loss": 0.7297595143318176, "Full-finetune/Loss (Raw)": 0.7569693326950073, "Full-finetune/Step": 4431, "Full-finetune/Step Time": 6.84196599572897} {"Full-finetune/Learning Rate": 1.1171014370744837e-06, "Full-finetune/Loss": 0.7336596250534058, "Full-finetune/Loss (Raw)": 0.8334677815437317, "Full-finetune/Step": 4432, "Full-finetune/Step Time": 6.831139482557774} {"Full-finetune/Learning Rate": 1.114242004884839e-06, "Full-finetune/Loss": 0.733222246170044, "Full-finetune/Loss (Raw)": 0.585197389125824, "Full-finetune/Step": 4433, "Full-finetune/Step Time": 6.833904972299933} {"Full-finetune/Learning Rate": 1.111386021084394e-06, "Full-finetune/Loss": 0.73435378074646, "Full-finetune/Loss (Raw)": 0.86018967628479, "Full-finetune/Step": 4434, "Full-finetune/Step Time": 6.829676762223244} {"Full-finetune/Learning Rate": 1.1085334867814946e-06, "Full-finetune/Loss": 0.7360647916793823, "Full-finetune/Loss (Raw)": 0.7913628816604614, "Full-finetune/Step": 4435, "Full-finetune/Step Time": 6.824986048042774} {"Full-finetune/Learning Rate": 1.1056844030831603e-06, "Full-finetune/Loss": 0.7356007695198059, "Full-finetune/Loss (Raw)": 0.6185725927352905, "Full-finetune/Step": 4436, "Full-finetune/Step Time": 6.8376448806375265} {"Full-finetune/Learning Rate": 1.1028387710950616e-06, "Full-finetune/Loss": 0.7354195713996887, "Full-finetune/Loss (Raw)": 0.7080124020576477, "Full-finetune/Step": 4437, "Full-finetune/Step Time": 6.8386888559907675} {"Full-finetune/Learning Rate": 1.0999965919215327e-06, "Full-finetune/Loss": 0.7389964461326599, "Full-finetune/Loss (Raw)": 0.8087199330329895, "Full-finetune/Step": 4438, "Full-finetune/Step Time": 6.823493937030435} {"Full-finetune/Learning Rate": 1.0971578666655713e-06, "Full-finetune/Loss": 0.7401038408279419, "Full-finetune/Loss (Raw)": 0.9292157888412476, "Full-finetune/Step": 4439, "Full-finetune/Step Time": 6.818876635283232} {"Full-finetune/Learning Rate": 1.0943225964288272e-06, "Full-finetune/Loss": 0.7401138544082642, "Full-finetune/Loss (Raw)": 0.8547558188438416, "Full-finetune/Step": 4440, "Full-finetune/Step Time": 6.828777762129903} {"Full-finetune/Learning Rate": 1.091490782311614e-06, "Full-finetune/Loss": 0.739987313747406, "Full-finetune/Loss (Raw)": 0.869922399520874, "Full-finetune/Step": 4441, "Full-finetune/Step Time": 6.812888374552131} {"Full-finetune/Learning Rate": 1.088662425412903e-06, "Full-finetune/Loss": 0.7393351793289185, "Full-finetune/Loss (Raw)": 0.7663823962211609, "Full-finetune/Step": 4442, "Full-finetune/Step Time": 6.827642563730478} {"Full-finetune/Learning Rate": 1.0858375268303267e-06, "Full-finetune/Loss": 0.7396721839904785, "Full-finetune/Loss (Raw)": 0.7977249026298523, "Full-finetune/Step": 4443, "Full-finetune/Step Time": 6.828042320907116} {"Full-finetune/Learning Rate": 1.0830160876601737e-06, "Full-finetune/Loss": 0.73982173204422, "Full-finetune/Loss (Raw)": 0.6908673644065857, "Full-finetune/Step": 4444, "Full-finetune/Step Time": 6.809206735342741} {"Full-finetune/Learning Rate": 1.0801981089973844e-06, "Full-finetune/Loss": 0.7391304969787598, "Full-finetune/Loss (Raw)": 0.6814196109771729, "Full-finetune/Step": 4445, "Full-finetune/Step Time": 6.8120703380554914} {"Full-finetune/Learning Rate": 1.0773835919355669e-06, "Full-finetune/Loss": 0.7411755323410034, "Full-finetune/Loss (Raw)": 0.8646515011787415, "Full-finetune/Step": 4446, "Full-finetune/Step Time": 6.803828397765756} {"Full-finetune/Learning Rate": 1.0745725375669768e-06, "Full-finetune/Loss": 0.7426809072494507, "Full-finetune/Loss (Raw)": 0.7652511596679688, "Full-finetune/Step": 4447, "Full-finetune/Step Time": 6.794572444632649} {"Full-finetune/Learning Rate": 1.0717649469825298e-06, "Full-finetune/Loss": 0.7429467439651489, "Full-finetune/Loss (Raw)": 0.769266664981842, "Full-finetune/Step": 4448, "Full-finetune/Step Time": 6.793165730312467} {"Full-finetune/Learning Rate": 1.0689608212718005e-06, "Full-finetune/Loss": 0.7453194260597229, "Full-finetune/Loss (Raw)": 0.8867442011833191, "Full-finetune/Step": 4449, "Full-finetune/Step Time": 6.780793292447925} {"Full-finetune/Learning Rate": 1.0661601615230122e-06, "Full-finetune/Loss": 0.7438668012619019, "Full-finetune/Loss (Raw)": 0.6891553997993469, "Full-finetune/Step": 4450, "Full-finetune/Step Time": 6.785240422934294} {"Full-finetune/Learning Rate": 1.0633629688230452e-06, "Full-finetune/Loss": 0.7441718578338623, "Full-finetune/Loss (Raw)": 0.6157028079032898, "Full-finetune/Step": 4451, "Full-finetune/Step Time": 6.790843423455954} {"Full-finetune/Learning Rate": 1.0605692442574356e-06, "Full-finetune/Loss": 0.7410489916801453, "Full-finetune/Loss (Raw)": 0.3522758483886719, "Full-finetune/Step": 4452, "Full-finetune/Step Time": 6.807527786120772} {"Full-finetune/Learning Rate": 1.0577789889103763e-06, "Full-finetune/Loss": 0.7424218654632568, "Full-finetune/Loss (Raw)": 0.6970332264900208, "Full-finetune/Step": 4453, "Full-finetune/Step Time": 6.801919570192695} {"Full-finetune/Learning Rate": 1.054992203864712e-06, "Full-finetune/Loss": 0.7430948615074158, "Full-finetune/Loss (Raw)": 0.72175133228302, "Full-finetune/Step": 4454, "Full-finetune/Step Time": 6.80011023953557} {"Full-finetune/Learning Rate": 1.0522088902019324e-06, "Full-finetune/Loss": 0.7438790202140808, "Full-finetune/Loss (Raw)": 0.6018140316009521, "Full-finetune/Step": 4455, "Full-finetune/Step Time": 6.79752997867763} {"Full-finetune/Learning Rate": 1.049429049002193e-06, "Full-finetune/Loss": 0.7437617778778076, "Full-finetune/Loss (Raw)": 0.7194872498512268, "Full-finetune/Step": 4456, "Full-finetune/Step Time": 6.799276934936643} {"Full-finetune/Learning Rate": 1.0466526813442957e-06, "Full-finetune/Loss": 0.7427360415458679, "Full-finetune/Loss (Raw)": 0.6571909189224243, "Full-finetune/Step": 4457, "Full-finetune/Step Time": 6.810537779703736} {"Full-finetune/Learning Rate": 1.0438797883056927e-06, "Full-finetune/Loss": 0.7413408160209656, "Full-finetune/Loss (Raw)": 0.6652183532714844, "Full-finetune/Step": 4458, "Full-finetune/Step Time": 6.834915092214942} {"Full-finetune/Learning Rate": 1.0411103709624903e-06, "Full-finetune/Loss": 0.7402787804603577, "Full-finetune/Loss (Raw)": 0.6954031586647034, "Full-finetune/Step": 4459, "Full-finetune/Step Time": 6.841010835021734} {"Full-finetune/Learning Rate": 1.0383444303894453e-06, "Full-finetune/Loss": 0.7408210635185242, "Full-finetune/Loss (Raw)": 0.8623088002204895, "Full-finetune/Step": 4460, "Full-finetune/Step Time": 6.83992094360292} {"Full-finetune/Learning Rate": 1.0355819676599655e-06, "Full-finetune/Loss": 0.7397031784057617, "Full-finetune/Loss (Raw)": 0.6890275478363037, "Full-finetune/Step": 4461, "Full-finetune/Step Time": 6.8388914708048105} {"Full-finetune/Learning Rate": 1.0328229838461069e-06, "Full-finetune/Loss": 0.7388370037078857, "Full-finetune/Loss (Raw)": 0.7675256133079529, "Full-finetune/Step": 4462, "Full-finetune/Step Time": 6.840844387188554} {"Full-finetune/Learning Rate": 1.0300674800185795e-06, "Full-finetune/Loss": 0.7367298603057861, "Full-finetune/Loss (Raw)": 0.5201115012168884, "Full-finetune/Step": 4463, "Full-finetune/Step Time": 6.84936460852623} {"Full-finetune/Learning Rate": 1.0273154572467436e-06, "Full-finetune/Loss": 0.7356728315353394, "Full-finetune/Loss (Raw)": 0.7536857724189758, "Full-finetune/Step": 4464, "Full-finetune/Step Time": 6.852612739428878} {"Full-finetune/Learning Rate": 1.0245669165985973e-06, "Full-finetune/Loss": 0.7372337579727173, "Full-finetune/Loss (Raw)": 0.7227444052696228, "Full-finetune/Step": 4465, "Full-finetune/Step Time": 6.830304853618145} {"Full-finetune/Learning Rate": 1.0218218591408013e-06, "Full-finetune/Loss": 0.7369176149368286, "Full-finetune/Loss (Raw)": 0.7451177835464478, "Full-finetune/Step": 4466, "Full-finetune/Step Time": 6.828627500683069} {"Full-finetune/Learning Rate": 1.0190802859386585e-06, "Full-finetune/Loss": 0.7375937700271606, "Full-finetune/Loss (Raw)": 0.8284064531326294, "Full-finetune/Step": 4467, "Full-finetune/Step Time": 6.823283225297928} {"Full-finetune/Learning Rate": 1.0163421980561184e-06, "Full-finetune/Loss": 0.737888514995575, "Full-finetune/Loss (Raw)": 0.7393285632133484, "Full-finetune/Step": 4468, "Full-finetune/Step Time": 6.810612544417381} {"Full-finetune/Learning Rate": 1.0136075965557812e-06, "Full-finetune/Loss": 0.7366747856140137, "Full-finetune/Loss (Raw)": 0.7489322423934937, "Full-finetune/Step": 4469, "Full-finetune/Step Time": 6.810692464932799} {"Full-finetune/Learning Rate": 1.0108764824988903e-06, "Full-finetune/Loss": 0.7372097373008728, "Full-finetune/Loss (Raw)": 0.7611187696456909, "Full-finetune/Step": 4470, "Full-finetune/Step Time": 6.772702546790242} {"Full-finetune/Learning Rate": 1.008148856945339e-06, "Full-finetune/Loss": 0.7366670370101929, "Full-finetune/Loss (Raw)": 0.6110922694206238, "Full-finetune/Step": 4471, "Full-finetune/Step Time": 6.775473793968558} {"Full-finetune/Learning Rate": 1.0054247209536628e-06, "Full-finetune/Loss": 0.7368035316467285, "Full-finetune/Loss (Raw)": 0.8060557842254639, "Full-finetune/Step": 4472, "Full-finetune/Step Time": 6.7668775357306} {"Full-finetune/Learning Rate": 1.0027040755810513e-06, "Full-finetune/Loss": 0.7387745380401611, "Full-finetune/Loss (Raw)": 0.9872617125511169, "Full-finetune/Step": 4473, "Full-finetune/Step Time": 6.765653824433684} {"Full-finetune/Learning Rate": 9.999869218833313e-07, "Full-finetune/Loss": 0.7390680313110352, "Full-finetune/Loss (Raw)": 0.8186447620391846, "Full-finetune/Step": 4474, "Full-finetune/Step Time": 6.764092653989792} {"Full-finetune/Learning Rate": 9.972732609149738e-07, "Full-finetune/Loss": 0.7375764846801758, "Full-finetune/Loss (Raw)": 0.6779290437698364, "Full-finetune/Step": 4475, "Full-finetune/Step Time": 6.794745011255145} {"Full-finetune/Learning Rate": 9.945630937291006e-07, "Full-finetune/Loss": 0.7378537058830261, "Full-finetune/Loss (Raw)": 0.8420225977897644, "Full-finetune/Step": 4476, "Full-finetune/Step Time": 6.796818271279335} {"Full-finetune/Learning Rate": 9.918564213774762e-07, "Full-finetune/Loss": 0.7365820407867432, "Full-finetune/Loss (Raw)": 0.7057050466537476, "Full-finetune/Step": 4477, "Full-finetune/Step Time": 6.7963791359215975} {"Full-finetune/Learning Rate": 9.891532449105045e-07, "Full-finetune/Loss": 0.7390721440315247, "Full-finetune/Loss (Raw)": 0.851803719997406, "Full-finetune/Step": 4478, "Full-finetune/Step Time": 6.779303988441825} {"Full-finetune/Learning Rate": 9.864535653772367e-07, "Full-finetune/Loss": 0.7387933731079102, "Full-finetune/Loss (Raw)": 0.7738380432128906, "Full-finetune/Step": 4479, "Full-finetune/Step Time": 6.788284290581942} {"Full-finetune/Learning Rate": 9.837573838253655e-07, "Full-finetune/Loss": 0.7377102971076965, "Full-finetune/Loss (Raw)": 0.6548596024513245, "Full-finetune/Step": 4480, "Full-finetune/Step Time": 6.793140882626176} {"Full-finetune/Learning Rate": 9.810647013012265e-07, "Full-finetune/Loss": 0.7367212772369385, "Full-finetune/Loss (Raw)": 0.699705958366394, "Full-finetune/Step": 4481, "Full-finetune/Step Time": 6.796122435480356} {"Full-finetune/Learning Rate": 9.783755188497944e-07, "Full-finetune/Loss": 0.7365261316299438, "Full-finetune/Loss (Raw)": 0.6965478658676147, "Full-finetune/Step": 4482, "Full-finetune/Step Time": 6.81006289459765} {"Full-finetune/Learning Rate": 9.75689837514696e-07, "Full-finetune/Loss": 0.7364709377288818, "Full-finetune/Loss (Raw)": 0.7940153479576111, "Full-finetune/Step": 4483, "Full-finetune/Step Time": 6.80703149177134} {"Full-finetune/Learning Rate": 9.730076583381841e-07, "Full-finetune/Loss": 0.7370192408561707, "Full-finetune/Loss (Raw)": 0.8750205039978027, "Full-finetune/Step": 4484, "Full-finetune/Step Time": 6.822734925895929} {"Full-finetune/Learning Rate": 9.703289823611606e-07, "Full-finetune/Loss": 0.736142635345459, "Full-finetune/Loss (Raw)": 0.756222665309906, "Full-finetune/Step": 4485, "Full-finetune/Step Time": 6.8234650287777185} {"Full-finetune/Learning Rate": 9.676538106231725e-07, "Full-finetune/Loss": 0.7355465888977051, "Full-finetune/Loss (Raw)": 0.7259410619735718, "Full-finetune/Step": 4486, "Full-finetune/Step Time": 6.819537496194243} {"Full-finetune/Learning Rate": 9.649821441623986e-07, "Full-finetune/Loss": 0.7321025729179382, "Full-finetune/Loss (Raw)": 0.327701210975647, "Full-finetune/Step": 4487, "Full-finetune/Step Time": 6.837970511987805} {"Full-finetune/Learning Rate": 9.623139840156604e-07, "Full-finetune/Loss": 0.7327876687049866, "Full-finetune/Loss (Raw)": 0.7922025322914124, "Full-finetune/Step": 4488, "Full-finetune/Step Time": 6.834242958575487} {"Full-finetune/Learning Rate": 9.596493312184197e-07, "Full-finetune/Loss": 0.7288590669631958, "Full-finetune/Loss (Raw)": 0.4800715148448944, "Full-finetune/Step": 4489, "Full-finetune/Step Time": 6.824238250032067} {"Full-finetune/Learning Rate": 9.569881868047747e-07, "Full-finetune/Loss": 0.7279795408248901, "Full-finetune/Loss (Raw)": 0.6503310799598694, "Full-finetune/Step": 4490, "Full-finetune/Step Time": 6.828058259561658} {"Full-finetune/Learning Rate": 9.543305518074652e-07, "Full-finetune/Loss": 0.727893590927124, "Full-finetune/Loss (Raw)": 0.7320540547370911, "Full-finetune/Step": 4491, "Full-finetune/Step Time": 6.828631654381752} {"Full-finetune/Learning Rate": 9.516764272578638e-07, "Full-finetune/Loss": 0.7288147211074829, "Full-finetune/Loss (Raw)": 0.8141196370124817, "Full-finetune/Step": 4492, "Full-finetune/Step Time": 6.8020640686154366} {"Full-finetune/Learning Rate": 9.49025814185992e-07, "Full-finetune/Loss": 0.7289966344833374, "Full-finetune/Loss (Raw)": 0.6994732618331909, "Full-finetune/Step": 4493, "Full-finetune/Step Time": 6.797682778909802} {"Full-finetune/Learning Rate": 9.463787136204938e-07, "Full-finetune/Loss": 0.727106511592865, "Full-finetune/Loss (Raw)": 0.5991459488868713, "Full-finetune/Step": 4494, "Full-finetune/Step Time": 6.803593097254634} {"Full-finetune/Learning Rate": 9.437351265886585e-07, "Full-finetune/Loss": 0.7275132536888123, "Full-finetune/Loss (Raw)": 0.7728894948959351, "Full-finetune/Step": 4495, "Full-finetune/Step Time": 6.804356196895242} {"Full-finetune/Learning Rate": 9.410950541164143e-07, "Full-finetune/Loss": 0.7266253232955933, "Full-finetune/Loss (Raw)": 0.697765052318573, "Full-finetune/Step": 4496, "Full-finetune/Step Time": 6.812682665884495} {"Full-finetune/Learning Rate": 9.384584972283184e-07, "Full-finetune/Loss": 0.7268216609954834, "Full-finetune/Loss (Raw)": 0.7623571157455444, "Full-finetune/Step": 4497, "Full-finetune/Step Time": 6.826095093041658} {"Full-finetune/Learning Rate": 9.358254569475689e-07, "Full-finetune/Loss": 0.7253743410110474, "Full-finetune/Loss (Raw)": 0.6695846319198608, "Full-finetune/Step": 4498, "Full-finetune/Step Time": 6.831606682389975} {"Full-finetune/Learning Rate": 9.331959342959951e-07, "Full-finetune/Loss": 0.7253355383872986, "Full-finetune/Loss (Raw)": 0.872025728225708, "Full-finetune/Step": 4499, "Full-finetune/Step Time": 6.8319414258003235} {"Full-finetune/Learning Rate": 9.305699302940662e-07, "Full-finetune/Loss": 0.7287400364875793, "Full-finetune/Loss (Raw)": 0.6771984696388245, "Full-finetune/Step": 4500, "Full-finetune/Step Time": 6.821800390258431} {"Full-finetune/Learning Rate": 9.279474459608806e-07, "Full-finetune/Loss": 0.7300273776054382, "Full-finetune/Loss (Raw)": 0.8691641092300415, "Full-finetune/Step": 4501, "Full-finetune/Step Time": 6.8200462982058525} {"Full-finetune/Learning Rate": 9.253284823141728e-07, "Full-finetune/Loss": 0.7290131449699402, "Full-finetune/Loss (Raw)": 0.6406733393669128, "Full-finetune/Step": 4502, "Full-finetune/Step Time": 6.8251308389008045} {"Full-finetune/Learning Rate": 9.227130403703178e-07, "Full-finetune/Loss": 0.7299191951751709, "Full-finetune/Loss (Raw)": 0.8513467907905579, "Full-finetune/Step": 4503, "Full-finetune/Step Time": 6.8147946279495955} {"Full-finetune/Learning Rate": 9.201011211443101e-07, "Full-finetune/Loss": 0.7300729751586914, "Full-finetune/Loss (Raw)": 0.7568578720092773, "Full-finetune/Step": 4504, "Full-finetune/Step Time": 6.8055949080735445} {"Full-finetune/Learning Rate": 9.174927256497846e-07, "Full-finetune/Loss": 0.7298799157142639, "Full-finetune/Loss (Raw)": 0.7504798173904419, "Full-finetune/Step": 4505, "Full-finetune/Step Time": 6.804794814437628} {"Full-finetune/Learning Rate": 9.148878548990137e-07, "Full-finetune/Loss": 0.7316402196884155, "Full-finetune/Loss (Raw)": 0.9503413438796997, "Full-finetune/Step": 4506, "Full-finetune/Step Time": 6.7950038611888885} {"Full-finetune/Learning Rate": 9.122865099028944e-07, "Full-finetune/Loss": 0.7293041944503784, "Full-finetune/Loss (Raw)": 0.3133023679256439, "Full-finetune/Step": 4507, "Full-finetune/Step Time": 6.799998041242361} {"Full-finetune/Learning Rate": 9.09688691670958e-07, "Full-finetune/Loss": 0.7283655405044556, "Full-finetune/Loss (Raw)": 0.7545530200004578, "Full-finetune/Step": 4508, "Full-finetune/Step Time": 6.7939990274608135} {"Full-finetune/Learning Rate": 9.070944012113669e-07, "Full-finetune/Loss": 0.727457582950592, "Full-finetune/Loss (Raw)": 0.7092200517654419, "Full-finetune/Step": 4509, "Full-finetune/Step Time": 6.793446900323033} {"Full-finetune/Learning Rate": 9.045036395309148e-07, "Full-finetune/Loss": 0.7271838188171387, "Full-finetune/Loss (Raw)": 0.842673122882843, "Full-finetune/Step": 4510, "Full-finetune/Step Time": 6.7895519603043795} {"Full-finetune/Learning Rate": 9.019164076350262e-07, "Full-finetune/Loss": 0.7298039197921753, "Full-finetune/Loss (Raw)": 0.9372761249542236, "Full-finetune/Step": 4511, "Full-finetune/Step Time": 6.776669023558497} {"Full-finetune/Learning Rate": 8.993327065277524e-07, "Full-finetune/Loss": 0.7266002297401428, "Full-finetune/Loss (Raw)": 0.3612333834171295, "Full-finetune/Step": 4512, "Full-finetune/Step Time": 6.777784628793597} {"Full-finetune/Learning Rate": 8.967525372117825e-07, "Full-finetune/Loss": 0.7267105579376221, "Full-finetune/Loss (Raw)": 0.775107204914093, "Full-finetune/Step": 4513, "Full-finetune/Step Time": 6.784906577318907} {"Full-finetune/Learning Rate": 8.941759006884265e-07, "Full-finetune/Loss": 0.72684246301651, "Full-finetune/Loss (Raw)": 0.6475391983985901, "Full-finetune/Step": 4514, "Full-finetune/Step Time": 6.766024772077799} {"Full-finetune/Learning Rate": 8.916027979576247e-07, "Full-finetune/Loss": 0.7254384160041809, "Full-finetune/Loss (Raw)": 0.7585707902908325, "Full-finetune/Step": 4515, "Full-finetune/Step Time": 6.769020013511181} {"Full-finetune/Learning Rate": 8.890332300179516e-07, "Full-finetune/Loss": 0.7238657474517822, "Full-finetune/Loss (Raw)": 0.6066796183586121, "Full-finetune/Step": 4516, "Full-finetune/Step Time": 6.763361122459173} {"Full-finetune/Learning Rate": 8.864671978666084e-07, "Full-finetune/Loss": 0.7229020595550537, "Full-finetune/Loss (Raw)": 0.6642953157424927, "Full-finetune/Step": 4517, "Full-finetune/Step Time": 6.766164228320122} {"Full-finetune/Learning Rate": 8.839047024994152e-07, "Full-finetune/Loss": 0.7240279316902161, "Full-finetune/Loss (Raw)": 0.9132649898529053, "Full-finetune/Step": 4518, "Full-finetune/Step Time": 6.761258468031883} {"Full-finetune/Learning Rate": 8.813457449108309e-07, "Full-finetune/Loss": 0.724911093711853, "Full-finetune/Loss (Raw)": 0.876130998134613, "Full-finetune/Step": 4519, "Full-finetune/Step Time": 6.754304213449359} {"Full-finetune/Learning Rate": 8.787903260939368e-07, "Full-finetune/Loss": 0.7288336157798767, "Full-finetune/Loss (Raw)": 0.8578110933303833, "Full-finetune/Step": 4520, "Full-finetune/Step Time": 6.735607076436281} {"Full-finetune/Learning Rate": 8.762384470404417e-07, "Full-finetune/Loss": 0.7309088706970215, "Full-finetune/Loss (Raw)": 0.891742467880249, "Full-finetune/Step": 4521, "Full-finetune/Step Time": 6.731403410434723} {"Full-finetune/Learning Rate": 8.736901087406779e-07, "Full-finetune/Loss": 0.73082035779953, "Full-finetune/Loss (Raw)": 0.7131019830703735, "Full-finetune/Step": 4522, "Full-finetune/Step Time": 6.727974634617567} {"Full-finetune/Learning Rate": 8.711453121836066e-07, "Full-finetune/Loss": 0.730855405330658, "Full-finetune/Loss (Raw)": 0.7769646048545837, "Full-finetune/Step": 4523, "Full-finetune/Step Time": 6.731427801772952} {"Full-finetune/Learning Rate": 8.686040583568134e-07, "Full-finetune/Loss": 0.7306768298149109, "Full-finetune/Loss (Raw)": 0.8023337125778198, "Full-finetune/Step": 4524, "Full-finetune/Step Time": 6.725850882008672} {"Full-finetune/Learning Rate": 8.660663482465093e-07, "Full-finetune/Loss": 0.7334117293357849, "Full-finetune/Loss (Raw)": 0.7689270973205566, "Full-finetune/Step": 4525, "Full-finetune/Step Time": 6.699372161179781} {"Full-finetune/Learning Rate": 8.63532182837531e-07, "Full-finetune/Loss": 0.7318974733352661, "Full-finetune/Loss (Raw)": 0.5652433037757874, "Full-finetune/Step": 4526, "Full-finetune/Step Time": 6.693269187584519} {"Full-finetune/Learning Rate": 8.610015631133395e-07, "Full-finetune/Loss": 0.7298654317855835, "Full-finetune/Loss (Raw)": 0.4416801631450653, "Full-finetune/Step": 4527, "Full-finetune/Step Time": 6.711999803781509} {"Full-finetune/Learning Rate": 8.584744900560149e-07, "Full-finetune/Loss": 0.7294955253601074, "Full-finetune/Loss (Raw)": 0.713067352771759, "Full-finetune/Step": 4528, "Full-finetune/Step Time": 6.703884162008762} {"Full-finetune/Learning Rate": 8.559509646462693e-07, "Full-finetune/Loss": 0.7331563234329224, "Full-finetune/Loss (Raw)": 0.8015538454055786, "Full-finetune/Step": 4529, "Full-finetune/Step Time": 6.688866298645735} {"Full-finetune/Learning Rate": 8.534309878634317e-07, "Full-finetune/Loss": 0.7349386811256409, "Full-finetune/Loss (Raw)": 0.8637140393257141, "Full-finetune/Step": 4530, "Full-finetune/Step Time": 6.671241518110037} {"Full-finetune/Learning Rate": 8.509145606854552e-07, "Full-finetune/Loss": 0.7354845404624939, "Full-finetune/Loss (Raw)": 0.8875830769538879, "Full-finetune/Step": 4531, "Full-finetune/Step Time": 6.68731245957315} {"Full-finetune/Learning Rate": 8.484016840889176e-07, "Full-finetune/Loss": 0.7329345345497131, "Full-finetune/Loss (Raw)": 0.6067726016044617, "Full-finetune/Step": 4532, "Full-finetune/Step Time": 6.696591533720493} {"Full-finetune/Learning Rate": 8.458923590490165e-07, "Full-finetune/Loss": 0.7335270643234253, "Full-finetune/Loss (Raw)": 0.8060877919197083, "Full-finetune/Step": 4533, "Full-finetune/Step Time": 6.689897065982223} {"Full-finetune/Learning Rate": 8.433865865395718e-07, "Full-finetune/Loss": 0.7313023209571838, "Full-finetune/Loss (Raw)": 0.33964502811431885, "Full-finetune/Step": 4534, "Full-finetune/Step Time": 6.707218114286661} {"Full-finetune/Learning Rate": 8.408843675330225e-07, "Full-finetune/Loss": 0.7311558723449707, "Full-finetune/Loss (Raw)": 0.6317160725593567, "Full-finetune/Step": 4535, "Full-finetune/Step Time": 6.716343678534031} {"Full-finetune/Learning Rate": 8.383857030004361e-07, "Full-finetune/Loss": 0.7298239469528198, "Full-finetune/Loss (Raw)": 0.7614887952804565, "Full-finetune/Step": 4536, "Full-finetune/Step Time": 6.7260967791080475} {"Full-finetune/Learning Rate": 8.358905939114948e-07, "Full-finetune/Loss": 0.730892539024353, "Full-finetune/Loss (Raw)": 0.7995285987854004, "Full-finetune/Step": 4537, "Full-finetune/Step Time": 6.7243114691227674} {"Full-finetune/Learning Rate": 8.333990412344961e-07, "Full-finetune/Loss": 0.7311933636665344, "Full-finetune/Loss (Raw)": 0.8313778638839722, "Full-finetune/Step": 4538, "Full-finetune/Step Time": 6.723893417045474} {"Full-finetune/Learning Rate": 8.30911045936369e-07, "Full-finetune/Loss": 0.7302736639976501, "Full-finetune/Loss (Raw)": 0.6422773599624634, "Full-finetune/Step": 4539, "Full-finetune/Step Time": 6.730848357081413} {"Full-finetune/Learning Rate": 8.284266089826531e-07, "Full-finetune/Loss": 0.7269887924194336, "Full-finetune/Loss (Raw)": 0.30948618054389954, "Full-finetune/Step": 4540, "Full-finetune/Step Time": 6.749529525637627} {"Full-finetune/Learning Rate": 8.259457313375097e-07, "Full-finetune/Loss": 0.7247165441513062, "Full-finetune/Loss (Raw)": 0.6855475902557373, "Full-finetune/Step": 4541, "Full-finetune/Step Time": 6.772739199921489} {"Full-finetune/Learning Rate": 8.234684139637205e-07, "Full-finetune/Loss": 0.7240048050880432, "Full-finetune/Loss (Raw)": 0.6380029320716858, "Full-finetune/Step": 4542, "Full-finetune/Step Time": 6.77439827658236} {"Full-finetune/Learning Rate": 8.209946578226836e-07, "Full-finetune/Loss": 0.7233284711837769, "Full-finetune/Loss (Raw)": 0.767891526222229, "Full-finetune/Step": 4543, "Full-finetune/Step Time": 6.768484599888325} {"Full-finetune/Learning Rate": 8.185244638744139e-07, "Full-finetune/Loss": 0.7249674797058105, "Full-finetune/Loss (Raw)": 0.8711366057395935, "Full-finetune/Step": 4544, "Full-finetune/Step Time": 6.759665263816714} {"Full-finetune/Learning Rate": 8.160578330775459e-07, "Full-finetune/Loss": 0.7239187955856323, "Full-finetune/Loss (Raw)": 0.7466931343078613, "Full-finetune/Step": 4545, "Full-finetune/Step Time": 6.77069877833128} {"Full-finetune/Learning Rate": 8.135947663893329e-07, "Full-finetune/Loss": 0.7245261669158936, "Full-finetune/Loss (Raw)": 0.8571000695228577, "Full-finetune/Step": 4546, "Full-finetune/Step Time": 6.770572559908032} {"Full-finetune/Learning Rate": 8.111352647656423e-07, "Full-finetune/Loss": 0.7243975400924683, "Full-finetune/Loss (Raw)": 0.8243436217308044, "Full-finetune/Step": 4547, "Full-finetune/Step Time": 6.752856096252799} {"Full-finetune/Learning Rate": 8.086793291609562e-07, "Full-finetune/Loss": 0.7235270738601685, "Full-finetune/Loss (Raw)": 0.5679090023040771, "Full-finetune/Step": 4548, "Full-finetune/Step Time": 6.776405541226268} {"Full-finetune/Learning Rate": 8.062269605283779e-07, "Full-finetune/Loss": 0.7222127914428711, "Full-finetune/Loss (Raw)": 0.580849289894104, "Full-finetune/Step": 4549, "Full-finetune/Step Time": 6.777459582313895} {"Full-finetune/Learning Rate": 8.037781598196225e-07, "Full-finetune/Loss": 0.7232490181922913, "Full-finetune/Loss (Raw)": 0.8955230712890625, "Full-finetune/Step": 4550, "Full-finetune/Step Time": 6.764510419219732} {"Full-finetune/Learning Rate": 8.013329279850235e-07, "Full-finetune/Loss": 0.7249449491500854, "Full-finetune/Loss (Raw)": 0.8194847702980042, "Full-finetune/Step": 4551, "Full-finetune/Step Time": 6.76331558637321} {"Full-finetune/Learning Rate": 7.988912659735249e-07, "Full-finetune/Loss": 0.7268942594528198, "Full-finetune/Loss (Raw)": 0.5843057036399841, "Full-finetune/Step": 4552, "Full-finetune/Step Time": 6.755929209291935} {"Full-finetune/Learning Rate": 7.964531747326898e-07, "Full-finetune/Loss": 0.7283994555473328, "Full-finetune/Loss (Raw)": 0.7984762191772461, "Full-finetune/Step": 4553, "Full-finetune/Step Time": 6.735603164881468} {"Full-finetune/Learning Rate": 7.940186552086937e-07, "Full-finetune/Loss": 0.7289273142814636, "Full-finetune/Loss (Raw)": 0.894571840763092, "Full-finetune/Step": 4554, "Full-finetune/Step Time": 6.734215557575226} {"Full-finetune/Learning Rate": 7.915877083463242e-07, "Full-finetune/Loss": 0.7305172681808472, "Full-finetune/Loss (Raw)": 0.8053350448608398, "Full-finetune/Step": 4555, "Full-finetune/Step Time": 6.735621118918061} {"Full-finetune/Learning Rate": 7.891603350889888e-07, "Full-finetune/Loss": 0.7315691709518433, "Full-finetune/Loss (Raw)": 0.9335075616836548, "Full-finetune/Step": 4556, "Full-finetune/Step Time": 6.744664339348674} {"Full-finetune/Learning Rate": 7.867365363787005e-07, "Full-finetune/Loss": 0.730827808380127, "Full-finetune/Loss (Raw)": 0.6791530251502991, "Full-finetune/Step": 4557, "Full-finetune/Step Time": 6.7473198510706425} {"Full-finetune/Learning Rate": 7.843163131560871e-07, "Full-finetune/Loss": 0.7306910753250122, "Full-finetune/Loss (Raw)": 0.5721906423568726, "Full-finetune/Step": 4558, "Full-finetune/Step Time": 6.734239438548684} {"Full-finetune/Learning Rate": 7.818996663603917e-07, "Full-finetune/Loss": 0.7318487167358398, "Full-finetune/Loss (Raw)": 0.9051483273506165, "Full-finetune/Step": 4559, "Full-finetune/Step Time": 6.73584122210741} {"Full-finetune/Learning Rate": 7.794865969294685e-07, "Full-finetune/Loss": 0.7310366630554199, "Full-finetune/Loss (Raw)": 0.7295291423797607, "Full-finetune/Step": 4560, "Full-finetune/Step Time": 6.730476235970855} {"Full-finetune/Learning Rate": 7.770771057997828e-07, "Full-finetune/Loss": 0.7325958609580994, "Full-finetune/Loss (Raw)": 0.7847712635993958, "Full-finetune/Step": 4561, "Full-finetune/Step Time": 6.721023187041283} {"Full-finetune/Learning Rate": 7.746711939064066e-07, "Full-finetune/Loss": 0.7324236035346985, "Full-finetune/Loss (Raw)": 0.8381428718566895, "Full-finetune/Step": 4562, "Full-finetune/Step Time": 6.720470283180475} {"Full-finetune/Learning Rate": 7.722688621830333e-07, "Full-finetune/Loss": 0.7286776304244995, "Full-finetune/Loss (Raw)": 0.3118806779384613, "Full-finetune/Step": 4563, "Full-finetune/Step Time": 6.736569378525019} {"Full-finetune/Learning Rate": 7.698701115619578e-07, "Full-finetune/Loss": 0.7297853827476501, "Full-finetune/Loss (Raw)": 0.7603644728660583, "Full-finetune/Step": 4564, "Full-finetune/Step Time": 6.722534457221627} {"Full-finetune/Learning Rate": 7.674749429740868e-07, "Full-finetune/Loss": 0.7296084761619568, "Full-finetune/Loss (Raw)": 0.6853660941123962, "Full-finetune/Step": 4565, "Full-finetune/Step Time": 6.717430718243122} {"Full-finetune/Learning Rate": 7.650833573489436e-07, "Full-finetune/Loss": 0.7292912602424622, "Full-finetune/Loss (Raw)": 0.7681156992912292, "Full-finetune/Step": 4566, "Full-finetune/Step Time": 6.719169562682509} {"Full-finetune/Learning Rate": 7.626953556146521e-07, "Full-finetune/Loss": 0.7288813591003418, "Full-finetune/Loss (Raw)": 0.8767507076263428, "Full-finetune/Step": 4567, "Full-finetune/Step Time": 6.717426234856248} {"Full-finetune/Learning Rate": 7.603109386979501e-07, "Full-finetune/Loss": 0.7284945249557495, "Full-finetune/Loss (Raw)": 0.8052350282669067, "Full-finetune/Step": 4568, "Full-finetune/Step Time": 6.703974410891533} {"Full-finetune/Learning Rate": 7.579301075241808e-07, "Full-finetune/Loss": 0.728266179561615, "Full-finetune/Loss (Raw)": 0.8406928777694702, "Full-finetune/Step": 4569, "Full-finetune/Step Time": 6.70265338383615} {"Full-finetune/Learning Rate": 7.555528630173014e-07, "Full-finetune/Loss": 0.7287899255752563, "Full-finetune/Loss (Raw)": 0.8334307670593262, "Full-finetune/Step": 4570, "Full-finetune/Step Time": 6.690198605880141} {"Full-finetune/Learning Rate": 7.531792060998766e-07, "Full-finetune/Loss": 0.7274322509765625, "Full-finetune/Loss (Raw)": 0.6239405870437622, "Full-finetune/Step": 4571, "Full-finetune/Step Time": 6.708370365202427} {"Full-finetune/Learning Rate": 7.508091376930682e-07, "Full-finetune/Loss": 0.7277613282203674, "Full-finetune/Loss (Raw)": 0.7329875826835632, "Full-finetune/Step": 4572, "Full-finetune/Step Time": 6.714619619771838} {"Full-finetune/Learning Rate": 7.484426587166615e-07, "Full-finetune/Loss": 0.7276625633239746, "Full-finetune/Loss (Raw)": 0.6687718033790588, "Full-finetune/Step": 4573, "Full-finetune/Step Time": 6.729852819815278} {"Full-finetune/Learning Rate": 7.460797700890365e-07, "Full-finetune/Loss": 0.7271409630775452, "Full-finetune/Loss (Raw)": 0.7978896498680115, "Full-finetune/Step": 4574, "Full-finetune/Step Time": 6.7295031771063805} {"Full-finetune/Learning Rate": 7.437204727271852e-07, "Full-finetune/Loss": 0.7267029285430908, "Full-finetune/Loss (Raw)": 0.7091831564903259, "Full-finetune/Step": 4575, "Full-finetune/Step Time": 6.744238488376141} {"Full-finetune/Learning Rate": 7.41364767546705e-07, "Full-finetune/Loss": 0.7259829044342041, "Full-finetune/Loss (Raw)": 0.6771007776260376, "Full-finetune/Step": 4576, "Full-finetune/Step Time": 6.7624186258763075} {"Full-finetune/Learning Rate": 7.390126554617982e-07, "Full-finetune/Loss": 0.7261168956756592, "Full-finetune/Loss (Raw)": 0.9038958549499512, "Full-finetune/Step": 4577, "Full-finetune/Step Time": 6.7658411134034395} {"Full-finetune/Learning Rate": 7.366641373852745e-07, "Full-finetune/Loss": 0.7272758483886719, "Full-finetune/Loss (Raw)": 0.8375060558319092, "Full-finetune/Step": 4578, "Full-finetune/Step Time": 6.758910974487662} {"Full-finetune/Learning Rate": 7.343192142285471e-07, "Full-finetune/Loss": 0.7257306575775146, "Full-finetune/Loss (Raw)": 0.41791895031929016, "Full-finetune/Step": 4579, "Full-finetune/Step Time": 6.765332700684667} {"Full-finetune/Learning Rate": 7.319778869016358e-07, "Full-finetune/Loss": 0.7250794768333435, "Full-finetune/Loss (Raw)": 0.2689240276813507, "Full-finetune/Step": 4580, "Full-finetune/Step Time": 6.768026316538453} {"Full-finetune/Learning Rate": 7.296401563131661e-07, "Full-finetune/Loss": 0.7247080206871033, "Full-finetune/Loss (Raw)": 0.6494855880737305, "Full-finetune/Step": 4581, "Full-finetune/Step Time": 6.776344792917371} {"Full-finetune/Learning Rate": 7.273060233703599e-07, "Full-finetune/Loss": 0.7246949076652527, "Full-finetune/Loss (Raw)": 0.7200706601142883, "Full-finetune/Step": 4582, "Full-finetune/Step Time": 6.780477037653327} {"Full-finetune/Learning Rate": 7.249754889790539e-07, "Full-finetune/Loss": 0.72693932056427, "Full-finetune/Loss (Raw)": 0.8891089558601379, "Full-finetune/Step": 4583, "Full-finetune/Step Time": 6.7671132273972034} {"Full-finetune/Learning Rate": 7.2264855404368e-07, "Full-finetune/Loss": 0.7267774343490601, "Full-finetune/Loss (Raw)": 0.6987581253051758, "Full-finetune/Step": 4584, "Full-finetune/Step Time": 6.767731238156557} {"Full-finetune/Learning Rate": 7.203252194672783e-07, "Full-finetune/Loss": 0.7265636920928955, "Full-finetune/Loss (Raw)": 0.6298338770866394, "Full-finetune/Step": 4585, "Full-finetune/Step Time": 6.756752606481314} {"Full-finetune/Learning Rate": 7.180054861514885e-07, "Full-finetune/Loss": 0.7266663312911987, "Full-finetune/Loss (Raw)": 0.6783502101898193, "Full-finetune/Step": 4586, "Full-finetune/Step Time": 6.749269437044859} {"Full-finetune/Learning Rate": 7.156893549965538e-07, "Full-finetune/Loss": 0.7280488014221191, "Full-finetune/Loss (Raw)": 0.8723617196083069, "Full-finetune/Step": 4587, "Full-finetune/Step Time": 6.744608810171485} {"Full-finetune/Learning Rate": 7.133768269013186e-07, "Full-finetune/Loss": 0.7279371023178101, "Full-finetune/Loss (Raw)": 0.8480180501937866, "Full-finetune/Step": 4588, "Full-finetune/Step Time": 6.7686333656311035} {"Full-finetune/Learning Rate": 7.110679027632295e-07, "Full-finetune/Loss": 0.7285344004631042, "Full-finetune/Loss (Raw)": 0.7654773592948914, "Full-finetune/Step": 4589, "Full-finetune/Step Time": 6.779607109725475} {"Full-finetune/Learning Rate": 7.087625834783385e-07, "Full-finetune/Loss": 0.728083610534668, "Full-finetune/Loss (Raw)": 0.7098209857940674, "Full-finetune/Step": 4590, "Full-finetune/Step Time": 6.769616277888417} {"Full-finetune/Learning Rate": 7.064608699412944e-07, "Full-finetune/Loss": 0.7303436994552612, "Full-finetune/Loss (Raw)": 0.8094055652618408, "Full-finetune/Step": 4591, "Full-finetune/Step Time": 6.753617728129029} {"Full-finetune/Learning Rate": 7.041627630453418e-07, "Full-finetune/Loss": 0.7267289161682129, "Full-finetune/Loss (Raw)": 0.2909941077232361, "Full-finetune/Step": 4592, "Full-finetune/Step Time": 6.752621926367283} {"Full-finetune/Learning Rate": 7.018682636823371e-07, "Full-finetune/Loss": 0.7275399565696716, "Full-finetune/Loss (Raw)": 0.8265577554702759, "Full-finetune/Step": 4593, "Full-finetune/Step Time": 6.758581411093473} {"Full-finetune/Learning Rate": 6.99577372742728e-07, "Full-finetune/Loss": 0.7249376177787781, "Full-finetune/Loss (Raw)": 0.4120197892189026, "Full-finetune/Step": 4594, "Full-finetune/Step Time": 6.775334298610687} {"Full-finetune/Learning Rate": 6.972900911155656e-07, "Full-finetune/Loss": 0.724622905254364, "Full-finetune/Loss (Raw)": 0.788123369216919, "Full-finetune/Step": 4595, "Full-finetune/Step Time": 6.778594836592674} {"Full-finetune/Learning Rate": 6.950064196884987e-07, "Full-finetune/Loss": 0.724680483341217, "Full-finetune/Loss (Raw)": 0.7466952800750732, "Full-finetune/Step": 4596, "Full-finetune/Step Time": 6.785322403535247} {"Full-finetune/Learning Rate": 6.927263593477751e-07, "Full-finetune/Loss": 0.7243185639381409, "Full-finetune/Loss (Raw)": 0.7026070356369019, "Full-finetune/Step": 4597, "Full-finetune/Step Time": 6.787654243409634} {"Full-finetune/Learning Rate": 6.904499109782426e-07, "Full-finetune/Loss": 0.7243915796279907, "Full-finetune/Loss (Raw)": 0.7704610228538513, "Full-finetune/Step": 4598, "Full-finetune/Step Time": 6.785504087805748} {"Full-finetune/Learning Rate": 6.881770754633444e-07, "Full-finetune/Loss": 0.7251180410385132, "Full-finetune/Loss (Raw)": 0.7040859460830688, "Full-finetune/Step": 4599, "Full-finetune/Step Time": 6.7791059128940105} {"Full-finetune/Learning Rate": 6.859078536851283e-07, "Full-finetune/Loss": 0.7248431444168091, "Full-finetune/Loss (Raw)": 0.7708733677864075, "Full-finetune/Step": 4600, "Full-finetune/Step Time": 6.771308993920684} {"Full-finetune/Learning Rate": 6.836422465242298e-07, "Full-finetune/Loss": 0.7223446369171143, "Full-finetune/Loss (Raw)": 0.6674493551254272, "Full-finetune/Step": 4601, "Full-finetune/Step Time": 6.776692282408476} {"Full-finetune/Learning Rate": 6.81380254859888e-07, "Full-finetune/Loss": 0.7235409021377563, "Full-finetune/Loss (Raw)": 0.971763551235199, "Full-finetune/Step": 4602, "Full-finetune/Step Time": 6.7782214265316725} {"Full-finetune/Learning Rate": 6.791218795699406e-07, "Full-finetune/Loss": 0.7230677604675293, "Full-finetune/Loss (Raw)": 0.6173667907714844, "Full-finetune/Step": 4603, "Full-finetune/Step Time": 6.761661788448691} {"Full-finetune/Learning Rate": 6.768671215308165e-07, "Full-finetune/Loss": 0.7228379249572754, "Full-finetune/Loss (Raw)": 0.8126038908958435, "Full-finetune/Step": 4604, "Full-finetune/Step Time": 6.755140336230397} {"Full-finetune/Learning Rate": 6.746159816175435e-07, "Full-finetune/Loss": 0.7198683619499207, "Full-finetune/Loss (Raw)": 0.3256024718284607, "Full-finetune/Step": 4605, "Full-finetune/Step Time": 6.773128384724259} {"Full-finetune/Learning Rate": 6.723684607037451e-07, "Full-finetune/Loss": 0.7196416854858398, "Full-finetune/Loss (Raw)": 0.8227831721305847, "Full-finetune/Step": 4606, "Full-finetune/Step Time": 6.782546538859606} {"Full-finetune/Learning Rate": 6.701245596616402e-07, "Full-finetune/Loss": 0.7192837595939636, "Full-finetune/Loss (Raw)": 0.7280311584472656, "Full-finetune/Step": 4607, "Full-finetune/Step Time": 6.774608043953776} {"Full-finetune/Learning Rate": 6.678842793620433e-07, "Full-finetune/Loss": 0.7201700210571289, "Full-finetune/Loss (Raw)": 0.7683016657829285, "Full-finetune/Step": 4608, "Full-finetune/Step Time": 6.7706332970410585} {"Full-finetune/Learning Rate": 6.656476206743623e-07, "Full-finetune/Loss": 0.7221336364746094, "Full-finetune/Loss (Raw)": 0.9510498046875, "Full-finetune/Step": 4609, "Full-finetune/Step Time": 6.768799841403961} {"Full-finetune/Learning Rate": 6.634145844666051e-07, "Full-finetune/Loss": 0.7223310470581055, "Full-finetune/Loss (Raw)": 0.721817135810852, "Full-finetune/Step": 4610, "Full-finetune/Step Time": 6.75177701190114} {"Full-finetune/Learning Rate": 6.611851716053641e-07, "Full-finetune/Loss": 0.718380331993103, "Full-finetune/Loss (Raw)": 0.28831276297569275, "Full-finetune/Step": 4611, "Full-finetune/Step Time": 6.769328603520989} {"Full-finetune/Learning Rate": 6.589593829558316e-07, "Full-finetune/Loss": 0.7176938652992249, "Full-finetune/Loss (Raw)": 0.7871541976928711, "Full-finetune/Step": 4612, "Full-finetune/Step Time": 6.751699918881059} {"Full-finetune/Learning Rate": 6.567372193817967e-07, "Full-finetune/Loss": 0.7167022228240967, "Full-finetune/Loss (Raw)": 0.6292929649353027, "Full-finetune/Step": 4613, "Full-finetune/Step Time": 6.7862746473401785} {"Full-finetune/Learning Rate": 6.545186817456339e-07, "Full-finetune/Loss": 0.7165317535400391, "Full-finetune/Loss (Raw)": 0.7041298747062683, "Full-finetune/Step": 4614, "Full-finetune/Step Time": 6.79333265684545} {"Full-finetune/Learning Rate": 6.523037709083158e-07, "Full-finetune/Loss": 0.7209376096725464, "Full-finetune/Loss (Raw)": 0.8916472792625427, "Full-finetune/Step": 4615, "Full-finetune/Step Time": 6.784417971968651} {"Full-finetune/Learning Rate": 6.500924877294057e-07, "Full-finetune/Loss": 0.7204245328903198, "Full-finetune/Loss (Raw)": 0.7265311479568481, "Full-finetune/Step": 4616, "Full-finetune/Step Time": 6.785205157473683} {"Full-finetune/Learning Rate": 6.478848330670607e-07, "Full-finetune/Loss": 0.7218093276023865, "Full-finetune/Loss (Raw)": 0.6573164463043213, "Full-finetune/Step": 4617, "Full-finetune/Step Time": 6.777039799839258} {"Full-finetune/Learning Rate": 6.456808077780263e-07, "Full-finetune/Loss": 0.723247766494751, "Full-finetune/Loss (Raw)": 0.8344589471817017, "Full-finetune/Step": 4618, "Full-finetune/Step Time": 6.76851430721581} {"Full-finetune/Learning Rate": 6.434804127176409e-07, "Full-finetune/Loss": 0.7244399785995483, "Full-finetune/Loss (Raw)": 0.8846567273139954, "Full-finetune/Step": 4619, "Full-finetune/Step Time": 6.7705984339118} {"Full-finetune/Learning Rate": 6.4128364873984e-07, "Full-finetune/Loss": 0.7236751914024353, "Full-finetune/Loss (Raw)": 0.716225802898407, "Full-finetune/Step": 4620, "Full-finetune/Step Time": 6.77381855994463} {"Full-finetune/Learning Rate": 6.390905166971406e-07, "Full-finetune/Loss": 0.7250732183456421, "Full-finetune/Loss (Raw)": 0.8784161806106567, "Full-finetune/Step": 4621, "Full-finetune/Step Time": 6.764697656035423} {"Full-finetune/Learning Rate": 6.369010174406531e-07, "Full-finetune/Loss": 0.7267507910728455, "Full-finetune/Loss (Raw)": 0.8138770461082458, "Full-finetune/Step": 4622, "Full-finetune/Step Time": 6.761970084160566} {"Full-finetune/Learning Rate": 6.347151518200834e-07, "Full-finetune/Loss": 0.7273532152175903, "Full-finetune/Loss (Raw)": 0.8500052690505981, "Full-finetune/Step": 4623, "Full-finetune/Step Time": 6.759136650711298} {"Full-finetune/Learning Rate": 6.325329206837217e-07, "Full-finetune/Loss": 0.7275906801223755, "Full-finetune/Loss (Raw)": 0.7281562089920044, "Full-finetune/Step": 4624, "Full-finetune/Step Time": 6.751255916431546} {"Full-finetune/Learning Rate": 6.303543248784483e-07, "Full-finetune/Loss": 0.7282498478889465, "Full-finetune/Loss (Raw)": 0.8467345833778381, "Full-finetune/Step": 4625, "Full-finetune/Step Time": 6.74197349883616} {"Full-finetune/Learning Rate": 6.281793652497359e-07, "Full-finetune/Loss": 0.7257365584373474, "Full-finetune/Loss (Raw)": 0.34788382053375244, "Full-finetune/Step": 4626, "Full-finetune/Step Time": 6.754691295325756} {"Full-finetune/Learning Rate": 6.260080426416415e-07, "Full-finetune/Loss": 0.7247552871704102, "Full-finetune/Loss (Raw)": 0.7464198470115662, "Full-finetune/Step": 4627, "Full-finetune/Step Time": 6.765674974769354} {"Full-finetune/Learning Rate": 6.238403578968144e-07, "Full-finetune/Loss": 0.7245916724205017, "Full-finetune/Loss (Raw)": 0.6562528610229492, "Full-finetune/Step": 4628, "Full-finetune/Step Time": 6.7691435515880585} {"Full-finetune/Learning Rate": 6.216763118564883e-07, "Full-finetune/Loss": 0.7235674858093262, "Full-finetune/Loss (Raw)": 0.7380757331848145, "Full-finetune/Step": 4629, "Full-finetune/Step Time": 6.766400422900915} {"Full-finetune/Learning Rate": 6.195159053604926e-07, "Full-finetune/Loss": 0.723842442035675, "Full-finetune/Loss (Raw)": 0.6758610010147095, "Full-finetune/Step": 4630, "Full-finetune/Step Time": 6.775634096935391} {"Full-finetune/Learning Rate": 6.173591392472334e-07, "Full-finetune/Loss": 0.7215589880943298, "Full-finetune/Loss (Raw)": 0.5590633749961853, "Full-finetune/Step": 4631, "Full-finetune/Step Time": 6.795345028862357} {"Full-finetune/Learning Rate": 6.152060143537109e-07, "Full-finetune/Loss": 0.7177703380584717, "Full-finetune/Loss (Raw)": 0.2719103991985321, "Full-finetune/Step": 4632, "Full-finetune/Step Time": 6.807669220492244} {"Full-finetune/Learning Rate": 6.130565315155124e-07, "Full-finetune/Loss": 0.7164618968963623, "Full-finetune/Loss (Raw)": 0.5829977989196777, "Full-finetune/Step": 4633, "Full-finetune/Step Time": 6.822275064885616} {"Full-finetune/Learning Rate": 6.109106915668106e-07, "Full-finetune/Loss": 0.7157527208328247, "Full-finetune/Loss (Raw)": 0.859573483467102, "Full-finetune/Step": 4634, "Full-finetune/Step Time": 6.819826098158956} {"Full-finetune/Learning Rate": 6.087684953403616e-07, "Full-finetune/Loss": 0.7177073955535889, "Full-finetune/Loss (Raw)": 0.5634933710098267, "Full-finetune/Step": 4635, "Full-finetune/Step Time": 6.822261605411768} {"Full-finetune/Learning Rate": 6.066299436675117e-07, "Full-finetune/Loss": 0.7179566621780396, "Full-finetune/Loss (Raw)": 0.7864679098129272, "Full-finetune/Step": 4636, "Full-finetune/Step Time": 6.824597952887416} {"Full-finetune/Learning Rate": 6.04495037378191e-07, "Full-finetune/Loss": 0.7190725207328796, "Full-finetune/Loss (Raw)": 0.8520476818084717, "Full-finetune/Step": 4637, "Full-finetune/Step Time": 6.829597033560276} {"Full-finetune/Learning Rate": 6.023637773009139e-07, "Full-finetune/Loss": 0.7202168703079224, "Full-finetune/Loss (Raw)": 0.9891474843025208, "Full-finetune/Step": 4638, "Full-finetune/Step Time": 6.834669316187501} {"Full-finetune/Learning Rate": 6.002361642627807e-07, "Full-finetune/Loss": 0.7175891995429993, "Full-finetune/Loss (Raw)": 0.6009365916252136, "Full-finetune/Step": 4639, "Full-finetune/Step Time": 6.859424438327551} {"Full-finetune/Learning Rate": 5.981121990894789e-07, "Full-finetune/Loss": 0.7214462757110596, "Full-finetune/Loss (Raw)": 0.854945182800293, "Full-finetune/Step": 4640, "Full-finetune/Step Time": 6.849533272907138} {"Full-finetune/Learning Rate": 5.959918826052757e-07, "Full-finetune/Loss": 0.7215392589569092, "Full-finetune/Loss (Raw)": 0.7870021462440491, "Full-finetune/Step": 4641, "Full-finetune/Step Time": 6.8463003896176815} {"Full-finetune/Learning Rate": 5.938752156330241e-07, "Full-finetune/Loss": 0.7226313352584839, "Full-finetune/Loss (Raw)": 0.7873328328132629, "Full-finetune/Step": 4642, "Full-finetune/Step Time": 6.848949721083045} {"Full-finetune/Learning Rate": 5.917621989941624e-07, "Full-finetune/Loss": 0.7235183119773865, "Full-finetune/Loss (Raw)": 0.872094452381134, "Full-finetune/Step": 4643, "Full-finetune/Step Time": 6.846744649112225} {"Full-finetune/Learning Rate": 5.896528335087137e-07, "Full-finetune/Loss": 0.7217311859130859, "Full-finetune/Loss (Raw)": 0.37792930006980896, "Full-finetune/Step": 4644, "Full-finetune/Step Time": 6.865544147789478} {"Full-finetune/Learning Rate": 5.875471199952754e-07, "Full-finetune/Loss": 0.7229159474372864, "Full-finetune/Loss (Raw)": 0.8159458041191101, "Full-finetune/Step": 4645, "Full-finetune/Step Time": 6.875427199527621} {"Full-finetune/Learning Rate": 5.854450592710382e-07, "Full-finetune/Loss": 0.7206134796142578, "Full-finetune/Loss (Raw)": 0.6185512542724609, "Full-finetune/Step": 4646, "Full-finetune/Step Time": 6.889109184965491} {"Full-finetune/Learning Rate": 5.833466521517695e-07, "Full-finetune/Loss": 0.7199351191520691, "Full-finetune/Loss (Raw)": 0.7892962098121643, "Full-finetune/Step": 4647, "Full-finetune/Step Time": 6.894545424729586} {"Full-finetune/Learning Rate": 5.812518994518213e-07, "Full-finetune/Loss": 0.7188615798950195, "Full-finetune/Loss (Raw)": 0.7203969359397888, "Full-finetune/Step": 4648, "Full-finetune/Step Time": 6.908331690356135} {"Full-finetune/Learning Rate": 5.791608019841244e-07, "Full-finetune/Loss": 0.7183969020843506, "Full-finetune/Loss (Raw)": 0.8322640657424927, "Full-finetune/Step": 4649, "Full-finetune/Step Time": 6.89508774317801} {"Full-finetune/Learning Rate": 5.770733605601953e-07, "Full-finetune/Loss": 0.7177326679229736, "Full-finetune/Loss (Raw)": 0.6280850172042847, "Full-finetune/Step": 4650, "Full-finetune/Step Time": 6.903122089803219} {"Full-finetune/Learning Rate": 5.74989575990127e-07, "Full-finetune/Loss": 0.7177565097808838, "Full-finetune/Loss (Raw)": 0.7800097465515137, "Full-finetune/Step": 4651, "Full-finetune/Step Time": 6.896884199231863} {"Full-finetune/Learning Rate": 5.729094490825948e-07, "Full-finetune/Loss": 0.7174282670021057, "Full-finetune/Loss (Raw)": 0.7603220343589783, "Full-finetune/Step": 4652, "Full-finetune/Step Time": 6.894488701596856} {"Full-finetune/Learning Rate": 5.708329806448587e-07, "Full-finetune/Loss": 0.7137981057167053, "Full-finetune/Loss (Raw)": 0.3042665719985962, "Full-finetune/Step": 4653, "Full-finetune/Step Time": 6.909697813913226} {"Full-finetune/Learning Rate": 5.687601714827551e-07, "Full-finetune/Loss": 0.7150663733482361, "Full-finetune/Loss (Raw)": 0.7275791764259338, "Full-finetune/Step": 4654, "Full-finetune/Step Time": 6.915329050272703} {"Full-finetune/Learning Rate": 5.666910224006971e-07, "Full-finetune/Loss": 0.7164244651794434, "Full-finetune/Loss (Raw)": 0.6155171990394592, "Full-finetune/Step": 4655, "Full-finetune/Step Time": 6.899308523163199} {"Full-finetune/Learning Rate": 5.646255342016848e-07, "Full-finetune/Loss": 0.7142625451087952, "Full-finetune/Loss (Raw)": 0.4363488256931305, "Full-finetune/Step": 4656, "Full-finetune/Step Time": 6.914600361138582} {"Full-finetune/Learning Rate": 5.625637076872914e-07, "Full-finetune/Loss": 0.713741660118103, "Full-finetune/Loss (Raw)": 0.7348752021789551, "Full-finetune/Step": 4657, "Full-finetune/Step Time": 6.913160840049386} {"Full-finetune/Learning Rate": 5.605055436576746e-07, "Full-finetune/Loss": 0.7135710120201111, "Full-finetune/Loss (Raw)": 0.8418744206428528, "Full-finetune/Step": 4658, "Full-finetune/Step Time": 6.908780807629228} {"Full-finetune/Learning Rate": 5.584510429115641e-07, "Full-finetune/Loss": 0.713362455368042, "Full-finetune/Loss (Raw)": 0.8608887791633606, "Full-finetune/Step": 4659, "Full-finetune/Step Time": 6.896164268255234} {"Full-finetune/Learning Rate": 5.564002062462737e-07, "Full-finetune/Loss": 0.7150907516479492, "Full-finetune/Loss (Raw)": 0.827991247177124, "Full-finetune/Step": 4660, "Full-finetune/Step Time": 6.902421901002526} {"Full-finetune/Learning Rate": 5.54353034457693e-07, "Full-finetune/Loss": 0.7139749526977539, "Full-finetune/Loss (Raw)": 0.6632585525512695, "Full-finetune/Step": 4661, "Full-finetune/Step Time": 6.903878157958388} {"Full-finetune/Learning Rate": 5.523095283402868e-07, "Full-finetune/Loss": 0.7178826332092285, "Full-finetune/Loss (Raw)": 0.8398396968841553, "Full-finetune/Step": 4662, "Full-finetune/Step Time": 6.88561824336648} {"Full-finetune/Learning Rate": 5.502696886871039e-07, "Full-finetune/Loss": 0.7193188071250916, "Full-finetune/Loss (Raw)": 0.8155403733253479, "Full-finetune/Step": 4663, "Full-finetune/Step Time": 6.873098189011216} {"Full-finetune/Learning Rate": 5.482335162897657e-07, "Full-finetune/Loss": 0.7194509506225586, "Full-finetune/Loss (Raw)": 0.7784032225608826, "Full-finetune/Step": 4664, "Full-finetune/Step Time": 6.860706157982349} {"Full-finetune/Learning Rate": 5.462010119384665e-07, "Full-finetune/Loss": 0.7185998558998108, "Full-finetune/Loss (Raw)": 0.6905917525291443, "Full-finetune/Step": 4665, "Full-finetune/Step Time": 6.860248493030667} {"Full-finetune/Learning Rate": 5.441721764219853e-07, "Full-finetune/Loss": 0.7165435552597046, "Full-finetune/Loss (Raw)": 0.5681635141372681, "Full-finetune/Step": 4666, "Full-finetune/Step Time": 6.878228895366192} {"Full-finetune/Learning Rate": 5.42147010527675e-07, "Full-finetune/Loss": 0.7179957628250122, "Full-finetune/Loss (Raw)": 0.8281629085540771, "Full-finetune/Step": 4667, "Full-finetune/Step Time": 6.865935193374753} {"Full-finetune/Learning Rate": 5.401255150414598e-07, "Full-finetune/Loss": 0.7212421894073486, "Full-finetune/Loss (Raw)": 0.7250365018844604, "Full-finetune/Step": 4668, "Full-finetune/Step Time": 6.857550650835037} {"Full-finetune/Learning Rate": 5.381076907478433e-07, "Full-finetune/Loss": 0.7186564207077026, "Full-finetune/Loss (Raw)": 0.3545665442943573, "Full-finetune/Step": 4669, "Full-finetune/Step Time": 6.864478530362248} {"Full-finetune/Learning Rate": 5.360935384299049e-07, "Full-finetune/Loss": 0.7199439406394958, "Full-finetune/Loss (Raw)": 0.8027976751327515, "Full-finetune/Step": 4670, "Full-finetune/Step Time": 6.873999508097768} {"Full-finetune/Learning Rate": 5.34083058869297e-07, "Full-finetune/Loss": 0.7194544076919556, "Full-finetune/Loss (Raw)": 0.705243706703186, "Full-finetune/Step": 4671, "Full-finetune/Step Time": 6.868913605809212} {"Full-finetune/Learning Rate": 5.320762528462453e-07, "Full-finetune/Loss": 0.7190083861351013, "Full-finetune/Loss (Raw)": 0.8140431642532349, "Full-finetune/Step": 4672, "Full-finetune/Step Time": 6.882438721135259} {"Full-finetune/Learning Rate": 5.30073121139556e-07, "Full-finetune/Loss": 0.7193830609321594, "Full-finetune/Loss (Raw)": 0.7946476340293884, "Full-finetune/Step": 4673, "Full-finetune/Step Time": 6.871253767982125} {"Full-finetune/Learning Rate": 5.28073664526606e-07, "Full-finetune/Loss": 0.7194454669952393, "Full-finetune/Loss (Raw)": 0.8650894165039062, "Full-finetune/Step": 4674, "Full-finetune/Step Time": 6.893575733527541} {"Full-finetune/Learning Rate": 5.260778837833391e-07, "Full-finetune/Loss": 0.719551682472229, "Full-finetune/Loss (Raw)": 0.837935745716095, "Full-finetune/Step": 4675, "Full-finetune/Step Time": 6.8956367913633585} {"Full-finetune/Learning Rate": 5.240857796842846e-07, "Full-finetune/Loss": 0.7208952903747559, "Full-finetune/Loss (Raw)": 0.7398954629898071, "Full-finetune/Step": 4676, "Full-finetune/Step Time": 6.856303904205561} {"Full-finetune/Learning Rate": 5.220973530025375e-07, "Full-finetune/Loss": 0.7229498624801636, "Full-finetune/Loss (Raw)": 0.8438357710838318, "Full-finetune/Step": 4677, "Full-finetune/Step Time": 6.846055814996362} {"Full-finetune/Learning Rate": 5.201126045097671e-07, "Full-finetune/Loss": 0.721847653388977, "Full-finetune/Loss (Raw)": 0.754438042640686, "Full-finetune/Step": 4678, "Full-finetune/Step Time": 6.8706607315689325} {"Full-finetune/Learning Rate": 5.181315349762162e-07, "Full-finetune/Loss": 0.7203406691551208, "Full-finetune/Loss (Raw)": 0.6265897154808044, "Full-finetune/Step": 4679, "Full-finetune/Step Time": 6.869693322107196} {"Full-finetune/Learning Rate": 5.161541451706975e-07, "Full-finetune/Loss": 0.7226970195770264, "Full-finetune/Loss (Raw)": 0.8859209418296814, "Full-finetune/Step": 4680, "Full-finetune/Step Time": 6.861109901219606} {"Full-finetune/Learning Rate": 5.141804358605995e-07, "Full-finetune/Loss": 0.722244381904602, "Full-finetune/Loss (Raw)": 0.7405406832695007, "Full-finetune/Step": 4681, "Full-finetune/Step Time": 6.877320662140846} {"Full-finetune/Learning Rate": 5.122104078118773e-07, "Full-finetune/Loss": 0.7209345102310181, "Full-finetune/Loss (Raw)": 0.7268989682197571, "Full-finetune/Step": 4682, "Full-finetune/Step Time": 6.888365572318435} {"Full-finetune/Learning Rate": 5.102440617890658e-07, "Full-finetune/Loss": 0.7206701636314392, "Full-finetune/Loss (Raw)": 0.7715052962303162, "Full-finetune/Step": 4683, "Full-finetune/Step Time": 6.872913837432861} {"Full-finetune/Learning Rate": 5.082813985552604e-07, "Full-finetune/Loss": 0.7201555371284485, "Full-finetune/Loss (Raw)": 0.8676334023475647, "Full-finetune/Step": 4684, "Full-finetune/Step Time": 6.860053608193994} {"Full-finetune/Learning Rate": 5.063224188721327e-07, "Full-finetune/Loss": 0.7214182615280151, "Full-finetune/Loss (Raw)": 0.8407773971557617, "Full-finetune/Step": 4685, "Full-finetune/Step Time": 6.862778117880225} {"Full-finetune/Learning Rate": 5.043671234999271e-07, "Full-finetune/Loss": 0.7219858765602112, "Full-finetune/Loss (Raw)": 0.6448448300361633, "Full-finetune/Step": 4686, "Full-finetune/Step Time": 6.849964015185833} {"Full-finetune/Learning Rate": 5.02415513197454e-07, "Full-finetune/Loss": 0.7210242748260498, "Full-finetune/Loss (Raw)": 0.7820636630058289, "Full-finetune/Step": 4687, "Full-finetune/Step Time": 6.841362262144685} {"Full-finetune/Learning Rate": 5.004675887220978e-07, "Full-finetune/Loss": 0.7232660055160522, "Full-finetune/Loss (Raw)": 1.0164762735366821, "Full-finetune/Step": 4688, "Full-finetune/Step Time": 6.839628182351589} {"Full-finetune/Learning Rate": 4.985233508298037e-07, "Full-finetune/Loss": 0.7239521741867065, "Full-finetune/Loss (Raw)": 0.8725983500480652, "Full-finetune/Step": 4689, "Full-finetune/Step Time": 6.831087909638882} {"Full-finetune/Learning Rate": 4.965828002750994e-07, "Full-finetune/Loss": 0.7223933935165405, "Full-finetune/Loss (Raw)": 0.6386260986328125, "Full-finetune/Step": 4690, "Full-finetune/Step Time": 6.845338789746165} {"Full-finetune/Learning Rate": 4.946459378110701e-07, "Full-finetune/Loss": 0.7269114255905151, "Full-finetune/Loss (Raw)": 0.8901845812797546, "Full-finetune/Step": 4691, "Full-finetune/Step Time": 6.829949166625738} {"Full-finetune/Learning Rate": 4.927127641893759e-07, "Full-finetune/Loss": 0.7281962037086487, "Full-finetune/Loss (Raw)": 0.9248136281967163, "Full-finetune/Step": 4692, "Full-finetune/Step Time": 6.8294045105576515} {"Full-finetune/Learning Rate": 4.907832801602452e-07, "Full-finetune/Loss": 0.7288681268692017, "Full-finetune/Loss (Raw)": 0.7713766694068909, "Full-finetune/Step": 4693, "Full-finetune/Step Time": 6.8404121696949005} {"Full-finetune/Learning Rate": 4.888574864724715e-07, "Full-finetune/Loss": 0.7297878265380859, "Full-finetune/Loss (Raw)": 0.8858359456062317, "Full-finetune/Step": 4694, "Full-finetune/Step Time": 6.845461040735245} {"Full-finetune/Learning Rate": 4.869353838734169e-07, "Full-finetune/Loss": 0.7292438745498657, "Full-finetune/Loss (Raw)": 0.8071221709251404, "Full-finetune/Step": 4695, "Full-finetune/Step Time": 6.861518256366253} {"Full-finetune/Learning Rate": 4.850169731090127e-07, "Full-finetune/Loss": 0.7281524538993835, "Full-finetune/Loss (Raw)": 0.6655327081680298, "Full-finetune/Step": 4696, "Full-finetune/Step Time": 6.871672661975026} {"Full-finetune/Learning Rate": 4.831022549237574e-07, "Full-finetune/Loss": 0.7273775935173035, "Full-finetune/Loss (Raw)": 0.7415130138397217, "Full-finetune/Step": 4697, "Full-finetune/Step Time": 6.866428064182401} {"Full-finetune/Learning Rate": 4.811912300607169e-07, "Full-finetune/Loss": 0.7275606989860535, "Full-finetune/Loss (Raw)": 0.8568665385246277, "Full-finetune/Step": 4698, "Full-finetune/Step Time": 6.870158564299345} {"Full-finetune/Learning Rate": 4.792838992615179e-07, "Full-finetune/Loss": 0.7296297550201416, "Full-finetune/Loss (Raw)": 0.8887786865234375, "Full-finetune/Step": 4699, "Full-finetune/Step Time": 6.87470487318933} {"Full-finetune/Learning Rate": 4.773802632663627e-07, "Full-finetune/Loss": 0.7299805879592896, "Full-finetune/Loss (Raw)": 0.7778962254524231, "Full-finetune/Step": 4700, "Full-finetune/Step Time": 6.870111199095845} {"Full-finetune/Learning Rate": 4.7548032281401367e-07, "Full-finetune/Loss": 0.7311657667160034, "Full-finetune/Loss (Raw)": 0.82047039270401, "Full-finetune/Step": 4701, "Full-finetune/Step Time": 6.857922378927469} {"Full-finetune/Learning Rate": 4.735840786418011e-07, "Full-finetune/Loss": 0.7307741641998291, "Full-finetune/Loss (Raw)": 0.7477629780769348, "Full-finetune/Step": 4702, "Full-finetune/Step Time": 6.85966245085001} {"Full-finetune/Learning Rate": 4.716915314856196e-07, "Full-finetune/Loss": 0.7322157025337219, "Full-finetune/Loss (Raw)": 0.893703043460846, "Full-finetune/Step": 4703, "Full-finetune/Step Time": 6.841755528002977} {"Full-finetune/Learning Rate": 4.698026820799306e-07, "Full-finetune/Loss": 0.7320849895477295, "Full-finetune/Loss (Raw)": 0.6603692173957825, "Full-finetune/Step": 4704, "Full-finetune/Step Time": 6.825577054172754} {"Full-finetune/Learning Rate": 4.6791753115775884e-07, "Full-finetune/Loss": 0.7318538427352905, "Full-finetune/Loss (Raw)": 0.8743106722831726, "Full-finetune/Step": 4705, "Full-finetune/Step Time": 6.826459454372525} {"Full-finetune/Learning Rate": 4.660360794506946e-07, "Full-finetune/Loss": 0.7309038639068604, "Full-finetune/Loss (Raw)": 0.7159036993980408, "Full-finetune/Step": 4706, "Full-finetune/Step Time": 6.8200414180755615} {"Full-finetune/Learning Rate": 4.641583276888939e-07, "Full-finetune/Loss": 0.7329833507537842, "Full-finetune/Loss (Raw)": 0.6841052174568176, "Full-finetune/Step": 4707, "Full-finetune/Step Time": 6.80431804805994} {"Full-finetune/Learning Rate": 4.622842766010749e-07, "Full-finetune/Loss": 0.7362018823623657, "Full-finetune/Loss (Raw)": 0.6808887720108032, "Full-finetune/Step": 4708, "Full-finetune/Step Time": 6.785457156598568} {"Full-finetune/Learning Rate": 4.604139269145191e-07, "Full-finetune/Loss": 0.7363767623901367, "Full-finetune/Loss (Raw)": 0.6718724370002747, "Full-finetune/Step": 4709, "Full-finetune/Step Time": 6.788403971120715} {"Full-finetune/Learning Rate": 4.585472793550738e-07, "Full-finetune/Loss": 0.7381491661071777, "Full-finetune/Loss (Raw)": 0.94693523645401, "Full-finetune/Step": 4710, "Full-finetune/Step Time": 6.778484705835581} {"Full-finetune/Learning Rate": 4.5668433464714835e-07, "Full-finetune/Loss": 0.7372757196426392, "Full-finetune/Loss (Raw)": 0.7773096561431885, "Full-finetune/Step": 4711, "Full-finetune/Step Time": 6.778721084818244} {"Full-finetune/Learning Rate": 4.548250935137144e-07, "Full-finetune/Loss": 0.7372266054153442, "Full-finetune/Loss (Raw)": 0.6924750208854675, "Full-finetune/Step": 4712, "Full-finetune/Step Time": 6.7914179395884275} {"Full-finetune/Learning Rate": 4.5296955667630926e-07, "Full-finetune/Loss": 0.7388807535171509, "Full-finetune/Loss (Raw)": 0.8415641784667969, "Full-finetune/Step": 4713, "Full-finetune/Step Time": 6.793704705312848} {"Full-finetune/Learning Rate": 4.51117724855028e-07, "Full-finetune/Loss": 0.7393643260002136, "Full-finetune/Loss (Raw)": 0.7402450442314148, "Full-finetune/Step": 4714, "Full-finetune/Step Time": 6.785574838519096} {"Full-finetune/Learning Rate": 4.492695987685314e-07, "Full-finetune/Loss": 0.738594651222229, "Full-finetune/Loss (Raw)": 0.7738401889801025, "Full-finetune/Step": 4715, "Full-finetune/Step Time": 6.788057040423155} {"Full-finetune/Learning Rate": 4.474251791340412e-07, "Full-finetune/Loss": 0.7377325296401978, "Full-finetune/Loss (Raw)": 0.7376642227172852, "Full-finetune/Step": 4716, "Full-finetune/Step Time": 6.764394540339708} {"Full-finetune/Learning Rate": 4.4558446666734147e-07, "Full-finetune/Loss": 0.7380945682525635, "Full-finetune/Loss (Raw)": 0.8118203282356262, "Full-finetune/Step": 4717, "Full-finetune/Step Time": 6.762006338685751} {"Full-finetune/Learning Rate": 4.437474620827764e-07, "Full-finetune/Loss": 0.7369221448898315, "Full-finetune/Loss (Raw)": 0.5597537159919739, "Full-finetune/Step": 4718, "Full-finetune/Step Time": 6.775735709816217} {"Full-finetune/Learning Rate": 4.419141660932491e-07, "Full-finetune/Loss": 0.7356507778167725, "Full-finetune/Loss (Raw)": 0.6466690301895142, "Full-finetune/Step": 4719, "Full-finetune/Step Time": 6.775671925395727} {"Full-finetune/Learning Rate": 4.4008457941022933e-07, "Full-finetune/Loss": 0.7371646761894226, "Full-finetune/Loss (Raw)": 0.4847727417945862, "Full-finetune/Step": 4720, "Full-finetune/Step Time": 6.793345743790269} {"Full-finetune/Learning Rate": 4.382587027437435e-07, "Full-finetune/Loss": 0.7353087663650513, "Full-finetune/Loss (Raw)": 0.5890020132064819, "Full-finetune/Step": 4721, "Full-finetune/Step Time": 6.795726675540209} {"Full-finetune/Learning Rate": 4.364365368023771e-07, "Full-finetune/Loss": 0.7374986410140991, "Full-finetune/Loss (Raw)": 0.6923267245292664, "Full-finetune/Step": 4722, "Full-finetune/Step Time": 6.780633712187409} {"Full-finetune/Learning Rate": 4.3461808229327773e-07, "Full-finetune/Loss": 0.7373789548873901, "Full-finetune/Loss (Raw)": 0.772804319858551, "Full-finetune/Step": 4723, "Full-finetune/Step Time": 6.777698069810867} {"Full-finetune/Learning Rate": 4.3280333992215206e-07, "Full-finetune/Loss": 0.736380398273468, "Full-finetune/Loss (Raw)": 0.6188763976097107, "Full-finetune/Step": 4724, "Full-finetune/Step Time": 6.798900628462434} {"Full-finetune/Learning Rate": 4.3099231039326785e-07, "Full-finetune/Loss": 0.7363836765289307, "Full-finetune/Loss (Raw)": 0.7030253410339355, "Full-finetune/Step": 4725, "Full-finetune/Step Time": 6.800705585628748} {"Full-finetune/Learning Rate": 4.291849944094462e-07, "Full-finetune/Loss": 0.7355936765670776, "Full-finetune/Loss (Raw)": 0.6693429946899414, "Full-finetune/Step": 4726, "Full-finetune/Step Time": 6.802871063351631} {"Full-finetune/Learning Rate": 4.273813926720771e-07, "Full-finetune/Loss": 0.7356821894645691, "Full-finetune/Loss (Raw)": 0.7154179215431213, "Full-finetune/Step": 4727, "Full-finetune/Step Time": 6.810759631916881} {"Full-finetune/Learning Rate": 4.2558150588109747e-07, "Full-finetune/Loss": 0.7351226806640625, "Full-finetune/Loss (Raw)": 0.6992587447166443, "Full-finetune/Step": 4728, "Full-finetune/Step Time": 6.813198013231158} {"Full-finetune/Learning Rate": 4.237853347350107e-07, "Full-finetune/Loss": 0.7379459142684937, "Full-finetune/Loss (Raw)": 1.0288195610046387, "Full-finetune/Step": 4729, "Full-finetune/Step Time": 6.809321068227291} {"Full-finetune/Learning Rate": 4.219928799308759e-07, "Full-finetune/Loss": 0.7363743185997009, "Full-finetune/Loss (Raw)": 0.7705981135368347, "Full-finetune/Step": 4730, "Full-finetune/Step Time": 6.804837210103869} {"Full-finetune/Learning Rate": 4.2020414216431106e-07, "Full-finetune/Loss": 0.7364024519920349, "Full-finetune/Loss (Raw)": 0.6209661960601807, "Full-finetune/Step": 4731, "Full-finetune/Step Time": 6.803251124918461} {"Full-finetune/Learning Rate": 4.184191221294875e-07, "Full-finetune/Loss": 0.736229658126831, "Full-finetune/Loss (Raw)": 0.7904853224754333, "Full-finetune/Step": 4732, "Full-finetune/Step Time": 6.802903264760971} {"Full-finetune/Learning Rate": 4.16637820519139e-07, "Full-finetune/Loss": 0.7402587532997131, "Full-finetune/Loss (Raw)": 0.8413294553756714, "Full-finetune/Step": 4733, "Full-finetune/Step Time": 6.7795144729316235} {"Full-finetune/Learning Rate": 4.1486023802455253e-07, "Full-finetune/Loss": 0.7391811609268188, "Full-finetune/Loss (Raw)": 0.6848430037498474, "Full-finetune/Step": 4734, "Full-finetune/Step Time": 6.784273711964488} {"Full-finetune/Learning Rate": 4.1308637533557294e-07, "Full-finetune/Loss": 0.7401164770126343, "Full-finetune/Loss (Raw)": 0.847755491733551, "Full-finetune/Step": 4735, "Full-finetune/Step Time": 6.784819979220629} {"Full-finetune/Learning Rate": 4.1131623314060175e-07, "Full-finetune/Loss": 0.7387406826019287, "Full-finetune/Loss (Raw)": 0.5922008156776428, "Full-finetune/Step": 4736, "Full-finetune/Step Time": 6.7942510433495045} {"Full-finetune/Learning Rate": 4.095498121265995e-07, "Full-finetune/Loss": 0.736595094203949, "Full-finetune/Loss (Raw)": 0.6764177083969116, "Full-finetune/Step": 4737, "Full-finetune/Step Time": 6.803407529368997} {"Full-finetune/Learning Rate": 4.077871129790767e-07, "Full-finetune/Loss": 0.7374255657196045, "Full-finetune/Loss (Raw)": 0.8281195163726807, "Full-finetune/Step": 4738, "Full-finetune/Step Time": 6.812410797923803} {"Full-finetune/Learning Rate": 4.060281363821017e-07, "Full-finetune/Loss": 0.7406299114227295, "Full-finetune/Loss (Raw)": 0.6984714865684509, "Full-finetune/Step": 4739, "Full-finetune/Step Time": 6.798506708815694} {"Full-finetune/Learning Rate": 4.042728830183018e-07, "Full-finetune/Loss": 0.7397277355194092, "Full-finetune/Loss (Raw)": 0.6716699004173279, "Full-finetune/Step": 4740, "Full-finetune/Step Time": 6.800492711365223} {"Full-finetune/Learning Rate": 4.025213535688544e-07, "Full-finetune/Loss": 0.7390131950378418, "Full-finetune/Loss (Raw)": 0.5378326773643494, "Full-finetune/Step": 4741, "Full-finetune/Step Time": 6.772719278931618} {"Full-finetune/Learning Rate": 4.007735487134956e-07, "Full-finetune/Loss": 0.7387341856956482, "Full-finetune/Loss (Raw)": 0.6684136986732483, "Full-finetune/Step": 4742, "Full-finetune/Step Time": 6.783357467502356} {"Full-finetune/Learning Rate": 3.99029469130513e-07, "Full-finetune/Loss": 0.7368195056915283, "Full-finetune/Loss (Raw)": 0.6465760469436646, "Full-finetune/Step": 4743, "Full-finetune/Step Time": 6.789959378540516} {"Full-finetune/Learning Rate": 3.972891154967506e-07, "Full-finetune/Loss": 0.7384791970252991, "Full-finetune/Loss (Raw)": 0.9389694333076477, "Full-finetune/Step": 4744, "Full-finetune/Step Time": 6.791586462408304} {"Full-finetune/Learning Rate": 3.95552488487605e-07, "Full-finetune/Loss": 0.7387877106666565, "Full-finetune/Loss (Raw)": 0.6968000531196594, "Full-finetune/Step": 4745, "Full-finetune/Step Time": 6.810915378853679} {"Full-finetune/Learning Rate": 3.9381958877702684e-07, "Full-finetune/Loss": 0.7376636266708374, "Full-finetune/Loss (Raw)": 0.6905831694602966, "Full-finetune/Step": 4746, "Full-finetune/Step Time": 6.8348807245492935} {"Full-finetune/Learning Rate": 3.920904170375239e-07, "Full-finetune/Loss": 0.7377492189407349, "Full-finetune/Loss (Raw)": 0.8956125378608704, "Full-finetune/Step": 4747, "Full-finetune/Step Time": 6.8341145273298025} {"Full-finetune/Learning Rate": 3.903649739401494e-07, "Full-finetune/Loss": 0.7386376857757568, "Full-finetune/Loss (Raw)": 0.8299521207809448, "Full-finetune/Step": 4748, "Full-finetune/Step Time": 6.829497497528791} {"Full-finetune/Learning Rate": 3.8864326015451536e-07, "Full-finetune/Loss": 0.7379007935523987, "Full-finetune/Loss (Raw)": 0.7840853333473206, "Full-finetune/Step": 4749, "Full-finetune/Step Time": 6.825599106028676} {"Full-finetune/Learning Rate": 3.8692527634878853e-07, "Full-finetune/Loss": 0.7372785806655884, "Full-finetune/Loss (Raw)": 0.7342315912246704, "Full-finetune/Step": 4750, "Full-finetune/Step Time": 6.828116737306118} {"Full-finetune/Learning Rate": 3.8521102318968106e-07, "Full-finetune/Loss": 0.7364145517349243, "Full-finetune/Loss (Raw)": 0.7394214868545532, "Full-finetune/Step": 4751, "Full-finetune/Step Time": 6.842959189787507} {"Full-finetune/Learning Rate": 3.835005013424631e-07, "Full-finetune/Loss": 0.7372115850448608, "Full-finetune/Loss (Raw)": 0.8301656246185303, "Full-finetune/Step": 4752, "Full-finetune/Step Time": 6.847269147634506} {"Full-finetune/Learning Rate": 3.8179371147095354e-07, "Full-finetune/Loss": 0.7360255718231201, "Full-finetune/Loss (Raw)": 0.6949208974838257, "Full-finetune/Step": 4753, "Full-finetune/Step Time": 6.838975021615624} {"Full-finetune/Learning Rate": 3.80090654237526e-07, "Full-finetune/Loss": 0.739437997341156, "Full-finetune/Loss (Raw)": 0.7846788167953491, "Full-finetune/Step": 4754, "Full-finetune/Step Time": 6.826131634414196} {"Full-finetune/Learning Rate": 3.7839133030310174e-07, "Full-finetune/Loss": 0.7395682334899902, "Full-finetune/Loss (Raw)": 0.7630948424339294, "Full-finetune/Step": 4755, "Full-finetune/Step Time": 6.824516512453556} {"Full-finetune/Learning Rate": 3.766957403271554e-07, "Full-finetune/Loss": 0.7403777837753296, "Full-finetune/Loss (Raw)": 0.7598745822906494, "Full-finetune/Step": 4756, "Full-finetune/Step Time": 6.818277781829238} {"Full-finetune/Learning Rate": 3.750038849677162e-07, "Full-finetune/Loss": 0.7406257390975952, "Full-finetune/Loss (Raw)": 0.7698106169700623, "Full-finetune/Step": 4757, "Full-finetune/Step Time": 6.816362712532282} {"Full-finetune/Learning Rate": 3.7331576488135456e-07, "Full-finetune/Loss": 0.7412198781967163, "Full-finetune/Loss (Raw)": 0.7519168853759766, "Full-finetune/Step": 4758, "Full-finetune/Step Time": 6.807227537035942} {"Full-finetune/Learning Rate": 3.716313807231986e-07, "Full-finetune/Loss": 0.7423275113105774, "Full-finetune/Loss (Raw)": 0.700839102268219, "Full-finetune/Step": 4759, "Full-finetune/Step Time": 6.79196877963841} {"Full-finetune/Learning Rate": 3.699507331469276e-07, "Full-finetune/Loss": 0.7463474273681641, "Full-finetune/Loss (Raw)": 0.7864595055580139, "Full-finetune/Step": 4760, "Full-finetune/Step Time": 6.777130367234349} {"Full-finetune/Learning Rate": 3.682738228047655e-07, "Full-finetune/Loss": 0.7464855313301086, "Full-finetune/Loss (Raw)": 0.6006702184677124, "Full-finetune/Step": 4761, "Full-finetune/Step Time": 6.778772048652172} {"Full-finetune/Learning Rate": 3.666006503474895e-07, "Full-finetune/Loss": 0.7453614473342896, "Full-finetune/Loss (Raw)": 0.7157024145126343, "Full-finetune/Step": 4762, "Full-finetune/Step Time": 6.7775335889309645} {"Full-finetune/Learning Rate": 3.6493121642442454e-07, "Full-finetune/Loss": 0.7469949722290039, "Full-finetune/Loss (Raw)": 0.7725744843482971, "Full-finetune/Step": 4763, "Full-finetune/Step Time": 6.772019639611244} {"Full-finetune/Learning Rate": 3.6326552168344574e-07, "Full-finetune/Loss": 0.7461129426956177, "Full-finetune/Loss (Raw)": 0.6735734343528748, "Full-finetune/Step": 4764, "Full-finetune/Step Time": 6.775200540199876} {"Full-finetune/Learning Rate": 3.616035667709761e-07, "Full-finetune/Loss": 0.7444994449615479, "Full-finetune/Loss (Raw)": 0.6455128788948059, "Full-finetune/Step": 4765, "Full-finetune/Step Time": 6.764126816764474} {"Full-finetune/Learning Rate": 3.5994535233198844e-07, "Full-finetune/Loss": 0.743939220905304, "Full-finetune/Loss (Raw)": 0.9174441695213318, "Full-finetune/Step": 4766, "Full-finetune/Step Time": 6.765829848125577} {"Full-finetune/Learning Rate": 3.582908790100037e-07, "Full-finetune/Loss": 0.7455615997314453, "Full-finetune/Loss (Raw)": 0.8085991144180298, "Full-finetune/Step": 4767, "Full-finetune/Step Time": 6.745923336595297} {"Full-finetune/Learning Rate": 3.566401474470904e-07, "Full-finetune/Loss": 0.7445738315582275, "Full-finetune/Loss (Raw)": 0.7285137176513672, "Full-finetune/Step": 4768, "Full-finetune/Step Time": 6.7499614134430885} {"Full-finetune/Learning Rate": 3.5499315828386393e-07, "Full-finetune/Loss": 0.7443098425865173, "Full-finetune/Loss (Raw)": 0.7532065510749817, "Full-finetune/Step": 4769, "Full-finetune/Step Time": 6.748285489156842} {"Full-finetune/Learning Rate": 3.533499121594919e-07, "Full-finetune/Loss": 0.7417702674865723, "Full-finetune/Loss (Raw)": 0.46226823329925537, "Full-finetune/Step": 4770, "Full-finetune/Step Time": 6.751484528183937} {"Full-finetune/Learning Rate": 3.5171040971168414e-07, "Full-finetune/Loss": 0.741937518119812, "Full-finetune/Loss (Raw)": 0.8935053944587708, "Full-finetune/Step": 4771, "Full-finetune/Step Time": 6.7543953862041235} {"Full-finetune/Learning Rate": 3.500746515766973e-07, "Full-finetune/Loss": 0.7443200349807739, "Full-finetune/Loss (Raw)": 0.6828876733779907, "Full-finetune/Step": 4772, "Full-finetune/Step Time": 6.74921553209424} {"Full-finetune/Learning Rate": 3.4844263838934023e-07, "Full-finetune/Loss": 0.7422404289245605, "Full-finetune/Loss (Raw)": 0.5497593879699707, "Full-finetune/Step": 4773, "Full-finetune/Step Time": 6.7374266013503075} {"Full-finetune/Learning Rate": 3.468143707829641e-07, "Full-finetune/Loss": 0.7435388565063477, "Full-finetune/Loss (Raw)": 0.7847563624382019, "Full-finetune/Step": 4774, "Full-finetune/Step Time": 6.728422157466412} {"Full-finetune/Learning Rate": 3.45189849389469e-07, "Full-finetune/Loss": 0.7441116571426392, "Full-finetune/Loss (Raw)": 0.8626099824905396, "Full-finetune/Step": 4775, "Full-finetune/Step Time": 6.724723951891065} {"Full-finetune/Learning Rate": 3.4356907483929835e-07, "Full-finetune/Loss": 0.7435743808746338, "Full-finetune/Loss (Raw)": 0.6516205668449402, "Full-finetune/Step": 4776, "Full-finetune/Step Time": 6.728339467197657} {"Full-finetune/Learning Rate": 3.419520477614435e-07, "Full-finetune/Loss": 0.7429630756378174, "Full-finetune/Loss (Raw)": 0.754019558429718, "Full-finetune/Step": 4777, "Full-finetune/Step Time": 6.72987818531692} {"Full-finetune/Learning Rate": 3.403387687834425e-07, "Full-finetune/Loss": 0.740318775177002, "Full-finetune/Loss (Raw)": 0.289614200592041, "Full-finetune/Step": 4778, "Full-finetune/Step Time": 6.733642015606165} {"Full-finetune/Learning Rate": 3.3872923853137453e-07, "Full-finetune/Loss": 0.7414125204086304, "Full-finetune/Loss (Raw)": 0.9200143814086914, "Full-finetune/Step": 4779, "Full-finetune/Step Time": 6.736539056524634} {"Full-finetune/Learning Rate": 3.371234576298699e-07, "Full-finetune/Loss": 0.7408802509307861, "Full-finetune/Loss (Raw)": 0.6921851634979248, "Full-finetune/Step": 4780, "Full-finetune/Step Time": 6.742088124155998} {"Full-finetune/Learning Rate": 3.355214267021012e-07, "Full-finetune/Loss": 0.743457555770874, "Full-finetune/Loss (Raw)": 0.6341582536697388, "Full-finetune/Step": 4781, "Full-finetune/Step Time": 6.743327854201198} {"Full-finetune/Learning Rate": 3.3392314636978117e-07, "Full-finetune/Loss": 0.742899477481842, "Full-finetune/Loss (Raw)": 0.6561524271965027, "Full-finetune/Step": 4782, "Full-finetune/Step Time": 6.7444165628403425} {"Full-finetune/Learning Rate": 3.323286172531748e-07, "Full-finetune/Loss": 0.7409787774085999, "Full-finetune/Loss (Raw)": 0.3696666955947876, "Full-finetune/Step": 4783, "Full-finetune/Step Time": 6.757667468860745} {"Full-finetune/Learning Rate": 3.30737839971087e-07, "Full-finetune/Loss": 0.7437022924423218, "Full-finetune/Loss (Raw)": 0.784961462020874, "Full-finetune/Step": 4784, "Full-finetune/Step Time": 6.733617475256324} {"Full-finetune/Learning Rate": 3.2915081514086624e-07, "Full-finetune/Loss": 0.7453919649124146, "Full-finetune/Loss (Raw)": 0.951149046421051, "Full-finetune/Step": 4785, "Full-finetune/Step Time": 6.744854075834155} {"Full-finetune/Learning Rate": 3.2756754337840757e-07, "Full-finetune/Loss": 0.7451897263526917, "Full-finetune/Loss (Raw)": 0.8159896731376648, "Full-finetune/Step": 4786, "Full-finetune/Step Time": 6.747083768248558} {"Full-finetune/Learning Rate": 3.2598802529814734e-07, "Full-finetune/Loss": 0.7443417310714722, "Full-finetune/Loss (Raw)": 0.7523471117019653, "Full-finetune/Step": 4787, "Full-finetune/Step Time": 6.743343763053417} {"Full-finetune/Learning Rate": 3.2441226151306403e-07, "Full-finetune/Loss": 0.7428544759750366, "Full-finetune/Loss (Raw)": 0.6376161575317383, "Full-finetune/Step": 4788, "Full-finetune/Step Time": 6.752696990966797} {"Full-finetune/Learning Rate": 3.228402526346819e-07, "Full-finetune/Loss": 0.742209255695343, "Full-finetune/Loss (Raw)": 0.5806811451911926, "Full-finetune/Step": 4789, "Full-finetune/Step Time": 6.752001259475946} {"Full-finetune/Learning Rate": 3.212719992730684e-07, "Full-finetune/Loss": 0.7419092059135437, "Full-finetune/Loss (Raw)": 0.8014282584190369, "Full-finetune/Step": 4790, "Full-finetune/Step Time": 6.758130118250847} {"Full-finetune/Learning Rate": 3.197075020368312e-07, "Full-finetune/Loss": 0.7379436492919922, "Full-finetune/Loss (Raw)": 0.30794835090637207, "Full-finetune/Step": 4791, "Full-finetune/Step Time": 6.777111653238535} {"Full-finetune/Learning Rate": 3.1814676153311907e-07, "Full-finetune/Loss": 0.7394563555717468, "Full-finetune/Loss (Raw)": 0.9720320701599121, "Full-finetune/Step": 4792, "Full-finetune/Step Time": 6.7832996509969234} {"Full-finetune/Learning Rate": 3.165897783676275e-07, "Full-finetune/Loss": 0.7405588030815125, "Full-finetune/Loss (Raw)": 0.8317039608955383, "Full-finetune/Step": 4793, "Full-finetune/Step Time": 6.781762043014169} {"Full-finetune/Learning Rate": 3.1503655314458983e-07, "Full-finetune/Loss": 0.7414823174476624, "Full-finetune/Loss (Raw)": 0.6863731741905212, "Full-finetune/Step": 4794, "Full-finetune/Step Time": 6.774451479315758} {"Full-finetune/Learning Rate": 3.134870864667827e-07, "Full-finetune/Loss": 0.7402402758598328, "Full-finetune/Loss (Raw)": 0.6691816449165344, "Full-finetune/Step": 4795, "Full-finetune/Step Time": 6.805945500731468} {"Full-finetune/Learning Rate": 3.1194137893552523e-07, "Full-finetune/Loss": 0.7411820888519287, "Full-finetune/Loss (Raw)": 0.8455840945243835, "Full-finetune/Step": 4796, "Full-finetune/Step Time": 6.800495384261012} {"Full-finetune/Learning Rate": 3.1039943115067417e-07, "Full-finetune/Loss": 0.7446472644805908, "Full-finetune/Loss (Raw)": 0.7981085777282715, "Full-finetune/Step": 4797, "Full-finetune/Step Time": 6.77256141230464} {"Full-finetune/Learning Rate": 3.088612437106309e-07, "Full-finetune/Loss": 0.7430020570755005, "Full-finetune/Loss (Raw)": 0.5922137498855591, "Full-finetune/Step": 4798, "Full-finetune/Step Time": 6.777773145586252} {"Full-finetune/Learning Rate": 3.0732681721233446e-07, "Full-finetune/Loss": 0.7440272569656372, "Full-finetune/Loss (Raw)": 0.836471438407898, "Full-finetune/Step": 4799, "Full-finetune/Step Time": 6.783107237890363} {"Full-finetune/Learning Rate": 3.057961522512676e-07, "Full-finetune/Loss": 0.7450064420700073, "Full-finetune/Loss (Raw)": 0.9393765330314636, "Full-finetune/Step": 4800, "Full-finetune/Step Time": 6.76719580963254} {"Full-finetune/Learning Rate": 3.0426924942145166e-07, "Full-finetune/Loss": 0.7454631328582764, "Full-finetune/Loss (Raw)": 0.8531086444854736, "Full-finetune/Step": 4801, "Full-finetune/Step Time": 6.760095547884703} {"Full-finetune/Learning Rate": 3.027461093154449e-07, "Full-finetune/Loss": 0.7448446154594421, "Full-finetune/Loss (Raw)": 0.7859129905700684, "Full-finetune/Step": 4802, "Full-finetune/Step Time": 6.734867667779326} {"Full-finetune/Learning Rate": 3.012267325243512e-07, "Full-finetune/Loss": 0.7449643015861511, "Full-finetune/Loss (Raw)": 0.8532564640045166, "Full-finetune/Step": 4803, "Full-finetune/Step Time": 6.739951426163316} {"Full-finetune/Learning Rate": 2.9971111963780994e-07, "Full-finetune/Loss": 0.7455399632453918, "Full-finetune/Loss (Raw)": 0.8135829567909241, "Full-finetune/Step": 4804, "Full-finetune/Step Time": 6.742049174383283} {"Full-finetune/Learning Rate": 2.981992712440007e-07, "Full-finetune/Loss": 0.7456172108650208, "Full-finetune/Loss (Raw)": 0.8537260293960571, "Full-finetune/Step": 4805, "Full-finetune/Step Time": 6.74825806543231} {"Full-finetune/Learning Rate": 2.96691187929643e-07, "Full-finetune/Loss": 0.7458832263946533, "Full-finetune/Loss (Raw)": 0.7884901165962219, "Full-finetune/Step": 4806, "Full-finetune/Step Time": 6.729495549574494} {"Full-finetune/Learning Rate": 2.9518687027999427e-07, "Full-finetune/Loss": 0.7469054460525513, "Full-finetune/Loss (Raw)": 0.7574295997619629, "Full-finetune/Step": 4807, "Full-finetune/Step Time": 6.731638083234429} {"Full-finetune/Learning Rate": 2.936863188788497e-07, "Full-finetune/Loss": 0.7470634579658508, "Full-finetune/Loss (Raw)": 0.9061503410339355, "Full-finetune/Step": 4808, "Full-finetune/Step Time": 6.728657329455018} {"Full-finetune/Learning Rate": 2.921895343085446e-07, "Full-finetune/Loss": 0.745657205581665, "Full-finetune/Loss (Raw)": 0.5605306625366211, "Full-finetune/Step": 4809, "Full-finetune/Step Time": 6.717682383954525} {"Full-finetune/Learning Rate": 2.9069651714995426e-07, "Full-finetune/Loss": 0.7458518147468567, "Full-finetune/Loss (Raw)": 0.7518138289451599, "Full-finetune/Step": 4810, "Full-finetune/Step Time": 6.708656072616577} {"Full-finetune/Learning Rate": 2.8920726798248645e-07, "Full-finetune/Loss": 0.7460020780563354, "Full-finetune/Loss (Raw)": 0.7907333374023438, "Full-finetune/Step": 4811, "Full-finetune/Step Time": 6.724510980769992} {"Full-finetune/Learning Rate": 2.8772178738408985e-07, "Full-finetune/Loss": 0.7453634738922119, "Full-finetune/Loss (Raw)": 0.7858932018280029, "Full-finetune/Step": 4812, "Full-finetune/Step Time": 6.723811201751232} {"Full-finetune/Learning Rate": 2.862400759312522e-07, "Full-finetune/Loss": 0.7425224184989929, "Full-finetune/Loss (Raw)": 0.4771277904510498, "Full-finetune/Step": 4813, "Full-finetune/Step Time": 6.737366354092956} {"Full-finetune/Learning Rate": 2.847621341989959e-07, "Full-finetune/Loss": 0.7438673973083496, "Full-finetune/Loss (Raw)": 0.8170037865638733, "Full-finetune/Step": 4814, "Full-finetune/Step Time": 6.729808934032917} {"Full-finetune/Learning Rate": 2.832879627608831e-07, "Full-finetune/Loss": 0.7446211576461792, "Full-finetune/Loss (Raw)": 0.8785443902015686, "Full-finetune/Step": 4815, "Full-finetune/Step Time": 6.726988598704338} {"Full-finetune/Learning Rate": 2.8181756218900626e-07, "Full-finetune/Loss": 0.7419136762619019, "Full-finetune/Loss (Raw)": 0.6699121594429016, "Full-finetune/Step": 4816, "Full-finetune/Step Time": 6.742712484672666} {"Full-finetune/Learning Rate": 2.803509330540033e-07, "Full-finetune/Loss": 0.7415726184844971, "Full-finetune/Loss (Raw)": 0.8289466500282288, "Full-finetune/Step": 4817, "Full-finetune/Step Time": 6.747043401002884} {"Full-finetune/Learning Rate": 2.788880759250423e-07, "Full-finetune/Loss": 0.7430922985076904, "Full-finetune/Loss (Raw)": 0.8331405520439148, "Full-finetune/Step": 4818, "Full-finetune/Step Time": 6.737642930820584} {"Full-finetune/Learning Rate": 2.7742899136983026e-07, "Full-finetune/Loss": 0.7429913878440857, "Full-finetune/Loss (Raw)": 0.8772723078727722, "Full-finetune/Step": 4819, "Full-finetune/Step Time": 6.735744718462229} {"Full-finetune/Learning Rate": 2.759736799546109e-07, "Full-finetune/Loss": 0.7425097227096558, "Full-finetune/Loss (Raw)": 0.8631556630134583, "Full-finetune/Step": 4820, "Full-finetune/Step Time": 6.737551024183631} {"Full-finetune/Learning Rate": 2.745221422441602e-07, "Full-finetune/Loss": 0.743346095085144, "Full-finetune/Loss (Raw)": 0.878441572189331, "Full-finetune/Step": 4821, "Full-finetune/Step Time": 6.735373931005597} {"Full-finetune/Learning Rate": 2.73074378801792e-07, "Full-finetune/Loss": 0.742904543876648, "Full-finetune/Loss (Raw)": 0.8293110132217407, "Full-finetune/Step": 4822, "Full-finetune/Step Time": 6.733258722350001} {"Full-finetune/Learning Rate": 2.7163039018935357e-07, "Full-finetune/Loss": 0.7432324290275574, "Full-finetune/Loss (Raw)": 0.8490926623344421, "Full-finetune/Step": 4823, "Full-finetune/Step Time": 6.7188152223825455} {"Full-finetune/Learning Rate": 2.701901769672322e-07, "Full-finetune/Loss": 0.7402821779251099, "Full-finetune/Loss (Raw)": 0.287899374961853, "Full-finetune/Step": 4824, "Full-finetune/Step Time": 6.721592467278242} {"Full-finetune/Learning Rate": 2.687537396943451e-07, "Full-finetune/Loss": 0.7402974367141724, "Full-finetune/Loss (Raw)": 0.7434700727462769, "Full-finetune/Step": 4825, "Full-finetune/Step Time": 6.736495945602655} {"Full-finetune/Learning Rate": 2.673210789281433e-07, "Full-finetune/Loss": 0.7396186590194702, "Full-finetune/Loss (Raw)": 0.7699756026268005, "Full-finetune/Step": 4826, "Full-finetune/Step Time": 6.734429758042097} {"Full-finetune/Learning Rate": 2.658921952246163e-07, "Full-finetune/Loss": 0.7393827438354492, "Full-finetune/Loss (Raw)": 0.8585874438285828, "Full-finetune/Step": 4827, "Full-finetune/Step Time": 6.714143363758922} {"Full-finetune/Learning Rate": 2.6446708913828725e-07, "Full-finetune/Loss": 0.7400350570678711, "Full-finetune/Loss (Raw)": 0.861393928527832, "Full-finetune/Step": 4828, "Full-finetune/Step Time": 6.719516508281231} {"Full-finetune/Learning Rate": 2.6304576122221035e-07, "Full-finetune/Loss": 0.7397282123565674, "Full-finetune/Loss (Raw)": 0.7811880707740784, "Full-finetune/Step": 4829, "Full-finetune/Step Time": 6.725852154195309} {"Full-finetune/Learning Rate": 2.616282120279745e-07, "Full-finetune/Loss": 0.739656925201416, "Full-finetune/Loss (Raw)": 0.738646924495697, "Full-finetune/Step": 4830, "Full-finetune/Step Time": 6.723398691043258} {"Full-finetune/Learning Rate": 2.602144421057051e-07, "Full-finetune/Loss": 0.7354246377944946, "Full-finetune/Loss (Raw)": 0.35195884108543396, "Full-finetune/Step": 4831, "Full-finetune/Step Time": 6.744427628815174} {"Full-finetune/Learning Rate": 2.5880445200405667e-07, "Full-finetune/Loss": 0.735533595085144, "Full-finetune/Loss (Raw)": 0.6743239164352417, "Full-finetune/Step": 4832, "Full-finetune/Step Time": 6.76762761361897} {"Full-finetune/Learning Rate": 2.573982422702193e-07, "Full-finetune/Loss": 0.7348215579986572, "Full-finetune/Loss (Raw)": 0.7831676006317139, "Full-finetune/Step": 4833, "Full-finetune/Step Time": 6.770254738628864} {"Full-finetune/Learning Rate": 2.559958134499163e-07, "Full-finetune/Loss": 0.7364423871040344, "Full-finetune/Loss (Raw)": 0.9233686327934265, "Full-finetune/Step": 4834, "Full-finetune/Step Time": 6.773457568138838} {"Full-finetune/Learning Rate": 2.545971660874036e-07, "Full-finetune/Loss": 0.7377032041549683, "Full-finetune/Loss (Raw)": 0.8454883694648743, "Full-finetune/Step": 4835, "Full-finetune/Step Time": 6.769610134884715} {"Full-finetune/Learning Rate": 2.532023007254658e-07, "Full-finetune/Loss": 0.7392470240592957, "Full-finetune/Loss (Raw)": 0.8784983158111572, "Full-finetune/Step": 4836, "Full-finetune/Step Time": 6.767372462898493} {"Full-finetune/Learning Rate": 2.518112179054266e-07, "Full-finetune/Loss": 0.7399972081184387, "Full-finetune/Loss (Raw)": 0.7678967714309692, "Full-finetune/Step": 4837, "Full-finetune/Step Time": 6.758487096056342} {"Full-finetune/Learning Rate": 2.504239181671353e-07, "Full-finetune/Loss": 0.7393319606781006, "Full-finetune/Loss (Raw)": 0.8617785573005676, "Full-finetune/Step": 4838, "Full-finetune/Step Time": 6.759290775284171} {"Full-finetune/Learning Rate": 2.490404020489767e-07, "Full-finetune/Loss": 0.7385727167129517, "Full-finetune/Loss (Raw)": 0.6801273822784424, "Full-finetune/Step": 4839, "Full-finetune/Step Time": 6.76030583307147} {"Full-finetune/Learning Rate": 2.476606700878681e-07, "Full-finetune/Loss": 0.7391935586929321, "Full-finetune/Loss (Raw)": 0.7719423174858093, "Full-finetune/Step": 4840, "Full-finetune/Step Time": 6.749790854752064} {"Full-finetune/Learning Rate": 2.4628472281925443e-07, "Full-finetune/Loss": 0.7382323741912842, "Full-finetune/Loss (Raw)": 0.7185434699058533, "Full-finetune/Step": 4841, "Full-finetune/Step Time": 6.764670148491859} {"Full-finetune/Learning Rate": 2.449125607771141e-07, "Full-finetune/Loss": 0.7374089360237122, "Full-finetune/Loss (Raw)": 0.6348385810852051, "Full-finetune/Step": 4842, "Full-finetune/Step Time": 6.773654483258724} {"Full-finetune/Learning Rate": 2.435441844939557e-07, "Full-finetune/Loss": 0.738645076751709, "Full-finetune/Loss (Raw)": 0.9320586323738098, "Full-finetune/Step": 4843, "Full-finetune/Step Time": 6.770759131759405} {"Full-finetune/Learning Rate": 2.4217959450082207e-07, "Full-finetune/Loss": 0.738123893737793, "Full-finetune/Loss (Raw)": 0.6709627509117126, "Full-finetune/Step": 4844, "Full-finetune/Step Time": 6.789806639775634} {"Full-finetune/Learning Rate": 2.408187913272841e-07, "Full-finetune/Loss": 0.7386026382446289, "Full-finetune/Loss (Raw)": 0.8730995059013367, "Full-finetune/Step": 4845, "Full-finetune/Step Time": 6.780163025483489} {"Full-finetune/Learning Rate": 2.3946177550143835e-07, "Full-finetune/Loss": 0.7403669357299805, "Full-finetune/Loss (Raw)": 0.7855783700942993, "Full-finetune/Step": 4846, "Full-finetune/Step Time": 6.79039466008544} {"Full-finetune/Learning Rate": 2.3810854754992008e-07, "Full-finetune/Loss": 0.7378813028335571, "Full-finetune/Loss (Raw)": 0.3285094201564789, "Full-finetune/Step": 4847, "Full-finetune/Step Time": 6.808983584865928} {"Full-finetune/Learning Rate": 2.3675910799789036e-07, "Full-finetune/Loss": 0.739364504814148, "Full-finetune/Loss (Raw)": 0.6746206879615784, "Full-finetune/Step": 4848, "Full-finetune/Step Time": 6.773831412196159} {"Full-finetune/Learning Rate": 2.354134573690392e-07, "Full-finetune/Loss": 0.7399821281433105, "Full-finetune/Loss (Raw)": 0.6680678725242615, "Full-finetune/Step": 4849, "Full-finetune/Step Time": 6.78836883418262} {"Full-finetune/Learning Rate": 2.3407159618558772e-07, "Full-finetune/Loss": 0.7413645386695862, "Full-finetune/Loss (Raw)": 0.8692695498466492, "Full-finetune/Step": 4850, "Full-finetune/Step Time": 6.795900275930762} {"Full-finetune/Learning Rate": 2.3273352496828605e-07, "Full-finetune/Loss": 0.7419738173484802, "Full-finetune/Loss (Raw)": 0.8507878184318542, "Full-finetune/Step": 4851, "Full-finetune/Step Time": 6.801999816671014} {"Full-finetune/Learning Rate": 2.3139924423641325e-07, "Full-finetune/Loss": 0.7422571182250977, "Full-finetune/Loss (Raw)": 0.6551371812820435, "Full-finetune/Step": 4852, "Full-finetune/Step Time": 6.793518699705601} {"Full-finetune/Learning Rate": 2.3006875450777733e-07, "Full-finetune/Loss": 0.7431248426437378, "Full-finetune/Loss (Raw)": 0.8140946626663208, "Full-finetune/Step": 4853, "Full-finetune/Step Time": 6.799446098506451} {"Full-finetune/Learning Rate": 2.287420562987175e-07, "Full-finetune/Loss": 0.7420191764831543, "Full-finetune/Loss (Raw)": 0.5278189182281494, "Full-finetune/Step": 4854, "Full-finetune/Step Time": 6.801705727353692} {"Full-finetune/Learning Rate": 2.2741915012409744e-07, "Full-finetune/Loss": 0.7424750328063965, "Full-finetune/Loss (Raw)": 0.7737671136856079, "Full-finetune/Step": 4855, "Full-finetune/Step Time": 6.7913123574107885} {"Full-finetune/Learning Rate": 2.2610003649731093e-07, "Full-finetune/Loss": 0.7433003783226013, "Full-finetune/Loss (Raw)": 0.8049068450927734, "Full-finetune/Step": 4856, "Full-finetune/Step Time": 6.785829829052091} {"Full-finetune/Learning Rate": 2.2478471593028073e-07, "Full-finetune/Loss": 0.7420918941497803, "Full-finetune/Loss (Raw)": 0.8741377592086792, "Full-finetune/Step": 4857, "Full-finetune/Step Time": 6.794452017173171} {"Full-finetune/Learning Rate": 2.234731889334585e-07, "Full-finetune/Loss": 0.7435499429702759, "Full-finetune/Loss (Raw)": 0.9572317600250244, "Full-finetune/Step": 4858, "Full-finetune/Step Time": 6.803329238668084} {"Full-finetune/Learning Rate": 2.221654560158204e-07, "Full-finetune/Loss": 0.7448222637176514, "Full-finetune/Loss (Raw)": 0.7838164567947388, "Full-finetune/Step": 4859, "Full-finetune/Step Time": 6.7892052717506886} {"Full-finetune/Learning Rate": 2.2086151768487273e-07, "Full-finetune/Loss": 0.7441525459289551, "Full-finetune/Loss (Raw)": 0.7047562599182129, "Full-finetune/Step": 4860, "Full-finetune/Step Time": 6.787662981078029} {"Full-finetune/Learning Rate": 2.195613744466485e-07, "Full-finetune/Loss": 0.7432107329368591, "Full-finetune/Loss (Raw)": 0.7207807898521423, "Full-finetune/Step": 4861, "Full-finetune/Step Time": 6.8017186392098665} {"Full-finetune/Learning Rate": 2.182650268057085e-07, "Full-finetune/Loss": 0.7450318336486816, "Full-finetune/Loss (Raw)": 0.9179415106773376, "Full-finetune/Step": 4862, "Full-finetune/Step Time": 6.785993894562125} {"Full-finetune/Learning Rate": 2.1697247526513698e-07, "Full-finetune/Loss": 0.745673418045044, "Full-finetune/Loss (Raw)": 0.9298797249794006, "Full-finetune/Step": 4863, "Full-finetune/Step Time": 6.79087134078145} {"Full-finetune/Learning Rate": 2.1568372032655384e-07, "Full-finetune/Loss": 0.7468246221542358, "Full-finetune/Loss (Raw)": 0.73955237865448, "Full-finetune/Step": 4864, "Full-finetune/Step Time": 6.794717093929648} {"Full-finetune/Learning Rate": 2.143987624900945e-07, "Full-finetune/Loss": 0.7480552196502686, "Full-finetune/Loss (Raw)": 0.8339460492134094, "Full-finetune/Step": 4865, "Full-finetune/Step Time": 6.787998612970114} {"Full-finetune/Learning Rate": 2.1311760225442678e-07, "Full-finetune/Loss": 0.7472256422042847, "Full-finetune/Loss (Raw)": 0.7219225168228149, "Full-finetune/Step": 4866, "Full-finetune/Step Time": 6.7793930768966675} {"Full-finetune/Learning Rate": 2.118402401167452e-07, "Full-finetune/Loss": 0.7445331811904907, "Full-finetune/Loss (Raw)": 0.35383373498916626, "Full-finetune/Step": 4867, "Full-finetune/Step Time": 6.796342169865966} {"Full-finetune/Learning Rate": 2.1056667657276876e-07, "Full-finetune/Loss": 0.7446771860122681, "Full-finetune/Loss (Raw)": 0.690102756023407, "Full-finetune/Step": 4868, "Full-finetune/Step Time": 6.796786013990641} {"Full-finetune/Learning Rate": 2.0929691211674319e-07, "Full-finetune/Loss": 0.7454249262809753, "Full-finetune/Loss (Raw)": 0.6335508227348328, "Full-finetune/Step": 4869, "Full-finetune/Step Time": 6.799673901870847} {"Full-finetune/Learning Rate": 2.0803094724143879e-07, "Full-finetune/Loss": 0.7457186579704285, "Full-finetune/Loss (Raw)": 0.7060070037841797, "Full-finetune/Step": 4870, "Full-finetune/Step Time": 6.7842822428792715} {"Full-finetune/Learning Rate": 2.0676878243815145e-07, "Full-finetune/Loss": 0.746842622756958, "Full-finetune/Loss (Raw)": 0.7904489040374756, "Full-finetune/Step": 4871, "Full-finetune/Step Time": 6.776800114661455} {"Full-finetune/Learning Rate": 2.055104181967027e-07, "Full-finetune/Loss": 0.7455077171325684, "Full-finetune/Loss (Raw)": 0.768101155757904, "Full-finetune/Step": 4872, "Full-finetune/Step Time": 6.765693938359618} {"Full-finetune/Learning Rate": 2.0425585500543854e-07, "Full-finetune/Loss": 0.7454079985618591, "Full-finetune/Loss (Raw)": 0.6840308904647827, "Full-finetune/Step": 4873, "Full-finetune/Step Time": 6.7617966048419476} {"Full-finetune/Learning Rate": 2.0300509335123286e-07, "Full-finetune/Loss": 0.744318425655365, "Full-finetune/Loss (Raw)": 0.551124632358551, "Full-finetune/Step": 4874, "Full-finetune/Step Time": 6.745197381824255} {"Full-finetune/Learning Rate": 2.0175813371948072e-07, "Full-finetune/Loss": 0.7430343627929688, "Full-finetune/Loss (Raw)": 0.7312490344047546, "Full-finetune/Step": 4875, "Full-finetune/Step Time": 6.754007114097476} {"Full-finetune/Learning Rate": 2.0051497659409947e-07, "Full-finetune/Loss": 0.743503212928772, "Full-finetune/Loss (Raw)": 0.8899668455123901, "Full-finetune/Step": 4876, "Full-finetune/Step Time": 6.745838023722172} {"Full-finetune/Learning Rate": 1.9927562245753874e-07, "Full-finetune/Loss": 0.744198203086853, "Full-finetune/Loss (Raw)": 0.8730403780937195, "Full-finetune/Step": 4877, "Full-finetune/Step Time": 6.747380385175347} {"Full-finetune/Learning Rate": 1.9804007179076602e-07, "Full-finetune/Loss": 0.7440246343612671, "Full-finetune/Loss (Raw)": 0.7120152115821838, "Full-finetune/Step": 4878, "Full-finetune/Step Time": 6.74765301682055} {"Full-finetune/Learning Rate": 1.968083250732744e-07, "Full-finetune/Loss": 0.7447855472564697, "Full-finetune/Loss (Raw)": 0.8368185758590698, "Full-finetune/Step": 4879, "Full-finetune/Step Time": 6.728989977389574} {"Full-finetune/Learning Rate": 1.9558038278308044e-07, "Full-finetune/Loss": 0.7451236248016357, "Full-finetune/Loss (Raw)": 0.873437762260437, "Full-finetune/Step": 4880, "Full-finetune/Step Time": 6.735345609486103} {"Full-finetune/Learning Rate": 1.9435624539672515e-07, "Full-finetune/Loss": 0.7472668886184692, "Full-finetune/Loss (Raw)": 0.9692544937133789, "Full-finetune/Step": 4881, "Full-finetune/Step Time": 6.729358732700348} {"Full-finetune/Learning Rate": 1.9313591338927185e-07, "Full-finetune/Loss": 0.7462401390075684, "Full-finetune/Loss (Raw)": 0.6532663702964783, "Full-finetune/Step": 4882, "Full-finetune/Step Time": 6.730949385091662} {"Full-finetune/Learning Rate": 1.9191938723430614e-07, "Full-finetune/Loss": 0.7455763816833496, "Full-finetune/Loss (Raw)": 0.678126871585846, "Full-finetune/Step": 4883, "Full-finetune/Step Time": 6.73166180588305} {"Full-finetune/Learning Rate": 1.9070666740394151e-07, "Full-finetune/Loss": 0.7463639974594116, "Full-finetune/Loss (Raw)": 0.8606874346733093, "Full-finetune/Step": 4884, "Full-finetune/Step Time": 6.733334871008992} {"Full-finetune/Learning Rate": 1.8949775436880924e-07, "Full-finetune/Loss": 0.7461204528808594, "Full-finetune/Loss (Raw)": 0.738645076751709, "Full-finetune/Step": 4885, "Full-finetune/Step Time": 6.73712058365345} {"Full-finetune/Learning Rate": 1.8829264859806296e-07, "Full-finetune/Loss": 0.7469639182090759, "Full-finetune/Loss (Raw)": 0.8598712086677551, "Full-finetune/Step": 4886, "Full-finetune/Step Time": 6.741817019879818} {"Full-finetune/Learning Rate": 1.8709135055938298e-07, "Full-finetune/Loss": 0.7479052543640137, "Full-finetune/Loss (Raw)": 0.8213410973548889, "Full-finetune/Step": 4887, "Full-finetune/Step Time": 6.748064670711756} {"Full-finetune/Learning Rate": 1.8589386071896975e-07, "Full-finetune/Loss": 0.7481914758682251, "Full-finetune/Loss (Raw)": 0.82309490442276, "Full-finetune/Step": 4888, "Full-finetune/Step Time": 6.756947921589017} {"Full-finetune/Learning Rate": 1.8470017954154262e-07, "Full-finetune/Loss": 0.7496241331100464, "Full-finetune/Loss (Raw)": 0.7840412259101868, "Full-finetune/Step": 4889, "Full-finetune/Step Time": 6.746178828179836} {"Full-finetune/Learning Rate": 1.8351030749034992e-07, "Full-finetune/Loss": 0.7494648694992065, "Full-finetune/Loss (Raw)": 0.6953256130218506, "Full-finetune/Step": 4890, "Full-finetune/Step Time": 6.763550130650401} {"Full-finetune/Learning Rate": 1.8232424502715672e-07, "Full-finetune/Loss": 0.7490469217300415, "Full-finetune/Loss (Raw)": 0.719074547290802, "Full-finetune/Step": 4891, "Full-finetune/Step Time": 6.750229695811868} {"Full-finetune/Learning Rate": 1.811419926122493e-07, "Full-finetune/Loss": 0.7498111724853516, "Full-finetune/Loss (Raw)": 0.7713973522186279, "Full-finetune/Step": 4892, "Full-finetune/Step Time": 6.74589834548533} {"Full-finetune/Learning Rate": 1.7996355070443727e-07, "Full-finetune/Loss": 0.7516756057739258, "Full-finetune/Loss (Raw)": 0.8841567635536194, "Full-finetune/Step": 4893, "Full-finetune/Step Time": 6.756375547498465} {"Full-finetune/Learning Rate": 1.787889197610515e-07, "Full-finetune/Loss": 0.7501466274261475, "Full-finetune/Loss (Raw)": 0.721739649772644, "Full-finetune/Step": 4894, "Full-finetune/Step Time": 6.761442523449659} {"Full-finetune/Learning Rate": 1.776181002379429e-07, "Full-finetune/Loss": 0.7489538192749023, "Full-finetune/Loss (Raw)": 0.6559174060821533, "Full-finetune/Step": 4895, "Full-finetune/Step Time": 6.753798935562372} {"Full-finetune/Learning Rate": 1.7645109258948245e-07, "Full-finetune/Loss": 0.7499322891235352, "Full-finetune/Loss (Raw)": 0.853755533695221, "Full-finetune/Step": 4896, "Full-finetune/Step Time": 6.746148651465774} {"Full-finetune/Learning Rate": 1.7528789726856566e-07, "Full-finetune/Loss": 0.7489378452301025, "Full-finetune/Loss (Raw)": 0.6259181499481201, "Full-finetune/Step": 4897, "Full-finetune/Step Time": 6.752423174679279} {"Full-finetune/Learning Rate": 1.7412851472660585e-07, "Full-finetune/Loss": 0.7505580186843872, "Full-finetune/Loss (Raw)": 0.669655442237854, "Full-finetune/Step": 4898, "Full-finetune/Step Time": 6.753827186301351} {"Full-finetune/Learning Rate": 1.7297294541353317e-07, "Full-finetune/Loss": 0.7502878904342651, "Full-finetune/Loss (Raw)": 0.8589324355125427, "Full-finetune/Step": 4899, "Full-finetune/Step Time": 6.751393955200911} {"Full-finetune/Learning Rate": 1.7182118977780548e-07, "Full-finetune/Loss": 0.7511196732521057, "Full-finetune/Loss (Raw)": 0.789350688457489, "Full-finetune/Step": 4900, "Full-finetune/Step Time": 6.738604808226228} {"Full-finetune/Learning Rate": 1.706732482663942e-07, "Full-finetune/Loss": 0.7525803446769714, "Full-finetune/Loss (Raw)": 0.7367267608642578, "Full-finetune/Step": 4901, "Full-finetune/Step Time": 6.736621975898743} {"Full-finetune/Learning Rate": 1.6952912132479404e-07, "Full-finetune/Loss": 0.7522094249725342, "Full-finetune/Loss (Raw)": 0.7372727990150452, "Full-finetune/Step": 4902, "Full-finetune/Step Time": 6.747478194534779} {"Full-finetune/Learning Rate": 1.6838880939701875e-07, "Full-finetune/Loss": 0.7511321306228638, "Full-finetune/Loss (Raw)": 0.7247225046157837, "Full-finetune/Step": 4903, "Full-finetune/Step Time": 6.746993966400623} {"Full-finetune/Learning Rate": 1.6725231292559985e-07, "Full-finetune/Loss": 0.7521001100540161, "Full-finetune/Loss (Raw)": 0.7755172252655029, "Full-finetune/Step": 4904, "Full-finetune/Step Time": 6.73169475607574} {"Full-finetune/Learning Rate": 1.661196323515901e-07, "Full-finetune/Loss": 0.7523730993270874, "Full-finetune/Loss (Raw)": 0.7889620065689087, "Full-finetune/Step": 4905, "Full-finetune/Step Time": 6.734986055642366} {"Full-finetune/Learning Rate": 1.6499076811456015e-07, "Full-finetune/Loss": 0.7557414770126343, "Full-finetune/Loss (Raw)": 0.7207733392715454, "Full-finetune/Step": 4906, "Full-finetune/Step Time": 6.7297987304627895} {"Full-finetune/Learning Rate": 1.6386572065260065e-07, "Full-finetune/Loss": 0.7532185912132263, "Full-finetune/Loss (Raw)": 0.597082257270813, "Full-finetune/Step": 4907, "Full-finetune/Step Time": 6.735195817425847} {"Full-finetune/Learning Rate": 1.6274449040232233e-07, "Full-finetune/Loss": 0.7537480592727661, "Full-finetune/Loss (Raw)": 0.7599509358406067, "Full-finetune/Step": 4908, "Full-finetune/Step Time": 6.746721815317869} {"Full-finetune/Learning Rate": 1.6162707779884933e-07, "Full-finetune/Loss": 0.754716157913208, "Full-finetune/Loss (Raw)": 0.7580810785293579, "Full-finetune/Step": 4909, "Full-finetune/Step Time": 6.732535170391202} {"Full-finetune/Learning Rate": 1.6051348327583038e-07, "Full-finetune/Loss": 0.7558780908584595, "Full-finetune/Loss (Raw)": 0.8048780560493469, "Full-finetune/Step": 4910, "Full-finetune/Step Time": 6.7220749370753765} {"Full-finetune/Learning Rate": 1.5940370726542864e-07, "Full-finetune/Loss": 0.7589588165283203, "Full-finetune/Loss (Raw)": 0.7639985680580139, "Full-finetune/Step": 4911, "Full-finetune/Step Time": 6.70427961833775} {"Full-finetune/Learning Rate": 1.5829775019832628e-07, "Full-finetune/Loss": 0.7587438821792603, "Full-finetune/Loss (Raw)": 0.7574557065963745, "Full-finetune/Step": 4912, "Full-finetune/Step Time": 6.709536796435714} {"Full-finetune/Learning Rate": 1.571956125037244e-07, "Full-finetune/Loss": 0.7575515508651733, "Full-finetune/Loss (Raw)": 0.7985202074050903, "Full-finetune/Step": 4913, "Full-finetune/Step Time": 6.708164131268859} {"Full-finetune/Learning Rate": 1.5609729460934087e-07, "Full-finetune/Loss": 0.7543742656707764, "Full-finetune/Loss (Raw)": 0.4093051552772522, "Full-finetune/Step": 4914, "Full-finetune/Step Time": 6.723226698115468} {"Full-finetune/Learning Rate": 1.550027969414114e-07, "Full-finetune/Loss": 0.7536023855209351, "Full-finetune/Loss (Raw)": 0.6535384058952332, "Full-finetune/Step": 4915, "Full-finetune/Step Time": 6.727072931826115} {"Full-finetune/Learning Rate": 1.5391211992468736e-07, "Full-finetune/Loss": 0.7544926404953003, "Full-finetune/Loss (Raw)": 0.7515822052955627, "Full-finetune/Step": 4916, "Full-finetune/Step Time": 6.701938001438975} {"Full-finetune/Learning Rate": 1.5282526398244236e-07, "Full-finetune/Loss": 0.7531795501708984, "Full-finetune/Loss (Raw)": 0.41259804368019104, "Full-finetune/Step": 4917, "Full-finetune/Step Time": 6.7154924385249615} {"Full-finetune/Learning Rate": 1.5174222953646344e-07, "Full-finetune/Loss": 0.7531270980834961, "Full-finetune/Loss (Raw)": 0.7947190403938293, "Full-finetune/Step": 4918, "Full-finetune/Step Time": 6.711171397939324} {"Full-finetune/Learning Rate": 1.5066301700705334e-07, "Full-finetune/Loss": 0.7567279934883118, "Full-finetune/Loss (Raw)": 0.7688629031181335, "Full-finetune/Step": 4919, "Full-finetune/Step Time": 6.6939277816563845} {"Full-finetune/Learning Rate": 1.4958762681303362e-07, "Full-finetune/Loss": 0.7559845447540283, "Full-finetune/Loss (Raw)": 0.8768627643585205, "Full-finetune/Step": 4920, "Full-finetune/Step Time": 6.687641533091664} {"Full-finetune/Learning Rate": 1.4851605937174384e-07, "Full-finetune/Loss": 0.7552965879440308, "Full-finetune/Loss (Raw)": 0.7436507344245911, "Full-finetune/Step": 4921, "Full-finetune/Step Time": 6.699721086770296} {"Full-finetune/Learning Rate": 1.474483150990369e-07, "Full-finetune/Loss": 0.7552253007888794, "Full-finetune/Loss (Raw)": 0.6772528886795044, "Full-finetune/Step": 4922, "Full-finetune/Step Time": 6.690814632922411} {"Full-finetune/Learning Rate": 1.4638439440928354e-07, "Full-finetune/Loss": 0.7557121515274048, "Full-finetune/Loss (Raw)": 0.7314959168434143, "Full-finetune/Step": 4923, "Full-finetune/Step Time": 6.665205363184214} {"Full-finetune/Learning Rate": 1.4532429771537237e-07, "Full-finetune/Loss": 0.7547200918197632, "Full-finetune/Loss (Raw)": 0.7185966968536377, "Full-finetune/Step": 4924, "Full-finetune/Step Time": 6.6658023763448} {"Full-finetune/Learning Rate": 1.4426802542870433e-07, "Full-finetune/Loss": 0.7541550397872925, "Full-finetune/Loss (Raw)": 0.7257858514785767, "Full-finetune/Step": 4925, "Full-finetune/Step Time": 6.6656612902879715} {"Full-finetune/Learning Rate": 1.4321557795919705e-07, "Full-finetune/Loss": 0.755240261554718, "Full-finetune/Loss (Raw)": 0.7311177849769592, "Full-finetune/Step": 4926, "Full-finetune/Step Time": 6.648872639983892} {"Full-finetune/Learning Rate": 1.4216695571528828e-07, "Full-finetune/Loss": 0.7545843124389648, "Full-finetune/Loss (Raw)": 0.7525107264518738, "Full-finetune/Step": 4927, "Full-finetune/Step Time": 6.646417867392302} {"Full-finetune/Learning Rate": 1.411221591039269e-07, "Full-finetune/Loss": 0.7516173124313354, "Full-finetune/Loss (Raw)": 0.5596045255661011, "Full-finetune/Step": 4928, "Full-finetune/Step Time": 6.6682029236108065} {"Full-finetune/Learning Rate": 1.4008118853057529e-07, "Full-finetune/Loss": 0.7505910396575928, "Full-finetune/Loss (Raw)": 0.7217384576797485, "Full-finetune/Step": 4929, "Full-finetune/Step Time": 6.672687727957964} {"Full-finetune/Learning Rate": 1.3904404439921582e-07, "Full-finetune/Loss": 0.7505077123641968, "Full-finetune/Loss (Raw)": 0.775253415107727, "Full-finetune/Step": 4930, "Full-finetune/Step Time": 6.67125990986824} {"Full-finetune/Learning Rate": 1.3801072711234541e-07, "Full-finetune/Loss": 0.7503626346588135, "Full-finetune/Loss (Raw)": 0.8346840739250183, "Full-finetune/Step": 4931, "Full-finetune/Step Time": 6.668422261252999} {"Full-finetune/Learning Rate": 1.369812370709711e-07, "Full-finetune/Loss": 0.7504507899284363, "Full-finetune/Loss (Raw)": 0.8248679041862488, "Full-finetune/Step": 4932, "Full-finetune/Step Time": 6.665612831711769} {"Full-finetune/Learning Rate": 1.3595557467461994e-07, "Full-finetune/Loss": 0.7492727041244507, "Full-finetune/Loss (Raw)": 0.702926754951477, "Full-finetune/Step": 4933, "Full-finetune/Step Time": 6.6754460241645575} {"Full-finetune/Learning Rate": 1.3493374032133134e-07, "Full-finetune/Loss": 0.7495934367179871, "Full-finetune/Loss (Raw)": 0.8295468091964722, "Full-finetune/Step": 4934, "Full-finetune/Step Time": 6.670724177733064} {"Full-finetune/Learning Rate": 1.3391573440765915e-07, "Full-finetune/Loss": 0.7502913475036621, "Full-finetune/Loss (Raw)": 0.8467627763748169, "Full-finetune/Step": 4935, "Full-finetune/Step Time": 6.6686507519334555} {"Full-finetune/Learning Rate": 1.329015573286696e-07, "Full-finetune/Loss": 0.7485227584838867, "Full-finetune/Loss (Raw)": 0.6797700524330139, "Full-finetune/Step": 4936, "Full-finetune/Step Time": 6.678907558321953} {"Full-finetune/Learning Rate": 1.3189120947794897e-07, "Full-finetune/Loss": 0.7512645125389099, "Full-finetune/Loss (Raw)": 0.9114798903465271, "Full-finetune/Step": 4937, "Full-finetune/Step Time": 6.681879505515099} {"Full-finetune/Learning Rate": 1.308846912475914e-07, "Full-finetune/Loss": 0.7516711950302124, "Full-finetune/Loss (Raw)": 0.8038700222969055, "Full-finetune/Step": 4938, "Full-finetune/Step Time": 6.692245576530695} {"Full-finetune/Learning Rate": 1.2988200302820552e-07, "Full-finetune/Loss": 0.7482086420059204, "Full-finetune/Loss (Raw)": 0.34752127528190613, "Full-finetune/Step": 4939, "Full-finetune/Step Time": 6.698188213631511} {"Full-finetune/Learning Rate": 1.2888314520891676e-07, "Full-finetune/Loss": 0.7468816637992859, "Full-finetune/Loss (Raw)": 0.6160427927970886, "Full-finetune/Step": 4940, "Full-finetune/Step Time": 6.698876949027181} {"Full-finetune/Learning Rate": 1.2788811817736168e-07, "Full-finetune/Loss": 0.7483646869659424, "Full-finetune/Loss (Raw)": 0.6669570803642273, "Full-finetune/Step": 4941, "Full-finetune/Step Time": 6.681049101054668} {"Full-finetune/Learning Rate": 1.2689692231969364e-07, "Full-finetune/Loss": 0.7473656535148621, "Full-finetune/Loss (Raw)": 0.689117431640625, "Full-finetune/Step": 4942, "Full-finetune/Step Time": 6.685049343854189} {"Full-finetune/Learning Rate": 1.2590955802057047e-07, "Full-finetune/Loss": 0.7458377480506897, "Full-finetune/Loss (Raw)": 0.682977557182312, "Full-finetune/Step": 4943, "Full-finetune/Step Time": 6.695501102134585} {"Full-finetune/Learning Rate": 1.2492602566317457e-07, "Full-finetune/Loss": 0.7468123435974121, "Full-finetune/Loss (Raw)": 0.7946533560752869, "Full-finetune/Step": 4944, "Full-finetune/Step Time": 6.681564217433333} {"Full-finetune/Learning Rate": 1.2394632562919283e-07, "Full-finetune/Loss": 0.7471325993537903, "Full-finetune/Loss (Raw)": 0.8699502944946289, "Full-finetune/Step": 4945, "Full-finetune/Step Time": 6.681261233985424} {"Full-finetune/Learning Rate": 1.2297045829882892e-07, "Full-finetune/Loss": 0.7462160587310791, "Full-finetune/Loss (Raw)": 0.715825617313385, "Full-finetune/Step": 4946, "Full-finetune/Step Time": 6.680590268224478} {"Full-finetune/Learning Rate": 1.2199842405079766e-07, "Full-finetune/Loss": 0.7454249858856201, "Full-finetune/Loss (Raw)": 0.7760128378868103, "Full-finetune/Step": 4947, "Full-finetune/Step Time": 6.692864732816815} {"Full-finetune/Learning Rate": 1.2103022326232505e-07, "Full-finetune/Loss": 0.7444276809692383, "Full-finetune/Loss (Raw)": 0.735494077205658, "Full-finetune/Step": 4948, "Full-finetune/Step Time": 6.689383262768388} {"Full-finetune/Learning Rate": 1.2006585630915391e-07, "Full-finetune/Loss": 0.7431122064590454, "Full-finetune/Loss (Raw)": 0.7100666165351868, "Full-finetune/Step": 4949, "Full-finetune/Step Time": 6.67674514837563} {"Full-finetune/Learning Rate": 1.1910532356553261e-07, "Full-finetune/Loss": 0.7416811585426331, "Full-finetune/Loss (Raw)": 0.6461332440376282, "Full-finetune/Step": 4950, "Full-finetune/Step Time": 6.682665748521686} {"Full-finetune/Learning Rate": 1.1814862540422744e-07, "Full-finetune/Loss": 0.7415906190872192, "Full-finetune/Loss (Raw)": 0.8375027179718018, "Full-finetune/Step": 4951, "Full-finetune/Step Time": 6.68213564530015} {"Full-finetune/Learning Rate": 1.1719576219651585e-07, "Full-finetune/Loss": 0.7455390095710754, "Full-finetune/Loss (Raw)": 0.7932929992675781, "Full-finetune/Step": 4952, "Full-finetune/Step Time": 6.674465315416455} {"Full-finetune/Learning Rate": 1.1624673431218092e-07, "Full-finetune/Loss": 0.7450892329216003, "Full-finetune/Loss (Raw)": 0.685904324054718, "Full-finetune/Step": 4953, "Full-finetune/Step Time": 6.664242133498192} {"Full-finetune/Learning Rate": 1.1530154211952583e-07, "Full-finetune/Loss": 0.7451796531677246, "Full-finetune/Loss (Raw)": 0.7815412282943726, "Full-finetune/Step": 4954, "Full-finetune/Step Time": 6.698768036440015} {"Full-finetune/Learning Rate": 1.1436018598535937e-07, "Full-finetune/Loss": 0.744638204574585, "Full-finetune/Loss (Raw)": 0.7892805337905884, "Full-finetune/Step": 4955, "Full-finetune/Step Time": 6.701645163819194} {"Full-finetune/Learning Rate": 1.1342266627500375e-07, "Full-finetune/Loss": 0.7444760203361511, "Full-finetune/Loss (Raw)": 0.8406417369842529, "Full-finetune/Step": 4956, "Full-finetune/Step Time": 6.68994957767427} {"Full-finetune/Learning Rate": 1.1248898335229241e-07, "Full-finetune/Loss": 0.7449309825897217, "Full-finetune/Loss (Raw)": 0.8394263982772827, "Full-finetune/Step": 4957, "Full-finetune/Step Time": 6.692209713160992} {"Full-finetune/Learning Rate": 1.1155913757956881e-07, "Full-finetune/Loss": 0.7444633841514587, "Full-finetune/Loss (Raw)": 0.6787846684455872, "Full-finetune/Step": 4958, "Full-finetune/Step Time": 6.708840811625123} {"Full-finetune/Learning Rate": 1.1063312931768877e-07, "Full-finetune/Loss": 0.7451799511909485, "Full-finetune/Loss (Raw)": 0.44368693232536316, "Full-finetune/Step": 4959, "Full-finetune/Step Time": 6.7089501321315765} {"Full-finetune/Learning Rate": 1.0971095892601702e-07, "Full-finetune/Loss": 0.7457184195518494, "Full-finetune/Loss (Raw)": 0.7432442903518677, "Full-finetune/Step": 4960, "Full-finetune/Step Time": 6.683107651770115} {"Full-finetune/Learning Rate": 1.0879262676243174e-07, "Full-finetune/Loss": 0.7454274296760559, "Full-finetune/Loss (Raw)": 0.7459229826927185, "Full-finetune/Step": 4961, "Full-finetune/Step Time": 6.681441752240062} {"Full-finetune/Learning Rate": 1.0787813318331786e-07, "Full-finetune/Loss": 0.744510293006897, "Full-finetune/Loss (Raw)": 0.8059792518615723, "Full-finetune/Step": 4962, "Full-finetune/Step Time": 6.680573757737875} {"Full-finetune/Learning Rate": 1.0696747854357259e-07, "Full-finetune/Loss": 0.743780255317688, "Full-finetune/Loss (Raw)": 0.7520334124565125, "Full-finetune/Step": 4963, "Full-finetune/Step Time": 6.686291817575693} {"Full-finetune/Learning Rate": 1.0606066319660435e-07, "Full-finetune/Loss": 0.7436708807945251, "Full-finetune/Loss (Raw)": 0.8645049929618835, "Full-finetune/Step": 4964, "Full-finetune/Step Time": 6.68932812102139} {"Full-finetune/Learning Rate": 1.0515768749432942e-07, "Full-finetune/Loss": 0.7447236180305481, "Full-finetune/Loss (Raw)": 0.9026420712471008, "Full-finetune/Step": 4965, "Full-finetune/Step Time": 6.691136427223682} {"Full-finetune/Learning Rate": 1.0425855178717636e-07, "Full-finetune/Loss": 0.7436442375183105, "Full-finetune/Loss (Raw)": 0.7236236929893494, "Full-finetune/Step": 4966, "Full-finetune/Step Time": 6.694567637518048} {"Full-finetune/Learning Rate": 1.0336325642408052e-07, "Full-finetune/Loss": 0.7450302243232727, "Full-finetune/Loss (Raw)": 0.8575260639190674, "Full-finetune/Step": 4967, "Full-finetune/Step Time": 6.6926839631050825} {"Full-finetune/Learning Rate": 1.024718017524895e-07, "Full-finetune/Loss": 0.7449511885643005, "Full-finetune/Loss (Raw)": 0.7618293166160583, "Full-finetune/Step": 4968, "Full-finetune/Step Time": 6.689155913889408} {"Full-finetune/Learning Rate": 1.015841881183588e-07, "Full-finetune/Loss": 0.7449575066566467, "Full-finetune/Loss (Raw)": 0.7193527817726135, "Full-finetune/Step": 4969, "Full-finetune/Step Time": 6.68257662653923} {"Full-finetune/Learning Rate": 1.0070041586615398e-07, "Full-finetune/Loss": 0.7461696863174438, "Full-finetune/Loss (Raw)": 0.7899914383888245, "Full-finetune/Step": 4970, "Full-finetune/Step Time": 6.665954817086458} {"Full-finetune/Learning Rate": 9.98204853388518e-08, "Full-finetune/Loss": 0.7459573149681091, "Full-finetune/Loss (Raw)": 0.9048781394958496, "Full-finetune/Step": 4971, "Full-finetune/Step Time": 6.676650473847985} {"Full-finetune/Learning Rate": 9.894439687793466e-08, "Full-finetune/Loss": 0.746658444404602, "Full-finetune/Loss (Raw)": 0.7607060074806213, "Full-finetune/Step": 4972, "Full-finetune/Step Time": 6.671433825045824} {"Full-finetune/Learning Rate": 9.807215082339395e-08, "Full-finetune/Loss": 0.7471014261245728, "Full-finetune/Loss (Raw)": 0.9298006892204285, "Full-finetune/Step": 4973, "Full-finetune/Step Time": 6.672465270385146} {"Full-finetune/Learning Rate": 9.720374751373329e-08, "Full-finetune/Loss": 0.7433403730392456, "Full-finetune/Loss (Raw)": 0.3041660189628601, "Full-finetune/Step": 4974, "Full-finetune/Step Time": 6.66403329372406} {"Full-finetune/Learning Rate": 9.633918728596203e-08, "Full-finetune/Loss": 0.7461856007575989, "Full-finetune/Loss (Raw)": 0.6927040815353394, "Full-finetune/Step": 4975, "Full-finetune/Step Time": 6.651841092854738} {"Full-finetune/Learning Rate": 9.547847047559955e-08, "Full-finetune/Loss": 0.7479416131973267, "Full-finetune/Loss (Raw)": 0.8993914127349854, "Full-finetune/Step": 4976, "Full-finetune/Step Time": 6.654275480657816} {"Full-finetune/Learning Rate": 9.462159741667309e-08, "Full-finetune/Loss": 0.7480590343475342, "Full-finetune/Loss (Raw)": 0.6830909252166748, "Full-finetune/Step": 4977, "Full-finetune/Step Time": 6.654395071789622} {"Full-finetune/Learning Rate": 9.376856844171889e-08, "Full-finetune/Loss": 0.7469833493232727, "Full-finetune/Loss (Raw)": 0.7315855026245117, "Full-finetune/Step": 4978, "Full-finetune/Step Time": 6.64592725597322} {"Full-finetune/Learning Rate": 9.291938388178102e-08, "Full-finetune/Loss": 0.7454057335853577, "Full-finetune/Loss (Raw)": 0.6488497257232666, "Full-finetune/Step": 4979, "Full-finetune/Step Time": 6.647915963083506} {"Full-finetune/Learning Rate": 9.207404406640919e-08, "Full-finetune/Loss": 0.745368480682373, "Full-finetune/Loss (Raw)": 0.650376558303833, "Full-finetune/Step": 4980, "Full-finetune/Step Time": 6.642858864739537} {"Full-finetune/Learning Rate": 9.123254932366655e-08, "Full-finetune/Loss": 0.7445510029792786, "Full-finetune/Loss (Raw)": 0.7094491124153137, "Full-finetune/Step": 4981, "Full-finetune/Step Time": 6.635236820206046} {"Full-finetune/Learning Rate": 9.039489998011853e-08, "Full-finetune/Loss": 0.7471895217895508, "Full-finetune/Loss (Raw)": 0.8655520677566528, "Full-finetune/Step": 4982, "Full-finetune/Step Time": 6.630589401349425} {"Full-finetune/Learning Rate": 8.956109636084065e-08, "Full-finetune/Loss": 0.7444087862968445, "Full-finetune/Loss (Raw)": 0.4178283214569092, "Full-finetune/Step": 4983, "Full-finetune/Step Time": 6.64716131798923} {"Full-finetune/Learning Rate": 8.873113878941631e-08, "Full-finetune/Loss": 0.7429561614990234, "Full-finetune/Loss (Raw)": 0.6189758777618408, "Full-finetune/Step": 4984, "Full-finetune/Step Time": 6.659479441121221} {"Full-finetune/Learning Rate": 8.790502758793562e-08, "Full-finetune/Loss": 0.7429883480072021, "Full-finetune/Loss (Raw)": 0.8782587647438049, "Full-finetune/Step": 4985, "Full-finetune/Step Time": 6.645918959751725} {"Full-finetune/Learning Rate": 8.70827630769977e-08, "Full-finetune/Loss": 0.7401461601257324, "Full-finetune/Loss (Raw)": 0.5934299230575562, "Full-finetune/Step": 4986, "Full-finetune/Step Time": 6.639134863391519} {"Full-finetune/Learning Rate": 8.626434557570507e-08, "Full-finetune/Loss": 0.7403408288955688, "Full-finetune/Loss (Raw)": 0.8087337613105774, "Full-finetune/Step": 4987, "Full-finetune/Step Time": 6.64677376486361} {"Full-finetune/Learning Rate": 8.544977540167032e-08, "Full-finetune/Loss": 0.7413728833198547, "Full-finetune/Loss (Raw)": 0.836859941482544, "Full-finetune/Step": 4988, "Full-finetune/Step Time": 6.655173933133483} {"Full-finetune/Learning Rate": 8.463905287101282e-08, "Full-finetune/Loss": 0.7421133518218994, "Full-finetune/Loss (Raw)": 0.815558910369873, "Full-finetune/Step": 4989, "Full-finetune/Step Time": 6.646940179169178} {"Full-finetune/Learning Rate": 8.383217829835755e-08, "Full-finetune/Loss": 0.7409948110580444, "Full-finetune/Loss (Raw)": 0.7747644782066345, "Full-finetune/Step": 4990, "Full-finetune/Step Time": 6.650217989459634} {"Full-finetune/Learning Rate": 8.302915199683737e-08, "Full-finetune/Loss": 0.7396694421768188, "Full-finetune/Loss (Raw)": 0.7602411508560181, "Full-finetune/Step": 4991, "Full-finetune/Step Time": 6.643108332529664} {"Full-finetune/Learning Rate": 8.222997427809076e-08, "Full-finetune/Loss": 0.7410445213317871, "Full-finetune/Loss (Raw)": 0.9155597686767578, "Full-finetune/Step": 4992, "Full-finetune/Step Time": 6.644173555076122} {"Full-finetune/Learning Rate": 8.143464545226298e-08, "Full-finetune/Loss": 0.7384033799171448, "Full-finetune/Loss (Raw)": 0.49588069319725037, "Full-finetune/Step": 4993, "Full-finetune/Step Time": 6.657975511625409} {"Full-finetune/Learning Rate": 8.064316582800491e-08, "Full-finetune/Loss": 0.7389021515846252, "Full-finetune/Loss (Raw)": 0.7857621312141418, "Full-finetune/Step": 4994, "Full-finetune/Step Time": 6.676885716617107} {"Full-finetune/Learning Rate": 7.98555357124764e-08, "Full-finetune/Loss": 0.7405356168746948, "Full-finetune/Loss (Raw)": 0.5629167556762695, "Full-finetune/Step": 4995, "Full-finetune/Step Time": 6.661431601271033} {"Full-finetune/Learning Rate": 7.907175541133849e-08, "Full-finetune/Loss": 0.7421436309814453, "Full-finetune/Loss (Raw)": 0.8959372043609619, "Full-finetune/Step": 4996, "Full-finetune/Step Time": 6.671540901064873} {"Full-finetune/Learning Rate": 7.829182522876344e-08, "Full-finetune/Loss": 0.7437412738800049, "Full-finetune/Loss (Raw)": 0.8380419015884399, "Full-finetune/Step": 4997, "Full-finetune/Step Time": 6.657014895230532} {"Full-finetune/Learning Rate": 7.75157454674258e-08, "Full-finetune/Loss": 0.7461659908294678, "Full-finetune/Loss (Raw)": 1.0163705348968506, "Full-finetune/Step": 4998, "Full-finetune/Step Time": 6.654499737545848} {"Full-finetune/Learning Rate": 7.674351642850797e-08, "Full-finetune/Loss": 0.7468733787536621, "Full-finetune/Loss (Raw)": 0.8809971213340759, "Full-finetune/Step": 4999, "Full-finetune/Step Time": 6.646168349310756} {"Full-finetune/Learning Rate": 7.597513841169469e-08, "Full-finetune/Loss": 0.7436519861221313, "Full-finetune/Loss (Raw)": 0.3557605445384979, "Full-finetune/Step": 5000, "Full-finetune/Step Time": 6.674900645390153} {"Full-finetune/Learning Rate": 7.521061171518184e-08, "Full-finetune/Loss": 0.7445657253265381, "Full-finetune/Loss (Raw)": 0.800992488861084, "Full-finetune/Step": 5001, "Full-finetune/Step Time": 6.669706668704748} {"Full-finetune/Learning Rate": 7.444993663566325e-08, "Full-finetune/Loss": 0.7443352937698364, "Full-finetune/Loss (Raw)": 0.5216233134269714, "Full-finetune/Step": 5002, "Full-finetune/Step Time": 6.67128406278789} {"Full-finetune/Learning Rate": 7.369311346834496e-08, "Full-finetune/Loss": 0.744123101234436, "Full-finetune/Loss (Raw)": 0.7040957808494568, "Full-finetune/Step": 5003, "Full-finetune/Step Time": 6.659169660881162} {"Full-finetune/Learning Rate": 7.294014250693538e-08, "Full-finetune/Loss": 0.7418603897094727, "Full-finetune/Loss (Raw)": 0.6003391146659851, "Full-finetune/Step": 5004, "Full-finetune/Step Time": 6.6735343132168055} {"Full-finetune/Learning Rate": 7.21910240436463e-08, "Full-finetune/Loss": 0.7410693168640137, "Full-finetune/Loss (Raw)": 0.7717879414558411, "Full-finetune/Step": 5005, "Full-finetune/Step Time": 6.675069021061063} {"Full-finetune/Learning Rate": 7.144575836919632e-08, "Full-finetune/Loss": 0.741597056388855, "Full-finetune/Loss (Raw)": 0.7795616388320923, "Full-finetune/Step": 5006, "Full-finetune/Step Time": 6.667463917285204} {"Full-finetune/Learning Rate": 7.07043457728096e-08, "Full-finetune/Loss": 0.7393545508384705, "Full-finetune/Loss (Raw)": 0.549776554107666, "Full-finetune/Step": 5007, "Full-finetune/Step Time": 6.67735617980361} {"Full-finetune/Learning Rate": 6.996678654221267e-08, "Full-finetune/Loss": 0.7379140853881836, "Full-finetune/Loss (Raw)": 0.6890513896942139, "Full-finetune/Step": 5008, "Full-finetune/Step Time": 6.674091476947069} {"Full-finetune/Learning Rate": 6.923308096363879e-08, "Full-finetune/Loss": 0.7375847697257996, "Full-finetune/Loss (Raw)": 0.9271122813224792, "Full-finetune/Step": 5009, "Full-finetune/Step Time": 6.676071964204311} {"Full-finetune/Learning Rate": 6.850322932182352e-08, "Full-finetune/Loss": 0.7385852336883545, "Full-finetune/Loss (Raw)": 0.781321108341217, "Full-finetune/Step": 5010, "Full-finetune/Step Time": 6.6674959771335125} {"Full-finetune/Learning Rate": 6.777723190000918e-08, "Full-finetune/Loss": 0.7390789985656738, "Full-finetune/Loss (Raw)": 0.7413268089294434, "Full-finetune/Step": 5011, "Full-finetune/Step Time": 6.6587068643420935} {"Full-finetune/Learning Rate": 6.705508897994151e-08, "Full-finetune/Loss": 0.7383629083633423, "Full-finetune/Loss (Raw)": 0.769027590751648, "Full-finetune/Step": 5012, "Full-finetune/Step Time": 6.66669687628746} {"Full-finetune/Learning Rate": 6.633680084186856e-08, "Full-finetune/Loss": 0.7375915050506592, "Full-finetune/Loss (Raw)": 0.6399132013320923, "Full-finetune/Step": 5013, "Full-finetune/Step Time": 6.66557302698493} {"Full-finetune/Learning Rate": 6.562236776454512e-08, "Full-finetune/Loss": 0.7364110350608826, "Full-finetune/Loss (Raw)": 0.7087646126747131, "Full-finetune/Step": 5014, "Full-finetune/Step Time": 6.6565455961972475} {"Full-finetune/Learning Rate": 6.491179002522829e-08, "Full-finetune/Loss": 0.7354204654693604, "Full-finetune/Loss (Raw)": 0.694548487663269, "Full-finetune/Step": 5015, "Full-finetune/Step Time": 6.646013142541051} {"Full-finetune/Learning Rate": 6.420506789967973e-08, "Full-finetune/Loss": 0.734623908996582, "Full-finetune/Loss (Raw)": 0.7211304306983948, "Full-finetune/Step": 5016, "Full-finetune/Step Time": 6.661321394145489} {"Full-finetune/Learning Rate": 6.350220166216337e-08, "Full-finetune/Loss": 0.733666718006134, "Full-finetune/Loss (Raw)": 0.6615265607833862, "Full-finetune/Step": 5017, "Full-finetune/Step Time": 6.660597471520305} {"Full-finetune/Learning Rate": 6.280319158544989e-08, "Full-finetune/Loss": 0.7333596348762512, "Full-finetune/Loss (Raw)": 0.6560227274894714, "Full-finetune/Step": 5018, "Full-finetune/Step Time": 6.649605957791209} {"Full-finetune/Learning Rate": 6.21080379408101e-08, "Full-finetune/Loss": 0.7333847284317017, "Full-finetune/Loss (Raw)": 0.7222854495048523, "Full-finetune/Step": 5019, "Full-finetune/Step Time": 6.648471603170037} {"Full-finetune/Learning Rate": 6.14167409980193e-08, "Full-finetune/Loss": 0.7301108837127686, "Full-finetune/Loss (Raw)": 0.3523438572883606, "Full-finetune/Step": 5020, "Full-finetune/Step Time": 6.6647614277899265} {"Full-finetune/Learning Rate": 6.072930102535734e-08, "Full-finetune/Loss": 0.7297400236129761, "Full-finetune/Loss (Raw)": 0.8366831541061401, "Full-finetune/Step": 5021, "Full-finetune/Step Time": 6.664590574800968} {"Full-finetune/Learning Rate": 6.004571828960525e-08, "Full-finetune/Loss": 0.7303193807601929, "Full-finetune/Loss (Raw)": 0.7958986163139343, "Full-finetune/Step": 5022, "Full-finetune/Step Time": 6.647798925638199} {"Full-finetune/Learning Rate": 5.936599305604862e-08, "Full-finetune/Loss": 0.7310808897018433, "Full-finetune/Loss (Raw)": 0.7533895969390869, "Full-finetune/Step": 5023, "Full-finetune/Step Time": 6.654052523896098} {"Full-finetune/Learning Rate": 5.869012558847642e-08, "Full-finetune/Loss": 0.7302950620651245, "Full-finetune/Loss (Raw)": 0.753172755241394, "Full-finetune/Step": 5024, "Full-finetune/Step Time": 6.657682526856661} {"Full-finetune/Learning Rate": 5.801811614917885e-08, "Full-finetune/Loss": 0.7311408519744873, "Full-finetune/Loss (Raw)": 0.7341790795326233, "Full-finetune/Step": 5025, "Full-finetune/Step Time": 6.662307625636458} {"Full-finetune/Learning Rate": 5.73499649989484e-08, "Full-finetune/Loss": 0.7324113845825195, "Full-finetune/Loss (Raw)": 0.8322820663452148, "Full-finetune/Step": 5026, "Full-finetune/Step Time": 6.654111826792359} {"Full-finetune/Learning Rate": 5.6685672397083226e-08, "Full-finetune/Loss": 0.7321887016296387, "Full-finetune/Loss (Raw)": 0.8304277658462524, "Full-finetune/Step": 5027, "Full-finetune/Step Time": 6.657387044280767} {"Full-finetune/Learning Rate": 5.602523860138154e-08, "Full-finetune/Loss": 0.7330321073532104, "Full-finetune/Loss (Raw)": 0.8973021507263184, "Full-finetune/Step": 5028, "Full-finetune/Step Time": 6.659962119534612} {"Full-finetune/Learning Rate": 5.536866386814499e-08, "Full-finetune/Loss": 0.7336945533752441, "Full-finetune/Loss (Raw)": 0.8215248584747314, "Full-finetune/Step": 5029, "Full-finetune/Step Time": 6.661696782335639} {"Full-finetune/Learning Rate": 5.471594845217865e-08, "Full-finetune/Loss": 0.7338053584098816, "Full-finetune/Loss (Raw)": 0.7514562010765076, "Full-finetune/Step": 5030, "Full-finetune/Step Time": 6.654079580679536} {"Full-finetune/Learning Rate": 5.4067092606785445e-08, "Full-finetune/Loss": 0.734052836894989, "Full-finetune/Loss (Raw)": 0.7563958168029785, "Full-finetune/Step": 5031, "Full-finetune/Step Time": 6.656101476401091} {"Full-finetune/Learning Rate": 5.342209658377728e-08, "Full-finetune/Loss": 0.7344244718551636, "Full-finetune/Loss (Raw)": 0.8230910301208496, "Full-finetune/Step": 5032, "Full-finetune/Step Time": 6.653979405760765} {"Full-finetune/Learning Rate": 5.278096063346172e-08, "Full-finetune/Loss": 0.7343314290046692, "Full-finetune/Loss (Raw)": 0.7770512104034424, "Full-finetune/Step": 5033, "Full-finetune/Step Time": 6.654620828106999} {"Full-finetune/Learning Rate": 5.214368500465305e-08, "Full-finetune/Loss": 0.7350383996963501, "Full-finetune/Loss (Raw)": 0.8112690448760986, "Full-finetune/Step": 5034, "Full-finetune/Step Time": 6.640480047091842} {"Full-finetune/Learning Rate": 5.151026994466457e-08, "Full-finetune/Loss": 0.7359548211097717, "Full-finetune/Loss (Raw)": 0.7143807411193848, "Full-finetune/Step": 5035, "Full-finetune/Step Time": 6.638776553794742} {"Full-finetune/Learning Rate": 5.088071569931185e-08, "Full-finetune/Loss": 0.7358299493789673, "Full-finetune/Loss (Raw)": 0.7439725399017334, "Full-finetune/Step": 5036, "Full-finetune/Step Time": 6.619078727439046} {"Full-finetune/Learning Rate": 5.025502251291392e-08, "Full-finetune/Loss": 0.7370594143867493, "Full-finetune/Loss (Raw)": 0.9154456853866577, "Full-finetune/Step": 5037, "Full-finetune/Step Time": 6.614553717896342} {"Full-finetune/Learning Rate": 4.963319062828875e-08, "Full-finetune/Loss": 0.7377980947494507, "Full-finetune/Loss (Raw)": 0.899427592754364, "Full-finetune/Step": 5038, "Full-finetune/Step Time": 6.623523358255625} {"Full-finetune/Learning Rate": 4.901522028675776e-08, "Full-finetune/Loss": 0.7393122315406799, "Full-finetune/Loss (Raw)": 0.9578139185905457, "Full-finetune/Step": 5039, "Full-finetune/Step Time": 6.628304662182927} {"Full-finetune/Learning Rate": 4.8401111728143545e-08, "Full-finetune/Loss": 0.7376682758331299, "Full-finetune/Loss (Raw)": 0.5470229983329773, "Full-finetune/Step": 5040, "Full-finetune/Step Time": 6.631998263299465} {"Full-finetune/Learning Rate": 4.779086519076992e-08, "Full-finetune/Loss": 0.7386239767074585, "Full-finetune/Loss (Raw)": 0.9208546280860901, "Full-finetune/Step": 5041, "Full-finetune/Step Time": 6.619892507791519} {"Full-finetune/Learning Rate": 4.7184480911459665e-08, "Full-finetune/Loss": 0.7422463893890381, "Full-finetune/Loss (Raw)": 0.8729695081710815, "Full-finetune/Step": 5042, "Full-finetune/Step Time": 6.615191405639052} {"Full-finetune/Learning Rate": 4.65819591255412e-08, "Full-finetune/Loss": 0.7434409856796265, "Full-finetune/Loss (Raw)": 0.806453287601471, "Full-finetune/Step": 5043, "Full-finetune/Step Time": 6.615690449252725} {"Full-finetune/Learning Rate": 4.5983300066839706e-08, "Full-finetune/Loss": 0.7437902688980103, "Full-finetune/Loss (Raw)": 0.7962842583656311, "Full-finetune/Step": 5044, "Full-finetune/Step Time": 6.619157820940018} {"Full-finetune/Learning Rate": 4.538850396768379e-08, "Full-finetune/Loss": 0.7429102659225464, "Full-finetune/Loss (Raw)": 0.2999541759490967, "Full-finetune/Step": 5045, "Full-finetune/Step Time": 6.616915995255113} {"Full-finetune/Learning Rate": 4.479757105890104e-08, "Full-finetune/Loss": 0.7393200397491455, "Full-finetune/Loss (Raw)": 0.33517777919769287, "Full-finetune/Step": 5046, "Full-finetune/Step Time": 6.633171694353223} {"Full-finetune/Learning Rate": 4.4210501569823536e-08, "Full-finetune/Loss": 0.739537239074707, "Full-finetune/Loss (Raw)": 0.7966614961624146, "Full-finetune/Step": 5047, "Full-finetune/Step Time": 6.63453072309494} {"Full-finetune/Learning Rate": 4.362729572827906e-08, "Full-finetune/Loss": 0.739710807800293, "Full-finetune/Loss (Raw)": 0.899081289768219, "Full-finetune/Step": 5048, "Full-finetune/Step Time": 6.64609001763165} {"Full-finetune/Learning Rate": 4.304795376059989e-08, "Full-finetune/Loss": 0.7394992113113403, "Full-finetune/Loss (Raw)": 0.7165613174438477, "Full-finetune/Step": 5049, "Full-finetune/Step Time": 6.638431357219815} {"Full-finetune/Learning Rate": 4.2472475891617295e-08, "Full-finetune/Loss": 0.7398891448974609, "Full-finetune/Loss (Raw)": 0.7271704077720642, "Full-finetune/Step": 5050, "Full-finetune/Step Time": 6.6483466774225235} {"Full-finetune/Learning Rate": 4.1900862344661506e-08, "Full-finetune/Loss": 0.7389264702796936, "Full-finetune/Loss (Raw)": 0.6082741022109985, "Full-finetune/Step": 5051, "Full-finetune/Step Time": 6.663054080680013} {"Full-finetune/Learning Rate": 4.133311334156509e-08, "Full-finetune/Loss": 0.7362135052680969, "Full-finetune/Loss (Raw)": 0.371338427066803, "Full-finetune/Step": 5052, "Full-finetune/Step Time": 6.675222586840391} {"Full-finetune/Learning Rate": 4.076922910266179e-08, "Full-finetune/Loss": 0.7351012229919434, "Full-finetune/Loss (Raw)": 0.5834032893180847, "Full-finetune/Step": 5053, "Full-finetune/Step Time": 6.684236124157906} {"Full-finetune/Learning Rate": 4.020920984678323e-08, "Full-finetune/Loss": 0.736229419708252, "Full-finetune/Loss (Raw)": 0.8755291104316711, "Full-finetune/Step": 5054, "Full-finetune/Step Time": 6.691719325259328} {"Full-finetune/Learning Rate": 3.965305579126111e-08, "Full-finetune/Loss": 0.7361440658569336, "Full-finetune/Loss (Raw)": 0.7415937185287476, "Full-finetune/Step": 5055, "Full-finetune/Step Time": 6.701427990570664} {"Full-finetune/Learning Rate": 3.9100767151928345e-08, "Full-finetune/Loss": 0.7381185293197632, "Full-finetune/Loss (Raw)": 0.812332034111023, "Full-finetune/Step": 5056, "Full-finetune/Step Time": 6.687198404222727} {"Full-finetune/Learning Rate": 3.855234414311904e-08, "Full-finetune/Loss": 0.738250732421875, "Full-finetune/Loss (Raw)": 0.7386630177497864, "Full-finetune/Step": 5057, "Full-finetune/Step Time": 6.690806096419692} {"Full-finetune/Learning Rate": 3.800778697766294e-08, "Full-finetune/Loss": 0.7383544445037842, "Full-finetune/Loss (Raw)": 0.7885333299636841, "Full-finetune/Step": 5058, "Full-finetune/Step Time": 6.704574966803193} {"Full-finetune/Learning Rate": 3.746709586689323e-08, "Full-finetune/Loss": 0.7386251091957092, "Full-finetune/Loss (Raw)": 0.8693218231201172, "Full-finetune/Step": 5059, "Full-finetune/Step Time": 6.701655086129904} {"Full-finetune/Learning Rate": 3.693027102064206e-08, "Full-finetune/Loss": 0.7393316030502319, "Full-finetune/Loss (Raw)": 0.9153057932853699, "Full-finetune/Step": 5060, "Full-finetune/Step Time": 6.702565345913172} {"Full-finetune/Learning Rate": 3.639731264724056e-08, "Full-finetune/Loss": 0.7404738664627075, "Full-finetune/Loss (Raw)": 0.849131166934967, "Full-finetune/Step": 5061, "Full-finetune/Step Time": 6.6852552611380816} {"Full-finetune/Learning Rate": 3.586822095351772e-08, "Full-finetune/Loss": 0.7387555837631226, "Full-finetune/Loss (Raw)": 0.6096076369285583, "Full-finetune/Step": 5062, "Full-finetune/Step Time": 6.689416328445077} {"Full-finetune/Learning Rate": 3.5342996144805966e-08, "Full-finetune/Loss": 0.7389845848083496, "Full-finetune/Loss (Raw)": 0.8760749101638794, "Full-finetune/Step": 5063, "Full-finetune/Step Time": 6.686423050239682} {"Full-finetune/Learning Rate": 3.482163842493447e-08, "Full-finetune/Loss": 0.7385420799255371, "Full-finetune/Loss (Raw)": 0.6231327056884766, "Full-finetune/Step": 5064, "Full-finetune/Step Time": 6.681372545659542} {"Full-finetune/Learning Rate": 3.4304147996230275e-08, "Full-finetune/Loss": 0.7381600141525269, "Full-finetune/Loss (Raw)": 0.862577497959137, "Full-finetune/Step": 5065, "Full-finetune/Step Time": 6.677606210112572} {"Full-finetune/Learning Rate": 3.379052505952274e-08, "Full-finetune/Loss": 0.7386808395385742, "Full-finetune/Loss (Raw)": 0.870527446269989, "Full-finetune/Step": 5066, "Full-finetune/Step Time": 6.669804375618696} {"Full-finetune/Learning Rate": 3.328076981413908e-08, "Full-finetune/Loss": 0.7417075037956238, "Full-finetune/Loss (Raw)": 0.7349370718002319, "Full-finetune/Step": 5067, "Full-finetune/Step Time": 6.644953865557909} {"Full-finetune/Learning Rate": 3.2774882457904386e-08, "Full-finetune/Loss": 0.7431948781013489, "Full-finetune/Loss (Raw)": 0.806427538394928, "Full-finetune/Step": 5068, "Full-finetune/Step Time": 6.637212704867125} {"Full-finetune/Learning Rate": 3.227286318714384e-08, "Full-finetune/Loss": 0.7434436082839966, "Full-finetune/Loss (Raw)": 0.698788583278656, "Full-finetune/Step": 5069, "Full-finetune/Step Time": 6.63899589329958} {"Full-finetune/Learning Rate": 3.177471219668271e-08, "Full-finetune/Loss": 0.7436509132385254, "Full-finetune/Loss (Raw)": 0.7156533598899841, "Full-finetune/Step": 5070, "Full-finetune/Step Time": 6.636542268097401} {"Full-finetune/Learning Rate": 3.1280429679841905e-08, "Full-finetune/Loss": 0.7439795136451721, "Full-finetune/Loss (Raw)": 0.7250419855117798, "Full-finetune/Step": 5071, "Full-finetune/Step Time": 6.653842018917203} {"Full-finetune/Learning Rate": 3.079001582844354e-08, "Full-finetune/Loss": 0.7450737953186035, "Full-finetune/Loss (Raw)": 0.9347211718559265, "Full-finetune/Step": 5072, "Full-finetune/Step Time": 6.656154617667198} {"Full-finetune/Learning Rate": 3.0303470832806494e-08, "Full-finetune/Loss": 0.7447964549064636, "Full-finetune/Loss (Raw)": 0.8344554901123047, "Full-finetune/Step": 5073, "Full-finetune/Step Time": 6.660283036530018} {"Full-finetune/Learning Rate": 2.9820794881751934e-08, "Full-finetune/Loss": 0.7460209131240845, "Full-finetune/Loss (Raw)": 0.8725548386573792, "Full-finetune/Step": 5074, "Full-finetune/Step Time": 6.656218575313687} {"Full-finetune/Learning Rate": 2.9341988162595593e-08, "Full-finetune/Loss": 0.7429539561271667, "Full-finetune/Loss (Raw)": 0.38343772292137146, "Full-finetune/Step": 5075, "Full-finetune/Step Time": 6.661012515425682} {"Full-finetune/Learning Rate": 2.886705086115216e-08, "Full-finetune/Loss": 0.7433210015296936, "Full-finetune/Loss (Raw)": 0.7824798226356506, "Full-finetune/Step": 5076, "Full-finetune/Step Time": 6.661991072818637} {"Full-finetune/Learning Rate": 2.8395983161736418e-08, "Full-finetune/Loss": 0.743605375289917, "Full-finetune/Loss (Raw)": 0.7464643716812134, "Full-finetune/Step": 5077, "Full-finetune/Step Time": 6.683270098641515} {"Full-finetune/Learning Rate": 2.7928785247161027e-08, "Full-finetune/Loss": 0.7441089749336243, "Full-finetune/Loss (Raw)": 0.710589587688446, "Full-finetune/Step": 5078, "Full-finetune/Step Time": 6.667140768840909} {"Full-finetune/Learning Rate": 2.7465457298736508e-08, "Full-finetune/Loss": 0.7443838119506836, "Full-finetune/Loss (Raw)": 0.8726915121078491, "Full-finetune/Step": 5079, "Full-finetune/Step Time": 6.6672643050551414} {"Full-finetune/Learning Rate": 2.700599949626903e-08, "Full-finetune/Loss": 0.7447651624679565, "Full-finetune/Loss (Raw)": 0.8420958518981934, "Full-finetune/Step": 5080, "Full-finetune/Step Time": 6.659580197185278} {"Full-finetune/Learning Rate": 2.6550412018067072e-08, "Full-finetune/Loss": 0.7441754937171936, "Full-finetune/Loss (Raw)": 0.6104341745376587, "Full-finetune/Step": 5081, "Full-finetune/Step Time": 6.656550848856568} {"Full-finetune/Learning Rate": 2.6098695040935874e-08, "Full-finetune/Loss": 0.7407897710800171, "Full-finetune/Loss (Raw)": 0.34817060828208923, "Full-finetune/Step": 5082, "Full-finetune/Step Time": 6.635155085474253} {"Full-finetune/Learning Rate": 2.5650848740177425e-08, "Full-finetune/Loss": 0.7402474880218506, "Full-finetune/Loss (Raw)": 0.7198636531829834, "Full-finetune/Step": 5083, "Full-finetune/Step Time": 6.641045251861215} {"Full-finetune/Learning Rate": 2.5206873289590484e-08, "Full-finetune/Loss": 0.7390950918197632, "Full-finetune/Loss (Raw)": 0.6931424140930176, "Full-finetune/Step": 5084, "Full-finetune/Step Time": 6.651351263746619} {"Full-finetune/Learning Rate": 2.4766768861475e-08, "Full-finetune/Loss": 0.7383359670639038, "Full-finetune/Loss (Raw)": 0.7422473430633545, "Full-finetune/Step": 5085, "Full-finetune/Step Time": 6.642386427149177} {"Full-finetune/Learning Rate": 2.4330535626627684e-08, "Full-finetune/Loss": 0.7401326298713684, "Full-finetune/Loss (Raw)": 0.9087585210800171, "Full-finetune/Step": 5086, "Full-finetune/Step Time": 6.640558043494821} {"Full-finetune/Learning Rate": 2.389817375433978e-08, "Full-finetune/Loss": 0.7426047921180725, "Full-finetune/Loss (Raw)": 0.7601256370544434, "Full-finetune/Step": 5087, "Full-finetune/Step Time": 6.644763343036175} {"Full-finetune/Learning Rate": 2.3469683412403745e-08, "Full-finetune/Loss": 0.7441773414611816, "Full-finetune/Loss (Raw)": 0.9445284605026245, "Full-finetune/Step": 5088, "Full-finetune/Step Time": 6.6433592066168785} {"Full-finetune/Learning Rate": 2.304506476710877e-08, "Full-finetune/Loss": 0.7449626922607422, "Full-finetune/Loss (Raw)": 0.8464605212211609, "Full-finetune/Step": 5089, "Full-finetune/Step Time": 6.641891647130251} {"Full-finetune/Learning Rate": 2.262431798323972e-08, "Full-finetune/Loss": 0.7442489266395569, "Full-finetune/Loss (Raw)": 0.7146100997924805, "Full-finetune/Step": 5090, "Full-finetune/Step Time": 6.644279783591628} {"Full-finetune/Learning Rate": 2.2207443224081528e-08, "Full-finetune/Loss": 0.7446593046188354, "Full-finetune/Loss (Raw)": 0.8045632839202881, "Full-finetune/Step": 5091, "Full-finetune/Step Time": 6.65119088999927} {"Full-finetune/Learning Rate": 2.1794440651414782e-08, "Full-finetune/Loss": 0.7404175996780396, "Full-finetune/Loss (Raw)": 0.32156962156295776, "Full-finetune/Step": 5092, "Full-finetune/Step Time": 6.665533408522606} {"Full-finetune/Learning Rate": 2.1385310425517945e-08, "Full-finetune/Loss": 0.7397451996803284, "Full-finetune/Loss (Raw)": 0.8165678381919861, "Full-finetune/Step": 5093, "Full-finetune/Step Time": 6.6578346118330956} {"Full-finetune/Learning Rate": 2.0980052705165123e-08, "Full-finetune/Loss": 0.7385208010673523, "Full-finetune/Loss (Raw)": 0.5668980479240417, "Full-finetune/Step": 5094, "Full-finetune/Step Time": 6.667181953787804} {"Full-finetune/Learning Rate": 2.05786676476305e-08, "Full-finetune/Loss": 0.736251711845398, "Full-finetune/Loss (Raw)": 0.567092776298523, "Full-finetune/Step": 5095, "Full-finetune/Step Time": 6.673938684165478} {"Full-finetune/Learning Rate": 2.0181155408682816e-08, "Full-finetune/Loss": 0.7355890274047852, "Full-finetune/Loss (Raw)": 0.6770042181015015, "Full-finetune/Step": 5096, "Full-finetune/Step Time": 6.70124770142138} {"Full-finetune/Learning Rate": 1.978751614258978e-08, "Full-finetune/Loss": 0.7334450483322144, "Full-finetune/Loss (Raw)": 0.444919615983963, "Full-finetune/Step": 5097, "Full-finetune/Step Time": 6.69166823476553} {"Full-finetune/Learning Rate": 1.9397750002113634e-08, "Full-finetune/Loss": 0.7321972250938416, "Full-finetune/Loss (Raw)": 0.6302667856216431, "Full-finetune/Step": 5098, "Full-finetune/Step Time": 6.693299632519484} {"Full-finetune/Learning Rate": 1.901185713851672e-08, "Full-finetune/Loss": 0.7306805849075317, "Full-finetune/Loss (Raw)": 0.7107540965080261, "Full-finetune/Step": 5099, "Full-finetune/Step Time": 6.681611265987158} {"Full-finetune/Learning Rate": 1.8629837701554797e-08, "Full-finetune/Loss": 0.7299121618270874, "Full-finetune/Loss (Raw)": 0.6623496413230896, "Full-finetune/Step": 5100, "Full-finetune/Step Time": 6.668406827375293} {"Full-finetune/Learning Rate": 1.825169183948372e-08, "Full-finetune/Loss": 0.7281632423400879, "Full-finetune/Loss (Raw)": 0.7059413194656372, "Full-finetune/Step": 5101, "Full-finetune/Step Time": 6.670008070766926} {"Full-finetune/Learning Rate": 1.7877419699053876e-08, "Full-finetune/Loss": 0.7315058708190918, "Full-finetune/Loss (Raw)": 0.7320118546485901, "Full-finetune/Step": 5102, "Full-finetune/Step Time": 6.65482366271317} {"Full-finetune/Learning Rate": 1.7507021425512417e-08, "Full-finetune/Loss": 0.7330514192581177, "Full-finetune/Loss (Raw)": 0.8905379772186279, "Full-finetune/Step": 5103, "Full-finetune/Step Time": 6.648583117872477} {"Full-finetune/Learning Rate": 1.7140497162605464e-08, "Full-finetune/Loss": 0.729977011680603, "Full-finetune/Loss (Raw)": 0.505870521068573, "Full-finetune/Step": 5104, "Full-finetune/Step Time": 6.6502877958118916} {"Full-finetune/Learning Rate": 1.677784705257146e-08, "Full-finetune/Loss": 0.7310835123062134, "Full-finetune/Loss (Raw)": 0.8247191309928894, "Full-finetune/Step": 5105, "Full-finetune/Step Time": 6.629848813638091} {"Full-finetune/Learning Rate": 1.6419071236148943e-08, "Full-finetune/Loss": 0.7326738238334656, "Full-finetune/Loss (Raw)": 0.9351420998573303, "Full-finetune/Step": 5106, "Full-finetune/Step Time": 6.647060662508011} {"Full-finetune/Learning Rate": 1.6064169852572088e-08, "Full-finetune/Loss": 0.7342472076416016, "Full-finetune/Loss (Raw)": 0.8502504229545593, "Full-finetune/Step": 5107, "Full-finetune/Step Time": 6.645461747422814} {"Full-finetune/Learning Rate": 1.5713143039571833e-08, "Full-finetune/Loss": 0.7351700067520142, "Full-finetune/Loss (Raw)": 0.7684858441352844, "Full-finetune/Step": 5108, "Full-finetune/Step Time": 6.6403804663568735} {"Full-finetune/Learning Rate": 1.5365990933373653e-08, "Full-finetune/Loss": 0.7365022301673889, "Full-finetune/Loss (Raw)": 0.8799769282341003, "Full-finetune/Step": 5109, "Full-finetune/Step Time": 6.643035568296909} {"Full-finetune/Learning Rate": 1.5022713668699785e-08, "Full-finetune/Loss": 0.7355207800865173, "Full-finetune/Loss (Raw)": 0.7399265170097351, "Full-finetune/Step": 5110, "Full-finetune/Step Time": 6.654529627412558} {"Full-finetune/Learning Rate": 1.4683311378771437e-08, "Full-finetune/Loss": 0.7372680902481079, "Full-finetune/Loss (Raw)": 0.6414884924888611, "Full-finetune/Step": 5111, "Full-finetune/Step Time": 6.6707930117845535} {"Full-finetune/Learning Rate": 1.434778419530325e-08, "Full-finetune/Loss": 0.7351638674736023, "Full-finetune/Loss (Raw)": 0.3496342897415161, "Full-finetune/Step": 5112, "Full-finetune/Step Time": 6.676729083061218} {"Full-finetune/Learning Rate": 1.401613224850551e-08, "Full-finetune/Loss": 0.7305996417999268, "Full-finetune/Loss (Raw)": 0.2940400540828705, "Full-finetune/Step": 5113, "Full-finetune/Step Time": 6.700920403003693} {"Full-finetune/Learning Rate": 1.3688355667087483e-08, "Full-finetune/Loss": 0.7317256927490234, "Full-finetune/Loss (Raw)": 0.7375599145889282, "Full-finetune/Step": 5114, "Full-finetune/Step Time": 6.712950207293034} {"Full-finetune/Learning Rate": 1.3364454578252972e-08, "Full-finetune/Loss": 0.7311190962791443, "Full-finetune/Loss (Raw)": 0.7310899496078491, "Full-finetune/Step": 5115, "Full-finetune/Step Time": 6.7049919329583645} {"Full-finetune/Learning Rate": 1.3044429107700319e-08, "Full-finetune/Loss": 0.7266243100166321, "Full-finetune/Loss (Raw)": 0.26152294874191284, "Full-finetune/Step": 5116, "Full-finetune/Step Time": 6.71627408079803} {"Full-finetune/Learning Rate": 1.272827937962573e-08, "Full-finetune/Loss": 0.7276583909988403, "Full-finetune/Loss (Raw)": 0.9479294419288635, "Full-finetune/Step": 5117, "Full-finetune/Step Time": 6.719781959429383} {"Full-finetune/Learning Rate": 1.2416005516722174e-08, "Full-finetune/Loss": 0.7276853919029236, "Full-finetune/Loss (Raw)": 0.7782185673713684, "Full-finetune/Step": 5118, "Full-finetune/Step Time": 6.721409559249878} {"Full-finetune/Learning Rate": 1.2107607640174934e-08, "Full-finetune/Loss": 0.7286381721496582, "Full-finetune/Loss (Raw)": 0.8821932077407837, "Full-finetune/Step": 5119, "Full-finetune/Step Time": 6.722307488322258} {"Full-finetune/Learning Rate": 1.1803085869667164e-08, "Full-finetune/Loss": 0.7241878509521484, "Full-finetune/Loss (Raw)": 0.3459182679653168, "Full-finetune/Step": 5120, "Full-finetune/Step Time": 6.72455126978457} {"Full-finetune/Learning Rate": 1.1502440323379882e-08, "Full-finetune/Loss": 0.7259488105773926, "Full-finetune/Loss (Raw)": 0.7212860584259033, "Full-finetune/Step": 5121, "Full-finetune/Step Time": 6.711908197030425} {"Full-finetune/Learning Rate": 1.1205671117985318e-08, "Full-finetune/Loss": 0.7252216339111328, "Full-finetune/Loss (Raw)": 0.6926825642585754, "Full-finetune/Step": 5122, "Full-finetune/Step Time": 6.691628377884626} {"Full-finetune/Learning Rate": 1.0912778368655786e-08, "Full-finetune/Loss": 0.7270458936691284, "Full-finetune/Loss (Raw)": 0.7964199781417847, "Full-finetune/Step": 5123, "Full-finetune/Step Time": 6.6946231834590435} {"Full-finetune/Learning Rate": 1.0623762189057029e-08, "Full-finetune/Loss": 0.726969838142395, "Full-finetune/Loss (Raw)": 0.8861985206604004, "Full-finetune/Step": 5124, "Full-finetune/Step Time": 6.686566894873977} {"Full-finetune/Learning Rate": 1.0338622691349332e-08, "Full-finetune/Loss": 0.7270722985267639, "Full-finetune/Loss (Raw)": 0.8511610627174377, "Full-finetune/Step": 5125, "Full-finetune/Step Time": 6.691555015742779} {"Full-finetune/Learning Rate": 1.005735998619084e-08, "Full-finetune/Loss": 0.7222067713737488, "Full-finetune/Loss (Raw)": 0.39358270168304443, "Full-finetune/Step": 5126, "Full-finetune/Step Time": 6.707530660554767} {"Full-finetune/Learning Rate": 9.779974182733131e-09, "Full-finetune/Loss": 0.7206103801727295, "Full-finetune/Loss (Raw)": 0.6766614317893982, "Full-finetune/Step": 5127, "Full-finetune/Step Time": 6.721768341958523} {"Full-finetune/Learning Rate": 9.506465388624541e-09, "Full-finetune/Loss": 0.7238145470619202, "Full-finetune/Loss (Raw)": 0.7658934593200684, "Full-finetune/Step": 5128, "Full-finetune/Step Time": 6.698156168684363} {"Full-finetune/Learning Rate": 9.236833710010163e-09, "Full-finetune/Loss": 0.7198484539985657, "Full-finetune/Loss (Raw)": 0.2933370769023895, "Full-finetune/Step": 5129, "Full-finetune/Step Time": 6.704462621361017} {"Full-finetune/Learning Rate": 8.971079251526293e-09, "Full-finetune/Loss": 0.7217592000961304, "Full-finetune/Loss (Raw)": 0.7661929130554199, "Full-finetune/Step": 5130, "Full-finetune/Step Time": 6.701548378914595} {"Full-finetune/Learning Rate": 8.709202116308214e-09, "Full-finetune/Loss": 0.72087562084198, "Full-finetune/Loss (Raw)": 0.5909935832023621, "Full-finetune/Step": 5131, "Full-finetune/Step Time": 6.717221366241574} {"Full-finetune/Learning Rate": 8.451202405985736e-09, "Full-finetune/Loss": 0.7207126021385193, "Full-finetune/Loss (Raw)": 0.5794776082038879, "Full-finetune/Step": 5132, "Full-finetune/Step Time": 6.711543584242463} {"Full-finetune/Learning Rate": 8.197080220683217e-09, "Full-finetune/Loss": 0.7202599048614502, "Full-finetune/Loss (Raw)": 0.7138405442237854, "Full-finetune/Step": 5133, "Full-finetune/Step Time": 6.711847739294171} {"Full-finetune/Learning Rate": 7.946835659020657e-09, "Full-finetune/Loss": 0.7205075621604919, "Full-finetune/Loss (Raw)": 0.8112643361091614, "Full-finetune/Step": 5134, "Full-finetune/Step Time": 6.715293522924185} {"Full-finetune/Learning Rate": 7.700468818111483e-09, "Full-finetune/Loss": 0.7211620211601257, "Full-finetune/Loss (Raw)": 0.6335453391075134, "Full-finetune/Step": 5135, "Full-finetune/Step Time": 6.709634508937597} {"Full-finetune/Learning Rate": 7.457979793568105e-09, "Full-finetune/Loss": 0.7228978276252747, "Full-finetune/Loss (Raw)": 0.911232590675354, "Full-finetune/Step": 5136, "Full-finetune/Step Time": 6.714268105104566} {"Full-finetune/Learning Rate": 7.219368679495242e-09, "Full-finetune/Loss": 0.7216827869415283, "Full-finetune/Loss (Raw)": 0.7715866565704346, "Full-finetune/Step": 5137, "Full-finetune/Step Time": 6.718132248148322} {"Full-finetune/Learning Rate": 6.984635568492159e-09, "Full-finetune/Loss": 0.7213537096977234, "Full-finetune/Loss (Raw)": 0.7392043471336365, "Full-finetune/Step": 5138, "Full-finetune/Step Time": 6.731420518830419} {"Full-finetune/Learning Rate": 6.753780551654876e-09, "Full-finetune/Loss": 0.7182588577270508, "Full-finetune/Loss (Raw)": 0.34518560767173767, "Full-finetune/Step": 5139, "Full-finetune/Step Time": 6.747927695512772} {"Full-finetune/Learning Rate": 6.526803718572838e-09, "Full-finetune/Loss": 0.7184698581695557, "Full-finetune/Loss (Raw)": 0.7960406541824341, "Full-finetune/Step": 5140, "Full-finetune/Step Time": 6.734111053869128} {"Full-finetune/Learning Rate": 6.3037051573333615e-09, "Full-finetune/Loss": 0.7203900218009949, "Full-finetune/Loss (Raw)": 0.8856887221336365, "Full-finetune/Step": 5141, "Full-finetune/Step Time": 6.731974335387349} {"Full-finetune/Learning Rate": 6.0844849545149685e-09, "Full-finetune/Loss": 0.7201024889945984, "Full-finetune/Loss (Raw)": 0.6719585657119751, "Full-finetune/Step": 5142, "Full-finetune/Step Time": 6.735921481624246} {"Full-finetune/Learning Rate": 5.869143195192939e-09, "Full-finetune/Loss": 0.7216067314147949, "Full-finetune/Loss (Raw)": 0.8870922923088074, "Full-finetune/Step": 5143, "Full-finetune/Step Time": 6.749462066218257} {"Full-finetune/Learning Rate": 5.657679962938201e-09, "Full-finetune/Loss": 0.720358669757843, "Full-finetune/Loss (Raw)": 0.5613805055618286, "Full-finetune/Step": 5144, "Full-finetune/Step Time": 6.754788728430867} {"Full-finetune/Learning Rate": 5.4500953398139985e-09, "Full-finetune/Loss": 0.7210983037948608, "Full-finetune/Loss (Raw)": 0.7562030553817749, "Full-finetune/Step": 5145, "Full-finetune/Step Time": 6.755731010809541} {"Full-finetune/Learning Rate": 5.246389406381447e-09, "Full-finetune/Loss": 0.7229222059249878, "Full-finetune/Loss (Raw)": 0.8894824981689453, "Full-finetune/Step": 5146, "Full-finetune/Step Time": 6.750849785283208} {"Full-finetune/Learning Rate": 5.046562241693975e-09, "Full-finetune/Loss": 0.7233971357345581, "Full-finetune/Loss (Raw)": 0.78307044506073, "Full-finetune/Step": 5147, "Full-finetune/Step Time": 6.753073649480939} {"Full-finetune/Learning Rate": 4.850613923300662e-09, "Full-finetune/Loss": 0.7245619297027588, "Full-finetune/Loss (Raw)": 0.5014414191246033, "Full-finetune/Step": 5148, "Full-finetune/Step Time": 6.740869017317891} {"Full-finetune/Learning Rate": 4.658544527244013e-09, "Full-finetune/Loss": 0.7239360809326172, "Full-finetune/Loss (Raw)": 0.7565680742263794, "Full-finetune/Step": 5149, "Full-finetune/Step Time": 6.735541500151157} {"Full-finetune/Learning Rate": 4.470354128065513e-09, "Full-finetune/Loss": 0.7237313985824585, "Full-finetune/Loss (Raw)": 0.7697038650512695, "Full-finetune/Step": 5150, "Full-finetune/Step Time": 6.742072641849518} {"Full-finetune/Learning Rate": 4.286042798794521e-09, "Full-finetune/Loss": 0.7233400344848633, "Full-finetune/Loss (Raw)": 0.70329749584198, "Full-finetune/Step": 5151, "Full-finetune/Step Time": 6.739506108686328} {"Full-finetune/Learning Rate": 4.105610610961597e-09, "Full-finetune/Loss": 0.7229611277580261, "Full-finetune/Loss (Raw)": 0.7046696543693542, "Full-finetune/Step": 5152, "Full-finetune/Step Time": 6.741324840113521} {"Full-finetune/Learning Rate": 3.929057634587397e-09, "Full-finetune/Loss": 0.7205033302307129, "Full-finetune/Loss (Raw)": 0.41957926750183105, "Full-finetune/Step": 5153, "Full-finetune/Step Time": 6.747528837993741} {"Full-finetune/Learning Rate": 3.756383938189334e-09, "Full-finetune/Loss": 0.7169396281242371, "Full-finetune/Loss (Raw)": 0.3761269748210907, "Full-finetune/Step": 5154, "Full-finetune/Step Time": 6.767274046316743} {"Full-finetune/Learning Rate": 3.587589588778251e-09, "Full-finetune/Loss": 0.715954065322876, "Full-finetune/Loss (Raw)": 0.7042813301086426, "Full-finetune/Step": 5155, "Full-finetune/Step Time": 6.769211852923036} {"Full-finetune/Learning Rate": 3.4226746518606356e-09, "Full-finetune/Loss": 0.7154186964035034, "Full-finetune/Loss (Raw)": 0.8287755250930786, "Full-finetune/Step": 5156, "Full-finetune/Step Time": 6.781843880191445} {"Full-finetune/Learning Rate": 3.2616391914364056e-09, "Full-finetune/Loss": 0.7150029540061951, "Full-finetune/Loss (Raw)": 0.7683072686195374, "Full-finetune/Step": 5157, "Full-finetune/Step Time": 6.781773330643773} {"Full-finetune/Learning Rate": 3.1044832699989035e-09, "Full-finetune/Loss": 0.715058445930481, "Full-finetune/Loss (Raw)": 0.7585544586181641, "Full-finetune/Step": 5158, "Full-finetune/Step Time": 6.777895571663976} {"Full-finetune/Learning Rate": 2.9512069485393426e-09, "Full-finetune/Loss": 0.7160266637802124, "Full-finetune/Loss (Raw)": 0.8803324699401855, "Full-finetune/Step": 5159, "Full-finetune/Step Time": 6.7808828223496675} {"Full-finetune/Learning Rate": 2.801810286540141e-09, "Full-finetune/Loss": 0.7165981531143188, "Full-finetune/Loss (Raw)": 0.8962461352348328, "Full-finetune/Step": 5160, "Full-finetune/Step Time": 6.779611296951771} {"Full-finetune/Learning Rate": 2.656293341980476e-09, "Full-finetune/Loss": 0.7168576717376709, "Full-finetune/Loss (Raw)": 0.810269296169281, "Full-finetune/Step": 5161, "Full-finetune/Step Time": 6.789489021524787} {"Full-finetune/Learning Rate": 2.5146561713307315e-09, "Full-finetune/Loss": 0.7160113453865051, "Full-finetune/Loss (Raw)": 0.7029334306716919, "Full-finetune/Step": 5162, "Full-finetune/Step Time": 6.805666485801339} {"Full-finetune/Learning Rate": 2.3768988295580497e-09, "Full-finetune/Loss": 0.7166951298713684, "Full-finetune/Loss (Raw)": 0.801909863948822, "Full-finetune/Step": 5163, "Full-finetune/Step Time": 6.800579406321049} {"Full-finetune/Learning Rate": 2.2430213701252203e-09, "Full-finetune/Loss": 0.7172856330871582, "Full-finetune/Loss (Raw)": 0.8195492029190063, "Full-finetune/Step": 5164, "Full-finetune/Step Time": 6.813478112220764} {"Full-finetune/Learning Rate": 2.1130238449851292e-09, "Full-finetune/Loss": 0.7144590616226196, "Full-finetune/Loss (Raw)": 0.5536495447158813, "Full-finetune/Step": 5165, "Full-finetune/Step Time": 6.8222868498414755} {"Full-finetune/Learning Rate": 1.986906304588532e-09, "Full-finetune/Loss": 0.7131036520004272, "Full-finetune/Loss (Raw)": 0.7259399890899658, "Full-finetune/Step": 5166, "Full-finetune/Step Time": 6.8276687152683735} {"Full-finetune/Learning Rate": 1.8646687978784992e-09, "Full-finetune/Loss": 0.7106348276138306, "Full-finetune/Loss (Raw)": 0.6418049335479736, "Full-finetune/Step": 5167, "Full-finetune/Step Time": 6.830687979236245} {"Full-finetune/Learning Rate": 1.7463113722937518e-09, "Full-finetune/Loss": 0.7131550312042236, "Full-finetune/Loss (Raw)": 0.8696011304855347, "Full-finetune/Step": 5168, "Full-finetune/Step Time": 6.82121872343123} {"Full-finetune/Learning Rate": 1.631834073765326e-09, "Full-finetune/Loss": 0.7126800417900085, "Full-finetune/Loss (Raw)": 0.8600592017173767, "Full-finetune/Step": 5169, "Full-finetune/Step Time": 6.832041682675481} {"Full-finetune/Learning Rate": 1.5212369467210164e-09, "Full-finetune/Loss": 0.7107746601104736, "Full-finetune/Loss (Raw)": 0.629082441329956, "Full-finetune/Step": 5170, "Full-finetune/Step Time": 6.828396685421467} {"Full-finetune/Learning Rate": 1.4145200340809352e-09, "Full-finetune/Loss": 0.7118112444877625, "Full-finetune/Loss (Raw)": 0.9391332268714905, "Full-finetune/Step": 5171, "Full-finetune/Step Time": 6.834143493324518} {"Full-finetune/Learning Rate": 1.3116833772597315e-09, "Full-finetune/Loss": 0.7112562656402588, "Full-finetune/Loss (Raw)": 0.7252444624900818, "Full-finetune/Step": 5172, "Full-finetune/Step Time": 6.829312132671475} {"Full-finetune/Learning Rate": 1.2127270161654824e-09, "Full-finetune/Loss": 0.7143991589546204, "Full-finetune/Loss (Raw)": 0.7022435069084167, "Full-finetune/Step": 5173, "Full-finetune/Step Time": 6.812624961137772} {"Full-finetune/Learning Rate": 1.1176509892030229e-09, "Full-finetune/Loss": 0.7192984819412231, "Full-finetune/Loss (Raw)": 0.962294340133667, "Full-finetune/Step": 5174, "Full-finetune/Step Time": 6.794000709429383} {"Full-finetune/Learning Rate": 1.0264553332683946e-09, "Full-finetune/Loss": 0.7187541127204895, "Full-finetune/Loss (Raw)": 0.7269831299781799, "Full-finetune/Step": 5175, "Full-finetune/Step Time": 6.798912251368165} {"Full-finetune/Learning Rate": 9.39140083752177e-10, "Full-finetune/Loss": 0.7142754793167114, "Full-finetune/Loss (Raw)": 0.3258092999458313, "Full-finetune/Step": 5176, "Full-finetune/Step Time": 6.804531564936042} {"Full-finetune/Learning Rate": 8.557052745405969e-10, "Full-finetune/Loss": 0.7151118516921997, "Full-finetune/Loss (Raw)": 0.8236227035522461, "Full-finetune/Step": 5177, "Full-finetune/Step Time": 6.799753610044718} {"Full-finetune/Learning Rate": 7.761509380133092e-10, "Full-finetune/Loss": 0.7144873738288879, "Full-finetune/Loss (Raw)": 0.6472402215003967, "Full-finetune/Step": 5178, "Full-finetune/Step Time": 6.785518141463399} {"Full-finetune/Learning Rate": 7.004771050433956e-10, "Full-finetune/Loss": 0.7150044441223145, "Full-finetune/Loss (Raw)": 0.674461841583252, "Full-finetune/Step": 5179, "Full-finetune/Step Time": 6.780285311862826} {"Full-finetune/Learning Rate": 6.286838049984755e-10, "Full-finetune/Loss": 0.7167967557907104, "Full-finetune/Loss (Raw)": 0.6007487177848816, "Full-finetune/Step": 5180, "Full-finetune/Step Time": 6.77127562277019} {"Full-finetune/Learning Rate": 5.607710657407061e-10, "Full-finetune/Loss": 0.7184963822364807, "Full-finetune/Loss (Raw)": 0.800956130027771, "Full-finetune/Step": 5181, "Full-finetune/Step Time": 6.76529180444777} {"Full-finetune/Learning Rate": 4.967389136245615e-10, "Full-finetune/Loss": 0.7163044214248657, "Full-finetune/Loss (Raw)": 0.5949622392654419, "Full-finetune/Step": 5182, "Full-finetune/Step Time": 6.767560785636306} {"Full-finetune/Learning Rate": 4.3658737350016354e-10, "Full-finetune/Loss": 0.7169547080993652, "Full-finetune/Loss (Raw)": 0.8248239755630493, "Full-finetune/Step": 5183, "Full-finetune/Step Time": 6.772649109363556} {"Full-finetune/Learning Rate": 3.8031646871106166e-10, "Full-finetune/Loss": 0.715758204460144, "Full-finetune/Loss (Raw)": 0.6591762900352478, "Full-finetune/Step": 5184, "Full-finetune/Step Time": 6.771892320364714} {"Full-finetune/Learning Rate": 3.279262210953427e-10, "Full-finetune/Loss": 0.7164114713668823, "Full-finetune/Loss (Raw)": 0.8222860097885132, "Full-finetune/Step": 5185, "Full-finetune/Step Time": 6.782777618616819} {"Full-finetune/Learning Rate": 2.794166509845209e-10, "Full-finetune/Loss": 0.7132034301757812, "Full-finetune/Loss (Raw)": 0.37790846824645996, "Full-finetune/Step": 5186, "Full-finetune/Step Time": 6.787738040089607} {"Full-finetune/Learning Rate": 2.3478777720353785e-10, "Full-finetune/Loss": 0.7112098336219788, "Full-finetune/Loss (Raw)": 0.614142894744873, "Full-finetune/Step": 5187, "Full-finetune/Step Time": 6.807879522442818} {"Full-finetune/Learning Rate": 1.940396170718728e-10, "Full-finetune/Loss": 0.709322988986969, "Full-finetune/Loss (Raw)": 0.6737805008888245, "Full-finetune/Step": 5188, "Full-finetune/Step Time": 6.8093437403440475} {"Full-finetune/Learning Rate": 1.5717218640465272e-10, "Full-finetune/Loss": 0.7071919441223145, "Full-finetune/Loss (Raw)": 0.5763542652130127, "Full-finetune/Step": 5189, "Full-finetune/Step Time": 6.821600127965212} {"Full-finetune/Learning Rate": 1.2418549950821145e-10, "Full-finetune/Loss": 0.7071464657783508, "Full-finetune/Loss (Raw)": 0.6037996411323547, "Full-finetune/Step": 5190, "Full-finetune/Step Time": 6.8200962562114} {"Full-finetune/Learning Rate": 9.507956918453076e-11, "Full-finetune/Loss": 0.7051472663879395, "Full-finetune/Loss (Raw)": 0.6201711297035217, "Full-finetune/Step": 5191, "Full-finetune/Step Time": 6.8323249369859695} {"Full-finetune/Learning Rate": 6.985440672790945e-11, "Full-finetune/Loss": 0.7058964967727661, "Full-finetune/Loss (Raw)": 0.7190332412719727, "Full-finetune/Step": 5192, "Full-finetune/Step Time": 6.828975239768624} {"Full-finetune/Learning Rate": 4.851002192940435e-11, "Full-finetune/Loss": 0.7043271660804749, "Full-finetune/Loss (Raw)": 0.6616999506950378, "Full-finetune/Step": 5193, "Full-finetune/Step Time": 6.840839413926005} {"Full-finetune/Learning Rate": 3.1046423071279254e-11, "Full-finetune/Loss": 0.7001548409461975, "Full-finetune/Loss (Raw)": 0.33647528290748596, "Full-finetune/Step": 5194, "Full-finetune/Step Time": 6.857359318062663} {"Full-finetune/Learning Rate": 1.7463616930335493e-11, "Full-finetune/Loss": 0.6999322772026062, "Full-finetune/Loss (Raw)": 0.7064501643180847, "Full-finetune/Step": 5195, "Full-finetune/Step Time": 6.857781944796443} {"Full-finetune/Learning Rate": 7.761608780132434e-12, "Full-finetune/Loss": 0.6990768909454346, "Full-finetune/Loss (Raw)": 0.6969334483146667, "Full-finetune/Step": 5196, "Full-finetune/Step Time": 6.8642480578273535} {"Full-finetune/Learning Rate": 1.9404023832159115e-12, "Full-finetune/Loss": 0.699927806854248, "Full-finetune/Loss (Raw)": 0.8077055811882019, "Full-finetune/Step": 5197, "Full-finetune/Step Time": 6.856661895290017} {"Full-finetune/Learning Rate": 0.0, "Full-finetune/Loss": 0.6998637914657593, "Full-finetune/Loss (Raw)": 0.7074692845344543, "Full-finetune/Step": 5198, "Full-finetune/Step Time": 6.853703048080206}