Model save

README.md

@@ -1,16 +1,12 @@
 ---
 license: apache-2.0
-base_model: alignment-handbook/zephyr-7b-sft-qlora
+library_name: peft
 tags:
-- alignment-handbook
-- trl
-- dpo
-- generated_from_trainer
 - trl
 - dpo
+- alignment-handbook
 - generated_from_trainer
-datasets:
-- HuggingFaceH4/ultrafeedback_binarized
+base_model: mistralai/Mistral-7B-v0.1
 model-index:
 - name: zephyr-7b-dpo-qlora-fsdp
   results: []
@@ -21,7 +17,7 @@ should probably proofread and complete it, then remove this comment. -->
 
 # zephyr-7b-dpo-qlora-fsdp
 
-This model is a fine-tuned version of [alignment-handbook/zephyr-7b-sft-qlora](https://huggingface.co/alignment-handbook/zephyr-7b-sft-qlora) on the HuggingFaceH4/ultrafeedback_binarized dataset.
+This model is a fine-tuned version of [mistralai/Mistral-7B-v0.1](https://huggingface.co/mistralai/Mistral-7B-v0.1) on an unknown dataset.
 
 ## Model description
 
@@ -54,7 +50,8 @@ The following hyperparameters were used during training:
 
 ### Framework versions
 
+- PEFT 0.9.0
 - Transformers 4.38.1
 - Pytorch 2.2.0+cu118
 - Datasets 2.17.1
-- Tokenizers 0.15.2
+- Tokenizers 0.15.2

adapter_config.json

@@ -0,0 +1,33 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "mistralai/Mistral-7B-v0.1",
+  "bias": "none",
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_dropout": 0.05,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "v_proj",
+    "q_proj",
+    "o_proj",
+    "gate_proj",
+    "down_proj",
+    "k_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "use_dora": false,
+  "use_rslora": false
+}

adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d11c467ea2ced3ece27c4663979d7c2e938981d71182ed8d30aedc7f981b8b34
+size 671150064

trainer_state.json

@@ -1,2142 +1,20 @@
 {
   "best_metric": null,
   "best_model_checkpoint": null,
-  "epoch": 4.985279685966634,
-  "eval_steps": 100,
-  "global_step": 1270,
+  "epoch": null,
+  "eval_steps": 500,
+  "global_step": 0,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
-  "log_history": [
-    {
-      "epoch": 0.0,
-      "grad_norm": 2.734375,
-      "learning_rate": 3.9370078740157486e-08,
-      "logits/chosen": -2.356706142425537,
-      "logits/rejected": -2.3367161750793457,
-      "logps/chosen": -287.937255859375,
-      "logps/rejected": -266.50421142578125,
-      "loss": 0.0001,
-      "rewards/accuracies": 0.0,
-      "rewards/chosen": 0.0,
-      "rewards/margins": 0.0,
-      "rewards/rejected": 0.0,
-      "step": 1
-    },
-    {
-      "epoch": 0.04,
-      "grad_norm": 106.5,
-      "learning_rate": 3.937007874015748e-07,
-      "logits/chosen": -2.3996241092681885,
-      "logits/rejected": -2.353182554244995,
-      "logps/chosen": -273.30889892578125,
-      "logps/rejected": -240.43850708007812,
-      "loss": 0.118,
-      "rewards/accuracies": 0.46388885378837585,
-      "rewards/chosen": 0.00026795046869665384,
-      "rewards/margins": 0.00038285815389826894,
-      "rewards/rejected": -0.00011490769247757271,
-      "step": 10
-    },
-    {
-      "epoch": 0.08,
-      "grad_norm": 129.0,
-      "learning_rate": 7.874015748031496e-07,
-      "logits/chosen": -2.38779878616333,
-      "logits/rejected": -2.3378489017486572,
-      "logps/chosen": -266.8452453613281,
-      "logps/rejected": -258.0478515625,
-      "loss": 0.1346,
-      "rewards/accuracies": 0.5525000095367432,
-      "rewards/chosen": 0.0013033099239692092,
-      "rewards/margins": 0.0008336402243003249,
-      "rewards/rejected": 0.0004696696996688843,
-      "step": 20
-    },
-    {
-      "epoch": 0.12,
-      "grad_norm": 108.0,
-      "learning_rate": 1.1811023622047246e-06,
-      "logits/chosen": -2.43515682220459,
-      "logits/rejected": -2.382305145263672,
-      "logps/chosen": -299.69366455078125,
-      "logps/rejected": -271.0403747558594,
-      "loss": 0.1585,
-      "rewards/accuracies": 0.4999999403953552,
-      "rewards/chosen": 0.002506708027794957,
-      "rewards/margins": 0.0005399176734499633,
-      "rewards/rejected": 0.0019667900633066893,
-      "step": 30
-    },
-    {
-      "epoch": 0.16,
-      "grad_norm": 95.0,
-      "learning_rate": 1.5748031496062992e-06,
-      "logits/chosen": -2.3607640266418457,
-      "logits/rejected": -2.313088893890381,
-      "logps/chosen": -288.7026672363281,
-      "logps/rejected": -253.4452667236328,
-      "loss": 0.1404,
-      "rewards/accuracies": 0.5774999856948853,
-      "rewards/chosen": 0.002137306611984968,
-      "rewards/margins": 0.0012014020467177033,
-      "rewards/rejected": 0.0009359045652672648,
-      "step": 40
-    },
-    {
-      "epoch": 0.2,
-      "grad_norm": 109.0,
-      "learning_rate": 1.968503937007874e-06,
-      "logits/chosen": -2.407731533050537,
-      "logits/rejected": -2.3856372833251953,
-      "logps/chosen": -267.39801025390625,
-      "logps/rejected": -264.8341064453125,
-      "loss": 0.1726,
-      "rewards/accuracies": 0.6200000047683716,
-      "rewards/chosen": 0.0022850066889077425,
-      "rewards/margins": 0.0014212832320481539,
-      "rewards/rejected": 0.0008637232822366059,
-      "step": 50
-    },
-    {
-      "epoch": 0.24,
-      "grad_norm": 135.0,
-      "learning_rate": 2.362204724409449e-06,
-      "logits/chosen": -2.3825833797454834,
-      "logits/rejected": -2.3412365913391113,
-      "logps/chosen": -272.9526672363281,
-      "logps/rejected": -239.8430633544922,
-      "loss": 0.1314,
-      "rewards/accuracies": 0.6150000095367432,
-      "rewards/chosen": 0.002347386907786131,
-      "rewards/margins": 0.001956633059307933,
-      "rewards/rejected": 0.00039075379027053714,
-      "step": 60
-    },
-    {
-      "epoch": 0.27,
-      "grad_norm": 89.5,
-      "learning_rate": 2.755905511811024e-06,
-      "logits/chosen": -2.382422924041748,
-      "logits/rejected": -2.3432180881500244,
-      "logps/chosen": -273.59881591796875,
-      "logps/rejected": -257.91790771484375,
-      "loss": 0.1777,
-      "rewards/accuracies": 0.6025000214576721,
-      "rewards/chosen": 0.002635209821164608,
-      "rewards/margins": 0.0014770927373319864,
-      "rewards/rejected": 0.0011581169674172997,
-      "step": 70
-    },
-    {
-      "epoch": 0.31,
-      "grad_norm": 125.0,
-      "learning_rate": 3.1496062992125985e-06,
-      "logits/chosen": -2.394813060760498,
-      "logits/rejected": -2.3611693382263184,
-      "logps/chosen": -279.48638916015625,
-      "logps/rejected": -260.6809997558594,
-      "loss": 0.1944,
-      "rewards/accuracies": 0.6100000739097595,
-      "rewards/chosen": 0.0013221392873674631,
-      "rewards/margins": 0.0018236342584714293,
-      "rewards/rejected": -0.0005014949128963053,
-      "step": 80
-    },
-    {
-      "epoch": 0.35,
-      "grad_norm": 146.0,
-      "learning_rate": 3.5433070866141735e-06,
-      "logits/chosen": -2.3927130699157715,
-      "logits/rejected": -2.3439621925354004,
-      "logps/chosen": -264.44195556640625,
-      "logps/rejected": -236.5146942138672,
-      "loss": 0.2032,
-      "rewards/accuracies": 0.5475000143051147,
-      "rewards/chosen": 0.001684651942923665,
-      "rewards/margins": 0.0011215232079848647,
-      "rewards/rejected": 0.0005631285603158176,
-      "step": 90
-    },
-    {
-      "epoch": 0.39,
-      "grad_norm": 145.0,
-      "learning_rate": 3.937007874015748e-06,
-      "logits/chosen": -2.4192750453948975,
-      "logits/rejected": -2.3637657165527344,
-      "logps/chosen": -280.88238525390625,
-      "logps/rejected": -253.8561553955078,
-      "loss": 0.2536,
-      "rewards/accuracies": 0.5900000333786011,
-      "rewards/chosen": 0.003346907440572977,
-      "rewards/margins": 0.002142687328159809,
-      "rewards/rejected": 0.001204220112413168,
-      "step": 100
-    },
-    {
-      "epoch": 0.39,
-      "eval_logits/chosen": -2.412179470062256,
-      "eval_logits/rejected": -2.3735485076904297,
-      "eval_logps/chosen": -267.4339904785156,
-      "eval_logps/rejected": -242.3385467529297,
-      "eval_loss": 0.27917271852493286,
-      "eval_rewards/accuracies": 0.6041666865348816,
-      "eval_rewards/chosen": 0.002446565078571439,
-      "eval_rewards/margins": 0.0018827051389962435,
-      "eval_rewards/rejected": 0.0005638597067445517,
-      "eval_runtime": 124.6972,
-      "eval_samples_per_second": 16.039,
-      "eval_steps_per_second": 0.337,
-      "step": 100
-    },
-    {
-      "epoch": 0.43,
-      "grad_norm": 196.0,
-      "learning_rate": 4.330708661417324e-06,
-      "logits/chosen": -2.383650064468384,
-      "logits/rejected": -2.345914125442505,
-      "logps/chosen": -298.79864501953125,
-      "logps/rejected": -272.7557373046875,
-      "loss": 0.2924,
-      "rewards/accuracies": 0.637499988079071,
-      "rewards/chosen": 0.004891454242169857,
-      "rewards/margins": 0.002777325687929988,
-      "rewards/rejected": 0.0021141283214092255,
-      "step": 110
-    },
-    {
-      "epoch": 0.47,
-      "grad_norm": 138.0,
-      "learning_rate": 4.724409448818898e-06,
-      "logits/chosen": -2.3592543601989746,
-      "logits/rejected": -2.29665470123291,
-      "logps/chosen": -285.70538330078125,
-      "logps/rejected": -253.1526336669922,
-      "loss": 0.3272,
-      "rewards/accuracies": 0.6274999380111694,
-      "rewards/chosen": 0.005152740981429815,
-      "rewards/margins": 0.0025280567351728678,
-      "rewards/rejected": 0.002624684479087591,
-      "step": 120
-    },
-    {
-      "epoch": 0.51,
-      "grad_norm": 227.0,
-      "learning_rate": 4.999915012051437e-06,
-      "logits/chosen": -2.3975679874420166,
-      "logits/rejected": -2.3674397468566895,
-      "logps/chosen": -261.0414733886719,
-      "logps/rejected": -249.3390350341797,
-      "loss": 0.3621,
-      "rewards/accuracies": 0.5925000309944153,
-      "rewards/chosen": 0.004668754059821367,
-      "rewards/margins": 0.0027659868355840445,
-      "rewards/rejected": 0.0019027665257453918,
-      "step": 130
-    },
-    {
-      "epoch": 0.55,
-      "grad_norm": 226.0,
-      "learning_rate": 4.9984042759305375e-06,
-      "logits/chosen": -2.4002552032470703,
-      "logits/rejected": -2.34761381149292,
-      "logps/chosen": -273.1309509277344,
-      "logps/rejected": -247.2257843017578,
-      "loss": 0.387,
-      "rewards/accuracies": 0.5449999570846558,
-      "rewards/chosen": 0.0018628574907779694,
-      "rewards/margins": 0.0016136768972501159,
-      "rewards/rejected": 0.00024918062263168395,
-      "step": 140
-    },
-    {
-      "epoch": 0.59,
-      "grad_norm": 196.0,
-      "learning_rate": 4.9950062323425556e-06,
-      "logits/chosen": -2.3986167907714844,
-      "logits/rejected": -2.358851194381714,
-      "logps/chosen": -271.7406921386719,
-      "logps/rejected": -249.7233428955078,
-      "loss": 0.3955,
-      "rewards/accuracies": 0.6075000166893005,
-      "rewards/chosen": 0.00026552577037364244,
-      "rewards/margins": 0.0028056656010448933,
-      "rewards/rejected": -0.002540139714255929,
-      "step": 150
-    },
-    {
-      "epoch": 0.63,
-      "grad_norm": 190.0,
-      "learning_rate": 4.989723448187132e-06,
-      "logits/chosen": -2.3923754692077637,
-      "logits/rejected": -2.3684566020965576,
-      "logps/chosen": -285.33184814453125,
-      "logps/rejected": -282.5887756347656,
-      "loss": 0.4099,
-      "rewards/accuracies": 0.5924999713897705,
-      "rewards/chosen": 0.0046120877377688885,
-      "rewards/margins": 0.0034965365193784237,
-      "rewards/rejected": 0.0011155509855598211,
-      "step": 160
-    },
-    {
-      "epoch": 0.67,
-      "grad_norm": 222.0,
-      "learning_rate": 4.982559914106645e-06,
-      "logits/chosen": -2.416792392730713,
-      "logits/rejected": -2.3636841773986816,
-      "logps/chosen": -297.1885681152344,
-      "logps/rejected": -281.99481201171875,
-      "loss": 0.5527,
-      "rewards/accuracies": 0.6299999952316284,
-      "rewards/chosen": 0.004418404307216406,
-      "rewards/margins": 0.0038651120848953724,
-      "rewards/rejected": 0.0005532926879823208,
-      "step": 170
-    },
-    {
-      "epoch": 0.71,
-      "grad_norm": 189.0,
-      "learning_rate": 4.973521041471662e-06,
-      "logits/chosen": -2.443068504333496,
-      "logits/rejected": -2.4043824672698975,
-      "logps/chosen": -284.85546875,
-      "logps/rejected": -246.3272705078125,
-      "loss": 0.467,
-      "rewards/accuracies": 0.5925000309944153,
-      "rewards/chosen": 0.003813292132690549,
-      "rewards/margins": 0.0034597956109791994,
-      "rewards/rejected": 0.00035349628888070583,
-      "step": 180
-    },
-    {
-      "epoch": 0.75,
-      "grad_norm": 186.0,
-      "learning_rate": 4.962613658293158e-06,
-      "logits/chosen": -2.364473581314087,
-      "logits/rejected": -2.336862802505493,
-      "logps/chosen": -260.23297119140625,
-      "logps/rejected": -244.9732666015625,
-      "loss": 0.4326,
-      "rewards/accuracies": 0.6074999570846558,
-      "rewards/chosen": 0.0008181848679669201,
-      "rewards/margins": 0.003228846937417984,
-      "rewards/rejected": -0.0024106616619974375,
-      "step": 190
-    },
-    {
-      "epoch": 0.79,
-      "grad_norm": 196.0,
-      "learning_rate": 4.949846004064605e-06,
-      "logits/chosen": -2.414769411087036,
-      "logits/rejected": -2.394395351409912,
-      "logps/chosen": -281.43670654296875,
-      "logps/rejected": -265.92974853515625,
-      "loss": 0.5352,
-      "rewards/accuracies": 0.5974999666213989,
-      "rewards/chosen": 0.00015548830560874194,
-      "rewards/margins": 0.003786542685702443,
-      "rewards/rejected": -0.003631054190918803,
-      "step": 200
-    },
-    {
-      "epoch": 0.79,
-      "eval_logits/chosen": -2.4014458656311035,
-      "eval_logits/rejected": -2.362912893295288,
-      "eval_logps/chosen": -267.5639953613281,
-      "eval_logps/rejected": -242.58323669433594,
-      "eval_loss": 0.5010271072387695,
-      "eval_rewards/accuracies": 0.574404776096344,
-      "eval_rewards/chosen": 0.0011464261915534735,
-      "eval_rewards/margins": 0.003029454033821821,
-      "eval_rewards/rejected": -0.001883027609437704,
-      "eval_runtime": 123.1297,
-      "eval_samples_per_second": 16.243,
-      "eval_steps_per_second": 0.341,
-      "step": 200
-    },
-    {
-      "epoch": 0.82,
-      "grad_norm": 254.0,
-      "learning_rate": 4.935227723537811e-06,
-      "logits/chosen": -2.406309127807617,
-      "logits/rejected": -2.360525131225586,
-      "logps/chosen": -296.66070556640625,
-      "logps/rejected": -266.5440979003906,
-      "loss": 0.5154,
-      "rewards/accuracies": 0.6399999856948853,
-      "rewards/chosen": 0.0017778485780581832,
-      "rewards/margins": 0.004943528212606907,
-      "rewards/rejected": -0.0031656797509640455,
-      "step": 210
-    },
-    {
-      "epoch": 0.86,
-      "grad_norm": 288.0,
-      "learning_rate": 4.918769859437233e-06,
-      "logits/chosen": -2.3826467990875244,
-      "logits/rejected": -2.3224892616271973,
-      "logps/chosen": -274.14825439453125,
-      "logps/rejected": -252.51400756835938,
-      "loss": 0.5475,
-      "rewards/accuracies": 0.6050000190734863,
-      "rewards/chosen": 0.004731293302029371,
-      "rewards/margins": 0.0043090214021503925,
-      "rewards/rejected": 0.0004222726565785706,
-      "step": 220
-    },
-    {
-      "epoch": 0.9,
-      "grad_norm": 182.0,
-      "learning_rate": 4.900484844118235e-06,
-      "logits/chosen": -2.3914456367492676,
-      "logits/rejected": -2.3371217250823975,
-      "logps/chosen": -280.33282470703125,
-      "logps/rejected": -240.6001739501953,
-      "loss": 0.561,
-      "rewards/accuracies": 0.6175000071525574,
-      "rewards/chosen": 0.002294857520610094,
-      "rewards/margins": 0.0033697611652314663,
-      "rewards/rejected": -0.0010749038774520159,
-      "step": 230
-    },
-    {
-      "epoch": 0.94,
-      "grad_norm": 196.0,
-      "learning_rate": 4.880386490175634e-06,
-      "logits/chosen": -2.359574794769287,
-      "logits/rejected": -2.327416181564331,
-      "logps/chosen": -290.989013671875,
-      "logps/rejected": -268.49395751953125,
-      "loss": 0.5661,
-      "rewards/accuracies": 0.5975000262260437,
-      "rewards/chosen": -0.00016075666644610465,
-      "rewards/margins": 0.003780897008255124,
-      "rewards/rejected": -0.003941653296351433,
-      "step": 240
-    },
-    {
-      "epoch": 0.98,
-      "grad_norm": 232.0,
-      "learning_rate": 4.8584899800095865e-06,
-      "logits/chosen": -2.4217326641082764,
-      "logits/rejected": -2.3503971099853516,
-      "logps/chosen": -288.73858642578125,
-      "logps/rejected": -258.19317626953125,
-      "loss": 0.5762,
-      "rewards/accuracies": 0.6399999856948853,
-      "rewards/chosen": 0.00230691721662879,
-      "rewards/margins": 0.004124250262975693,
-      "rewards/rejected": -0.0018173331627622247,
-      "step": 250
-    },
-    {
-      "epoch": 1.02,
-      "grad_norm": 162.0,
-      "learning_rate": 4.834811854356729e-06,
-      "logits/chosen": -2.406905174255371,
-      "logits/rejected": -2.3617682456970215,
-      "logps/chosen": -266.1622619628906,
-      "logps/rejected": -245.754150390625,
-      "loss": 0.4313,
-      "rewards/accuracies": 0.6574999690055847,
-      "rewards/chosen": 0.006862832698971033,
-      "rewards/margins": 0.010333456099033356,
-      "rewards/rejected": -0.0034706243313848972,
-      "step": 260
-    },
-    {
-      "epoch": 1.06,
-      "grad_norm": 119.0,
-      "learning_rate": 4.809369999795219e-06,
-      "logits/chosen": -2.367124080657959,
-      "logits/rejected": -2.352210760116577,
-      "logps/chosen": -271.3493957519531,
-      "logps/rejected": -274.4390563964844,
-      "loss": 0.2246,
-      "rewards/accuracies": 0.7775000333786011,
-      "rewards/chosen": 0.011581487953662872,
-      "rewards/margins": 0.02090141549706459,
-      "rewards/rejected": -0.009319926612079144,
-      "step": 270
-    },
-    {
-      "epoch": 1.1,
-      "grad_norm": 152.0,
-      "learning_rate": 4.7821836352331235e-06,
-      "logits/chosen": -2.4189422130584717,
-      "logits/rejected": -2.362922430038452,
-      "logps/chosen": -276.532470703125,
-      "logps/rejected": -254.9628143310547,
-      "loss": 0.3068,
-      "rewards/accuracies": 0.7899999618530273,
-      "rewards/chosen": 0.010409007780253887,
-      "rewards/margins": 0.019158251583576202,
-      "rewards/rejected": -0.008749241940677166,
-      "step": 280
-    },
-    {
-      "epoch": 1.14,
-      "grad_norm": 122.5,
-      "learning_rate": 4.7532732973903525e-06,
-      "logits/chosen": -2.392087936401367,
-      "logits/rejected": -2.3331363201141357,
-      "logps/chosen": -281.2344970703125,
-      "logps/rejected": -266.2068176269531,
-      "loss": 0.2544,
-      "rewards/accuracies": 0.7375000715255737,
-      "rewards/chosen": 0.01263010036200285,
-      "rewards/margins": 0.01806234009563923,
-      "rewards/rejected": -0.0054322415962815285,
-      "step": 290
-    },
-    {
-      "epoch": 1.18,
-      "grad_norm": 177.0,
-      "learning_rate": 4.722660825285122e-06,
-      "logits/chosen": -2.413367509841919,
-      "logits/rejected": -2.3771374225616455,
-      "logps/chosen": -278.57904052734375,
-      "logps/rejected": -270.5982360839844,
-      "loss": 0.3676,
-      "rewards/accuracies": 0.762499988079071,
-      "rewards/chosen": 0.014171945862472057,
-      "rewards/margins": 0.020834611728787422,
-      "rewards/rejected": -0.0066626654006540775,
-      "step": 300
-    },
-    {
-      "epoch": 1.18,
-      "eval_logits/chosen": -2.416761636734009,
-      "eval_logits/rejected": -2.3788468837738037,
-      "eval_logps/chosen": -266.88555908203125,
-      "eval_logps/rejected": -242.12106323242188,
-      "eval_loss": 0.8293091654777527,
-      "eval_rewards/accuracies": 0.5982142686843872,
-      "eval_rewards/chosen": 0.00793052464723587,
-      "eval_rewards/margins": 0.005191552918404341,
-      "eval_rewards/rejected": 0.0027389726601541042,
-      "eval_runtime": 123.1487,
-      "eval_samples_per_second": 16.241,
-      "eval_steps_per_second": 0.341,
-      "step": 300
-    },
-    {
-      "epoch": 1.22,
-      "grad_norm": 140.0,
-      "learning_rate": 4.690369343736637e-06,
-      "logits/chosen": -2.4030745029449463,
-      "logits/rejected": -2.368807077407837,
-      "logps/chosen": -278.1623229980469,
-      "logps/rejected": -264.98541259765625,
-      "loss": 0.3085,
-      "rewards/accuracies": 0.7649999856948853,
-      "rewards/chosen": 0.01418610941618681,
-      "rewards/margins": 0.01857883669435978,
-      "rewards/rejected": -0.004392724949866533,
-      "step": 310
-    },
-    {
-      "epoch": 1.26,
-      "grad_norm": 185.0,
-      "learning_rate": 4.656423245896494e-06,
-      "logits/chosen": -2.4111828804016113,
-      "logits/rejected": -2.3645176887512207,
-      "logps/chosen": -270.58477783203125,
-      "logps/rejected": -256.71099853515625,
-      "loss": 0.2759,
-      "rewards/accuracies": 0.7575000524520874,
-      "rewards/chosen": 0.008453365415334702,
-      "rewards/margins": 0.01515539176762104,
-      "rewards/rejected": -0.0067020258866250515,
-      "step": 320
-    },
-    {
-      "epoch": 1.3,
-      "grad_norm": 163.0,
-      "learning_rate": 4.6208481748219645e-06,
-      "logits/chosen": -2.392019271850586,
-      "logits/rejected": -2.367377519607544,
-      "logps/chosen": -271.10015869140625,
-      "logps/rejected": -256.1971435546875,
-      "loss": 0.2666,
-      "rewards/accuracies": 0.7675000429153442,
-      "rewards/chosen": 0.007386817596852779,
-      "rewards/margins": 0.019648974761366844,
-      "rewards/rejected": -0.01226215623319149,
-      "step": 330
-    },
-    {
-      "epoch": 1.33,
-      "grad_norm": 142.0,
-      "learning_rate": 4.583671004105096e-06,
-      "logits/chosen": -2.3817129135131836,
-      "logits/rejected": -2.342694044113159,
-      "logps/chosen": -275.9081115722656,
-      "logps/rejected": -251.59848022460938,
-      "loss": 0.2849,
-      "rewards/accuracies": 0.75,
-      "rewards/chosen": 0.0086748655885458,
-      "rewards/margins": 0.01924619823694229,
-      "rewards/rejected": -0.010571330785751343,
-      "step": 340
-    },
-    {
-      "epoch": 1.37,
-      "grad_norm": 113.0,
-      "learning_rate": 4.544919817572262e-06,
-      "logits/chosen": -2.3859992027282715,
-      "logits/rejected": -2.325930118560791,
-      "logps/chosen": -272.119873046875,
-      "logps/rejected": -247.908935546875,
-      "loss": 0.2871,
-      "rewards/accuracies": 0.7649999856948853,
-      "rewards/chosen": 0.012000922113656998,
-      "rewards/margins": 0.019705070182681084,
-      "rewards/rejected": -0.007704148534685373,
-      "step": 350
-    },
-    {
-      "epoch": 1.41,
-      "grad_norm": 154.0,
-      "learning_rate": 4.504623888069497e-06,
-      "logits/chosen": -2.397146701812744,
-      "logits/rejected": -2.3492813110351562,
-      "logps/chosen": -271.32171630859375,
-      "logps/rejected": -247.8098907470703,
-      "loss": 0.3405,
-      "rewards/accuracies": 0.7600000500679016,
-      "rewards/chosen": 0.013489668257534504,
-      "rewards/margins": 0.017752837389707565,
-      "rewards/rejected": -0.0042631677351891994,
-      "step": 360
-    },
-    {
-      "epoch": 1.45,
-      "grad_norm": 139.0,
-      "learning_rate": 4.462813655349637e-06,
-      "logits/chosen": -2.372323751449585,
-      "logits/rejected": -2.3170628547668457,
-      "logps/chosen": -268.582275390625,
-      "logps/rejected": -245.34097290039062,
-      "loss": 0.3015,
-      "rewards/accuracies": 0.7275000214576721,
-      "rewards/chosen": 0.013011058792471886,
-      "rewards/margins": 0.018547596409916878,
-      "rewards/rejected": -0.005536535754799843,
-      "step": 370
-    },
-    {
-      "epoch": 1.49,
-      "grad_norm": 195.0,
-      "learning_rate": 4.419520703077975e-06,
-      "logits/chosen": -2.3980116844177246,
-      "logits/rejected": -2.3060975074768066,
-      "logps/chosen": -284.5008544921875,
-      "logps/rejected": -232.28515625,
-      "loss": 0.2953,
-      "rewards/accuracies": 0.7300000190734863,
-      "rewards/chosen": 0.009602969512343407,
-      "rewards/margins": 0.01879434660077095,
-      "rewards/rejected": -0.00919137429445982,
-      "step": 380
-    },
-    {
-      "epoch": 1.53,
-      "grad_norm": 184.0,
-      "learning_rate": 4.3747777349737905e-06,
-      "logits/chosen": -2.394030809402466,
-      "logits/rejected": -2.3477187156677246,
-      "logps/chosen": -299.9769592285156,
-      "logps/rejected": -268.77301025390625,
-      "loss": 0.318,
-      "rewards/accuracies": 0.7575000524520874,
-      "rewards/chosen": 0.011008193716406822,
-      "rewards/margins": 0.020197119563817978,
-      "rewards/rejected": -0.009188923053443432,
-      "step": 390
-    },
-    {
-      "epoch": 1.57,
-      "grad_norm": 153.0,
-      "learning_rate": 4.328618550105802e-06,
-      "logits/chosen": -2.3696258068084717,
-      "logits/rejected": -2.341409206390381,
-      "logps/chosen": -271.8193664550781,
-      "logps/rejected": -264.6459045410156,
-      "loss": 0.366,
-      "rewards/accuracies": 0.7400000691413879,
-      "rewards/chosen": 0.012812617234885693,
-      "rewards/margins": 0.018871381878852844,
-      "rewards/rejected": -0.006058765109628439,
-      "step": 400
-    },
-    {
-      "epoch": 1.57,
-      "eval_logits/chosen": -2.4146392345428467,
-      "eval_logits/rejected": -2.377370595932007,
-      "eval_logps/chosen": -267.025634765625,
-      "eval_logps/rejected": -242.3221435546875,
-      "eval_loss": 0.8238700032234192,
-      "eval_rewards/accuracies": 0.6398809552192688,
-      "eval_rewards/chosen": 0.006530104670673609,
-      "eval_rewards/margins": 0.005802116356790066,
-      "eval_rewards/rejected": 0.0007279877318069339,
-      "eval_runtime": 123.0966,
-      "eval_samples_per_second": 16.247,
-      "eval_steps_per_second": 0.341,
-      "step": 400
-    },
-    {
-      "epoch": 1.61,
-      "grad_norm": 91.0,
-      "learning_rate": 4.2810780173601675e-06,
-      "logits/chosen": -2.3998053073883057,
-      "logits/rejected": -2.341407060623169,
-      "logps/chosen": -285.62054443359375,
-      "logps/rejected": -247.3552703857422,
-      "loss": 0.3234,
-      "rewards/accuracies": 0.762499988079071,
-      "rewards/chosen": 0.008955768309533596,
-      "rewards/margins": 0.020532304421067238,
-      "rewards/rejected": -0.011576534248888493,
-      "step": 410
-    },
-    {
-      "epoch": 1.65,
-      "grad_norm": 174.0,
-      "learning_rate": 4.232192049100351e-06,
-      "logits/chosen": -2.411689043045044,
-      "logits/rejected": -2.384003162384033,
-      "logps/chosen": -242.8949737548828,
-      "logps/rejected": -236.754150390625,
-      "loss": 0.292,
-      "rewards/accuracies": 0.747499942779541,
-      "rewards/chosen": 0.005577466916292906,
-      "rewards/margins": 0.016019560396671295,
-      "rewards/rejected": -0.010442093946039677,
-      "step": 420
-    },
-    {
-      "epoch": 1.69,
-      "grad_norm": 110.0,
-      "learning_rate": 4.1819975740387406e-06,
-      "logits/chosen": -2.4044318199157715,
-      "logits/rejected": -2.3681979179382324,
-      "logps/chosen": -276.5671691894531,
-      "logps/rejected": -259.7832336425781,
-      "loss": 0.3016,
-      "rewards/accuracies": 0.7475000619888306,
-      "rewards/chosen": 0.011015561409294605,
-      "rewards/margins": 0.02647540345788002,
-      "rewards/rejected": -0.01545984111726284,
-      "step": 430
-    },
-    {
-      "epoch": 1.73,
-      "grad_norm": 158.0,
-      "learning_rate": 4.1305325093405045e-06,
-      "logits/chosen": -2.4186065196990967,
-      "logits/rejected": -2.406249523162842,
-      "logps/chosen": -295.4107971191406,
-      "logps/rejected": -280.9339294433594,
-      "loss": 0.4026,
-      "rewards/accuracies": 0.7675000429153442,
-      "rewards/chosen": 0.010993788950145245,
-      "rewards/margins": 0.020805999636650085,
-      "rewards/rejected": -0.00981221068650484,
-      "step": 440
-    },
-    {
-      "epoch": 1.77,
-      "grad_norm": 175.0,
-      "learning_rate": 4.077835731980775e-06,
-      "logits/chosen": -2.416654348373413,
-      "logits/rejected": -2.368619203567505,
-      "logps/chosen": -279.9720764160156,
-      "logps/rejected": -245.94345092773438,
-      "loss": 0.3414,
-      "rewards/accuracies": 0.7575000524520874,
-      "rewards/chosen": 0.006698101758956909,
-      "rewards/margins": 0.016826082020998,
-      "rewards/rejected": -0.010127981193363667,
-      "step": 450
-    },
-    {
-      "epoch": 1.81,
-      "grad_norm": 175.0,
-      "learning_rate": 4.02394704937677e-06,
-      "logits/chosen": -2.3919434547424316,
-      "logits/rejected": -2.3505940437316895,
-      "logps/chosen": -280.6643981933594,
-      "logps/rejected": -252.3192901611328,
-      "loss": 0.3603,
-      "rewards/accuracies": 0.7575000524520874,
-      "rewards/chosen": 0.00701780105009675,
-      "rewards/margins": 0.01795104146003723,
-      "rewards/rejected": -0.010933240875601768,
-      "step": 460
-    },
-    {
-      "epoch": 1.84,
-      "grad_norm": 176.0,
-      "learning_rate": 3.96890716931708e-06,
-      "logits/chosen": -2.381404399871826,
-      "logits/rejected": -2.369319438934326,
-      "logps/chosen": -251.976806640625,
-      "logps/rejected": -239.9644775390625,
-      "loss": 0.3975,
-      "rewards/accuracies": 0.7425000071525574,
-      "rewards/chosen": 0.00673043355345726,
-      "rewards/margins": 0.01468564011156559,
-      "rewards/rejected": -0.00795520469546318,
-      "step": 470
-    },
-    {
-      "epoch": 1.88,
-      "grad_norm": 129.0,
-      "learning_rate": 3.912757669210783e-06,
-      "logits/chosen": -2.4172403812408447,
-      "logits/rejected": -2.354468584060669,
-      "logps/chosen": -258.93780517578125,
-      "logps/rejected": -234.1327362060547,
-      "loss": 0.354,
-      "rewards/accuracies": 0.7425000071525574,
-      "rewards/chosen": 0.015040628612041473,
-      "rewards/margins": 0.020659491419792175,
-      "rewards/rejected": -0.005618864204734564,
-      "step": 480
-    },
-    {
-      "epoch": 1.92,
-      "grad_norm": 127.5,
-      "learning_rate": 3.855540964679658e-06,
-      "logits/chosen": -2.3677306175231934,
-      "logits/rejected": -2.323366641998291,
-      "logps/chosen": -239.04776000976562,
-      "logps/rejected": -228.19580078125,
-      "loss": 0.2687,
-      "rewards/accuracies": 0.7850000262260437,
-      "rewards/chosen": 0.0076850662007927895,
-      "rewards/margins": 0.01919684186577797,
-      "rewards/rejected": -0.01151177566498518,
-      "step": 490
-    },
-    {
-      "epoch": 1.96,
-      "grad_norm": 100.5,
-      "learning_rate": 3.797300277517212e-06,
-      "logits/chosen": -2.412917137145996,
-      "logits/rejected": -2.38498592376709,
-      "logps/chosen": -285.268310546875,
-      "logps/rejected": -264.0386962890625,
-      "loss": 0.292,
-      "rewards/accuracies": 0.7850000262260437,
-      "rewards/chosen": 0.010482062585651875,
-      "rewards/margins": 0.020365219563245773,
-      "rewards/rejected": -0.009883158840239048,
-      "step": 500
-    },
-    {
-      "epoch": 1.96,
-      "eval_logits/chosen": -2.4342868328094482,
-      "eval_logits/rejected": -2.3977575302124023,
-      "eval_logps/chosen": -267.17938232421875,
-      "eval_logps/rejected": -242.4461669921875,
-      "eval_loss": 0.8145859837532043,
-      "eval_rewards/accuracies": 0.6398809552192688,
-      "eval_rewards/chosen": 0.004992412868887186,
-      "eval_rewards/margins": 0.005504657980054617,
-      "eval_rewards/rejected": -0.0005122453439980745,
-      "eval_runtime": 123.1687,
-      "eval_samples_per_second": 16.238,
-      "eval_steps_per_second": 0.341,
-      "step": 500
-    },
-    {
-      "epoch": 2.0,
-      "grad_norm": 150.0,
-      "learning_rate": 3.7380796030387035e-06,
-      "logits/chosen": -2.4117255210876465,
-      "logits/rejected": -2.3580873012542725,
-      "logps/chosen": -288.262451171875,
-      "logps/rejected": -250.64486694335938,
-      "loss": 0.2919,
-      "rewards/accuracies": 0.7850000262260437,
-      "rewards/chosen": 0.012420935556292534,
-      "rewards/margins": 0.022166112437844276,
-      "rewards/rejected": -0.009745175018906593,
-      "step": 510
-    },
-    {
-      "epoch": 2.04,
-      "grad_norm": 72.0,
-      "learning_rate": 3.6779236768468647e-06,
-      "logits/chosen": -2.416080951690674,
-      "logits/rejected": -2.3825385570526123,
-      "logps/chosen": -266.83453369140625,
-      "logps/rejected": -257.94635009765625,
-      "loss": 0.0944,
-      "rewards/accuracies": 0.8500000238418579,
-      "rewards/chosen": 0.017391610890626907,
-      "rewards/margins": 0.030041953548789024,
-      "rewards/rejected": -0.012650340795516968,
-      "step": 520
-    },
-    {
-      "epoch": 2.08,
-      "grad_norm": 40.75,
-      "learning_rate": 3.6168779410383905e-06,
-      "logits/chosen": -2.4022631645202637,
-      "logits/rejected": -2.366995334625244,
-      "logps/chosen": -274.7615661621094,
-      "logps/rejected": -253.46835327148438,
-      "loss": 0.0944,
-      "rewards/accuracies": 0.824999988079071,
-      "rewards/chosen": 0.014151136390864849,
-      "rewards/margins": 0.029283767566084862,
-      "rewards/rejected": -0.015132628381252289,
-      "step": 530
-    },
-    {
-      "epoch": 2.12,
-      "grad_norm": 109.0,
-      "learning_rate": 3.554988509876747e-06,
-      "logits/chosen": -2.411635637283325,
-      "logits/rejected": -2.379657030105591,
-      "logps/chosen": -264.20758056640625,
-      "logps/rejected": -248.7351837158203,
-      "loss": 0.1176,
-      "rewards/accuracies": 0.8475000262260437,
-      "rewards/chosen": 0.01760762929916382,
-      "rewards/margins": 0.03199198096990585,
-      "rewards/rejected": -0.014384354464709759,
-      "step": 540
-    },
-    {
-      "epoch": 2.16,
-      "grad_norm": 162.0,
-      "learning_rate": 3.4923021349572183e-06,
-      "logits/chosen": -2.4204351902008057,
-      "logits/rejected": -2.342064619064331,
-      "logps/chosen": -293.338623046875,
-      "logps/rejected": -249.83984375,
-      "loss": 0.1199,
-      "rewards/accuracies": 0.8449999690055847,
-      "rewards/chosen": 0.016306212171912193,
-      "rewards/margins": 0.03437874838709831,
-      "rewards/rejected": -0.01807253621518612,
-      "step": 550
-    },
-    {
-      "epoch": 2.2,
-      "grad_norm": 156.0,
-      "learning_rate": 3.428866169890511e-06,
-      "logits/chosen": -2.4187042713165283,
-      "logits/rejected": -2.3788833618164062,
-      "logps/chosen": -280.169921875,
-      "logps/rejected": -266.49139404296875,
-      "loss": 0.1396,
-      "rewards/accuracies": 0.8550000190734863,
-      "rewards/chosen": 0.020975306630134583,
-      "rewards/margins": 0.0341680608689785,
-      "rewards/rejected": -0.013192756101489067,
-      "step": 560
-    },
-    {
-      "epoch": 2.24,
-      "grad_norm": 52.5,
-      "learning_rate": 3.3647285345315933e-06,
-      "logits/chosen": -2.426948308944702,
-      "logits/rejected": -2.3513236045837402,
-      "logps/chosen": -301.64703369140625,
-      "logps/rejected": -252.04910278320312,
-      "loss": 0.1179,
-      "rewards/accuracies": 0.8324999809265137,
-      "rewards/chosen": 0.02232900820672512,
-      "rewards/margins": 0.03726055473089218,
-      "rewards/rejected": -0.01493154652416706,
-      "step": 570
-    },
-    {
-      "epoch": 2.28,
-      "grad_norm": 63.0,
-      "learning_rate": 3.299937678780786e-06,
-      "logits/chosen": -2.3919901847839355,
-      "logits/rejected": -2.376873016357422,
-      "logps/chosen": -270.5113830566406,
-      "logps/rejected": -262.92120361328125,
-      "loss": 0.1103,
-      "rewards/accuracies": 0.8525000810623169,
-      "rewards/chosen": 0.01391147542744875,
-      "rewards/margins": 0.03057609498500824,
-      "rewards/rejected": -0.016664620488882065,
-      "step": 580
-    },
-    {
-      "epoch": 2.32,
-      "grad_norm": 38.25,
-      "learning_rate": 3.234542545984464e-06,
-      "logits/chosen": -2.3860366344451904,
-      "logits/rejected": -2.3532588481903076,
-      "logps/chosen": -279.0345764160156,
-      "logps/rejected": -268.3638000488281,
-      "loss": 0.113,
-      "rewards/accuracies": 0.8125,
-      "rewards/chosen": 0.013167209923267365,
-      "rewards/margins": 0.03231758996844292,
-      "rewards/rejected": -0.0191503819078207,
-      "step": 590
-    },
-    {
-      "epoch": 2.36,
-      "grad_norm": 95.5,
-      "learning_rate": 3.1685925359629928e-06,
-      "logits/chosen": -2.382845401763916,
-      "logits/rejected": -2.345613479614258,
-      "logps/chosen": -270.888427734375,
-      "logps/rejected": -262.33251953125,
-      "loss": 0.1355,
-      "rewards/accuracies": 0.8575000762939453,
-      "rewards/chosen": 0.01834903098642826,
-      "rewards/margins": 0.034947365522384644,
-      "rewards/rejected": -0.016598336398601532,
-      "step": 600
-    },
-    {
-      "epoch": 2.36,
-      "eval_logits/chosen": -2.4177558422088623,
-      "eval_logits/rejected": -2.3796002864837646,
-      "eval_logps/chosen": -267.20611572265625,
-      "eval_logps/rejected": -242.52117919921875,
-      "eval_loss": 0.9650812745094299,
-      "eval_rewards/accuracies": 0.6160714030265808,
-      "eval_rewards/chosen": 0.004725400358438492,
-      "eval_rewards/margins": 0.00598777923732996,
-      "eval_rewards/rejected": -0.0012623785296455026,
-      "eval_runtime": 123.1103,
-      "eval_samples_per_second": 16.246,
-      "eval_steps_per_second": 0.341,
-      "step": 600
-    },
-    {
-      "epoch": 2.39,
-      "grad_norm": 79.5,
-      "learning_rate": 3.102137467693858e-06,
-      "logits/chosen": -2.3922505378723145,
-      "logits/rejected": -2.3382246494293213,
-      "logps/chosen": -273.4150390625,
-      "logps/rejected": -258.9840393066406,
-      "loss": 0.252,
-      "rewards/accuracies": 0.8725000619888306,
-      "rewards/chosen": 0.01791740581393242,
-      "rewards/margins": 0.032240770757198334,
-      "rewards/rejected": -0.014323368668556213,
-      "step": 610
-    },
-    {
-      "epoch": 2.43,
-      "grad_norm": 120.0,
-      "learning_rate": 3.0352275416781465e-06,
-      "logits/chosen": -2.416335344314575,
-      "logits/rejected": -2.379333019256592,
-      "logps/chosen": -273.5201110839844,
-      "logps/rejected": -258.6203918457031,
-      "loss": 0.1567,
-      "rewards/accuracies": 0.8400000333786011,
-      "rewards/chosen": 0.02479313686490059,
-      "rewards/margins": 0.0343189537525177,
-      "rewards/rejected": -0.00952581875026226,
-      "step": 620
-    },
-    {
-      "epoch": 2.47,
-      "grad_norm": 106.0,
-      "learning_rate": 2.96791330201883e-06,
-      "logits/chosen": -2.421025514602661,
-      "logits/rejected": -2.3913843631744385,
-      "logps/chosen": -266.0569763183594,
-      "logps/rejected": -255.9671173095703,
-      "loss": 0.1255,
-      "rewards/accuracies": 0.8274999856948853,
-      "rewards/chosen": 0.021568376570940018,
-      "rewards/margins": 0.03504693880677223,
-      "rewards/rejected": -0.013478565029799938,
-      "step": 630
-    },
-    {
-      "epoch": 2.51,
-      "grad_norm": 94.5,
-      "learning_rate": 2.9002455982394946e-06,
-      "logits/chosen": -2.3834731578826904,
-      "logits/rejected": -2.3404629230499268,
-      "logps/chosen": -279.171630859375,
-      "logps/rejected": -251.27841186523438,
-      "loss": 0.1115,
-      "rewards/accuracies": 0.8674999475479126,
-      "rewards/chosen": 0.020785773172974586,
-      "rewards/margins": 0.03259057179093361,
-      "rewards/rejected": -0.011804800480604172,
-      "step": 640
-    },
-    {
-      "epoch": 2.55,
-      "grad_norm": 61.0,
-      "learning_rate": 2.832275546872339e-06,
-      "logits/chosen": -2.401367664337158,
-      "logits/rejected": -2.3691532611846924,
-      "logps/chosen": -261.18377685546875,
-      "logps/rejected": -267.6328125,
-      "loss": 0.0953,
-      "rewards/accuracies": 0.8850000500679016,
-      "rewards/chosen": 0.018077706918120384,
-      "rewards/margins": 0.03268102556467056,
-      "rewards/rejected": -0.014603319577872753,
-      "step": 650
-    },
-    {
-      "epoch": 2.59,
-      "grad_norm": 130.0,
-      "learning_rate": 2.7640544928444927e-06,
-      "logits/chosen": -2.418788194656372,
-      "logits/rejected": -2.3343942165374756,
-      "logps/chosen": -288.7831726074219,
-      "logps/rejected": -252.0687713623047,
-      "loss": 0.1191,
-      "rewards/accuracies": 0.8675000071525574,
-      "rewards/chosen": 0.020471712574362755,
-      "rewards/margins": 0.03600749000906944,
-      "rewards/rejected": -0.015535781159996986,
-      "step": 660
-    },
-    {
-      "epoch": 2.63,
-      "grad_norm": 86.0,
-      "learning_rate": 2.695633970691786e-06,
-      "logits/chosen": -2.3701846599578857,
-      "logits/rejected": -2.351933240890503,
-      "logps/chosen": -257.39752197265625,
-      "logps/rejected": -252.69287109375,
-      "loss": 0.0849,
-      "rewards/accuracies": 0.8550000190734863,
-      "rewards/chosen": 0.019604947417974472,
-      "rewards/margins": 0.031910307705402374,
-      "rewards/rejected": -0.012305359356105328,
-      "step": 670
-    },
-    {
-      "epoch": 2.67,
-      "grad_norm": 101.5,
-      "learning_rate": 2.6270656656293007e-06,
-      "logits/chosen": -2.394273281097412,
-      "logits/rejected": -2.348475694656372,
-      "logps/chosen": -264.9925537109375,
-      "logps/rejected": -248.74368286132812,
-      "loss": 0.0955,
-      "rewards/accuracies": 0.8574999570846558,
-      "rewards/chosen": 0.020610950887203217,
-      "rewards/margins": 0.03306712210178375,
-      "rewards/rejected": -0.012456170283257961,
-      "step": 680
-    },
-    {
-      "epoch": 2.71,
-      "grad_norm": 63.25,
-      "learning_rate": 2.558401374508089e-06,
-      "logits/chosen": -2.402439594268799,
-      "logits/rejected": -2.3409905433654785,
-      "logps/chosen": -276.2132873535156,
-      "logps/rejected": -251.0520782470703,
-      "loss": 0.0996,
-      "rewards/accuracies": 0.8650000691413879,
-      "rewards/chosen": 0.02115057222545147,
-      "rewards/margins": 0.030901487916707993,
-      "rewards/rejected": -0.009750919416546822,
-      "step": 690
-    },
-    {
-      "epoch": 2.75,
-      "grad_norm": 57.75,
-      "learning_rate": 2.4896929666875665e-06,
-      "logits/chosen": -2.4019179344177246,
-      "logits/rejected": -2.3663971424102783,
-      "logps/chosen": -274.6024475097656,
-      "logps/rejected": -264.2455139160156,
-      "loss": 0.1327,
-      "rewards/accuracies": 0.8574999570846558,
-      "rewards/chosen": 0.016096513718366623,
-      "rewards/margins": 0.03081604465842247,
-      "rewards/rejected": -0.014719529077410698,
-      "step": 700
-    },
-    {
-      "epoch": 2.75,
-      "eval_logits/chosen": -2.4065868854522705,
-      "eval_logits/rejected": -2.369014263153076,
-      "eval_logps/chosen": -267.2229919433594,
-      "eval_logps/rejected": -242.58834838867188,
-      "eval_loss": 0.9984952211380005,
-      "eval_rewards/accuracies": 0.6339285969734192,
-      "eval_rewards/chosen": 0.004556288011372089,
-      "eval_rewards/margins": 0.006490407045930624,
-      "eval_rewards/rejected": -0.001934119500219822,
-      "eval_runtime": 122.9942,
-      "eval_samples_per_second": 16.261,
-      "eval_steps_per_second": 0.341,
-      "step": 700
-    },
-    {
-      "epoch": 2.79,
-      "grad_norm": 104.0,
-      "learning_rate": 2.420992344853132e-06,
-      "logits/chosen": -2.4031834602355957,
-      "logits/rejected": -2.380056142807007,
-      "logps/chosen": -276.49700927734375,
-      "logps/rejected": -262.06341552734375,
-      "loss": 0.1394,
-      "rewards/accuracies": 0.8399999737739563,
-      "rewards/chosen": 0.019898083060979843,
-      "rewards/margins": 0.033728718757629395,
-      "rewards/rejected": -0.013830636627972126,
-      "step": 710
-    },
-    {
-      "epoch": 2.83,
-      "grad_norm": 111.0,
-      "learning_rate": 2.3523514058086093e-06,
-      "logits/chosen": -2.410182237625122,
-      "logits/rejected": -2.326798915863037,
-      "logps/chosen": -288.55609130859375,
-      "logps/rejected": -250.171630859375,
-      "loss": 0.1191,
-      "rewards/accuracies": 0.8675000071525574,
-      "rewards/chosen": 0.01868962123990059,
-      "rewards/margins": 0.02993825078010559,
-      "rewards/rejected": -0.011248626746237278,
-      "step": 720
-    },
-    {
-      "epoch": 2.87,
-      "grad_norm": 63.5,
-      "learning_rate": 2.2838220012731365e-06,
-      "logits/chosen": -2.3818917274475098,
-      "logits/rejected": -2.3685965538024902,
-      "logps/chosen": -270.9535217285156,
-      "logps/rejected": -267.3423767089844,
-      "loss": 0.1279,
-      "rewards/accuracies": 0.8825000524520874,
-      "rewards/chosen": 0.023929597809910774,
-      "rewards/margins": 0.041905276477336884,
-      "rewards/rejected": -0.01797567494213581,
-      "step": 730
-    },
-    {
-      "epoch": 2.9,
-      "grad_norm": 136.0,
-      "learning_rate": 2.2154558987121054e-06,
-      "logits/chosen": -2.3983840942382812,
-      "logits/rejected": -2.3515264987945557,
-      "logps/chosen": -274.8974609375,
-      "logps/rejected": -253.7775115966797,
-      "loss": 0.1044,
-      "rewards/accuracies": 0.8500000238418579,
-      "rewards/chosen": 0.025471847504377365,
-      "rewards/margins": 0.03200577199459076,
-      "rewards/rejected": -0.006533923093229532,
-      "step": 740
-    },
-    {
-      "epoch": 2.94,
-      "grad_norm": 208.0,
-      "learning_rate": 2.147304742231758e-06,
-      "logits/chosen": -2.3778913021087646,
-      "logits/rejected": -2.3485236167907715,
-      "logps/chosen": -254.48635864257812,
-      "logps/rejected": -267.0797119140625,
-      "loss": 0.1637,
-      "rewards/accuracies": 0.8550001382827759,
-      "rewards/chosen": 0.018020575866103172,
-      "rewards/margins": 0.030409198254346848,
-      "rewards/rejected": -0.012388622388243675,
-      "step": 750
-    },
-    {
-      "epoch": 2.98,
-      "grad_norm": 62.5,
-      "learning_rate": 2.0794200135669586e-06,
-      "logits/chosen": -2.399770498275757,
-      "logits/rejected": -2.364065647125244,
-      "logps/chosen": -277.4521484375,
-      "logps/rejected": -267.6507568359375,
-      "loss": 0.1466,
-      "rewards/accuracies": 0.8574999570846558,
-      "rewards/chosen": 0.021130980923771858,
-      "rewards/margins": 0.036186493933200836,
-      "rewards/rejected": -0.015055513009428978,
-      "step": 760
-    },
-    {
-      "epoch": 3.02,
-      "grad_norm": 43.25,
-      "learning_rate": 2.011852993191625e-06,
-      "logits/chosen": -2.3711681365966797,
-      "logits/rejected": -2.331266403198242,
-      "logps/chosen": -284.1138916015625,
-      "logps/rejected": -270.9895324707031,
-      "loss": 0.0563,
-      "rewards/accuracies": 0.9100000262260437,
-      "rewards/chosen": 0.02099769189953804,
-      "rewards/margins": 0.03560823202133179,
-      "rewards/rejected": -0.014610541984438896,
-      "step": 770
-    },
-    {
-      "epoch": 3.06,
-      "grad_norm": 91.5,
-      "learning_rate": 1.944654721581196e-06,
-      "logits/chosen": -2.3276844024658203,
-      "logits/rejected": -2.3009562492370605,
-      "logps/chosen": -260.98626708984375,
-      "logps/rejected": -247.24685668945312,
-      "loss": 0.0353,
-      "rewards/accuracies": 0.9300001263618469,
-      "rewards/chosen": 0.024031776934862137,
-      "rewards/margins": 0.04021488502621651,
-      "rewards/rejected": -0.01618310809135437,
-      "step": 780
-    },
-    {
-      "epoch": 3.1,
-      "grad_norm": 29.5,
-      "learning_rate": 1.877875960656394e-06,
-      "logits/chosen": -2.3537003993988037,
-      "logits/rejected": -2.3234972953796387,
-      "logps/chosen": -275.46453857421875,
-      "logps/rejected": -260.1392517089844,
-      "loss": 0.0298,
-      "rewards/accuracies": 0.9125000238418579,
-      "rewards/chosen": 0.021290091797709465,
-      "rewards/margins": 0.032937195152044296,
-      "rewards/rejected": -0.011647104285657406,
-      "step": 790
-    },
-    {
-      "epoch": 3.14,
-      "grad_norm": 42.5,
-      "learning_rate": 1.8115671554374067e-06,
-      "logits/chosen": -2.399651050567627,
-      "logits/rejected": -2.3859167098999023,
-      "logps/chosen": -268.92803955078125,
-      "logps/rejected": -275.9405822753906,
-      "loss": 0.0389,
-      "rewards/accuracies": 0.9375,
-      "rewards/chosen": 0.025255614891648293,
-      "rewards/margins": 0.04190623760223389,
-      "rewards/rejected": -0.016650624573230743,
-      "step": 800
-    },
-    {
-      "epoch": 3.14,
-      "eval_logits/chosen": -2.3946948051452637,
-      "eval_logits/rejected": -2.3562896251678467,
-      "eval_logps/chosen": -266.8748474121094,
-      "eval_logps/rejected": -242.36962890625,
-      "eval_loss": 0.8932417035102844,
-      "eval_rewards/accuracies": 0.6517857313156128,
-      "eval_rewards/chosen": 0.008037895895540714,
-      "eval_rewards/margins": 0.007784782908856869,
-      "eval_rewards/rejected": 0.00025311243371106684,
-      "eval_runtime": 123.1142,
-      "eval_samples_per_second": 16.245,
-      "eval_steps_per_second": 0.341,
-      "step": 800
-    },
-    {
-      "epoch": 3.18,
-      "grad_norm": 19.25,
-      "learning_rate": 1.7457783959374585e-06,
-      "logits/chosen": -2.404486894607544,
-      "logits/rejected": -2.3604061603546143,
-      "logps/chosen": -278.95855712890625,
-      "logps/rejected": -251.1371307373047,
-      "loss": 0.0348,
-      "rewards/accuracies": 0.9275000691413879,
-      "rewards/chosen": 0.025924455374479294,
-      "rewards/margins": 0.037114791572093964,
-      "rewards/rejected": -0.011190338991582394,
-      "step": 810
-    },
-    {
-      "epoch": 3.22,
-      "grad_norm": 32.25,
-      "learning_rate": 1.680559379324558e-06,
-      "logits/chosen": -2.390227794647217,
-      "logits/rejected": -2.3385822772979736,
-      "logps/chosen": -292.6279296875,
-      "logps/rejected": -254.8878173828125,
-      "loss": 0.0299,
-      "rewards/accuracies": 0.9099999666213989,
-      "rewards/chosen": 0.02545427717268467,
-      "rewards/margins": 0.03745580464601517,
-      "rewards/rejected": -0.012001526542007923,
-      "step": 820
-    },
-    {
-      "epoch": 3.26,
-      "grad_norm": 60.5,
-      "learning_rate": 1.6159593723800013e-06,
-      "logits/chosen": -2.4059481620788574,
-      "logits/rejected": -2.3442025184631348,
-      "logps/chosen": -264.06060791015625,
-      "logps/rejected": -246.68521118164062,
-      "loss": 0.0307,
-      "rewards/accuracies": 0.9125000238418579,
-      "rewards/chosen": 0.0275394506752491,
-      "rewards/margins": 0.036764778196811676,
-      "rewards/rejected": -0.00922533217817545,
-      "step": 830
-    },
-    {
-      "epoch": 3.3,
-      "grad_norm": 49.75,
-      "learning_rate": 1.5520271742819883e-06,
-      "logits/chosen": -2.389446973800659,
-      "logits/rejected": -2.3524551391601562,
-      "logps/chosen": -271.63641357421875,
-      "logps/rejected": -255.3404998779297,
-      "loss": 0.0301,
-      "rewards/accuracies": 0.877500057220459,
-      "rewards/chosen": 0.02547125145792961,
-      "rewards/margins": 0.03725877031683922,
-      "rewards/rejected": -0.01178752165287733,
-      "step": 840
-    },
-    {
-      "epoch": 3.34,
-      "grad_norm": 24.625,
-      "learning_rate": 1.4888110797424783e-06,
-      "logits/chosen": -2.4469919204711914,
-      "logits/rejected": -2.3791050910949707,
-      "logps/chosen": -315.74066162109375,
-      "logps/rejected": -277.9239501953125,
-      "loss": 0.0354,
-      "rewards/accuracies": 0.9025000333786011,
-      "rewards/chosen": 0.03072303533554077,
-      "rewards/margins": 0.048735830932855606,
-      "rewards/rejected": -0.018012793734669685,
-      "step": 850
-    },
-    {
-      "epoch": 3.38,
-      "grad_norm": 54.0,
-      "learning_rate": 1.4263588425251052e-06,
-      "logits/chosen": -2.4028658866882324,
-      "logits/rejected": -2.3509697914123535,
-      "logps/chosen": -289.78485107421875,
-      "logps/rejected": -251.5,
-      "loss": 0.0267,
-      "rewards/accuracies": 0.9375001192092896,
-      "rewards/chosen": 0.028129303827881813,
-      "rewards/margins": 0.04094386473298073,
-      "rewards/rejected": -0.01281456183642149,
-      "step": 860
-    },
-    {
-      "epoch": 3.42,
-      "grad_norm": 25.625,
-      "learning_rate": 1.3647176393717509e-06,
-      "logits/chosen": -2.4022791385650635,
-      "logits/rejected": -2.3641726970672607,
-      "logps/chosen": -278.85333251953125,
-      "logps/rejected": -270.0050964355469,
-      "loss": 0.0217,
-      "rewards/accuracies": 0.9199999570846558,
-      "rewards/chosen": 0.025644132867455482,
-      "rewards/margins": 0.03728828951716423,
-      "rewards/rejected": -0.011644158512353897,
-      "step": 870
-    },
-    {
-      "epoch": 3.45,
-      "grad_norm": 30.875,
-      "learning_rate": 1.303934034364983e-06,
-      "logits/chosen": -2.3777599334716797,
-      "logits/rejected": -2.3236684799194336,
-      "logps/chosen": -261.53216552734375,
-      "logps/rejected": -239.6273651123047,
-      "loss": 0.0216,
-      "rewards/accuracies": 0.9025000333786011,
-      "rewards/chosen": 0.024678941816091537,
-      "rewards/margins": 0.03661385923624039,
-      "rewards/rejected": -0.011934916488826275,
-      "step": 880
-    },
-    {
-      "epoch": 3.49,
-      "grad_norm": 87.0,
-      "learning_rate": 1.2440539437533075e-06,
-      "logits/chosen": -2.352806806564331,
-      "logits/rejected": -2.3354265689849854,
-      "logps/chosen": -269.5535583496094,
-      "logps/rejected": -268.3061218261719,
-      "loss": 0.0273,
-      "rewards/accuracies": 0.9300000071525574,
-      "rewards/chosen": 0.022415757179260254,
-      "rewards/margins": 0.036310791969299316,
-      "rewards/rejected": -0.013895031996071339,
-      "step": 890
-    },
-    {
-      "epoch": 3.53,
-      "grad_norm": 33.25,
-      "learning_rate": 1.1851226012658015e-06,
-      "logits/chosen": -2.366988182067871,
-      "logits/rejected": -2.323378562927246,
-      "logps/chosen": -264.4871520996094,
-      "logps/rejected": -251.3701934814453,
-      "loss": 0.029,
-      "rewards/accuracies": 0.9175000190734863,
-      "rewards/chosen": 0.023964881896972656,
-      "rewards/margins": 0.039505355060100555,
-      "rewards/rejected": -0.015540470369160175,
-      "step": 900
-    },
-    {
-      "epoch": 3.53,
-      "eval_logits/chosen": -2.411829948425293,
-      "eval_logits/rejected": -2.3752424716949463,
-      "eval_logps/chosen": -266.7797546386719,
-      "eval_logps/rejected": -242.3114013671875,
-      "eval_loss": 0.9391952157020569,
-      "eval_rewards/accuracies": 0.6577380895614624,
-      "eval_rewards/chosen": 0.008988723158836365,
-      "eval_rewards/margins": 0.008153370581567287,
-      "eval_rewards/rejected": 0.0008353526936843991,
-      "eval_runtime": 123.1579,
-      "eval_samples_per_second": 16.239,
-      "eval_steps_per_second": 0.341,
-      "step": 900
-    },
-    {
-      "epoch": 3.57,
-      "grad_norm": 16.5,
-      "learning_rate": 1.1271845239423196e-06,
-      "logits/chosen": -2.4022092819213867,
-      "logits/rejected": -2.357339382171631,
-      "logps/chosen": -289.2544250488281,
-      "logps/rejected": -258.82379150390625,
-      "loss": 0.0269,
-      "rewards/accuracies": 0.9175000190734863,
-      "rewards/chosen": 0.022589916363358498,
-      "rewards/margins": 0.03589435666799545,
-      "rewards/rejected": -0.013304440304636955,
-      "step": 910
-    },
-    {
-      "epoch": 3.61,
-      "grad_norm": 24.25,
-      "learning_rate": 1.0702834785050893e-06,
-      "logits/chosen": -2.3661084175109863,
-      "logits/rejected": -2.3483455181121826,
-      "logps/chosen": -276.6420593261719,
-      "logps/rejected": -278.82080078125,
-      "loss": 0.033,
-      "rewards/accuracies": 0.9250000715255737,
-      "rewards/chosen": 0.021236615255475044,
-      "rewards/margins": 0.038433950394392014,
-      "rewards/rejected": -0.017197338864207268,
-      "step": 920
-    },
-    {
-      "epoch": 3.65,
-      "grad_norm": 40.5,
-      "learning_rate": 1.0144624482971082e-06,
-      "logits/chosen": -2.4388625621795654,
-      "logits/rejected": -2.380392551422119,
-      "logps/chosen": -271.2420349121094,
-      "logps/rejected": -256.2777404785156,
-      "loss": 0.0275,
-      "rewards/accuracies": 0.9350000619888306,
-      "rewards/chosen": 0.022215455770492554,
-      "rewards/margins": 0.03798101097345352,
-      "rewards/rejected": -0.01576555334031582,
-      "step": 930
-    },
-    {
-      "epoch": 3.69,
-      "grad_norm": 14.0,
-      "learning_rate": 9.597636008123052e-07,
-      "logits/chosen": -2.4123058319091797,
-      "logits/rejected": -2.3628923892974854,
-      "logps/chosen": -308.3174743652344,
-      "logps/rejected": -278.7733459472656,
-      "loss": 0.0278,
-      "rewards/accuracies": 0.942500114440918,
-      "rewards/chosen": 0.024744439870119095,
-      "rewards/margins": 0.038902923464775085,
-      "rewards/rejected": -0.014158482663333416,
-      "step": 940
-    },
-    {
-      "epoch": 3.73,
-      "grad_norm": 20.375,
-      "learning_rate": 9.06228255841991e-07,
-      "logits/chosen": -2.382359504699707,
-      "logits/rejected": -2.3492817878723145,
-      "logps/chosen": -264.2523193359375,
-      "logps/rejected": -256.5533752441406,
-      "loss": 0.0279,
-      "rewards/accuracies": 0.8999999761581421,
-      "rewards/chosen": 0.022538714110851288,
-      "rewards/margins": 0.03617415204644203,
-      "rewards/rejected": -0.013635434210300446,
-      "step": 950
-    },
-    {
-      "epoch": 3.77,
-      "grad_norm": 36.25,
-      "learning_rate": 8.538968542616846e-07,
-      "logits/chosen": -2.41325044631958,
-      "logits/rejected": -2.375253677368164,
-      "logps/chosen": -281.48248291015625,
-      "logps/rejected": -265.0972900390625,
-      "loss": 0.0179,
-      "rewards/accuracies": 0.9075000882148743,
-      "rewards/chosen": 0.025317683815956116,
-      "rewards/margins": 0.038605697453022,
-      "rewards/rejected": -0.013288016431033611,
-      "step": 960
-    },
-    {
-      "epoch": 3.81,
-      "grad_norm": 29.625,
-      "learning_rate": 8.028089274818624e-07,
-      "logits/chosen": -2.4157140254974365,
-      "logits/rejected": -2.377472162246704,
-      "logps/chosen": -278.72564697265625,
-      "logps/rejected": -257.7362365722656,
-      "loss": 0.034,
-      "rewards/accuracies": 0.9274999499320984,
-      "rewards/chosen": 0.023796474561095238,
-      "rewards/margins": 0.03968465328216553,
-      "rewards/rejected": -0.01588817685842514,
-      "step": 970
-    },
-    {
-      "epoch": 3.85,
-      "grad_norm": 24.5,
-      "learning_rate": 7.530030675857252e-07,
-      "logits/chosen": -2.371452569961548,
-      "logits/rejected": -2.33616304397583,
-      "logps/chosen": -278.2137756347656,
-      "logps/rejected": -254.2333984375,
-      "loss": 0.0245,
-      "rewards/accuracies": 0.9350000619888306,
-      "rewards/chosen": 0.028386935591697693,
-      "rewards/margins": 0.044838014990091324,
-      "rewards/rejected": -0.01645107939839363,
-      "step": 980
-    },
-    {
-      "epoch": 3.89,
-      "grad_norm": 31.5,
-      "learning_rate": 7.045168981765427e-07,
-      "logits/chosen": -2.4061717987060547,
-      "logits/rejected": -2.367501735687256,
-      "logps/chosen": -277.0457763671875,
-      "logps/rejected": -248.8456268310547,
-      "loss": 0.0248,
-      "rewards/accuracies": 0.9325000047683716,
-      "rewards/chosen": 0.025927498936653137,
-      "rewards/margins": 0.03844950348138809,
-      "rewards/rejected": -0.012522002682089806,
-      "step": 990
-    },
-    {
-      "epoch": 3.93,
-      "grad_norm": 12.875,
-      "learning_rate": 6.573870459565907e-07,
-      "logits/chosen": -2.381437301635742,
-      "logits/rejected": -2.3325188159942627,
-      "logps/chosen": -293.3424377441406,
-      "logps/rejected": -262.5635681152344,
-      "loss": 0.0198,
-      "rewards/accuracies": 0.9225000143051147,
-      "rewards/chosen": 0.02644011378288269,
-      "rewards/margins": 0.04168093949556351,
-      "rewards/rejected": -0.015240825712680817,
-      "step": 1000
-    },
-    {
-      "epoch": 3.93,
-      "eval_logits/chosen": -2.4145004749298096,
-      "eval_logits/rejected": -2.3780155181884766,
-      "eval_logps/chosen": -266.8046875,
-      "eval_logps/rejected": -242.29171752929688,
-      "eval_loss": 0.820038914680481,
-      "eval_rewards/accuracies": 0.6577380895614624,
-      "eval_rewards/chosen": 0.008739516139030457,
-      "eval_rewards/margins": 0.007707077078521252,
-      "eval_rewards/rejected": 0.0010324395261704922,
-      "eval_runtime": 123.2067,
-      "eval_samples_per_second": 16.233,
-      "eval_steps_per_second": 0.341,
-      "step": 1000
-    },
-    {
-      "epoch": 3.96,
-      "grad_norm": 33.75,
-      "learning_rate": 6.116491130591478e-07,
-      "logits/chosen": -2.410226821899414,
-      "logits/rejected": -2.366502285003662,
-      "logps/chosen": -279.39166259765625,
-      "logps/rejected": -251.89364624023438,
-      "loss": 0.0233,
-      "rewards/accuracies": 0.9175001382827759,
-      "rewards/chosen": 0.024083226919174194,
-      "rewards/margins": 0.03586304560303688,
-      "rewards/rejected": -0.011779818683862686,
-      "step": 1010
-    },
-    {
-      "epoch": 4.0,
-      "grad_norm": 3.4375,
-      "learning_rate": 5.673376501544641e-07,
-      "logits/chosen": -2.41102933883667,
-      "logits/rejected": -2.36838960647583,
-      "logps/chosen": -274.3736877441406,
-      "logps/rejected": -243.4824981689453,
-      "loss": 0.017,
-      "rewards/accuracies": 0.9175000190734863,
-      "rewards/chosen": 0.023644987493753433,
-      "rewards/margins": 0.0368778295814991,
-      "rewards/rejected": -0.013232842087745667,
-      "step": 1020
-    },
-    {
-      "epoch": 4.04,
-      "grad_norm": 20.5,
-      "learning_rate": 5.244861303500026e-07,
-      "logits/chosen": -2.413541793823242,
-      "logits/rejected": -2.362837314605713,
-      "logps/chosen": -272.99359130859375,
-      "logps/rejected": -239.32608032226562,
-      "loss": 0.008,
-      "rewards/accuracies": 0.9325000643730164,
-      "rewards/chosen": 0.021824661642313004,
-      "rewards/margins": 0.0353974774479866,
-      "rewards/rejected": -0.013572819530963898,
-      "step": 1030
-    },
-    {
-      "epoch": 4.08,
-      "grad_norm": 3.765625,
-      "learning_rate": 4.831269239046851e-07,
-      "logits/chosen": -2.4089815616607666,
-      "logits/rejected": -2.366555690765381,
-      "logps/chosen": -266.49798583984375,
-      "logps/rejected": -252.52145385742188,
-      "loss": 0.0043,
-      "rewards/accuracies": 0.949999988079071,
-      "rewards/chosen": 0.025229623541235924,
-      "rewards/margins": 0.03910643607378006,
-      "rewards/rejected": -0.013876812532544136,
-      "step": 1040
-    },
-    {
-      "epoch": 4.12,
-      "grad_norm": 18.75,
-      "learning_rate": 4.4329127377623127e-07,
-      "logits/chosen": -2.380474328994751,
-      "logits/rejected": -2.3557381629943848,
-      "logps/chosen": -275.6029968261719,
-      "logps/rejected": -260.39776611328125,
-      "loss": 0.0054,
-      "rewards/accuracies": 0.9550000429153442,
-      "rewards/chosen": 0.025818094611167908,
-      "rewards/margins": 0.038945622742176056,
-      "rewards/rejected": -0.013127523474395275,
-      "step": 1050
-    },
-    {
-      "epoch": 4.16,
-      "grad_norm": 14.8125,
-      "learning_rate": 4.050092720200638e-07,
-      "logits/chosen": -2.384019374847412,
-      "logits/rejected": -2.338085889816284,
-      "logps/chosen": -280.4701232910156,
-      "logps/rejected": -250.2019805908203,
-      "loss": 0.0052,
-      "rewards/accuracies": 0.9575001001358032,
-      "rewards/chosen": 0.028378132730722427,
-      "rewards/margins": 0.042529620230197906,
-      "rewards/rejected": -0.014151493087410927,
-      "step": 1060
-    },
-    {
-      "epoch": 4.2,
-      "grad_norm": 8.375,
-      "learning_rate": 3.683098370576196e-07,
-      "logits/chosen": -2.406979560852051,
-      "logits/rejected": -2.3714377880096436,
-      "logps/chosen": -287.4407958984375,
-      "logps/rejected": -259.043212890625,
-      "loss": 0.0062,
-      "rewards/accuracies": 0.9449999928474426,
-      "rewards/chosen": 0.024499880149960518,
-      "rewards/margins": 0.03765181452035904,
-      "rewards/rejected": -0.013151939027011395,
-      "step": 1070
-    },
-    {
-      "epoch": 4.24,
-      "grad_norm": 11.6875,
-      "learning_rate": 3.3322069183122253e-07,
-      "logits/chosen": -2.415555477142334,
-      "logits/rejected": -2.3626112937927246,
-      "logps/chosen": -270.87249755859375,
-      "logps/rejected": -251.2595977783203,
-      "loss": 0.0049,
-      "rewards/accuracies": 0.9399999380111694,
-      "rewards/chosen": 0.028504956513643265,
-      "rewards/margins": 0.043677303940057755,
-      "rewards/rejected": -0.015172350220382214,
-      "step": 1080
-    },
-    {
-      "epoch": 4.28,
-      "grad_norm": 60.75,
-      "learning_rate": 2.997683428620296e-07,
-      "logits/chosen": -2.4119620323181152,
-      "logits/rejected": -2.3438758850097656,
-      "logps/chosen": -288.2101745605469,
-      "logps/rejected": -261.5445251464844,
-      "loss": 0.0101,
-      "rewards/accuracies": 0.9550000429153442,
-      "rewards/chosen": 0.026679161936044693,
-      "rewards/margins": 0.042851291596889496,
-      "rewards/rejected": -0.016172129660844803,
-      "step": 1090
-    },
-    {
-      "epoch": 4.32,
-      "grad_norm": 23.25,
-      "learning_rate": 2.6797806022686835e-07,
-      "logits/chosen": -2.3794476985931396,
-      "logits/rejected": -2.350893974304199,
-      "logps/chosen": -262.9915771484375,
-      "logps/rejected": -261.4454650878906,
-      "loss": 0.0059,
-      "rewards/accuracies": 0.9449998736381531,
-      "rewards/chosen": 0.024703029543161392,
-      "rewards/margins": 0.04311930388212204,
-      "rewards/rejected": -0.018416276201605797,
-      "step": 1100
-    },
-    {
-      "epoch": 4.32,
-      "eval_logits/chosen": -2.410806894302368,
-      "eval_logits/rejected": -2.3743793964385986,
-      "eval_logps/chosen": -266.8759765625,
-      "eval_logps/rejected": -242.3739471435547,
-      "eval_loss": 0.8903548717498779,
-      "eval_rewards/accuracies": 0.6577380895614624,
-      "eval_rewards/chosen": 0.008026321418583393,
-      "eval_rewards/margins": 0.007816384546458721,
-      "eval_rewards/rejected": 0.00020993576617911458,
-      "eval_runtime": 123.126,
-      "eval_samples_per_second": 16.244,
-      "eval_steps_per_second": 0.341,
-      "step": 1100
-    },
-    {
-      "epoch": 4.36,
-      "grad_norm": 33.5,
-      "learning_rate": 2.378738584690926e-07,
-      "logits/chosen": -2.3809187412261963,
-      "logits/rejected": -2.345142126083374,
-      "logps/chosen": -274.513916015625,
-      "logps/rejected": -259.05767822265625,
-      "loss": 0.0073,
-      "rewards/accuracies": 0.9474999308586121,
-      "rewards/chosen": 0.024536920711398125,
-      "rewards/margins": 0.0396781824529171,
-      "rewards/rejected": -0.015141261741518974,
-      "step": 1110
-    },
-    {
-      "epoch": 4.4,
-      "grad_norm": 6.09375,
-      "learning_rate": 2.0947847845787073e-07,
-      "logits/chosen": -2.380807638168335,
-      "logits/rejected": -2.3750548362731934,
-      "logps/chosen": -273.5718688964844,
-      "logps/rejected": -271.99774169921875,
-      "loss": 0.0056,
-      "rewards/accuracies": 0.9699999690055847,
-      "rewards/chosen": 0.025182534009218216,
-      "rewards/margins": 0.04230727255344391,
-      "rewards/rejected": -0.017124736681580544,
-      "step": 1120
-    },
-    {
-      "epoch": 4.44,
-      "grad_norm": 12.75,
-      "learning_rate": 1.828133702096152e-07,
-      "logits/chosen": -2.4007954597473145,
-      "logits/rejected": -2.3510959148406982,
-      "logps/chosen": -297.80023193359375,
-      "logps/rejected": -267.09271240234375,
-      "loss": 0.0058,
-      "rewards/accuracies": 0.9350000619888306,
-      "rewards/chosen": 0.027246862649917603,
-      "rewards/margins": 0.04566134512424469,
-      "rewards/rejected": -0.01841447874903679,
-      "step": 1130
-    },
-    {
-      "epoch": 4.47,
-      "grad_norm": 6.8125,
-      "learning_rate": 1.5789867668453224e-07,
-      "logits/chosen": -2.359222650527954,
-      "logits/rejected": -2.316779613494873,
-      "logps/chosen": -256.35321044921875,
-      "logps/rejected": -244.2194366455078,
-      "loss": 0.005,
-      "rewards/accuracies": 0.940000057220459,
-      "rewards/chosen": 0.02366521954536438,
-      "rewards/margins": 0.03908833488821983,
-      "rewards/rejected": -0.015423113480210304,
-      "step": 1140
-    },
-    {
-      "epoch": 4.51,
-      "grad_norm": 10.625,
-      "learning_rate": 1.3475321857052387e-07,
-      "logits/chosen": -2.3972084522247314,
-      "logits/rejected": -2.3675570487976074,
-      "logps/chosen": -278.0527038574219,
-      "logps/rejected": -256.3372497558594,
-      "loss": 0.0053,
-      "rewards/accuracies": 0.9575001001358032,
-      "rewards/chosen": 0.025035608559846878,
-      "rewards/margins": 0.039394162595272064,
-      "rewards/rejected": -0.014358552172780037,
-      "step": 1150
-    },
-    {
-      "epoch": 4.55,
-      "grad_norm": 24.5,
-      "learning_rate": 1.1339448006594284e-07,
-      "logits/chosen": -2.3742661476135254,
-      "logits/rejected": -2.3620691299438477,
-      "logps/chosen": -269.88934326171875,
-      "logps/rejected": -263.372314453125,
-      "loss": 0.0056,
-      "rewards/accuracies": 0.9574999809265137,
-      "rewards/chosen": 0.02747185155749321,
-      "rewards/margins": 0.04529104381799698,
-      "rewards/rejected": -0.01781919226050377,
-      "step": 1160
-    },
-    {
-      "epoch": 4.59,
-      "grad_norm": 15.125,
-      "learning_rate": 9.383859567194148e-08,
-      "logits/chosen": -2.4092886447906494,
-      "logits/rejected": -2.3820230960845947,
-      "logps/chosen": -292.05755615234375,
-      "logps/rejected": -274.00262451171875,
-      "loss": 0.0076,
-      "rewards/accuracies": 0.9600000381469727,
-      "rewards/chosen": 0.02904806099832058,
-      "rewards/margins": 0.04705999046564102,
-      "rewards/rejected": -0.018011927604675293,
-      "step": 1170
-    },
-    {
-      "epoch": 4.63,
-      "grad_norm": 6.125,
-      "learning_rate": 7.610033800438343e-08,
-      "logits/chosen": -2.404353141784668,
-      "logits/rejected": -2.3535656929016113,
-      "logps/chosen": -279.27001953125,
-      "logps/rejected": -261.3420104980469,
-      "loss": 0.0036,
-      "rewards/accuracies": 0.9624999761581421,
-      "rewards/chosen": 0.023024918511509895,
-      "rewards/margins": 0.038382213562726974,
-      "rewards/rejected": -0.015357298776507378,
-      "step": 1180
-    },
-    {
-      "epoch": 4.67,
-      "grad_norm": 15.3125,
-      "learning_rate": 6.019310663453654e-08,
-      "logits/chosen": -2.379361867904663,
-      "logits/rejected": -2.348564624786377,
-      "logps/chosen": -272.57965087890625,
-      "logps/rejected": -280.6869201660156,
-      "loss": 0.0045,
-      "rewards/accuracies": 0.9474999308586121,
-      "rewards/chosen": 0.02704106830060482,
-      "rewards/margins": 0.04761399328708649,
-      "rewards/rejected": -0.02057291939854622,
-      "step": 1190
-    },
-    {
-      "epoch": 4.71,
-      "grad_norm": 6.5625,
-      "learning_rate": 4.6128917966964394e-08,
-      "logits/chosen": -2.3975164890289307,
-      "logits/rejected": -2.354431390762329,
-      "logps/chosen": -264.81683349609375,
-      "logps/rejected": -240.6919403076172,
-      "loss": 0.0042,
-      "rewards/accuracies": 0.9375,
-      "rewards/chosen": 0.023663988336920738,
-      "rewards/margins": 0.037333834916353226,
-      "rewards/rejected": -0.013669842854142189,
-      "step": 1200
-    },
-    {
-      "epoch": 4.71,
-      "eval_logits/chosen": -2.4119250774383545,
-      "eval_logits/rejected": -2.375300407409668,
-      "eval_logps/chosen": -266.87713623046875,
-      "eval_logps/rejected": -242.38916015625,
-      "eval_loss": 0.8778771162033081,
-      "eval_rewards/accuracies": 0.6517857313156128,
-      "eval_rewards/chosen": 0.00801478698849678,
-      "eval_rewards/margins": 0.007957086898386478,
-      "eval_rewards/rejected": 5.7699922763276845e-05,
-      "eval_runtime": 123.014,
-      "eval_samples_per_second": 16.258,
-      "eval_steps_per_second": 0.341,
-      "step": 1200
-    },
-    {
-      "epoch": 4.75,
-      "grad_norm": 27.0,
-      "learning_rate": 3.3918396162275214e-08,
-      "logits/chosen": -2.429567337036133,
-      "logits/rejected": -2.401862621307373,
-      "logps/chosen": -265.3700866699219,
-      "logps/rejected": -255.75082397460938,
-      "loss": 0.0077,
-      "rewards/accuracies": 0.9475001096725464,
-      "rewards/chosen": 0.02286478877067566,
-      "rewards/margins": 0.037350136786699295,
-      "rewards/rejected": -0.01448534894734621,
-      "step": 1210
-    },
-    {
-      "epoch": 4.79,
-      "grad_norm": 29.625,
-      "learning_rate": 2.3570765111574357e-08,
-      "logits/chosen": -2.420926570892334,
-      "logits/rejected": -2.3869175910949707,
-      "logps/chosen": -275.02593994140625,
-      "logps/rejected": -250.9207000732422,
-      "loss": 0.0076,
-      "rewards/accuracies": 0.9325000643730164,
-      "rewards/chosen": 0.025391753762960434,
-      "rewards/margins": 0.04114841669797897,
-      "rewards/rejected": -0.01575666293501854,
-      "step": 1220
-    },
-    {
-      "epoch": 4.83,
-      "grad_norm": 12.8125,
-      "learning_rate": 1.5093841468690473e-08,
-      "logits/chosen": -2.378108501434326,
-      "logits/rejected": -2.3375566005706787,
-      "logps/chosen": -278.16180419921875,
-      "logps/rejected": -249.80557250976562,
-      "loss": 0.0043,
-      "rewards/accuracies": 0.9550000429153442,
-      "rewards/chosen": 0.02722669579088688,
-      "rewards/margins": 0.042906779795885086,
-      "rewards/rejected": -0.015680085867643356,
-      "step": 1230
-    },
-    {
-      "epoch": 4.87,
-      "grad_norm": 6.65625,
-      "learning_rate": 8.494028745434368e-09,
-      "logits/chosen": -2.4071974754333496,
-      "logits/rejected": -2.3638107776641846,
-      "logps/chosen": -272.5426940917969,
-      "logps/rejected": -254.20095825195312,
-      "loss": 0.0031,
-      "rewards/accuracies": 0.9424999952316284,
-      "rewards/chosen": 0.024697447195649147,
-      "rewards/margins": 0.0481376014649868,
-      "rewards/rejected": -0.023440156131982803,
-      "step": 1240
-    },
-    {
-      "epoch": 4.91,
-      "grad_norm": 13.625,
-      "learning_rate": 3.776312474353394e-09,
-      "logits/chosen": -2.382949113845825,
-      "logits/rejected": -2.3320465087890625,
-      "logps/chosen": -262.14154052734375,
-      "logps/rejected": -247.5254669189453,
-      "loss": 0.0039,
-      "rewards/accuracies": 0.9550000429153442,
-      "rewards/chosen": 0.022636910900473595,
-      "rewards/margins": 0.03510580584406853,
-      "rewards/rejected": -0.012468894943594933,
-      "step": 1250
-    },
-    {
-      "epoch": 4.95,
-      "grad_norm": 6.78125,
-      "learning_rate": 9.442564426342949e-10,
-      "logits/chosen": -2.3739681243896484,
-      "logits/rejected": -2.3734383583068848,
-      "logps/chosen": -248.2006072998047,
-      "logps/rejected": -255.94161987304688,
-      "loss": 0.0033,
-      "rewards/accuracies": 0.9275000691413879,
-      "rewards/chosen": 0.025133823975920677,
-      "rewards/margins": 0.04240426793694496,
-      "rewards/rejected": -0.017270449548959732,
-      "step": 1260
-    },
-    {
-      "epoch": 4.99,
-      "grad_norm": 13.1875,
-      "learning_rate": 0.0,
-      "logits/chosen": -2.393690586090088,
-      "logits/rejected": -2.3561182022094727,
-      "logps/chosen": -263.6891174316406,
-      "logps/rejected": -243.8829345703125,
-      "loss": 0.0063,
-      "rewards/accuracies": 0.949999988079071,
-      "rewards/chosen": 0.024886978790163994,
-      "rewards/margins": 0.0397338829934597,
-      "rewards/rejected": -0.014846903271973133,
-      "step": 1270
-    },
-    {
-      "epoch": 4.99,
-      "step": 1270,
-      "total_flos": 0.0,
-      "train_loss": 0.164279118501603,
-      "train_runtime": 43545.4617,
-      "train_samples_per_second": 7.02,
-      "train_steps_per_second": 0.029
-    }
-  ],
-  "logging_steps": 10,
-  "max_steps": 1270,
+  "log_history": [],
+  "logging_steps": 500,
+  "max_steps": 0,
   "num_input_tokens_seen": 0,
-  "num_train_epochs": 5,
-  "save_steps": 100,
-  "total_flos": 0.0,
-  "train_batch_size": 10,
+  "num_train_epochs": 0,
+  "save_steps": 500,
+  "total_flos": 0,
+  "train_batch_size": null,
   "trial_name": null,
   "trial_params": null
 }

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3d774fcd77a3ac8f054a155d242f866a208af24764821461a89749487a46613a
+oid sha256:92fee4626f14c7e9e46522e5b22431404ce3eaa07753dde47b190bdd90f9420d
 size 5112