Training in progress, step 23000, checkpoint

checkpoint-23000/model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:de2b1d2a6797a5b840da642d6b9bcb645578127b823147ba58a2b58b1816240b
+oid sha256:ae7b44924427168fed2c3f405a09d1323aac5b116746113da926c2cfbe9b15db
 size 1711848436

checkpoint-23000/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3ebc81e6b46a4b9f15c3ae4ee1e55624008b67babb0c50a1a0bf3abd0331d141
+oid sha256:1006d855237654104542906eb19fd23990e908858ff26d8a12ee8ca5b62c2bff
 size 3424043887

checkpoint-23000/rng_state.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:083b40d2763349ea43639ef6beaf7ecb1c9bbf079e2cab234bf89f6a0aa0ec40
+oid sha256:808a643a84510ed5faf1ca887880b5fa202c1a5fa5442b2da6283dd4ccf04802
 size 14503

checkpoint-23000/scheduler.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2cf46d2c83cb9e3aa75ffc4be717b14490e93d2b37cc7dfb5313e19a3240d9df
+oid sha256:777f5fb2b300ed6e9b68e9391b120c977f6ae587a79dd5ac1a93c72f10907d4a
 size 623

checkpoint-23000/trainer_state.json

@@ -1,7 +1,7 @@
 {
-  "best_metric": 0.17669138312339783,
-  "best_model_checkpoint": "./output/clip-finetuned-csu-p14-336-e4l57-l/checkpoint-22500",
-  "epoch": 3.851951096968682,
+  "best_metric": 0.48340705037117004,
+  "best_model_checkpoint": "./output/clip-finetuned-csu-p14-336-e4l57-l/checkpoint-23000",
+  "epoch": 2.453071672354949,
   "eval_steps": 500,
   "global_step": 23000,
   "is_hyper_param_search": false,
@@ -9,700 +9,700 @@
   "is_world_process_zero": true,
   "log_history": [
     {
-      "epoch": 0.09211495946941783,
-      "grad_norm": 3.0293657779693604,
-      "learning_rate": 4.884856300663227e-07,
-      "loss": 0.271,
+      "epoch": 0.05332764505119454,
+      "grad_norm": 414.2552795410156,
+      "learning_rate": 4.911120591581342e-07,
+      "loss": 0.3809,
       "step": 500
     },
     {
-      "epoch": 0.09211495946941783,
-      "eval_loss": 1.069305419921875,
-      "eval_runtime": 75.1409,
-      "eval_samples_per_second": 16.063,
-      "eval_steps_per_second": 2.01,
+      "epoch": 0.05332764505119454,
+      "eval_loss": 1.1109352111816406,
+      "eval_runtime": 63.6734,
+      "eval_samples_per_second": 15.501,
+      "eval_steps_per_second": 1.947,
       "step": 500
     },
     {
-      "epoch": 0.18422991893883567,
-      "grad_norm": 7.3040852546691895,
-      "learning_rate": 4.769712601326456e-07,
-      "loss": 0.2493,
+      "epoch": 0.10665529010238908,
+      "grad_norm": 45.880027770996094,
+      "learning_rate": 4.822241183162685e-07,
+      "loss": 0.2649,
       "step": 1000
     },
     {
-      "epoch": 0.18422991893883567,
-      "eval_loss": 0.9427077770233154,
-      "eval_runtime": 74.8926,
-      "eval_samples_per_second": 16.116,
-      "eval_steps_per_second": 2.016,
+      "epoch": 0.10665529010238908,
+      "eval_loss": 0.9588962197303772,
+      "eval_runtime": 62.6825,
+      "eval_samples_per_second": 15.746,
+      "eval_steps_per_second": 1.978,
       "step": 1000
     },
     {
-      "epoch": 0.2763448784082535,
-      "grad_norm": 0.020736657083034515,
-      "learning_rate": 4.654568901989683e-07,
-      "loss": 0.2348,
+      "epoch": 0.1599829351535836,
+      "grad_norm": 432.9915466308594,
+      "learning_rate": 4.733361774744027e-07,
+      "loss": 0.2104,
       "step": 1500
     },
     {
-      "epoch": 0.2763448784082535,
-      "eval_loss": 0.8726764917373657,
-      "eval_runtime": 76.6621,
-      "eval_samples_per_second": 15.744,
-      "eval_steps_per_second": 1.97,
+      "epoch": 0.1599829351535836,
+      "eval_loss": 0.90110844373703,
+      "eval_runtime": 62.6757,
+      "eval_samples_per_second": 15.748,
+      "eval_steps_per_second": 1.978,
       "step": 1500
     },
     {
-      "epoch": 0.36845983787767134,
-      "grad_norm": 5.3738909628009424e-05,
-      "learning_rate": 4.5394252026529107e-07,
-      "loss": 0.1552,
+      "epoch": 0.21331058020477817,
+      "grad_norm": 0.014208819717168808,
+      "learning_rate": 4.6444823663253695e-07,
+      "loss": 0.1849,
       "step": 2000
     },
     {
-      "epoch": 0.36845983787767134,
-      "eval_loss": 0.8325753808021545,
-      "eval_runtime": 76.6127,
-      "eval_samples_per_second": 15.755,
-      "eval_steps_per_second": 1.971,
+      "epoch": 0.21331058020477817,
+      "eval_loss": 0.8570474982261658,
+      "eval_runtime": 63.7296,
+      "eval_samples_per_second": 15.487,
+      "eval_steps_per_second": 1.946,
       "step": 2000
     },
     {
-      "epoch": 0.46057479734708917,
-      "grad_norm": 404.510009765625,
-      "learning_rate": 4.4242815033161386e-07,
-      "loss": 0.1753,
+      "epoch": 0.2666382252559727,
+      "grad_norm": 174.73324584960938,
+      "learning_rate": 4.5556029579067116e-07,
+      "loss": 0.2056,
       "step": 2500
     },
     {
-      "epoch": 0.46057479734708917,
-      "eval_loss": 0.7550467848777771,
-      "eval_runtime": 76.8024,
-      "eval_samples_per_second": 15.716,
-      "eval_steps_per_second": 1.966,
+      "epoch": 0.2666382252559727,
+      "eval_loss": 0.7895939350128174,
+      "eval_runtime": 63.8387,
+      "eval_samples_per_second": 15.461,
+      "eval_steps_per_second": 1.942,
       "step": 2500
     },
     {
-      "epoch": 0.552689756816507,
-      "grad_norm": 370.7579650878906,
-      "learning_rate": 4.3091378039793665e-07,
-      "loss": 0.1659,
+      "epoch": 0.3199658703071672,
+      "grad_norm": 9.852073823424234e-09,
+      "learning_rate": 4.4667235494880547e-07,
+      "loss": 0.159,
       "step": 3000
     },
     {
-      "epoch": 0.552689756816507,
-      "eval_loss": 0.719167172908783,
-      "eval_runtime": 76.6327,
-      "eval_samples_per_second": 15.75,
-      "eval_steps_per_second": 1.97,
+      "epoch": 0.3199658703071672,
+      "eval_loss": 0.7647623419761658,
+      "eval_runtime": 63.9155,
+      "eval_samples_per_second": 15.442,
+      "eval_steps_per_second": 1.94,
       "step": 3000
     },
     {
-      "epoch": 0.6448047162859248,
-      "grad_norm": 24.97766876220703,
-      "learning_rate": 4.193994104642594e-07,
-      "loss": 0.105,
+      "epoch": 0.37329351535836175,
+      "grad_norm": 0.0025215104687958956,
+      "learning_rate": 4.377844141069397e-07,
+      "loss": 0.1696,
       "step": 3500
     },
     {
-      "epoch": 0.6448047162859248,
-      "eval_loss": 0.7117515802383423,
-      "eval_runtime": 77.1286,
-      "eval_samples_per_second": 15.649,
-      "eval_steps_per_second": 1.958,
+      "epoch": 0.37329351535836175,
+      "eval_loss": 0.7638036012649536,
+      "eval_runtime": 63.8003,
+      "eval_samples_per_second": 15.47,
+      "eval_steps_per_second": 1.944,
       "step": 3500
     },
     {
-      "epoch": 0.7369196757553427,
-      "grad_norm": 305.62060546875,
-      "learning_rate": 4.0788504053058217e-07,
-      "loss": 0.1336,
+      "epoch": 0.42662116040955633,
+      "grad_norm": 2.4611830711364746,
+      "learning_rate": 4.2889647326507393e-07,
+      "loss": 0.1125,
       "step": 4000
     },
     {
-      "epoch": 0.7369196757553427,
-      "eval_loss": 0.6952972412109375,
-      "eval_runtime": 76.7253,
-      "eval_samples_per_second": 15.731,
-      "eval_steps_per_second": 1.968,
+      "epoch": 0.42662116040955633,
+      "eval_loss": 0.7486374378204346,
+      "eval_runtime": 63.9847,
+      "eval_samples_per_second": 15.426,
+      "eval_steps_per_second": 1.938,
       "step": 4000
     },
     {
-      "epoch": 0.8290346352247605,
-      "grad_norm": 506.12237548828125,
-      "learning_rate": 3.9637067059690496e-07,
-      "loss": 0.1154,
+      "epoch": 0.47994880546075086,
+      "grad_norm": 5.230295658111572,
+      "learning_rate": 4.2000853242320814e-07,
+      "loss": 0.1322,
       "step": 4500
     },
     {
-      "epoch": 0.8290346352247605,
-      "eval_loss": 0.6744509935379028,
-      "eval_runtime": 76.9493,
-      "eval_samples_per_second": 15.686,
-      "eval_steps_per_second": 1.962,
+      "epoch": 0.47994880546075086,
+      "eval_loss": 0.7251659035682678,
+      "eval_runtime": 64.4782,
+      "eval_samples_per_second": 15.308,
+      "eval_steps_per_second": 1.923,
       "step": 4500
     },
     {
-      "epoch": 0.9211495946941783,
-      "grad_norm": 8.823301322991028e-05,
-      "learning_rate": 3.8485630066322774e-07,
-      "loss": 0.108,
+      "epoch": 0.5332764505119454,
+      "grad_norm": 3.821002974291332e-05,
+      "learning_rate": 4.1112059158134245e-07,
+      "loss": 0.1633,
       "step": 5000
     },
     {
-      "epoch": 0.9211495946941783,
-      "eval_loss": 0.6559990048408508,
-      "eval_runtime": 75.2391,
-      "eval_samples_per_second": 16.042,
-      "eval_steps_per_second": 2.007,
+      "epoch": 0.5332764505119454,
+      "eval_loss": 0.7035414576530457,
+      "eval_runtime": 62.6186,
+      "eval_samples_per_second": 15.762,
+      "eval_steps_per_second": 1.98,
       "step": 5000
     },
     {
-      "epoch": 1.013264554163596,
-      "grad_norm": 0.0008560372516512871,
-      "learning_rate": 3.733419307295504e-07,
-      "loss": 0.1106,
+      "epoch": 0.58660409556314,
+      "grad_norm": 5.916939244343666e-06,
+      "learning_rate": 4.0223265073947665e-07,
+      "loss": 0.125,
       "step": 5500
     },
     {
-      "epoch": 1.013264554163596,
-      "eval_loss": 0.6367496848106384,
-      "eval_runtime": 77.0031,
-      "eval_samples_per_second": 15.675,
-      "eval_steps_per_second": 1.961,
+      "epoch": 0.58660409556314,
+      "eval_loss": 0.6792380809783936,
+      "eval_runtime": 65.0551,
+      "eval_samples_per_second": 15.172,
+      "eval_steps_per_second": 1.906,
       "step": 5500
     },
     {
-      "epoch": 1.105379513633014,
-      "grad_norm": 0.04510360211133957,
-      "learning_rate": 3.618275607958732e-07,
-      "loss": 0.0591,
+      "epoch": 0.6399317406143344,
+      "grad_norm": 92.8386459350586,
+      "learning_rate": 3.933447098976109e-07,
+      "loss": 0.1297,
       "step": 6000
     },
     {
-      "epoch": 1.105379513633014,
-      "eval_loss": 0.6259325742721558,
-      "eval_runtime": 75.2787,
-      "eval_samples_per_second": 16.034,
-      "eval_steps_per_second": 2.006,
+      "epoch": 0.6399317406143344,
+      "eval_loss": 0.6549482345581055,
+      "eval_runtime": 64.0827,
+      "eval_samples_per_second": 15.402,
+      "eval_steps_per_second": 1.935,
       "step": 6000
     },
     {
-      "epoch": 1.1974944731024317,
-      "grad_norm": 0.0005158882704563439,
-      "learning_rate": 3.50313190862196e-07,
-      "loss": 0.0745,
+      "epoch": 0.693259385665529,
+      "grad_norm": 0.00014717792510055006,
+      "learning_rate": 3.8445676905574517e-07,
+      "loss": 0.1114,
       "step": 6500
     },
     {
-      "epoch": 1.1974944731024317,
-      "eval_loss": 0.6209793090820312,
-      "eval_runtime": 77.1989,
-      "eval_samples_per_second": 15.635,
-      "eval_steps_per_second": 1.956,
+      "epoch": 0.693259385665529,
+      "eval_loss": 0.6223254799842834,
+      "eval_runtime": 64.89,
+      "eval_samples_per_second": 15.21,
+      "eval_steps_per_second": 1.911,
       "step": 6500
     },
     {
-      "epoch": 1.2896094325718497,
-      "grad_norm": 3.5883346072296263e-07,
-      "learning_rate": 3.387988209285188e-07,
-      "loss": 0.0502,
+      "epoch": 0.7465870307167235,
+      "grad_norm": 4.789559397977428e-07,
+      "learning_rate": 3.755688282138794e-07,
+      "loss": 0.144,
       "step": 7000
     },
     {
-      "epoch": 1.2896094325718497,
-      "eval_loss": 0.6133300065994263,
-      "eval_runtime": 75.1116,
-      "eval_samples_per_second": 16.069,
-      "eval_steps_per_second": 2.01,
+      "epoch": 0.7465870307167235,
+      "eval_loss": 0.6204637289047241,
+      "eval_runtime": 62.7674,
+      "eval_samples_per_second": 15.725,
+      "eval_steps_per_second": 1.976,
       "step": 7000
     },
     {
-      "epoch": 1.3817243920412676,
-      "grad_norm": 21.66597557067871,
-      "learning_rate": 3.272844509948415e-07,
-      "loss": 0.079,
+      "epoch": 0.7999146757679181,
+      "grad_norm": 1.7078508138656616,
+      "learning_rate": 3.6668088737201363e-07,
+      "loss": 0.1399,
       "step": 7500
     },
     {
-      "epoch": 1.3817243920412676,
-      "eval_loss": 0.6006803512573242,
-      "eval_runtime": 75.0086,
-      "eval_samples_per_second": 16.091,
-      "eval_steps_per_second": 2.013,
+      "epoch": 0.7999146757679181,
+      "eval_loss": 0.6074371933937073,
+      "eval_runtime": 63.4694,
+      "eval_samples_per_second": 15.551,
+      "eval_steps_per_second": 1.954,
       "step": 7500
     },
     {
-      "epoch": 1.4738393515106853,
-      "grad_norm": 7.992870814632624e-05,
-      "learning_rate": 3.157700810611643e-07,
-      "loss": 0.0776,
+      "epoch": 0.8532423208191127,
+      "grad_norm": 870.1183471679688,
+      "learning_rate": 3.5779294653014783e-07,
+      "loss": 0.141,
       "step": 8000
     },
     {
-      "epoch": 1.4738393515106853,
-      "eval_loss": 0.5866427421569824,
-      "eval_runtime": 75.6702,
-      "eval_samples_per_second": 15.951,
-      "eval_steps_per_second": 1.996,
+      "epoch": 0.8532423208191127,
+      "eval_loss": 0.5950366854667664,
+      "eval_runtime": 62.915,
+      "eval_samples_per_second": 15.688,
+      "eval_steps_per_second": 1.971,
       "step": 8000
     },
     {
-      "epoch": 1.565954310980103,
-      "grad_norm": 0.0010381735628470778,
-      "learning_rate": 3.042557111274871e-07,
-      "loss": 0.0492,
+      "epoch": 0.9065699658703071,
+      "grad_norm": 5.781071013188921e-05,
+      "learning_rate": 3.4890500568828214e-07,
+      "loss": 0.1291,
       "step": 8500
     },
     {
-      "epoch": 1.565954310980103,
-      "eval_loss": 0.5679212212562561,
-      "eval_runtime": 75.0643,
-      "eval_samples_per_second": 16.08,
-      "eval_steps_per_second": 2.012,
+      "epoch": 0.9065699658703071,
+      "eval_loss": 0.5800224542617798,
+      "eval_runtime": 62.3336,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.989,
       "step": 8500
     },
     {
-      "epoch": 1.658069270449521,
-      "grad_norm": 1.733377956725235e-07,
-      "learning_rate": 2.927413411938099e-07,
-      "loss": 0.0794,
+      "epoch": 0.9598976109215017,
+      "grad_norm": 1.3669992685317993,
+      "learning_rate": 3.4001706484641635e-07,
+      "loss": 0.1246,
       "step": 9000
     },
     {
-      "epoch": 1.658069270449521,
-      "eval_loss": 0.5761673450469971,
-      "eval_runtime": 76.2738,
-      "eval_samples_per_second": 15.825,
-      "eval_steps_per_second": 1.98,
+      "epoch": 0.9598976109215017,
+      "eval_loss": 0.5702486038208008,
+      "eval_runtime": 64.055,
+      "eval_samples_per_second": 15.409,
+      "eval_steps_per_second": 1.936,
       "step": 9000
     },
     {
-      "epoch": 1.750184229918939,
-      "grad_norm": 3.1063287359023306e-11,
-      "learning_rate": 2.812269712601326e-07,
-      "loss": 0.0677,
+      "epoch": 1.0132252559726962,
+      "grad_norm": 644.68115234375,
+      "learning_rate": 3.311291240045506e-07,
+      "loss": 0.1257,
       "step": 9500
     },
     {
-      "epoch": 1.750184229918939,
-      "eval_loss": 0.5565799474716187,
-      "eval_runtime": 75.6897,
-      "eval_samples_per_second": 15.947,
-      "eval_steps_per_second": 1.995,
+      "epoch": 1.0132252559726962,
+      "eval_loss": 0.5807780623435974,
+      "eval_runtime": 62.5312,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.983,
       "step": 9500
     },
     {
-      "epoch": 1.8422991893883567,
-      "grad_norm": 0.00010032750287791714,
-      "learning_rate": 2.697126013264554e-07,
-      "loss": 0.0566,
+      "epoch": 1.0665529010238908,
+      "grad_norm": 27.854568481445312,
+      "learning_rate": 3.2224118316268486e-07,
+      "loss": 0.0336,
       "step": 10000
     },
     {
-      "epoch": 1.8422991893883567,
-      "eval_loss": 0.5482199192047119,
-      "eval_runtime": 75.0029,
-      "eval_samples_per_second": 16.093,
-      "eval_steps_per_second": 2.013,
+      "epoch": 1.0665529010238908,
+      "eval_loss": 0.5713164210319519,
+      "eval_runtime": 62.5622,
+      "eval_samples_per_second": 15.776,
+      "eval_steps_per_second": 1.982,
       "step": 10000
     },
     {
-      "epoch": 1.9344141488577744,
-      "grad_norm": 5.553330595375883e-08,
-      "learning_rate": 2.581982313927782e-07,
-      "loss": 0.0828,
+      "epoch": 1.1198805460750854,
+      "grad_norm": 0.00045933027286082506,
+      "learning_rate": 3.133532423208191e-07,
+      "loss": 0.0663,
       "step": 10500
     },
     {
-      "epoch": 1.9344141488577744,
-      "eval_loss": 0.5500391125679016,
-      "eval_runtime": 76.2612,
-      "eval_samples_per_second": 15.827,
-      "eval_steps_per_second": 1.98,
+      "epoch": 1.1198805460750854,
+      "eval_loss": 0.5601483583450317,
+      "eval_runtime": 63.4675,
+      "eval_samples_per_second": 15.551,
+      "eval_steps_per_second": 1.954,
       "step": 10500
     },
     {
-      "epoch": 2.026529108327192,
-      "grad_norm": 1.5537947319899104e-06,
-      "learning_rate": 2.466838614591009e-07,
-      "loss": 0.0573,
+      "epoch": 1.17320819112628,
+      "grad_norm": 0.14330387115478516,
+      "learning_rate": 3.044653014789533e-07,
+      "loss": 0.078,
       "step": 11000
     },
     {
-      "epoch": 2.026529108327192,
-      "eval_loss": 0.5341849327087402,
-      "eval_runtime": 75.6131,
-      "eval_samples_per_second": 15.963,
-      "eval_steps_per_second": 1.997,
+      "epoch": 1.17320819112628,
+      "eval_loss": 0.5580261945724487,
+      "eval_runtime": 63.3927,
+      "eval_samples_per_second": 15.57,
+      "eval_steps_per_second": 1.956,
       "step": 11000
     },
     {
-      "epoch": 2.1186440677966103,
-      "grad_norm": 36.10622024536133,
-      "learning_rate": 2.3516949152542374e-07,
-      "loss": 0.0401,
+      "epoch": 1.2265358361774745,
+      "grad_norm": 4.066005518388316e-11,
+      "learning_rate": 2.955773606370876e-07,
+      "loss": 0.1118,
       "step": 11500
     },
     {
-      "epoch": 2.1186440677966103,
-      "eval_loss": 0.5351412296295166,
-      "eval_runtime": 74.8873,
-      "eval_samples_per_second": 16.118,
-      "eval_steps_per_second": 2.016,
+      "epoch": 1.2265358361774745,
+      "eval_loss": 0.5491370558738708,
+      "eval_runtime": 63.7293,
+      "eval_samples_per_second": 15.487,
+      "eval_steps_per_second": 1.946,
       "step": 11500
     },
     {
-      "epoch": 2.210759027266028,
-      "grad_norm": 0.49580878019332886,
-      "learning_rate": 2.236551215917465e-07,
-      "loss": 0.0152,
+      "epoch": 1.2798634812286689,
+      "grad_norm": 7.519358769059181e-05,
+      "learning_rate": 2.8668941979522184e-07,
+      "loss": 0.0534,
       "step": 12000
     },
     {
-      "epoch": 2.210759027266028,
-      "eval_loss": 0.5349487662315369,
-      "eval_runtime": 76.3109,
-      "eval_samples_per_second": 15.817,
-      "eval_steps_per_second": 1.979,
+      "epoch": 1.2798634812286689,
+      "eval_loss": 0.5478764772415161,
+      "eval_runtime": 62.7547,
+      "eval_samples_per_second": 15.728,
+      "eval_steps_per_second": 1.976,
       "step": 12000
     },
     {
-      "epoch": 2.3028739867354457,
-      "grad_norm": 1.7469781637191772,
-      "learning_rate": 2.1214075165806928e-07,
-      "loss": 0.0638,
+      "epoch": 1.3331911262798635,
+      "grad_norm": 8.400617446113756e-08,
+      "learning_rate": 2.778014789533561e-07,
+      "loss": 0.0507,
       "step": 12500
     },
     {
-      "epoch": 2.3028739867354457,
-      "eval_loss": 0.531768798828125,
-      "eval_runtime": 75.5712,
-      "eval_samples_per_second": 15.972,
-      "eval_steps_per_second": 1.998,
+      "epoch": 1.3331911262798635,
+      "eval_loss": 0.576600968837738,
+      "eval_runtime": 64.301,
+      "eval_samples_per_second": 15.35,
+      "eval_steps_per_second": 1.928,
       "step": 12500
     },
     {
-      "epoch": 2.3949889462048635,
-      "grad_norm": 0.0006845183088444173,
-      "learning_rate": 2.0062638172439202e-07,
-      "loss": 0.0488,
+      "epoch": 1.386518771331058,
+      "grad_norm": 2.5385464130084356e-12,
+      "learning_rate": 2.689135381114903e-07,
+      "loss": 0.0784,
       "step": 13000
     },
     {
-      "epoch": 2.3949889462048635,
-      "eval_loss": 0.530552089214325,
-      "eval_runtime": 75.6479,
-      "eval_samples_per_second": 15.955,
-      "eval_steps_per_second": 1.996,
+      "epoch": 1.386518771331058,
+      "eval_loss": 0.5729931592941284,
+      "eval_runtime": 63.652,
+      "eval_samples_per_second": 15.506,
+      "eval_steps_per_second": 1.948,
       "step": 13000
     },
     {
-      "epoch": 2.4871039056742816,
-      "grad_norm": 0.0002451306500006467,
-      "learning_rate": 1.891120117907148e-07,
-      "loss": 0.0456,
+      "epoch": 1.4398464163822526,
+      "grad_norm": 0.08409273624420166,
+      "learning_rate": 2.600255972696246e-07,
+      "loss": 0.079,
       "step": 13500
     },
     {
-      "epoch": 2.4871039056742816,
-      "eval_loss": 0.5211473107337952,
-      "eval_runtime": 76.4401,
-      "eval_samples_per_second": 15.79,
-      "eval_steps_per_second": 1.975,
+      "epoch": 1.4398464163822526,
+      "eval_loss": 0.5694039463996887,
+      "eval_runtime": 62.8247,
+      "eval_samples_per_second": 15.71,
+      "eval_steps_per_second": 1.974,
       "step": 13500
     },
     {
-      "epoch": 2.5792188651436994,
-      "grad_norm": 0.003609905717894435,
-      "learning_rate": 1.7759764185703757e-07,
-      "loss": 0.0264,
+      "epoch": 1.493174061433447,
+      "grad_norm": 3.7334253022436314e-08,
+      "learning_rate": 2.511376564277588e-07,
+      "loss": 0.0507,
       "step": 14000
     },
     {
-      "epoch": 2.5792188651436994,
-      "eval_loss": 0.5193639993667603,
-      "eval_runtime": 74.595,
-      "eval_samples_per_second": 16.181,
-      "eval_steps_per_second": 2.024,
+      "epoch": 1.493174061433447,
+      "eval_loss": 0.5713907480239868,
+      "eval_runtime": 62.3577,
+      "eval_samples_per_second": 15.828,
+      "eval_steps_per_second": 1.989,
       "step": 14000
     },
     {
-      "epoch": 2.671333824613117,
-      "grad_norm": 6.751325276610487e-09,
-      "learning_rate": 1.6608327192336035e-07,
-      "loss": 0.0381,
+      "epoch": 1.5465017064846416,
+      "grad_norm": 1.5219894647598267,
+      "learning_rate": 2.422497155858931e-07,
+      "loss": 0.0774,
       "step": 14500
     },
     {
-      "epoch": 2.671333824613117,
-      "eval_loss": 0.520633339881897,
-      "eval_runtime": 75.2937,
-      "eval_samples_per_second": 16.031,
-      "eval_steps_per_second": 2.005,
+      "epoch": 1.5465017064846416,
+      "eval_loss": 0.5674533843994141,
+      "eval_runtime": 62.3753,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.988,
       "step": 14500
     },
     {
-      "epoch": 2.7634487840825352,
-      "grad_norm": 8.452875022157968e-07,
-      "learning_rate": 1.5456890198968311e-07,
-      "loss": 0.0413,
+      "epoch": 1.5998293515358362,
+      "grad_norm": 879.1193237304688,
+      "learning_rate": 2.333617747440273e-07,
+      "loss": 0.0376,
       "step": 15000
     },
     {
-      "epoch": 2.7634487840825352,
-      "eval_loss": 0.5167532563209534,
-      "eval_runtime": 75.615,
-      "eval_samples_per_second": 15.962,
-      "eval_steps_per_second": 1.997,
+      "epoch": 1.5998293515358362,
+      "eval_loss": 0.561253547668457,
+      "eval_runtime": 62.4788,
+      "eval_samples_per_second": 15.797,
+      "eval_steps_per_second": 1.985,
       "step": 15000
     },
     {
-      "epoch": 2.855563743551953,
-      "grad_norm": 2.365756202049596e-11,
-      "learning_rate": 1.430545320560059e-07,
-      "loss": 0.0392,
+      "epoch": 1.6531569965870307,
+      "grad_norm": 0.0006221308722160757,
+      "learning_rate": 2.2447383390216154e-07,
+      "loss": 0.0929,
       "step": 15500
     },
     {
-      "epoch": 2.855563743551953,
-      "eval_loss": 0.5148531198501587,
-      "eval_runtime": 74.6217,
-      "eval_samples_per_second": 16.175,
-      "eval_steps_per_second": 2.024,
+      "epoch": 1.6531569965870307,
+      "eval_loss": 0.5395999550819397,
+      "eval_runtime": 62.5651,
+      "eval_samples_per_second": 15.776,
+      "eval_steps_per_second": 1.982,
       "step": 15500
     },
     {
-      "epoch": 2.9476787030213707,
-      "grad_norm": 0.32707780599594116,
-      "learning_rate": 1.3154016212232866e-07,
-      "loss": 0.0352,
+      "epoch": 1.7064846416382253,
+      "grad_norm": 37.85109329223633,
+      "learning_rate": 2.155858930602958e-07,
+      "loss": 0.0442,
       "step": 16000
     },
     {
-      "epoch": 2.9476787030213707,
-      "eval_loss": 0.5112494826316833,
-      "eval_runtime": 75.142,
-      "eval_samples_per_second": 16.063,
-      "eval_steps_per_second": 2.01,
+      "epoch": 1.7064846416382253,
+      "eval_loss": 0.5307884216308594,
+      "eval_runtime": 62.4681,
+      "eval_samples_per_second": 15.8,
+      "eval_steps_per_second": 1.985,
       "step": 16000
     },
     {
-      "epoch": 3.0397936624907884,
-      "grad_norm": 4.2568670810538833e-11,
-      "learning_rate": 1.2002579218865142e-07,
-      "loss": 0.0467,
+      "epoch": 1.75981228668942,
+      "grad_norm": 9.600337885773413e-11,
+      "learning_rate": 2.0669795221843002e-07,
+      "loss": 0.0593,
       "step": 16500
     },
     {
-      "epoch": 3.0397936624907884,
-      "eval_loss": 0.5097618699073792,
-      "eval_runtime": 75.705,
-      "eval_samples_per_second": 15.943,
-      "eval_steps_per_second": 1.995,
+      "epoch": 1.75981228668942,
+      "eval_loss": 0.5230685472488403,
+      "eval_runtime": 63.8124,
+      "eval_samples_per_second": 15.467,
+      "eval_steps_per_second": 1.943,
       "step": 16500
     },
     {
-      "epoch": 3.131908621960206,
-      "grad_norm": 7.157668679980134e-09,
-      "learning_rate": 1.085114222549742e-07,
-      "loss": 0.0366,
+      "epoch": 1.8131399317406145,
+      "grad_norm": 9.067448940129619e-13,
+      "learning_rate": 1.9781001137656428e-07,
+      "loss": 0.0578,
       "step": 17000
     },
     {
-      "epoch": 3.131908621960206,
-      "eval_loss": 0.5088639855384827,
-      "eval_runtime": 74.5863,
-      "eval_samples_per_second": 16.183,
-      "eval_steps_per_second": 2.025,
+      "epoch": 1.8131399317406145,
+      "eval_loss": 0.5200281739234924,
+      "eval_runtime": 62.3767,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.988,
       "step": 17000
     },
     {
-      "epoch": 3.2240235814296243,
-      "grad_norm": 0.5768880248069763,
-      "learning_rate": 9.699705232129697e-08,
-      "loss": 0.0454,
+      "epoch": 1.8664675767918089,
+      "grad_norm": 0.00843018852174282,
+      "learning_rate": 1.889220705346985e-07,
+      "loss": 0.0928,
       "step": 17500
     },
     {
-      "epoch": 3.2240235814296243,
-      "eval_loss": 0.5103679299354553,
-      "eval_runtime": 74.9953,
-      "eval_samples_per_second": 16.094,
-      "eval_steps_per_second": 2.013,
+      "epoch": 1.8664675767918089,
+      "eval_loss": 0.509168803691864,
+      "eval_runtime": 62.4108,
+      "eval_samples_per_second": 15.815,
+      "eval_steps_per_second": 1.987,
       "step": 17500
     },
     {
-      "epoch": 3.316138540899042,
-      "grad_norm": 0.029806463047862053,
-      "learning_rate": 8.548268238761974e-08,
-      "loss": 0.0209,
+      "epoch": 1.9197952218430034,
+      "grad_norm": 4.264631314754297e-09,
+      "learning_rate": 1.8003412969283277e-07,
+      "loss": 0.0633,
       "step": 18000
     },
     {
-      "epoch": 3.316138540899042,
-      "eval_loss": 0.5070606470108032,
-      "eval_runtime": 75.7962,
-      "eval_samples_per_second": 15.924,
-      "eval_steps_per_second": 1.992,
+      "epoch": 1.9197952218430034,
+      "eval_loss": 0.4954444468021393,
+      "eval_runtime": 63.3639,
+      "eval_samples_per_second": 15.577,
+      "eval_steps_per_second": 1.957,
       "step": 18000
     },
     {
-      "epoch": 3.4082535003684598,
-      "grad_norm": 0.09116169810295105,
-      "learning_rate": 7.396831245394252e-08,
-      "loss": 0.0636,
+      "epoch": 1.9731228668941978,
+      "grad_norm": 459.5137939453125,
+      "learning_rate": 1.7114618885096697e-07,
+      "loss": 0.067,
       "step": 18500
     },
     {
-      "epoch": 3.4082535003684598,
-      "eval_loss": 0.5044638514518738,
-      "eval_runtime": 74.7518,
-      "eval_samples_per_second": 16.147,
-      "eval_steps_per_second": 2.02,
+      "epoch": 1.9731228668941978,
+      "eval_loss": 0.49844062328338623,
+      "eval_runtime": 63.6573,
+      "eval_samples_per_second": 15.505,
+      "eval_steps_per_second": 1.948,
       "step": 18500
     },
     {
-      "epoch": 3.500368459837878,
-      "grad_norm": 0.13625642657279968,
-      "learning_rate": 6.245394252026529e-08,
-      "loss": 0.0159,
+      "epoch": 2.0264505119453924,
+      "grad_norm": 0.27590689063072205,
+      "learning_rate": 1.6225824800910123e-07,
+      "loss": 0.0525,
       "step": 19000
     },
     {
-      "epoch": 3.500368459837878,
-      "eval_loss": 0.5019211769104004,
-      "eval_runtime": 75.0004,
-      "eval_samples_per_second": 16.093,
-      "eval_steps_per_second": 2.013,
+      "epoch": 2.0264505119453924,
+      "eval_loss": 0.4999145269393921,
+      "eval_runtime": 63.1727,
+      "eval_samples_per_second": 15.624,
+      "eval_steps_per_second": 1.963,
       "step": 19000
     },
     {
-      "epoch": 3.5924834193072956,
-      "grad_norm": 2.3882080313342158e-06,
-      "learning_rate": 5.0939572586588065e-08,
-      "loss": 0.0303,
+      "epoch": 2.079778156996587,
+      "grad_norm": 1.349623744317796e-06,
+      "learning_rate": 1.533703071672355e-07,
+      "loss": 0.0301,
       "step": 19500
     },
     {
-      "epoch": 3.5924834193072956,
-      "eval_loss": 0.49851301312446594,
-      "eval_runtime": 75.7613,
-      "eval_samples_per_second": 15.932,
-      "eval_steps_per_second": 1.993,
+      "epoch": 2.079778156996587,
+      "eval_loss": 0.4926412105560303,
+      "eval_runtime": 63.4421,
+      "eval_samples_per_second": 15.557,
+      "eval_steps_per_second": 1.955,
       "step": 19500
     },
     {
-      "epoch": 3.6845983787767134,
-      "grad_norm": 8.883699109674126e-09,
-      "learning_rate": 3.942520265291083e-08,
-      "loss": 0.0353,
+      "epoch": 2.1331058020477816,
+      "grad_norm": 4.928138878312893e-06,
+      "learning_rate": 1.4448236632536972e-07,
+      "loss": 0.0408,
       "step": 20000
     },
     {
-      "epoch": 3.6845983787767134,
-      "eval_loss": 0.49747875332832336,
-      "eval_runtime": 74.5978,
-      "eval_samples_per_second": 16.18,
-      "eval_steps_per_second": 2.024,
+      "epoch": 2.1331058020477816,
+      "eval_loss": 0.4872625172138214,
+      "eval_runtime": 63.2047,
+      "eval_samples_per_second": 15.616,
+      "eval_steps_per_second": 1.962,
       "step": 20000
     },
     {
-      "epoch": 3.776713338246131,
-      "grad_norm": 0.12466106563806534,
-      "learning_rate": 2.7910832719233602e-08,
-      "loss": 0.0261,
+      "epoch": 2.186433447098976,
+      "grad_norm": 0.003605826525017619,
+      "learning_rate": 1.3559442548350398e-07,
+      "loss": 0.0489,
       "step": 20500
     },
     {
-      "epoch": 3.776713338246131,
-      "eval_loss": 0.4962019622325897,
-      "eval_runtime": 75.2409,
-      "eval_samples_per_second": 16.042,
-      "eval_steps_per_second": 2.007,
+      "epoch": 2.186433447098976,
+      "eval_loss": 0.48834049701690674,
+      "eval_runtime": 63.5831,
+      "eval_samples_per_second": 15.523,
+      "eval_steps_per_second": 1.95,
       "step": 20500
     },
     {
-      "epoch": 3.868828297715549,
-      "grad_norm": 1.0646399317693067e-07,
-      "learning_rate": 1.6396462785556373e-08,
-      "loss": 0.0291,
+      "epoch": 2.2397610921501707,
+      "grad_norm": 0.0027391298208385706,
+      "learning_rate": 1.267064846416382e-07,
+      "loss": 0.0354,
       "step": 21000
     },
     {
-      "epoch": 3.868828297715549,
-      "eval_loss": 0.49561700224876404,
-      "eval_runtime": 75.8041,
-      "eval_samples_per_second": 15.923,
-      "eval_steps_per_second": 1.992,
+      "epoch": 2.2397610921501707,
+      "eval_loss": 0.4887964129447937,
+      "eval_runtime": 63.0462,
+      "eval_samples_per_second": 15.655,
+      "eval_steps_per_second": 1.967,
       "step": 21000
     },
     {
-      "epoch": 3.960943257184967,
-      "grad_norm": 3.997461042737349e-11,
-      "learning_rate": 4.882092851879145e-09,
-      "loss": 0.0338,
+      "epoch": 2.2930887372013653,
+      "grad_norm": 1.5921540352792363e-07,
+      "learning_rate": 1.1781854379977247e-07,
+      "loss": 0.0667,
       "step": 21500
     },
     {
-      "epoch": 3.960943257184967,
-      "eval_loss": 0.495645672082901,
-      "eval_runtime": 74.5834,
-      "eval_samples_per_second": 16.183,
-      "eval_steps_per_second": 2.025,
+      "epoch": 2.2930887372013653,
+      "eval_loss": 0.4866587221622467,
+      "eval_runtime": 63.4143,
+      "eval_samples_per_second": 15.564,
+      "eval_steps_per_second": 1.955,
       "step": 21500
     },
     {
-      "epoch": 3.6844749623178696,
-      "grad_norm": 0.022144177928566933,
-      "learning_rate": 3.944062971026629e-08,
-      "loss": 0.0432,
+      "epoch": 2.34641638225256,
+      "grad_norm": 3.954861771404694e-09,
+      "learning_rate": 1.0893060295790671e-07,
+      "loss": 0.0315,
       "step": 22000
     },
     {
-      "epoch": 3.6844749623178696,
-      "eval_loss": 0.17835669219493866,
-      "eval_runtime": 7.4507,
-      "eval_samples_per_second": 16.24,
-      "eval_steps_per_second": 2.147,
+      "epoch": 2.34641638225256,
+      "eval_loss": 0.48805058002471924,
+      "eval_runtime": 63.4731,
+      "eval_samples_per_second": 15.55,
+      "eval_steps_per_second": 1.954,
       "step": 22000
     },
     {
-      "epoch": 3.7682130296432756,
-      "grad_norm": 2.4324751279891643e-07,
-      "learning_rate": 2.897337129459052e-08,
-      "loss": 0.0461,
+      "epoch": 2.399744027303754,
+      "grad_norm": 1.405540217547241e-07,
+      "learning_rate": 1.0004266211604096e-07,
+      "loss": 0.0579,
       "step": 22500
     },
     {
-      "epoch": 3.7682130296432756,
-      "eval_loss": 0.17669138312339783,
-      "eval_runtime": 7.5623,
-      "eval_samples_per_second": 16.0,
-      "eval_steps_per_second": 2.116,
+      "epoch": 2.399744027303754,
+      "eval_loss": 0.48956525325775146,
+      "eval_runtime": 63.5675,
+      "eval_samples_per_second": 15.527,
+      "eval_steps_per_second": 1.951,
       "step": 22500
     },
     {
-      "epoch": 3.851951096968682,
-      "grad_norm": 4.172260048562748e-07,
-      "learning_rate": 1.8506112878914754e-08,
-      "loss": 0.0513,
+      "epoch": 2.453071672354949,
+      "grad_norm": 2.2330443698592717e-06,
+      "learning_rate": 9.11547212741752e-08,
+      "loss": 0.0387,
       "step": 23000
     },
     {
-      "epoch": 3.851951096968682,
-      "eval_loss": 0.1774093359708786,
-      "eval_runtime": 7.8193,
-      "eval_samples_per_second": 15.475,
-      "eval_steps_per_second": 2.046,
+      "epoch": 2.453071672354949,
+      "eval_loss": 0.48340705037117004,
+      "eval_runtime": 62.238,
+      "eval_samples_per_second": 15.858,
+      "eval_steps_per_second": 1.992,
       "step": 23000
     }
   ],
   "logging_steps": 500,
-  "max_steps": 23884,
+  "max_steps": 28128,
   "num_input_tokens_seen": 0,
-  "num_train_epochs": 4,
+  "num_train_epochs": 3,
   "save_steps": 500,
   "stateful_callbacks": {
     "TrainerControl": {

checkpoint-23000/training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f71c1b7b2e3d63d5d414705f142c4860d9045ea3e8f1377d3c6bfb13c6fefb73
+oid sha256:057f6bbf1ff3e34abda0863071bbf5fdcc7edb1102f04dba05c7a15087bc3396
 size 4847