Model save

README.md

@@ -2,15 +2,9 @@
 license: apache-2.0
 base_model: Minbyul/biomistral-7b-wo-kqa_golden-iter-dpo-step2
 tags:
-- alignment-handbook
 - trl
 - dpo
 - generated_from_trainer
-- trl
-- dpo
-- generated_from_trainer
-datasets:
-- HuggingFaceH4/ultrafeedback_binarized
 model-index:
 - name: biomistral-7b-wo-kqa_golden-iter-dpo-step3
   results: []
@@ -21,17 +15,7 @@ should probably proofread and complete it, then remove this comment. -->
 
 # biomistral-7b-wo-kqa_golden-iter-dpo-step3
 
-This model is a fine-tuned version of [Minbyul/biomistral-7b-wo-kqa_golden-iter-dpo-step2](https://huggingface.co/Minbyul/biomistral-7b-wo-kqa_golden-iter-dpo-step2) on the HuggingFaceH4/ultrafeedback_binarized dataset.
-It achieves the following results on the evaluation set:
-- Loss: 0.6829
-- Rewards/chosen: -0.0502
-- Rewards/rejected: -0.0800
-- Rewards/accuracies: 0.6300
-- Rewards/margins: 0.0298
-- Logps/rejected: -60.1185
-- Logps/chosen: -40.1264
-- Logits/rejected: -1.5228
-- Logits/chosen: -0.8710
+This model is a fine-tuned version of [Minbyul/biomistral-7b-wo-kqa_golden-iter-dpo-step2](https://huggingface.co/Minbyul/biomistral-7b-wo-kqa_golden-iter-dpo-step2) on an unknown dataset.
 
 ## Model description
 
@@ -66,15 +50,11 @@ The following hyperparameters were used during training:
 
 ### Training results
 
-| Training Loss | Epoch | Step | Logits/chosen | Logits/rejected | Logps/chosen | Logps/rejected | Validation Loss | Rewards/accuracies | Rewards/chosen | Rewards/margins | Rewards/rejected |
-|:-------------:|:-----:|:----:|:-------------:|:---------------:|:------------:|:--------------:|:---------------:|:------------------:|:--------------:|:---------------:|:----------------:|
-| 0.6794        | 0.37  | 100  | -0.8266       | -1.4757         | -35.5860     | -53.0765       | 0.6906          | 0.5900             | -0.0048        | 0.0048          | -0.0096          |
-| 0.6555        | 0.74  | 200  | -0.8589       | -1.5130         | -39.0432     | -58.7210       | 0.6837          | 0.6400             | -0.0394        | 0.0267          | -0.0661          |
 
 
 ### Framework versions
 
-- Transformers 4.39.0.dev0
-- Pytorch 2.1.2
+- Transformers 4.38.2
+- Pytorch 2.1.2+cu121
 - Datasets 2.14.6
 - Tokenizers 0.15.2

all_results.json

@@ -1,21 +1,8 @@
 {
-    "epoch": 1.0,
-    "eval_logits/chosen": -0.8710112571716309,
-    "eval_logits/rejected": -1.522813320159912,
-    "eval_logps/chosen": -40.126407623291016,
-    "eval_logps/rejected": -60.11854934692383,
-    "eval_loss": 0.6829047799110413,
-    "eval_rewards/accuracies": 0.6299999952316284,
-    "eval_rewards/chosen": -0.050214797258377075,
-    "eval_rewards/margins": 0.02983289770781994,
-    "eval_rewards/rejected": -0.08004770427942276,
-    "eval_runtime": 65.4744,
-    "eval_samples": 785,
-    "eval_samples_per_second": 11.989,
-    "eval_steps_per_second": 0.382,
-    "train_loss": 0.1692200431117305,
-    "train_runtime": 798.9647,
-    "train_samples": 17264,
-    "train_samples_per_second": 21.608,
-    "train_steps_per_second": 0.338
+    "epoch": 0.99,
+    "train_loss": 0.6816894855905087,
+    "train_runtime": 1070.9433,
+    "train_samples": 6039,
+    "train_samples_per_second": 5.639,
+    "train_steps_per_second": 0.088
 }

config.json

@@ -20,7 +20,7 @@
   "sliding_window": 4096,
   "tie_word_embeddings": false,
   "torch_dtype": "bfloat16",
-  "transformers_version": "4.39.0.dev0",
-  "use_cache": true,
+  "transformers_version": "4.38.2",
+  "use_cache": false,
   "vocab_size": 32000
 }

generation_config.json

@@ -2,5 +2,5 @@
   "_from_model_config": true,
   "bos_token_id": 1,
   "eos_token_id": 2,
-  "transformers_version": "4.39.0.dev0"
+  "transformers_version": "4.38.2"
 }

model-00001-of-00003.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:cce409cff4722b9a731610eee4285f50c9f2d5b36cf9e7d2481b206ae57402a9
+oid sha256:75ddb36d2d5d2523812b20bb5a38e56c6ac47fb4338d8969c27b520509a2f32e
 size 4943162336

model-00002-of-00003.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8187324d72ff88a59d6bc9c26004792edd0b87978a47298a811599baa6429504
+oid sha256:105af67768a394930d011db6704a0f3abc9425e70d63e13d82d2c002e8efd97c
 size 4999819336

model-00003-of-00003.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a53d580d5ce1226972d503f03c4c98a758c0b21e66e24742d3e4d812d193cc5d
+oid sha256:5421445557816c21ef40cf96ce4d7a8058cac224c2456363aac443ed153aca39
 size 4540516344

train_results.json

@@ -1,8 +1,8 @@
 {
-    "epoch": 1.0,
-    "train_loss": 0.1692200431117305,
-    "train_runtime": 798.9647,
-    "train_samples": 17264,
-    "train_samples_per_second": 21.608,
-    "train_steps_per_second": 0.338
+    "epoch": 0.99,
+    "train_loss": 0.6816894855905087,
+    "train_runtime": 1070.9433,
+    "train_samples": 6039,
+    "train_samples_per_second": 5.639,
+    "train_steps_per_second": 0.088
 }

trainer_state.json

@@ -1,21 +1,21 @@
 {
   "best_metric": null,
   "best_model_checkpoint": null,
-  "epoch": 1.0,
-  "eval_steps": 100,
-  "global_step": 270,
+  "epoch": 0.9947089947089947,
+  "eval_steps": 500,
+  "global_step": 94,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
-      "epoch": 0.0,
-      "grad_norm": 41.70182769977995,
-      "learning_rate": 3.703703703703703e-09,
-      "logits/chosen": -1.5467987060546875,
-      "logits/rejected": -1.308578372001648,
-      "logps/chosen": -37.12041473388672,
-      "logps/rejected": -61.68842315673828,
+      "epoch": 0.01,
+      "grad_norm": 25.005165779900267,
+      "learning_rate": 1e-08,
+      "logits/chosen": -1.9501205682754517,
+      "logits/rejected": -2.513594388961792,
+      "logps/chosen": -348.5884704589844,
+      "logps/rejected": -166.58517456054688,
       "loss": 0.6931,
       "rewards/accuracies": 0.0,
       "rewards/chosen": 0.0,
@@ -24,457 +24,155 @@
       "step": 1
     },
     {
-      "epoch": 0.04,
-      "grad_norm": 37.047557637653945,
-      "learning_rate": 3.7037037037037036e-08,
-      "logits/chosen": -0.9453011751174927,
-      "logits/rejected": -1.405588984489441,
-      "logps/chosen": -28.646324157714844,
-      "logps/rejected": -40.27184295654297,
-      "loss": 0.693,
+      "epoch": 0.11,
+      "grad_norm": 23.68906488321,
+      "learning_rate": 1e-07,
+      "logits/chosen": -2.640984535217285,
+      "logits/rejected": -2.219906806945801,
+      "logps/chosen": -213.95584106445312,
+      "logps/rejected": -198.17874145507812,
+      "loss": 0.6932,
       "rewards/accuracies": 0.4930555522441864,
-      "rewards/chosen": -0.00016396207502111793,
-      "rewards/margins": 0.001127531286329031,
-      "rewards/rejected": -0.0012914934195578098,
+      "rewards/chosen": 0.0007135343039408326,
+      "rewards/margins": 0.0006534373969770968,
+      "rewards/rejected": 6.009703065501526e-05,
       "step": 10
     },
     {
-      "epoch": 0.07,
-      "grad_norm": 38.71986120422308,
-      "learning_rate": 7.407407407407407e-08,
-      "logits/chosen": -1.144275188446045,
-      "logits/rejected": -1.4153748750686646,
-      "logps/chosen": -38.298439025878906,
-      "logps/rejected": -41.69558334350586,
-      "loss": 0.6927,
-      "rewards/accuracies": 0.5625,
-      "rewards/chosen": 0.0009220183710567653,
-      "rewards/margins": 0.0017043445259332657,
-      "rewards/rejected": -0.0007823262130841613,
+      "epoch": 0.21,
+      "grad_norm": 22.52231020802604,
+      "learning_rate": 9.65436874322102e-08,
+      "logits/chosen": -2.397062301635742,
+      "logits/rejected": -2.3303606510162354,
+      "logps/chosen": -257.8389587402344,
+      "logps/rejected": -222.90444946289062,
+      "loss": 0.6917,
+      "rewards/accuracies": 0.6187499761581421,
+      "rewards/chosen": 0.0030278589110821486,
+      "rewards/margins": 0.002860091160982847,
+      "rewards/rejected": 0.00016776802658569068,
       "step": 20
     },
     {
-      "epoch": 0.11,
-      "grad_norm": 39.765817458266895,
-      "learning_rate": 9.99623976252115e-08,
-      "logits/chosen": -0.8312944173812866,
-      "logits/rejected": -1.4380576610565186,
-      "logps/chosen": -32.726478576660156,
-      "logps/rejected": -43.810638427734375,
-      "loss": 0.6925,
-      "rewards/accuracies": 0.59375,
-      "rewards/chosen": 0.001944520859979093,
-      "rewards/margins": 0.0023826442193239927,
-      "rewards/rejected": -0.00043812341755256057,
+      "epoch": 0.32,
+      "grad_norm": 24.931881082321315,
+      "learning_rate": 8.665259359149131e-08,
+      "logits/chosen": -2.534593105316162,
+      "logits/rejected": -2.4346184730529785,
+      "logps/chosen": -227.6776885986328,
+      "logps/rejected": -204.8966064453125,
+      "loss": 0.689,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": 0.00954088568687439,
+      "rewards/margins": 0.00823338981717825,
+      "rewards/rejected": 0.0013074951712042093,
       "step": 30
     },
     {
-      "epoch": 0.15,
-      "grad_norm": 35.39090968218046,
-      "learning_rate": 9.929548316723981e-08,
-      "logits/chosen": -1.0166804790496826,
-      "logits/rejected": -1.4055747985839844,
-      "logps/chosen": -31.433521270751953,
-      "logps/rejected": -46.84731674194336,
-      "loss": 0.6915,
-      "rewards/accuracies": 0.5874999761581421,
-      "rewards/chosen": 0.0015541177708655596,
-      "rewards/margins": 0.002054566517472267,
-      "rewards/rejected": -0.0005004488630220294,
+      "epoch": 0.42,
+      "grad_norm": 24.657287539381663,
+      "learning_rate": 7.16941869558779e-08,
+      "logits/chosen": -2.358189105987549,
+      "logits/rejected": -2.440410614013672,
+      "logps/chosen": -230.328857421875,
+      "logps/rejected": -210.7056427001953,
+      "loss": 0.6846,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": 0.018901044502854347,
+      "rewards/margins": 0.01690484955906868,
+      "rewards/rejected": 0.0019961954094469547,
       "step": 40
     },
     {
-      "epoch": 0.19,
-      "grad_norm": 42.118488242592456,
-      "learning_rate": 9.780577808869398e-08,
-      "logits/chosen": -1.0024656057357788,
-      "logits/rejected": -1.3138850927352905,
-      "logps/chosen": -30.340801239013672,
-      "logps/rejected": -43.08925247192383,
-      "loss": 0.6905,
-      "rewards/accuracies": 0.606249988079071,
-      "rewards/chosen": 0.0033007909078150988,
-      "rewards/margins": 0.003635294735431671,
-      "rewards/rejected": -0.000334503740305081,
+      "epoch": 0.53,
+      "grad_norm": 24.026917311824274,
+      "learning_rate": 5.373650467932121e-08,
+      "logits/chosen": -2.379296064376831,
+      "logits/rejected": -2.6683709621429443,
+      "logps/chosen": -233.249267578125,
+      "logps/rejected": -199.204833984375,
+      "loss": 0.6805,
+      "rewards/accuracies": 0.8187500238418579,
+      "rewards/chosen": 0.02693324163556099,
+      "rewards/margins": 0.024523768573999405,
+      "rewards/rejected": 0.002409472828730941,
       "step": 50
     },
     {
-      "epoch": 0.22,
-      "grad_norm": 36.92382364235435,
-      "learning_rate": 9.551814704830733e-08,
-      "logits/chosen": -1.1217851638793945,
-      "logits/rejected": -1.4166934490203857,
-      "logps/chosen": -31.992273330688477,
-      "logps/rejected": -46.66759490966797,
-      "loss": 0.6886,
-      "rewards/accuracies": 0.6625000238418579,
-      "rewards/chosen": 0.008065198548138142,
-      "rewards/margins": 0.010263124480843544,
-      "rewards/rejected": -0.0021979263983666897,
+      "epoch": 0.63,
+      "grad_norm": 23.069042390909715,
+      "learning_rate": 3.5262241279454787e-08,
+      "logits/chosen": -2.233121156692505,
+      "logits/rejected": -2.644624710083008,
+      "logps/chosen": -250.5662078857422,
+      "logps/rejected": -173.79067993164062,
+      "loss": 0.6762,
+      "rewards/accuracies": 0.90625,
+      "rewards/chosen": 0.03840692713856697,
+      "rewards/margins": 0.03592415899038315,
+      "rewards/rejected": 0.0024827648885548115,
       "step": 60
     },
     {
-      "epoch": 0.26,
-      "grad_norm": 39.24669749459966,
-      "learning_rate": 9.247077288236486e-08,
-      "logits/chosen": -0.9909595251083374,
-      "logits/rejected": -1.514561653137207,
-      "logps/chosen": -31.16408348083496,
-      "logps/rejected": -57.032684326171875,
-      "loss": 0.6863,
-      "rewards/accuracies": 0.71875,
-      "rewards/chosen": 0.009227448143064976,
-      "rewards/margins": 0.01421318482607603,
-      "rewards/rejected": -0.004985739476978779,
+      "epoch": 0.74,
+      "grad_norm": 23.60279896856753,
+      "learning_rate": 1.8825509907063325e-08,
+      "logits/chosen": -2.354218006134033,
+      "logits/rejected": -2.420661449432373,
+      "logps/chosen": -241.6086883544922,
+      "logps/rejected": -207.097900390625,
+      "loss": 0.6725,
+      "rewards/accuracies": 0.824999988079071,
+      "rewards/chosen": 0.04547766596078873,
+      "rewards/margins": 0.04097073897719383,
+      "rewards/rejected": 0.004506924655288458,
       "step": 70
     },
     {
-      "epoch": 0.3,
-      "grad_norm": 38.72478614258279,
-      "learning_rate": 8.871451929520661e-08,
-      "logits/chosen": -0.7630345225334167,
-      "logits/rejected": -1.5050407648086548,
-      "logps/chosen": -25.78152084350586,
-      "logps/rejected": -38.437538146972656,
-      "loss": 0.6839,
-      "rewards/accuracies": 0.800000011920929,
-      "rewards/chosen": 0.013479883782565594,
-      "rewards/margins": 0.018440086394548416,
-      "rewards/rejected": -0.004960204008966684,
+      "epoch": 0.85,
+      "grad_norm": 23.640093404183123,
+      "learning_rate": 6.698729810778064e-09,
+      "logits/chosen": -2.2659950256347656,
+      "logits/rejected": -2.467071056365967,
+      "logps/chosen": -243.5118408203125,
+      "logps/rejected": -205.4396209716797,
+      "loss": 0.6739,
+      "rewards/accuracies": 0.8812500238418579,
+      "rewards/chosen": 0.0479864627122879,
+      "rewards/margins": 0.04288201406598091,
+      "rewards/rejected": 0.005104447714984417,
       "step": 80
     },
     {
-      "epoch": 0.33,
-      "grad_norm": 35.63155458414573,
-      "learning_rate": 8.431208189343668e-08,
-      "logits/chosen": -0.8956977725028992,
-      "logits/rejected": -1.463505506515503,
-      "logps/chosen": -30.033924102783203,
-      "logps/rejected": -56.11266326904297,
-      "loss": 0.6827,
-      "rewards/accuracies": 0.7749999761581421,
-      "rewards/chosen": 0.013467254117131233,
-      "rewards/margins": 0.022690553218126297,
-      "rewards/rejected": -0.00922329816967249,
+      "epoch": 0.95,
+      "grad_norm": 22.97821730211384,
+      "learning_rate": 5.584586887435739e-10,
+      "logits/chosen": -2.3436455726623535,
+      "logits/rejected": -2.367281436920166,
+      "logps/chosen": -226.5314178466797,
+      "logps/rejected": -200.30862426757812,
+      "loss": 0.6745,
+      "rewards/accuracies": 0.8374999761581421,
+      "rewards/chosen": 0.043580617755651474,
+      "rewards/margins": 0.037381939589977264,
+      "rewards/rejected": 0.006198678631335497,
       "step": 90
     },
     {
-      "epoch": 0.37,
-      "grad_norm": 37.21064690820017,
-      "learning_rate": 7.93369417339209e-08,
-      "logits/chosen": -0.7459134459495544,
-      "logits/rejected": -1.5061514377593994,
-      "logps/chosen": -28.494699478149414,
-      "logps/rejected": -49.45507049560547,
-      "loss": 0.6794,
-      "rewards/accuracies": 0.800000011920929,
-      "rewards/chosen": 0.015813838690519333,
-      "rewards/margins": 0.02535676583647728,
-      "rewards/rejected": -0.009542929008603096,
-      "step": 100
-    },
-    {
-      "epoch": 0.37,
-      "eval_logits/chosen": -0.8265538215637207,
-      "eval_logits/rejected": -1.4756767749786377,
-      "eval_logps/chosen": -35.586036682128906,
-      "eval_logps/rejected": -53.076499938964844,
-      "eval_loss": 0.6905515193939209,
-      "eval_rewards/accuracies": 0.5899999737739563,
-      "eval_rewards/chosen": -0.004811165854334831,
-      "eval_rewards/margins": 0.004816049709916115,
-      "eval_rewards/rejected": -0.009627215564250946,
-      "eval_runtime": 65.669,
-      "eval_samples_per_second": 11.954,
-      "eval_steps_per_second": 0.381,
-      "step": 100
-    },
-    {
-      "epoch": 0.41,
-      "grad_norm": 41.489362342224126,
-      "learning_rate": 7.387213885189744e-08,
-      "logits/chosen": -0.9582328796386719,
-      "logits/rejected": -1.4757133722305298,
-      "logps/chosen": -29.652179718017578,
-      "logps/rejected": -51.44434356689453,
-      "loss": 0.6783,
-      "rewards/accuracies": 0.824999988079071,
-      "rewards/chosen": 0.01538840215653181,
-      "rewards/margins": 0.030315887182950974,
-      "rewards/rejected": -0.014927484095096588,
-      "step": 110
-    },
-    {
-      "epoch": 0.44,
-      "grad_norm": 34.811996422375785,
-      "learning_rate": 6.800888624023552e-08,
-      "logits/chosen": -1.1017180681228638,
-      "logits/rejected": -1.2910267114639282,
-      "logps/chosen": -34.17676544189453,
-      "logps/rejected": -47.27839660644531,
-      "loss": 0.6752,
-      "rewards/accuracies": 0.8062499761581421,
-      "rewards/chosen": 0.019322626292705536,
-      "rewards/margins": 0.03619503602385521,
-      "rewards/rejected": -0.01687241531908512,
-      "step": 120
-    },
-    {
-      "epoch": 0.48,
-      "grad_norm": 36.517002002282,
-      "learning_rate": 6.184504741390595e-08,
-      "logits/chosen": -1.0513687133789062,
-      "logits/rejected": -1.4773876667022705,
-      "logps/chosen": -30.05930519104004,
-      "logps/rejected": -49.12663269042969,
-      "loss": 0.6722,
-      "rewards/accuracies": 0.856249988079071,
-      "rewards/chosen": 0.017959367483854294,
-      "rewards/margins": 0.038190923631191254,
-      "rewards/rejected": -0.02023155428469181,
-      "step": 130
-    },
-    {
-      "epoch": 0.52,
-      "grad_norm": 45.92150266428131,
-      "learning_rate": 5.5483502970626584e-08,
-      "logits/chosen": -1.0855646133422852,
-      "logits/rejected": -1.4738510847091675,
-      "logps/chosen": -35.4333610534668,
-      "logps/rejected": -48.60808563232422,
-      "loss": 0.6713,
-      "rewards/accuracies": 0.893750011920929,
-      "rewards/chosen": 0.01625698432326317,
-      "rewards/margins": 0.04006917029619217,
-      "rewards/rejected": -0.023812180384993553,
-      "step": 140
-    },
-    {
-      "epoch": 0.56,
-      "grad_norm": 52.17233360297021,
-      "learning_rate": 4.9030433411408785e-08,
-      "logits/chosen": -1.0967937707901,
-      "logits/rejected": -1.4282653331756592,
-      "logps/chosen": -27.93039321899414,
-      "logps/rejected": -48.556251525878906,
-      "loss": 0.6666,
-      "rewards/accuracies": 0.875,
-      "rewards/chosen": 0.01784222200512886,
-      "rewards/margins": 0.05219421908259392,
-      "rewards/rejected": -0.03435199707746506,
-      "step": 150
-    },
-    {
-      "epoch": 0.59,
-      "grad_norm": 36.1012077969649,
-      "learning_rate": 4.259354688243757e-08,
-      "logits/chosen": -1.088585615158081,
-      "logits/rejected": -1.4759548902511597,
-      "logps/chosen": -30.474609375,
-      "logps/rejected": -50.4084358215332,
-      "loss": 0.668,
-      "rewards/accuracies": 0.824999988079071,
-      "rewards/chosen": 0.007844483479857445,
-      "rewards/margins": 0.05492537096142769,
-      "rewards/rejected": -0.04708089306950569,
-      "step": 160
-    },
-    {
-      "epoch": 0.63,
-      "grad_norm": 39.4057901333892,
-      "learning_rate": 3.628028141903493e-08,
-      "logits/chosen": -1.1803226470947266,
-      "logits/rejected": -1.41080641746521,
-      "logps/chosen": -32.83087158203125,
-      "logps/rejected": -45.69880676269531,
-      "loss": 0.6612,
-      "rewards/accuracies": 0.8187500238418579,
-      "rewards/chosen": 0.01473814807832241,
-      "rewards/margins": 0.06375907361507416,
-      "rewards/rejected": -0.0490209236741066,
-      "step": 170
-    },
-    {
-      "epoch": 0.67,
-      "grad_norm": 38.966283555722626,
-      "learning_rate": 3.0196011698042156e-08,
-      "logits/chosen": -1.0210983753204346,
-      "logits/rejected": -1.4852098226547241,
-      "logps/chosen": -29.679798126220703,
-      "logps/rejected": -50.24894332885742,
-      "loss": 0.6648,
-      "rewards/accuracies": 0.8687499761581421,
-      "rewards/chosen": 0.006426820065826178,
-      "rewards/margins": 0.06615011394023895,
-      "rewards/rejected": -0.059723298996686935,
-      "step": 180
-    },
-    {
-      "epoch": 0.7,
-      "grad_norm": 40.704072746648535,
-      "learning_rate": 2.444229022970634e-08,
-      "logits/chosen": -0.9898138046264648,
-      "logits/rejected": -1.4493308067321777,
-      "logps/chosen": -29.71084976196289,
-      "logps/rejected": -52.1125602722168,
-      "loss": 0.6587,
-      "rewards/accuracies": 0.8187500238418579,
-      "rewards/chosen": 0.005389463622123003,
-      "rewards/margins": 0.06682445108890533,
-      "rewards/rejected": -0.06143498420715332,
-      "step": 190
-    },
-    {
-      "epoch": 0.74,
-      "grad_norm": 48.9298479981709,
-      "learning_rate": 1.9115152345327152e-08,
-      "logits/chosen": -1.0230295658111572,
-      "logits/rejected": -1.5533816814422607,
-      "logps/chosen": -30.142093658447266,
-      "logps/rejected": -58.23160934448242,
-      "loss": 0.6555,
-      "rewards/accuracies": 0.862500011920929,
-      "rewards/chosen": 0.004340417217463255,
-      "rewards/margins": 0.08527159690856934,
-      "rewards/rejected": -0.08093118667602539,
-      "step": 200
-    },
-    {
-      "epoch": 0.74,
-      "eval_logits/chosen": -0.8588806390762329,
-      "eval_logits/rejected": -1.5130012035369873,
-      "eval_logps/chosen": -39.04316711425781,
-      "eval_logps/rejected": -58.72098159790039,
-      "eval_loss": 0.6837000846862793,
-      "eval_rewards/accuracies": 0.6399999856948853,
-      "eval_rewards/chosen": -0.0393824502825737,
-      "eval_rewards/margins": 0.026689546182751656,
-      "eval_rewards/rejected": -0.0660719946026802,
-      "eval_runtime": 65.4972,
-      "eval_samples_per_second": 11.985,
-      "eval_steps_per_second": 0.382,
-      "step": 200
-    },
-    {
-      "epoch": 0.78,
-      "grad_norm": 38.64475256798845,
-      "learning_rate": 1.4303513272105055e-08,
-      "logits/chosen": -0.9661673307418823,
-      "logits/rejected": -1.5310702323913574,
-      "logps/chosen": -31.798084259033203,
-      "logps/rejected": -55.2061882019043,
-      "loss": 0.6543,
-      "rewards/accuracies": 0.856249988079071,
-      "rewards/chosen": 0.008609571494162083,
-      "rewards/margins": 0.0721936896443367,
-      "rewards/rejected": -0.06358411908149719,
-      "step": 210
-    },
-    {
-      "epoch": 0.81,
-      "grad_norm": 46.11579146010557,
-      "learning_rate": 1.0087684049605349e-08,
-      "logits/chosen": -1.0623886585235596,
-      "logits/rejected": -1.484731912612915,
-      "logps/chosen": -33.05571365356445,
-      "logps/rejected": -68.4007568359375,
-      "loss": 0.6512,
-      "rewards/accuracies": 0.84375,
-      "rewards/chosen": 0.002279428532347083,
-      "rewards/margins": 0.09988276660442352,
-      "rewards/rejected": -0.09760333597660065,
-      "step": 220
-    },
-    {
-      "epoch": 0.85,
-      "grad_norm": 42.366279277913364,
-      "learning_rate": 6.538031058667609e-09,
-      "logits/chosen": -0.9526311159133911,
-      "logits/rejected": -1.5703321695327759,
-      "logps/chosen": -32.05732727050781,
-      "logps/rejected": -69.91355895996094,
-      "loss": 0.6539,
-      "rewards/accuracies": 0.8687499761581421,
-      "rewards/chosen": -0.0006113927811384201,
-      "rewards/margins": 0.10779329389333725,
-      "rewards/rejected": -0.10840468108654022,
-      "step": 230
-    },
-    {
-      "epoch": 0.89,
-      "grad_norm": 42.96686513067178,
-      "learning_rate": 3.713801536555483e-09,
-      "logits/chosen": -1.1161487102508545,
-      "logits/rejected": -1.5149171352386475,
-      "logps/chosen": -31.100393295288086,
-      "logps/rejected": -60.296607971191406,
-      "loss": 0.6558,
-      "rewards/accuracies": 0.800000011920929,
-      "rewards/chosen": -0.0010345360497012734,
-      "rewards/margins": 0.09006427228450775,
-      "rewards/rejected": -0.09109880030155182,
-      "step": 240
-    },
-    {
-      "epoch": 0.93,
-      "grad_norm": 46.95737595476474,
-      "learning_rate": 1.662134681666899e-09,
-      "logits/chosen": -0.9997628927230835,
-      "logits/rejected": -1.6180452108383179,
-      "logps/chosen": -27.377477645874023,
-      "logps/rejected": -54.080360412597656,
-      "loss": 0.6512,
-      "rewards/accuracies": 0.8999999761581421,
-      "rewards/chosen": 0.0026857522316277027,
-      "rewards/margins": 0.09887233376502991,
-      "rewards/rejected": -0.0961865708231926,
-      "step": 250
-    },
-    {
-      "epoch": 0.96,
-      "grad_norm": 42.2077659611999,
-      "learning_rate": 4.172748534499448e-10,
-      "logits/chosen": -1.158440113067627,
-      "logits/rejected": -1.4339531660079956,
-      "logps/chosen": -32.833595275878906,
-      "logps/rejected": -55.285552978515625,
-      "loss": 0.6543,
-      "rewards/accuracies": 0.856249988079071,
-      "rewards/chosen": -0.0024300869554281235,
-      "rewards/margins": 0.09633214771747589,
-      "rewards/rejected": -0.09876222908496857,
-      "step": 260
-    },
-    {
-      "epoch": 1.0,
-      "grad_norm": 44.84738059556724,
-      "learning_rate": 0.0,
-      "logits/chosen": -1.0697628259658813,
-      "logits/rejected": -1.563280463218689,
-      "logps/chosen": -33.97581481933594,
-      "logps/rejected": -64.04183197021484,
-      "loss": 0.6482,
-      "rewards/accuracies": 0.8187500238418579,
-      "rewards/chosen": -0.0003928671940229833,
-      "rewards/margins": 0.0867936760187149,
-      "rewards/rejected": -0.0871865525841713,
-      "step": 270
-    },
-    {
-      "epoch": 1.0,
-      "step": 270,
+      "epoch": 0.99,
+      "step": 94,
       "total_flos": 0.0,
-      "train_loss": 0.1692200431117305,
-      "train_runtime": 798.9647,
-      "train_samples_per_second": 21.608,
-      "train_steps_per_second": 0.338
+      "train_loss": 0.6816894855905087,
+      "train_runtime": 1070.9433,
+      "train_samples_per_second": 5.639,
+      "train_steps_per_second": 0.088
     }
   ],
   "logging_steps": 10,
-  "max_steps": 270,
+  "max_steps": 94,
   "num_input_tokens_seen": 0,
   "num_train_epochs": 1,
-  "save_steps": 100,
+  "save_steps": 500,
   "total_flos": 0.0,
   "train_batch_size": 8,
   "trial_name": null,

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:0e423e6281d937957b182db4274e151dd830ea7f134ddb99760f67ce8bed87a0
+oid sha256:5966de10a9274f8fe235eed3be0e3d0ae289568d178ec8f7314f12dcf1ab1f8e
 size 6200