Model save

Browse files

Files changed (11) hide show

README.md +19 -22
all_results.json +5 -18
generation_config.json +1 -1
model-00001-of-00003.safetensors +2 -2
model-00002-of-00003.safetensors +1 -1
model-00003-of-00003.safetensors +2 -2
model.safetensors.index.json +1 -1
runs/May09_18-46-08_n136-144-082/events.out.tfevents.1715252270.n136-144-082.2278777.0 +2 -2
train_results.json +5 -5
trainer_state.json +1428 -628
training_args.bin +1 -1

README.md CHANGED Viewed

@@ -1,16 +1,8 @@
 ---
-license: apache-2.0
-base_model: alignment-handbook/zephyr-7b-sft-full
 tags:
-- alignment-handbook
 - trl
 - dpo
 - generated_from_trainer
-- trl
-- dpo
-- generated_from_trainer
-datasets:
-- HuggingFaceH4/ultrafeedback_binarized
 model-index:
 - name: zephyr-7b-dpo-full
   results: []
@@ -21,17 +13,17 @@ should probably proofread and complete it, then remove this comment. -->
 # zephyr-7b-dpo-full
-This model is a fine-tuned version of [alignment-handbook/zephyr-7b-sft-full](https://huggingface.co/alignment-handbook/zephyr-7b-sft-full) on the HuggingFaceH4/ultrafeedback_binarized dataset.
 It achieves the following results on the evaluation set:
-- Loss: 0.5352
-- Rewards/chosen: 0.5766
-- Rewards/rejected: -0.3207
-- Rewards/accuracies: 0.7617
-- Rewards/margins: 0.8972
-- Logps/rejected: -269.0807
-- Logps/chosen: -251.0624
-- Logits/rejected: -2.4374
-- Logits/chosen: -2.4784
 ## Model description
@@ -68,10 +60,15 @@ The following hyperparameters were used during training:
 | Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
 |:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
-| 0.565         | 0.21  | 100  | 0.5718          | 0.5950         | -0.0056          | 0.7383             | 0.6006          | -262.7792      | -250.6930    | -2.5105         | -2.5504       |
-| 0.5467        | 0.42  | 200  | 0.5433          | 0.6478         | -0.1115          | 0.7461             | 0.7594          | -264.8979      | -249.6371    | -2.4783         | -2.5179       |
-| 0.517         | 0.63  | 300  | 0.5370          | 0.5686         | -0.2689          | 0.7695             | 0.8374          | -268.0445      | -251.2220    | -2.5203         | -2.5623       |
-| 0.518         | 0.84  | 400  | 0.5348          | 0.6286         | -0.2212          | 0.7539             | 0.8498          | -267.0915      | -250.0218    | -2.4324         | -2.4731       |
 ### Framework versions

 ---
 tags:
 - trl
 - dpo
 - generated_from_trainer
 model-index:
 - name: zephyr-7b-dpo-full
   results: []
 # zephyr-7b-dpo-full
+This model was trained from scratch on the None dataset.
 It achieves the following results on the evaluation set:
+- Loss: 1.8979
+- Rewards/chosen: -6.9869
+- Rewards/rejected: -8.4701
+- Rewards/accuracies: 0.6094
+- Rewards/margins: 1.4832
+- Logps/rejected: -1164.5387
+- Logps/chosen: -1010.4669
+- Logits/rejected: -0.5643
+- Logits/chosen: -0.7199
 ## Model description
 | Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
 |:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
+| 0.2555        | 0.1   | 100  | 1.4172          | -4.8884        | -5.6701          | 0.5898             | 0.7817          | -884.5335      | -800.6121    | -1.3358         | -1.3942       |
+| 0.1854        | 0.21  | 200  | 1.6754          | -6.1508        | -7.3259          | 0.6211             | 1.1752          | -1050.1200     | -926.8517    | -1.1088         | -1.1853       |
+| 0.1799        | 0.31  | 300  | 1.5590          | -5.9157        | -6.9794          | 0.5977             | 1.0637          | -1015.4615     | -903.3419    | -1.0193         | -1.1110       |
+| 0.1679        | 0.42  | 400  | 2.1030          | -7.8503        | -9.2060          | 0.6094             | 1.3557          | -1238.1252     | -1096.8108   | -0.5753         | -0.7096       |
+| 0.1693        | 0.52  | 500  | 1.6563          | -6.3408        | -7.6718          | 0.625              | 1.3310          | -1084.7078     | -945.8611    | -0.8598         | -0.9873       |
+| 0.1609        | 0.63  | 600  | 1.6818          | -6.4795        | -7.7992          | 0.6211             | 1.3198          | -1097.4480     | -959.7227    | -0.4515         | -0.6164       |
+| 0.1559        | 0.73  | 700  | 1.9278          | -7.3485        | -8.7955          | 0.6133             | 1.4470          | -1197.0731     | -1046.6217   | -0.4166         | -0.5852       |
+| 0.1433        | 0.84  | 800  | 1.9050          | -7.1496        | -8.6252          | 0.6172             | 1.4756          | -1180.0403     | -1026.7318   | -0.5141         | -0.6745       |
+| 0.1479        | 0.94  | 900  | 1.8979          | -6.9869        | -8.4701          | 0.6094             | 1.4832          | -1164.5387     | -1010.4669   | -0.5643         | -0.7199       |
 ### Framework versions

all_results.json CHANGED Viewed

@@ -1,21 +1,8 @@
 {
     "epoch": 1.0,
-    "eval_logits/chosen": -2.4783902168273926,
-    "eval_logits/rejected": -2.4373888969421387,
-    "eval_logps/chosen": -251.0624237060547,
-    "eval_logps/rejected": -269.08074951171875,
-    "eval_loss": 0.5351515412330627,
-    "eval_rewards/accuracies": 0.76171875,
-    "eval_rewards/chosen": 0.5765520930290222,
-    "eval_rewards/margins": 0.8972306251525879,
-    "eval_rewards/rejected": -0.3206784725189209,
-    "eval_runtime": 96.1948,
-    "eval_samples": 2000,
-    "eval_samples_per_second": 20.791,
-    "eval_steps_per_second": 0.333,
-    "train_loss": 0.5478911828795238,
-    "train_runtime": 7553.9268,
-    "train_samples": 61134,
-    "train_samples_per_second": 8.093,
-    "train_steps_per_second": 0.063
 }

 {
     "epoch": 1.0,
+    "train_loss": 0.1961859940234279,
+    "train_runtime": 15468.9338,
+    "train_samples": 122270,
+    "train_samples_per_second": 7.904,
+    "train_steps_per_second": 0.062
 }

generation_config.json CHANGED Viewed

@@ -1,6 +1,6 @@
 {
   "_from_model_config": true,
   "bos_token_id": 1,
-  "eos_token_id": 2,
   "transformers_version": "4.38.2"
 }

 {
   "_from_model_config": true,
   "bos_token_id": 1,
+  "eos_token_id": 32000,
   "transformers_version": "4.38.2"
 }

model-00001-of-00003.safetensors CHANGED Viewed

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4b668f8f887f1df175ea3fede818f9956dde3a4f82b012ed40e1f75f528510af
-size 4943162336

 version https://git-lfs.github.com/spec/v1
+oid sha256:995180ab1bfccd03f3ed43211962300537543bf349fa09701bdfed2446682de0
+size 4943178720

model-00002-of-00003.safetensors CHANGED Viewed

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5ab93f64cdd679c94fb82526772aa1ddff9407c1f21541bf00dadf1482df2e87
 size 4999819336

 version https://git-lfs.github.com/spec/v1
+oid sha256:1bcfc7b87b51e023f9120c05ca16d79dc2363a9b1014b9e836f1d230e2035fc3
 size 4999819336

model-00003-of-00003.safetensors CHANGED Viewed

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:dabcfcdd26b2c41b402da15a5fc4afd601be2f77a064f1dacd450e2fbd05129f
-size 4540516344

 version https://git-lfs.github.com/spec/v1
+oid sha256:93c22b80739cca4d51c122983391f1fc2fe409c9718974e8823aff56b6815dfe
+size 4540532728

model.safetensors.index.json CHANGED Viewed

@@ -1,6 +1,6 @@
 {
   "metadata": {
-    "total_size": 14483464192
   },
   "weight_map": {
     "lm_head.weight": "model-00003-of-00003.safetensors",

 {
   "metadata": {
+    "total_size": 14483496960
   },
   "weight_map": {
     "lm_head.weight": "model-00003-of-00003.safetensors",

runs/May09_18-46-08_n136-144-082/events.out.tfevents.1715252270.n136-144-082.2278777.0 CHANGED Viewed

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ccd9051f9116773434c18da461972c5355c7205a2dec6e5a1ad38752d7c43032
-size 74047

 version https://git-lfs.github.com/spec/v1
+oid sha256:aa23ceecdb3712ec605ab90931ebf9878e5bbf973580c3b549d65cb4da191613
+size 77841

train_results.json CHANGED Viewed

@@ -1,8 +1,8 @@
 {
     "epoch": 1.0,
-    "train_loss": 0.5478911828795238,
-    "train_runtime": 7553.9268,
-    "train_samples": 61134,
-    "train_samples_per_second": 8.093,
-    "train_steps_per_second": 0.063
 }

 {
     "epoch": 1.0,
+    "train_loss": 0.1961859940234279,
+    "train_runtime": 15468.9338,
+    "train_samples": 122270,
+    "train_samples_per_second": 7.904,
+    "train_steps_per_second": 0.062
 }

trainer_state.json CHANGED Viewed

@@ -1,21 +1,21 @@
 {
   "best_metric": null,
   "best_model_checkpoint": null,
-  "epoch": 1.0,
   "eval_steps": 100,
-  "global_step": 478,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
       "epoch": 0.0,
-      "grad_norm": 49.891043665102934,
-      "learning_rate": 1.0416666666666666e-08,
-      "logits/chosen": -2.7660439014434814,
-      "logits/rejected": -2.717564582824707,
-      "logps/chosen": -269.8568420410156,
-      "logps/rejected": -360.52459716796875,
       "loss": 0.6931,
       "rewards/accuracies": 0.0,
       "rewards/chosen": 0.0,
@@ -23,787 +23,1587 @@
       "rewards/rejected": 0.0,
       "step": 1
     },
     {
       "epoch": 0.02,
-      "grad_norm": 46.946091297352105,
       "learning_rate": 1.0416666666666667e-07,
-      "logits/chosen": -2.592543125152588,
-      "logits/rejected": -2.56319522857666,
-      "logps/chosen": -264.7040100097656,
-      "logps/rejected": -251.515625,
-      "loss": 0.6933,
-      "rewards/accuracies": 0.4791666567325592,
-      "rewards/chosen": 0.004693002440035343,
-      "rewards/margins": 0.0028277651872485876,
-      "rewards/rejected": 0.0018652371363714337,
-      "step": 10
     },
     {
       "epoch": 0.04,
-      "grad_norm": 41.817724108185395,
       "learning_rate": 2.0833333333333333e-07,
-      "logits/chosen": -2.65449595451355,
-      "logits/rejected": -2.6068952083587646,
-      "logps/chosen": -280.5221252441406,
-      "logps/rejected": -295.92376708984375,
-      "loss": 0.689,
-      "rewards/accuracies": 0.581250011920929,
-      "rewards/chosen": 0.05156273767352104,
-      "rewards/margins": 0.00740828737616539,
-      "rewards/rejected": 0.04415445029735565,
-      "step": 20
     },
     {
       "epoch": 0.06,
-      "grad_norm": 39.81553425430633,
       "learning_rate": 3.1249999999999997e-07,
-      "logits/chosen": -2.6671488285064697,
-      "logits/rejected": -2.5955922603607178,
-      "logps/chosen": -296.41644287109375,
-      "logps/rejected": -260.6401672363281,
-      "loss": 0.6733,
-      "rewards/accuracies": 0.59375,
-      "rewards/chosen": 0.2127685844898224,
-      "rewards/margins": 0.04726782441139221,
-      "rewards/rejected": 0.16550076007843018,
-      "step": 30
     },
     {
       "epoch": 0.08,
-      "grad_norm": 38.58454153774096,
       "learning_rate": 4.1666666666666667e-07,
-      "logits/chosen": -2.5658886432647705,
-      "logits/rejected": -2.5324325561523438,
-      "logps/chosen": -259.78594970703125,
-      "logps/rejected": -241.00991821289062,
-      "loss": 0.6399,
-      "rewards/accuracies": 0.7250000238418579,
-      "rewards/chosen": 0.3669721484184265,
-      "rewards/margins": 0.19786901772022247,
-      "rewards/rejected": 0.16910310089588165,
-      "step": 40
     },
     {
       "epoch": 0.1,
-      "grad_norm": 37.351752662935816,
-      "learning_rate": 4.999733114418725e-07,
-      "logits/chosen": -2.5195257663726807,
-      "logits/rejected": -2.4827651977539062,
-      "logps/chosen": -273.65081787109375,
-      "logps/rejected": -290.78680419921875,
-      "loss": 0.6094,
-      "rewards/accuracies": 0.6312500238418579,
-      "rewards/chosen": 0.304054319858551,
-      "rewards/margins": 0.2041884958744049,
-      "rewards/rejected": 0.09986577928066254,
-      "step": 50
     },
     {
       "epoch": 0.13,
-      "grad_norm": 39.61129660699584,
-      "learning_rate": 4.990398100856366e-07,
-      "logits/chosen": -2.567991018295288,
-      "logits/rejected": -2.5036864280700684,
-      "logps/chosen": -260.38055419921875,
-      "logps/rejected": -294.011474609375,
-      "loss": 0.6013,
-      "rewards/accuracies": 0.6937500238418579,
-      "rewards/chosen": 0.5612996220588684,
-      "rewards/margins": 0.3578048348426819,
-      "rewards/rejected": 0.20349478721618652,
-      "step": 60
     },
     {
       "epoch": 0.15,
-      "grad_norm": 41.460696281749556,
-      "learning_rate": 4.967775735898179e-07,
-      "logits/chosen": -2.460195302963257,
-      "logits/rejected": -2.46120023727417,
-      "logps/chosen": -253.1399383544922,
-      "logps/rejected": -253.4242706298828,
-      "loss": 0.5693,
-      "rewards/accuracies": 0.78125,
-      "rewards/chosen": 0.6238263845443726,
-      "rewards/margins": 0.4591788649559021,
-      "rewards/rejected": 0.16464750468730927,
-      "step": 70
     },
     {
       "epoch": 0.17,
-      "grad_norm": 61.37030849441711,
-      "learning_rate": 4.931986719649298e-07,
-      "logits/chosen": -2.615948438644409,
-      "logits/rejected": -2.5394978523254395,
-      "logps/chosen": -311.7240295410156,
-      "logps/rejected": -263.1805725097656,
-      "loss": 0.5671,
-      "rewards/accuracies": 0.6937500238418579,
-      "rewards/chosen": 0.5546952486038208,
-      "rewards/margins": 0.5107932686805725,
-      "rewards/rejected": 0.04390193149447441,
-      "step": 80
     },
     {
       "epoch": 0.19,
-      "grad_norm": 39.59727717104598,
-      "learning_rate": 4.883222001996351e-07,
-      "logits/chosen": -2.5085294246673584,
-      "logits/rejected": -2.4543616771698,
-      "logps/chosen": -251.203369140625,
-      "logps/rejected": -259.8647766113281,
-      "loss": 0.5646,
-      "rewards/accuracies": 0.7562500238418579,
-      "rewards/chosen": 0.3953971564769745,
-      "rewards/margins": 0.7687323689460754,
-      "rewards/rejected": -0.37333518266677856,
-      "step": 90
     },
     {
       "epoch": 0.21,
-      "grad_norm": 36.57841721590594,
-      "learning_rate": 4.821741763807186e-07,
-      "logits/chosen": -2.499514102935791,
-      "logits/rejected": -2.4649369716644287,
-      "logps/chosen": -248.44363403320312,
-      "logps/rejected": -257.64776611328125,
-      "loss": 0.565,
-      "rewards/accuracies": 0.731249988079071,
-      "rewards/chosen": 0.5957446694374084,
-      "rewards/margins": 0.6267115473747253,
-      "rewards/rejected": -0.03096688725054264,
-      "step": 100
     },
     {
       "epoch": 0.21,
-      "eval_logits/chosen": -2.550398111343384,
-      "eval_logits/rejected": -2.5104503631591797,
-      "eval_logps/chosen": -250.69297790527344,
-      "eval_logps/rejected": -262.7791748046875,
-      "eval_loss": 0.5717624425888062,
-      "eval_rewards/accuracies": 0.73828125,
-      "eval_rewards/chosen": 0.5950239300727844,
-      "eval_rewards/margins": 0.6006231904029846,
-      "eval_rewards/rejected": -0.005599223077297211,
-      "eval_runtime": 96.9486,
-      "eval_samples_per_second": 20.629,
-      "eval_steps_per_second": 0.33,
-      "step": 100
     },
     {
       "epoch": 0.23,
-      "grad_norm": 51.91494841998397,
-      "learning_rate": 4.747874028753375e-07,
-      "logits/chosen": -2.55851149559021,
-      "logits/rejected": -2.4656014442443848,
-      "logps/chosen": -292.62615966796875,
-      "logps/rejected": -258.59661865234375,
-      "loss": 0.5713,
-      "rewards/accuracies": 0.75,
-      "rewards/chosen": 0.5976964831352234,
-      "rewards/margins": 0.6357330083847046,
-      "rewards/rejected": -0.0380365327000618,
-      "step": 110
     },
     {
       "epoch": 0.25,
-      "grad_norm": 70.69069363258822,
-      "learning_rate": 4.662012913161997e-07,
-      "logits/chosen": -2.4600424766540527,
-      "logits/rejected": -2.4324684143066406,
-      "logps/chosen": -270.7308349609375,
-      "logps/rejected": -260.5433349609375,
-      "loss": 0.5497,
-      "rewards/accuracies": 0.737500011920929,
-      "rewards/chosen": 0.5218156576156616,
-      "rewards/margins": 0.5561539530754089,
-      "rewards/rejected": -0.03433822840452194,
-      "step": 120
     },
     {
       "epoch": 0.27,
-      "grad_norm": 42.312370253489476,
-      "learning_rate": 4.5646165232345103e-07,
-      "logits/chosen": -2.464791774749756,
-      "logits/rejected": -2.439894676208496,
-      "logps/chosen": -268.9382019042969,
-      "logps/rejected": -269.9627685546875,
-      "loss": 0.5423,
-      "rewards/accuracies": 0.6875,
-      "rewards/chosen": 0.6532469987869263,
-      "rewards/margins": 0.7295945882797241,
-      "rewards/rejected": -0.07634757459163666,
-      "step": 130
     },
     {
       "epoch": 0.29,
-      "grad_norm": 40.45859260542855,
-      "learning_rate": 4.456204510851956e-07,
-      "logits/chosen": -2.5265681743621826,
-      "logits/rejected": -2.485774517059326,
-      "logps/chosen": -303.1440124511719,
-      "logps/rejected": -301.68914794921875,
-      "loss": 0.5376,
-      "rewards/accuracies": 0.768750011920929,
-      "rewards/chosen": 0.6410696506500244,
-      "rewards/margins": 0.6916864514350891,
-      "rewards/rejected": -0.0506168007850647,
-      "step": 140
     },
     {
       "epoch": 0.31,
-      "grad_norm": 41.1747855806655,
-      "learning_rate": 4.337355301007335e-07,
-      "logits/chosen": -2.5189616680145264,
-      "logits/rejected": -2.4531705379486084,
-      "logps/chosen": -272.0736999511719,
-      "logps/rejected": -276.2969055175781,
-      "loss": 0.5442,
-      "rewards/accuracies": 0.737500011920929,
-      "rewards/chosen": 0.5575242042541504,
-      "rewards/margins": 0.5619192719459534,
-      "rewards/rejected": -0.004395070485770702,
-      "step": 150
     },
     {
       "epoch": 0.33,
-      "grad_norm": 48.726180323544725,
-      "learning_rate": 4.2087030056579986e-07,
-      "logits/chosen": -2.5174994468688965,
-      "logits/rejected": -2.43558406829834,
-      "logps/chosen": -260.0892028808594,
-      "logps/rejected": -260.7149658203125,
-      "loss": 0.5652,
-      "rewards/accuracies": 0.737500011920929,
-      "rewards/chosen": 0.6489425897598267,
-      "rewards/margins": 0.8134964108467102,
-      "rewards/rejected": -0.16455380618572235,
-      "step": 160
     },
     {
       "epoch": 0.36,
-      "grad_norm": 49.53825953706789,
-      "learning_rate": 4.070934040463998e-07,
-      "logits/chosen": -2.4509148597717285,
-      "logits/rejected": -2.3897039890289307,
-      "logps/chosen": -239.52261352539062,
-      "logps/rejected": -233.6277618408203,
-      "loss": 0.5489,
-      "rewards/accuracies": 0.7250000238418579,
-      "rewards/chosen": 0.49063143134117126,
-      "rewards/margins": 0.6157802939414978,
-      "rewards/rejected": -0.12514881789684296,
-      "step": 170
     },
     {
       "epoch": 0.38,
-      "grad_norm": 51.08561061111303,
-      "learning_rate": 3.9247834624635404e-07,
-      "logits/chosen": -2.3483898639678955,
-      "logits/rejected": -2.306784152984619,
-      "logps/chosen": -247.6396026611328,
-      "logps/rejected": -231.8523406982422,
-      "loss": 0.5181,
-      "rewards/accuracies": 0.7250000238418579,
-      "rewards/chosen": 0.43596941232681274,
-      "rewards/margins": 0.6500319242477417,
-      "rewards/rejected": -0.21406252682209015,
-      "step": 180
     },
     {
       "epoch": 0.4,
-      "grad_norm": 42.31027201995276,
-      "learning_rate": 3.7710310482256523e-07,
-      "logits/chosen": -2.41634464263916,
-      "logits/rejected": -2.378105878829956,
-      "logps/chosen": -260.20306396484375,
-      "logps/rejected": -261.46502685546875,
-      "loss": 0.5392,
-      "rewards/accuracies": 0.7124999761581421,
-      "rewards/chosen": 0.5519876480102539,
-      "rewards/margins": 0.6375012993812561,
-      "rewards/rejected": -0.08551368862390518,
-      "step": 190
     },
     {
       "epoch": 0.42,
-      "grad_norm": 102.86207924802177,
-      "learning_rate": 3.610497133404795e-07,
-      "logits/chosen": -2.392763614654541,
-      "logits/rejected": -2.381993532180786,
-      "logps/chosen": -249.912109375,
-      "logps/rejected": -256.75439453125,
-      "loss": 0.5467,
-      "rewards/accuracies": 0.731249988079071,
-      "rewards/chosen": 0.49922746419906616,
-      "rewards/margins": 0.7344967126846313,
-      "rewards/rejected": -0.235269233584404,
-      "step": 200
     },
     {
       "epoch": 0.42,
-      "eval_logits/chosen": -2.517864942550659,
-      "eval_logits/rejected": -2.4783387184143066,
-      "eval_logps/chosen": -249.6370849609375,
-      "eval_logps/rejected": -264.89788818359375,
-      "eval_loss": 0.5432960391044617,
-      "eval_rewards/accuracies": 0.74609375,
-      "eval_rewards/chosen": 0.6478186845779419,
-      "eval_rewards/margins": 0.759353518486023,
-      "eval_rewards/rejected": -0.11153475195169449,
-      "eval_runtime": 96.4207,
-      "eval_samples_per_second": 20.742,
-      "eval_steps_per_second": 0.332,
-      "step": 200
     },
     {
       "epoch": 0.44,
-      "grad_norm": 45.308290366409736,
-      "learning_rate": 3.4440382358952115e-07,
-      "logits/chosen": -2.4460113048553467,
-      "logits/rejected": -2.391810894012451,
-      "logps/chosen": -278.56781005859375,
-      "logps/rejected": -257.2254943847656,
-      "loss": 0.5436,
-      "rewards/accuracies": 0.731249988079071,
-      "rewards/chosen": 0.5562152862548828,
-      "rewards/margins": 0.8551079034805298,
-      "rewards/rejected": -0.29889267683029175,
-      "step": 210
     },
     {
       "epoch": 0.46,
-      "grad_norm": 50.1182470431882,
-      "learning_rate": 3.272542485937368e-07,
-      "logits/chosen": -2.4605488777160645,
-      "logits/rejected": -2.42708683013916,
-      "logps/chosen": -257.90826416015625,
-      "logps/rejected": -253.3182830810547,
-      "loss": 0.54,
-      "rewards/accuracies": 0.793749988079071,
-      "rewards/chosen": 0.3734419643878937,
-      "rewards/margins": 0.7561658024787903,
-      "rewards/rejected": -0.382723867893219,
-      "step": 220
     },
     {
       "epoch": 0.48,
-      "grad_norm": 43.71024962971359,
-      "learning_rate": 3.096924887558854e-07,
-      "logits/chosen": -2.490509510040283,
-      "logits/rejected": -2.4491913318634033,
-      "logps/chosen": -237.17898559570312,
-      "logps/rejected": -251.81686401367188,
-      "loss": 0.5441,
-      "rewards/accuracies": 0.7437499761581421,
-      "rewards/chosen": 0.5843235850334167,
-      "rewards/margins": 0.7882751226425171,
-      "rewards/rejected": -0.20395155251026154,
-      "step": 230
     },
     {
       "epoch": 0.5,
-      "grad_norm": 44.93616969967234,
-      "learning_rate": 2.9181224366319943e-07,
-      "logits/chosen": -2.533695697784424,
-      "logits/rejected": -2.500807285308838,
-      "logps/chosen": -253.635498046875,
-      "logps/rejected": -253.0944061279297,
-      "loss": 0.5142,
-      "rewards/accuracies": 0.699999988079071,
-      "rewards/chosen": 0.5032340884208679,
-      "rewards/margins": 0.7045356035232544,
-      "rewards/rejected": -0.2013014256954193,
-      "step": 240
     },
     {
       "epoch": 0.52,
-      "grad_norm": 42.68904256130122,
-      "learning_rate": 2.7370891215954565e-07,
-      "logits/chosen": -2.483025074005127,
-      "logits/rejected": -2.4015185832977295,
-      "logps/chosen": -285.0963439941406,
-      "logps/rejected": -263.43560791015625,
-      "loss": 0.5198,
-      "rewards/accuracies": 0.762499988079071,
-      "rewards/chosen": 0.6547069549560547,
-      "rewards/margins": 0.850358784198761,
-      "rewards/rejected": -0.19565197825431824,
-      "step": 250
     },
     {
       "epoch": 0.54,
-      "grad_norm": 45.43502171857602,
-      "learning_rate": 2.55479083351317e-07,
-      "logits/chosen": -2.516913890838623,
-      "logits/rejected": -2.478473424911499,
-      "logps/chosen": -282.80230712890625,
-      "logps/rejected": -258.77288818359375,
-      "loss": 0.5235,
-      "rewards/accuracies": 0.768750011920929,
-      "rewards/chosen": 0.5754625797271729,
-      "rewards/margins": 0.8150871396064758,
-      "rewards/rejected": -0.23962458968162537,
-      "step": 260
     },
     {
-      "epoch": 0.56,
-      "grad_norm": 41.73526734917468,
-      "learning_rate": 2.3722002126275822e-07,
-      "logits/chosen": -2.5381340980529785,
-      "logits/rejected": -2.4941086769104004,
-      "logps/chosen": -267.4333190917969,
-      "logps/rejected": -260.50677490234375,
-      "loss": 0.5406,
-      "rewards/accuracies": 0.675000011920929,
-      "rewards/chosen": 0.4781308174133301,
-      "rewards/margins": 0.6212563514709473,
-      "rewards/rejected": -0.14312560856342316,
-      "step": 270
     },
     {
       "epoch": 0.59,
-      "grad_norm": 48.561323508433155,
-      "learning_rate": 2.19029145890313e-07,
-      "logits/chosen": -2.510133743286133,
-      "logits/rejected": -2.4422435760498047,
-      "logps/chosen": -250.73855590820312,
-      "logps/rejected": -247.487060546875,
-      "loss": 0.5599,
-      "rewards/accuracies": 0.737500011920929,
-      "rewards/chosen": 0.5469261407852173,
-      "rewards/margins": 0.8119627833366394,
-      "rewards/rejected": -0.2650366425514221,
-      "step": 280
     },
     {
       "epoch": 0.61,
-      "grad_norm": 44.504093632124075,
-      "learning_rate": 2.0100351342479216e-07,
-      "logits/chosen": -2.5589568614959717,
-      "logits/rejected": -2.5217483043670654,
-      "logps/chosen": -240.7520751953125,
-      "logps/rejected": -244.8422088623047,
-      "loss": 0.5354,
-      "rewards/accuracies": 0.7250000238418579,
-      "rewards/chosen": 0.5167636871337891,
-      "rewards/margins": 0.661081075668335,
-      "rewards/rejected": -0.14431743323802948,
-      "step": 290
     },
     {
       "epoch": 0.63,
-      "grad_norm": 52.52022669231452,
-      "learning_rate": 1.8323929841460178e-07,
-      "logits/chosen": -2.5429511070251465,
-      "logits/rejected": -2.472679376602173,
-      "logps/chosen": -292.2240905761719,
-      "logps/rejected": -266.68658447265625,
-      "loss": 0.517,
-      "rewards/accuracies": 0.731249988079071,
-      "rewards/chosen": 0.4385985732078552,
-      "rewards/margins": 0.7676541209220886,
-      "rewards/rejected": -0.329055517911911,
-      "step": 300
     },
     {
       "epoch": 0.63,
-      "eval_logits/chosen": -2.5622596740722656,
-      "eval_logits/rejected": -2.520256280899048,
-      "eval_logps/chosen": -251.2219696044922,
-      "eval_logps/rejected": -268.04449462890625,
-      "eval_loss": 0.53697669506073,
-      "eval_rewards/accuracies": 0.76953125,
-      "eval_rewards/chosen": 0.5685745477676392,
-      "eval_rewards/margins": 0.8374388217926025,
-      "eval_rewards/rejected": -0.2688642740249634,
-      "eval_runtime": 96.3678,
-      "eval_samples_per_second": 20.754,
-      "eval_steps_per_second": 0.332,
-      "step": 300
     },
     {
       "epoch": 0.65,
-      "grad_norm": 43.866661437938184,
-      "learning_rate": 1.6583128063291573e-07,
-      "logits/chosen": -2.4593474864959717,
-      "logits/rejected": -2.443233013153076,
-      "logps/chosen": -285.5498046875,
-      "logps/rejected": -263.8379821777344,
-      "loss": 0.5077,
-      "rewards/accuracies": 0.768750011920929,
-      "rewards/chosen": 0.4893050193786621,
-      "rewards/margins": 0.7553777098655701,
-      "rewards/rejected": -0.26607269048690796,
-      "step": 310
     },
     {
       "epoch": 0.67,
-      "grad_norm": 43.407860217947494,
-      "learning_rate": 1.488723393865766e-07,
-      "logits/chosen": -2.4746253490448,
-      "logits/rejected": -2.4388270378112793,
-      "logps/chosen": -283.4583740234375,
-      "logps/rejected": -250.38204956054688,
-      "loss": 0.504,
-      "rewards/accuracies": 0.731249988079071,
-      "rewards/chosen": 0.5105848908424377,
-      "rewards/margins": 0.788524329662323,
-      "rewards/rejected": -0.2779393792152405,
-      "step": 320
     },
     {
       "epoch": 0.69,
-      "grad_norm": 40.302692173545196,
-      "learning_rate": 1.3245295796480788e-07,
-      "logits/chosen": -2.4712371826171875,
-      "logits/rejected": -2.4099698066711426,
-      "logps/chosen": -252.349853515625,
-      "logps/rejected": -264.03912353515625,
-      "loss": 0.5242,
-      "rewards/accuracies": 0.737500011920929,
-      "rewards/chosen": 0.4930170178413391,
-      "rewards/margins": 0.7200408577919006,
-      "rewards/rejected": -0.2270239144563675,
-      "step": 330
     },
     {
       "epoch": 0.71,
-      "grad_norm": 50.168955016672676,
-      "learning_rate": 1.1666074087171627e-07,
-      "logits/chosen": -2.467729091644287,
-      "logits/rejected": -2.4046943187713623,
-      "logps/chosen": -278.697509765625,
-      "logps/rejected": -285.4507141113281,
-      "loss": 0.524,
-      "rewards/accuracies": 0.7437499761581421,
-      "rewards/chosen": 0.5648467540740967,
-      "rewards/margins": 0.8352931141853333,
-      "rewards/rejected": -0.2704463601112366,
-      "step": 340
     },
     {
       "epoch": 0.73,
-      "grad_norm": 46.15971070553052,
-      "learning_rate": 1.0157994641835734e-07,
-      "logits/chosen": -2.445666790008545,
-      "logits/rejected": -2.377004384994507,
-      "logps/chosen": -248.63241577148438,
-      "logps/rejected": -248.23904418945312,
-      "loss": 0.4924,
-      "rewards/accuracies": 0.75,
-      "rewards/chosen": 0.47894006967544556,
-      "rewards/margins": 0.8554509878158569,
-      "rewards/rejected": -0.37651100754737854,
-      "step": 350
     },
     {
       "epoch": 0.75,
-      "grad_norm": 54.17198760484943,
-      "learning_rate": 8.729103716819111e-08,
-      "logits/chosen": -2.4745380878448486,
-      "logits/rejected": -2.376185178756714,
-      "logps/chosen": -292.89483642578125,
-      "logps/rejected": -269.1952209472656,
-      "loss": 0.5388,
-      "rewards/accuracies": 0.7562500238418579,
-      "rewards/chosen": 0.4898607134819031,
-      "rewards/margins": 0.8843740224838257,
-      "rewards/rejected": -0.3945133090019226,
-      "step": 360
     },
     {
       "epoch": 0.77,
-      "grad_norm": 44.15468601338237,
-      "learning_rate": 7.387025063449081e-08,
-      "logits/chosen": -2.409170150756836,
-      "logits/rejected": -2.367518186569214,
-      "logps/chosen": -266.35430908203125,
-      "logps/rejected": -242.5480194091797,
-      "loss": 0.5384,
-      "rewards/accuracies": 0.6875,
-      "rewards/chosen": 0.44893550872802734,
-      "rewards/margins": 0.6646324992179871,
-      "rewards/rejected": -0.21569697558879852,
-      "step": 370
     },
     {
-      "epoch": 0.79,
-      "grad_norm": 39.47383320898196,
-      "learning_rate": 6.138919252022435e-08,
-      "logits/chosen": -2.3523006439208984,
-      "logits/rejected": -2.3420968055725098,
-      "logps/chosen": -230.9795379638672,
-      "logps/rejected": -267.8912658691406,
-      "loss": 0.5181,
-      "rewards/accuracies": 0.737500011920929,
-      "rewards/chosen": 0.4252557158470154,
-      "rewards/margins": 0.8839966058731079,
-      "rewards/rejected": -0.45874080061912537,
-      "step": 380
     },
     {
       "epoch": 0.82,
-      "grad_norm": 48.64961363299689,
-      "learning_rate": 4.991445467064689e-08,
-      "logits/chosen": -2.4286305904388428,
-      "logits/rejected": -2.394604206085205,
-      "logps/chosen": -293.20440673828125,
-      "logps/rejected": -287.0997009277344,
-      "loss": 0.5149,
-      "rewards/accuracies": 0.7124999761581421,
-      "rewards/chosen": 0.6216251850128174,
-      "rewards/margins": 0.7780872583389282,
-      "rewards/rejected": -0.15646204352378845,
-      "step": 390
     },
     {
       "epoch": 0.84,
-      "grad_norm": 44.38080817079193,
-      "learning_rate": 3.9507259776993954e-08,
-      "logits/chosen": -2.4102301597595215,
-      "logits/rejected": -2.3357295989990234,
-      "logps/chosen": -259.7147521972656,
-      "logps/rejected": -273.10699462890625,
-      "loss": 0.518,
-      "rewards/accuracies": 0.762499988079071,
-      "rewards/chosen": 0.541024386882782,
-      "rewards/margins": 0.8457515835762024,
-      "rewards/rejected": -0.30472710728645325,
-      "step": 400
     },
     {
       "epoch": 0.84,
-      "eval_logits/chosen": -2.4731171131134033,
-      "eval_logits/rejected": -2.4323782920837402,
-      "eval_logps/chosen": -250.02178955078125,
-      "eval_logps/rejected": -267.0915222167969,
-      "eval_loss": 0.5348160862922668,
-      "eval_rewards/accuracies": 0.75390625,
-      "eval_rewards/chosen": 0.6285843849182129,
-      "eval_rewards/margins": 0.8498014211654663,
-      "eval_rewards/rejected": -0.22121697664260864,
-      "eval_runtime": 96.4764,
-      "eval_samples_per_second": 20.73,
-      "eval_steps_per_second": 0.332,
-      "step": 400
     },
     {
       "epoch": 0.86,
-      "grad_norm": 48.15981350653104,
-      "learning_rate": 3.022313472693447e-08,
-      "logits/chosen": -2.444577932357788,
-      "logits/rejected": -2.3699073791503906,
-      "logps/chosen": -286.5138854980469,
-      "logps/rejected": -274.3666687011719,
-      "loss": 0.5226,
-      "rewards/accuracies": 0.768750011920929,
-      "rewards/chosen": 0.6201778650283813,
-      "rewards/margins": 0.8976529240608215,
-      "rewards/rejected": -0.2774750590324402,
-      "step": 410
     },
     {
       "epoch": 0.88,
-      "grad_norm": 48.573506313099124,
-      "learning_rate": 2.2111614344599684e-08,
-      "logits/chosen": -2.429912805557251,
-      "logits/rejected": -2.3931796550750732,
-      "logps/chosen": -287.13067626953125,
-      "logps/rejected": -279.46844482421875,
-      "loss": 0.5212,
-      "rewards/accuracies": 0.675000011920929,
-      "rewards/chosen": 0.4971562325954437,
-      "rewards/margins": 0.7474662065505981,
-      "rewards/rejected": -0.25030994415283203,
-      "step": 420
     },
     {
       "epoch": 0.9,
-      "grad_norm": 41.98038749926915,
-      "learning_rate": 1.521597710086439e-08,
-      "logits/chosen": -2.3573684692382812,
-      "logits/rejected": -2.3092567920684814,
-      "logps/chosen": -269.9436950683594,
-      "logps/rejected": -265.4564514160156,
-      "loss": 0.501,
-      "rewards/accuracies": 0.7875000238418579,
-      "rewards/chosen": 0.45472264289855957,
-      "rewards/margins": 0.838543713092804,
-      "rewards/rejected": -0.38382115960121155,
-      "step": 430
     },
     {
       "epoch": 0.92,
-      "grad_norm": 44.22650678163462,
-      "learning_rate": 9.57301420397924e-09,
-      "logits/chosen": -2.4332785606384277,
-      "logits/rejected": -2.3776473999023438,
-      "logps/chosen": -272.65960693359375,
-      "logps/rejected": -271.44329833984375,
-      "loss": 0.5213,
-      "rewards/accuracies": 0.6875,
-      "rewards/chosen": 0.5838757157325745,
-      "rewards/margins": 0.700454831123352,
-      "rewards/rejected": -0.1165790781378746,
-      "step": 440
     },
     {
       "epoch": 0.94,
-      "grad_norm": 43.00589727019739,
-      "learning_rate": 5.212833302556258e-09,
-      "logits/chosen": -2.3836779594421387,
-      "logits/rejected": -2.360665798187256,
-      "logps/chosen": -284.2134704589844,
-      "logps/rejected": -312.9830627441406,
-      "loss": 0.5099,
-      "rewards/accuracies": 0.6812499761581421,
-      "rewards/chosen": 0.4735330641269684,
-      "rewards/margins": 0.7689631581306458,
-      "rewards/rejected": -0.29543009400367737,
-      "step": 450
     },
     {
       "epoch": 0.96,
-      "grad_norm": 46.86754726240038,
-      "learning_rate": 2.158697848236607e-09,
-      "logits/chosen": -2.417273998260498,
-      "logits/rejected": -2.377349376678467,
-      "logps/chosen": -262.1804504394531,
-      "logps/rejected": -247.7431182861328,
-      "loss": 0.5264,
-      "rewards/accuracies": 0.699999988079071,
-      "rewards/chosen": 0.48920711874961853,
-      "rewards/margins": 0.7419728636741638,
-      "rewards/rejected": -0.2527657151222229,
-      "step": 460
     },
     {
       "epoch": 0.98,
-      "grad_norm": 45.055740606082026,
-      "learning_rate": 4.269029751107489e-10,
-      "logits/chosen": -2.4338390827178955,
-      "logits/rejected": -2.3758208751678467,
-      "logps/chosen": -268.4836730957031,
-      "logps/rejected": -289.60205078125,
-      "loss": 0.4974,
-      "rewards/accuracies": 0.762499988079071,
-      "rewards/chosen": 0.5107508897781372,
-      "rewards/margins": 0.8732994794845581,
-      "rewards/rejected": -0.3625485301017761,
-      "step": 470
     },
     {
       "epoch": 1.0,
-      "step": 478,
       "total_flos": 0.0,
-      "train_loss": 0.5478911828795238,
-      "train_runtime": 7553.9268,
-      "train_samples_per_second": 8.093,
-      "train_steps_per_second": 0.063
     }
   ],
   "logging_steps": 10,
-  "max_steps": 478,
   "num_input_tokens_seen": 0,
   "num_train_epochs": 1,
   "save_steps": 100,

 {
   "best_metric": null,
   "best_model_checkpoint": null,
+  "epoch": 0.9994767137624281,
   "eval_steps": 100,
+  "global_step": 955,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
       "epoch": 0.0,
+      "grad_norm": 50.348802908690914,
+      "learning_rate": 5.208333333333333e-09,
+      "logits/chosen": -1.8382859230041504,
+      "logits/rejected": -1.788834810256958,
+      "logps/chosen": -119.0692138671875,
+      "logps/rejected": -76.35714721679688,
       "loss": 0.6931,
       "rewards/accuracies": 0.0,
       "rewards/chosen": 0.0,
       "rewards/rejected": 0.0,
       "step": 1
     },
+    {
+      "epoch": 0.01,
+      "grad_norm": 46.56256136532365,
+      "learning_rate": 5.208333333333333e-08,
+      "logits/chosen": -1.6623330116271973,
+      "logits/rejected": -1.5512146949768066,
+      "logps/chosen": -129.29541015625,
+      "logps/rejected": -82.82817840576172,
+      "loss": 0.6931,
+      "rewards/accuracies": 0.4375,
+      "rewards/chosen": -0.00018554604321252555,
+      "rewards/margins": 0.00036421266850084066,
+      "rewards/rejected": -0.0005497586098499596,
+      "step": 10
+    },
     {
       "epoch": 0.02,
+      "grad_norm": 47.9759657771044,
       "learning_rate": 1.0416666666666667e-07,
+      "logits/chosen": -1.6969773769378662,
+      "logits/rejected": -1.5268709659576416,
+      "logps/chosen": -140.09909057617188,
+      "logps/rejected": -80.9607162475586,
+      "loss": 0.6866,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": 0.013138635084033012,
+      "rewards/margins": 0.015350555069744587,
+      "rewards/rejected": -0.002211919752880931,
+      "step": 20
+    },
+    {
+      "epoch": 0.03,
+      "grad_norm": 37.210502761575135,
+      "learning_rate": 1.5624999999999999e-07,
+      "logits/chosen": -1.7122924327850342,
+      "logits/rejected": -1.6340242624282837,
+      "logps/chosen": -119.13212585449219,
+      "logps/rejected": -84.83552551269531,
+      "loss": 0.657,
+      "rewards/accuracies": 0.862500011920929,
+      "rewards/chosen": 0.05337844043970108,
+      "rewards/margins": 0.0730198472738266,
+      "rewards/rejected": -0.01964140310883522,
+      "step": 30
     },
     {
       "epoch": 0.04,
+      "grad_norm": 38.29612001708921,
       "learning_rate": 2.0833333333333333e-07,
+      "logits/chosen": -1.8123095035552979,
+      "logits/rejected": -1.716663122177124,
+      "logps/chosen": -130.39593505859375,
+      "logps/rejected": -98.6071548461914,
+      "loss": 0.5915,
+      "rewards/accuracies": 0.90625,
+      "rewards/chosen": 0.1293005496263504,
+      "rewards/margins": 0.2863259017467499,
+      "rewards/rejected": -0.15702535212039948,
+      "step": 40
+    },
+    {
+      "epoch": 0.05,
+      "grad_norm": 25.169270167372616,
+      "learning_rate": 2.604166666666667e-07,
+      "logits/chosen": -1.6766433715820312,
+      "logits/rejected": -1.6306852102279663,
+      "logps/chosen": -127.97342681884766,
+      "logps/rejected": -136.70712280273438,
+      "loss": 0.484,
+      "rewards/accuracies": 0.8812500238418579,
+      "rewards/chosen": 0.0137257669121027,
+      "rewards/margins": 0.6059251427650452,
+      "rewards/rejected": -0.5921992063522339,
+      "step": 50
     },
     {
       "epoch": 0.06,
+      "grad_norm": 26.650306729386905,
       "learning_rate": 3.1249999999999997e-07,
+      "logits/chosen": -1.6139761209487915,
+      "logits/rejected": -1.610743761062622,
+      "logps/chosen": -173.83758544921875,
+      "logps/rejected": -218.09707641601562,
+      "loss": 0.3946,
+      "rewards/accuracies": 0.887499988079071,
+      "rewards/chosen": -0.24665935337543488,
+      "rewards/margins": 1.0333704948425293,
+      "rewards/rejected": -1.280029535293579,
+      "step": 60
+    },
+    {
+      "epoch": 0.07,
+      "grad_norm": 32.401140738769385,
+      "learning_rate": 3.645833333333333e-07,
+      "logits/chosen": -1.4975354671478271,
+      "logits/rejected": -1.4718661308288574,
+      "logps/chosen": -191.6230010986328,
+      "logps/rejected": -297.1890563964844,
+      "loss": 0.3474,
+      "rewards/accuracies": 0.84375,
+      "rewards/chosen": -0.6469367146492004,
+      "rewards/margins": 1.586439847946167,
+      "rewards/rejected": -2.2333762645721436,
+      "step": 70
     },
     {
       "epoch": 0.08,
+      "grad_norm": 31.488475681745037,
       "learning_rate": 4.1666666666666667e-07,
+      "logits/chosen": -1.4592185020446777,
+      "logits/rejected": -1.3416706323623657,
+      "logps/chosen": -249.4734649658203,
+      "logps/rejected": -405.06207275390625,
+      "loss": 0.2966,
+      "rewards/accuracies": 0.8812500238418579,
+      "rewards/chosen": -1.1794850826263428,
+      "rewards/margins": 2.0616517066955566,
+      "rewards/rejected": -3.2411365509033203,
+      "step": 80
+    },
+    {
+      "epoch": 0.09,
+      "grad_norm": 34.896761047309674,
+      "learning_rate": 4.6874999999999996e-07,
+      "logits/chosen": -1.2935254573822021,
+      "logits/rejected": -1.215850591659546,
+      "logps/chosen": -236.42935180664062,
+      "logps/rejected": -427.88079833984375,
+      "loss": 0.2792,
+      "rewards/accuracies": 0.862500011920929,
+      "rewards/chosen": -1.0514061450958252,
+      "rewards/margins": 2.4284262657165527,
+      "rewards/rejected": -3.479832410812378,
+      "step": 90
     },
     {
       "epoch": 0.1,
+      "grad_norm": 40.95979792917679,
+      "learning_rate": 4.999732492681437e-07,
+      "logits/chosen": -1.425644040107727,
+      "logits/rejected": -1.2574630975723267,
+      "logps/chosen": -236.2029571533203,
+      "logps/rejected": -470.1878967285156,
+      "loss": 0.2555,
+      "rewards/accuracies": 0.893750011920929,
+      "rewards/chosen": -1.0542919635772705,
+      "rewards/margins": 2.8619871139526367,
+      "rewards/rejected": -3.916278839111328,
+      "step": 100
+    },
+    {
+      "epoch": 0.1,
+      "eval_logits/chosen": -1.394241213798523,
+      "eval_logits/rejected": -1.3358198404312134,
+      "eval_logps/chosen": -800.6121215820312,
+      "eval_logps/rejected": -884.5335083007812,
+      "eval_loss": 1.4172171354293823,
+      "eval_rewards/accuracies": 0.58984375,
+      "eval_rewards/chosen": -4.888356685638428,
+      "eval_rewards/margins": 0.7817266583442688,
+      "eval_rewards/rejected": -5.670083045959473,
+      "eval_runtime": 97.8793,
+      "eval_samples_per_second": 20.433,
+      "eval_steps_per_second": 0.327,
+      "step": 100
+    },
+    {
+      "epoch": 0.12,
+      "grad_norm": 32.26210278499666,
+      "learning_rate": 4.996723692767926e-07,
+      "logits/chosen": -1.300011157989502,
+      "logits/rejected": -1.1855316162109375,
+      "logps/chosen": -253.3606719970703,
+      "logps/rejected": -546.6198120117188,
+      "loss": 0.2322,
+      "rewards/accuracies": 0.862500011920929,
+      "rewards/chosen": -1.3233938217163086,
+      "rewards/margins": 3.418271541595459,
+      "rewards/rejected": -4.741665840148926,
+      "step": 110
     },
     {
       "epoch": 0.13,
+      "grad_norm": 47.80125591905201,
+      "learning_rate": 4.990375746213598e-07,
+      "logits/chosen": -1.4404656887054443,
+      "logits/rejected": -1.311632513999939,
+      "logps/chosen": -286.67169189453125,
+      "logps/rejected": -644.265625,
+      "loss": 0.2299,
+      "rewards/accuracies": 0.893750011920929,
+      "rewards/chosen": -1.5356190204620361,
+      "rewards/margins": 4.171111583709717,
+      "rewards/rejected": -5.706730842590332,
+      "step": 120
+    },
+    {
+      "epoch": 0.14,
+      "grad_norm": 103.21873941787233,
+      "learning_rate": 4.980697142834314e-07,
+      "logits/chosen": -1.4422317743301392,
+      "logits/rejected": -1.295238733291626,
+      "logps/chosen": -258.7148742675781,
+      "logps/rejected": -593.8055419921875,
+      "loss": 0.2288,
+      "rewards/accuracies": 0.875,
+      "rewards/chosen": -1.2843689918518066,
+      "rewards/margins": 3.8400402069091797,
+      "rewards/rejected": -5.1244096755981445,
+      "step": 130
     },
     {
       "epoch": 0.15,
+      "grad_norm": 40.83439235144051,
+      "learning_rate": 4.967700826904229e-07,
+      "logits/chosen": -1.4604966640472412,
+      "logits/rejected": -1.289422631263733,
+      "logps/chosen": -291.0477600097656,
+      "logps/rejected": -753.3111572265625,
+      "loss": 0.2219,
+      "rewards/accuracies": 0.8687499761581421,
+      "rewards/chosen": -1.6053016185760498,
+      "rewards/margins": 5.150030612945557,
+      "rewards/rejected": -6.755332946777344,
+      "step": 140
+    },
+    {
+      "epoch": 0.16,
+      "grad_norm": 35.47272090007365,
+      "learning_rate": 4.951404179843962e-07,
+      "logits/chosen": -1.439012885093689,
+      "logits/rejected": -1.197693109512329,
+      "logps/chosen": -317.4605407714844,
+      "logps/rejected": -829.1416015625,
+      "loss": 0.1987,
+      "rewards/accuracies": 0.887499988079071,
+      "rewards/chosen": -1.6784883737564087,
+      "rewards/margins": 5.721285820007324,
+      "rewards/rejected": -7.399774074554443,
+      "step": 150
     },
     {
       "epoch": 0.17,
+      "grad_norm": 41.95405531315011,
+      "learning_rate": 4.931828996974498e-07,
+      "logits/chosen": -1.3928442001342773,
+      "logits/rejected": -1.1220872402191162,
+      "logps/chosen": -289.9758605957031,
+      "logps/rejected": -711.052001953125,
+      "loss": 0.21,
+      "rewards/accuracies": 0.875,
+      "rewards/chosen": -1.6032127141952515,
+      "rewards/margins": 4.786148548126221,
+      "rewards/rejected": -6.389361381530762,
+      "step": 160
+    },
+    {
+      "epoch": 0.18,
+      "grad_norm": 44.30975250132851,
+      "learning_rate": 4.909001458367866e-07,
+      "logits/chosen": -1.4432841539382935,
+      "logits/rejected": -1.237818956375122,
+      "logps/chosen": -394.95751953125,
+      "logps/rejected": -859.7473754882812,
+      "loss": 0.1885,
+      "rewards/accuracies": 0.856249988079071,
+      "rewards/chosen": -2.7817580699920654,
+      "rewards/margins": 5.132468223571777,
+      "rewards/rejected": -7.914226531982422,
+      "step": 170
     },
     {
       "epoch": 0.19,
+      "grad_norm": 42.31623927536496,
+      "learning_rate": 4.882952093833627e-07,
+      "logits/chosen": -1.450234055519104,
+      "logits/rejected": -1.2738139629364014,
+      "logps/chosen": -342.06048583984375,
+      "logps/rejected": -785.0755004882812,
+      "loss": 0.2045,
+      "rewards/accuracies": 0.8687499761581421,
+      "rewards/chosen": -2.051614999771118,
+      "rewards/margins": 4.953377723693848,
+      "rewards/rejected": -7.0049920082092285,
+      "step": 180
+    },
+    {
+      "epoch": 0.2,
+      "grad_norm": 39.855754537956905,
+      "learning_rate": 4.853715742087946e-07,
+      "logits/chosen": -1.3248379230499268,
+      "logits/rejected": -1.09066903591156,
+      "logps/chosen": -368.9568786621094,
+      "logps/rejected": -883.1036987304688,
+      "loss": 0.1906,
+      "rewards/accuracies": 0.893750011920929,
+      "rewards/chosen": -2.428023099899292,
+      "rewards/margins": 5.649343967437744,
+      "rewards/rejected": -8.077366828918457,
+      "step": 190
     },
     {
       "epoch": 0.21,
+      "grad_norm": 43.44905869134368,
+      "learning_rate": 4.821331504159906e-07,
+      "logits/chosen": -1.122195839881897,
+      "logits/rejected": -0.9366437196731567,
+      "logps/chosen": -330.4066467285156,
+      "logps/rejected": -748.525634765625,
+      "loss": 0.1854,
+      "rewards/accuracies": 0.856249988079071,
+      "rewards/chosen": -2.157914161682129,
+      "rewards/margins": 4.577654838562012,
+      "rewards/rejected": -6.735569000244141,
+      "step": 200
     },
     {
       "epoch": 0.21,
+      "eval_logits/chosen": -1.1852705478668213,
+      "eval_logits/rejected": -1.108778953552246,
+      "eval_logps/chosen": -926.8517456054688,
+      "eval_logps/rejected": -1050.1199951171875,
+      "eval_loss": 1.6753935813903809,
+      "eval_rewards/accuracies": 0.62109375,
+      "eval_rewards/chosen": -6.15075159072876,
+      "eval_rewards/margins": 1.1751970052719116,
+      "eval_rewards/rejected": -7.325948715209961,
+      "eval_runtime": 97.5501,
+      "eval_samples_per_second": 20.502,
+      "eval_steps_per_second": 0.328,
+      "step": 200
+    },
+    {
+      "epoch": 0.22,
+      "grad_norm": 36.690007199237776,
+      "learning_rate": 4.785842691097342e-07,
+      "logits/chosen": -1.129880666732788,
+      "logits/rejected": -0.8643050193786621,
+      "logps/chosen": -371.864013671875,
+      "logps/rejected": -938.4737548828125,
+      "loss": 0.174,
+      "rewards/accuracies": 0.875,
+      "rewards/chosen": -2.416229009628296,
+      "rewards/margins": 6.226431369781494,
+      "rewards/rejected": -8.642660140991211,
+      "step": 210
     },
     {
       "epoch": 0.23,
+      "grad_norm": 50.37285904600086,
+      "learning_rate": 4.7472967660421603e-07,
+      "logits/chosen": -1.2257055044174194,
+      "logits/rejected": -0.964257538318634,
+      "logps/chosen": -339.07196044921875,
+      "logps/rejected": -897.4797973632812,
+      "loss": 0.1776,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -2.065497875213623,
+      "rewards/margins": 6.163290023803711,
+      "rewards/rejected": -8.228787422180176,
+      "step": 220
+    },
+    {
+      "epoch": 0.24,
+      "grad_norm": 41.9657054722659,
+      "learning_rate": 4.705745280752585e-07,
+      "logits/chosen": -1.2483174800872803,
+      "logits/rejected": -0.9970757365226746,
+      "logps/chosen": -328.69775390625,
+      "logps/rejected": -740.644775390625,
+      "loss": 0.1898,
+      "rewards/accuracies": 0.856249988079071,
+      "rewards/chosen": -1.9448823928833008,
+      "rewards/margins": 4.71053409576416,
+      "rewards/rejected": -6.655416965484619,
+      "step": 230
     },
     {
       "epoch": 0.25,
+      "grad_norm": 44.41737576885688,
+      "learning_rate": 4.6612438066572555e-07,
+      "logits/chosen": -1.1642967462539673,
+      "logits/rejected": -0.9903414845466614,
+      "logps/chosen": -318.5704650878906,
+      "logps/rejected": -773.8700561523438,
+      "loss": 0.1919,
+      "rewards/accuracies": 0.875,
+      "rewards/chosen": -1.9761199951171875,
+      "rewards/margins": 5.03547477722168,
+      "rewards/rejected": -7.011595249176025,
+      "step": 240
+    },
+    {
+      "epoch": 0.26,
+      "grad_norm": 35.762754087270444,
+      "learning_rate": 4.6138518605333664e-07,
+      "logits/chosen": -1.106542706489563,
+      "logits/rejected": -0.8413636088371277,
+      "logps/chosen": -434.7865295410156,
+      "logps/rejected": -1112.9827880859375,
+      "loss": 0.1773,
+      "rewards/accuracies": 0.887499988079071,
+      "rewards/chosen": -2.952258825302124,
+      "rewards/margins": 7.3310394287109375,
+      "rewards/rejected": -10.283297538757324,
+      "step": 250
     },
     {
       "epoch": 0.27,
+      "grad_norm": 35.92606291569474,
+      "learning_rate": 4.5636328249082514e-07,
+      "logits/chosen": -0.869143009185791,
+      "logits/rejected": -0.5741620063781738,
+      "logps/chosen": -467.0962829589844,
+      "logps/rejected": -1112.403564453125,
+      "loss": 0.162,
+      "rewards/accuracies": 0.875,
+      "rewards/chosen": -3.3054141998291016,
+      "rewards/margins": 7.030813694000244,
+      "rewards/rejected": -10.336227416992188,
+      "step": 260
+    },
+    {
+      "epoch": 0.28,
+      "grad_norm": 53.50950070995327,
+      "learning_rate": 4.510653863290871e-07,
+      "logits/chosen": -0.9721959829330444,
+      "logits/rejected": -0.6126078367233276,
+      "logps/chosen": -439.41436767578125,
+      "logps/rejected": -1052.373291015625,
+      "loss": 0.1814,
+      "rewards/accuracies": 0.893750011920929,
+      "rewards/chosen": -2.9460256099700928,
+      "rewards/margins": 6.7303972244262695,
+      "rewards/rejected": -9.676422119140625,
+      "step": 270
     },
     {
       "epoch": 0.29,
+      "grad_norm": 39.535885045091504,
+      "learning_rate": 4.4549858303465737e-07,
+      "logits/chosen": -1.0113680362701416,
+      "logits/rejected": -0.6861897706985474,
+      "logps/chosen": -475.51251220703125,
+      "logps/rejected": -1126.9036865234375,
+      "loss": 0.1758,
+      "rewards/accuracies": 0.8812500238418579,
+      "rewards/chosen": -3.39562726020813,
+      "rewards/margins": 7.080619812011719,
+      "rewards/rejected": -10.47624683380127,
+      "step": 280
+    },
+    {
+      "epoch": 0.3,
+      "grad_norm": 27.716897646836834,
+      "learning_rate": 4.396703177135261e-07,
+      "logits/chosen": -0.9551790952682495,
+      "logits/rejected": -0.6880333423614502,
+      "logps/chosen": -410.09857177734375,
+      "logps/rejected": -904.8678588867188,
+      "loss": 0.1956,
+      "rewards/accuracies": 0.875,
+      "rewards/chosen": -2.7216389179229736,
+      "rewards/margins": 5.489422798156738,
+      "rewards/rejected": -8.21106243133545,
+      "step": 290
     },
     {
       "epoch": 0.31,
+      "grad_norm": 39.15676652312449,
+      "learning_rate": 4.335883851539693e-07,
+      "logits/chosen": -1.0925099849700928,
+      "logits/rejected": -0.763215184211731,
+      "logps/chosen": -293.86553955078125,
+      "logps/rejected": -759.0989990234375,
+      "loss": 0.1799,
+      "rewards/accuracies": 0.893750011920929,
+      "rewards/chosen": -1.6122547388076782,
+      "rewards/margins": 5.217433452606201,
+      "rewards/rejected": -6.82968807220459,
+      "step": 300
+    },
+    {
+      "epoch": 0.31,
+      "eval_logits/chosen": -1.1109943389892578,
+      "eval_logits/rejected": -1.0192608833312988,
+      "eval_logps/chosen": -903.3418579101562,
+      "eval_logps/rejected": -1015.4615478515625,
+      "eval_loss": 1.5589935779571533,
+      "eval_rewards/accuracies": 0.59765625,
+      "eval_rewards/chosen": -5.915654182434082,
+      "eval_rewards/margins": 1.0637093782424927,
+      "eval_rewards/rejected": -6.979363441467285,
+      "eval_runtime": 97.4807,
+      "eval_samples_per_second": 20.517,
+      "eval_steps_per_second": 0.328,
+      "step": 300
+    },
+    {
+      "epoch": 0.32,
+      "grad_norm": 31.761072657280515,
+      "learning_rate": 4.272609194017105e-07,
+      "logits/chosen": -1.161056637763977,
+      "logits/rejected": -0.8523654937744141,
+      "logps/chosen": -387.35833740234375,
+      "logps/rejected": -977.5528564453125,
+      "loss": 0.1806,
+      "rewards/accuracies": 0.8687499761581421,
+      "rewards/chosen": -2.436070442199707,
+      "rewards/margins": 6.559242248535156,
+      "rewards/rejected": -8.99531364440918,
+      "step": 310
     },
     {
       "epoch": 0.33,
+      "grad_norm": 36.03487432587838,
+      "learning_rate": 4.2069638288135547e-07,
+      "logits/chosen": -0.8736389875411987,
+      "logits/rejected": -0.6802955865859985,
+      "logps/chosen": -391.59564208984375,
+      "logps/rejected": -1047.556884765625,
+      "loss": 0.1698,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -2.664311170578003,
+      "rewards/margins": 7.013079643249512,
+      "rewards/rejected": -9.677392959594727,
+      "step": 320
+    },
+    {
+      "epoch": 0.35,
+      "grad_norm": 31.572371493306342,
+      "learning_rate": 4.139035550786494e-07,
+      "logits/chosen": -1.1484278440475464,
+      "logits/rejected": -0.7356959581375122,
+      "logps/chosen": -385.1824645996094,
+      "logps/rejected": -1041.6712646484375,
+      "loss": 0.1467,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -2.5319342613220215,
+      "rewards/margins": 7.115927219390869,
+      "rewards/rejected": -9.647860527038574,
+      "step": 330
     },
     {
       "epoch": 0.36,
+      "grad_norm": 30.493158942186962,
+      "learning_rate": 4.0689152079869306e-07,
+      "logits/chosen": -1.111301064491272,
+      "logits/rejected": -0.7459093332290649,
+      "logps/chosen": -419.9774475097656,
+      "logps/rejected": -1145.664306640625,
+      "loss": 0.1589,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -2.8832128047943115,
+      "rewards/margins": 7.796321868896484,
+      "rewards/rejected": -10.679533958435059,
+      "step": 340
+    },
+    {
+      "epoch": 0.37,
+      "grad_norm": 35.73091881437617,
+      "learning_rate": 3.99669658015821e-07,
+      "logits/chosen": -0.9885336756706238,
+      "logits/rejected": -0.6827311515808105,
+      "logps/chosen": -417.18621826171875,
+      "logps/rejected": -1244.3375244140625,
+      "loss": 0.1607,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -2.9024839401245117,
+      "rewards/margins": 8.765107154846191,
+      "rewards/rejected": -11.667591094970703,
+      "step": 350
     },
     {
       "epoch": 0.38,
+      "grad_norm": 26.448274353945774,
+      "learning_rate": 3.92247625331392e-07,
+      "logits/chosen": -0.9021160006523132,
+      "logits/rejected": -0.46636825799942017,
+      "logps/chosen": -396.3219299316406,
+      "logps/rejected": -1072.203857421875,
+      "loss": 0.168,
+      "rewards/accuracies": 0.918749988079071,
+      "rewards/chosen": -2.622156858444214,
+      "rewards/margins": 7.3311333656311035,
+      "rewards/rejected": -9.953289985656738,
+      "step": 360
+    },
+    {
+      "epoch": 0.39,
+      "grad_norm": 26.57225676522975,
+      "learning_rate": 3.846353490562664e-07,
+      "logits/chosen": -0.7423251867294312,
+      "logits/rejected": -0.3315241038799286,
+      "logps/chosen": -403.2605285644531,
+      "logps/rejected": -1123.38623046875,
+      "loss": 0.1524,
+      "rewards/accuracies": 0.893750011920929,
+      "rewards/chosen": -2.6948583126068115,
+      "rewards/margins": 7.7378740310668945,
+      "rewards/rejected": -10.432731628417969,
+      "step": 370
     },
     {
       "epoch": 0.4,
+      "grad_norm": 31.253155533057107,
+      "learning_rate": 3.768430099352445e-07,
+      "logits/chosen": -0.6284547448158264,
+      "logits/rejected": -0.2557411789894104,
+      "logps/chosen": -438.34454345703125,
+      "logps/rejected": -1187.157958984375,
+      "loss": 0.161,
+      "rewards/accuracies": 0.893750011920929,
+      "rewards/chosen": -3.113786220550537,
+      "rewards/margins": 7.986392974853516,
+      "rewards/rejected": -11.100178718566895,
+      "step": 380
+    },
+    {
+      "epoch": 0.41,
+      "grad_norm": 30.747112252887895,
+      "learning_rate": 3.6888102953122304e-07,
+      "logits/chosen": -0.888546347618103,
+      "logits/rejected": -0.5933178663253784,
+      "logps/chosen": -410.141357421875,
+      "logps/rejected": -1144.3177490234375,
+      "loss": 0.1578,
+      "rewards/accuracies": 0.8687499761581421,
+      "rewards/chosen": -2.877619981765747,
+      "rewards/margins": 7.820855617523193,
+      "rewards/rejected": -10.69847583770752,
+      "step": 390
     },
     {
       "epoch": 0.42,
+      "grad_norm": 46.52274946683974,
+      "learning_rate": 3.607600562872785e-07,
+      "logits/chosen": -0.6739174127578735,
+      "logits/rejected": -0.4188503324985504,
+      "logps/chosen": -447.11083984375,
+      "logps/rejected": -1131.9605712890625,
+      "loss": 0.1679,
+      "rewards/accuracies": 0.887499988079071,
+      "rewards/chosen": -3.236891508102417,
+      "rewards/margins": 7.345515251159668,
+      "rewards/rejected": -10.582406997680664,
+      "step": 400
     },
     {
       "epoch": 0.42,
+      "eval_logits/chosen": -0.7096253037452698,
+      "eval_logits/rejected": -0.5753335356712341,
+      "eval_logps/chosen": -1096.810791015625,
+      "eval_logps/rejected": -1238.125244140625,
+      "eval_loss": 2.102952241897583,
+      "eval_rewards/accuracies": 0.609375,
+      "eval_rewards/chosen": -7.850342750549316,
+      "eval_rewards/margins": 1.355657696723938,
+      "eval_rewards/rejected": -9.206000328063965,
+      "eval_runtime": 97.4138,
+      "eval_samples_per_second": 20.531,
+      "eval_steps_per_second": 0.328,
+      "step": 400
+    },
+    {
+      "epoch": 0.43,
+      "grad_norm": 39.27703884155879,
+      "learning_rate": 3.5249095128531856e-07,
+      "logits/chosen": -0.6762361526489258,
+      "logits/rejected": -0.2780495882034302,
+      "logps/chosen": -492.8828125,
+      "logps/rejected": -1225.805419921875,
+      "loss": 0.1705,
+      "rewards/accuracies": 0.856249988079071,
+      "rewards/chosen": -3.5533318519592285,
+      "rewards/margins": 7.891423225402832,
+      "rewards/rejected": -11.444755554199219,
+      "step": 410
     },
     {
       "epoch": 0.44,
+      "grad_norm": 29.296123751306272,
+      "learning_rate": 3.4408477372034736e-07,
+      "logits/chosen": -0.6263514757156372,
+      "logits/rejected": -0.36908912658691406,
+      "logps/chosen": -436.956298828125,
+      "logps/rejected": -1203.846923828125,
+      "loss": 0.1542,
+      "rewards/accuracies": 0.90625,
+      "rewards/chosen": -3.0062294006347656,
+      "rewards/margins": 8.188032150268555,
+      "rewards/rejected": -11.19426155090332,
+      "step": 420
+    },
+    {
+      "epoch": 0.45,
+      "grad_norm": 38.53939686623868,
+      "learning_rate": 3.3555276610977276e-07,
+      "logits/chosen": -0.7492274045944214,
+      "logits/rejected": -0.3440536856651306,
+      "logps/chosen": -440.6358337402344,
+      "logps/rejected": -1202.001953125,
+      "loss": 0.1624,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -3.0986926555633545,
+      "rewards/margins": 8.180191040039062,
+      "rewards/rejected": -11.278883934020996,
+      "step": 430
     },
     {
       "epoch": 0.46,
+      "grad_norm": 44.81056957165417,
+      "learning_rate": 3.269063392575352e-07,
+      "logits/chosen": -0.7927631139755249,
+      "logits/rejected": -0.4414951801300049,
+      "logps/chosen": -404.3497009277344,
+      "logps/rejected": -1111.9443359375,
+      "loss": 0.1883,
+      "rewards/accuracies": 0.862500011920929,
+      "rewards/chosen": -2.7665016651153564,
+      "rewards/margins": 7.598056793212891,
+      "rewards/rejected": -10.364558219909668,
+      "step": 440
+    },
+    {
+      "epoch": 0.47,
+      "grad_norm": 46.720798060289724,
+      "learning_rate": 3.1815705699316964e-07,
+      "logits/chosen": -0.7731117010116577,
+      "logits/rejected": -0.4938638210296631,
+      "logps/chosen": -400.8499755859375,
+      "logps/rejected": -1148.82177734375,
+      "loss": 0.1656,
+      "rewards/accuracies": 0.84375,
+      "rewards/chosen": -2.6965749263763428,
+      "rewards/margins": 8.003564834594727,
+      "rewards/rejected": -10.700139045715332,
+      "step": 450
     },
     {
       "epoch": 0.48,
+      "grad_norm": 35.01981360851454,
+      "learning_rate": 3.0931662070620794e-07,
+      "logits/chosen": -0.787259578704834,
+      "logits/rejected": -0.5048869848251343,
+      "logps/chosen": -389.9725036621094,
+      "logps/rejected": -1115.496337890625,
+      "loss": 0.1644,
+      "rewards/accuracies": 0.893750011920929,
+      "rewards/chosen": -2.5107271671295166,
+      "rewards/margins": 7.763253688812256,
+      "rewards/rejected": -10.273981094360352,
+      "step": 460
+    },
+    {
+      "epoch": 0.49,
+      "grad_norm": 32.88620144282501,
+      "learning_rate": 3.003968536966078e-07,
+      "logits/chosen": -0.8425847887992859,
+      "logits/rejected": -0.507870614528656,
+      "logps/chosen": -377.3370666503906,
+      "logps/rejected": -1063.682861328125,
+      "loss": 0.1581,
+      "rewards/accuracies": 0.8812500238418579,
+      "rewards/chosen": -2.452817440032959,
+      "rewards/margins": 7.454998970031738,
+      "rewards/rejected": -9.907815933227539,
+      "step": 470
     },
     {
       "epoch": 0.5,
+      "grad_norm": 55.426706045689585,
+      "learning_rate": 2.9140968536213693e-07,
+      "logits/chosen": -0.915787398815155,
+      "logits/rejected": -0.6263198852539062,
+      "logps/chosen": -401.2126770019531,
+      "logps/rejected": -1090.992431640625,
+      "loss": 0.176,
+      "rewards/accuracies": 0.893750011920929,
+      "rewards/chosen": -2.7542808055877686,
+      "rewards/margins": 7.400801181793213,
+      "rewards/rejected": -10.155081748962402,
+      "step": 480
+    },
+    {
+      "epoch": 0.51,
+      "grad_norm": 22.799610213421833,
+      "learning_rate": 2.823671352438608e-07,
+      "logits/chosen": -0.8439705967903137,
+      "logits/rejected": -0.6034272313117981,
+      "logps/chosen": -354.22772216796875,
+      "logps/rejected": -956.9412841796875,
+      "loss": 0.1624,
+      "rewards/accuracies": 0.90625,
+      "rewards/chosen": -2.377206802368164,
+      "rewards/margins": 6.480790615081787,
+      "rewards/rejected": -8.85799789428711,
+      "step": 490
     },
     {
       "epoch": 0.52,
+      "grad_norm": 39.88720282221281,
+      "learning_rate": 2.73281296951072e-07,
+      "logits/chosen": -1.0180326700210571,
+      "logits/rejected": -0.7310226559638977,
+      "logps/chosen": -385.81866455078125,
+      "logps/rejected": -1014.95458984375,
+      "loss": 0.1693,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -2.5694196224212646,
+      "rewards/margins": 6.7894439697265625,
+      "rewards/rejected": -9.358863830566406,
+      "step": 500
+    },
+    {
+      "epoch": 0.52,
+      "eval_logits/chosen": -0.987293541431427,
+      "eval_logits/rejected": -0.8597699403762817,
+      "eval_logps/chosen": -945.861083984375,
+      "eval_logps/rejected": -1084.707763671875,
+      "eval_loss": 1.6562931537628174,
+      "eval_rewards/accuracies": 0.625,
+      "eval_rewards/chosen": -6.340845108032227,
+      "eval_rewards/margins": 1.3309805393218994,
+      "eval_rewards/rejected": -7.671825885772705,
+      "eval_runtime": 97.6234,
+      "eval_samples_per_second": 20.487,
+      "eval_steps_per_second": 0.328,
+      "step": 500
+    },
+    {
+      "epoch": 0.53,
+      "grad_norm": 36.2723513365496,
+      "learning_rate": 2.641643219871597e-07,
+      "logits/chosen": -1.0296061038970947,
+      "logits/rejected": -0.6009940505027771,
+      "logps/chosen": -367.12713623046875,
+      "logps/rejected": -957.3023681640625,
+      "loss": 0.1488,
+      "rewards/accuracies": 0.918749988079071,
+      "rewards/chosen": -2.196636199951172,
+      "rewards/margins": 6.570919990539551,
+      "rewards/rejected": -8.767557144165039,
+      "step": 510
     },
     {
       "epoch": 0.54,
+      "grad_norm": 41.00764971947524,
+      "learning_rate": 2.550284034980507e-07,
+      "logits/chosen": -0.9317744970321655,
+      "logits/rejected": -0.4665905833244324,
+      "logps/chosen": -389.6575012207031,
+      "logps/rejected": -1017.7044067382812,
+      "loss": 0.1731,
+      "rewards/accuracies": 0.875,
+      "rewards/chosen": -2.4715576171875,
+      "rewards/margins": 6.838615417480469,
+      "rewards/rejected": -9.310173034667969,
+      "step": 520
     },
     {
+      "epoch": 0.55,
+      "grad_norm": 24.760088451093488,
+      "learning_rate": 2.4588575996495794e-07,
+      "logits/chosen": -1.0027564764022827,
+      "logits/rejected": -0.5636709928512573,
+      "logps/chosen": -361.3106384277344,
+      "logps/rejected": -1032.2296142578125,
+      "loss": 0.1409,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -2.317450761795044,
+      "rewards/margins": 7.240323066711426,
+      "rewards/rejected": -9.557774543762207,
+      "step": 530
+    },
+    {
+      "epoch": 0.57,
+      "grad_norm": 29.45673267405302,
+      "learning_rate": 2.367486188632446e-07,
+      "logits/chosen": -0.6549164652824402,
+      "logits/rejected": -0.2623385787010193,
+      "logps/chosen": -400.9425048828125,
+      "logps/rejected": -1065.4869384765625,
+      "loss": 0.1447,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -2.686591863632202,
+      "rewards/margins": 7.183202266693115,
+      "rewards/rejected": -9.869793891906738,
+      "step": 540
+    },
+    {
+      "epoch": 0.58,
+      "grad_norm": 28.112202270911702,
+      "learning_rate": 2.276292003092593e-07,
+      "logits/chosen": -0.7201762199401855,
+      "logits/rejected": -0.20686273276805878,
+      "logps/chosen": -414.7852478027344,
+      "logps/rejected": -1124.4974365234375,
+      "loss": 0.154,
+      "rewards/accuracies": 0.90625,
+      "rewards/chosen": -2.9022722244262695,
+      "rewards/margins": 7.623268127441406,
+      "rewards/rejected": -10.525540351867676,
+      "step": 550
     },
     {
       "epoch": 0.59,
+      "grad_norm": 39.46687358645063,
+      "learning_rate": 2.185397007170141e-07,
+      "logits/chosen": -0.40733662247657776,
+      "logits/rejected": 0.02392803505063057,
+      "logps/chosen": -366.5719299316406,
+      "logps/rejected": -931.62841796875,
+      "loss": 0.1679,
+      "rewards/accuracies": 0.856249988079071,
+      "rewards/chosen": -2.599259853363037,
+      "rewards/margins": 6.089108467102051,
+      "rewards/rejected": -8.68836784362793,
+      "step": 560
+    },
+    {
+      "epoch": 0.6,
+      "grad_norm": 45.19451118355015,
+      "learning_rate": 2.094922764865619e-07,
+      "logits/chosen": -0.6102726459503174,
+      "logits/rejected": -0.25020402669906616,
+      "logps/chosen": -418.000732421875,
+      "logps/rejected": -1118.455322265625,
+      "loss": 0.1527,
+      "rewards/accuracies": 0.8812500238418579,
+      "rewards/chosen": -2.813910722732544,
+      "rewards/margins": 7.487911224365234,
+      "rewards/rejected": -10.3018217086792,
+      "step": 570
     },
     {
       "epoch": 0.61,
+      "grad_norm": 27.296281473923543,
+      "learning_rate": 2.0049902774588797e-07,
+      "logits/chosen": -0.5131772756576538,
+      "logits/rejected": -0.07118790596723557,
+      "logps/chosen": -358.7304992675781,
+      "logps/rejected": -981.1790161132812,
+      "loss": 0.1379,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -2.248889923095703,
+      "rewards/margins": 6.8233489990234375,
+      "rewards/rejected": -9.072239875793457,
+      "step": 580
+    },
+    {
+      "epoch": 0.62,
+      "grad_norm": 28.400791802256467,
+      "learning_rate": 1.9157198216806238e-07,
+      "logits/chosen": -0.5218924283981323,
+      "logits/rejected": -0.11522813141345978,
+      "logps/chosen": -385.2206115722656,
+      "logps/rejected": -1059.663330078125,
+      "loss": 0.1526,
+      "rewards/accuracies": 0.90625,
+      "rewards/chosen": -2.5596184730529785,
+      "rewards/margins": 7.243247032165527,
+      "rewards/rejected": -9.802865028381348,
+      "step": 590
     },
     {
       "epoch": 0.63,
+      "grad_norm": 36.01204411609108,
+      "learning_rate": 1.8272307888529274e-07,
+      "logits/chosen": -0.6575510501861572,
+      "logits/rejected": -0.13078925013542175,
+      "logps/chosen": -393.07891845703125,
+      "logps/rejected": -1083.641357421875,
+      "loss": 0.1609,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -2.5931639671325684,
+      "rewards/margins": 7.458055019378662,
+      "rewards/rejected": -10.05121898651123,
+      "step": 600
     },
     {
       "epoch": 0.63,
+      "eval_logits/chosen": -0.616359293460846,
+      "eval_logits/rejected": -0.45154044032096863,
+      "eval_logps/chosen": -959.7227172851562,
+      "eval_logps/rejected": -1097.447998046875,
+      "eval_loss": 1.6818441152572632,
+      "eval_rewards/accuracies": 0.62109375,
+      "eval_rewards/chosen": -6.479461669921875,
+      "eval_rewards/margins": 1.3197669982910156,
+      "eval_rewards/rejected": -7.799228668212891,
+      "eval_runtime": 97.4083,
+      "eval_samples_per_second": 20.532,
+      "eval_steps_per_second": 0.329,
+      "step": 600
+    },
+    {
+      "epoch": 0.64,
+      "grad_norm": 25.93993016029263,
+      "learning_rate": 1.7396415252139288e-07,
+      "logits/chosen": -0.9318272471427917,
+      "logits/rejected": -0.5557786226272583,
+      "logps/chosen": -372.4070129394531,
+      "logps/rejected": -1039.4412841796875,
+      "loss": 0.1523,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -2.3948473930358887,
+      "rewards/margins": 7.164391994476318,
+      "rewards/rejected": -9.559239387512207,
+      "step": 610
     },
     {
       "epoch": 0.65,
+      "grad_norm": 43.56570182026958,
+      "learning_rate": 1.6530691736402316e-07,
+      "logits/chosen": -0.6166467070579529,
+      "logits/rejected": -0.15916576981544495,
+      "logps/chosen": -382.71112060546875,
+      "logps/rejected": -1044.164794921875,
+      "loss": 0.1514,
+      "rewards/accuracies": 0.887499988079071,
+      "rewards/chosen": -2.561720848083496,
+      "rewards/margins": 7.1611833572387695,
+      "rewards/rejected": -9.722905158996582,
+      "step": 620
+    },
+    {
+      "epoch": 0.66,
+      "grad_norm": 28.928080380488158,
+      "learning_rate": 1.5676295169786864e-07,
+      "logits/chosen": -0.5803557634353638,
+      "logits/rejected": -0.2489413022994995,
+      "logps/chosen": -383.3107604980469,
+      "logps/rejected": -997.8870239257812,
+      "loss": 0.164,
+      "rewards/accuracies": 0.887499988079071,
+      "rewards/chosen": -2.578981876373291,
+      "rewards/margins": 6.61221170425415,
+      "rewards/rejected": -9.191193580627441,
+      "step": 630
     },
     {
       "epoch": 0.67,
+      "grad_norm": 20.471962287027644,
+      "learning_rate": 1.483436823197092e-07,
+      "logits/chosen": -0.6895365118980408,
+      "logits/rejected": -0.1990533173084259,
+      "logps/chosen": -362.12237548828125,
+      "logps/rejected": -1061.60400390625,
+      "loss": 0.1475,
+      "rewards/accuracies": 0.956250011920929,
+      "rewards/chosen": -2.2892990112304688,
+      "rewards/margins": 7.556910514831543,
+      "rewards/rejected": -9.846209526062012,
+      "step": 640
+    },
+    {
+      "epoch": 0.68,
+      "grad_norm": 26.665412290672624,
+      "learning_rate": 1.4006036925609243e-07,
+      "logits/chosen": -0.6053385734558105,
+      "logits/rejected": -0.19510070979595184,
+      "logps/chosen": -381.46514892578125,
+      "logps/rejected": -1129.877685546875,
+      "loss": 0.1432,
+      "rewards/accuracies": 0.90625,
+      "rewards/chosen": -2.4206106662750244,
+      "rewards/margins": 7.994618892669678,
+      "rewards/rejected": -10.415228843688965,
+      "step": 650
     },
     {
       "epoch": 0.69,
+      "grad_norm": 33.292289854748574,
+      "learning_rate": 1.319240907040458e-07,
+      "logits/chosen": -0.5973755717277527,
+      "logits/rejected": -0.16952477395534515,
+      "logps/chosen": -410.3275451660156,
+      "logps/rejected": -1123.840087890625,
+      "loss": 0.1388,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -2.748767852783203,
+      "rewards/margins": 7.701995849609375,
+      "rewards/rejected": -10.450764656066895,
+      "step": 660
+    },
+    {
+      "epoch": 0.7,
+      "grad_norm": 30.896865825928337,
+      "learning_rate": 1.239457282149695e-07,
+      "logits/chosen": -0.5566490292549133,
+      "logits/rejected": -0.23316040635108948,
+      "logps/chosen": -428.1212463378906,
+      "logps/rejected": -1241.2476806640625,
+      "loss": 0.1475,
+      "rewards/accuracies": 0.887499988079071,
+      "rewards/chosen": -3.0948407649993896,
+      "rewards/margins": 8.58587646484375,
+      "rewards/rejected": -11.680717468261719,
+      "step": 670
     },
     {
       "epoch": 0.71,
+      "grad_norm": 35.84504872343951,
+      "learning_rate": 1.1613595214152711e-07,
+      "logits/chosen": -0.5821831822395325,
+      "logits/rejected": -0.07458965480327606,
+      "logps/chosen": -399.4475402832031,
+      "logps/rejected": -1166.2537841796875,
+      "loss": 0.1518,
+      "rewards/accuracies": 0.9312499761581421,
+      "rewards/chosen": -2.7263970375061035,
+      "rewards/margins": 8.200857162475586,
+      "rewards/rejected": -10.927252769470215,
+      "step": 680
+    },
+    {
+      "epoch": 0.72,
+      "grad_norm": 33.84684500335279,
+      "learning_rate": 1.0850520736699362e-07,
+      "logits/chosen": -0.5655028223991394,
+      "logits/rejected": -0.15712787210941315,
+      "logps/chosen": -377.57208251953125,
+      "logps/rejected": -1100.81884765625,
+      "loss": 0.1454,
+      "rewards/accuracies": 0.90625,
+      "rewards/chosen": -2.5599417686462402,
+      "rewards/margins": 7.693644046783447,
+      "rewards/rejected": -10.253584861755371,
+      "step": 690
     },
     {
       "epoch": 0.73,
+      "grad_norm": 47.388927709457874,
+      "learning_rate": 1.0106369933615042e-07,
+      "logits/chosen": -0.2697560787200928,
+      "logits/rejected": 0.21001645922660828,
+      "logps/chosen": -389.8284606933594,
+      "logps/rejected": -1081.4793701171875,
+      "loss": 0.1559,
+      "rewards/accuracies": 0.893750011920929,
+      "rewards/chosen": -2.6655356884002686,
+      "rewards/margins": 7.382843971252441,
+      "rewards/rejected": -10.048379898071289,
+      "step": 700
+    },
+    {
+      "epoch": 0.73,
+      "eval_logits/chosen": -0.5852116346359253,
+      "eval_logits/rejected": -0.4166191518306732,
+      "eval_logps/chosen": -1046.6217041015625,
+      "eval_logps/rejected": -1197.0731201171875,
+      "eval_loss": 1.9277653694152832,
+      "eval_rewards/accuracies": 0.61328125,
+      "eval_rewards/chosen": -7.348450660705566,
+      "eval_rewards/margins": 1.447028398513794,
+      "eval_rewards/rejected": -8.795478820800781,
+      "eval_runtime": 97.4242,
+      "eval_samples_per_second": 20.529,
+      "eval_steps_per_second": 0.328,
+      "step": 700
+    },
+    {
+      "epoch": 0.74,
+      "grad_norm": 21.54892454805239,
+      "learning_rate": 9.382138040640714e-08,
+      "logits/chosen": -0.6082885265350342,
+      "logits/rejected": 0.03385084122419357,
+      "logps/chosen": -389.99542236328125,
+      "logps/rejected": -1186.9339599609375,
+      "loss": 0.1358,
+      "rewards/accuracies": 0.9312499761581421,
+      "rewards/chosen": -2.5867698192596436,
+      "rewards/margins": 8.532285690307617,
+      "rewards/rejected": -11.11905574798584,
+      "step": 710
     },
     {
       "epoch": 0.75,
+      "grad_norm": 45.770131366856944,
+      "learning_rate": 8.678793653740632e-08,
+      "logits/chosen": -0.45069313049316406,
+      "logits/rejected": -0.01569805108010769,
+      "logps/chosen": -418.41015625,
+      "logps/rejected": -1100.5294189453125,
+      "loss": 0.1518,
+      "rewards/accuracies": 0.8812500238418579,
+      "rewards/chosen": -2.9045753479003906,
+      "rewards/margins": 7.348256587982178,
+      "rewards/rejected": -10.252832412719727,
+      "step": 720
+    },
+    {
+      "epoch": 0.76,
+      "grad_norm": 26.24794912329421,
+      "learning_rate": 7.997277433690983e-08,
+      "logits/chosen": -0.6921511292457581,
+      "logits/rejected": -0.1756693720817566,
+      "logps/chosen": -402.31463623046875,
+      "logps/rejected": -1098.74951171875,
+      "loss": 0.1545,
+      "rewards/accuracies": 0.918749988079071,
+      "rewards/chosen": -2.674762725830078,
+      "rewards/margins": 7.446639060974121,
+      "rewards/rejected": -10.1214017868042,
+      "step": 730
     },
     {
       "epoch": 0.77,
+      "grad_norm": 42.21104117559166,
+      "learning_rate": 7.338500848029602e-08,
+      "logits/chosen": -0.5230437517166138,
+      "logits/rejected": -0.08038869500160217,
+      "logps/chosen": -374.2379150390625,
+      "logps/rejected": -1173.2691650390625,
+      "loss": 0.1524,
+      "rewards/accuracies": 0.90625,
+      "rewards/chosen": -2.3602004051208496,
+      "rewards/margins": 8.614425659179688,
+      "rewards/rejected": -10.974625587463379,
+      "step": 740
     },
     {
+      "epoch": 0.78,
+      "grad_norm": 45.35711154178482,
+      "learning_rate": 6.70334495204884e-08,
+      "logits/chosen": -0.4201095998287201,
+      "logits/rejected": 0.112357497215271,
+      "logps/chosen": -399.91766357421875,
+      "logps/rejected": -1149.41357421875,
+      "loss": 0.1425,
+      "rewards/accuracies": 0.918749988079071,
+      "rewards/chosen": -2.750382900238037,
+      "rewards/margins": 8.02344799041748,
+      "rewards/rejected": -10.773832321166992,
+      "step": 750
+    },
+    {
+      "epoch": 0.8,
+      "grad_norm": 26.89122562285564,
+      "learning_rate": 6.092659210462231e-08,
+      "logits/chosen": -0.6679720878601074,
+      "logits/rejected": -0.1197366863489151,
+      "logps/chosen": -417.52996826171875,
+      "logps/rejected": -1188.0615234375,
+      "loss": 0.1345,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -2.673654079437256,
+      "rewards/margins": 8.336530685424805,
+      "rewards/rejected": -11.010185241699219,
+      "step": 760
+    },
+    {
+      "epoch": 0.81,
+      "grad_norm": 34.7731803926411,
+      "learning_rate": 5.507260361320737e-08,
+      "logits/chosen": -0.6739757657051086,
+      "logits/rejected": -0.05404149740934372,
+      "logps/chosen": -402.13238525390625,
+      "logps/rejected": -1078.5826416015625,
+      "loss": 0.1516,
+      "rewards/accuracies": 0.893750011920929,
+      "rewards/chosen": -2.7083559036254883,
+      "rewards/margins": 7.324588775634766,
+      "rewards/rejected": -10.032943725585938,
+      "step": 770
     },
     {
       "epoch": 0.82,
+      "grad_norm": 25.35847618120489,
+      "learning_rate": 4.947931323697982e-08,
+      "logits/chosen": -0.5962761044502258,
+      "logits/rejected": -0.05018671602010727,
+      "logps/chosen": -392.66363525390625,
+      "logps/rejected": -1066.8021240234375,
+      "loss": 0.1551,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -2.6160168647766113,
+      "rewards/margins": 7.251564025878906,
+      "rewards/rejected": -9.86758041381836,
+      "step": 780
+    },
+    {
+      "epoch": 0.83,
+      "grad_norm": 27.140797859113846,
+      "learning_rate": 4.415420150605398e-08,
+      "logits/chosen": -0.6300855278968811,
+      "logits/rejected": -0.1746017038822174,
+      "logps/chosen": -400.11431884765625,
+      "logps/rejected": -1069.177001953125,
+      "loss": 0.1616,
+      "rewards/accuracies": 0.875,
+      "rewards/chosen": -2.7479443550109863,
+      "rewards/margins": 7.171761989593506,
+      "rewards/rejected": -9.919707298278809,
+      "step": 790
     },
     {
       "epoch": 0.84,
+      "grad_norm": 20.145953031877735,
+      "learning_rate": 3.9104390285376374e-08,
+      "logits/chosen": -0.44818106293678284,
+      "logits/rejected": -0.035945743322372437,
+      "logps/chosen": -396.0670471191406,
+      "logps/rejected": -1147.34619140625,
+      "loss": 0.1433,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -2.6665611267089844,
+      "rewards/margins": 8.064103126525879,
+      "rewards/rejected": -10.730664253234863,
+      "step": 800
     },
     {
       "epoch": 0.84,
+      "eval_logits/chosen": -0.6744760870933533,
+      "eval_logits/rejected": -0.5140590667724609,
+      "eval_logps/chosen": -1026.7318115234375,
+      "eval_logps/rejected": -1180.040283203125,
+      "eval_loss": 1.9049861431121826,
+      "eval_rewards/accuracies": 0.6171875,
+      "eval_rewards/chosen": -7.149553298950195,
+      "eval_rewards/margins": 1.4755982160568237,
+      "eval_rewards/rejected": -8.625151634216309,
+      "eval_runtime": 97.364,
+      "eval_samples_per_second": 20.541,
+      "eval_steps_per_second": 0.329,
+      "step": 800
+    },
+    {
+      "epoch": 0.85,
+      "grad_norm": 21.539181167828147,
+      "learning_rate": 3.433663324986208e-08,
+      "logits/chosen": -0.5423880815505981,
+      "logits/rejected": -0.08916589617729187,
+      "logps/chosen": -378.41876220703125,
+      "logps/rejected": -1105.344970703125,
+      "loss": 0.156,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -2.467062473297119,
+      "rewards/margins": 7.7889909744262695,
+      "rewards/rejected": -10.256052017211914,
+      "step": 810
     },
     {
       "epoch": 0.86,
+      "grad_norm": 25.60967931137277,
+      "learning_rate": 2.9857306851953897e-08,
+      "logits/chosen": -0.5412378907203674,
+      "logits/rejected": -0.05382275581359863,
+      "logps/chosen": -373.0191345214844,
+      "logps/rejected": -1065.0948486328125,
+      "loss": 0.1524,
+      "rewards/accuracies": 0.8812500238418579,
+      "rewards/chosen": -2.4381911754608154,
+      "rewards/margins": 7.450464725494385,
+      "rewards/rejected": -9.888655662536621,
+      "step": 820
+    },
+    {
+      "epoch": 0.87,
+      "grad_norm": 49.176125880409266,
+      "learning_rate": 2.567240179368185e-08,
+      "logits/chosen": -0.5491518378257751,
+      "logits/rejected": 0.00046962351188994944,
+      "logps/chosen": -376.9706115722656,
+      "logps/rejected": -1090.8355712890625,
+      "loss": 0.1429,
+      "rewards/accuracies": 0.8812500238418579,
+      "rewards/chosen": -2.460528612136841,
+      "rewards/margins": 7.668572902679443,
+      "rewards/rejected": -10.129101753234863,
+      "step": 830
     },
     {
       "epoch": 0.88,
+      "grad_norm": 37.134305931663185,
+      "learning_rate": 2.1787515014630357e-08,
+      "logits/chosen": -0.8858461380004883,
+      "logits/rejected": -0.30200493335723877,
+      "logps/chosen": -372.07769775390625,
+      "logps/rejected": -1134.0087890625,
+      "loss": 0.1405,
+      "rewards/accuracies": 0.9437500238418579,
+      "rewards/chosen": -2.2380013465881348,
+      "rewards/margins": 8.35414981842041,
+      "rewards/rejected": -10.592150688171387,
+      "step": 840
+    },
+    {
+      "epoch": 0.89,
+      "grad_norm": 44.332126805939915,
+      "learning_rate": 1.820784220652766e-08,
+      "logits/chosen": -0.44481903314590454,
+      "logits/rejected": -0.00442737340927124,
+      "logps/chosen": -393.5993957519531,
+      "logps/rejected": -1024.645751953125,
+      "loss": 0.1641,
+      "rewards/accuracies": 0.8500000238418579,
+      "rewards/chosen": -2.714064121246338,
+      "rewards/margins": 6.799878120422363,
+      "rewards/rejected": -9.513941764831543,
+      "step": 850
     },
     {
       "epoch": 0.9,
+      "grad_norm": 27.781110281391676,
+      "learning_rate": 1.4938170864468636e-08,
+      "logits/chosen": -0.713314414024353,
+      "logits/rejected": -0.28505033254623413,
+      "logps/chosen": -372.9283447265625,
+      "logps/rejected": -1127.688232421875,
+      "loss": 0.151,
+      "rewards/accuracies": 0.9437500238418579,
+      "rewards/chosen": -2.429081678390503,
+      "rewards/margins": 8.078619003295898,
+      "rewards/rejected": -10.507699966430664,
+      "step": 860
+    },
+    {
+      "epoch": 0.91,
+      "grad_norm": 40.33670958727325,
+      "learning_rate": 1.1982873884064465e-08,
+      "logits/chosen": -0.642200767993927,
+      "logits/rejected": -0.2615371346473694,
+      "logps/chosen": -382.0066223144531,
+      "logps/rejected": -1205.7874755859375,
+      "loss": 0.1425,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -2.4844939708709717,
+      "rewards/margins": 8.8142728805542,
+      "rewards/rejected": -11.298765182495117,
+      "step": 870
     },
     {
       "epoch": 0.92,
+      "grad_norm": 41.06397869005653,
+      "learning_rate": 9.345903713082304e-09,
+      "logits/chosen": -0.6282533407211304,
+      "logits/rejected": -0.19878247380256653,
+      "logps/chosen": -400.76605224609375,
+      "logps/rejected": -1277.481689453125,
+      "loss": 0.1417,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -2.5621111392974854,
+      "rewards/margins": 9.376588821411133,
+      "rewards/rejected": -11.938699722290039,
+      "step": 880
+    },
+    {
+      "epoch": 0.93,
+      "grad_norm": 59.83522561090434,
+      "learning_rate": 7.030787065396865e-09,
+      "logits/chosen": -0.5553107857704163,
+      "logits/rejected": -0.15107165277004242,
+      "logps/chosen": -375.3813781738281,
+      "logps/rejected": -1131.446533203125,
+      "loss": 0.1561,
+      "rewards/accuracies": 0.875,
+      "rewards/chosen": -2.4943184852600098,
+      "rewards/margins": 8.074764251708984,
+      "rewards/rejected": -10.56908130645752,
+      "step": 890
     },
     {
       "epoch": 0.94,
+      "grad_norm": 20.123097535281502,
+      "learning_rate": 5.04062020432286e-09,
+      "logits/chosen": -0.5520753860473633,
+      "logits/rejected": -0.19162021577358246,
+      "logps/chosen": -359.1523742675781,
+      "logps/rejected": -1050.653076171875,
+      "loss": 0.1479,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -2.378551959991455,
+      "rewards/margins": 7.357401371002197,
+      "rewards/rejected": -9.735953330993652,
+      "step": 900
+    },
+    {
+      "epoch": 0.94,
+      "eval_logits/chosen": -0.7198767066001892,
+      "eval_logits/rejected": -0.5643453598022461,
+      "eval_logps/chosen": -1010.4668579101562,
+      "eval_logps/rejected": -1164.5386962890625,
+      "eval_loss": 1.8978877067565918,
+      "eval_rewards/accuracies": 0.609375,
+      "eval_rewards/chosen": -6.986903190612793,
+      "eval_rewards/margins": 1.483232855796814,
+      "eval_rewards/rejected": -8.470136642456055,
+      "eval_runtime": 97.3726,
+      "eval_samples_per_second": 20.54,
+      "eval_steps_per_second": 0.329,
+      "step": 900
+    },
+    {
+      "epoch": 0.95,
+      "grad_norm": 27.829801809728036,
+      "learning_rate": 3.3780648016376866e-09,
+      "logits/chosen": -0.5426167249679565,
+      "logits/rejected": -0.20517143607139587,
+      "logps/chosen": -367.1676330566406,
+      "logps/rejected": -1091.75390625,
+      "loss": 0.1338,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -2.459343910217285,
+      "rewards/margins": 7.6989898681640625,
+      "rewards/rejected": -10.158334732055664,
+      "step": 910
     },
     {
       "epoch": 0.96,
+      "grad_norm": 28.33961947218242,
+      "learning_rate": 2.0453443778310766e-09,
+      "logits/chosen": -0.6683856844902039,
+      "logits/rejected": -0.24923817813396454,
+      "logps/chosen": -397.66192626953125,
+      "logps/rejected": -1160.3267822265625,
+      "loss": 0.1607,
+      "rewards/accuracies": 0.887499988079071,
+      "rewards/chosen": -2.5613341331481934,
+      "rewards/margins": 8.2199125289917,
+      "rewards/rejected": -10.781246185302734,
+      "step": 920
+    },
+    {
+      "epoch": 0.97,
+      "grad_norm": 31.245120668993923,
+      "learning_rate": 1.0442413283435758e-09,
+      "logits/chosen": -0.5892433524131775,
+      "logits/rejected": -0.16971439123153687,
+      "logps/chosen": -382.60394287109375,
+      "logps/rejected": -1098.789306640625,
+      "loss": 0.1559,
+      "rewards/accuracies": 0.893750011920929,
+      "rewards/chosen": -2.5811736583709717,
+      "rewards/margins": 7.667700290679932,
+      "rewards/rejected": -10.248873710632324,
+      "step": 930
     },
     {
       "epoch": 0.98,
+      "grad_norm": 34.32124556738926,
+      "learning_rate": 3.760945397705828e-10,
+      "logits/chosen": -0.6785596013069153,
+      "logits/rejected": -0.16107648611068726,
+      "logps/chosen": -382.91876220703125,
+      "logps/rejected": -1063.984375,
+      "loss": 0.152,
+      "rewards/accuracies": 0.875,
+      "rewards/chosen": -2.6222147941589355,
+      "rewards/margins": 7.256742000579834,
+      "rewards/rejected": -9.878957748413086,
+      "step": 940
+    },
+    {
+      "epoch": 0.99,
+      "grad_norm": 33.81539431852237,
+      "learning_rate": 4.17975992204056e-11,
+      "logits/chosen": -0.4908576011657715,
+      "logits/rejected": -0.08039870113134384,
+      "logps/chosen": -358.55047607421875,
+      "logps/rejected": -1048.7579345703125,
+      "loss": 0.1541,
+      "rewards/accuracies": 0.90625,
+      "rewards/chosen": -2.3112599849700928,
+      "rewards/margins": 7.406725883483887,
+      "rewards/rejected": -9.717985153198242,
+      "step": 950
     },
     {
       "epoch": 1.0,
+      "step": 955,
       "total_flos": 0.0,
+      "train_loss": 0.1961859940234279,
+      "train_runtime": 15468.9338,
+      "train_samples_per_second": 7.904,
+      "train_steps_per_second": 0.062
     }
   ],
   "logging_steps": 10,
+  "max_steps": 955,
   "num_input_tokens_seen": 0,
   "num_train_epochs": 1,
   "save_steps": 100,

training_args.bin CHANGED Viewed

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5a0eaf9e5538cf0fdfc4e8fe8d9145b675dbaffdc612df10f426c4a87429492c
 size 6264

 version https://git-lfs.github.com/spec/v1
+oid sha256:8eecd791d9666b4084eb7e99fb09071ac1531764dcbc88a460e7ab273a6a26f1
 size 6264