Model save

README.md

@@ -1,13 +1,11 @@
 ---
+license: apache-2.0
 library_name: peft
 tags:
-- alignment-handbook
 - trl
 - dpo
 - generated_from_trainer
 base_model: norallm/normistral-7b-warm
-datasets:
-- hugodk-sch/aftonposten_title_prefs
 model-index:
 - name: ap-normistral-7b-align-scan
   results: []
@@ -18,17 +16,17 @@ should probably proofread and complete it, then remove this comment. -->
 
 # ap-normistral-7b-align-scan
 
-This model is a fine-tuned version of [data/ap-normistral-7b-sft-qlora](https://huggingface.co/data/ap-normistral-7b-sft-qlora) on the hugodk-sch/aftonposten_title_prefs dataset.
+This model is a fine-tuned version of [norallm/normistral-7b-warm](https://huggingface.co/norallm/normistral-7b-warm) on the None dataset.
 It achieves the following results on the evaluation set:
-- Loss: 0.7822
-- Rewards/chosen: -0.1005
-- Rewards/rejected: -0.2047
-- Rewards/accuracies: 0.5249
-- Rewards/margins: 0.1042
-- Logps/rejected: -36.2224
-- Logps/chosen: -32.5688
-- Logits/rejected: 98.7467
-- Logits/chosen: 98.7645
+- Loss: 0.7290
+- Rewards/chosen: -0.1187
+- Rewards/rejected: -0.1591
+- Rewards/accuracies: 0.5453
+- Rewards/margins: 0.0404
+- Logps/rejected: -36.2848
+- Logps/chosen: -32.6805
+- Logits/rejected: 98.5351
+- Logits/chosen: 98.5572
 
 ## Model description
 
@@ -63,9 +61,9 @@ 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 |
 |:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
-| 0.7017        | 0.26  | 100  | 0.8030          | -0.0408        | -0.0544          | 0.5137             | 0.0136          | -36.0346       | -32.4942     | 98.7637         | 98.7765       |
-| 0.6236        | 0.52  | 200  | 0.7719          | -0.1720        | -0.2931          | 0.5216             | 0.1211          | -36.3329       | -32.6582     | 98.7039         | 98.7271       |
-| 0.5655        | 0.78  | 300  | 0.7744          | -0.0855        | -0.2061          | 0.5428             | 0.1206          | -36.2242       | -32.5501     | 98.7517         | 98.7722       |
+| 0.6859        | 0.26  | 100  | 0.7229          | -0.0559        | -0.0850          | 0.5282             | 0.0290          | -36.1364       | -32.5550     | 98.6727         | 98.6826       |
+| 0.5453        | 0.52  | 200  | 0.7068          | -0.1291        | -0.2159          | 0.5332             | 0.0868          | -36.3984       | -32.7014     | 98.4732         | 98.4956       |
+| 0.5511        | 0.78  | 300  | 0.7290          | -0.1187        | -0.1591          | 0.5453             | 0.0404          | -36.2848       | -32.6805     | 98.5351         | 98.5572       |
 
 
 ### Framework versions

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c739e512a4ad39bb3b4d3e837377f2819ba4a6e296af61c74c999df0d4a7f3cb
+oid sha256:5c0ea97376da5afb222e6a83a4557406c60f92c0d7ecff04ae353fe63ef5dae6
 size 671150064

all_results.json

@@ -1,21 +1,8 @@
 {
     "epoch": 1.0,
-    "eval_logits/chosen": 98.7645034790039,
-    "eval_logits/rejected": 98.74665832519531,
-    "eval_logps/chosen": -32.56877517700195,
-    "eval_logps/rejected": -36.22236251831055,
-    "eval_loss": 0.7821906208992004,
-    "eval_rewards/accuracies": 0.52491694688797,
-    "eval_rewards/chosen": -0.10047715157270432,
-    "eval_rewards/margins": 0.1041894406080246,
-    "eval_rewards/rejected": -0.20466658473014832,
-    "eval_runtime": 103.6205,
-    "eval_samples": 343,
-    "eval_samples_per_second": 3.31,
-    "eval_steps_per_second": 0.415,
-    "train_loss": 0.6050039254225694,
-    "train_runtime": 2556.17,
+    "train_loss": 0.5860598966672823,
+    "train_runtime": 2558.949,
     "train_samples": 3079,
-    "train_samples_per_second": 1.205,
-    "train_steps_per_second": 0.151
+    "train_samples_per_second": 1.203,
+    "train_steps_per_second": 0.15
 }

train_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 1.0,
-    "train_loss": 0.6050039254225694,
-    "train_runtime": 2556.17,
+    "train_loss": 0.5860598966672823,
+    "train_runtime": 2558.949,
     "train_samples": 3079,
-    "train_samples_per_second": 1.205,
-    "train_steps_per_second": 0.151
+    "train_samples_per_second": 1.203,
+    "train_steps_per_second": 0.15
 }

trainer_state.json

@@ -10,7 +10,7 @@
   "log_history": [
     {
       "epoch": 0.0,
-      "grad_norm": 52.25,
+      "grad_norm": 32.75,
       "learning_rate": 1.282051282051282e-07,
       "logits/chosen": 88.18099975585938,
       "logits/rejected": 88.25153350830078,
@@ -25,630 +25,630 @@
     },
     {
       "epoch": 0.03,
-      "grad_norm": 40.75,
+      "grad_norm": 25.5,
       "learning_rate": 1.282051282051282e-06,
-      "logits/chosen": 81.06388092041016,
-      "logits/rejected": 80.76617431640625,
-      "logps/chosen": -34.315635681152344,
-      "logps/rejected": -33.083072662353516,
-      "loss": 0.7277,
-      "rewards/accuracies": 0.4305555522441864,
-      "rewards/chosen": -0.05912955477833748,
-      "rewards/margins": 0.04060414060950279,
-      "rewards/rejected": -0.09973368048667908,
+      "logits/chosen": 81.08353424072266,
+      "logits/rejected": 80.7865219116211,
+      "logps/chosen": -34.20063781738281,
+      "logps/rejected": -33.02577590942383,
+      "loss": 0.684,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": 0.020544879138469696,
+      "rewards/margins": 0.05422975495457649,
+      "rewards/rejected": -0.033684875816106796,
       "step": 10
     },
     {
       "epoch": 0.05,
-      "grad_norm": 36.25,
+      "grad_norm": 25.0,
       "learning_rate": 2.564102564102564e-06,
-      "logits/chosen": 80.6710433959961,
-      "logits/rejected": 80.55816650390625,
-      "logps/chosen": -33.6015625,
-      "logps/rejected": -30.793548583984375,
-      "loss": 0.724,
-      "rewards/accuracies": 0.4749999940395355,
-      "rewards/chosen": 0.07067807018756866,
-      "rewards/margins": 0.08352533727884293,
-      "rewards/rejected": -0.012847280129790306,
+      "logits/chosen": 80.67228698730469,
+      "logits/rejected": 80.55956268310547,
+      "logps/chosen": -33.523399353027344,
+      "logps/rejected": -30.766698837280273,
+      "loss": 0.6867,
+      "rewards/accuracies": 0.512499988079071,
+      "rewards/chosen": 0.0832558423280716,
+      "rewards/margins": 0.0778605192899704,
+      "rewards/rejected": 0.00539533281698823,
       "step": 20
     },
     {
       "epoch": 0.08,
-      "grad_norm": 48.0,
+      "grad_norm": 29.375,
       "learning_rate": 3.846153846153847e-06,
-      "logits/chosen": 82.53105163574219,
-      "logits/rejected": 82.56202697753906,
-      "logps/chosen": -33.846534729003906,
-      "logps/rejected": -31.217113494873047,
-      "loss": 0.823,
-      "rewards/accuracies": 0.44999998807907104,
-      "rewards/chosen": 0.13303251564502716,
-      "rewards/margins": -0.03234311193227768,
-      "rewards/rejected": 0.16537563502788544,
+      "logits/chosen": 82.52102661132812,
+      "logits/rejected": 82.54798889160156,
+      "logps/chosen": -33.94279861450195,
+      "logps/rejected": -31.214736938476562,
+      "loss": 0.7575,
+      "rewards/accuracies": 0.4375,
+      "rewards/chosen": 0.035010773688554764,
+      "rewards/margins": -0.06953731179237366,
+      "rewards/rejected": 0.10454808175563812,
       "step": 30
     },
     {
       "epoch": 0.1,
-      "grad_norm": 49.0,
+      "grad_norm": 24.25,
       "learning_rate": 4.999896948438434e-06,
-      "logits/chosen": 81.03451538085938,
-      "logits/rejected": 81.03263092041016,
-      "logps/chosen": -32.87987518310547,
-      "logps/rejected": -33.08030700683594,
-      "loss": 0.7898,
-      "rewards/accuracies": 0.5,
-      "rewards/chosen": 0.2353484332561493,
-      "rewards/margins": 0.004774211905896664,
-      "rewards/rejected": 0.23057425022125244,
+      "logits/chosen": 81.06597137451172,
+      "logits/rejected": 81.06147766113281,
+      "logps/chosen": -32.76929473876953,
+      "logps/rejected": -33.212310791015625,
+      "loss": 0.6836,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.20238244533538818,
+      "rewards/margins": 0.12427721172571182,
+      "rewards/rejected": 0.07810524851083755,
       "step": 40
     },
     {
       "epoch": 0.13,
-      "grad_norm": 37.0,
+      "grad_norm": 18.75,
       "learning_rate": 4.987541037542187e-06,
-      "logits/chosen": 78.62713623046875,
-      "logits/rejected": 78.64068603515625,
-      "logps/chosen": -30.693645477294922,
-      "logps/rejected": -30.871618270874023,
-      "loss": 0.7362,
-      "rewards/accuracies": 0.550000011920929,
-      "rewards/chosen": 0.40928253531455994,
-      "rewards/margins": 0.24794897437095642,
-      "rewards/rejected": 0.1613335758447647,
+      "logits/chosen": 78.68827819824219,
+      "logits/rejected": 78.70021057128906,
+      "logps/chosen": -30.579212188720703,
+      "logps/rejected": -30.904521942138672,
+      "loss": 0.6716,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.3130193054676056,
+      "rewards/margins": 0.22863590717315674,
+      "rewards/rejected": 0.08438339829444885,
       "step": 50
     },
     {
       "epoch": 0.16,
-      "grad_norm": 35.75,
+      "grad_norm": 25.375,
       "learning_rate": 4.954691471941119e-06,
-      "logits/chosen": 83.1414566040039,
-      "logits/rejected": 83.20195007324219,
-      "logps/chosen": -30.973648071289062,
-      "logps/rejected": -29.62088394165039,
-      "loss": 0.7214,
-      "rewards/accuracies": 0.5,
-      "rewards/chosen": 0.16120928525924683,
-      "rewards/margins": 0.18549844622612,
-      "rewards/rejected": -0.02428916096687317,
+      "logits/chosen": 83.18531799316406,
+      "logits/rejected": 83.2416763305664,
+      "logps/chosen": -30.874561309814453,
+      "logps/rejected": -29.523483276367188,
+      "loss": 0.6904,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": 0.15030017495155334,
+      "rewards/margins": 0.11678256839513779,
+      "rewards/rejected": 0.03351757675409317,
       "step": 60
     },
     {
       "epoch": 0.18,
-      "grad_norm": 72.5,
+      "grad_norm": 40.75,
       "learning_rate": 4.901618883413549e-06,
-      "logits/chosen": 83.78003692626953,
-      "logits/rejected": 83.8111343383789,
-      "logps/chosen": -30.537906646728516,
-      "logps/rejected": -33.126197814941406,
-      "loss": 0.7291,
-      "rewards/accuracies": 0.612500011920929,
-      "rewards/chosen": 0.07289580255746841,
-      "rewards/margins": 0.14368507266044617,
-      "rewards/rejected": -0.07078926265239716,
+      "logits/chosen": 83.8118896484375,
+      "logits/rejected": 83.84835815429688,
+      "logps/chosen": -30.521596908569336,
+      "logps/rejected": -33.058387756347656,
+      "loss": 0.7083,
+      "rewards/accuracies": 0.512499988079071,
+      "rewards/chosen": 0.05371224880218506,
+      "rewards/margins": 0.06404958665370941,
+      "rewards/rejected": -0.010337340645492077,
       "step": 70
     },
     {
       "epoch": 0.21,
-      "grad_norm": 46.25,
+      "grad_norm": 25.625,
       "learning_rate": 4.828760511501322e-06,
-      "logits/chosen": 81.44255065917969,
-      "logits/rejected": 81.430908203125,
-      "logps/chosen": -31.439010620117188,
-      "logps/rejected": -30.995372772216797,
-      "loss": 0.7039,
-      "rewards/accuracies": 0.48750001192092896,
-      "rewards/chosen": 0.0527455098927021,
-      "rewards/margins": 0.24292418360710144,
-      "rewards/rejected": -0.19017870724201202,
+      "logits/chosen": 81.4435043334961,
+      "logits/rejected": 81.42585754394531,
+      "logps/chosen": -31.408594131469727,
+      "logps/rejected": -31.023513793945312,
+      "loss": 0.6594,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.04817419499158859,
+      "rewards/margins": 0.18110866844654083,
+      "rewards/rejected": -0.13293446600437164,
       "step": 80
     },
     {
       "epoch": 0.23,
-      "grad_norm": 61.75,
+      "grad_norm": 31.75,
       "learning_rate": 4.7367166013034295e-06,
-      "logits/chosen": 78.15830993652344,
-      "logits/rejected": 78.127197265625,
-      "logps/chosen": -32.702423095703125,
-      "logps/rejected": -31.252826690673828,
-      "loss": 0.7773,
-      "rewards/accuracies": 0.5625,
-      "rewards/chosen": -0.0513942614197731,
-      "rewards/margins": 0.18969133496284485,
-      "rewards/rejected": -0.24108560383319855,
+      "logits/chosen": 78.11305236816406,
+      "logits/rejected": 78.08197784423828,
+      "logps/chosen": -32.596641540527344,
+      "logps/rejected": -31.254547119140625,
+      "loss": 0.677,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.02076822891831398,
+      "rewards/margins": 0.17230577766895294,
+      "rewards/rejected": -0.15153756737709045,
       "step": 90
     },
     {
       "epoch": 0.26,
-      "grad_norm": 39.0,
+      "grad_norm": 24.25,
       "learning_rate": 4.626245458345211e-06,
-      "logits/chosen": 83.44832611083984,
-      "logits/rejected": 83.47621154785156,
-      "logps/chosen": -34.1397819519043,
-      "logps/rejected": -31.86690330505371,
-      "loss": 0.7017,
-      "rewards/accuracies": 0.574999988079071,
-      "rewards/chosen": 0.1321648806333542,
-      "rewards/margins": 0.225250244140625,
-      "rewards/rejected": -0.09308534860610962,
+      "logits/chosen": 83.37003326416016,
+      "logits/rejected": 83.38694763183594,
+      "logps/chosen": -34.16094207763672,
+      "logps/rejected": -31.8507137298584,
+      "loss": 0.6859,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": 0.0720243752002716,
+      "rewards/margins": 0.12210811674594879,
+      "rewards/rejected": -0.05008373409509659,
       "step": 100
     },
     {
       "epoch": 0.26,
-      "eval_logits/chosen": 98.7764892578125,
-      "eval_logits/rejected": 98.76370239257812,
-      "eval_logps/chosen": -32.49420166015625,
-      "eval_logps/rejected": -36.03456497192383,
-      "eval_loss": 0.8030149340629578,
-      "eval_rewards/accuracies": 0.5137043595314026,
-      "eval_rewards/chosen": -0.04081834852695465,
-      "eval_rewards/margins": 0.013612199574708939,
-      "eval_rewards/rejected": -0.05443055182695389,
-      "eval_runtime": 104.1164,
-      "eval_samples_per_second": 3.294,
-      "eval_steps_per_second": 0.413,
+      "eval_logits/chosen": 98.68260955810547,
+      "eval_logits/rejected": 98.67269134521484,
+      "eval_logps/chosen": -32.555023193359375,
+      "eval_logps/rejected": -36.136436462402344,
+      "eval_loss": 0.7229137420654297,
+      "eval_rewards/accuracies": 0.5282392501831055,
+      "eval_rewards/chosen": -0.05592246353626251,
+      "eval_rewards/margins": 0.029031021520495415,
+      "eval_rewards/rejected": -0.08495348691940308,
+      "eval_runtime": 104.2675,
+      "eval_samples_per_second": 3.29,
+      "eval_steps_per_second": 0.412,
       "step": 100
     },
     {
       "epoch": 0.29,
-      "grad_norm": 48.25,
+      "grad_norm": 35.25,
       "learning_rate": 4.498257201263691e-06,
-      "logits/chosen": 83.67167663574219,
-      "logits/rejected": 83.56147766113281,
-      "logps/chosen": -32.48063659667969,
-      "logps/rejected": -32.763511657714844,
-      "loss": 0.6258,
-      "rewards/accuracies": 0.6000000238418579,
-      "rewards/chosen": 0.43373650312423706,
-      "rewards/margins": 0.5198318362236023,
-      "rewards/rejected": -0.08609537780284882,
+      "logits/chosen": 83.4867172241211,
+      "logits/rejected": 83.372802734375,
+      "logps/chosen": -32.426979064941406,
+      "logps/rejected": -32.882118225097656,
+      "loss": 0.5808,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": 0.2979138493537903,
+      "rewards/margins": 0.4110265374183655,
+      "rewards/rejected": -0.11311274766921997,
       "step": 110
     },
     {
       "epoch": 0.31,
-      "grad_norm": 53.75,
+      "grad_norm": 38.5,
       "learning_rate": 4.353806263777678e-06,
-      "logits/chosen": 83.80461120605469,
-      "logits/rejected": 83.91490173339844,
-      "logps/chosen": -28.273183822631836,
-      "logps/rejected": -35.56282043457031,
-      "loss": 0.5439,
-      "rewards/accuracies": 0.7250000238418579,
-      "rewards/chosen": 0.5251210927963257,
-      "rewards/margins": 0.608059287071228,
-      "rewards/rejected": -0.08293820172548294,
+      "logits/chosen": 83.65151977539062,
+      "logits/rejected": 83.7606201171875,
+      "logps/chosen": -28.169925689697266,
+      "logps/rejected": -35.475196838378906,
+      "loss": 0.6054,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": 0.37982815504074097,
+      "rewards/margins": 0.3878515660762787,
+      "rewards/rejected": -0.008023401722311974,
       "step": 120
     },
     {
       "epoch": 0.34,
-      "grad_norm": 28.75,
+      "grad_norm": 24.0,
       "learning_rate": 4.1940827077152755e-06,
-      "logits/chosen": 81.06788635253906,
-      "logits/rejected": 81.09136962890625,
-      "logps/chosen": -30.509033203125,
-      "logps/rejected": -32.07707595825195,
-      "loss": 0.6608,
-      "rewards/accuracies": 0.5874999761581421,
-      "rewards/chosen": 0.3186565339565277,
-      "rewards/margins": 0.4350167214870453,
-      "rewards/rejected": -0.11636020243167877,
+      "logits/chosen": 80.93061065673828,
+      "logits/rejected": 80.95037841796875,
+      "logps/chosen": -30.428030014038086,
+      "logps/rejected": -32.12748336791992,
+      "loss": 0.6058,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.23966185748577118,
+      "rewards/margins": 0.33758825063705444,
+      "rewards/rejected": -0.09792639315128326,
       "step": 130
     },
     {
       "epoch": 0.36,
-      "grad_norm": 24.875,
+      "grad_norm": 17.75,
       "learning_rate": 4.0204024186666215e-06,
-      "logits/chosen": 82.33757019042969,
-      "logits/rejected": 82.34171295166016,
-      "logps/chosen": -26.971057891845703,
-      "logps/rejected": -32.84387969970703,
-      "loss": 0.5554,
-      "rewards/accuracies": 0.7250000238418579,
-      "rewards/chosen": 0.38101473450660706,
-      "rewards/margins": 0.7174161076545715,
-      "rewards/rejected": -0.33640143275260925,
+      "logits/chosen": 82.01722717285156,
+      "logits/rejected": 82.02776336669922,
+      "logps/chosen": -27.013805389404297,
+      "logps/rejected": -33.108734130859375,
+      "loss": 0.5348,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": 0.2167574167251587,
+      "rewards/margins": 0.5594369173049927,
+      "rewards/rejected": -0.3426794409751892,
       "step": 140
     },
     {
       "epoch": 0.39,
-      "grad_norm": 30.125,
+      "grad_norm": 26.125,
       "learning_rate": 3.834196265035119e-06,
-      "logits/chosen": 80.85295104980469,
-      "logits/rejected": 80.82408142089844,
-      "logps/chosen": -28.91988754272461,
-      "logps/rejected": -33.04113006591797,
-      "loss": 0.5578,
-      "rewards/accuracies": 0.699999988079071,
-      "rewards/chosen": 0.375379741191864,
-      "rewards/margins": 0.7550806403160095,
-      "rewards/rejected": -0.37970098853111267,
+      "logits/chosen": 80.49574279785156,
+      "logits/rejected": 80.46543884277344,
+      "logps/chosen": -28.96371078491211,
+      "logps/rejected": -33.28276443481445,
+      "loss": 0.5393,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": 0.21269837021827698,
+      "rewards/margins": 0.5708296895027161,
+      "rewards/rejected": -0.3581313192844391,
       "step": 150
     },
     {
       "epoch": 0.42,
-      "grad_norm": 60.5,
+      "grad_norm": 36.75,
       "learning_rate": 3.636998309800573e-06,
-      "logits/chosen": 82.69292449951172,
-      "logits/rejected": 82.70225524902344,
-      "logps/chosen": -33.54145050048828,
-      "logps/rejected": -30.415197372436523,
-      "loss": 0.6193,
-      "rewards/accuracies": 0.699999988079071,
-      "rewards/chosen": 0.4762394428253174,
-      "rewards/margins": 0.8463689684867859,
-      "rewards/rejected": -0.37012940645217896,
+      "logits/chosen": 82.2945556640625,
+      "logits/rejected": 82.3069839477539,
+      "logps/chosen": -33.60367202758789,
+      "logps/rejected": -30.597631454467773,
+      "loss": 0.5835,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": 0.266539990901947,
+      "rewards/margins": 0.5890872478485107,
+      "rewards/rejected": -0.3225473165512085,
       "step": 160
     },
     {
       "epoch": 0.44,
-      "grad_norm": 30.375,
+      "grad_norm": 26.25,
       "learning_rate": 3.4304331721118078e-06,
-      "logits/chosen": 83.42604064941406,
-      "logits/rejected": 83.37831115722656,
-      "logps/chosen": -30.8799991607666,
-      "logps/rejected": -32.62963104248047,
-      "loss": 0.5673,
-      "rewards/accuracies": 0.6625000238418579,
-      "rewards/chosen": 0.3034316599369049,
-      "rewards/margins": 0.7910871505737305,
-      "rewards/rejected": -0.4876554012298584,
+      "logits/chosen": 83.09800720214844,
+      "logits/rejected": 83.04072570800781,
+      "logps/chosen": -30.741390228271484,
+      "logps/rejected": -32.57554626464844,
+      "loss": 0.5698,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": 0.25894853472709656,
+      "rewards/margins": 0.5366927981376648,
+      "rewards/rejected": -0.27774426341056824,
       "step": 170
     },
     {
       "epoch": 0.47,
-      "grad_norm": 32.0,
+      "grad_norm": 23.625,
       "learning_rate": 3.2162026428305436e-06,
-      "logits/chosen": 80.95600891113281,
-      "logits/rejected": 80.93622589111328,
-      "logps/chosen": -30.429906845092773,
-      "logps/rejected": -31.591838836669922,
-      "loss": 0.4992,
-      "rewards/accuracies": 0.737500011920929,
-      "rewards/chosen": 0.6137940287590027,
-      "rewards/margins": 0.9614995718002319,
-      "rewards/rejected": -0.34770551323890686,
+      "logits/chosen": 80.62547302246094,
+      "logits/rejected": 80.6036605834961,
+      "logps/chosen": -30.658870697021484,
+      "logps/rejected": -31.662927627563477,
+      "loss": 0.55,
+      "rewards/accuracies": 0.7250000238418579,
+      "rewards/chosen": 0.2691388726234436,
+      "rewards/margins": 0.5219991207122803,
+      "rewards/rejected": -0.25286024808883667,
       "step": 180
     },
     {
       "epoch": 0.49,
-      "grad_norm": 17.25,
+      "grad_norm": 12.4375,
       "learning_rate": 2.996071664294641e-06,
-      "logits/chosen": 82.7037124633789,
-      "logits/rejected": 82.69169616699219,
-      "logps/chosen": -30.465667724609375,
-      "logps/rejected": -30.747249603271484,
-      "loss": 0.7493,
-      "rewards/accuracies": 0.574999988079071,
-      "rewards/chosen": 0.24060054123401642,
-      "rewards/margins": 0.463981568813324,
-      "rewards/rejected": -0.22338099777698517,
+      "logits/chosen": 82.23011779785156,
+      "logits/rejected": 82.23200225830078,
+      "logps/chosen": -30.246551513671875,
+      "logps/rejected": -30.743694305419922,
+      "loss": 0.6343,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.25993233919143677,
+      "rewards/margins": 0.3977679908275604,
+      "rewards/rejected": -0.13783565163612366,
       "step": 190
     },
     {
       "epoch": 0.52,
-      "grad_norm": 21.5,
+      "grad_norm": 14.625,
       "learning_rate": 2.7718537898066833e-06,
-      "logits/chosen": 78.2643814086914,
-      "logits/rejected": 78.21286010742188,
-      "logps/chosen": -33.79939270019531,
-      "logps/rejected": -32.67485046386719,
-      "loss": 0.6236,
-      "rewards/accuracies": 0.675000011920929,
-      "rewards/chosen": 0.7626792192459106,
-      "rewards/margins": 0.9078337550163269,
-      "rewards/rejected": -0.14515459537506104,
+      "logits/chosen": 77.68287658691406,
+      "logits/rejected": 77.6307144165039,
+      "logps/chosen": -33.85746383666992,
+      "logps/rejected": -32.90498352050781,
+      "loss": 0.5453,
+      "rewards/accuracies": 0.7250000238418579,
+      "rewards/chosen": 0.44763606786727905,
+      "rewards/margins": 0.6534249186515808,
+      "rewards/rejected": -0.20578882098197937,
       "step": 200
     },
     {
       "epoch": 0.52,
-      "eval_logits/chosen": 98.72710418701172,
-      "eval_logits/rejected": 98.70391082763672,
-      "eval_logps/chosen": -32.6581916809082,
-      "eval_logps/rejected": -36.33292770385742,
-      "eval_loss": 0.7719493508338928,
-      "eval_rewards/accuracies": 0.5215947031974792,
-      "eval_rewards/chosen": -0.17201441526412964,
-      "eval_rewards/margins": 0.12110573798418045,
-      "eval_rewards/rejected": -0.2931201457977295,
-      "eval_runtime": 103.8817,
-      "eval_samples_per_second": 3.302,
+      "eval_logits/chosen": 98.49555206298828,
+      "eval_logits/rejected": 98.47319030761719,
+      "eval_logps/chosen": -32.7014274597168,
+      "eval_logps/rejected": -36.39837646484375,
+      "eval_loss": 0.7067598700523376,
+      "eval_rewards/accuracies": 0.5332226157188416,
+      "eval_rewards/chosen": -0.12912562489509583,
+      "eval_rewards/margins": 0.0867983028292656,
+      "eval_rewards/rejected": -0.21592393517494202,
+      "eval_runtime": 103.9124,
+      "eval_samples_per_second": 3.301,
       "eval_steps_per_second": 0.414,
       "step": 200
     },
     {
       "epoch": 0.55,
-      "grad_norm": 78.5,
+      "grad_norm": 46.5,
       "learning_rate": 2.5453962426402006e-06,
-      "logits/chosen": 80.8266830444336,
-      "logits/rejected": 80.73197174072266,
-      "logps/chosen": -33.31422805786133,
-      "logps/rejected": -35.28189468383789,
-      "loss": 0.6059,
-      "rewards/accuracies": 0.6625000238418579,
-      "rewards/chosen": 0.5051938891410828,
-      "rewards/margins": 0.7418341636657715,
-      "rewards/rejected": -0.23664025962352753,
+      "logits/chosen": 80.28144836425781,
+      "logits/rejected": 80.18412780761719,
+      "logps/chosen": -33.1484489440918,
+      "logps/rejected": -35.34803009033203,
+      "loss": 0.5401,
+      "rewards/accuracies": 0.762499988079071,
+      "rewards/chosen": 0.39863476157188416,
+      "rewards/margins": 0.5796052813529968,
+      "rewards/rejected": -0.18097054958343506,
       "step": 210
     },
     {
       "epoch": 0.57,
-      "grad_norm": 21.0,
+      "grad_norm": 18.875,
       "learning_rate": 2.3185646976551794e-06,
-      "logits/chosen": 83.00508117675781,
-      "logits/rejected": 83.09242248535156,
-      "logps/chosen": -31.03920555114746,
-      "logps/rejected": -30.981882095336914,
-      "loss": 0.5212,
-      "rewards/accuracies": 0.800000011920929,
-      "rewards/chosen": 0.7277995347976685,
-      "rewards/margins": 0.930141270160675,
-      "rewards/rejected": -0.20234176516532898,
+      "logits/chosen": 82.44231414794922,
+      "logits/rejected": 82.5237808227539,
+      "logps/chosen": -31.035247802734375,
+      "logps/rejected": -31.218652725219727,
+      "loss": 0.5206,
+      "rewards/accuracies": 0.762499988079071,
+      "rewards/chosen": 0.45685210824012756,
+      "rewards/margins": 0.7017010450363159,
+      "rewards/rejected": -0.24484892189502716,
       "step": 220
     },
     {
       "epoch": 0.6,
-      "grad_norm": 49.25,
+      "grad_norm": 27.5,
       "learning_rate": 2.0932279108998323e-06,
-      "logits/chosen": 80.0581283569336,
-      "logits/rejected": 80.11595153808594,
-      "logps/chosen": -32.08351516723633,
-      "logps/rejected": -34.33815002441406,
-      "loss": 0.594,
+      "logits/chosen": 79.57551574707031,
+      "logits/rejected": 79.63117218017578,
+      "logps/chosen": -32.342308044433594,
+      "logps/rejected": -34.377655029296875,
+      "loss": 0.6337,
       "rewards/accuracies": 0.637499988079071,
-      "rewards/chosen": 0.5777846574783325,
-      "rewards/margins": 0.8385850787162781,
-      "rewards/rejected": -0.26080039143562317,
+      "rewards/chosen": 0.23171910643577576,
+      "rewards/margins": 0.4144717752933502,
+      "rewards/rejected": -0.18275271356105804,
       "step": 230
     },
     {
       "epoch": 0.62,
-      "grad_norm": 28.0,
+      "grad_norm": 29.375,
       "learning_rate": 1.8712423238279358e-06,
-      "logits/chosen": 82.56427001953125,
-      "logits/rejected": 82.85057830810547,
-      "logps/chosen": -30.730464935302734,
-      "logps/rejected": -31.813983917236328,
-      "loss": 0.4701,
-      "rewards/accuracies": 0.75,
-      "rewards/chosen": 0.7660826444625854,
-      "rewards/margins": 0.9916342496871948,
-      "rewards/rejected": -0.22555160522460938,
+      "logits/chosen": 81.96827697753906,
+      "logits/rejected": 82.25470733642578,
+      "logps/chosen": -30.665319442749023,
+      "logps/rejected": -31.84256362915039,
+      "loss": 0.5035,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": 0.5113734006881714,
+      "rewards/margins": 0.6666342616081238,
+      "rewards/rejected": -0.1552608758211136,
       "step": 240
     },
     {
       "epoch": 0.65,
-      "grad_norm": 50.5,
+      "grad_norm": 27.125,
       "learning_rate": 1.6544367689701824e-06,
-      "logits/chosen": 81.2629623413086,
-      "logits/rejected": 81.32804870605469,
-      "logps/chosen": -26.856103897094727,
-      "logps/rejected": -30.107614517211914,
-      "loss": 0.6072,
-      "rewards/accuracies": 0.6499999761581421,
-      "rewards/chosen": 0.5918346047401428,
-      "rewards/margins": 0.7573351263999939,
-      "rewards/rejected": -0.1655004471540451,
+      "logits/chosen": 80.69163513183594,
+      "logits/rejected": 80.74885559082031,
+      "logps/chosen": -26.948383331298828,
+      "logps/rejected": -30.213348388671875,
+      "loss": 0.6037,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.32375770807266235,
+      "rewards/margins": 0.48006218671798706,
+      "rewards/rejected": -0.1563045233488083,
       "step": 250
     },
     {
       "epoch": 0.68,
-      "grad_norm": 36.25,
+      "grad_norm": 27.75,
       "learning_rate": 1.4445974030621963e-06,
-      "logits/chosen": 78.54646301269531,
-      "logits/rejected": 78.68824768066406,
-      "logps/chosen": -30.280277252197266,
-      "logps/rejected": -36.395103454589844,
-      "loss": 0.4323,
-      "rewards/accuracies": 0.800000011920929,
-      "rewards/chosen": 1.0245726108551025,
-      "rewards/margins": 1.259353756904602,
-      "rewards/rejected": -0.23478105664253235,
+      "logits/chosen": 77.91614532470703,
+      "logits/rejected": 78.04708099365234,
+      "logps/chosen": -30.49196434020996,
+      "logps/rejected": -36.51909255981445,
+      "loss": 0.4936,
+      "rewards/accuracies": 0.7250000238418579,
+      "rewards/chosen": 0.5345155000686646,
+      "rewards/margins": 0.7432474493980408,
+      "rewards/rejected": -0.2087319791316986,
       "step": 260
     },
     {
       "epoch": 0.7,
-      "grad_norm": 22.625,
+      "grad_norm": 16.75,
       "learning_rate": 1.243452991757889e-06,
-      "logits/chosen": 77.9116439819336,
-      "logits/rejected": 77.93907165527344,
-      "logps/chosen": -30.94207763671875,
-      "logps/rejected": -31.825298309326172,
-      "loss": 0.5298,
-      "rewards/accuracies": 0.7124999761581421,
-      "rewards/chosen": 0.6694773435592651,
-      "rewards/margins": 0.9253811836242676,
-      "rewards/rejected": -0.2559038996696472,
+      "logits/chosen": 77.25732421875,
+      "logits/rejected": 77.28456115722656,
+      "logps/chosen": -30.877111434936523,
+      "logps/rejected": -31.975589752197266,
+      "loss": 0.5167,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": 0.4509071707725525,
+      "rewards/margins": 0.685990035533905,
+      "rewards/rejected": -0.23508290946483612,
       "step": 270
     },
     {
       "epoch": 0.73,
-      "grad_norm": 48.0,
+      "grad_norm": 25.125,
       "learning_rate": 1.0526606671603523e-06,
-      "logits/chosen": 80.6185531616211,
-      "logits/rejected": 80.40409851074219,
-      "logps/chosen": -31.1258487701416,
-      "logps/rejected": -29.83230972290039,
-      "loss": 0.6589,
-      "rewards/accuracies": 0.6625000238418579,
-      "rewards/chosen": 0.5472350716590881,
-      "rewards/margins": 0.6794873476028442,
-      "rewards/rejected": -0.1322522908449173,
+      "logits/chosen": 80.03535461425781,
+      "logits/rejected": 79.8197250366211,
+      "logps/chosen": -31.083843231201172,
+      "logps/rejected": -29.844961166381836,
+      "loss": 0.6198,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": 0.3630252182483673,
+      "rewards/margins": 0.45200806856155396,
+      "rewards/rejected": -0.08898283541202545,
       "step": 280
     },
     {
       "epoch": 0.75,
-      "grad_norm": 17.25,
+      "grad_norm": 16.625,
       "learning_rate": 8.737922755071455e-07,
-      "logits/chosen": 80.76725006103516,
-      "logits/rejected": 80.67757415771484,
-      "logps/chosen": -33.04716491699219,
-      "logps/rejected": -32.507728576660156,
-      "loss": 0.4727,
-      "rewards/accuracies": 0.7250000238418579,
-      "rewards/chosen": 0.8456665277481079,
-      "rewards/margins": 1.238355040550232,
-      "rewards/rejected": -0.39268869161605835,
+      "logits/chosen": 80.13519287109375,
+      "logits/rejected": 80.05864715576172,
+      "logps/chosen": -33.047523498535156,
+      "logps/rejected": -32.68352127075195,
+      "loss": 0.4787,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": 0.5283600091934204,
+      "rewards/margins": 0.8616873025894165,
+      "rewards/rejected": -0.3333272337913513,
       "step": 290
     },
     {
       "epoch": 0.78,
-      "grad_norm": 44.75,
+      "grad_norm": 26.875,
       "learning_rate": 7.08321427484816e-07,
-      "logits/chosen": 76.41008758544922,
-      "logits/rejected": 76.48558044433594,
-      "logps/chosen": -32.2567253112793,
-      "logps/rejected": -29.211252212524414,
-      "loss": 0.5655,
-      "rewards/accuracies": 0.75,
-      "rewards/chosen": 0.9191843271255493,
-      "rewards/margins": 1.0065667629241943,
-      "rewards/rejected": -0.08738242089748383,
+      "logits/chosen": 75.82341003417969,
+      "logits/rejected": 75.91938781738281,
+      "logps/chosen": -32.24137878417969,
+      "logps/rejected": -29.33247947692871,
+      "loss": 0.5511,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.5821641087532043,
+      "rewards/margins": 0.6973918676376343,
+      "rewards/rejected": -0.11522769927978516,
       "step": 300
     },
     {
       "epoch": 0.78,
-      "eval_logits/chosen": 98.77224731445312,
-      "eval_logits/rejected": 98.75171661376953,
-      "eval_logps/chosen": -32.550052642822266,
-      "eval_logps/rejected": -36.224205017089844,
-      "eval_loss": 0.7744476795196533,
-      "eval_rewards/accuracies": 0.5427741408348083,
-      "eval_rewards/chosen": -0.08550228923559189,
-      "eval_rewards/margins": 0.1206393763422966,
-      "eval_rewards/rejected": -0.2061416506767273,
-      "eval_runtime": 104.0332,
-      "eval_samples_per_second": 3.297,
+      "eval_logits/chosen": 98.5571517944336,
+      "eval_logits/rejected": 98.53510284423828,
+      "eval_logps/chosen": -32.68053436279297,
+      "eval_logps/rejected": -36.284751892089844,
+      "eval_loss": 0.7289925813674927,
+      "eval_rewards/accuracies": 0.545265793800354,
+      "eval_rewards/chosen": -0.11868056654930115,
+      "eval_rewards/margins": 0.04043150320649147,
+      "eval_rewards/rejected": -0.1591120809316635,
+      "eval_runtime": 104.072,
+      "eval_samples_per_second": 3.296,
       "eval_steps_per_second": 0.413,
       "step": 300
     },
     {
       "epoch": 0.81,
-      "grad_norm": 31.875,
+      "grad_norm": 21.625,
       "learning_rate": 5.576113578589035e-07,
-      "logits/chosen": 83.43260192871094,
-      "logits/rejected": 83.46700286865234,
-      "logps/chosen": -30.02802085876465,
-      "logps/rejected": -32.44657897949219,
-      "loss": 0.5905,
-      "rewards/accuracies": 0.699999988079071,
-      "rewards/chosen": 0.6684112548828125,
-      "rewards/margins": 0.8658668398857117,
-      "rewards/rejected": -0.19745555520057678,
+      "logits/chosen": 82.95035552978516,
+      "logits/rejected": 82.97549438476562,
+      "logps/chosen": -30.2395076751709,
+      "logps/rejected": -32.58390808105469,
+      "loss": 0.5895,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": 0.31201133131980896,
+      "rewards/margins": 0.5040835738182068,
+      "rewards/rejected": -0.19207225739955902,
       "step": 310
     },
     {
       "epoch": 0.83,
-      "grad_norm": 24.875,
+      "grad_norm": 18.0,
       "learning_rate": 4.229036944380913e-07,
-      "logits/chosen": 80.97013092041016,
-      "logits/rejected": 80.96881103515625,
-      "logps/chosen": -30.49709701538086,
-      "logps/rejected": -29.160675048828125,
-      "loss": 0.4662,
-      "rewards/accuracies": 0.7250000238418579,
-      "rewards/chosen": 0.8811699748039246,
-      "rewards/margins": 1.0741162300109863,
-      "rewards/rejected": -0.19294631481170654,
+      "logits/chosen": 80.41062927246094,
+      "logits/rejected": 80.41141510009766,
+      "logps/chosen": -30.43392562866211,
+      "logps/rejected": -29.273609161376953,
+      "loss": 0.4816,
+      "rewards/accuracies": 0.762499988079071,
+      "rewards/chosen": 0.5823162794113159,
+      "rewards/margins": 0.7593742609024048,
+      "rewards/rejected": -0.1770579218864441,
       "step": 320
     },
     {
       "epoch": 0.86,
-      "grad_norm": 19.625,
+      "grad_norm": 16.625,
       "learning_rate": 3.053082288996112e-07,
-      "logits/chosen": 78.14985656738281,
-      "logits/rejected": 78.19586181640625,
-      "logps/chosen": -29.189855575561523,
-      "logps/rejected": -32.966094970703125,
-      "loss": 0.4527,
-      "rewards/accuracies": 0.7749999761581421,
-      "rewards/chosen": 0.9302013516426086,
-      "rewards/margins": 1.195185661315918,
-      "rewards/rejected": -0.26498422026634216,
+      "logits/chosen": 77.58735656738281,
+      "logits/rejected": 77.6371841430664,
+      "logps/chosen": -29.118694305419922,
+      "logps/rejected": -33.023460388183594,
+      "loss": 0.4851,
+      "rewards/accuracies": 0.762499988079071,
+      "rewards/chosen": 0.6169556975364685,
+      "rewards/margins": 0.8112555742263794,
+      "rewards/rejected": -0.19429990649223328,
       "step": 330
     },
     {
       "epoch": 0.88,
-      "grad_norm": 61.5,
+      "grad_norm": 33.5,
       "learning_rate": 2.0579377374915805e-07,
-      "logits/chosen": 82.46419525146484,
-      "logits/rejected": 82.48377990722656,
-      "logps/chosen": -32.1870002746582,
-      "logps/rejected": -33.69792175292969,
-      "loss": 0.5537,
-      "rewards/accuracies": 0.737500011920929,
-      "rewards/chosen": 0.8201173543930054,
-      "rewards/margins": 1.0834996700286865,
-      "rewards/rejected": -0.26338234543800354,
+      "logits/chosen": 81.87162780761719,
+      "logits/rejected": 81.90576171875,
+      "logps/chosen": -32.27037048339844,
+      "logps/rejected": -33.92435836791992,
+      "loss": 0.5157,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": 0.4708884358406067,
+      "rewards/margins": 0.7487217783927917,
+      "rewards/rejected": -0.27783337235450745,
       "step": 340
     },
     {
       "epoch": 0.91,
-      "grad_norm": 16.75,
+      "grad_norm": 13.0,
       "learning_rate": 1.2518018074041684e-07,
-      "logits/chosen": 81.44327545166016,
-      "logits/rejected": 81.4502944946289,
-      "logps/chosen": -32.606056213378906,
-      "logps/rejected": -33.298622131347656,
-      "loss": 0.5676,
-      "rewards/accuracies": 0.6875,
-      "rewards/chosen": 0.8301448822021484,
-      "rewards/margins": 0.9956042170524597,
-      "rewards/rejected": -0.16545933485031128,
+      "logits/chosen": 80.89490509033203,
+      "logits/rejected": 80.90391540527344,
+      "logps/chosen": -32.372520446777344,
+      "logps/rejected": -33.53419876098633,
+      "loss": 0.4983,
+      "rewards/accuracies": 0.762499988079071,
+      "rewards/chosen": 0.6356090307235718,
+      "rewards/margins": 0.8568087816238403,
+      "rewards/rejected": -0.22119958698749542,
       "step": 350
     },
     {
       "epoch": 0.94,
-      "grad_norm": 33.0,
+      "grad_norm": 20.125,
       "learning_rate": 6.41315865106129e-08,
-      "logits/chosen": 82.95188903808594,
-      "logits/rejected": 82.9841079711914,
-      "logps/chosen": -28.371601104736328,
-      "logps/rejected": -31.830394744873047,
-      "loss": 0.4872,
-      "rewards/accuracies": 0.7124999761581421,
-      "rewards/chosen": 0.9444905519485474,
-      "rewards/margins": 1.0889289379119873,
-      "rewards/rejected": -0.14443838596343994,
+      "logits/chosen": 82.41497802734375,
+      "logits/rejected": 82.44314575195312,
+      "logps/chosen": -28.537517547607422,
+      "logps/rejected": -31.826416015625,
+      "loss": 0.5313,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": 0.5073467493057251,
+      "rewards/margins": 0.5956294536590576,
+      "rewards/rejected": -0.08828280866146088,
       "step": 360
     },
     {
       "epoch": 0.96,
-      "grad_norm": 34.25,
+      "grad_norm": 21.75,
       "learning_rate": 2.3150941078050325e-08,
-      "logits/chosen": 82.37725067138672,
-      "logits/rejected": 82.40049743652344,
-      "logps/chosen": -31.749374389648438,
-      "logps/rejected": -35.2952766418457,
-      "loss": 0.508,
-      "rewards/accuracies": 0.7124999761581421,
-      "rewards/chosen": 0.9014309644699097,
-      "rewards/margins": 1.0501554012298584,
-      "rewards/rejected": -0.14872442185878754,
+      "logits/chosen": 81.85018920898438,
+      "logits/rejected": 81.87150573730469,
+      "logps/chosen": -31.847997665405273,
+      "logps/rejected": -35.508277893066406,
+      "loss": 0.5486,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": 0.5140814781188965,
+      "rewards/margins": 0.7135335206985474,
+      "rewards/rejected": -0.19945205748081207,
       "step": 370
     },
     {
       "epoch": 0.99,
-      "grad_norm": 38.0,
+      "grad_norm": 25.5,
       "learning_rate": 2.575864278703266e-09,
-      "logits/chosen": 76.2959976196289,
-      "logits/rejected": 76.17599487304688,
-      "logps/chosen": -29.714282989501953,
-      "logps/rejected": -28.307018280029297,
-      "loss": 0.57,
-      "rewards/accuracies": 0.6625000238418579,
-      "rewards/chosen": 0.6894746422767639,
-      "rewards/margins": 0.806629478931427,
-      "rewards/rejected": -0.11715485900640488,
+      "logits/chosen": 75.74685668945312,
+      "logits/rejected": 75.61808013916016,
+      "logps/chosen": -29.746694564819336,
+      "logps/rejected": -28.432201385498047,
+      "loss": 0.5715,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": 0.4147160053253174,
+      "rewards/margins": 0.5505291223526001,
+      "rewards/rejected": -0.1358131319284439,
       "step": 380
     },
     {
       "epoch": 1.0,
       "step": 385,
       "total_flos": 0.0,
-      "train_loss": 0.6050039254225694,
-      "train_runtime": 2556.17,
-      "train_samples_per_second": 1.205,
-      "train_steps_per_second": 0.151
+      "train_loss": 0.5860598966672823,
+      "train_runtime": 2558.949,
+      "train_samples_per_second": 1.203,
+      "train_steps_per_second": 0.15
     }
   ],
   "logging_steps": 10,