Model save

README.md

@@ -1,13 +1,10 @@
 ---
 library_name: peft
 tags:
-- alignment-handbook
 - trl
 - dpo
 - generated_from_trainer
 base_model: NbAiLab/nb-gpt-j-6B-v2
-datasets:
-- hugodk-sch/aftonposten_title_prefs
 model-index:
 - name: aftonposten-6b-align-scan
   results: []
@@ -18,17 +15,17 @@ should probably proofread and complete it, then remove this comment. -->
 
 # aftonposten-6b-align-scan
 
-This model is a fine-tuned version of [data/ap-gpt-j-6b-sft-qlora-04-08](https://huggingface.co/data/ap-gpt-j-6b-sft-qlora-04-08) on the hugodk-sch/aftonposten_title_prefs dataset.
+This model is a fine-tuned version of [NbAiLab/nb-gpt-j-6B-v2](https://huggingface.co/NbAiLab/nb-gpt-j-6B-v2) on an unknown dataset.
 It achieves the following results on the evaluation set:
-- Loss: 0.9991
-- Rewards/chosen: 0.0161
-- Rewards/rejected: 0.0150
-- Rewards/accuracies: 0.5365
-- Rewards/margins: 0.0010
-- Logps/rejected: -37.4999
-- Logps/chosen: -34.0167
-- Logits/rejected: -2.2391
-- Logits/chosen: -2.2439
+- Loss: 0.6929
+- Rewards/chosen: -0.0003
+- Rewards/rejected: -0.0007
+- Rewards/accuracies: 0.5419
+- Rewards/margins: 0.0004
+- Logps/rejected: -37.5868
+- Logps/chosen: -34.0648
+- Logits/rejected: -2.2163
+- Logits/chosen: -2.2211
 
 ## Model description
 
@@ -61,11 +58,11 @@ The following hyperparameters were used during training:
 
 ### Training results
 
-| Training Loss | Epoch | Step | Logits/chosen | Logits/rejected | Logps/chosen | Logps/rejected | Validation Loss | Rewards/accuracies | Rewards/chosen | Rewards/margins | Rewards/rejected |
-|:-------------:|:-----:|:----:|:-------------:|:---------------:|:------------:|:--------------:|:---------------:|:------------------:|:--------------:|:---------------:|:----------------:|
-| 0.9823        | 0.26  | 100  | -2.2440       | -2.2391         | -34.0216     | -37.5058       | 0.9982          | 0.5104             | 0.0116         | 0.0019          | 0.0097           |
-| 0.9696        | 0.52  | 200  | -2.2439       | -2.2390         | -34.0217     | -37.5024       | 1.0014          | 0.4693             | 0.0116         | -0.0012         | 0.0128           |
-| 0.9581        | 0.78  | 300  | -2.2437       | -2.2388         | -34.0306     | -37.4994       | 1.0120          | 0.4726             | 0.0035         | -0.0120         | 0.0155           |
+| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
+|:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
+| 0.6924        | 0.26  | 100  | 0.6931          | -0.0001        | -0.0002          | 0.5187             | 0.0001          | -37.5393       | -34.0468     | -2.2252         | -2.2301       |
+| 0.6914        | 0.52  | 200  | 0.6930          | -0.0001        | -0.0004          | 0.5245             | 0.0003          | -37.5530       | -34.0426     | -2.2201         | -2.2250       |
+| 0.6895        | 0.78  | 300  | 0.6929          | -0.0003        | -0.0007          | 0.5419             | 0.0004          | -37.5868       | -34.0648     | -2.2163         | -2.2211       |
 
 
 ### Framework versions

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b62113b40e07a216428627da02a746dc9cd8cf45c85cdf8f158153476744aada
+oid sha256:c85c938a116836c2be173e9bdc671f7190ee519e02c0173d3562a6d2d1044243
 size 176183216

all_results.json

@@ -1,21 +1,8 @@
 {
     "epoch": 1.0,
-    "eval_logits/chosen": -2.2439348697662354,
-    "eval_logits/rejected": -2.2390594482421875,
-    "eval_logps/chosen": -34.016693115234375,
-    "eval_logps/rejected": -37.499900817871094,
-    "eval_loss": 0.9991074204444885,
-    "eval_rewards/accuracies": 0.5365448594093323,
-    "eval_rewards/chosen": 0.01607290655374527,
-    "eval_rewards/margins": 0.0010251295752823353,
-    "eval_rewards/rejected": 0.015047776512801647,
-    "eval_runtime": 146.1502,
-    "eval_samples": 343,
-    "eval_samples_per_second": 2.347,
-    "eval_steps_per_second": 0.294,
-    "train_loss": 0.21873634759481853,
-    "train_runtime": 627.4596,
+    "train_loss": 0.6918259131443965,
+    "train_runtime": 3249.6401,
     "train_samples": 3079,
-    "train_samples_per_second": 4.907,
-    "train_steps_per_second": 0.614
+    "train_samples_per_second": 0.947,
+    "train_steps_per_second": 0.118
 }

train_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 1.0,
-    "train_loss": 0.21873634759481853,
-    "train_runtime": 627.4596,
+    "train_loss": 0.6918259131443965,
+    "train_runtime": 3249.6401,
     "train_samples": 3079,
-    "train_samples_per_second": 4.907,
-    "train_steps_per_second": 0.614
+    "train_samples_per_second": 0.947,
+    "train_steps_per_second": 0.118
 }

trainer_state.json

@@ -10,12 +10,12 @@
   "log_history": [
     {
       "epoch": 0.0,
-      "learning_rate": 1.282051282051282e-08,
+      "learning_rate": 1.282051282051282e-07,
       "logits/chosen": -1.7278180122375488,
       "logits/rejected": -1.7377450466156006,
       "logps/chosen": -29.553977966308594,
       "logps/rejected": -42.813133239746094,
-      "loss": 1.0,
+      "loss": 0.6931,
       "rewards/accuracies": 0.0,
       "rewards/chosen": 0.0,
       "rewards/margins": 0.0,
@@ -24,592 +24,592 @@
     },
     {
       "epoch": 0.03,
-      "learning_rate": 1.2820512820512818e-07,
-      "logits/chosen": -1.8661272525787354,
-      "logits/rejected": -1.8704373836517334,
-      "logps/chosen": -36.993247985839844,
-      "logps/rejected": -33.67000961303711,
-      "loss": 0.952,
-      "rewards/accuracies": 0.5138888955116272,
-      "rewards/chosen": 0.012012440711259842,
-      "rewards/margins": 0.047955650836229324,
-      "rewards/rejected": -0.035943206399679184,
+      "learning_rate": 1.282051282051282e-06,
+      "logits/chosen": -1.8666962385177612,
+      "logits/rejected": -1.8709977865219116,
+      "logps/chosen": -36.98939514160156,
+      "logps/rejected": -33.66963195800781,
+      "loss": 0.6929,
+      "rewards/accuracies": 0.5694444179534912,
+      "rewards/chosen": 0.00017197892884723842,
+      "rewards/margins": 0.0005675320862792432,
+      "rewards/rejected": -0.0003955531574320048,
       "step": 10
     },
     {
       "epoch": 0.05,
-      "learning_rate": 2.5641025641025636e-07,
-      "logits/chosen": -1.9975776672363281,
-      "logits/rejected": -2.000213146209717,
-      "logps/chosen": -29.647293090820312,
-      "logps/rejected": -29.056921005249023,
-      "loss": 1.0105,
-      "rewards/accuracies": 0.5,
-      "rewards/chosen": -0.00459072832018137,
-      "rewards/margins": -0.010462107136845589,
-      "rewards/rejected": 0.005871377885341644,
+      "learning_rate": 2.564102564102564e-06,
+      "logits/chosen": -1.9978935718536377,
+      "logits/rejected": -2.000532627105713,
+      "logps/chosen": -29.66562843322754,
+      "logps/rejected": -29.045883178710938,
+      "loss": 0.6934,
+      "rewards/accuracies": 0.36250001192092896,
+      "rewards/chosen": -0.00023434234026353806,
+      "rewards/margins": -0.0004099405778106302,
+      "rewards/rejected": 0.0001755982666509226,
       "step": 20
     },
     {
       "epoch": 0.08,
-      "learning_rate": 3.8461538461538463e-07,
-      "logits/chosen": -1.920013189315796,
-      "logits/rejected": -1.917337417602539,
-      "logps/chosen": -31.4055233001709,
-      "logps/rejected": -33.22685623168945,
-      "loss": 0.9829,
-      "rewards/accuracies": 0.5625,
-      "rewards/chosen": 0.009516814723610878,
-      "rewards/margins": 0.01707320474088192,
-      "rewards/rejected": -0.007556394673883915,
+      "learning_rate": 3.846153846153847e-06,
+      "logits/chosen": -1.9211324453353882,
+      "logits/rejected": -1.9184545278549194,
+      "logps/chosen": -31.41294288635254,
+      "logps/rejected": -33.23053741455078,
+      "loss": 0.6931,
+      "rewards/accuracies": 0.5249999761581421,
+      "rewards/chosen": 3.152530553052202e-05,
+      "rewards/margins": 0.000152341352077201,
+      "rewards/rejected": -0.00012081606837455183,
       "step": 30
     },
     {
       "epoch": 0.1,
-      "learning_rate": 4.999896948438433e-07,
-      "logits/chosen": -2.0167124271392822,
-      "logits/rejected": -2.0079712867736816,
-      "logps/chosen": -32.590545654296875,
-      "logps/rejected": -32.499671936035156,
-      "loss": 1.0241,
-      "rewards/accuracies": 0.4375,
-      "rewards/chosen": -0.01236567273736,
-      "rewards/margins": -0.024063030257821083,
-      "rewards/rejected": 0.011697360314428806,
+      "learning_rate": 4.999896948438434e-06,
+      "logits/chosen": -2.017341375350952,
+      "logits/rejected": -2.0086092948913574,
+      "logps/chosen": -32.60146713256836,
+      "logps/rejected": -32.49399185180664,
+      "loss": 0.6934,
+      "rewards/accuracies": 0.4124999940395355,
+      "rewards/chosen": -0.0002466029836796224,
+      "rewards/margins": -0.0004333632532507181,
+      "rewards/rejected": 0.00018676018225960433,
       "step": 40
     },
     {
       "epoch": 0.13,
-      "learning_rate": 4.987541037542186e-07,
-      "logits/chosen": -1.8641105890274048,
-      "logits/rejected": -1.8533456325531006,
-      "logps/chosen": -33.56208038330078,
-      "logps/rejected": -35.437400817871094,
-      "loss": 1.0176,
-      "rewards/accuracies": 0.4375,
-      "rewards/chosen": -0.004139312542974949,
-      "rewards/margins": -0.017586929723620415,
-      "rewards/rejected": 0.013447612524032593,
+      "learning_rate": 4.987541037542187e-06,
+      "logits/chosen": -1.862633466720581,
+      "logits/rejected": -1.8518692255020142,
+      "logps/chosen": -33.55931091308594,
+      "logps/rejected": -35.44870376586914,
+      "loss": 0.6932,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -1.831089502957184e-05,
+      "rewards/margins": -5.47249146620743e-05,
+      "rewards/rejected": 3.641402145149186e-05,
       "step": 50
     },
     {
       "epoch": 0.16,
-      "learning_rate": 4.954691471941118e-07,
-      "logits/chosen": -1.9452192783355713,
-      "logits/rejected": -1.9471514225006104,
-      "logps/chosen": -32.587867736816406,
-      "logps/rejected": -33.20810317993164,
-      "loss": 0.9559,
-      "rewards/accuracies": 0.5874999761581421,
-      "rewards/chosen": 0.01108462829142809,
-      "rewards/margins": 0.04413585737347603,
-      "rewards/rejected": -0.033051230013370514,
+      "learning_rate": 4.954691471941119e-06,
+      "logits/chosen": -1.9409154653549194,
+      "logits/rejected": -1.9428699016571045,
+      "logps/chosen": -32.53916549682617,
+      "logps/rejected": -33.24130630493164,
+      "loss": 0.6925,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": 0.0006101715262047946,
+      "rewards/margins": 0.0013094183523207903,
+      "rewards/rejected": -0.0006992466514930129,
       "step": 60
     },
     {
       "epoch": 0.18,
-      "learning_rate": 4.901618883413548e-07,
-      "logits/chosen": -2.0798697471618652,
-      "logits/rejected": -2.084836483001709,
-      "logps/chosen": -33.99953079223633,
-      "logps/rejected": -36.59130859375,
-      "loss": 0.9897,
-      "rewards/accuracies": 0.4749999940395355,
-      "rewards/chosen": -0.009940117597579956,
-      "rewards/margins": 0.010251833125948906,
-      "rewards/rejected": -0.020191945135593414,
+      "learning_rate": 4.901618883413549e-06,
+      "logits/chosen": -2.070591926574707,
+      "logits/rejected": -2.075544834136963,
+      "logps/chosen": -34.023067474365234,
+      "logps/rejected": -36.647151947021484,
+      "loss": 0.6929,
+      "rewards/accuracies": 0.5375000238418579,
+      "rewards/chosen": -0.00034581663203425705,
+      "rewards/margins": 0.0004369783273432404,
+      "rewards/rejected": -0.0007827949011698365,
       "step": 70
     },
     {
       "epoch": 0.21,
-      "learning_rate": 4.828760511501322e-07,
-      "logits/chosen": -1.9422235488891602,
-      "logits/rejected": -1.9453849792480469,
-      "logps/chosen": -34.411861419677734,
-      "logps/rejected": -34.599830627441406,
-      "loss": 0.979,
-      "rewards/accuracies": 0.5625,
-      "rewards/chosen": 0.014353242702782154,
-      "rewards/margins": 0.02102992869913578,
-      "rewards/rejected": -0.006676685996353626,
+      "learning_rate": 4.828760511501322e-06,
+      "logits/chosen": -1.9308092594146729,
+      "logits/rejected": -1.933943748474121,
+      "logps/chosen": -34.318023681640625,
+      "logps/rejected": -34.67802429199219,
+      "loss": 0.6922,
+      "rewards/accuracies": 0.5375000238418579,
+      "rewards/chosen": 0.0010978971840813756,
+      "rewards/margins": 0.0019540609791874886,
+      "rewards/rejected": -0.000856163795106113,
       "step": 80
     },
     {
       "epoch": 0.23,
-      "learning_rate": 4.736716601303429e-07,
-      "logits/chosen": -1.950791358947754,
-      "logits/rejected": -1.9553003311157227,
-      "logps/chosen": -32.45661163330078,
-      "logps/rejected": -32.35947799682617,
-      "loss": 0.9898,
-      "rewards/accuracies": 0.5375000238418579,
-      "rewards/chosen": 0.009730497375130653,
-      "rewards/margins": 0.010243801400065422,
-      "rewards/rejected": -0.0005133040249347687,
+      "learning_rate": 4.7367166013034295e-06,
+      "logits/chosen": -1.9380912780761719,
+      "logits/rejected": -1.9425855875015259,
+      "logps/chosen": -32.38385009765625,
+      "logps/rejected": -32.35346603393555,
+      "loss": 0.6928,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.0008357145707122982,
+      "rewards/margins": 0.0007813175907358527,
+      "rewards/rejected": 5.439693995867856e-05,
       "step": 90
     },
     {
       "epoch": 0.26,
-      "learning_rate": 4.62624545834521e-07,
-      "logits/chosen": -2.0489749908447266,
-      "logits/rejected": -2.046968460083008,
-      "logps/chosen": -32.233360290527344,
-      "logps/rejected": -31.278451919555664,
-      "loss": 0.9823,
-      "rewards/accuracies": 0.512499988079071,
-      "rewards/chosen": 0.006729811429977417,
-      "rewards/margins": 0.01774778962135315,
-      "rewards/rejected": -0.011017980054020882,
+      "learning_rate": 4.626245458345211e-06,
+      "logits/chosen": -2.035137891769409,
+      "logits/rejected": -2.0331528186798096,
+      "logps/chosen": -32.112831115722656,
+      "logps/rejected": -31.29166030883789,
+      "loss": 0.6924,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.0012800416443496943,
+      "rewards/margins": 0.0015345367137342691,
+      "rewards/rejected": -0.0002544948656577617,
       "step": 100
     },
     {
       "epoch": 0.26,
-      "eval_logits/chosen": -2.243974447250366,
-      "eval_logits/rejected": -2.239091634750366,
-      "eval_logps/chosen": -34.021610260009766,
-      "eval_logps/rejected": -37.50578689575195,
-      "eval_loss": 0.9981797933578491,
-      "eval_rewards/accuracies": 0.5103820562362671,
-      "eval_rewards/chosen": 0.011647163890302181,
-      "eval_rewards/margins": 0.0018998426385223866,
-      "eval_rewards/rejected": 0.009747318923473358,
-      "eval_runtime": 146.0735,
-      "eval_samples_per_second": 2.348,
-      "eval_steps_per_second": 0.294,
+      "eval_logits/chosen": -2.2300801277160645,
+      "eval_logits/rejected": -2.225238084793091,
+      "eval_logps/chosen": -34.04683303833008,
+      "eval_logps/rejected": -37.53927230834961,
+      "eval_loss": 0.6930972337722778,
+      "eval_rewards/accuracies": 0.5186877250671387,
+      "eval_rewards/chosen": -0.00012280470400583,
+      "eval_rewards/margins": 0.00010372586984885857,
+      "eval_rewards/rejected": -0.0002265305956825614,
+      "eval_runtime": 145.7259,
+      "eval_samples_per_second": 2.354,
+      "eval_steps_per_second": 0.295,
       "step": 100
     },
     {
       "epoch": 0.29,
-      "learning_rate": 4.4982572012636904e-07,
-      "logits/chosen": -2.0054612159729004,
-      "logits/rejected": -2.003042221069336,
-      "logps/chosen": -33.2550048828125,
-      "logps/rejected": -34.02202224731445,
-      "loss": 1.0167,
-      "rewards/accuracies": 0.44999998807907104,
-      "rewards/chosen": -0.010068392381072044,
-      "rewards/margins": -0.016683781519532204,
-      "rewards/rejected": 0.006615391466766596,
+      "learning_rate": 4.498257201263691e-06,
+      "logits/chosen": -1.989782691001892,
+      "logits/rejected": -1.9873950481414795,
+      "logps/chosen": -33.12385559082031,
+      "logps/rejected": -34.011810302734375,
+      "loss": 0.6926,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": 0.0011996207758784294,
+      "rewards/margins": 0.001024017808958888,
+      "rewards/rejected": 0.00017560287960805,
       "step": 110
     },
     {
       "epoch": 0.31,
-      "learning_rate": 4.353806263777677e-07,
-      "logits/chosen": -2.016511917114258,
-      "logits/rejected": -2.008146047592163,
-      "logps/chosen": -32.45395278930664,
-      "logps/rejected": -32.17681121826172,
-      "loss": 0.9944,
-      "rewards/accuracies": 0.5249999761581421,
-      "rewards/chosen": -0.008141009137034416,
-      "rewards/margins": 0.00563241308555007,
-      "rewards/rejected": -0.013773423619568348,
+      "learning_rate": 4.353806263777678e-06,
+      "logits/chosen": -2.0008151531219482,
+      "logits/rejected": -1.992500901222229,
+      "logps/chosen": -32.320838928222656,
+      "logps/rejected": -32.128170013427734,
+      "loss": 0.6927,
+      "rewards/accuracies": 0.512499988079071,
+      "rewards/chosen": 0.001240686746314168,
+      "rewards/margins": 0.0009073130786418915,
+      "rewards/rejected": 0.0003333735803607851,
       "step": 120
     },
     {
       "epoch": 0.34,
-      "learning_rate": 4.194082707715275e-07,
-      "logits/chosen": -2.0463106632232666,
-      "logits/rejected": -2.0382590293884277,
-      "logps/chosen": -30.476207733154297,
-      "logps/rejected": -32.020484924316406,
-      "loss": 1.0297,
-      "rewards/accuracies": 0.3499999940395355,
-      "rewards/chosen": -0.00044600124238058925,
-      "rewards/margins": -0.029714178293943405,
-      "rewards/rejected": 0.02926817536354065,
+      "learning_rate": 4.1940827077152755e-06,
+      "logits/chosen": -2.0289230346679688,
+      "logits/rejected": -2.020946502685547,
+      "logps/chosen": -30.313907623291016,
+      "logps/rejected": -32.086116790771484,
+      "loss": 0.6922,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.0016180993989109993,
+      "rewards/margins": 0.0019491963321343064,
+      "rewards/rejected": -0.00033109664218500257,
       "step": 130
     },
     {
       "epoch": 0.36,
-      "learning_rate": 4.020402418666621e-07,
-      "logits/chosen": -1.9767357110977173,
-      "logits/rejected": -1.986999273300171,
-      "logps/chosen": -31.381155014038086,
-      "logps/rejected": -32.54108428955078,
-      "loss": 0.9598,
-      "rewards/accuracies": 0.5375000238418579,
-      "rewards/chosen": 0.034655503928661346,
-      "rewards/margins": 0.04024083539843559,
-      "rewards/rejected": -0.005585335195064545,
+      "learning_rate": 4.0204024186666215e-06,
+      "logits/chosen": -1.9592479467391968,
+      "logits/rejected": -1.9694607257843018,
+      "logps/chosen": -31.223953247070312,
+      "logps/rejected": -32.547454833984375,
+      "loss": 0.6921,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.0019570994190871716,
+      "rewards/margins": 0.002082846825942397,
+      "rewards/rejected": -0.00012574761058203876,
       "step": 140
     },
     {
       "epoch": 0.39,
-      "learning_rate": 3.8341962650351185e-07,
-      "logits/chosen": -1.8906068801879883,
-      "logits/rejected": -1.8916860818862915,
-      "logps/chosen": -34.20075225830078,
-      "logps/rejected": -34.7618522644043,
-      "loss": 1.0031,
-      "rewards/accuracies": 0.4124999940395355,
-      "rewards/chosen": -0.007252261973917484,
-      "rewards/margins": -0.0031175222247838974,
-      "rewards/rejected": -0.004134741611778736,
+      "learning_rate": 3.834196265035119e-06,
+      "logits/chosen": -1.8708124160766602,
+      "logits/rejected": -1.8719879388809204,
+      "logps/chosen": -33.877174377441406,
+      "logps/rejected": -34.78774642944336,
+      "loss": 0.6914,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.003155181184411049,
+      "rewards/margins": 0.0034600873477756977,
+      "rewards/rejected": -0.00030490627977997065,
       "step": 150
     },
     {
       "epoch": 0.42,
-      "learning_rate": 3.636998309800572e-07,
-      "logits/chosen": -1.942251443862915,
-      "logits/rejected": -1.9387743473052979,
-      "logps/chosen": -36.145790100097656,
-      "logps/rejected": -32.73196029663086,
-      "loss": 0.9684,
-      "rewards/accuracies": 0.512499988079071,
-      "rewards/chosen": 0.023301539942622185,
-      "rewards/margins": 0.031586963683366776,
-      "rewards/rejected": -0.008285422809422016,
+      "learning_rate": 3.636998309800573e-06,
+      "logits/chosen": -1.9215673208236694,
+      "logits/rejected": -1.9181665182113647,
+      "logps/chosen": -36.011531829833984,
+      "logps/rejected": -32.685707092285156,
+      "loss": 0.6925,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": 0.001601455733180046,
+      "rewards/margins": 0.001231002388522029,
+      "rewards/rejected": 0.00037045328645035625,
       "step": 160
     },
     {
       "epoch": 0.44,
-      "learning_rate": 3.430433172111807e-07,
-      "logits/chosen": -2.041952610015869,
-      "logits/rejected": -2.0345559120178223,
-      "logps/chosen": -33.81856155395508,
-      "logps/rejected": -31.359323501586914,
-      "loss": 1.0119,
-      "rewards/accuracies": 0.4749999940395355,
-      "rewards/chosen": -0.018632039427757263,
-      "rewards/margins": -0.011905943043529987,
-      "rewards/rejected": -0.006726093590259552,
+      "learning_rate": 3.4304331721118078e-06,
+      "logits/chosen": -2.021604061126709,
+      "logits/rejected": -2.014291524887085,
+      "logps/chosen": -33.482086181640625,
+      "logps/rejected": -31.404422760009766,
+      "loss": 0.6913,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": 0.0031577465124428272,
+      "rewards/margins": 0.003683448536321521,
+      "rewards/rejected": -0.0005257020820863545,
       "step": 170
     },
     {
       "epoch": 0.47,
-      "learning_rate": 3.216202642830543e-07,
-      "logits/chosen": -2.0473170280456543,
-      "logits/rejected": -2.0525729656219482,
-      "logps/chosen": -32.52764129638672,
-      "logps/rejected": -32.50192642211914,
-      "loss": 0.9695,
-      "rewards/accuracies": 0.574999988079071,
-      "rewards/chosen": 0.015429094433784485,
-      "rewards/margins": 0.030534446239471436,
-      "rewards/rejected": -0.015105349011719227,
+      "learning_rate": 3.2162026428305436e-06,
+      "logits/chosen": -2.027444839477539,
+      "logits/rejected": -2.032665729522705,
+      "logps/chosen": -32.183101654052734,
+      "logps/rejected": -32.39936065673828,
+      "loss": 0.6918,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.0036168727092444897,
+      "rewards/margins": 0.0027590212412178516,
+      "rewards/rejected": 0.0008578516426496208,
       "step": 180
     },
     {
       "epoch": 0.49,
-      "learning_rate": 2.9960716642946403e-07,
-      "logits/chosen": -2.048530340194702,
-      "logits/rejected": -2.045767307281494,
-      "logps/chosen": -31.485219955444336,
-      "logps/rejected": -31.34262466430664,
-      "loss": 0.9747,
-      "rewards/accuracies": 0.5874999761581421,
-      "rewards/chosen": 0.003360496135428548,
-      "rewards/margins": 0.02533094584941864,
-      "rewards/rejected": -0.02197045087814331,
+      "learning_rate": 2.996071664294641e-06,
+      "logits/chosen": -2.027879238128662,
+      "logits/rejected": -2.025132656097412,
+      "logps/chosen": -31.258464813232422,
+      "logps/rejected": -31.348388671875,
+      "loss": 0.6919,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.0023048892617225647,
+      "rewards/margins": 0.0026066480204463005,
+      "rewards/rejected": -0.000301758642308414,
       "step": 190
     },
     {
       "epoch": 0.52,
-      "learning_rate": 2.771853789806683e-07,
-      "logits/chosen": -1.918745756149292,
-      "logits/rejected": -1.9234097003936768,
-      "logps/chosen": -31.576425552368164,
-      "logps/rejected": -32.79480743408203,
-      "loss": 0.9696,
-      "rewards/accuracies": 0.5375000238418579,
-      "rewards/chosen": 0.026870254427194595,
-      "rewards/margins": 0.03035963699221611,
-      "rewards/rejected": -0.0034893869888037443,
+      "learning_rate": 2.7718537898066833e-06,
+      "logits/chosen": -1.8983129262924194,
+      "logits/rejected": -1.902967095375061,
+      "logps/chosen": -31.276391983032227,
+      "logps/rejected": -32.81935119628906,
+      "loss": 0.6914,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": 0.0032989257015287876,
+      "rewards/margins": 0.00358308176510036,
+      "rewards/rejected": -0.00028415597626008093,
       "step": 200
     },
     {
       "epoch": 0.52,
-      "eval_logits/chosen": -2.243912696838379,
-      "eval_logits/rejected": -2.2390317916870117,
-      "eval_logps/chosen": -34.021690368652344,
-      "eval_logps/rejected": -37.502418518066406,
-      "eval_loss": 1.001371145248413,
-      "eval_rewards/accuracies": 0.4692690968513489,
-      "eval_rewards/chosen": 0.011576101183891296,
-      "eval_rewards/margins": -0.0012021720176562667,
-      "eval_rewards/rejected": 0.012778270989656448,
-      "eval_runtime": 145.8493,
-      "eval_samples_per_second": 2.352,
+      "eval_logits/chosen": -2.2249655723571777,
+      "eval_logits/rejected": -2.220139503479004,
+      "eval_logps/chosen": -34.04255294799805,
+      "eval_logps/rejected": -37.55300521850586,
+      "eval_loss": 0.6930080056190491,
+      "eval_rewards/accuracies": 0.5245016813278198,
+      "eval_rewards/chosen": -8.006239659152925e-05,
+      "eval_rewards/margins": 0.00028380370349623263,
+      "eval_rewards/rejected": -0.00036386612919159234,
+      "eval_runtime": 145.5269,
+      "eval_samples_per_second": 2.357,
       "eval_steps_per_second": 0.295,
       "step": 200
     },
     {
       "epoch": 0.55,
-      "learning_rate": 2.5453962426402e-07,
-      "logits/chosen": -2.0318515300750732,
-      "logits/rejected": -2.0425400733947754,
-      "logps/chosen": -31.96026039123535,
-      "logps/rejected": -33.8691291809082,
-      "loss": 0.9843,
-      "rewards/accuracies": 0.625,
-      "rewards/chosen": 0.009936656802892685,
-      "rewards/margins": 0.015721982344985008,
-      "rewards/rejected": -0.005785329267382622,
+      "learning_rate": 2.5453962426402006e-06,
+      "logits/chosen": -2.010593891143799,
+      "logits/rejected": -2.021207332611084,
+      "logps/chosen": -31.7437801361084,
+      "logps/rejected": -33.93886947631836,
+      "loss": 0.6916,
+      "rewards/accuracies": 0.574999988079071,
+      "rewards/chosen": 0.0022752191871404648,
+      "rewards/margins": 0.003036911366507411,
+      "rewards/rejected": -0.000761692295782268,
       "step": 210
     },
     {
       "epoch": 0.57,
-      "learning_rate": 2.318564697655179e-07,
-      "logits/chosen": -1.9256188869476318,
-      "logits/rejected": -1.9404585361480713,
-      "logps/chosen": -30.080001831054688,
-      "logps/rejected": -31.53633689880371,
-      "loss": 0.9913,
-      "rewards/accuracies": 0.512499988079071,
-      "rewards/chosen": 0.028384273871779442,
-      "rewards/margins": 0.00873211957514286,
-      "rewards/rejected": 0.01965215802192688,
+      "learning_rate": 2.3185646976551794e-06,
+      "logits/chosen": -1.9029136896133423,
+      "logits/rejected": -1.9176632165908813,
+      "logps/chosen": -29.78145408630371,
+      "logps/rejected": -31.63638687133789,
+      "loss": 0.6911,
+      "rewards/accuracies": 0.7749999761581421,
+      "rewards/chosen": 0.003300876123830676,
+      "rewards/margins": 0.0040829661302268505,
+      "rewards/rejected": -0.000782089657150209,
       "step": 220
     },
     {
       "epoch": 0.6,
-      "learning_rate": 2.093227910899832e-07,
-      "logits/chosen": -1.9830671548843384,
-      "logits/rejected": -1.9870269298553467,
-      "logps/chosen": -33.386474609375,
-      "logps/rejected": -31.55719566345215,
-      "loss": 0.965,
-      "rewards/accuracies": 0.5249999761581421,
-      "rewards/chosen": 0.02789575420320034,
-      "rewards/margins": 0.03499894589185715,
-      "rewards/rejected": -0.007103185169398785,
+      "learning_rate": 2.0932279108998323e-06,
+      "logits/chosen": -1.9591538906097412,
+      "logits/rejected": -1.9631026983261108,
+      "logps/chosen": -33.05189895629883,
+      "logps/rejected": -31.594707489013672,
+      "loss": 0.6911,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": 0.003655704203993082,
+      "rewards/margins": 0.004109731875360012,
+      "rewards/rejected": -0.00045402703108265996,
       "step": 230
     },
     {
       "epoch": 0.62,
-      "learning_rate": 1.8712423238279356e-07,
-      "logits/chosen": -1.9833042621612549,
-      "logits/rejected": -1.96132493019104,
-      "logps/chosen": -34.13536834716797,
-      "logps/rejected": -34.952049255371094,
-      "loss": 0.9823,
-      "rewards/accuracies": 0.5375000238418579,
-      "rewards/chosen": 0.016542484983801842,
-      "rewards/margins": 0.01767394319176674,
-      "rewards/rejected": -0.0011314575094729662,
+      "learning_rate": 1.8712423238279358e-06,
+      "logits/chosen": -1.9572566747665405,
+      "logits/rejected": -1.9354870319366455,
+      "logps/chosen": -33.83857727050781,
+      "logps/rejected": -35.12303924560547,
+      "loss": 0.6907,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": 0.0031517534516751766,
+      "rewards/margins": 0.004874187987297773,
+      "rewards/rejected": -0.0017224351176992059,
       "step": 240
     },
     {
       "epoch": 0.65,
-      "learning_rate": 1.654436768970182e-07,
-      "logits/chosen": -2.0247886180877686,
-      "logits/rejected": -2.0214812755584717,
-      "logps/chosen": -32.913856506347656,
-      "logps/rejected": -36.23151397705078,
-      "loss": 0.984,
-      "rewards/accuracies": 0.5625,
-      "rewards/chosen": 0.010092089883983135,
-      "rewards/margins": 0.016022875905036926,
-      "rewards/rejected": -0.0059307836927473545,
+      "learning_rate": 1.6544367689701824e-06,
+      "logits/chosen": -1.998875379562378,
+      "logits/rejected": -1.9955555200576782,
+      "logps/chosen": -32.72559356689453,
+      "logps/rejected": -36.2435417175293,
+      "loss": 0.6921,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": 0.0019947518594563007,
+      "rewards/margins": 0.0021809376776218414,
+      "rewards/rejected": -0.0001861859782366082,
       "step": 250
     },
     {
       "epoch": 0.68,
-      "learning_rate": 1.444597403062196e-07,
-      "logits/chosen": -1.8914467096328735,
-      "logits/rejected": -1.8890043497085571,
-      "logps/chosen": -34.18690872192383,
-      "logps/rejected": -35.524742126464844,
-      "loss": 0.9806,
-      "rewards/accuracies": 0.5375000238418579,
-      "rewards/chosen": 0.0013398710871115327,
-      "rewards/margins": 0.019445054233074188,
-      "rewards/rejected": -0.01810518093407154,
+      "learning_rate": 1.4445974030621963e-06,
+      "logits/chosen": -1.8661177158355713,
+      "logits/rejected": -1.8636993169784546,
+      "logps/chosen": -33.959014892578125,
+      "logps/rejected": -35.526344299316406,
+      "loss": 0.6919,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.0022938635665923357,
+      "rewards/margins": 0.0025111136492341757,
+      "rewards/rejected": -0.0002172500389860943,
       "step": 260
     },
     {
       "epoch": 0.7,
-      "learning_rate": 1.2434529917578887e-07,
-      "logits/chosen": -1.8761389255523682,
-      "logits/rejected": -1.8735923767089844,
-      "logps/chosen": -34.396785736083984,
-      "logps/rejected": -31.74139976501465,
-      "loss": 1.0262,
-      "rewards/accuracies": 0.4000000059604645,
-      "rewards/chosen": 0.000734941684640944,
-      "rewards/margins": -0.026202131062746048,
-      "rewards/rejected": 0.02693706750869751,
+      "learning_rate": 1.243452991757889e-06,
+      "logits/chosen": -1.8511241674423218,
+      "logits/rejected": -1.8487510681152344,
+      "logps/chosen": -34.16337585449219,
+      "logps/rejected": -31.830408096313477,
+      "loss": 0.6917,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.002342230873182416,
+      "rewards/margins": 0.0029330006800591946,
+      "rewards/rejected": -0.0005907699232921004,
       "step": 270
     },
     {
       "epoch": 0.73,
-      "learning_rate": 1.0526606671603521e-07,
-      "logits/chosen": -1.9801757335662842,
-      "logits/rejected": -1.9695546627044678,
-      "logps/chosen": -35.34246063232422,
-      "logps/rejected": -31.823230743408203,
-      "loss": 1.0032,
-      "rewards/accuracies": 0.4749999940395355,
-      "rewards/chosen": 0.012763315811753273,
-      "rewards/margins": -0.003222113475203514,
-      "rewards/rejected": 0.015985429286956787,
+      "learning_rate": 1.0526606671603523e-06,
+      "logits/chosen": -1.9535648822784424,
+      "logits/rejected": -1.943101167678833,
+      "logps/chosen": -35.01304244995117,
+      "logps/rejected": -31.87521743774414,
+      "loss": 0.6913,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.0034359837882220745,
+      "rewards/margins": 0.0037782168947160244,
+      "rewards/rejected": -0.0003422332229092717,
       "step": 280
     },
     {
       "epoch": 0.75,
-      "learning_rate": 8.737922755071453e-08,
-      "logits/chosen": -2.075316905975342,
-      "logits/rejected": -2.0602917671203613,
-      "logps/chosen": -30.897375106811523,
-      "logps/rejected": -32.64093780517578,
-      "loss": 0.9857,
-      "rewards/accuracies": 0.5249999761581421,
-      "rewards/chosen": 0.024072837084531784,
-      "rewards/margins": 0.014264687895774841,
-      "rewards/rejected": 0.009808152914047241,
+      "learning_rate": 8.737922755071455e-07,
+      "logits/chosen": -2.048783302307129,
+      "logits/rejected": -2.0339112281799316,
+      "logps/chosen": -30.716812133789062,
+      "logps/rejected": -32.62614059448242,
+      "loss": 0.6922,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": 0.0020731096155941486,
+      "rewards/margins": 0.001816184027120471,
+      "rewards/rejected": 0.00025692558847367764,
       "step": 290
     },
     {
       "epoch": 0.78,
-      "learning_rate": 7.08321427484816e-08,
-      "logits/chosen": -1.946758508682251,
-      "logits/rejected": -1.9442226886749268,
-      "logps/chosen": -32.89502716064453,
-      "logps/rejected": -30.831274032592773,
-      "loss": 0.9581,
-      "rewards/accuracies": 0.5625,
-      "rewards/chosen": 0.0288736280053854,
-      "rewards/margins": 0.04189068451523781,
-      "rewards/rejected": -0.013017055578529835,
+      "learning_rate": 7.08321427484816e-07,
+      "logits/chosen": -1.9198474884033203,
+      "logits/rejected": -1.917340636253357,
+      "logps/chosen": -32.29683303833008,
+      "logps/rejected": -30.91409683227539,
+      "loss": 0.6895,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": 0.0063027567230165005,
+      "rewards/margins": 0.007275627460330725,
+      "rewards/rejected": -0.0009728703880682588,
       "step": 300
     },
     {
       "epoch": 0.78,
-      "eval_logits/chosen": -2.2437047958374023,
-      "eval_logits/rejected": -2.2388339042663574,
-      "eval_logps/chosen": -34.03062438964844,
-      "eval_logps/rejected": -37.499412536621094,
-      "eval_loss": 1.0120128393173218,
-      "eval_rewards/accuracies": 0.472591370344162,
-      "eval_rewards/chosen": 0.003532935632392764,
-      "eval_rewards/margins": -0.011953731998801231,
-      "eval_rewards/rejected": 0.015486669726669788,
-      "eval_runtime": 145.7744,
+      "eval_logits/chosen": -2.221111297607422,
+      "eval_logits/rejected": -2.2162926197052,
+      "eval_logps/chosen": -34.0648307800293,
+      "eval_logps/rejected": -37.58684158325195,
+      "eval_loss": 0.6929495930671692,
+      "eval_rewards/accuracies": 0.5419435501098633,
+      "eval_rewards/chosen": -0.0003027978236787021,
+      "eval_rewards/margins": 0.0003993964346591383,
+      "eval_rewards/rejected": -0.0007021942874416709,
+      "eval_runtime": 145.7415,
       "eval_samples_per_second": 2.353,
       "eval_steps_per_second": 0.295,
       "step": 300
     },
     {
       "epoch": 0.81,
-      "learning_rate": 5.576113578589034e-08,
-      "logits/chosen": -1.928391695022583,
-      "logits/rejected": -1.925132393836975,
-      "logps/chosen": -31.573719024658203,
-      "logps/rejected": -33.733333587646484,
-      "loss": 0.9813,
-      "rewards/accuracies": 0.574999988079071,
-      "rewards/chosen": 0.026669049635529518,
-      "rewards/margins": 0.01873381994664669,
-      "rewards/rejected": 0.007935228757560253,
+      "learning_rate": 5.576113578589035e-07,
+      "logits/chosen": -1.905469298362732,
+      "logits/rejected": -1.9022337198257446,
+      "logps/chosen": -31.291845321655273,
+      "logps/rejected": -33.816497802734375,
+      "loss": 0.6912,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": 0.003115057712420821,
+      "rewards/margins": 0.003858551848679781,
+      "rewards/rejected": -0.0007434942526742816,
       "step": 310
     },
     {
       "epoch": 0.83,
-      "learning_rate": 4.229036944380912e-08,
-      "logits/chosen": -1.9808565378189087,
-      "logits/rejected": -1.9685465097427368,
-      "logps/chosen": -34.569026947021484,
-      "logps/rejected": -33.5385856628418,
-      "loss": 0.9759,
-      "rewards/accuracies": 0.5874999761581421,
-      "rewards/chosen": 0.02433854714035988,
-      "rewards/margins": 0.024108227342367172,
-      "rewards/rejected": 0.00023032091849017888,
+      "learning_rate": 4.229036944380913e-07,
+      "logits/chosen": -1.9548442363739014,
+      "logits/rejected": -1.9426848888397217,
+      "logps/chosen": -34.22846603393555,
+      "logps/rejected": -33.660888671875,
+      "loss": 0.6907,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.0036761020310223103,
+      "rewards/margins": 0.004896566737443209,
+      "rewards/rejected": -0.0012204641243442893,
       "step": 320
     },
     {
       "epoch": 0.86,
-      "learning_rate": 3.053082288996112e-08,
-      "logits/chosen": -2.0160164833068848,
-      "logits/rejected": -2.0145697593688965,
-      "logps/chosen": -33.470794677734375,
-      "logps/rejected": -32.465301513671875,
-      "loss": 0.9994,
-      "rewards/accuracies": 0.5,
-      "rewards/chosen": 0.013503548689186573,
-      "rewards/margins": 0.0006495133275166154,
-      "rewards/rejected": 0.012854036875069141,
+      "learning_rate": 3.053082288996112e-07,
+      "logits/chosen": -1.9896023273468018,
+      "logits/rejected": -1.9881556034088135,
+      "logps/chosen": -33.096466064453125,
+      "logps/rejected": -32.550086975097656,
+      "loss": 0.6909,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": 0.0038932852912694216,
+      "rewards/margins": 0.004598308354616165,
+      "rewards/rejected": -0.0007050234125927091,
       "step": 330
     },
     {
       "epoch": 0.88,
-      "learning_rate": 2.05793773749158e-08,
-      "logits/chosen": -2.1030516624450684,
-      "logits/rejected": -2.087252140045166,
-      "logps/chosen": -34.154754638671875,
-      "logps/rejected": -33.087013244628906,
-      "loss": 1.0057,
-      "rewards/accuracies": 0.512499988079071,
-      "rewards/chosen": 0.02217467688024044,
-      "rewards/margins": -0.005742859095335007,
-      "rewards/rejected": 0.027917543426156044,
+      "learning_rate": 2.0579377374915805e-07,
+      "logits/chosen": -2.0761990547180176,
+      "logits/rejected": -2.0605435371398926,
+      "logps/chosen": -33.7951774597168,
+      "logps/rejected": -33.09763717651367,
+      "loss": 0.6913,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.0038421000353991985,
+      "rewards/margins": 0.0036381154786795378,
+      "rewards/rejected": 0.00020398428023327142,
       "step": 340
     },
     {
       "epoch": 0.91,
-      "learning_rate": 1.251801807404168e-08,
-      "logits/chosen": -1.9748103618621826,
-      "logits/rejected": -1.9738750457763672,
-      "logps/chosen": -33.23518371582031,
-      "logps/rejected": -32.46066665649414,
-      "loss": 0.9618,
-      "rewards/accuracies": 0.574999988079071,
-      "rewards/chosen": 0.04222174361348152,
-      "rewards/margins": 0.03818847984075546,
-      "rewards/rejected": 0.004033264704048634,
+      "learning_rate": 1.2518018074041684e-07,
+      "logits/chosen": -1.9492791891098022,
+      "logits/rejected": -1.9484554529190063,
+      "logps/chosen": -32.777854919433594,
+      "logps/rejected": -32.48105239868164,
+      "loss": 0.6906,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.005042382050305605,
+      "rewards/margins": 0.00520141888409853,
+      "rewards/rejected": -0.00015903706662356853,
       "step": 350
     },
     {
       "epoch": 0.94,
-      "learning_rate": 6.41315865106129e-09,
-      "logits/chosen": -1.9303611516952515,
-      "logits/rejected": -1.940708875656128,
-      "logps/chosen": -32.200523376464844,
-      "logps/rejected": -35.27861785888672,
-      "loss": 1.0054,
-      "rewards/accuracies": 0.42500001192092896,
-      "rewards/chosen": 0.0037514485884457827,
-      "rewards/margins": -0.005373493768274784,
-      "rewards/rejected": 0.009124943986535072,
+      "learning_rate": 6.41315865106129e-08,
+      "logits/chosen": -1.9036123752593994,
+      "logits/rejected": -1.9138826131820679,
+      "logps/chosen": -31.845821380615234,
+      "logps/rejected": -35.32244873046875,
+      "loss": 0.6912,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": 0.003588717896491289,
+      "rewards/margins": 0.003925635013729334,
+      "rewards/rejected": -0.00033691702992655337,
       "step": 360
     },
     {
       "epoch": 0.96,
-      "learning_rate": 2.3150941078050324e-09,
-      "logits/chosen": -2.069342613220215,
-      "logits/rejected": -2.062788486480713,
-      "logps/chosen": -33.65803527832031,
-      "logps/rejected": -29.199758529663086,
-      "loss": 1.0295,
-      "rewards/accuracies": 0.5,
-      "rewards/chosen": -0.005459528882056475,
-      "rewards/margins": -0.029461050406098366,
-      "rewards/rejected": 0.02400151826441288,
+      "learning_rate": 2.3150941078050325e-08,
+      "logits/chosen": -2.043790102005005,
+      "logits/rejected": -2.0373566150665283,
+      "logps/chosen": -33.307098388671875,
+      "logps/rejected": -29.227703094482422,
+      "loss": 0.6914,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": 0.003448712872341275,
+      "rewards/margins": 0.0034615055192261934,
+      "rewards/rejected": -1.2792841516784392e-05,
       "step": 370
     },
     {
       "epoch": 0.99,
-      "learning_rate": 2.575864278703266e-10,
-      "logits/chosen": -1.9290788173675537,
-      "logits/rejected": -1.931239366531372,
-      "logps/chosen": -34.229915618896484,
-      "logps/rejected": -30.90093994140625,
-      "loss": 0.9663,
-      "rewards/accuracies": 0.5625,
-      "rewards/chosen": 0.00822196900844574,
-      "rewards/margins": 0.03370029479265213,
-      "rewards/rejected": -0.02547832392156124,
+      "learning_rate": 2.575864278703266e-09,
+      "logits/chosen": -1.9043537378311157,
+      "logits/rejected": -1.9065628051757812,
+      "logps/chosen": -33.83999252319336,
+      "logps/rejected": -30.96780776977539,
+      "loss": 0.6907,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": 0.003990581724792719,
+      "rewards/margins": 0.004942373372614384,
+      "rewards/rejected": -0.0009517916478216648,
       "step": 380
     },
     {
       "epoch": 1.0,
       "step": 385,
       "total_flos": 0.0,
-      "train_loss": 0.21873634759481853,
-      "train_runtime": 627.4596,
-      "train_samples_per_second": 4.907,
-      "train_steps_per_second": 0.614
+      "train_loss": 0.6918259131443965,
+      "train_runtime": 3249.6401,
+      "train_samples_per_second": 0.947,
+      "train_steps_per_second": 0.118
     }
   ],
   "logging_steps": 10,