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: 0.5060
-- Rewards/chosen: 0.0978
-- Rewards/rejected: 0.1310
-- Rewards/accuracies: 0.5220
-- Rewards/margins: -0.0333
-- Logps/rejected: -35.7793
-- Logps/chosen: -32.3035
-- Logits/rejected: 98.3593
-- Logits/chosen: 98.3666
+- Loss: 0.4962
+- Rewards/chosen: 0.0856
+- Rewards/rejected: 0.0600
+- Rewards/accuracies: 0.5104
+- Rewards/margins: 0.0256
+- Logps/rejected: -35.8166
+- Logps/chosen: -32.2293
+- Logits/rejected: 98.3704
+- Logits/chosen: 98.3655
 
 ## 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 |
 |:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
-| 0.461         | 0.26  | 100  | 0.5000          | 0.0677         | 0.0660           | 0.5158             | 0.0017          | -35.8723       | -32.3464     | 98.6883         | 98.7043       |
-| 0.3172        | 0.52  | 200  | 0.4956          | 0.0869         | 0.0578           | 0.5133             | 0.0291          | -35.8840       | -32.3190     | 98.4409         | 98.4463       |
-| 0.3303        | 0.78  | 300  | 0.4980          | 0.1165         | 0.1232           | 0.5017             | -0.0067         | -35.7905       | -32.2767     | 98.3630         | 98.3713       |
+| 0.4749        | 0.26  | 100  | 0.5047          | 0.0131         | 0.0430           | 0.4747             | -0.0299         | -35.8591       | -32.4105     | 98.7377         | 98.7500       |
+| 0.3843        | 0.52  | 200  | 0.4944          | 0.0238         | -0.0061          | 0.5307             | 0.0299          | -35.9817       | -32.3837     | 98.3784         | 98.3836       |
+| 0.363         | 0.78  | 300  | 0.4962          | 0.0856         | 0.0600           | 0.5104             | 0.0256          | -35.8166       | -32.2293     | 98.3704         | 98.3655       |
 
 
 ### Framework versions

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:968e193539784cf5983717c00c3e3960cd4875b80258301bd65477c94a0ea838
+oid sha256:4e09aeea11f61f666c4168ca9571269efa984ea3dde291fbc770977b3ec68bbb
 size 671150064

all_results.json

@@ -1,20 +1,7 @@
 {
     "epoch": 1.0,
-    "eval_logits/chosen": 98.3665542602539,
-    "eval_logits/rejected": 98.35928344726562,
-    "eval_logps/chosen": -32.303466796875,
-    "eval_logps/rejected": -35.77931594848633,
-    "eval_loss": 0.5059725642204285,
-    "eval_rewards/accuracies": 0.5220099687576294,
-    "eval_rewards/chosen": 0.09779632091522217,
-    "eval_rewards/margins": -0.033251672983169556,
-    "eval_rewards/rejected": 0.13104799389839172,
-    "eval_runtime": 103.819,
-    "eval_samples": 343,
-    "eval_samples_per_second": 3.304,
-    "eval_steps_per_second": 0.414,
-    "train_loss": 0.38857739250381274,
-    "train_runtime": 2559.074,
+    "train_loss": 0.4199758805237807,
+    "train_runtime": 2559.5376,
     "train_samples": 3079,
     "train_samples_per_second": 1.203,
     "train_steps_per_second": 0.15

train_results.json

@@ -1,7 +1,7 @@
 {
     "epoch": 1.0,
-    "train_loss": 0.38857739250381274,
-    "train_runtime": 2559.074,
+    "train_loss": 0.4199758805237807,
+    "train_runtime": 2559.5376,
     "train_samples": 3079,
     "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": 22.875,
+      "grad_norm": 13.0625,
       "learning_rate": 1.282051282051282e-07,
       "logits/chosen": 88.18099975585938,
       "logits/rejected": 88.25153350830078,
@@ -25,628 +25,628 @@
     },
     {
       "epoch": 0.03,
-      "grad_norm": 8.75,
+      "grad_norm": 5.375,
       "learning_rate": 1.282051282051282e-06,
-      "logits/chosen": 81.07839965820312,
-      "logits/rejected": 80.78140258789062,
-      "logps/chosen": -34.22772216796875,
-      "logps/rejected": -33.004886627197266,
-      "loss": 0.493,
-      "rewards/accuracies": 0.4583333432674408,
-      "rewards/chosen": 0.009805028326809406,
-      "rewards/margins": 0.0423436164855957,
-      "rewards/rejected": -0.03253858909010887,
+      "logits/chosen": 81.08451080322266,
+      "logits/rejected": 80.78488159179688,
+      "logps/chosen": -34.28126525878906,
+      "logps/rejected": -33.13351058959961,
+      "loss": 0.492,
+      "rewards/accuracies": 0.4861111044883728,
+      "rewards/chosen": -0.015815330669283867,
+      "rewards/margins": 0.05422591418027878,
+      "rewards/rejected": -0.0700412467122078,
       "step": 10
     },
     {
       "epoch": 0.05,
-      "grad_norm": 15.5,
+      "grad_norm": 9.6875,
       "learning_rate": 2.564102564102564e-06,
-      "logits/chosen": 80.6578369140625,
-      "logits/rejected": 80.54452514648438,
-      "logps/chosen": -33.567832946777344,
-      "logps/rejected": -30.769500732421875,
-      "loss": 0.4824,
-      "rewards/accuracies": 0.4749999940395355,
-      "rewards/chosen": 0.0854543000459671,
-      "rewards/margins": 0.07986308634281158,
-      "rewards/rejected": 0.005591208580881357,
+      "logits/chosen": 80.63957214355469,
+      "logits/rejected": 80.53150939941406,
+      "logps/chosen": -33.59695816040039,
+      "logps/rejected": -30.76776123046875,
+      "loss": 0.4944,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": 0.03718096390366554,
+      "rewards/margins": 0.03328876942396164,
+      "rewards/rejected": 0.003892195178195834,
       "step": 20
     },
     {
       "epoch": 0.08,
-      "grad_norm": 17.875,
+      "grad_norm": 10.5,
       "learning_rate": 3.846153846153847e-06,
-      "logits/chosen": 82.4737777709961,
-      "logits/rejected": 82.50477600097656,
-      "logps/chosen": -33.87335205078125,
-      "logps/rejected": -31.195053100585938,
-      "loss": 0.5111,
-      "rewards/accuracies": 0.4124999940395355,
-      "rewards/chosen": 0.09762730449438095,
-      "rewards/margins": -0.06251715123653412,
-      "rewards/rejected": 0.16014449298381805,
+      "logits/chosen": 82.47281646728516,
+      "logits/rejected": 82.50285339355469,
+      "logps/chosen": -33.86055374145508,
+      "logps/rejected": -31.097183227539062,
+      "loss": 0.5134,
+      "rewards/accuracies": 0.4000000059604645,
+      "rewards/chosen": 0.06090690940618515,
+      "rewards/margins": -0.0697535052895546,
+      "rewards/rejected": 0.13066044449806213,
       "step": 30
     },
     {
       "epoch": 0.1,
-      "grad_norm": 15.625,
+      "grad_norm": 10.875,
       "learning_rate": 4.999896948438434e-06,
-      "logits/chosen": 81.05291748046875,
-      "logits/rejected": 81.04745483398438,
-      "logps/chosen": -32.597801208496094,
-      "logps/rejected": -33.1552734375,
-      "loss": 0.4525,
-      "rewards/accuracies": 0.550000011920929,
-      "rewards/chosen": 0.40338650345802307,
-      "rewards/margins": 0.25411099195480347,
-      "rewards/rejected": 0.14927545189857483,
+      "logits/chosen": 81.05327606201172,
+      "logits/rejected": 81.04902648925781,
+      "logps/chosen": -32.787689208984375,
+      "logps/rejected": -33.05846405029297,
+      "loss": 0.4913,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": 0.15454962849617004,
+      "rewards/margins": 0.030526524409651756,
+      "rewards/rejected": 0.12402307987213135,
       "step": 40
     },
     {
       "epoch": 0.13,
-      "grad_norm": 14.5625,
+      "grad_norm": 8.125,
       "learning_rate": 4.987541037542187e-06,
-      "logits/chosen": 78.7928466796875,
-      "logits/rejected": 78.8023910522461,
-      "logps/chosen": -30.487594604492188,
-      "logps/rejected": -30.689586639404297,
-      "loss": 0.4635,
-      "rewards/accuracies": 0.6000000238418579,
-      "rewards/chosen": 0.5023558735847473,
-      "rewards/margins": 0.23376783728599548,
-      "rewards/rejected": 0.26858800649642944,
+      "logits/chosen": 78.79315185546875,
+      "logits/rejected": 78.80322265625,
+      "logps/chosen": -30.44384765625,
+      "logps/rejected": -30.80826759338379,
+      "loss": 0.4599,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": 0.30455905199050903,
+      "rewards/margins": 0.19855086505413055,
+      "rewards/rejected": 0.10600819438695908,
       "step": 50
     },
     {
       "epoch": 0.16,
-      "grad_norm": 13.375,
+      "grad_norm": 9.625,
       "learning_rate": 4.954691471941119e-06,
-      "logits/chosen": 83.43833923339844,
-      "logits/rejected": 83.5004653930664,
-      "logps/chosen": -30.851482391357422,
-      "logps/rejected": -29.280162811279297,
-      "loss": 0.501,
-      "rewards/accuracies": 0.42500001192092896,
-      "rewards/chosen": 0.22657600045204163,
-      "rewards/margins": 0.009325137361884117,
-      "rewards/rejected": 0.21725085377693176,
+      "logits/chosen": 83.44609069824219,
+      "logits/rejected": 83.50300598144531,
+      "logps/chosen": -30.81852149963379,
+      "logps/rejected": -29.231618881225586,
+      "loss": 0.4994,
+      "rewards/accuracies": 0.4124999940395355,
+      "rewards/chosen": 0.1426558792591095,
+      "rewards/margins": -0.0009051367524079978,
+      "rewards/rejected": 0.1435610055923462,
       "step": 60
     },
     {
       "epoch": 0.18,
-      "grad_norm": 13.5,
+      "grad_norm": 7.875,
       "learning_rate": 4.901618883413549e-06,
-      "logits/chosen": 84.01557159423828,
-      "logits/rejected": 84.04491424560547,
-      "logps/chosen": -30.311126708984375,
-      "logps/rejected": -32.73175811767578,
-      "loss": 0.4997,
-      "rewards/accuracies": 0.42500001192092896,
-      "rewards/chosen": 0.22253043949604034,
-      "rewards/margins": 0.008363036438822746,
-      "rewards/rejected": 0.21416740119457245,
+      "logits/chosen": 84.02252960205078,
+      "logits/rejected": 84.05018615722656,
+      "logps/chosen": -30.281408309936523,
+      "logps/rejected": -32.76702880859375,
+      "loss": 0.4981,
+      "rewards/accuracies": 0.4375,
+      "rewards/chosen": 0.13904651999473572,
+      "rewards/margins": 0.03077618218958378,
+      "rewards/rejected": 0.10827036201953888,
       "step": 70
     },
     {
       "epoch": 0.21,
-      "grad_norm": 16.375,
+      "grad_norm": 9.5,
       "learning_rate": 4.828760511501322e-06,
-      "logits/chosen": 81.79966735839844,
-      "logits/rejected": 81.77557373046875,
-      "logps/chosen": -31.009265899658203,
-      "logps/rejected": -30.690744400024414,
-      "loss": 0.4434,
-      "rewards/accuracies": 0.574999988079071,
-      "rewards/chosen": 0.34697332978248596,
-      "rewards/margins": 0.3001402020454407,
-      "rewards/rejected": 0.04683312773704529,
+      "logits/chosen": 81.84428405761719,
+      "logits/rejected": 81.82522583007812,
+      "logps/chosen": -31.032154083251953,
+      "logps/rejected": -30.712255477905273,
+      "loss": 0.4648,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": 0.18911631405353546,
+      "rewards/margins": 0.1709602177143097,
+      "rewards/rejected": 0.018156107515096664,
       "step": 80
     },
     {
       "epoch": 0.23,
-      "grad_norm": 20.75,
+      "grad_norm": 12.375,
       "learning_rate": 4.7367166013034295e-06,
-      "logits/chosen": 78.57239532470703,
-      "logits/rejected": 78.55184173583984,
-      "logps/chosen": -32.28063201904297,
-      "logps/rejected": -30.936914443969727,
-      "loss": 0.4612,
-      "rewards/accuracies": 0.5249999761581421,
-      "rewards/chosen": 0.2502840459346771,
-      "rewards/margins": 0.2400922328233719,
-      "rewards/rejected": 0.0101918401196599,
+      "logits/chosen": 78.68492126464844,
+      "logits/rejected": 78.66017150878906,
+      "logps/chosen": -32.12371063232422,
+      "logps/rejected": -30.837940216064453,
+      "loss": 0.4704,
+      "rewards/accuracies": 0.574999988079071,
+      "rewards/chosen": 0.20578794181346893,
+      "rewards/margins": 0.16037426888942719,
+      "rewards/rejected": 0.04541371017694473,
       "step": 90
     },
     {
       "epoch": 0.26,
-      "grad_norm": 15.5625,
+      "grad_norm": 9.9375,
       "learning_rate": 4.626245458345211e-06,
-      "logits/chosen": 83.69164276123047,
-      "logits/rejected": 83.69895935058594,
-      "logps/chosen": -33.99507141113281,
-      "logps/rejected": -31.751794815063477,
-      "loss": 0.461,
-      "rewards/accuracies": 0.512499988079071,
-      "rewards/chosen": 0.21694469451904297,
-      "rewards/margins": 0.21781782805919647,
-      "rewards/rejected": -0.0008731067064218223,
+      "logits/chosen": 83.77175903320312,
+      "logits/rejected": 83.79080963134766,
+      "logps/chosen": -33.829063415527344,
+      "logps/rejected": -31.544397354125977,
+      "loss": 0.4749,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": 0.1903691589832306,
+      "rewards/margins": 0.10790882259607315,
+      "rewards/rejected": 0.08246034383773804,
       "step": 100
     },
     {
       "epoch": 0.26,
-      "eval_logits/chosen": 98.70431518554688,
-      "eval_logits/rejected": 98.68829345703125,
-      "eval_logps/chosen": -32.346439361572266,
-      "eval_logps/rejected": -35.87228775024414,
-      "eval_loss": 0.5000348687171936,
-      "eval_rewards/accuracies": 0.5157807469367981,
-      "eval_rewards/chosen": 0.06771623343229294,
-      "eval_rewards/margins": 0.0017476709326729178,
-      "eval_rewards/rejected": 0.06596855819225311,
-      "eval_runtime": 104.2308,
-      "eval_samples_per_second": 3.291,
-      "eval_steps_per_second": 0.413,
+      "eval_logits/chosen": 98.75000762939453,
+      "eval_logits/rejected": 98.7376708984375,
+      "eval_logps/chosen": -32.41049575805664,
+      "eval_logps/rejected": -35.85907745361328,
+      "eval_loss": 0.504712700843811,
+      "eval_rewards/accuracies": 0.4746677577495575,
+      "eval_rewards/chosen": 0.013071590103209019,
+      "eval_rewards/margins": -0.02990747056901455,
+      "eval_rewards/rejected": 0.042979057878255844,
+      "eval_runtime": 104.3402,
+      "eval_samples_per_second": 3.287,
+      "eval_steps_per_second": 0.412,
       "step": 100
     },
     {
       "epoch": 0.29,
-      "grad_norm": 17.625,
+      "grad_norm": 11.875,
       "learning_rate": 4.498257201263691e-06,
-      "logits/chosen": 83.92507934570312,
-      "logits/rejected": 83.79533386230469,
-      "logps/chosen": -32.187767028808594,
-      "logps/rejected": -32.65043258666992,
-      "loss": 0.3988,
-      "rewards/accuracies": 0.637499988079071,
-      "rewards/chosen": 0.5845246911048889,
-      "rewards/margins": 0.5807043313980103,
-      "rewards/rejected": 0.0038203417789191008,
+      "logits/chosen": 84.02198028564453,
+      "logits/rejected": 83.90216064453125,
+      "logps/chosen": -32.067047119140625,
+      "logps/rejected": -32.60837936401367,
+      "loss": 0.4251,
+      "rewards/accuracies": 0.574999988079071,
+      "rewards/chosen": 0.382302463054657,
+      "rewards/margins": 0.36329856514930725,
+      "rewards/rejected": 0.01900387369096279,
       "step": 110
     },
     {
       "epoch": 0.31,
-      "grad_norm": 19.375,
+      "grad_norm": 10.9375,
       "learning_rate": 4.353806263777678e-06,
-      "logits/chosen": 83.96737670898438,
-      "logits/rejected": 84.08332061767578,
-      "logps/chosen": -28.09195327758789,
-      "logps/rejected": -35.17109680175781,
-      "loss": 0.4351,
-      "rewards/accuracies": 0.550000011920929,
-      "rewards/chosen": 0.5863426327705383,
-      "rewards/margins": 0.3847105801105499,
-      "rewards/rejected": 0.2016320526599884,
+      "logits/chosen": 84.06869506835938,
+      "logits/rejected": 84.17839050292969,
+      "logps/chosen": -28.029781341552734,
+      "logps/rejected": -35.15131378173828,
+      "loss": 0.4499,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": 0.35991987586021423,
+      "rewards/margins": 0.23678629100322723,
+      "rewards/rejected": 0.123133584856987,
       "step": 120
     },
     {
       "epoch": 0.34,
-      "grad_norm": 13.25,
+      "grad_norm": 9.125,
       "learning_rate": 4.1940827077152755e-06,
-      "logits/chosen": 81.23406219482422,
-      "logits/rejected": 81.2734603881836,
-      "logps/chosen": -30.209375381469727,
-      "logps/rejected": -31.70619773864746,
-      "loss": 0.4414,
-      "rewards/accuracies": 0.6000000238418579,
-      "rewards/chosen": 0.48858681321144104,
-      "rewards/margins": 0.330787718296051,
-      "rewards/rejected": 0.1577991098165512,
+      "logits/chosen": 81.26686096191406,
+      "logits/rejected": 81.29707336425781,
+      "logps/chosen": -30.0821533203125,
+      "logps/rejected": -31.722143173217773,
+      "loss": 0.4515,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.3300797641277313,
+      "rewards/margins": 0.24628591537475586,
+      "rewards/rejected": 0.08379384875297546,
       "step": 130
     },
     {
       "epoch": 0.36,
-      "grad_norm": 15.3125,
+      "grad_norm": 7.09375,
       "learning_rate": 4.0204024186666215e-06,
-      "logits/chosen": 82.31855773925781,
-      "logits/rejected": 82.33930206298828,
-      "logps/chosen": -26.636510848999023,
-      "logps/rejected": -32.51647186279297,
-      "loss": 0.3966,
-      "rewards/accuracies": 0.612500011920929,
-      "rewards/chosen": 0.5675702691078186,
-      "rewards/margins": 0.6327372789382935,
-      "rewards/rejected": -0.06516699492931366,
+      "logits/chosen": 82.26895904541016,
+      "logits/rejected": 82.28730773925781,
+      "logps/chosen": -26.675317764282227,
+      "logps/rejected": -32.535396575927734,
+      "loss": 0.4328,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": 0.30880242586135864,
+      "rewards/margins": 0.35360854864120483,
+      "rewards/rejected": -0.044806141406297684,
       "step": 140
     },
     {
       "epoch": 0.39,
-      "grad_norm": 14.5625,
+      "grad_norm": 8.75,
       "learning_rate": 3.834196265035119e-06,
-      "logits/chosen": 80.7647476196289,
-      "logits/rejected": 80.73397064208984,
-      "logps/chosen": -28.527297973632812,
-      "logps/rejected": -32.617584228515625,
-      "loss": 0.3943,
-      "rewards/accuracies": 0.7124999761581421,
-      "rewards/chosen": 0.603269100189209,
-      "rewards/margins": 0.6390293836593628,
-      "rewards/rejected": -0.035760145634412766,
+      "logits/chosen": 80.73310089111328,
+      "logits/rejected": 80.70652770996094,
+      "logps/chosen": -28.468103408813477,
+      "logps/rejected": -32.77440643310547,
+      "loss": 0.4125,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": 0.36840274930000305,
+      "rewards/margins": 0.45156532526016235,
+      "rewards/rejected": -0.08316260576248169,
       "step": 150
     },
     {
       "epoch": 0.42,
-      "grad_norm": 16.0,
+      "grad_norm": 8.5,
       "learning_rate": 3.636998309800573e-06,
-      "logits/chosen": 82.51421356201172,
-      "logits/rejected": 82.53550720214844,
-      "logps/chosen": -33.11312484741211,
-      "logps/rejected": -30.041656494140625,
-      "loss": 0.3649,
-      "rewards/accuracies": 0.675000011920929,
-      "rewards/chosen": 0.7165369987487793,
-      "rewards/margins": 0.7789196372032166,
-      "rewards/rejected": -0.06238250061869621,
+      "logits/chosen": 82.41865539550781,
+      "logits/rejected": 82.44725799560547,
+      "logps/chosen": -33.112274169921875,
+      "logps/rejected": -30.1284122467041,
+      "loss": 0.4055,
+      "rewards/accuracies": 0.7250000238418579,
+      "rewards/chosen": 0.40979090332984924,
+      "rewards/margins": 0.4801415503025055,
+      "rewards/rejected": -0.07035063952207565,
       "step": 160
     },
     {
       "epoch": 0.44,
-      "grad_norm": 11.75,
+      "grad_norm": 7.5625,
       "learning_rate": 3.4304331721118078e-06,
-      "logits/chosen": 83.22119903564453,
-      "logits/rejected": 83.18238067626953,
-      "logps/chosen": -30.43136978149414,
-      "logps/rejected": -32.384212493896484,
-      "loss": 0.3752,
+      "logits/chosen": 83.09947204589844,
+      "logits/rejected": 83.05732727050781,
+      "logps/chosen": -30.413660049438477,
+      "logps/rejected": -32.43395233154297,
+      "loss": 0.4144,
       "rewards/accuracies": 0.6499999761581421,
-      "rewards/chosen": 0.5795407295227051,
-      "rewards/margins": 0.8344497680664062,
-      "rewards/rejected": -0.25490903854370117,
+      "rewards/chosen": 0.33825185894966125,
+      "rewards/margins": 0.5038093328475952,
+      "rewards/rejected": -0.16555748879909515,
       "step": 170
     },
     {
       "epoch": 0.47,
-      "grad_norm": 7.625,
+      "grad_norm": 6.46875,
       "learning_rate": 3.2162026428305436e-06,
-      "logits/chosen": 80.77767181396484,
-      "logits/rejected": 80.74992370605469,
-      "logps/chosen": -30.34661293029785,
-      "logps/rejected": -31.299999237060547,
-      "loss": 0.3834,
-      "rewards/accuracies": 0.6625000238418579,
-      "rewards/chosen": 0.5953787565231323,
-      "rewards/margins": 0.6953333616256714,
-      "rewards/rejected": -0.09995462000370026,
+      "logits/chosen": 80.59300231933594,
+      "logits/rejected": 80.56774139404297,
+      "logps/chosen": -30.52907371520996,
+      "logps/rejected": -31.434350967407227,
+      "loss": 0.433,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.2672317624092102,
+      "rewards/margins": 0.3780880868434906,
+      "rewards/rejected": -0.1108563169836998,
       "step": 180
     },
     {
       "epoch": 0.49,
-      "grad_norm": 7.5,
+      "grad_norm": 5.40625,
       "learning_rate": 2.996071664294641e-06,
-      "logits/chosen": 82.40937805175781,
-      "logits/rejected": 82.38029479980469,
-      "logps/chosen": -29.8316593170166,
-      "logps/rejected": -30.288339614868164,
-      "loss": 0.4151,
+      "logits/chosen": 82.23339080810547,
+      "logits/rejected": 82.1915283203125,
+      "logps/chosen": -29.8317813873291,
+      "logps/rejected": -30.190067291259766,
+      "loss": 0.4509,
       "rewards/accuracies": 0.612500011920929,
-      "rewards/chosen": 0.6543310284614563,
-      "rewards/margins": 0.5285537838935852,
-      "rewards/rejected": 0.1257772147655487,
+      "rewards/chosen": 0.37385472655296326,
+      "rewards/margins": 0.26267164945602417,
+      "rewards/rejected": 0.1111830323934555,
       "step": 190
     },
     {
       "epoch": 0.52,
-      "grad_norm": 8.5625,
+      "grad_norm": 6.625,
       "learning_rate": 2.7718537898066833e-06,
-      "logits/chosen": 77.76971435546875,
-      "logits/rejected": 77.71305847167969,
-      "logps/chosen": -33.00738525390625,
-      "logps/rejected": -32.31935119628906,
-      "loss": 0.3172,
+      "logits/chosen": 77.5392074584961,
+      "logits/rejected": 77.48648834228516,
+      "logps/chosen": -33.0440788269043,
+      "logps/rejected": -32.247684478759766,
+      "loss": 0.3843,
       "rewards/accuracies": 0.699999988079071,
-      "rewards/chosen": 1.2217466831207275,
-      "rewards/margins": 1.0999095439910889,
-      "rewards/rejected": 0.12183725833892822,
+      "rewards/chosen": 0.6834636926651001,
+      "rewards/margins": 0.5851765871047974,
+      "rewards/rejected": 0.09828709810972214,
       "step": 200
     },
     {
       "epoch": 0.52,
-      "eval_logits/chosen": 98.44633483886719,
-      "eval_logits/rejected": 98.44091033935547,
-      "eval_logps/chosen": -32.319034576416016,
-      "eval_logps/rejected": -35.88397979736328,
-      "eval_loss": 0.49560824036598206,
-      "eval_rewards/accuracies": 0.5132890343666077,
-      "eval_rewards/chosen": 0.08689937740564346,
-      "eval_rewards/margins": 0.029115768149495125,
-      "eval_rewards/rejected": 0.057783618569374084,
-      "eval_runtime": 104.0206,
-      "eval_samples_per_second": 3.297,
+      "eval_logits/chosen": 98.38356018066406,
+      "eval_logits/rejected": 98.37836456298828,
+      "eval_logps/chosen": -32.3836784362793,
+      "eval_logps/rejected": -35.98166275024414,
+      "eval_loss": 0.4944417476654053,
+      "eval_rewards/accuracies": 0.5307309031486511,
+      "eval_rewards/chosen": 0.023798126727342606,
+      "eval_rewards/margins": 0.029852891340851784,
+      "eval_rewards/rejected": -0.00605476601049304,
+      "eval_runtime": 104.0926,
+      "eval_samples_per_second": 3.295,
       "eval_steps_per_second": 0.413,
       "step": 200
     },
     {
       "epoch": 0.55,
-      "grad_norm": 18.875,
+      "grad_norm": 12.3125,
       "learning_rate": 2.5453962426402006e-06,
-      "logits/chosen": 80.35867309570312,
-      "logits/rejected": 80.27238464355469,
-      "logps/chosen": -32.687442779541016,
-      "logps/rejected": -34.85881423950195,
-      "loss": 0.3718,
-      "rewards/accuracies": 0.6625000238418579,
-      "rewards/chosen": 0.8807934522628784,
-      "rewards/margins": 0.7916989922523499,
-      "rewards/rejected": 0.08909468352794647,
+      "logits/chosen": 80.09881591796875,
+      "logits/rejected": 80.01982879638672,
+      "logps/chosen": -32.544151306152344,
+      "logps/rejected": -35.001224517822266,
+      "loss": 0.3882,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": 0.5606266260147095,
+      "rewards/margins": 0.5666781663894653,
+      "rewards/rejected": -0.006051521748304367,
       "step": 210
     },
     {
       "epoch": 0.57,
-      "grad_norm": 15.625,
+      "grad_norm": 8.1875,
       "learning_rate": 2.3185646976551794e-06,
-      "logits/chosen": 82.43341064453125,
-      "logits/rejected": 82.50484466552734,
-      "logps/chosen": -30.421199798583984,
-      "logps/rejected": -30.724105834960938,
-      "loss": 0.3248,
-      "rewards/accuracies": 0.7250000238418579,
-      "rewards/chosen": 1.0694276094436646,
-      "rewards/margins": 1.0660343170166016,
-      "rewards/rejected": 0.0033930898644030094,
+      "logits/chosen": 82.21588134765625,
+      "logits/rejected": 82.28758239746094,
+      "logps/chosen": -30.38724136352539,
+      "logps/rejected": -30.862102508544922,
+      "loss": 0.3707,
+      "rewards/accuracies": 0.762499988079071,
+      "rewards/chosen": 0.6246866583824158,
+      "rewards/margins": 0.6779459714889526,
+      "rewards/rejected": -0.05325937271118164,
       "step": 220
     },
     {
       "epoch": 0.6,
-      "grad_norm": 11.6875,
+      "grad_norm": 8.5,
       "learning_rate": 2.0932279108998323e-06,
-      "logits/chosen": 79.52705383300781,
-      "logits/rejected": 79.57157897949219,
-      "logps/chosen": -31.819992065429688,
-      "logps/rejected": -33.83860397338867,
-      "loss": 0.4096,
-      "rewards/accuracies": 0.5874999761581421,
-      "rewards/chosen": 0.6900272369384766,
-      "rewards/margins": 0.5685457587242126,
-      "rewards/rejected": 0.1214815154671669,
+      "logits/chosen": 79.33607482910156,
+      "logits/rejected": 79.39179992675781,
+      "logps/chosen": -31.715646743774414,
+      "logps/rejected": -33.92882537841797,
+      "loss": 0.4251,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.4360387325286865,
+      "rewards/margins": 0.40270787477493286,
+      "rewards/rejected": 0.03333085775375366,
       "step": 230
     },
     {
       "epoch": 0.62,
-      "grad_norm": 17.25,
+      "grad_norm": 10.625,
       "learning_rate": 1.8712423238279358e-06,
-      "logits/chosen": 81.8853988647461,
-      "logits/rejected": 82.16355895996094,
-      "logps/chosen": -30.063451766967773,
-      "logps/rejected": -31.44558334350586,
-      "loss": 0.3113,
-      "rewards/accuracies": 0.762499988079071,
-      "rewards/chosen": 1.137229561805725,
-      "rewards/margins": 1.0767086744308472,
-      "rewards/rejected": 0.06052086502313614,
+      "logits/chosen": 81.82881164550781,
+      "logits/rejected": 82.10590362548828,
+      "logps/chosen": -29.962848663330078,
+      "logps/rejected": -31.462871551513672,
+      "loss": 0.3667,
+      "rewards/accuracies": 0.7875000238418579,
+      "rewards/chosen": 0.6900869607925415,
+      "rewards/margins": 0.6624192595481873,
+      "rewards/rejected": 0.027667587623000145,
       "step": 240
     },
     {
       "epoch": 0.65,
-      "grad_norm": 17.0,
+      "grad_norm": 11.25,
       "learning_rate": 1.6544367689701824e-06,
-      "logits/chosen": 80.61282348632812,
-      "logits/rejected": 80.67658996582031,
-      "logps/chosen": -26.506671905517578,
-      "logps/rejected": -29.721996307373047,
-      "loss": 0.4011,
-      "rewards/accuracies": 0.6000000238418579,
-      "rewards/chosen": 0.7624577283859253,
-      "rewards/margins": 0.6373406648635864,
-      "rewards/rejected": 0.12511704862117767,
+      "logits/chosen": 80.50505065917969,
+      "logits/rejected": 80.56990051269531,
+      "logps/chosen": -26.388418197631836,
+      "logps/rejected": -29.717565536499023,
+      "loss": 0.4245,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.48299089074134827,
+      "rewards/margins": 0.40972191095352173,
+      "rewards/rejected": 0.07326899468898773,
       "step": 250
     },
     {
       "epoch": 0.68,
-      "grad_norm": 12.25,
+      "grad_norm": 8.625,
       "learning_rate": 1.4445974030621963e-06,
-      "logits/chosen": 77.79762268066406,
-      "logits/rejected": 77.97430419921875,
-      "logps/chosen": -29.87813949584961,
-      "logps/rejected": -35.95857238769531,
-      "loss": 0.3311,
-      "rewards/accuracies": 0.675000011920929,
-      "rewards/chosen": 1.1779998540878296,
-      "rewards/margins": 1.0778592824935913,
-      "rewards/rejected": 0.10014041513204575,
+      "logits/chosen": 77.76419830322266,
+      "logits/rejected": 77.94038391113281,
+      "logps/chosen": -29.758447647094727,
+      "logps/rejected": -36.150657653808594,
+      "loss": 0.3604,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": 0.7210196256637573,
+      "rewards/margins": 0.7406320571899414,
+      "rewards/rejected": -0.0196123905479908,
       "step": 260
     },
     {
       "epoch": 0.7,
-      "grad_norm": 8.8125,
+      "grad_norm": 6.1875,
       "learning_rate": 1.243452991757889e-06,
-      "logits/chosen": 77.17610168457031,
-      "logits/rejected": 77.21318817138672,
-      "logps/chosen": -30.205867767333984,
-      "logps/rejected": -31.4431095123291,
-      "loss": 0.3276,
-      "rewards/accuracies": 0.737500011920929,
-      "rewards/chosen": 1.1011395454406738,
-      "rewards/margins": 1.0575218200683594,
-      "rewards/rejected": 0.04361782968044281,
+      "logits/chosen": 77.1490707397461,
+      "logits/rejected": 77.18915557861328,
+      "logps/chosen": -30.186153411865234,
+      "logps/rejected": -31.373126983642578,
+      "loss": 0.3871,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": 0.6371081471443176,
+      "rewards/margins": 0.5841903686523438,
+      "rewards/rejected": 0.05291769653558731,
       "step": 270
     },
     {
       "epoch": 0.73,
-      "grad_norm": 17.375,
+      "grad_norm": 12.5,
       "learning_rate": 1.0526606671603523e-06,
-      "logits/chosen": 79.84697723388672,
-      "logits/rejected": 79.62041473388672,
-      "logps/chosen": -30.549224853515625,
-      "logps/rejected": -29.2475643157959,
-      "loss": 0.3847,
-      "rewards/accuracies": 0.699999988079071,
-      "rewards/chosen": 0.8824693560600281,
-      "rewards/margins": 0.588868260383606,
-      "rewards/rejected": 0.2936010956764221,
+      "logits/chosen": 79.78591918945312,
+      "logits/rejected": 79.5605239868164,
+      "logps/chosen": -30.330188751220703,
+      "logps/rejected": -29.223669052124023,
+      "loss": 0.4182,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": 0.5918818712234497,
+      "rewards/margins": 0.41455134749412537,
+      "rewards/rejected": 0.17733044922351837,
       "step": 280
     },
     {
       "epoch": 0.75,
-      "grad_norm": 13.125,
+      "grad_norm": 8.4375,
       "learning_rate": 8.737922755071455e-07,
-      "logits/chosen": 80.05425262451172,
-      "logits/rejected": 79.96458435058594,
-      "logps/chosen": -32.43302536010742,
-      "logps/rejected": -32.11684036254883,
-      "loss": 0.3019,
-      "rewards/accuracies": 0.7875000238418579,
-      "rewards/chosen": 1.169856309890747,
-      "rewards/margins": 1.2398390769958496,
-      "rewards/rejected": -0.06998284161090851,
+      "logits/chosen": 80.03771209716797,
+      "logits/rejected": 79.94876861572266,
+      "logps/chosen": -32.170814514160156,
+      "logps/rejected": -32.12440872192383,
+      "loss": 0.3449,
+      "rewards/accuracies": 0.824999988079071,
+      "rewards/chosen": 0.7733729481697083,
+      "rewards/margins": 0.8163889646530151,
+      "rewards/rejected": -0.04301605746150017,
       "step": 290
     },
     {
       "epoch": 0.78,
-      "grad_norm": 9.5625,
+      "grad_norm": 5.5,
       "learning_rate": 7.08321427484816e-07,
-      "logits/chosen": 75.59251403808594,
-      "logits/rejected": 75.66484069824219,
-      "logps/chosen": -31.76091957092285,
-      "logps/rejected": -28.981002807617188,
-      "loss": 0.3303,
-      "rewards/accuracies": 0.7124999761581421,
-      "rewards/chosen": 1.151352882385254,
-      "rewards/margins": 1.0666378736495972,
-      "rewards/rejected": 0.08471502363681793,
+      "logits/chosen": 75.5302505493164,
+      "logits/rejected": 75.59835052490234,
+      "logps/chosen": -31.593181610107422,
+      "logps/rejected": -29.076126098632812,
+      "loss": 0.363,
+      "rewards/accuracies": 0.7749999761581421,
+      "rewards/chosen": 0.7250088453292847,
+      "rewards/margins": 0.714650571346283,
+      "rewards/rejected": 0.010358264669775963,
       "step": 300
     },
     {
       "epoch": 0.78,
-      "eval_logits/chosen": 98.37130737304688,
-      "eval_logits/rejected": 98.36297607421875,
-      "eval_logps/chosen": -32.2767219543457,
-      "eval_logps/rejected": -35.79054260253906,
-      "eval_loss": 0.49796876311302185,
-      "eval_rewards/accuracies": 0.5016611218452454,
-      "eval_rewards/chosen": 0.11651887744665146,
-      "eval_rewards/margins": -0.0066675543785095215,
-      "eval_rewards/rejected": 0.12318644672632217,
-      "eval_runtime": 104.1032,
-      "eval_samples_per_second": 3.295,
+      "eval_logits/chosen": 98.36553192138672,
+      "eval_logits/rejected": 98.37035369873047,
+      "eval_logps/chosen": -32.229251861572266,
+      "eval_logps/rejected": -35.81660079956055,
+      "eval_loss": 0.49616533517837524,
+      "eval_rewards/accuracies": 0.5103820562362671,
+      "eval_rewards/chosen": 0.0855708047747612,
+      "eval_rewards/margins": 0.025601176545023918,
+      "eval_rewards/rejected": 0.05996962636709213,
+      "eval_runtime": 104.1169,
+      "eval_samples_per_second": 3.294,
       "eval_steps_per_second": 0.413,
       "step": 300
     },
     {
       "epoch": 0.81,
-      "grad_norm": 10.625,
+      "grad_norm": 8.0625,
       "learning_rate": 5.576113578589035e-07,
-      "logits/chosen": 82.78644561767578,
-      "logits/rejected": 82.81319427490234,
-      "logps/chosen": -29.468921661376953,
-      "logps/rejected": -32.31016159057617,
-      "loss": 0.3414,
-      "rewards/accuracies": 0.7124999761581421,
-      "rewards/chosen": 0.9762304425239563,
-      "rewards/margins": 1.0535131692886353,
-      "rewards/rejected": -0.07728256285190582,
+      "logits/chosen": 82.7135238647461,
+      "logits/rejected": 82.74400329589844,
+      "logps/chosen": -29.19759178161621,
+      "logps/rejected": -32.07235336303711,
+      "loss": 0.3811,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": 0.666375994682312,
+      "rewards/margins": 0.6154125332832336,
+      "rewards/rejected": 0.05096355825662613,
       "step": 310
     },
     {
       "epoch": 0.83,
-      "grad_norm": 13.0625,
+      "grad_norm": 8.5,
       "learning_rate": 4.229036944380913e-07,
-      "logits/chosen": 80.16923522949219,
-      "logits/rejected": 80.17256164550781,
-      "logps/chosen": -29.88885498046875,
-      "logps/rejected": -28.86279296875,
-      "loss": 0.3116,
-      "rewards/accuracies": 0.699999988079071,
-      "rewards/chosen": 1.1967929601669312,
-      "rewards/margins": 1.157100796699524,
-      "rewards/rejected": 0.03969240561127663,
+      "logits/chosen": 80.10868835449219,
+      "logits/rejected": 80.10973358154297,
+      "logps/chosen": -29.693145751953125,
+      "logps/rejected": -28.737625122070312,
+      "loss": 0.3658,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": 0.7621666193008423,
+      "rewards/margins": 0.6894195675849915,
+      "rewards/rejected": 0.07274699211120605,
       "step": 320
     },
     {
       "epoch": 0.86,
-      "grad_norm": 11.6875,
+      "grad_norm": 7.8125,
       "learning_rate": 3.053082288996112e-07,
-      "logits/chosen": 77.39164733886719,
-      "logits/rejected": 77.41889953613281,
-      "logps/chosen": -28.371896743774414,
-      "logps/rejected": -32.67755889892578,
-      "loss": 0.2678,
-      "rewards/accuracies": 0.75,
-      "rewards/chosen": 1.3864963054656982,
-      "rewards/margins": 1.4163827896118164,
-      "rewards/rejected": -0.02988656982779503,
+      "logits/chosen": 77.36323547363281,
+      "logits/rejected": 77.391357421875,
+      "logps/chosen": -28.293853759765625,
+      "logps/rejected": -32.6038703918457,
+      "loss": 0.3393,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": 0.8235009908676147,
+      "rewards/margins": 0.8111019134521484,
+      "rewards/rejected": 0.012399068102240562,
       "step": 330
     },
     {
       "epoch": 0.88,
-      "grad_norm": 15.3125,
+      "grad_norm": 10.625,
       "learning_rate": 2.0579377374915805e-07,
-      "logits/chosen": 81.65691375732422,
-      "logits/rejected": 81.6773681640625,
-      "logps/chosen": -31.6644229888916,
-      "logps/rejected": -33.2541389465332,
-      "loss": 0.3491,
-      "rewards/accuracies": 0.737500011920929,
-      "rewards/chosen": 1.0834088325500488,
-      "rewards/margins": 1.003218412399292,
-      "rewards/rejected": 0.08019042015075684,
+      "logits/chosen": 81.62095642089844,
+      "logits/rejected": 81.64241027832031,
+      "logps/chosen": -31.521636962890625,
+      "logps/rejected": -33.3758659362793,
+      "loss": 0.377,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.6762043833732605,
+      "rewards/margins": 0.6790729761123657,
+      "rewards/rejected": -0.002868417650461197,
       "step": 340
     },
     {
       "epoch": 0.91,
-      "grad_norm": 9.5625,
+      "grad_norm": 7.71875,
       "learning_rate": 1.2518018074041684e-07,
-      "logits/chosen": 80.73246765136719,
-      "logits/rejected": 80.75402069091797,
-      "logps/chosen": -31.80560874938965,
-      "logps/rejected": -32.868934631347656,
-      "loss": 0.3373,
-      "rewards/accuracies": 0.7250000238418579,
-      "rewards/chosen": 1.2866908311843872,
-      "rewards/margins": 1.1306837797164917,
-      "rewards/rejected": 0.15600721538066864,
+      "logits/chosen": 80.68013763427734,
+      "logits/rejected": 80.69649505615234,
+      "logps/chosen": -31.641170501708984,
+      "logps/rejected": -32.8623161315918,
+      "loss": 0.3712,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": 0.8010267019271851,
+      "rewards/margins": 0.7092341780662537,
+      "rewards/rejected": 0.0917925089597702,
       "step": 350
     },
     {
       "epoch": 0.94,
-      "grad_norm": 10.5,
+      "grad_norm": 6.25,
       "learning_rate": 6.41315865106129e-08,
-      "logits/chosen": 82.23421478271484,
-      "logits/rejected": 82.29096221923828,
-      "logps/chosen": -27.619741439819336,
-      "logps/rejected": -31.5566349029541,
-      "loss": 0.2941,
-      "rewards/accuracies": 0.7749999761581421,
-      "rewards/chosen": 1.3527272939682007,
-      "rewards/margins": 1.2874782085418701,
-      "rewards/rejected": 0.06524892896413803,
+      "logits/chosen": 82.13801574707031,
+      "logits/rejected": 82.19556427001953,
+      "logps/chosen": -27.77614402770996,
+      "logps/rejected": -31.526113510131836,
+      "loss": 0.369,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": 0.7104269862174988,
+      "rewards/margins": 0.6609331965446472,
+      "rewards/rejected": 0.04949387162923813,
       "step": 360
     },
     {
       "epoch": 0.96,
-      "grad_norm": 11.5,
+      "grad_norm": 7.5,
       "learning_rate": 2.3150941078050325e-08,
-      "logits/chosen": 81.71080017089844,
-      "logits/rejected": 81.74567413330078,
-      "logps/chosen": -31.354503631591797,
-      "logps/rejected": -34.771522521972656,
-      "loss": 0.3671,
-      "rewards/accuracies": 0.612500011920929,
-      "rewards/chosen": 1.0651617050170898,
-      "rewards/margins": 0.8286684155464172,
-      "rewards/rejected": 0.23649325966835022,
+      "logits/chosen": 81.63580322265625,
+      "logits/rejected": 81.66941833496094,
+      "logps/chosen": -31.2968692779541,
+      "logps/rejected": -34.87516403198242,
+      "loss": 0.4021,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.6317180395126343,
+      "rewards/margins": 0.5380347967147827,
+      "rewards/rejected": 0.09368324279785156,
       "step": 370
     },
     {
       "epoch": 0.99,
-      "grad_norm": 13.8125,
+      "grad_norm": 8.625,
       "learning_rate": 2.575864278703266e-09,
-      "logits/chosen": 75.57669067382812,
-      "logits/rejected": 75.45292663574219,
-      "logps/chosen": -29.130199432373047,
-      "logps/rejected": -27.9501953125,
-      "loss": 0.3558,
-      "rewards/accuracies": 0.7250000238418579,
-      "rewards/chosen": 1.0121484994888306,
-      "rewards/margins": 0.8648824691772461,
-      "rewards/rejected": 0.14726600050926208,
+      "logits/chosen": 75.4756088256836,
+      "logits/rejected": 75.35409545898438,
+      "logps/chosen": -29.19217300415039,
+      "logps/rejected": -27.851299285888672,
+      "loss": 0.4126,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": 0.5535811185836792,
+      "rewards/margins": 0.4298717975616455,
+      "rewards/rejected": 0.1237092986702919,
       "step": 380
     },
     {
       "epoch": 1.0,
       "step": 385,
       "total_flos": 0.0,
-      "train_loss": 0.38857739250381274,
-      "train_runtime": 2559.074,
+      "train_loss": 0.4199758805237807,
+      "train_runtime": 2559.5376,
       "train_samples_per_second": 1.203,
       "train_steps_per_second": 0.15
     }