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: 0.9545
-- Rewards/chosen: -0.0743
-- Rewards/rejected: -0.1195
-- Rewards/accuracies: 0.5889
-- Rewards/margins: 0.0453
-- Logps/rejected: -37.1618
-- Logps/chosen: -33.1857
-- Logits/rejected: 97.6826
-- Logits/chosen: 97.7083
+- Loss: 0.9869
+- Rewards/chosen: -0.0473
+- Rewards/rejected: -0.1807
+- Rewards/accuracies: 0.5831
+- Rewards/margins: 0.1334
+- Logps/rejected: -36.1923
+- Logps/chosen: -32.5023
+- Logits/rejected: 98.8800
+- Logits/chosen: 98.9053
 
 ## 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 |
 |:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
-| 0.9577        | 0.26  | 100  | 0.9935          | 0.0116         | 0.0052           | 0.5066             | 0.0064          | -35.9146       | -32.3275     | 98.7059         | 98.7128       |
-| 0.8562        | 0.52  | 200  | 0.9647          | -0.0504        | -0.0855          | 0.5677             | 0.0351          | -36.8217       | -32.9469     | 97.9546         | 97.9719       |
-| 0.8271        | 0.78  | 300  | 0.9534          | -0.0764        | -0.1230          | 0.5714             | 0.0466          | -37.1962       | -33.2069     | 97.6898         | 97.7152       |
+| 0.9567        | 0.26  | 100  | 1.1574          | -0.0132        | 0.0638           | 0.4630             | -0.0769         | -35.8868       | -32.4596     | 98.7233         | 98.7349       |
+| 0.8098        | 0.52  | 200  | 1.0545          | -0.0943        | -0.1571          | 0.5278             | 0.0627          | -36.1629       | -32.5611     | 98.8438         | 98.8682       |
+| 0.6965        | 0.78  | 300  | 0.9869          | -0.0473        | -0.1807          | 0.5831             | 0.1334          | -36.1923       | -32.5023     | 98.8800         | 98.9053       |
 
 
 ### Framework versions

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9163b39dc6f617d133d06a7277c2cc09053d5524a776bdf2e8ee51d59383a53b
+oid sha256:4abece2ca5197d0f50cab042af90e7a41850f358f7df73bc0db074d29a562d23
 size 671150064

all_results.json

@@ -1,21 +1,8 @@
 {
     "epoch": 1.0,
-    "eval_logits/chosen": 97.70832824707031,
-    "eval_logits/rejected": 97.6825942993164,
-    "eval_logps/chosen": -33.185726165771484,
-    "eval_logps/rejected": -37.16183853149414,
-    "eval_loss": 0.9545474052429199,
-    "eval_rewards/accuracies": 0.5888704061508179,
-    "eval_rewards/chosen": -0.07425507158041,
-    "eval_rewards/margins": 0.045276571065187454,
-    "eval_rewards/rejected": -0.11953166127204895,
-    "eval_runtime": 103.7845,
-    "eval_samples": 343,
-    "eval_samples_per_second": 3.305,
-    "eval_steps_per_second": 0.414,
-    "train_loss": 0.88636748450143,
-    "train_runtime": 2554.6221,
+    "train_loss": 0.7616742146479619,
+    "train_runtime": 2551.1965,
     "train_samples": 3079,
-    "train_samples_per_second": 1.205,
+    "train_samples_per_second": 1.207,
     "train_steps_per_second": 0.151
 }

train_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 1.0,
-    "train_loss": 0.88636748450143,
-    "train_runtime": 2554.6221,
+    "train_loss": 0.7616742146479619,
+    "train_runtime": 2551.1965,
     "train_samples": 3079,
-    "train_samples_per_second": 1.205,
+    "train_samples_per_second": 1.207,
     "train_steps_per_second": 0.151
 }

trainer_state.json

@@ -10,7 +10,7 @@
   "log_history": [
     {
       "epoch": 0.0,
-      "grad_norm": 13.0625,
+      "grad_norm": 104.5,
       "learning_rate": 1.282051282051282e-07,
       "logits/chosen": 88.18099975585938,
       "logits/rejected": 88.25153350830078,
@@ -25,629 +25,629 @@
     },
     {
       "epoch": 0.03,
-      "grad_norm": 10.875,
+      "grad_norm": 78.0,
       "learning_rate": 1.282051282051282e-06,
-      "logits/chosen": 81.074951171875,
-      "logits/rejected": 80.7809829711914,
-      "logps/chosen": -34.20733642578125,
-      "logps/rejected": -32.97297668457031,
-      "loss": 0.9951,
-      "rewards/accuracies": 0.4583333432674408,
-      "rewards/chosen": 0.003438829444348812,
-      "rewards/margins": 0.004895869642496109,
-      "rewards/rejected": -0.0014570390339940786,
+      "logits/chosen": 81.08700561523438,
+      "logits/rejected": 80.79035186767578,
+      "logps/chosen": -34.29237365722656,
+      "logps/rejected": -33.04549026489258,
+      "loss": 0.9866,
+      "rewards/accuracies": 0.3888888955116272,
+      "rewards/chosen": -0.040518708527088165,
+      "rewards/margins": 0.029149238020181656,
+      "rewards/rejected": -0.06966794282197952,
       "step": 10
     },
     {
       "epoch": 0.05,
-      "grad_norm": 12.6875,
+      "grad_norm": 85.5,
       "learning_rate": 2.564102564102564e-06,
-      "logits/chosen": 80.64326477050781,
-      "logits/rejected": 80.53416442871094,
-      "logps/chosen": -33.720584869384766,
-      "logps/rejected": -30.82167625427246,
-      "loss": 0.9986,
-      "rewards/accuracies": 0.44999998807907104,
-      "rewards/chosen": -0.003067571669816971,
-      "rewards/margins": 0.0013512909645214677,
-      "rewards/rejected": -0.004418861120939255,
+      "logits/chosen": 80.68280029296875,
+      "logits/rejected": 80.57057189941406,
+      "logps/chosen": -33.53000259399414,
+      "logps/rejected": -30.8519287109375,
+      "loss": 0.8984,
+      "rewards/accuracies": 0.5375000238418579,
+      "rewards/chosen": 0.12792269885540009,
+      "rewards/margins": 0.18747423589229584,
+      "rewards/rejected": -0.05955154448747635,
       "step": 20
     },
     {
       "epoch": 0.08,
-      "grad_norm": 12.0,
+      "grad_norm": 73.0,
       "learning_rate": 3.846153846153847e-06,
-      "logits/chosen": 82.51445007324219,
-      "logits/rejected": 82.54810333251953,
-      "logps/chosen": -33.81728744506836,
-      "logps/rejected": -31.204355239868164,
-      "loss": 1.0024,
-      "rewards/accuracies": 0.44999998807907104,
-      "rewards/chosen": 0.019553204998373985,
-      "rewards/margins": -0.002394508570432663,
-      "rewards/rejected": 0.0219477117061615,
+      "logits/chosen": 82.52570343017578,
+      "logits/rejected": 82.552734375,
+      "logps/chosen": -33.96383285522461,
+      "logps/rejected": -31.202930450439453,
+      "loss": 1.1955,
+      "rewards/accuracies": 0.4000000059604645,
+      "rewards/chosen": 0.03918968886137009,
+      "rewards/margins": -0.1375342458486557,
+      "rewards/rejected": 0.1767239272594452,
       "step": 30
     },
     {
       "epoch": 0.1,
-      "grad_norm": 11.8125,
+      "grad_norm": 74.5,
       "learning_rate": 4.999896948438434e-06,
-      "logits/chosen": 81.101806640625,
-      "logits/rejected": 81.09938049316406,
-      "logps/chosen": -32.73223876953125,
-      "logps/rejected": -33.143699645996094,
-      "loss": 0.9783,
-      "rewards/accuracies": 0.5249999761581421,
-      "rewards/chosen": 0.04418279603123665,
-      "rewards/margins": 0.02170029655098915,
-      "rewards/rejected": 0.022482499480247498,
+      "logits/chosen": 80.97098541259766,
+      "logits/rejected": 80.9694595336914,
+      "logps/chosen": -32.8037109375,
+      "logps/rejected": -33.232460021972656,
+      "loss": 0.9309,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.2962803840637207,
+      "rewards/margins": 0.1874258816242218,
+      "rewards/rejected": 0.10885453224182129,
       "step": 40
     },
     {
       "epoch": 0.13,
-      "grad_norm": 11.9375,
+      "grad_norm": 70.0,
       "learning_rate": 4.987541037542187e-06,
-      "logits/chosen": 78.82136535644531,
-      "logits/rejected": 78.82911682128906,
-      "logps/chosen": -30.3783016204834,
-      "logps/rejected": -30.641677856445312,
-      "loss": 0.9605,
-      "rewards/accuracies": 0.5249999761581421,
-      "rewards/chosen": 0.082694411277771,
-      "rewards/margins": 0.039533428847789764,
-      "rewards/rejected": 0.04316098242998123,
+      "logits/chosen": 78.56432342529297,
+      "logits/rejected": 78.58203125,
+      "logps/chosen": -30.68828773498535,
+      "logps/rejected": -30.789108276367188,
+      "loss": 0.9855,
+      "rewards/accuracies": 0.550000011920929,
+      "rewards/chosen": 0.4135704040527344,
+      "rewards/margins": 0.18622872233390808,
+      "rewards/rejected": 0.2273416966199875,
       "step": 50
     },
     {
       "epoch": 0.16,
-      "grad_norm": 9.75,
+      "grad_norm": 84.0,
       "learning_rate": 4.954691471941119e-06,
-      "logits/chosen": 83.49494171142578,
-      "logits/rejected": 83.55232238769531,
-      "logps/chosen": -30.781469345092773,
-      "logps/rejected": -29.190662384033203,
-      "loss": 1.0006,
-      "rewards/accuracies": 0.44999998807907104,
-      "rewards/chosen": 0.03936903923749924,
-      "rewards/margins": -0.0006168211111798882,
-      "rewards/rejected": 0.039985861629247665,
+      "logits/chosen": 83.11712646484375,
+      "logits/rejected": 83.17086791992188,
+      "logps/chosen": -30.859710693359375,
+      "logps/rejected": -29.577678680419922,
+      "loss": 0.9236,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.25236082077026367,
+      "rewards/margins": 0.24208447337150574,
+      "rewards/rejected": 0.010276327840983868,
       "step": 60
     },
     {
       "epoch": 0.18,
-      "grad_norm": 12.0625,
+      "grad_norm": 107.0,
       "learning_rate": 4.901618883413549e-06,
-      "logits/chosen": 84.1231918334961,
-      "logits/rejected": 84.15650939941406,
-      "logps/chosen": -30.209863662719727,
-      "logps/rejected": -32.619781494140625,
-      "loss": 0.9999,
+      "logits/chosen": 83.80155944824219,
+      "logits/rejected": 83.8315658569336,
+      "logps/chosen": -30.606204986572266,
+      "logps/rejected": -32.91896057128906,
+      "loss": 1.141,
       "rewards/accuracies": 0.48750001192092896,
-      "rewards/chosen": 0.041915904730558395,
-      "rewards/margins": 0.000123101839562878,
-      "rewards/rejected": 0.04179280251264572,
+      "rewards/chosen": 0.01825830340385437,
+      "rewards/margins": -0.0767388790845871,
+      "rewards/rejected": 0.09499720484018326,
       "step": 70
     },
     {
       "epoch": 0.21,
-      "grad_norm": 12.375,
+      "grad_norm": 84.5,
       "learning_rate": 4.828760511501322e-06,
-      "logits/chosen": 81.8515853881836,
-      "logits/rejected": 81.83155822753906,
-      "logps/chosen": -30.982410430908203,
-      "logps/rejected": -30.617040634155273,
-      "loss": 0.9618,
-      "rewards/accuracies": 0.5249999761581421,
-      "rewards/chosen": 0.05225307494401932,
-      "rewards/margins": 0.03819245845079422,
-      "rewards/rejected": 0.014060619287192822,
+      "logits/chosen": 81.45521545410156,
+      "logits/rejected": 81.44440460205078,
+      "logps/chosen": -31.48111343383789,
+      "logps/rejected": -31.072092056274414,
+      "loss": 0.8995,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": 0.01906520128250122,
+      "rewards/margins": 0.2706204056739807,
+      "rewards/rejected": -0.2515551745891571,
       "step": 80
     },
     {
       "epoch": 0.23,
-      "grad_norm": 14.75,
+      "grad_norm": 102.0,
       "learning_rate": 4.7367166013034295e-06,
-      "logits/chosen": 78.68502807617188,
-      "logits/rejected": 78.65934753417969,
-      "logps/chosen": -32.18014144897461,
-      "logps/rejected": -30.878421783447266,
-      "loss": 0.9627,
-      "rewards/accuracies": 0.5249999761581421,
-      "rewards/chosen": 0.04580371826887131,
-      "rewards/margins": 0.03849860280752182,
-      "rewards/rejected": 0.007305114530026913,
+      "logits/chosen": 78.27713775634766,
+      "logits/rejected": 78.24540710449219,
+      "logps/chosen": -32.49984359741211,
+      "logps/rejected": -31.142908096313477,
+      "loss": 0.9467,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": 0.11066800355911255,
+      "rewards/margins": 0.2638159692287445,
+      "rewards/rejected": -0.15314793586730957,
       "step": 90
     },
     {
       "epoch": 0.26,
-      "grad_norm": 12.375,
+      "grad_norm": 73.5,
       "learning_rate": 4.626245458345211e-06,
-      "logits/chosen": 83.6978530883789,
-      "logits/rejected": 83.72080993652344,
-      "logps/chosen": -33.769813537597656,
-      "logps/rejected": -31.638240814208984,
-      "loss": 0.9577,
+      "logits/chosen": 83.56034851074219,
+      "logits/rejected": 83.58055877685547,
+      "logps/chosen": -34.057029724121094,
+      "logps/rejected": -31.78998374938965,
+      "loss": 0.9567,
       "rewards/accuracies": 0.574999988079071,
-      "rewards/chosen": 0.05351760983467102,
-      "rewards/margins": 0.04228735715150833,
-      "rewards/rejected": 0.01123025082051754,
+      "rewards/chosen": 0.19836829602718353,
+      "rewards/margins": 0.22992074489593506,
+      "rewards/rejected": -0.031552452594041824,
       "step": 100
     },
     {
       "epoch": 0.26,
-      "eval_logits/chosen": 98.71279907226562,
-      "eval_logits/rejected": 98.70589447021484,
-      "eval_logps/chosen": -32.32750701904297,
-      "eval_logps/rejected": -35.914642333984375,
-      "eval_loss": 0.993482768535614,
-      "eval_rewards/accuracies": 0.5066444873809814,
-      "eval_rewards/chosen": 0.011567190289497375,
-      "eval_rewards/margins": 0.006378817837685347,
-      "eval_rewards/rejected": 0.005188372451812029,
-      "eval_runtime": 104.0772,
-      "eval_samples_per_second": 3.296,
+      "eval_logits/chosen": 98.73486328125,
+      "eval_logits/rejected": 98.72332763671875,
+      "eval_logps/chosen": -32.45964431762695,
+      "eval_logps/rejected": -35.88682556152344,
+      "eval_loss": 1.1574316024780273,
+      "eval_rewards/accuracies": 0.4630398452281952,
+      "eval_rewards/chosen": -0.013175476342439651,
+      "eval_rewards/margins": -0.07693858444690704,
+      "eval_rewards/rejected": 0.0637631043791771,
+      "eval_runtime": 104.0187,
+      "eval_samples_per_second": 3.297,
       "eval_steps_per_second": 0.413,
       "step": 100
     },
     {
       "epoch": 0.29,
-      "grad_norm": 14.125,
+      "grad_norm": 104.0,
       "learning_rate": 4.498257201263691e-06,
-      "logits/chosen": 83.8444595336914,
-      "logits/rejected": 83.72880554199219,
-      "logps/chosen": -32.138458251953125,
-      "logps/rejected": -32.649452209472656,
-      "loss": 0.9122,
+      "logits/chosen": 83.91990661621094,
+      "logits/rejected": 83.79493713378906,
+      "logps/chosen": -32.23216247558594,
+      "logps/rejected": -32.69822692871094,
+      "loss": 0.7793,
       "rewards/accuracies": 0.625,
-      "rewards/chosen": 0.08843465149402618,
-      "rewards/margins": 0.0877910926938057,
-      "rewards/rejected": 0.0006435603136196733,
+      "rewards/chosen": 0.6325147151947021,
+      "rewards/margins": 0.6663830280303955,
+      "rewards/rejected": -0.03386829048395157,
       "step": 110
     },
     {
       "epoch": 0.31,
-      "grad_norm": 13.0,
+      "grad_norm": 108.5,
       "learning_rate": 4.353806263777678e-06,
-      "logits/chosen": 83.81868743896484,
-      "logits/rejected": 83.92826080322266,
-      "logps/chosen": -28.111730575561523,
-      "logps/rejected": -35.34394454956055,
-      "loss": 0.9297,
-      "rewards/accuracies": 0.5874999761581421,
-      "rewards/chosen": 0.08178504556417465,
-      "rewards/margins": 0.07026515156030655,
-      "rewards/rejected": 0.011519892141222954,
+      "logits/chosen": 84.13631439208984,
+      "logits/rejected": 84.24298858642578,
+      "logps/chosen": -28.422576904296875,
+      "logps/rejected": -35.379127502441406,
+      "loss": 0.8738,
+      "rewards/accuracies": 0.5249999761581421,
+      "rewards/chosen": 0.40560561418533325,
+      "rewards/margins": 0.34159213304519653,
+      "rewards/rejected": 0.06401350349187851,
       "step": 120
     },
     {
       "epoch": 0.34,
-      "grad_norm": 9.375,
+      "grad_norm": 55.5,
       "learning_rate": 4.1940827077152755e-06,
-      "logits/chosen": 80.93681335449219,
-      "logits/rejected": 80.97130584716797,
-      "logps/chosen": -30.195592880249023,
-      "logps/rejected": -31.842870712280273,
-      "loss": 0.939,
-      "rewards/accuracies": 0.6625000238418579,
-      "rewards/chosen": 0.07117608934640884,
-      "rewards/margins": 0.06230046600103378,
-      "rewards/rejected": 0.008875617757439613,
+      "logits/chosen": 81.52787017822266,
+      "logits/rejected": 81.55335998535156,
+      "logps/chosen": -30.235469818115234,
+      "logps/rejected": -31.976581573486328,
+      "loss": 0.8013,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": 0.5375091433525085,
+      "rewards/margins": 0.5734738707542419,
+      "rewards/rejected": -0.03596482425928116,
       "step": 130
     },
     {
       "epoch": 0.36,
-      "grad_norm": 11.0625,
+      "grad_norm": 57.75,
       "learning_rate": 4.0204024186666215e-06,
-      "logits/chosen": 81.7886962890625,
-      "logits/rejected": 81.80296325683594,
-      "logps/chosen": -26.81143569946289,
-      "logps/rejected": -32.920143127441406,
-      "loss": 0.8867,
-      "rewards/accuracies": 0.6499999761581421,
-      "rewards/chosen": 0.0635889396071434,
-      "rewards/margins": 0.11326569318771362,
-      "rewards/rejected": -0.04967674985527992,
+      "logits/chosen": 82.77738189697266,
+      "logits/rejected": 82.78349304199219,
+      "logps/chosen": -27.04022216796875,
+      "logps/rejected": -32.86003875732422,
+      "loss": 0.7387,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": 0.32568174600601196,
+      "rewards/margins": 0.6750079393386841,
+      "rewards/rejected": -0.34932616353034973,
       "step": 140
     },
     {
       "epoch": 0.39,
-      "grad_norm": 12.1875,
+      "grad_norm": 63.0,
       "learning_rate": 3.834196265035119e-06,
-      "logits/chosen": 80.10134887695312,
-      "logits/rejected": 80.06209564208984,
-      "logps/chosen": -28.821029663085938,
-      "logps/rejected": -33.226470947265625,
-      "loss": 0.8772,
-      "rewards/accuracies": 0.7124999761581421,
-      "rewards/chosen": 0.05680803209543228,
-      "rewards/margins": 0.1228049248456955,
-      "rewards/rejected": -0.06599690765142441,
+      "logits/chosen": 81.38068389892578,
+      "logits/rejected": 81.35902404785156,
+      "logps/chosen": -28.90557289123535,
+      "logps/rejected": -32.89466094970703,
+      "loss": 0.6488,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": 0.38682785630226135,
+      "rewards/margins": 0.6493567228317261,
+      "rewards/rejected": -0.26252883672714233,
       "step": 150
     },
     {
       "epoch": 0.42,
-      "grad_norm": 13.4375,
+      "grad_norm": 66.0,
       "learning_rate": 3.636998309800573e-06,
-      "logits/chosen": 81.63700866699219,
-      "logits/rejected": 81.66020202636719,
-      "logps/chosen": -34.080284118652344,
-      "logps/rejected": -30.817296981811523,
-      "loss": 0.9079,
-      "rewards/accuracies": 0.699999988079071,
-      "rewards/chosen": 0.005646559409797192,
-      "rewards/margins": 0.09212217479944229,
-      "rewards/rejected": -0.08647561073303223,
+      "logits/chosen": 83.2899169921875,
+      "logits/rejected": 83.30531311035156,
+      "logps/chosen": -33.436553955078125,
+      "logps/rejected": -30.43096351623535,
+      "loss": 0.6075,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": 0.5601544976234436,
+      "rewards/margins": 0.9428972005844116,
+      "rewards/rejected": -0.382742702960968,
       "step": 160
     },
     {
       "epoch": 0.44,
-      "grad_norm": 15.125,
+      "grad_norm": 70.5,
       "learning_rate": 3.4304331721118078e-06,
-      "logits/chosen": 82.27827453613281,
-      "logits/rejected": 82.23299407958984,
-      "logps/chosen": -31.016094207763672,
-      "logps/rejected": -33.03407669067383,
-      "loss": 0.8743,
-      "rewards/accuracies": 0.699999988079071,
-      "rewards/chosen": 0.024319300428032875,
-      "rewards/margins": 0.12572081387043,
-      "rewards/rejected": -0.10140150785446167,
+      "logits/chosen": 84.12855529785156,
+      "logits/rejected": 84.0693359375,
+      "logps/chosen": -30.63346290588379,
+      "logps/rejected": -32.29146194458008,
+      "loss": 0.7575,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.5006591081619263,
+      "rewards/margins": 0.717779278755188,
+      "rewards/rejected": -0.21712002158164978,
       "step": 170
     },
     {
       "epoch": 0.47,
-      "grad_norm": 12.5,
+      "grad_norm": 65.0,
       "learning_rate": 3.2162026428305436e-06,
-      "logits/chosen": 79.44108581542969,
-      "logits/rejected": 79.4173355102539,
-      "logps/chosen": -30.948467254638672,
-      "logps/rejected": -32.03376007080078,
-      "loss": 0.8875,
-      "rewards/accuracies": 0.6499999761581421,
-      "rewards/chosen": 0.02486853487789631,
-      "rewards/margins": 0.1125236377120018,
-      "rewards/rejected": -0.08765510469675064,
+      "logits/chosen": 81.81861114501953,
+      "logits/rejected": 81.79876708984375,
+      "logps/chosen": -30.488971710205078,
+      "logps/rejected": -31.519611358642578,
+      "loss": 0.6265,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": 0.5665422677993774,
+      "rewards/margins": 0.8564668893814087,
+      "rewards/rejected": -0.289924681186676,
       "step": 180
     },
     {
       "epoch": 0.49,
-      "grad_norm": 9.6875,
+      "grad_norm": 34.0,
       "learning_rate": 2.996071664294641e-06,
-      "logits/chosen": 80.90037536621094,
-      "logits/rejected": 80.8733901977539,
-      "logps/chosen": -30.740875244140625,
-      "logps/rejected": -31.226177215576172,
-      "loss": 0.9221,
-      "rewards/accuracies": 0.6000000238418579,
-      "rewards/chosen": 0.002553999889642,
-      "rewards/margins": 0.07836906611919403,
-      "rewards/rejected": -0.07581506669521332,
+      "logits/chosen": 83.52326965332031,
+      "logits/rejected": 83.50785064697266,
+      "logps/chosen": -30.285634994506836,
+      "logps/rejected": -30.630422592163086,
+      "loss": 0.8287,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": 0.38462626934051514,
+      "rewards/margins": 0.5145466327667236,
+      "rewards/rejected": -0.12992039322853088,
       "step": 190
     },
     {
       "epoch": 0.52,
-      "grad_norm": 14.4375,
+      "grad_norm": 50.25,
       "learning_rate": 2.7718537898066833e-06,
-      "logits/chosen": 76.02571105957031,
-      "logits/rejected": 75.9735107421875,
-      "logps/chosen": -34.12778854370117,
-      "logps/rejected": -33.30614471435547,
-      "loss": 0.8562,
-      "rewards/accuracies": 0.6625000238418579,
-      "rewards/chosen": 0.0624953992664814,
-      "rewards/margins": 0.14376921951770782,
-      "rewards/rejected": -0.08127383887767792,
+      "logits/chosen": 79.0837173461914,
+      "logits/rejected": 79.03185272216797,
+      "logps/chosen": -33.81468963623047,
+      "logps/rejected": -32.4266357421875,
+      "loss": 0.8098,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": 0.7504364252090454,
+      "rewards/margins": 0.6970199346542358,
+      "rewards/rejected": 0.053416453301906586,
       "step": 200
     },
     {
       "epoch": 0.52,
-      "eval_logits/chosen": 97.97191619873047,
-      "eval_logits/rejected": 97.95458984375,
-      "eval_logps/chosen": -32.94685363769531,
-      "eval_logps/rejected": -36.8216552734375,
-      "eval_loss": 0.9646754860877991,
-      "eval_rewards/accuracies": 0.5676910281181335,
-      "eval_rewards/chosen": -0.05036771669983864,
-      "eval_rewards/margins": 0.035145342350006104,
-      "eval_rewards/rejected": -0.08551305532455444,
-      "eval_runtime": 103.8205,
-      "eval_samples_per_second": 3.304,
+      "eval_logits/chosen": 98.86815643310547,
+      "eval_logits/rejected": 98.84379577636719,
+      "eval_logps/chosen": -32.561100006103516,
+      "eval_logps/rejected": -36.16287612915039,
+      "eval_loss": 1.0544954538345337,
+      "eval_rewards/accuracies": 0.5278239250183105,
+      "eval_rewards/chosen": -0.09434036910533905,
+      "eval_rewards/margins": 0.06273789703845978,
+      "eval_rewards/rejected": -0.15707828104496002,
+      "eval_runtime": 103.8402,
+      "eval_samples_per_second": 3.303,
       "eval_steps_per_second": 0.414,
       "step": 200
     },
     {
       "epoch": 0.55,
-      "grad_norm": 18.625,
+      "grad_norm": 104.0,
       "learning_rate": 2.5453962426402006e-06,
-      "logits/chosen": 78.56858825683594,
-      "logits/rejected": 78.47276306152344,
-      "logps/chosen": -33.586204528808594,
-      "logps/rejected": -36.05295944213867,
-      "loss": 0.8633,
-      "rewards/accuracies": 0.612500011920929,
-      "rewards/chosen": 0.0359516479074955,
-      "rewards/margins": 0.1426382064819336,
-      "rewards/rejected": -0.1066865548491478,
+      "logits/chosen": 81.58848571777344,
+      "logits/rejected": 81.49668884277344,
+      "logps/chosen": -33.169471740722656,
+      "logps/rejected": -34.991939544677734,
+      "loss": 0.7585,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": 0.620996356010437,
+      "rewards/margins": 0.6256735324859619,
+      "rewards/rejected": -0.004677181597799063,
       "step": 210
     },
     {
       "epoch": 0.57,
-      "grad_norm": 14.6875,
+      "grad_norm": 35.0,
       "learning_rate": 2.3185646976551794e-06,
-      "logits/chosen": 80.53726959228516,
-      "logits/rejected": 80.62950134277344,
-      "logps/chosen": -31.454355239868164,
-      "logps/rejected": -32.02970504760742,
-      "loss": 0.8286,
-      "rewards/accuracies": 0.6875,
-      "rewards/chosen": 0.049460187554359436,
-      "rewards/margins": 0.17953529953956604,
-      "rewards/rejected": -0.1300750970840454,
+      "logits/chosen": 83.66175079345703,
+      "logits/rejected": 83.74690246582031,
+      "logps/chosen": -31.094005584716797,
+      "logps/rejected": -31.061717987060547,
+      "loss": 0.626,
+      "rewards/accuracies": 0.7875000238418579,
+      "rewards/chosen": 0.683959424495697,
+      "rewards/margins": 0.9501722455024719,
+      "rewards/rejected": -0.2662127912044525,
       "step": 220
     },
     {
       "epoch": 0.6,
-      "grad_norm": 14.0,
+      "grad_norm": 69.5,
       "learning_rate": 2.0932279108998323e-06,
-      "logits/chosen": 77.57215881347656,
-      "logits/rejected": 77.61759948730469,
-      "logps/chosen": -32.74131393432617,
-      "logps/rejected": -35.29151153564453,
-      "loss": 0.8713,
-      "rewards/accuracies": 0.6625000238418579,
-      "rewards/chosen": 0.006443141493946314,
-      "rewards/margins": 0.13437876105308533,
-      "rewards/rejected": -0.12793561816215515,
+      "logits/chosen": 80.7747802734375,
+      "logits/rejected": 80.82144165039062,
+      "logps/chosen": -32.322105407714844,
+      "logps/rejected": -34.30824661254883,
+      "loss": 0.7928,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.38691413402557373,
+      "rewards/margins": 0.6237919926643372,
+      "rewards/rejected": -0.23687779903411865,
       "step": 230
     },
     {
       "epoch": 0.62,
-      "grad_norm": 16.5,
+      "grad_norm": 89.5,
       "learning_rate": 1.8712423238279358e-06,
-      "logits/chosen": 79.76509094238281,
-      "logits/rejected": 80.07305908203125,
-      "logps/chosen": -31.252161026000977,
-      "logps/rejected": -32.75459671020508,
-      "loss": 0.8345,
+      "logits/chosen": 83.18328094482422,
+      "logits/rejected": 83.4581069946289,
+      "logps/chosen": -30.6195125579834,
+      "logps/rejected": -31.637020111083984,
+      "loss": 0.5475,
       "rewards/accuracies": 0.737500011920929,
-      "rewards/chosen": 0.04359050467610359,
-      "rewards/margins": 0.16584597527980804,
-      "rewards/rejected": -0.12225550413131714,
+      "rewards/chosen": 0.854840099811554,
+      "rewards/margins": 0.9388192892074585,
+      "rewards/rejected": -0.08397923409938812,
       "step": 240
     },
     {
       "epoch": 0.65,
-      "grad_norm": 15.0625,
+      "grad_norm": 84.5,
       "learning_rate": 1.6544367689701824e-06,
-      "logits/chosen": 78.32225036621094,
-      "logits/rejected": 78.37224578857422,
-      "logps/chosen": -27.815731048583984,
-      "logps/rejected": -31.077539443969727,
-      "loss": 0.9043,
-      "rewards/accuracies": 0.6000000238418579,
-      "rewards/chosen": -0.02198370359838009,
-      "rewards/margins": 0.09569612890481949,
-      "rewards/rejected": -0.11767983436584473,
+      "logits/chosen": 81.9409408569336,
+      "logits/rejected": 82.00200653076172,
+      "logps/chosen": -26.908878326416016,
+      "logps/rejected": -30.06099510192871,
+      "loss": 0.8063,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": 0.5496150255203247,
+      "rewards/margins": 0.6778196692466736,
+      "rewards/rejected": -0.12820473313331604,
       "step": 250
     },
     {
       "epoch": 0.68,
-      "grad_norm": 13.75,
+      "grad_norm": 60.25,
       "learning_rate": 1.4445974030621963e-06,
-      "logits/chosen": 75.31268310546875,
-      "logits/rejected": 75.46823120117188,
-      "logps/chosen": -31.16106605529785,
-      "logps/rejected": -38.22980880737305,
-      "loss": 0.7589,
-      "rewards/accuracies": 0.7250000238418579,
-      "rewards/chosen": 0.039992958307266235,
-      "rewards/margins": 0.2528113126754761,
-      "rewards/rejected": -0.21281830966472626,
+      "logits/chosen": 79.24473571777344,
+      "logits/rejected": 79.37559509277344,
+      "logps/chosen": -30.283336639404297,
+      "logps/rejected": -36.4005241394043,
+      "loss": 0.4689,
+      "rewards/accuracies": 0.8125,
+      "rewards/chosen": 1.0221258401870728,
+      "rewards/margins": 1.2612489461898804,
+      "rewards/rejected": -0.23912319540977478,
       "step": 260
     },
     {
       "epoch": 0.7,
-      "grad_norm": 13.1875,
+      "grad_norm": 31.875,
       "learning_rate": 1.243452991757889e-06,
-      "logits/chosen": 74.42684173583984,
-      "logits/rejected": 74.45647430419922,
-      "logps/chosen": -31.796234130859375,
-      "logps/rejected": -33.113197326660156,
-      "loss": 0.8432,
-      "rewards/accuracies": 0.6875,
-      "rewards/chosen": -0.0017309554386883974,
-      "rewards/margins": 0.15904627740383148,
-      "rewards/rejected": -0.1607772409915924,
+      "logits/chosen": 78.6611328125,
+      "logits/rejected": 78.69209289550781,
+      "logps/chosen": -30.992252349853516,
+      "logps/rejected": -31.770212173461914,
+      "loss": 0.6471,
+      "rewards/accuracies": 0.7250000238418579,
+      "rewards/chosen": 0.6293355822563171,
+      "rewards/margins": 0.8411690592765808,
+      "rewards/rejected": -0.21183356642723083,
       "step": 270
     },
     {
       "epoch": 0.73,
-      "grad_norm": 19.75,
+      "grad_norm": 100.0,
       "learning_rate": 1.0526606671603523e-06,
-      "logits/chosen": 77.391845703125,
-      "logits/rejected": 77.1694564819336,
-      "logps/chosen": -32.110328674316406,
-      "logps/rejected": -30.959293365478516,
-      "loss": 0.9109,
-      "rewards/accuracies": 0.6499999761581421,
-      "rewards/chosen": -0.030043601989746094,
-      "rewards/margins": 0.09918614476919174,
-      "rewards/rejected": -0.12922975420951843,
+      "logits/chosen": 81.2646255493164,
+      "logits/rejected": 81.05351257324219,
+      "logps/chosen": -31.10286521911621,
+      "logps/rejected": -29.866031646728516,
+      "loss": 0.7333,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": 0.5656233429908752,
+      "rewards/margins": 0.7248517274856567,
+      "rewards/rejected": -0.15922844409942627,
       "step": 280
     },
     {
       "epoch": 0.75,
-      "grad_norm": 13.1875,
+      "grad_norm": 61.25,
       "learning_rate": 8.737922755071455e-07,
-      "logits/chosen": 77.39460754394531,
-      "logits/rejected": 77.31553649902344,
-      "logps/chosen": -33.84746551513672,
-      "logps/rejected": -34.28327178955078,
-      "loss": 0.7586,
-      "rewards/accuracies": 0.762499988079071,
-      "rewards/chosen": 0.02567802369594574,
-      "rewards/margins": 0.2523185908794403,
-      "rewards/rejected": -0.22664058208465576,
+      "logits/chosen": 81.45980072021484,
+      "logits/rejected": 81.37715148925781,
+      "logps/chosen": -33.03575134277344,
+      "logps/rejected": -32.42512512207031,
+      "loss": 0.6231,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": 0.8547972440719604,
+      "rewards/margins": 1.1814041137695312,
+      "rewards/rejected": -0.32660672068595886,
       "step": 290
     },
     {
       "epoch": 0.78,
-      "grad_norm": 14.8125,
+      "grad_norm": 78.0,
       "learning_rate": 7.08321427484816e-07,
-      "logits/chosen": 73.05286407470703,
-      "logits/rejected": 73.1805648803711,
-      "logps/chosen": -33.01818084716797,
-      "logps/rejected": -30.520904541015625,
-      "loss": 0.8271,
-      "rewards/accuracies": 0.7124999761581421,
-      "rewards/chosen": 0.03875284641981125,
-      "rewards/margins": 0.18064062297344208,
-      "rewards/rejected": -0.14188775420188904,
+      "logits/chosen": 77.12594604492188,
+      "logits/rejected": 77.21607971191406,
+      "logps/chosen": -32.209590911865234,
+      "logps/rejected": -29.23952865600586,
+      "loss": 0.6965,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": 0.9568928480148315,
+      "rewards/margins": 1.066897988319397,
+      "rewards/rejected": -0.11000506579875946,
       "step": 300
     },
     {
       "epoch": 0.78,
-      "eval_logits/chosen": 97.71524047851562,
-      "eval_logits/rejected": 97.68975067138672,
-      "eval_logps/chosen": -33.206886291503906,
-      "eval_logps/rejected": -37.196205139160156,
-      "eval_loss": 0.9533767700195312,
-      "eval_rewards/accuracies": 0.5714285373687744,
-      "eval_rewards/chosen": -0.07637124508619308,
-      "eval_rewards/margins": 0.04659651592373848,
-      "eval_rewards/rejected": -0.12296776473522186,
-      "eval_runtime": 104.0124,
-      "eval_samples_per_second": 3.298,
-      "eval_steps_per_second": 0.413,
+      "eval_logits/chosen": 98.90528106689453,
+      "eval_logits/rejected": 98.8800277709961,
+      "eval_logps/chosen": -32.50230407714844,
+      "eval_logps/rejected": -36.192344665527344,
+      "eval_loss": 0.9868877530097961,
+      "eval_rewards/accuracies": 0.5830564498901367,
+      "eval_rewards/chosen": -0.047303199768066406,
+      "eval_rewards/margins": 0.1333501785993576,
+      "eval_rewards/rejected": -0.180653378367424,
+      "eval_runtime": 103.7665,
+      "eval_samples_per_second": 3.305,
+      "eval_steps_per_second": 0.414,
       "step": 300
     },
     {
       "epoch": 0.81,
-      "grad_norm": 13.0,
+      "grad_norm": 59.25,
       "learning_rate": 5.576113578589035e-07,
-      "logits/chosen": 80.51168060302734,
-      "logits/rejected": 80.51959228515625,
-      "logps/chosen": -30.940738677978516,
-      "logps/rejected": -33.976158142089844,
-      "loss": 0.841,
-      "rewards/accuracies": 0.6875,
-      "rewards/chosen": -0.00772014120593667,
-      "rewards/margins": 0.1699191778898239,
-      "rewards/rejected": -0.17763930559158325,
+      "logits/chosen": 84.09150695800781,
+      "logits/rejected": 84.11959075927734,
+      "logps/chosen": -30.127155303955078,
+      "logps/rejected": -32.29438781738281,
+      "loss": 0.7446,
+      "rewards/accuracies": 0.7250000238418579,
+      "rewards/chosen": 0.5891044735908508,
+      "rewards/margins": 0.6648039817810059,
+      "rewards/rejected": -0.07569954544305801,
       "step": 310
     },
     {
       "epoch": 0.83,
-      "grad_norm": 14.9375,
+      "grad_norm": 60.25,
       "learning_rate": 4.229036944380913e-07,
-      "logits/chosen": 77.665771484375,
-      "logits/rejected": 77.6807861328125,
-      "logps/chosen": -31.073196411132812,
-      "logps/rejected": -30.321044921875,
-      "loss": 0.8202,
-      "rewards/accuracies": 0.699999988079071,
-      "rewards/chosen": 0.05253634601831436,
-      "rewards/margins": 0.19269177317619324,
-      "rewards/rejected": -0.14015543460845947,
+      "logits/chosen": 81.67195892333984,
+      "logits/rejected": 81.67171478271484,
+      "logps/chosen": -30.566574096679688,
+      "logps/rejected": -29.08144187927246,
+      "loss": 0.5518,
+      "rewards/accuracies": 0.762499988079071,
+      "rewards/chosen": 0.8255898356437683,
+      "rewards/margins": 0.9551518559455872,
+      "rewards/rejected": -0.12956194579601288,
       "step": 320
     },
     {
       "epoch": 0.86,
-      "grad_norm": 16.375,
+      "grad_norm": 58.0,
       "learning_rate": 3.053082288996112e-07,
-      "logits/chosen": 74.7007064819336,
-      "logits/rejected": 74.74031829833984,
-      "logps/chosen": -29.66641616821289,
-      "logps/rejected": -34.42815399169922,
-      "loss": 0.7549,
-      "rewards/accuracies": 0.762499988079071,
-      "rewards/chosen": 0.06861907988786697,
-      "rewards/margins": 0.24794825911521912,
-      "rewards/rejected": -0.17932915687561035,
+      "logits/chosen": 78.98292541503906,
+      "logits/rejected": 79.03089904785156,
+      "logps/chosen": -29.0233211517334,
+      "logps/rejected": -32.939231872558594,
+      "loss": 0.5283,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": 1.063429832458496,
+      "rewards/margins": 1.306926965713501,
+      "rewards/rejected": -0.24349698424339294,
       "step": 330
     },
     {
       "epoch": 0.88,
-      "grad_norm": 17.125,
+      "grad_norm": 88.5,
       "learning_rate": 2.0579377374915805e-07,
-      "logits/chosen": 79.32682800292969,
-      "logits/rejected": 79.36729431152344,
-      "logps/chosen": -33.05027770996094,
-      "logps/rejected": -35.377281188964844,
-      "loss": 0.7972,
-      "rewards/accuracies": 0.737500011920929,
-      "rewards/chosen": 0.01618681475520134,
-      "rewards/margins": 0.21704569458961487,
-      "rewards/rejected": -0.20085887610912323,
+      "logits/chosen": 83.09165954589844,
+      "logits/rejected": 83.11895751953125,
+      "logps/chosen": -32.228084564208984,
+      "logps/rejected": -33.656490325927734,
+      "loss": 0.6498,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": 0.7872547507286072,
+      "rewards/margins": 1.0174915790557861,
+      "rewards/rejected": -0.23023685812950134,
       "step": 340
     },
     {
       "epoch": 0.91,
-      "grad_norm": 15.5,
+      "grad_norm": 42.5,
       "learning_rate": 1.2518018074041684e-07,
-      "logits/chosen": 78.2139663696289,
-      "logits/rejected": 78.22695922851562,
-      "logps/chosen": -33.175575256347656,
-      "logps/rejected": -34.91777801513672,
-      "loss": 0.7788,
-      "rewards/accuracies": 0.737500011920929,
-      "rewards/chosen": 0.04681625962257385,
-      "rewards/margins": 0.22941403090953827,
-      "rewards/rejected": -0.18259775638580322,
+      "logits/chosen": 82.23709869384766,
+      "logits/rejected": 82.24420928955078,
+      "logps/chosen": -32.544349670410156,
+      "logps/rejected": -33.209228515625,
+      "loss": 0.6454,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": 0.8795096278190613,
+      "rewards/margins": 0.9734523892402649,
+      "rewards/rejected": -0.09394274652004242,
       "step": 350
     },
     {
       "epoch": 0.94,
-      "grad_norm": 11.6875,
+      "grad_norm": 66.0,
       "learning_rate": 6.41315865106129e-08,
-      "logits/chosen": 79.90480041503906,
-      "logits/rejected": 79.93946075439453,
-      "logps/chosen": -28.85941505432129,
-      "logps/rejected": -33.032447814941406,
-      "loss": 0.8014,
-      "rewards/accuracies": 0.6625000238418579,
-      "rewards/chosen": 0.06927979737520218,
-      "rewards/margins": 0.20753948390483856,
-      "rewards/rejected": -0.13825969398021698,
+      "logits/chosen": 83.5949935913086,
+      "logits/rejected": 83.62043762207031,
+      "logps/chosen": -28.472000122070312,
+      "logps/rejected": -31.736831665039062,
+      "loss": 0.6336,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": 0.8641688227653503,
+      "rewards/margins": 0.9337539672851562,
+      "rewards/rejected": -0.0695851519703865,
       "step": 360
     },
     {
       "epoch": 0.96,
-      "grad_norm": 12.8125,
+      "grad_norm": 78.5,
       "learning_rate": 2.3150941078050325e-08,
-      "logits/chosen": 79.3287582397461,
-      "logits/rejected": 79.33840942382812,
-      "logps/chosen": -33.099693298339844,
-      "logps/rejected": -37.02233123779297,
-      "loss": 0.8425,
-      "rewards/accuracies": 0.6625000238418579,
-      "rewards/chosen": -0.022353025153279305,
-      "rewards/margins": 0.16894304752349854,
-      "rewards/rejected": -0.1912960708141327,
+      "logits/chosen": 83.08488464355469,
+      "logits/rejected": 83.10858917236328,
+      "logps/chosen": -31.81646156311035,
+      "logps/rejected": -35.26220703125,
+      "loss": 0.6364,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": 0.847761332988739,
+      "rewards/margins": 0.9700264930725098,
+      "rewards/rejected": -0.1222650408744812,
       "step": 370
     },
     {
       "epoch": 0.99,
-      "grad_norm": 13.25,
+      "grad_norm": 82.5,
       "learning_rate": 2.575864278703266e-09,
-      "logits/chosen": 72.94795227050781,
-      "logits/rejected": 72.81517791748047,
-      "logps/chosen": -30.674768447875977,
-      "logps/rejected": -29.731273651123047,
-      "loss": 0.8564,
-      "rewards/accuracies": 0.6875,
-      "rewards/chosen": -0.009864235296845436,
-      "rewards/margins": 0.14720600843429565,
-      "rewards/rejected": -0.15707024931907654,
+      "logits/chosen": 77.07919311523438,
+      "logits/rejected": 76.94573974609375,
+      "logps/chosen": -29.726633071899414,
+      "logps/rejected": -28.218725204467773,
+      "loss": 0.7075,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.6795950531959534,
+      "rewards/margins": 0.7261168956756592,
+      "rewards/rejected": -0.046521805226802826,
       "step": 380
     },
     {
       "epoch": 1.0,
       "step": 385,
       "total_flos": 0.0,
-      "train_loss": 0.88636748450143,
-      "train_runtime": 2554.6221,
-      "train_samples_per_second": 1.205,
+      "train_loss": 0.7616742146479619,
+      "train_runtime": 2551.1965,
+      "train_samples_per_second": 1.207,
       "train_steps_per_second": 0.151
     }
   ],