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: 1.0237
-- Rewards/chosen: -0.0663
-- Rewards/rejected: -0.1206
-- Rewards/accuracies: 0.5311
-- Rewards/margins: 0.0543
-- Logps/rejected: -36.1388
-- Logps/chosen: -32.5378
-- Logits/rejected: 98.6966
-- Logits/chosen: 98.7185
+- Loss: 0.9809
+- Rewards/chosen: -0.0923
+- Rewards/rejected: -0.1298
+- Rewards/accuracies: 0.5158
+- Rewards/margins: 0.0375
+- Logps/rejected: -36.2911
+- Logps/chosen: -32.6740
+- Logits/rejected: 98.5972
+- Logits/chosen: 98.6205
 
 ## 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 |
 |:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
-| 0.9161        | 0.26  | 100  | 1.0662          | -0.0346        | -0.0243          | 0.5129             | -0.0103         | -36.0012       | -32.4926     | 98.7623         | 98.7729       |
-| 0.8398        | 0.52  | 200  | 1.0090          | -0.1355        | -0.2279          | 0.5141             | 0.0924          | -36.2922       | -32.6368     | 98.6916         | 98.7167       |
-| 0.733         | 0.78  | 300  | 1.0124          | -0.0636        | -0.1409          | 0.5399             | 0.0773          | -36.1679       | -32.5340     | 98.7098         | 98.7339       |
+| 0.8779        | 0.26  | 100  | 0.9411          | 0.0053         | -0.0639          | 0.5461             | 0.0692          | -36.1263       | -32.4300     | 98.7518         | 98.7613       |
+| 0.683         | 0.52  | 200  | 0.9410          | -0.1143        | -0.1935          | 0.5569             | 0.0793          | -36.4504       | -32.7289     | 98.5890         | 98.6121       |
+| 0.7349        | 0.78  | 300  | 0.9809          | -0.0923        | -0.1298          | 0.5158             | 0.0375          | -36.2911       | -32.6740     | 98.5972         | 98.6205       |
 
 
 ### Framework versions

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:12f217803f6082d2fac75d6b435e8620e698ba3512c507b5952bf5a12433d0f2
+oid sha256:b9f71e75b5b337c21d2dd78223a44fe5451a0c980e81b68799c41f830503d425
 size 671150064

all_results.json

@@ -1,21 +1,8 @@
 {
     "epoch": 1.0,
-    "eval_logits/chosen": 98.71852111816406,
-    "eval_logits/rejected": 98.69664001464844,
-    "eval_logps/chosen": -32.537837982177734,
-    "eval_logps/rejected": -36.13877868652344,
-    "eval_loss": 1.023661732673645,
-    "eval_rewards/accuracies": 0.531146228313446,
-    "eval_rewards/chosen": -0.06626088172197342,
-    "eval_rewards/margins": 0.054315000772476196,
-    "eval_rewards/rejected": -0.12057589739561081,
-    "eval_runtime": 103.7609,
-    "eval_samples": 343,
-    "eval_samples_per_second": 3.306,
-    "eval_steps_per_second": 0.414,
-    "train_loss": 0.7429099231571347,
-    "train_runtime": 2555.3187,
+    "train_loss": 0.741242148659446,
+    "train_runtime": 2553.0854,
     "train_samples": 3079,
-    "train_samples_per_second": 1.205,
+    "train_samples_per_second": 1.206,
     "train_steps_per_second": 0.151
 }

train_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 1.0,
-    "train_loss": 0.7429099231571347,
-    "train_runtime": 2555.3187,
+    "train_loss": 0.741242148659446,
+    "train_runtime": 2553.0854,
     "train_samples": 3079,
-    "train_samples_per_second": 1.205,
+    "train_samples_per_second": 1.206,
     "train_steps_per_second": 0.151
 }

trainer_state.json

@@ -10,7 +10,7 @@
   "log_history": [
     {
       "epoch": 0.0,
-      "grad_norm": 91.5,
+      "grad_norm": 52.25,
       "learning_rate": 1.282051282051282e-07,
       "logits/chosen": 88.18099975585938,
       "logits/rejected": 88.25153350830078,
@@ -25,629 +25,629 @@
     },
     {
       "epoch": 0.03,
-      "grad_norm": 76.5,
+      "grad_norm": 43.5,
       "learning_rate": 1.282051282051282e-06,
-      "logits/chosen": 81.08255004882812,
-      "logits/rejected": 80.78926086425781,
-      "logps/chosen": -34.20470428466797,
-      "logps/rejected": -33.038047790527344,
-      "loss": 0.9368,
-      "rewards/accuracies": 0.5,
-      "rewards/chosen": 0.02591484785079956,
-      "rewards/margins": 0.0816662609577179,
-      "rewards/rejected": -0.055751409381628036,
+      "logits/chosen": 81.09413146972656,
+      "logits/rejected": 80.79685974121094,
+      "logps/chosen": -34.14472961425781,
+      "logps/rejected": -33.07267761230469,
+      "loss": 0.9155,
+      "rewards/accuracies": 0.5555555820465088,
+      "rewards/chosen": 0.038799185305833817,
+      "rewards/margins": 0.08450882881879807,
+      "rewards/rejected": -0.04570963978767395,
       "step": 10
     },
     {
       "epoch": 0.05,
-      "grad_norm": 60.25,
+      "grad_norm": 50.5,
       "learning_rate": 2.564102564102564e-06,
-      "logits/chosen": 80.67174530029297,
-      "logits/rejected": 80.55998229980469,
-      "logps/chosen": -33.60923767089844,
-      "logps/rejected": -30.828128814697266,
-      "loss": 0.9439,
-      "rewards/accuracies": 0.5375000238418579,
-      "rewards/chosen": 0.05647152662277222,
-      "rewards/margins": 0.09191782772541046,
-      "rewards/rejected": -0.035446297377347946,
+      "logits/chosen": 80.67982482910156,
+      "logits/rejected": 80.56922149658203,
+      "logps/chosen": -33.638526916503906,
+      "logps/rejected": -30.80965805053711,
+      "loss": 0.9668,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": 0.020552771165966988,
+      "rewards/margins": 0.03342083841562271,
+      "rewards/rejected": -0.012868070974946022,
       "step": 20
     },
     {
       "epoch": 0.08,
-      "grad_norm": 67.5,
+      "grad_norm": 48.25,
       "learning_rate": 3.846153846153847e-06,
-      "logits/chosen": 82.49557495117188,
-      "logits/rejected": 82.5277099609375,
-      "logps/chosen": -33.90664291381836,
-      "logps/rejected": -31.188213348388672,
-      "loss": 1.0985,
-      "rewards/accuracies": 0.42500001192092896,
-      "rewards/chosen": 0.07432325184345245,
-      "rewards/margins": -0.0906088799238205,
-      "rewards/rejected": 0.16493213176727295,
+      "logits/chosen": 82.50788879394531,
+      "logits/rejected": 82.54085540771484,
+      "logps/chosen": -33.867698669433594,
+      "logps/rejected": -31.21173667907715,
+      "loss": 1.0398,
+      "rewards/accuracies": 0.44999998807907104,
+      "rewards/chosen": 0.05804871767759323,
+      "rewards/margins": -0.026789050549268723,
+      "rewards/rejected": 0.08483776450157166,
       "step": 30
     },
     {
       "epoch": 0.1,
-      "grad_norm": 75.5,
+      "grad_norm": 37.0,
       "learning_rate": 4.999896948438434e-06,
-      "logits/chosen": 81.06272888183594,
-      "logits/rejected": 81.05645751953125,
-      "logps/chosen": -32.722740173339844,
-      "logps/rejected": -33.15789031982422,
-      "loss": 0.8862,
-      "rewards/accuracies": 0.625,
-      "rewards/chosen": 0.3159271776676178,
-      "rewards/margins": 0.1684812754392624,
-      "rewards/rejected": 0.1474459171295166,
+      "logits/chosen": 81.12205505371094,
+      "logits/rejected": 81.11637878417969,
+      "logps/chosen": -32.64350128173828,
+      "logps/rejected": -33.15576934814453,
+      "loss": 0.9203,
+      "rewards/accuracies": 0.5249999761581421,
+      "rewards/chosen": 0.21222472190856934,
+      "rewards/margins": 0.1271229237318039,
+      "rewards/rejected": 0.08510179817676544,
       "step": 40
     },
     {
       "epoch": 0.13,
-      "grad_norm": 48.75,
+      "grad_norm": 28.125,
       "learning_rate": 4.987541037542187e-06,
-      "logits/chosen": 78.66886901855469,
-      "logits/rejected": 78.68830871582031,
-      "logps/chosen": -30.57431411743164,
-      "logps/rejected": -30.798864364624023,
-      "loss": 0.9351,
-      "rewards/accuracies": 0.6000000238418579,
-      "rewards/chosen": 0.44165611267089844,
-      "rewards/margins": 0.24956238269805908,
-      "rewards/rejected": 0.19209368526935577,
+      "logits/chosen": 78.77560424804688,
+      "logits/rejected": 78.78538513183594,
+      "logps/chosen": -30.54385757446289,
+      "logps/rejected": -30.8847599029541,
+      "loss": 0.8737,
+      "rewards/accuracies": 0.5375000238418579,
+      "rewards/chosen": 0.26455581188201904,
+      "rewards/margins": 0.18914423882961273,
+      "rewards/rejected": 0.07541157305240631,
       "step": 50
     },
     {
       "epoch": 0.16,
-      "grad_norm": 72.5,
+      "grad_norm": 40.25,
       "learning_rate": 4.954691471941119e-06,
-      "logits/chosen": 83.24879455566406,
-      "logits/rejected": 83.29969787597656,
-      "logps/chosen": -30.918521881103516,
-      "logps/rejected": -29.482006072998047,
-      "loss": 1.0359,
-      "rewards/accuracies": 0.48750001192092896,
-      "rewards/chosen": 0.17964980006217957,
-      "rewards/margins": 0.1036890521645546,
-      "rewards/rejected": 0.07596075534820557,
+      "logits/chosen": 83.37484741210938,
+      "logits/rejected": 83.43515014648438,
+      "logps/chosen": -30.861572265625,
+      "logps/rejected": -29.3466796875,
+      "loss": 1.0221,
+      "rewards/accuracies": 0.4749999940395355,
+      "rewards/chosen": 0.12543414533138275,
+      "rewards/margins": 0.027898848056793213,
+      "rewards/rejected": 0.09753530472517014,
       "step": 60
     },
     {
       "epoch": 0.18,
-      "grad_norm": 87.0,
+      "grad_norm": 51.25,
       "learning_rate": 4.901618883413549e-06,
-      "logits/chosen": 83.91014099121094,
-      "logits/rejected": 83.93647766113281,
-      "logps/chosen": -30.502422332763672,
-      "logps/rejected": -33.12609100341797,
-      "loss": 0.9567,
+      "logits/chosen": 84.05386352539062,
+      "logits/rejected": 84.0824966430664,
+      "logps/chosen": -30.508914947509766,
+      "logps/rejected": -33.01020431518555,
+      "loss": 0.9697,
       "rewards/accuracies": 0.5249999761581421,
-      "rewards/chosen": 0.08862228691577911,
-      "rewards/margins": 0.15049318969249725,
-      "rewards/rejected": -0.06187089532613754,
+      "rewards/chosen": 0.04804382845759392,
+      "rewards/margins": 0.0370425209403038,
+      "rewards/rejected": 0.011001313105225563,
       "step": 70
     },
     {
       "epoch": 0.21,
-      "grad_norm": 74.0,
+      "grad_norm": 40.25,
       "learning_rate": 4.828760511501322e-06,
-      "logits/chosen": 81.4886245727539,
-      "logits/rejected": 81.47552490234375,
-      "logps/chosen": -31.330123901367188,
-      "logps/rejected": -31.121978759765625,
-      "loss": 0.814,
-      "rewards/accuracies": 0.6499999761581421,
-      "rewards/chosen": 0.12237221002578735,
-      "rewards/margins": 0.3774043917655945,
-      "rewards/rejected": -0.25503218173980713,
+      "logits/chosen": 81.74610900878906,
+      "logits/rejected": 81.73453521728516,
+      "logps/chosen": -31.33428382873535,
+      "logps/rejected": -30.90468978881836,
+      "loss": 0.9004,
+      "rewards/accuracies": 0.5375000238418579,
+      "rewards/chosen": 0.06826337426900864,
+      "rewards/margins": 0.12708182632923126,
+      "rewards/rejected": -0.05881844088435173,
       "step": 80
     },
     {
       "epoch": 0.23,
-      "grad_norm": 83.0,
+      "grad_norm": 47.75,
       "learning_rate": 4.7367166013034295e-06,
-      "logits/chosen": 78.25160217285156,
-      "logits/rejected": 78.21989440917969,
-      "logps/chosen": -32.49519729614258,
-      "logps/rejected": -31.26288414001465,
-      "loss": 0.8564,
-      "rewards/accuracies": 0.6000000238418579,
-      "rewards/chosen": 0.10008607059717178,
-      "rewards/margins": 0.31807559728622437,
-      "rewards/rejected": -0.21798951923847198,
+      "logits/chosen": 78.39300537109375,
+      "logits/rejected": 78.36354064941406,
+      "logps/chosen": -32.50352478027344,
+      "logps/rejected": -31.204748153686523,
+      "loss": 0.8878,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": 0.05386130139231682,
+      "rewards/margins": 0.1551724374294281,
+      "rewards/rejected": -0.10131114721298218,
       "step": 90
     },
     {
       "epoch": 0.26,
-      "grad_norm": 82.5,
+      "grad_norm": 48.5,
       "learning_rate": 4.626245458345211e-06,
-      "logits/chosen": 83.49636840820312,
-      "logits/rejected": 83.53155517578125,
-      "logps/chosen": -34.187034606933594,
-      "logps/rejected": -31.907695770263672,
-      "loss": 0.9161,
-      "rewards/accuracies": 0.5375000238418579,
-      "rewards/chosen": 0.0825684443116188,
-      "rewards/margins": 0.19257497787475586,
-      "rewards/rejected": -0.11000655591487885,
+      "logits/chosen": 83.5033950805664,
+      "logits/rejected": 83.53245544433594,
+      "logps/chosen": -34.1702880859375,
+      "logps/rejected": -31.961788177490234,
+      "loss": 0.8779,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": 0.05388103798031807,
+      "rewards/margins": 0.13837869465351105,
+      "rewards/rejected": -0.08449766039848328,
       "step": 100
     },
     {
       "epoch": 0.26,
-      "eval_logits/chosen": 98.7728500366211,
-      "eval_logits/rejected": 98.76228332519531,
-      "eval_logps/chosen": -32.4925651550293,
-      "eval_logps/rejected": -36.00117492675781,
-      "eval_loss": 1.0662287473678589,
-      "eval_rewards/accuracies": 0.5128737688064575,
-      "eval_rewards/chosen": -0.03457321599125862,
-      "eval_rewards/margins": -0.010320436209440231,
-      "eval_rewards/rejected": -0.02425277978181839,
-      "eval_runtime": 104.2438,
-      "eval_samples_per_second": 3.29,
-      "eval_steps_per_second": 0.412,
+      "eval_logits/chosen": 98.76129913330078,
+      "eval_logits/rejected": 98.75178527832031,
+      "eval_logps/chosen": -32.43000030517578,
+      "eval_logps/rejected": -36.12626647949219,
+      "eval_loss": 0.9410777688026428,
+      "eval_rewards/accuracies": 0.5460963845252991,
+      "eval_rewards/chosen": 0.005270163994282484,
+      "eval_rewards/margins": 0.06916692852973938,
+      "eval_rewards/rejected": -0.06389675289392471,
+      "eval_runtime": 104.1853,
+      "eval_samples_per_second": 3.292,
+      "eval_steps_per_second": 0.413,
       "step": 100
     },
     {
       "epoch": 0.29,
-      "grad_norm": 91.5,
+      "grad_norm": 53.5,
       "learning_rate": 4.498257201263691e-06,
-      "logits/chosen": 83.7912368774414,
-      "logits/rejected": 83.67459106445312,
-      "logps/chosen": -32.45995330810547,
-      "logps/rejected": -32.786006927490234,
-      "loss": 0.8008,
-      "rewards/accuracies": 0.675000011920929,
-      "rewards/chosen": 0.39399290084838867,
-      "rewards/margins": 0.4850761294364929,
-      "rewards/rejected": -0.09108323603868484,
+      "logits/chosen": 83.67477416992188,
+      "logits/rejected": 83.56143951416016,
+      "logps/chosen": -32.46035385131836,
+      "logps/rejected": -32.912376403808594,
+      "loss": 0.752,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": 0.22498121857643127,
+      "rewards/margins": 0.3275759816169739,
+      "rewards/rejected": -0.102594755589962,
       "step": 110
     },
     {
       "epoch": 0.31,
-      "grad_norm": 95.0,
+      "grad_norm": 52.25,
       "learning_rate": 4.353806263777678e-06,
-      "logits/chosen": 84.00267028808594,
-      "logits/rejected": 84.11933898925781,
-      "logps/chosen": -28.27166175842285,
-      "logps/rejected": -35.5056037902832,
-      "loss": 0.7011,
-      "rewards/accuracies": 0.6875,
-      "rewards/chosen": 0.4605434536933899,
-      "rewards/margins": 0.4930640757083893,
-      "rewards/rejected": -0.03252064064145088,
+      "logits/chosen": 83.85575103759766,
+      "logits/rejected": 83.95751953125,
+      "logps/chosen": -28.3017520904541,
+      "logps/rejected": -35.58953094482422,
+      "loss": 0.7604,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.2511330246925354,
+      "rewards/margins": 0.30328577756881714,
+      "rewards/rejected": -0.05215279012918472,
       "step": 120
     },
     {
       "epoch": 0.34,
-      "grad_norm": 64.5,
+      "grad_norm": 38.0,
       "learning_rate": 4.1940827077152755e-06,
-      "logits/chosen": 81.23250579833984,
-      "logits/rejected": 81.24813842773438,
-      "logps/chosen": -30.38728904724121,
-      "logps/rejected": -32.12664031982422,
-      "loss": 0.7344,
+      "logits/chosen": 81.05609130859375,
+      "logits/rejected": 81.07179260253906,
+      "logps/chosen": -30.556743621826172,
+      "logps/rejected": -32.228946685791016,
+      "loss": 0.7882,
       "rewards/accuracies": 0.612500011920929,
-      "rewards/chosen": 0.3640448749065399,
-      "rewards/margins": 0.5005542039871216,
-      "rewards/rejected": -0.13650932908058167,
+      "rewards/chosen": 0.14024314284324646,
+      "rewards/margins": 0.25916972756385803,
+      "rewards/rejected": -0.11892654746770859,
       "step": 130
     },
     {
       "epoch": 0.36,
-      "grad_norm": 59.75,
+      "grad_norm": 45.25,
       "learning_rate": 4.0204024186666215e-06,
-      "logits/chosen": 82.44822692871094,
-      "logits/rejected": 82.46932220458984,
-      "logps/chosen": -27.172740936279297,
-      "logps/rejected": -33.0168571472168,
-      "loss": 0.7509,
-      "rewards/accuracies": 0.6625000238418579,
-      "rewards/chosen": 0.19220882654190063,
-      "rewards/margins": 0.6076450347900391,
-      "rewards/rejected": -0.4154362082481384,
+      "logits/chosen": 82.19850158691406,
+      "logits/rejected": 82.19775390625,
+      "logps/chosen": -26.9793701171875,
+      "logps/rejected": -32.99781036376953,
+      "loss": 0.7152,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": 0.18718235194683075,
+      "rewards/margins": 0.4169556498527527,
+      "rewards/rejected": -0.2297733575105667,
       "step": 140
     },
     {
       "epoch": 0.39,
-      "grad_norm": 54.0,
+      "grad_norm": 38.0,
       "learning_rate": 3.834196265035119e-06,
-      "logits/chosen": 80.97503662109375,
-      "logits/rejected": 80.94606018066406,
-      "logps/chosen": -29.037616729736328,
-      "logps/rejected": -33.28493118286133,
-      "loss": 0.6493,
-      "rewards/accuracies": 0.762499988079071,
-      "rewards/chosen": 0.2460467517375946,
-      "rewards/margins": 0.7489484548568726,
-      "rewards/rejected": -0.5029016733169556,
+      "logits/chosen": 80.65988159179688,
+      "logits/rejected": 80.63349914550781,
+      "logps/chosen": -28.97601890563965,
+      "logps/rejected": -33.183013916015625,
+      "loss": 0.6687,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": 0.16523592174053192,
+      "rewards/margins": 0.4118415415287018,
+      "rewards/rejected": -0.24660558998584747,
       "step": 150
     },
     {
       "epoch": 0.42,
-      "grad_norm": 63.0,
+      "grad_norm": 52.25,
       "learning_rate": 3.636998309800573e-06,
-      "logits/chosen": 82.87214660644531,
-      "logits/rejected": 82.87894439697266,
-      "logps/chosen": -33.497344970703125,
-      "logps/rejected": -30.375295639038086,
-      "loss": 0.7389,
-      "rewards/accuracies": 0.6875,
-      "rewards/chosen": 0.4475820064544678,
-      "rewards/margins": 0.7435113191604614,
-      "rewards/rejected": -0.2959292531013489,
+      "logits/chosen": 82.44218444824219,
+      "logits/rejected": 82.45413970947266,
+      "logps/chosen": -33.828277587890625,
+      "logps/rejected": -30.54928207397461,
+      "loss": 0.7433,
+      "rewards/accuracies": 0.7250000238418579,
+      "rewards/chosen": 0.12338707596063614,
+      "rewards/margins": 0.36208492517471313,
+      "rewards/rejected": -0.238697811961174,
       "step": 160
     },
     {
       "epoch": 0.44,
-      "grad_norm": 63.25,
+      "grad_norm": 40.25,
       "learning_rate": 3.4304331721118078e-06,
-      "logits/chosen": 83.60514831542969,
-      "logits/rejected": 83.55717468261719,
-      "logps/chosen": -30.812519073486328,
-      "logps/rejected": -32.62251281738281,
-      "loss": 0.7275,
+      "logits/chosen": 83.18556213378906,
+      "logits/rejected": 83.12220764160156,
+      "logps/chosen": -30.9090518951416,
+      "logps/rejected": -32.738372802734375,
+      "loss": 0.7036,
       "rewards/accuracies": 0.675000011920929,
-      "rewards/chosen": 0.31273749470710754,
-      "rewards/margins": 0.7344537377357483,
-      "rewards/rejected": -0.42171627283096313,
+      "rewards/chosen": 0.14009535312652588,
+      "rewards/margins": 0.4274210035800934,
+      "rewards/rejected": -0.28732559084892273,
       "step": 170
     },
     {
       "epoch": 0.47,
-      "grad_norm": 54.25,
+      "grad_norm": 28.125,
       "learning_rate": 3.2162026428305436e-06,
-      "logits/chosen": 81.30427551269531,
-      "logits/rejected": 81.28076171875,
-      "logps/chosen": -30.593530654907227,
-      "logps/rejected": -31.66329574584961,
-      "loss": 0.58,
-      "rewards/accuracies": 0.75,
-      "rewards/chosen": 0.4225357472896576,
-      "rewards/margins": 0.7767966985702515,
-      "rewards/rejected": -0.3542609214782715,
+      "logits/chosen": 80.74728393554688,
+      "logits/rejected": 80.72408294677734,
+      "logps/chosen": -30.6306095123291,
+      "logps/rejected": -31.813491821289062,
+      "loss": 0.6778,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": 0.2266170233488083,
+      "rewards/margins": 0.48913127183914185,
+      "rewards/rejected": -0.2625141739845276,
       "step": 180
     },
     {
       "epoch": 0.49,
-      "grad_norm": 37.5,
+      "grad_norm": 23.5,
       "learning_rate": 2.996071664294641e-06,
-      "logits/chosen": 82.88330078125,
-      "logits/rejected": 82.88325500488281,
-      "logps/chosen": -30.3848934173584,
-      "logps/rejected": -30.6480770111084,
-      "loss": 0.8789,
-      "rewards/accuracies": 0.625,
-      "rewards/chosen": 0.2670655846595764,
-      "rewards/margins": 0.39310184121131897,
-      "rewards/rejected": -0.12603625655174255,
+      "logits/chosen": 82.42842102050781,
+      "logits/rejected": 82.41558837890625,
+      "logps/chosen": -30.537174224853516,
+      "logps/rejected": -30.8133602142334,
+      "loss": 0.8564,
+      "rewards/accuracies": 0.48750001192092896,
+      "rewards/chosen": 0.09169743955135345,
+      "rewards/margins": 0.22983141243457794,
+      "rewards/rejected": -0.13813400268554688,
       "step": 190
     },
     {
       "epoch": 0.52,
-      "grad_norm": 56.25,
+      "grad_norm": 33.5,
       "learning_rate": 2.7718537898066833e-06,
-      "logits/chosen": 78.49826049804688,
-      "logits/rejected": 78.43463134765625,
-      "logps/chosen": -33.972293853759766,
-      "logps/rejected": -32.59092330932617,
-      "loss": 0.8398,
-      "rewards/accuracies": 0.574999988079071,
-      "rewards/chosen": 0.5463098287582397,
-      "rewards/margins": 0.6145724058151245,
-      "rewards/rejected": -0.06826266646385193,
+      "logits/chosen": 77.84940338134766,
+      "logits/rejected": 77.80615234375,
+      "logps/chosen": -33.916507720947266,
+      "logps/rejected": -32.88044738769531,
+      "loss": 0.683,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": 0.33449333906173706,
+      "rewards/margins": 0.4893106520175934,
+      "rewards/rejected": -0.1548173427581787,
       "step": 200
     },
     {
       "epoch": 0.52,
-      "eval_logits/chosen": 98.71674346923828,
-      "eval_logits/rejected": 98.69157409667969,
-      "eval_logps/chosen": -32.63681411743164,
-      "eval_logps/rejected": -36.29216003417969,
-      "eval_loss": 1.0089657306671143,
-      "eval_rewards/accuracies": 0.5141196250915527,
-      "eval_rewards/chosen": -0.13554596900939941,
-      "eval_rewards/margins": 0.09239647537469864,
-      "eval_rewards/rejected": -0.22794245183467865,
-      "eval_runtime": 104.0982,
-      "eval_samples_per_second": 3.295,
+      "eval_logits/chosen": 98.61212158203125,
+      "eval_logits/rejected": 98.5889663696289,
+      "eval_logps/chosen": -32.7288703918457,
+      "eval_logps/rejected": -36.450355529785156,
+      "eval_loss": 0.9409860372543335,
+      "eval_rewards/accuracies": 0.5568937063217163,
+      "eval_rewards/chosen": -0.11428017914295197,
+      "eval_rewards/margins": 0.07925106585025787,
+      "eval_rewards/rejected": -0.19353125989437103,
+      "eval_runtime": 104.0333,
+      "eval_samples_per_second": 3.297,
       "eval_steps_per_second": 0.413,
       "step": 200
     },
     {
       "epoch": 0.55,
-      "grad_norm": 73.5,
+      "grad_norm": 59.0,
       "learning_rate": 2.5453962426402006e-06,
-      "logits/chosen": 81.18994140625,
-      "logits/rejected": 81.10676574707031,
-      "logps/chosen": -33.214935302734375,
-      "logps/rejected": -35.187870025634766,
-      "loss": 0.6733,
-      "rewards/accuracies": 0.75,
-      "rewards/chosen": 0.511551022529602,
-      "rewards/margins": 0.6527955532073975,
-      "rewards/rejected": -0.14124450087547302,
+      "logits/chosen": 80.46003723144531,
+      "logits/rejected": 80.36625671386719,
+      "logps/chosen": -33.2611083984375,
+      "logps/rejected": -35.46075439453125,
+      "loss": 0.6831,
+      "rewards/accuracies": 0.7250000238418579,
+      "rewards/chosen": 0.27384519577026367,
+      "rewards/margins": 0.46370941400527954,
+      "rewards/rejected": -0.18986421823501587,
       "step": 210
     },
     {
       "epoch": 0.57,
-      "grad_norm": 53.5,
+      "grad_norm": 45.0,
       "learning_rate": 2.3185646976551794e-06,
-      "logits/chosen": 83.3000259399414,
-      "logits/rejected": 83.38983917236328,
-      "logps/chosen": -31.05356216430664,
-      "logps/rejected": -31.100086212158203,
-      "loss": 0.5573,
-      "rewards/accuracies": 0.737500011920929,
-      "rewards/chosen": 0.6267757415771484,
-      "rewards/margins": 0.8865677714347839,
-      "rewards/rejected": -0.2597920000553131,
+      "logits/chosen": 82.58009338378906,
+      "logits/rejected": 82.66780853271484,
+      "logps/chosen": -31.060461044311523,
+      "logps/rejected": -31.333576202392578,
+      "loss": 0.6185,
+      "rewards/accuracies": 0.7250000238418579,
+      "rewards/chosen": 0.35539740324020386,
+      "rewards/margins": 0.5972464084625244,
+      "rewards/rejected": -0.24184902012348175,
       "step": 220
     },
     {
       "epoch": 0.6,
-      "grad_norm": 59.5,
+      "grad_norm": 46.25,
       "learning_rate": 2.0932279108998323e-06,
-      "logits/chosen": 80.46002197265625,
-      "logits/rejected": 80.5117416381836,
-      "logps/chosen": -32.2435417175293,
-      "logps/rejected": -34.2344970703125,
-      "loss": 0.7635,
-      "rewards/accuracies": 0.6625000238418579,
-      "rewards/chosen": 0.393540620803833,
-      "rewards/margins": 0.5491863489151001,
-      "rewards/rejected": -0.15564575791358948,
+      "logits/chosen": 79.74901580810547,
+      "logits/rejected": 79.80045318603516,
+      "logps/chosen": -32.357086181640625,
+      "logps/rejected": -34.57195281982422,
+      "loss": 0.7442,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.17946358025074005,
+      "rewards/margins": 0.403384268283844,
+      "rewards/rejected": -0.22392070293426514,
       "step": 230
     },
     {
       "epoch": 0.62,
-      "grad_norm": 72.0,
+      "grad_norm": 47.75,
       "learning_rate": 1.8712423238279358e-06,
-      "logits/chosen": 82.81127166748047,
-      "logits/rejected": 83.10184478759766,
-      "logps/chosen": -30.6541748046875,
-      "logps/rejected": -31.797021865844727,
-      "loss": 0.5215,
-      "rewards/accuracies": 0.737500011920929,
-      "rewards/chosen": 0.7237205505371094,
-      "rewards/margins": 0.9092057347297668,
-      "rewards/rejected": -0.18548506498336792,
+      "logits/chosen": 82.10381317138672,
+      "logits/rejected": 82.40473937988281,
+      "logps/chosen": -30.87957763671875,
+      "logps/rejected": -32.0345344543457,
+      "loss": 0.5885,
+      "rewards/accuracies": 0.7749999761581421,
+      "rewards/chosen": 0.3233951926231384,
+      "rewards/margins": 0.5243898630142212,
+      "rewards/rejected": -0.20099470019340515,
       "step": 240
     },
     {
       "epoch": 0.65,
-      "grad_norm": 76.5,
+      "grad_norm": 39.75,
       "learning_rate": 1.6544367689701824e-06,
-      "logits/chosen": 81.55049133300781,
-      "logits/rejected": 81.6218032836914,
-      "logps/chosen": -26.869796752929688,
-      "logps/rejected": -30.16421890258789,
-      "loss": 0.7649,
-      "rewards/accuracies": 0.6499999761581421,
-      "rewards/chosen": 0.5082693696022034,
-      "rewards/margins": 0.6927058696746826,
-      "rewards/rejected": -0.18443644046783447,
+      "logits/chosen": 80.81050109863281,
+      "logits/rejected": 80.87081909179688,
+      "logps/chosen": -27.13393783569336,
+      "logps/rejected": -30.56355857849121,
+      "loss": 0.6984,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": 0.18478286266326904,
+      "rewards/margins": 0.44991129636764526,
+      "rewards/rejected": -0.2651284337043762,
       "step": 250
     },
     {
       "epoch": 0.68,
-      "grad_norm": 52.5,
+      "grad_norm": 30.25,
       "learning_rate": 1.4445974030621963e-06,
-      "logits/chosen": 78.84456634521484,
-      "logits/rejected": 78.9933090209961,
-      "logps/chosen": -30.304040908813477,
-      "logps/rejected": -36.30915069580078,
-      "loss": 0.4765,
-      "rewards/accuracies": 0.7875000238418579,
-      "rewards/chosen": 0.8798693418502808,
-      "rewards/margins": 1.0251331329345703,
-      "rewards/rejected": -0.1452637016773224,
+      "logits/chosen": 78.12925720214844,
+      "logits/rejected": 78.25952911376953,
+      "logps/chosen": -30.566410064697266,
+      "logps/rejected": -36.688663482666016,
+      "loss": 0.5996,
+      "rewards/accuracies": 0.7250000238418579,
+      "rewards/chosen": 0.3978345990180969,
+      "rewards/margins": 0.6326489448547363,
+      "rewards/rejected": -0.2348143756389618,
       "step": 260
     },
     {
       "epoch": 0.7,
-      "grad_norm": 57.0,
+      "grad_norm": 42.25,
       "learning_rate": 1.243452991757889e-06,
-      "logits/chosen": 78.18994903564453,
-      "logits/rejected": 78.21639251708984,
-      "logps/chosen": -30.825641632080078,
-      "logps/rejected": -31.800342559814453,
-      "loss": 0.6623,
-      "rewards/accuracies": 0.7250000238418579,
-      "rewards/chosen": 0.6672974228858948,
-      "rewards/margins": 0.8737428784370422,
-      "rewards/rejected": -0.20644548535346985,
+      "logits/chosen": 77.4239501953125,
+      "logits/rejected": 77.45281219482422,
+      "logps/chosen": -30.834758758544922,
+      "logps/rejected": -31.846826553344727,
+      "loss": 0.626,
+      "rewards/accuracies": 0.824999988079071,
+      "rewards/chosen": 0.37766537070274353,
+      "rewards/margins": 0.5142279863357544,
+      "rewards/rejected": -0.13656257092952728,
       "step": 270
     },
     {
       "epoch": 0.73,
-      "grad_norm": 76.0,
+      "grad_norm": 54.0,
       "learning_rate": 1.0526606671603523e-06,
-      "logits/chosen": 80.78196716308594,
-      "logits/rejected": 80.56169128417969,
-      "logps/chosen": -30.946773529052734,
-      "logps/rejected": -29.854522705078125,
-      "loss": 0.6812,
-      "rewards/accuracies": 0.6875,
-      "rewards/chosen": 0.6041832566261292,
-      "rewards/margins": 0.7354522943496704,
-      "rewards/rejected": -0.13126906752586365,
+      "logits/chosen": 80.0883560180664,
+      "logits/rejected": 79.87921905517578,
+      "logps/chosen": -31.279199600219727,
+      "logps/rejected": -29.9936580657959,
+      "loss": 0.7705,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": 0.2122771292924881,
+      "rewards/margins": 0.3429415822029114,
+      "rewards/rejected": -0.13066445291042328,
       "step": 280
     },
     {
       "epoch": 0.75,
-      "grad_norm": 60.25,
+      "grad_norm": 40.5,
       "learning_rate": 8.737922755071455e-07,
-      "logits/chosen": 80.99944305419922,
-      "logits/rejected": 80.91734313964844,
-      "logps/chosen": -32.94629669189453,
-      "logps/rejected": -32.49176025390625,
-      "loss": 0.5405,
-      "rewards/accuracies": 0.7250000238418579,
-      "rewards/chosen": 0.8105659484863281,
-      "rewards/margins": 1.1429924964904785,
-      "rewards/rejected": -0.3324265480041504,
+      "logits/chosen": 80.207763671875,
+      "logits/rejected": 80.12342834472656,
+      "logps/chosen": -33.07125473022461,
+      "logps/rejected": -32.81315612792969,
+      "loss": 0.5368,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": 0.41319671273231506,
+      "rewards/margins": 0.7317113280296326,
+      "rewards/rejected": -0.3185145854949951,
       "step": 290
     },
     {
       "epoch": 0.78,
-      "grad_norm": 72.5,
+      "grad_norm": 43.0,
       "learning_rate": 7.08321427484816e-07,
-      "logits/chosen": 76.61840057373047,
-      "logits/rejected": 76.69322204589844,
-      "logps/chosen": -32.09418487548828,
-      "logps/rejected": -29.156299591064453,
-      "loss": 0.733,
-      "rewards/accuracies": 0.675000011920929,
-      "rewards/chosen": 0.9180667996406555,
-      "rewards/margins": 0.9560586214065552,
-      "rewards/rejected": -0.03799174353480339,
+      "logits/chosen": 75.83465576171875,
+      "logits/rejected": 75.93038177490234,
+      "logps/chosen": -32.472633361816406,
+      "logps/rejected": -29.376873016357422,
+      "loss": 0.7349,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": 0.37323057651519775,
+      "rewards/margins": 0.48316845297813416,
+      "rewards/rejected": -0.109937883913517,
       "step": 300
     },
     {
       "epoch": 0.78,
-      "eval_logits/chosen": 98.73394012451172,
-      "eval_logits/rejected": 98.70979309082031,
-      "eval_logps/chosen": -32.53404235839844,
-      "eval_logps/rejected": -36.16787338256836,
-      "eval_loss": 1.0123846530914307,
-      "eval_rewards/accuracies": 0.5398671627044678,
-      "eval_rewards/chosen": -0.06360965222120285,
-      "eval_rewards/margins": 0.07733342051506042,
-      "eval_rewards/rejected": -0.14094306528568268,
-      "eval_runtime": 104.0616,
-      "eval_samples_per_second": 3.296,
-      "eval_steps_per_second": 0.413,
+      "eval_logits/chosen": 98.62049102783203,
+      "eval_logits/rejected": 98.59721374511719,
+      "eval_logps/chosen": -32.674007415771484,
+      "eval_logps/rejected": -36.29110336303711,
+      "eval_loss": 0.9808743000030518,
+      "eval_rewards/accuracies": 0.5157807469367981,
+      "eval_rewards/chosen": -0.0923333466053009,
+      "eval_rewards/margins": 0.03749755397439003,
+      "eval_rewards/rejected": -0.12983089685440063,
+      "eval_runtime": 103.8205,
+      "eval_samples_per_second": 3.304,
+      "eval_steps_per_second": 0.414,
       "step": 300
     },
     {
       "epoch": 0.81,
-      "grad_norm": 63.0,
+      "grad_norm": 40.0,
       "learning_rate": 5.576113578589035e-07,
-      "logits/chosen": 83.6187515258789,
-      "logits/rejected": 83.65609741210938,
-      "logps/chosen": -30.124670028686523,
-      "logps/rejected": -32.455326080322266,
-      "loss": 0.6913,
-      "rewards/accuracies": 0.7124999761581421,
-      "rewards/chosen": 0.5172038078308105,
-      "rewards/margins": 0.6960991024971008,
-      "rewards/rejected": -0.1788952797651291,
+      "logits/chosen": 82.97636413574219,
+      "logits/rejected": 83.00100708007812,
+      "logps/chosen": -30.166763305664062,
+      "logps/rejected": -32.79009246826172,
+      "loss": 0.6505,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": 0.278707891702652,
+      "rewards/margins": 0.5148409008979797,
+      "rewards/rejected": -0.23613300919532776,
       "step": 310
     },
     {
       "epoch": 0.83,
-      "grad_norm": 53.5,
+      "grad_norm": 32.5,
       "learning_rate": 4.229036944380913e-07,
-      "logits/chosen": 81.21861267089844,
-      "logits/rejected": 81.21974182128906,
-      "logps/chosen": -30.413522720336914,
-      "logps/rejected": -29.050277709960938,
-      "loss": 0.5746,
-      "rewards/accuracies": 0.699999988079071,
-      "rewards/chosen": 0.8295270204544067,
-      "rewards/margins": 0.921076774597168,
-      "rewards/rejected": -0.09154972434043884,
+      "logits/chosen": 80.45620727539062,
+      "logits/rejected": 80.45777893066406,
+      "logps/chosen": -30.465673446655273,
+      "logps/rejected": -29.35947036743164,
+      "loss": 0.5314,
+      "rewards/accuracies": 0.762499988079071,
+      "rewards/chosen": 0.45315417647361755,
+      "rewards/margins": 0.6291457414627075,
+      "rewards/rejected": -0.17599165439605713,
       "step": 320
     },
     {
       "epoch": 0.86,
-      "grad_norm": 42.5,
+      "grad_norm": 34.0,
       "learning_rate": 3.053082288996112e-07,
-      "logits/chosen": 78.42388153076172,
-      "logits/rejected": 78.47276306152344,
-      "logps/chosen": -28.912973403930664,
-      "logps/rejected": -32.85163116455078,
-      "loss": 0.5118,
-      "rewards/accuracies": 0.762499988079071,
-      "rewards/chosen": 1.007744550704956,
-      "rewards/margins": 1.1594822406768799,
-      "rewards/rejected": -0.151737779378891,
+      "logits/chosen": 77.60248565673828,
+      "logits/rejected": 77.65824890136719,
+      "logps/chosen": -29.066116333007812,
+      "logps/rejected": -33.186683654785156,
+      "loss": 0.4902,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": 0.5145958065986633,
+      "rewards/margins": 0.735322117805481,
+      "rewards/rejected": -0.2207263708114624,
       "step": 330
     },
     {
       "epoch": 0.88,
-      "grad_norm": 78.0,
+      "grad_norm": 46.75,
       "learning_rate": 2.0579377374915805e-07,
-      "logits/chosen": 82.58137512207031,
-      "logits/rejected": 82.63166046142578,
-      "logps/chosen": -32.287044525146484,
-      "logps/rejected": -33.664268493652344,
-      "loss": 0.7276,
-      "rewards/accuracies": 0.675000011920929,
-      "rewards/chosen": 0.6475721597671509,
-      "rewards/margins": 0.8544757962226868,
-      "rewards/rejected": -0.2069036066532135,
+      "logits/chosen": 81.93863677978516,
+      "logits/rejected": 81.96864318847656,
+      "logps/chosen": -32.32833480834961,
+      "logps/rejected": -33.94022750854492,
+      "loss": 0.6732,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": 0.35352635383605957,
+      "rewards/margins": 0.5821414589881897,
+      "rewards/rejected": -0.22861509025096893,
       "step": 340
     },
     {
       "epoch": 0.91,
-      "grad_norm": 56.75,
+      "grad_norm": 31.875,
       "learning_rate": 1.2518018074041684e-07,
-      "logits/chosen": 81.68583679199219,
-      "logits/rejected": 81.69776916503906,
-      "logps/chosen": -32.62708282470703,
-      "logps/rejected": -33.35443115234375,
-      "loss": 0.626,
+      "logits/chosen": 80.9842529296875,
+      "logits/rejected": 80.99238586425781,
+      "logps/chosen": -32.631919860839844,
+      "logps/rejected": -33.691749572753906,
+      "loss": 0.5834,
       "rewards/accuracies": 0.7250000238418579,
-      "rewards/chosen": 0.7116604447364807,
-      "rewards/margins": 0.8955025672912598,
-      "rewards/rejected": -0.18384216725826263,
+      "rewards/chosen": 0.4047267436981201,
+      "rewards/margins": 0.6447058320045471,
+      "rewards/rejected": -0.23997914791107178,
       "step": 350
     },
     {
       "epoch": 0.94,
-      "grad_norm": 45.5,
+      "grad_norm": 29.125,
       "learning_rate": 6.41315865106129e-08,
-      "logits/chosen": 83.15383911132812,
-      "logits/rejected": 83.18663024902344,
-      "logps/chosen": -28.314571380615234,
-      "logps/rejected": -31.6760196685791,
-      "loss": 0.6231,
-      "rewards/accuracies": 0.7124999761581421,
-      "rewards/chosen": 0.8663473129272461,
-      "rewards/margins": 0.8846683502197266,
-      "rewards/rejected": -0.01832098886370659,
+      "logits/chosen": 82.46900177001953,
+      "logits/rejected": 82.49642944335938,
+      "logps/chosen": -28.48472023010254,
+      "logps/rejected": -31.835073471069336,
+      "loss": 0.6674,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.42699599266052246,
+      "rewards/margins": 0.5010865926742554,
+      "rewards/rejected": -0.0740906298160553,
       "step": 360
     },
     {
       "epoch": 0.96,
-      "grad_norm": 76.0,
+      "grad_norm": 36.0,
       "learning_rate": 2.3150941078050325e-08,
-      "logits/chosen": 82.54973602294922,
-      "logits/rejected": 82.57438659667969,
-      "logps/chosen": -31.776432037353516,
-      "logps/rejected": -35.274253845214844,
-      "loss": 0.6836,
+      "logits/chosen": 81.88428497314453,
+      "logits/rejected": 81.90206909179688,
+      "logps/chosen": -31.865198135375977,
+      "logps/rejected": -35.7531623840332,
+      "loss": 0.5901,
       "rewards/accuracies": 0.7124999761581421,
-      "rewards/chosen": 0.7698130011558533,
-      "rewards/margins": 0.8852267265319824,
-      "rewards/rejected": -0.11541371047496796,
+      "rewards/chosen": 0.40438637137413025,
+      "rewards/margins": 0.6619016528129578,
+      "rewards/rejected": -0.2575153410434723,
       "step": 370
     },
     {
       "epoch": 0.99,
-      "grad_norm": 72.0,
+      "grad_norm": 36.0,
       "learning_rate": 2.575864278703266e-09,
-      "logits/chosen": 76.53568267822266,
-      "logits/rejected": 76.40200805664062,
-      "logps/chosen": -29.737590789794922,
-      "logps/rejected": -28.21639060974121,
-      "loss": 0.7558,
-      "rewards/accuracies": 0.6625000238418579,
-      "rewards/chosen": 0.5869752168655396,
-      "rewards/margins": 0.6260467767715454,
-      "rewards/rejected": -0.03907149285078049,
+      "logits/chosen": 75.7695083618164,
+      "logits/rejected": 75.64456939697266,
+      "logps/chosen": -29.696298599243164,
+      "logps/rejected": -28.395023345947266,
+      "loss": 0.6836,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": 0.35193198919296265,
+      "rewards/margins": 0.4457107484340668,
+      "rewards/rejected": -0.09377876669168472,
       "step": 380
     },
     {
       "epoch": 1.0,
       "step": 385,
       "total_flos": 0.0,
-      "train_loss": 0.7429099231571347,
-      "train_runtime": 2555.3187,
-      "train_samples_per_second": 1.205,
+      "train_loss": 0.741242148659446,
+      "train_runtime": 2553.0854,
+      "train_samples_per_second": 1.206,
       "train_steps_per_second": 0.151
     }
   ],