Model save

README.md

@@ -1,13 +1,10 @@
 ---
 library_name: peft
 tags:
-- alignment-handbook
 - trl
 - dpo
 - generated_from_trainer
 base_model: NorLLM-AI/NorMistral-7B
-datasets:
-- hugodk-sch/aftonposten_title_prefs
 model-index:
 - name: norllm-ai-normistral-7b-align-scan
   results: []
@@ -18,17 +15,17 @@ should probably proofread and complete it, then remove this comment. -->
 
 # norllm-ai-normistral-7b-align-scan
 
-This model is a fine-tuned version of [data/norllm-ai-normistral-7b-sft-qlora](https://huggingface.co/data/norllm-ai-normistral-7b-sft-qlora) on the hugodk-sch/aftonposten_title_prefs dataset.
+This model is a fine-tuned version of [NorLLM-AI/NorMistral-7B](https://huggingface.co/NorLLM-AI/NorMistral-7B) on the None dataset.
 It achieves the following results on the evaluation set:
-- Loss: 0.8988
-- Rewards/chosen: -0.0761
-- Rewards/rejected: -0.1908
-- Rewards/accuracies: 0.5772
-- Rewards/margins: 0.1147
-- Logps/rejected: -34.9707
-- Logps/chosen: -31.3912
-- Logits/rejected: -2.8188
-- Logits/chosen: -2.8213
+- Loss: 0.9202
+- Rewards/chosen: -0.0777
+- Rewards/rejected: -0.1629
+- Rewards/accuracies: 0.5656
+- Rewards/margins: 0.0852
+- Logps/rejected: -35.1054
+- Logps/chosen: -31.4766
+- Logits/rejected: -2.8221
+- Logits/chosen: -2.8242
 
 ## Model description
 
@@ -63,9 +60,9 @@ 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 |
 |:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
-| 0.8179        | 0.26  | 100  | 0.8995          | -0.0056        | -0.1202          | 0.6100             | 0.1146          | -34.8698       | -31.2905     | -2.8112         | -2.8139       |
-| 0.7257        | 0.52  | 200  | 0.9028          | -0.0688        | -0.1792          | 0.5573             | 0.1104          | -34.9542       | -31.3808     | -2.8244         | -2.8270       |
-| 0.6308        | 0.78  | 300  | 0.9012          | -0.0642        | -0.1770          | 0.5656             | 0.1128          | -34.9511       | -31.3742     | -2.8184         | -2.8208       |
+| 0.8436        | 0.26  | 100  | 0.9278          | -0.0021        | -0.0777          | 0.5984             | 0.0756          | -34.8925       | -31.2877     | -2.8097         | -2.8124       |
+| 0.7459        | 0.52  | 200  | 0.9159          | -0.0693        | -0.1587          | 0.5835             | 0.0894          | -35.0949       | -31.4556     | -2.8227         | -2.8250       |
+| 0.665         | 0.78  | 300  | 0.9202          | -0.0777        | -0.1629          | 0.5656             | 0.0852          | -35.1054       | -31.4766     | -2.8221         | -2.8242       |
 
 
 ### Framework versions

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:da97bf572af082befc70fa537e18d89feb8e151575a48d8688dbd5c101804305
+oid sha256:d42413d859fecd7330e3ba97a9779744e69562a35713be675f15c1e7c0893221
 size 671150064

all_results.json

@@ -1,21 +1,8 @@
 {
     "epoch": 1.0,
-    "eval_logits/chosen": -2.821291208267212,
-    "eval_logits/rejected": -2.818777084350586,
-    "eval_logps/chosen": -31.391199111938477,
-    "eval_logps/rejected": -34.970741271972656,
-    "eval_loss": 0.8988423347473145,
-    "eval_rewards/accuracies": 0.5772424936294556,
-    "eval_rewards/chosen": -0.07612106204032898,
-    "eval_rewards/margins": 0.11468725651502609,
-    "eval_rewards/rejected": -0.19080831110477448,
-    "eval_runtime": 112.7622,
-    "eval_samples": 343,
-    "eval_samples_per_second": 3.042,
-    "eval_steps_per_second": 0.381,
-    "train_loss": 0.6684874992866021,
-    "train_runtime": 2718.5428,
+    "train_loss": 0.7203964704042906,
+    "train_runtime": 2720.5927,
     "train_samples": 3079,
-    "train_samples_per_second": 1.133,
+    "train_samples_per_second": 1.132,
     "train_steps_per_second": 0.142
 }

train_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 1.0,
-    "train_loss": 0.6684874992866021,
-    "train_runtime": 2718.5428,
+    "train_loss": 0.7203964704042906,
+    "train_runtime": 2720.5927,
     "train_samples": 3079,
-    "train_samples_per_second": 1.133,
+    "train_samples_per_second": 1.132,
     "train_steps_per_second": 0.142
 }

trainer_state.json

@@ -10,7 +10,7 @@
   "log_history": [
     {
       "epoch": 0.0,
-      "grad_norm": 55.25,
+      "grad_norm": 31.625,
       "learning_rate": 1.282051282051282e-07,
       "logits/chosen": -2.7358343601226807,
       "logits/rejected": -2.7480404376983643,
@@ -25,629 +25,629 @@
     },
     {
       "epoch": 0.03,
-      "grad_norm": 94.0,
+      "grad_norm": 53.5,
       "learning_rate": 1.282051282051282e-06,
-      "logits/chosen": -3.009542942047119,
-      "logits/rejected": -2.998302698135376,
-      "logps/chosen": -33.17807388305664,
-      "logps/rejected": -31.984642028808594,
-      "loss": 0.9835,
-      "rewards/accuracies": 0.5,
-      "rewards/chosen": 0.002009392250329256,
-      "rewards/margins": 0.016525447368621826,
-      "rewards/rejected": -0.014516051858663559,
+      "logits/chosen": -3.0091097354888916,
+      "logits/rejected": -2.998112678527832,
+      "logps/chosen": -33.19804763793945,
+      "logps/rejected": -31.97610092163086,
+      "loss": 1.002,
+      "rewards/accuracies": 0.4722222089767456,
+      "rewards/chosen": -0.006841929163783789,
+      "rewards/margins": -0.0019655984360724688,
+      "rewards/rejected": -0.004876333754509687,
       "step": 10
     },
     {
       "epoch": 0.05,
-      "grad_norm": 64.5,
+      "grad_norm": 36.75,
       "learning_rate": 2.564102564102564e-06,
-      "logits/chosen": -2.899103879928589,
-      "logits/rejected": -2.8936753273010254,
-      "logps/chosen": -32.4957160949707,
-      "logps/rejected": -28.984477996826172,
-      "loss": 1.0126,
-      "rewards/accuracies": 0.44999998807907104,
-      "rewards/chosen": -0.014431963674724102,
-      "rewards/margins": -0.012606342323124409,
-      "rewards/rejected": -0.0018256217008456588,
+      "logits/chosen": -2.899024248123169,
+      "logits/rejected": -2.8939948081970215,
+      "logps/chosen": -32.47606658935547,
+      "logps/rejected": -28.96236228942871,
+      "loss": 1.0082,
+      "rewards/accuracies": 0.42500001192092896,
+      "rewards/chosen": -0.0003868088242597878,
+      "rewards/margins": -0.008188480511307716,
+      "rewards/rejected": 0.007801672909408808,
       "step": 20
     },
     {
       "epoch": 0.08,
-      "grad_norm": 62.0,
+      "grad_norm": 35.25,
       "learning_rate": 3.846153846153847e-06,
-      "logits/chosen": -3.0970749855041504,
-      "logits/rejected": -3.1094558238983154,
-      "logps/chosen": -32.78588104248047,
-      "logps/rejected": -30.1519775390625,
-      "loss": 0.9449,
-      "rewards/accuracies": 0.574999988079071,
-      "rewards/chosen": 0.08596263825893402,
-      "rewards/margins": 0.05510985851287842,
-      "rewards/rejected": 0.030852774158120155,
+      "logits/chosen": -3.097607135772705,
+      "logits/rejected": -3.109447956085205,
+      "logps/chosen": -32.788970947265625,
+      "logps/rejected": -30.134746551513672,
+      "loss": 0.9766,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": 0.047886691987514496,
+      "rewards/margins": 0.023363908752799034,
+      "rewards/rejected": 0.02452278696000576,
       "step": 30
     },
     {
       "epoch": 0.1,
-      "grad_norm": 50.5,
+      "grad_norm": 33.0,
       "learning_rate": 4.999896948438434e-06,
-      "logits/chosen": -2.862823486328125,
-      "logits/rejected": -2.853950023651123,
-      "logps/chosen": -31.538238525390625,
-      "logps/rejected": -32.37445068359375,
-      "loss": 0.8574,
-      "rewards/accuracies": 0.612500011920929,
-      "rewards/chosen": 0.20445938408374786,
-      "rewards/margins": 0.21635785698890686,
-      "rewards/rejected": -0.011898472905158997,
+      "logits/chosen": -2.8632471561431885,
+      "logits/rejected": -2.853929042816162,
+      "logps/chosen": -31.501651763916016,
+      "logps/rejected": -32.37067413330078,
+      "loss": 0.8747,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.1314682960510254,
+      "rewards/margins": 0.13675658404827118,
+      "rewards/rejected": -0.005288311745971441,
       "step": 40
     },
     {
       "epoch": 0.13,
-      "grad_norm": 39.0,
+      "grad_norm": 22.5,
       "learning_rate": 4.987541037542187e-06,
-      "logits/chosen": -2.881490707397461,
-      "logits/rejected": -2.8795459270477295,
-      "logps/chosen": -29.380319595336914,
-      "logps/rejected": -30.1256046295166,
-      "loss": 0.8579,
-      "rewards/accuracies": 0.6000000238418579,
-      "rewards/chosen": 0.2518887519836426,
-      "rewards/margins": 0.2775513231754303,
-      "rewards/rejected": -0.025662561878561974,
+      "logits/chosen": -2.881110191345215,
+      "logits/rejected": -2.878807783126831,
+      "logps/chosen": -29.374563217163086,
+      "logps/rejected": -30.105758666992188,
+      "loss": 0.886,
+      "rewards/accuracies": 0.512499988079071,
+      "rewards/chosen": 0.1462380588054657,
+      "rewards/margins": 0.15296456217765808,
+      "rewards/rejected": -0.006726491264998913,
       "step": 50
     },
     {
       "epoch": 0.16,
-      "grad_norm": 55.75,
+      "grad_norm": 32.75,
       "learning_rate": 4.954691471941119e-06,
-      "logits/chosen": -2.9107794761657715,
-      "logits/rejected": -2.912201166152954,
-      "logps/chosen": -29.885101318359375,
-      "logps/rejected": -28.084728240966797,
-      "loss": 0.804,
-      "rewards/accuracies": 0.625,
-      "rewards/chosen": 0.17126679420471191,
-      "rewards/margins": 0.26159828901290894,
-      "rewards/rejected": -0.0903315395116806,
+      "logits/chosen": -2.9093446731567383,
+      "logits/rejected": -2.9112706184387207,
+      "logps/chosen": -29.828176498413086,
+      "logps/rejected": -28.021154403686523,
+      "loss": 0.8701,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": 0.12063708156347275,
+      "rewards/margins": 0.14682583510875702,
+      "rewards/rejected": -0.026188766583800316,
       "step": 60
     },
     {
       "epoch": 0.18,
-      "grad_norm": 72.0,
+      "grad_norm": 43.5,
       "learning_rate": 4.901618883413549e-06,
-      "logits/chosen": -2.9932219982147217,
-      "logits/rejected": -2.999392032623291,
-      "logps/chosen": -29.26175308227539,
-      "logps/rejected": -31.045358657836914,
-      "loss": 0.8981,
+      "logits/chosen": -2.9932944774627686,
+      "logits/rejected": -2.999823808670044,
+      "logps/chosen": -29.177806854248047,
+      "logps/rejected": -30.96217918395996,
+      "loss": 0.9306,
       "rewards/accuracies": 0.612500011920929,
-      "rewards/chosen": 0.013451325707137585,
-      "rewards/margins": 0.12094201892614365,
-      "rewards/rejected": -0.10749070346355438,
+      "rewards/chosen": 0.041265495121479034,
+      "rewards/margins": 0.06941740214824677,
+      "rewards/rejected": -0.028151914477348328,
       "step": 70
     },
     {
       "epoch": 0.21,
-      "grad_norm": 61.75,
+      "grad_norm": 37.75,
       "learning_rate": 4.828760511501322e-06,
-      "logits/chosen": -2.808314561843872,
-      "logits/rejected": -2.8239831924438477,
-      "logps/chosen": -29.382043838500977,
-      "logps/rejected": -29.915264129638672,
-      "loss": 0.8143,
-      "rewards/accuracies": 0.6625000238418579,
-      "rewards/chosen": 0.11108281463384628,
-      "rewards/margins": 0.27977070212364197,
-      "rewards/rejected": -0.1686878651380539,
+      "logits/chosen": -2.8079781532287598,
+      "logits/rejected": -2.8230159282684326,
+      "logps/chosen": -29.30885887145996,
+      "logps/rejected": -29.878082275390625,
+      "loss": 0.8332,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.09274748712778091,
+      "rewards/margins": 0.1742672175168991,
+      "rewards/rejected": -0.08151975274085999,
       "step": 80
     },
     {
       "epoch": 0.23,
-      "grad_norm": 48.75,
+      "grad_norm": 28.125,
       "learning_rate": 4.7367166013034295e-06,
-      "logits/chosen": -2.898421049118042,
-      "logits/rejected": -2.8805272579193115,
-      "logps/chosen": -32.720703125,
-      "logps/rejected": -30.371923446655273,
-      "loss": 0.8446,
-      "rewards/accuracies": 0.6625000238418579,
-      "rewards/chosen": 0.08101295679807663,
-      "rewards/margins": 0.3262065052986145,
-      "rewards/rejected": -0.2451934814453125,
+      "logits/chosen": -2.897887945175171,
+      "logits/rejected": -2.8805594444274902,
+      "logps/chosen": -32.75906753540039,
+      "logps/rejected": -30.377796173095703,
+      "loss": 0.8621,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.030946413055062294,
+      "rewards/margins": 0.17340537905693054,
+      "rewards/rejected": -0.1424589455127716,
       "step": 90
     },
     {
       "epoch": 0.26,
-      "grad_norm": 48.25,
+      "grad_norm": 27.625,
       "learning_rate": 4.626245458345211e-06,
-      "logits/chosen": -3.005709409713745,
-      "logits/rejected": -3.0060715675354004,
-      "logps/chosen": -31.795307159423828,
-      "logps/rejected": -30.9265193939209,
-      "loss": 0.8179,
-      "rewards/accuracies": 0.612500011920929,
-      "rewards/chosen": 0.18309839069843292,
-      "rewards/margins": 0.26034125685691833,
-      "rewards/rejected": -0.07724283635616302,
+      "logits/chosen": -3.0043857097625732,
+      "logits/rejected": -3.005366086959839,
+      "logps/chosen": -31.849411010742188,
+      "logps/rejected": -31.039392471313477,
+      "loss": 0.8436,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.0829862430691719,
+      "rewards/margins": 0.1722743958234787,
+      "rewards/rejected": -0.08928816020488739,
       "step": 100
     },
     {
       "epoch": 0.26,
-      "eval_logits/chosen": -2.813856840133667,
-      "eval_logits/rejected": -2.8111777305603027,
-      "eval_logps/chosen": -31.290462493896484,
-      "eval_logps/rejected": -34.86981201171875,
-      "eval_loss": 0.8994626402854919,
-      "eval_rewards/accuracies": 0.610049843788147,
-      "eval_rewards/chosen": -0.005606824532151222,
-      "eval_rewards/margins": 0.11455333232879639,
-      "eval_rewards/rejected": -0.12016014754772186,
-      "eval_runtime": 113.4039,
+      "eval_logits/chosen": -2.812446117401123,
+      "eval_logits/rejected": -2.8096587657928467,
+      "eval_logps/chosen": -31.28770637512207,
+      "eval_logps/rejected": -34.89251708984375,
+      "eval_loss": 0.9278174042701721,
+      "eval_rewards/accuracies": 0.5984219312667847,
+      "eval_rewards/chosen": -0.0021020672284066677,
+      "eval_rewards/margins": 0.07564153522253036,
+      "eval_rewards/rejected": -0.07774360477924347,
+      "eval_runtime": 113.4025,
       "eval_samples_per_second": 3.025,
       "eval_steps_per_second": 0.379,
       "step": 100
     },
     {
       "epoch": 0.29,
-      "grad_norm": 73.0,
+      "grad_norm": 42.0,
       "learning_rate": 4.498257201263691e-06,
-      "logits/chosen": -2.9637725353240967,
-      "logits/rejected": -2.9400086402893066,
-      "logps/chosen": -31.802236557006836,
-      "logps/rejected": -31.42850685119629,
-      "loss": 0.7127,
-      "rewards/accuracies": 0.737500011920929,
-      "rewards/chosen": 0.29019027948379517,
-      "rewards/margins": 0.47895127534866333,
-      "rewards/rejected": -0.18876102566719055,
+      "logits/chosen": -2.961785078048706,
+      "logits/rejected": -2.9373226165771484,
+      "logps/chosen": -31.922677993774414,
+      "logps/rejected": -31.55402183532715,
+      "loss": 0.7753,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": 0.11764738708734512,
+      "rewards/margins": 0.2757166028022766,
+      "rewards/rejected": -0.1580692082643509,
       "step": 110
     },
     {
       "epoch": 0.31,
-      "grad_norm": 57.25,
+      "grad_norm": 35.5,
       "learning_rate": 4.353806263777678e-06,
-      "logits/chosen": -3.0474579334259033,
-      "logits/rejected": -3.076024055480957,
-      "logps/chosen": -28.733871459960938,
-      "logps/rejected": -34.316707611083984,
-      "loss": 0.7493,
-      "rewards/accuracies": 0.637499988079071,
-      "rewards/chosen": 0.3010229170322418,
-      "rewards/margins": 0.41943326592445374,
-      "rewards/rejected": -0.11841034889221191,
+      "logits/chosen": -3.0439839363098145,
+      "logits/rejected": -3.073005199432373,
+      "logps/chosen": -28.63894271850586,
+      "logps/rejected": -34.55387878417969,
+      "loss": 0.7242,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": 0.20998422801494598,
+      "rewards/margins": 0.372516393661499,
+      "rewards/rejected": -0.16253218054771423,
       "step": 120
     },
     {
       "epoch": 0.34,
-      "grad_norm": 46.25,
+      "grad_norm": 33.75,
       "learning_rate": 4.1940827077152755e-06,
-      "logits/chosen": -2.7504048347473145,
-      "logits/rejected": -2.7461349964141846,
-      "logps/chosen": -28.568979263305664,
-      "logps/rejected": -30.313528060913086,
-      "loss": 0.7475,
-      "rewards/accuracies": 0.637499988079071,
-      "rewards/chosen": 0.28399184346199036,
-      "rewards/margins": 0.4097979664802551,
-      "rewards/rejected": -0.12580609321594238,
+      "logits/chosen": -2.7482008934020996,
+      "logits/rejected": -2.7447853088378906,
+      "logps/chosen": -28.648107528686523,
+      "logps/rejected": -30.5201358795166,
+      "loss": 0.7756,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.13062894344329834,
+      "rewards/margins": 0.2851604223251343,
+      "rewards/rejected": -0.15453146398067474,
       "step": 130
     },
     {
       "epoch": 0.36,
-      "grad_norm": 40.75,
+      "grad_norm": 26.25,
       "learning_rate": 4.0204024186666215e-06,
-      "logits/chosen": -3.0258471965789795,
-      "logits/rejected": -3.022930860519409,
-      "logps/chosen": -27.234683990478516,
-      "logps/rejected": -31.9457950592041,
-      "loss": 0.7968,
-      "rewards/accuracies": 0.6000000238418579,
-      "rewards/chosen": 0.17608027160167694,
-      "rewards/margins": 0.3783981502056122,
-      "rewards/rejected": -0.20231786370277405,
+      "logits/chosen": -3.0264618396759033,
+      "logits/rejected": -3.023954391479492,
+      "logps/chosen": -27.322336196899414,
+      "logps/rejected": -32.06721496582031,
+      "loss": 0.8404,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": 0.06555580347776413,
+      "rewards/margins": 0.22973506152629852,
+      "rewards/rejected": -0.1641792207956314,
       "step": 140
     },
     {
       "epoch": 0.39,
-      "grad_norm": 47.25,
+      "grad_norm": 27.625,
       "learning_rate": 3.834196265035119e-06,
-      "logits/chosen": -2.823291063308716,
-      "logits/rejected": -2.818061113357544,
-      "logps/chosen": -27.286752700805664,
-      "logps/rejected": -31.543277740478516,
-      "loss": 0.6353,
-      "rewards/accuracies": 0.675000011920929,
-      "rewards/chosen": 0.3771280348300934,
-      "rewards/margins": 0.6008729934692383,
-      "rewards/rejected": -0.22374498844146729,
+      "logits/chosen": -2.823460817337036,
+      "logits/rejected": -2.818603038787842,
+      "logps/chosen": -27.322635650634766,
+      "logps/rejected": -31.723133087158203,
+      "loss": 0.711,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.2011496126651764,
+      "rewards/margins": 0.4009442925453186,
+      "rewards/rejected": -0.1997946947813034,
       "step": 150
     },
     {
       "epoch": 0.42,
-      "grad_norm": 61.0,
+      "grad_norm": 40.25,
       "learning_rate": 3.636998309800573e-06,
-      "logits/chosen": -3.141061305999756,
-      "logits/rejected": -3.1238036155700684,
-      "logps/chosen": -31.78842544555664,
-      "logps/rejected": -29.447246551513672,
-      "loss": 0.5473,
-      "rewards/accuracies": 0.699999988079071,
-      "rewards/chosen": 0.4923074245452881,
-      "rewards/margins": 0.8195418119430542,
-      "rewards/rejected": -0.32723429799079895,
+      "logits/chosen": -3.140930414199829,
+      "logits/rejected": -3.123213291168213,
+      "logps/chosen": -31.732044219970703,
+      "logps/rejected": -29.67775535583496,
+      "loss": 0.5884,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": 0.3038697838783264,
+      "rewards/margins": 0.5830621123313904,
+      "rewards/rejected": -0.27919238805770874,
       "step": 160
     },
     {
       "epoch": 0.44,
-      "grad_norm": 47.25,
+      "grad_norm": 33.5,
       "learning_rate": 3.4304331721118078e-06,
-      "logits/chosen": -2.9534718990325928,
-      "logits/rejected": -2.960083484649658,
-      "logps/chosen": -29.32253074645996,
-      "logps/rejected": -31.597339630126953,
-      "loss": 0.5934,
-      "rewards/accuracies": 0.737500011920929,
-      "rewards/chosen": 0.40046125650405884,
-      "rewards/margins": 0.7489619255065918,
-      "rewards/rejected": -0.34850066900253296,
+      "logits/chosen": -2.9524331092834473,
+      "logits/rejected": -2.960047960281372,
+      "logps/chosen": -29.4462947845459,
+      "logps/rejected": -31.882787704467773,
+      "loss": 0.6552,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": 0.1793280392885208,
+      "rewards/margins": 0.49264994263648987,
+      "rewards/rejected": -0.31332188844680786,
       "step": 170
     },
     {
       "epoch": 0.47,
-      "grad_norm": 51.0,
+      "grad_norm": 28.375,
       "learning_rate": 3.2162026428305436e-06,
-      "logits/chosen": -2.8042290210723877,
-      "logits/rejected": -2.801795244216919,
-      "logps/chosen": -29.153839111328125,
-      "logps/rejected": -30.13584327697754,
-      "loss": 0.6384,
-      "rewards/accuracies": 0.7250000238418579,
-      "rewards/chosen": 0.36120903491973877,
-      "rewards/margins": 0.618577778339386,
-      "rewards/rejected": -0.2573687434196472,
+      "logits/chosen": -2.8057327270507812,
+      "logits/rejected": -2.803337812423706,
+      "logps/chosen": -29.322677612304688,
+      "logps/rejected": -30.268985748291016,
+      "loss": 0.7292,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": 0.1388693004846573,
+      "rewards/margins": 0.3391944468021393,
+      "rewards/rejected": -0.2003251612186432,
       "step": 180
     },
     {
       "epoch": 0.49,
-      "grad_norm": 31.125,
+      "grad_norm": 23.625,
       "learning_rate": 2.996071664294641e-06,
-      "logits/chosen": -2.9180445671081543,
-      "logits/rejected": -2.914367914199829,
-      "logps/chosen": -29.715240478515625,
-      "logps/rejected": -28.60836410522461,
-      "loss": 0.7637,
-      "rewards/accuracies": 0.6499999761581421,
-      "rewards/chosen": 0.33003363013267517,
-      "rewards/margins": 0.5056184530258179,
-      "rewards/rejected": -0.1755848377943039,
+      "logits/chosen": -2.9184248447418213,
+      "logits/rejected": -2.9148693084716797,
+      "logps/chosen": -29.80621910095215,
+      "logps/rejected": -28.81686782836914,
+      "loss": 0.7753,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.1521972417831421,
+      "rewards/margins": 0.33593088388442993,
+      "rewards/rejected": -0.18373365700244904,
       "step": 190
     },
     {
       "epoch": 0.52,
-      "grad_norm": 46.0,
+      "grad_norm": 28.125,
       "learning_rate": 2.7718537898066833e-06,
-      "logits/chosen": -2.987438201904297,
-      "logits/rejected": -2.97584867477417,
-      "logps/chosen": -32.835445404052734,
-      "logps/rejected": -30.47810935974121,
-      "loss": 0.7257,
-      "rewards/accuracies": 0.675000011920929,
-      "rewards/chosen": 0.6045005917549133,
-      "rewards/margins": 0.6353167295455933,
-      "rewards/rejected": -0.030816053971648216,
+      "logits/chosen": -2.985816478729248,
+      "logits/rejected": -2.9742205142974854,
+      "logps/chosen": -32.966896057128906,
+      "logps/rejected": -30.653575897216797,
+      "loss": 0.7459,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": 0.29284805059432983,
+      "rewards/margins": 0.38064295053482056,
+      "rewards/rejected": -0.08779484033584595,
       "step": 200
     },
     {
       "epoch": 0.52,
-      "eval_logits/chosen": -2.8270444869995117,
-      "eval_logits/rejected": -2.8244259357452393,
-      "eval_logps/chosen": -31.380773544311523,
-      "eval_logps/rejected": -34.95420837402344,
-      "eval_loss": 0.9027834534645081,
-      "eval_rewards/accuracies": 0.5573089718818665,
-      "eval_rewards/chosen": -0.06882722675800323,
-      "eval_rewards/margins": 0.11040809750556946,
-      "eval_rewards/rejected": -0.1792353093624115,
-      "eval_runtime": 113.148,
-      "eval_samples_per_second": 3.031,
+      "eval_logits/chosen": -2.825000762939453,
+      "eval_logits/rejected": -2.8226914405822754,
+      "eval_logps/chosen": -31.455589294433594,
+      "eval_logps/rejected": -35.09489822387695,
+      "eval_loss": 0.9159321784973145,
+      "eval_rewards/accuracies": 0.5834717750549316,
+      "eval_rewards/chosen": -0.06925402581691742,
+      "eval_rewards/margins": 0.0894429013133049,
+      "eval_rewards/rejected": -0.15869693458080292,
+      "eval_runtime": 113.1231,
+      "eval_samples_per_second": 3.032,
       "eval_steps_per_second": 0.38,
       "step": 200
     },
     {
       "epoch": 0.55,
-      "grad_norm": 56.25,
+      "grad_norm": 34.0,
       "learning_rate": 2.5453962426402006e-06,
-      "logits/chosen": -2.9236907958984375,
-      "logits/rejected": -2.9235434532165527,
-      "logps/chosen": -32.24640655517578,
-      "logps/rejected": -34.160125732421875,
-      "loss": 0.5593,
-      "rewards/accuracies": 0.7749999761581421,
-      "rewards/chosen": 0.5509027242660522,
-      "rewards/margins": 0.7599622011184692,
-      "rewards/rejected": -0.2090594321489334,
+      "logits/chosen": -2.9240100383758545,
+      "logits/rejected": -2.9254937171936035,
+      "logps/chosen": -32.46898651123047,
+      "logps/rejected": -34.43748092651367,
+      "loss": 0.6514,
+      "rewards/accuracies": 0.7250000238418579,
+      "rewards/chosen": 0.22576837241649628,
+      "rewards/margins": 0.4561736583709717,
+      "rewards/rejected": -0.2304052859544754,
       "step": 210
     },
     {
       "epoch": 0.57,
-      "grad_norm": 42.25,
+      "grad_norm": 32.25,
       "learning_rate": 2.3185646976551794e-06,
-      "logits/chosen": -2.90382981300354,
-      "logits/rejected": -2.9191505908966064,
-      "logps/chosen": -29.37839698791504,
-      "logps/rejected": -28.82822036743164,
-      "loss": 0.6297,
-      "rewards/accuracies": 0.7124999761581421,
-      "rewards/chosen": 0.5597518682479858,
-      "rewards/margins": 0.7697170972824097,
-      "rewards/rejected": -0.20996519923210144,
+      "logits/chosen": -2.9049904346466064,
+      "logits/rejected": -2.9205589294433594,
+      "logps/chosen": -29.369049072265625,
+      "logps/rejected": -29.028148651123047,
+      "loss": 0.6606,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": 0.3235965967178345,
+      "rewards/margins": 0.523547351360321,
+      "rewards/rejected": -0.19995081424713135,
       "step": 220
     },
     {
       "epoch": 0.6,
-      "grad_norm": 44.5,
+      "grad_norm": 28.75,
       "learning_rate": 2.0932279108998323e-06,
-      "logits/chosen": -2.9497718811035156,
-      "logits/rejected": -2.954267740249634,
-      "logps/chosen": -30.704570770263672,
-      "logps/rejected": -32.00530242919922,
-      "loss": 0.7476,
-      "rewards/accuracies": 0.625,
-      "rewards/chosen": 0.3131088614463806,
-      "rewards/margins": 0.48611265420913696,
-      "rewards/rejected": -0.17300380766391754,
+      "logits/chosen": -2.950502395629883,
+      "logits/rejected": -2.9541735649108887,
+      "logps/chosen": -30.678882598876953,
+      "logps/rejected": -32.15251922607422,
+      "loss": 0.7668,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": 0.18919439613819122,
+      "rewards/margins": 0.3469431698322296,
+      "rewards/rejected": -0.1577487289905548,
       "step": 230
     },
     {
       "epoch": 0.62,
-      "grad_norm": 41.75,
+      "grad_norm": 34.0,
       "learning_rate": 1.8712423238279358e-06,
-      "logits/chosen": -3.004119396209717,
-      "logits/rejected": -3.012087106704712,
-      "logps/chosen": -30.007003784179688,
-      "logps/rejected": -30.454036712646484,
-      "loss": 0.4699,
-      "rewards/accuracies": 0.8500000238418579,
-      "rewards/chosen": 0.6285442113876343,
-      "rewards/margins": 0.7816993594169617,
-      "rewards/rejected": -0.15315520763397217,
+      "logits/chosen": -3.00547456741333,
+      "logits/rejected": -3.013357400894165,
+      "logps/chosen": -30.189453125,
+      "logps/rejected": -30.650592803955078,
+      "loss": 0.6128,
+      "rewards/accuracies": 0.7749999761581421,
+      "rewards/chosen": 0.2861887216567993,
+      "rewards/margins": 0.4523293375968933,
+      "rewards/rejected": -0.1661405861377716,
       "step": 240
     },
     {
       "epoch": 0.65,
-      "grad_norm": 48.25,
+      "grad_norm": 33.0,
       "learning_rate": 1.6544367689701824e-06,
-      "logits/chosen": -2.8357105255126953,
-      "logits/rejected": -2.825700283050537,
-      "logps/chosen": -26.410411834716797,
-      "logps/rejected": -29.5693416595459,
-      "loss": 0.719,
-      "rewards/accuracies": 0.737500011920929,
-      "rewards/chosen": 0.3806636929512024,
-      "rewards/margins": 0.40705281496047974,
-      "rewards/rejected": -0.02638910338282585,
+      "logits/chosen": -2.838496208190918,
+      "logits/rejected": -2.828334093093872,
+      "logps/chosen": -26.336994171142578,
+      "logps/rejected": -29.769458770751953,
+      "loss": 0.7551,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.24688871204853058,
+      "rewards/margins": 0.3420138955116272,
+      "rewards/rejected": -0.09512520581483841,
       "step": 250
     },
     {
       "epoch": 0.68,
-      "grad_norm": 19.25,
+      "grad_norm": 29.625,
       "learning_rate": 1.4445974030621963e-06,
-      "logits/chosen": -2.817229747772217,
-      "logits/rejected": -2.837646245956421,
-      "logps/chosen": -28.973642349243164,
-      "logps/rejected": -34.52573013305664,
-      "loss": 0.5428,
-      "rewards/accuracies": 0.824999988079071,
-      "rewards/chosen": 0.599422812461853,
-      "rewards/margins": 0.8455693125724792,
-      "rewards/rejected": -0.24614660441875458,
+      "logits/chosen": -2.8198904991149902,
+      "logits/rejected": -2.8399436473846436,
+      "logps/chosen": -29.0089168548584,
+      "logps/rejected": -34.85820388793945,
+      "loss": 0.5813,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": 0.32841864228248596,
+      "rewards/margins": 0.6020644903182983,
+      "rewards/rejected": -0.27364593744277954,
       "step": 260
     },
     {
       "epoch": 0.7,
-      "grad_norm": 62.25,
+      "grad_norm": 36.75,
       "learning_rate": 1.243452991757889e-06,
-      "logits/chosen": -2.959622859954834,
-      "logits/rejected": -2.9659740924835205,
-      "logps/chosen": -30.064538955688477,
-      "logps/rejected": -30.272811889648438,
-      "loss": 0.534,
-      "rewards/accuracies": 0.800000011920929,
-      "rewards/chosen": 0.5017729997634888,
-      "rewards/margins": 0.7870587110519409,
-      "rewards/rejected": -0.28528574109077454,
+      "logits/chosen": -2.962282657623291,
+      "logits/rejected": -2.969045639038086,
+      "logps/chosen": -30.215042114257812,
+      "logps/rejected": -30.566543579101562,
+      "loss": 0.5913,
+      "rewards/accuracies": 0.8125,
+      "rewards/chosen": 0.22652491927146912,
+      "rewards/margins": 0.5070374608039856,
+      "rewards/rejected": -0.2805125117301941,
       "step": 270
     },
     {
       "epoch": 0.73,
-      "grad_norm": 31.75,
+      "grad_norm": 29.5,
       "learning_rate": 1.0526606671603523e-06,
-      "logits/chosen": -2.974242687225342,
-      "logits/rejected": -2.960573673248291,
-      "logps/chosen": -30.0527400970459,
-      "logps/rejected": -28.732250213623047,
-      "loss": 0.7261,
-      "rewards/accuracies": 0.7749999761581421,
-      "rewards/chosen": 0.4682597219944,
-      "rewards/margins": 0.5388898253440857,
-      "rewards/rejected": -0.07063011080026627,
+      "logits/chosen": -2.9766674041748047,
+      "logits/rejected": -2.9624428749084473,
+      "logps/chosen": -30.22548484802246,
+      "logps/rejected": -28.989456176757812,
+      "loss": 0.7584,
+      "rewards/accuracies": 0.762499988079071,
+      "rewards/chosen": 0.19847916066646576,
+      "rewards/margins": 0.34172323346138,
+      "rewards/rejected": -0.14324405789375305,
       "step": 280
     },
     {
       "epoch": 0.75,
-      "grad_norm": 32.25,
+      "grad_norm": 30.0,
       "learning_rate": 8.737922755071455e-07,
-      "logits/chosen": -2.901106595993042,
-      "logits/rejected": -2.8834261894226074,
-      "logps/chosen": -31.619491577148438,
-      "logps/rejected": -31.00750160217285,
-      "loss": 0.4327,
-      "rewards/accuracies": 0.8500000238418579,
-      "rewards/chosen": 0.7117676734924316,
-      "rewards/margins": 1.0507009029388428,
-      "rewards/rejected": -0.3389332592487335,
+      "logits/chosen": -2.9045450687408447,
+      "logits/rejected": -2.887028932571411,
+      "logps/chosen": -31.592443466186523,
+      "logps/rejected": -31.358936309814453,
+      "loss": 0.4878,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": 0.4175439774990082,
+      "rewards/margins": 0.751793384552002,
+      "rewards/rejected": -0.3342495858669281,
       "step": 290
     },
     {
       "epoch": 0.78,
-      "grad_norm": 43.5,
+      "grad_norm": 24.75,
       "learning_rate": 7.08321427484816e-07,
-      "logits/chosen": -2.9032962322235107,
-      "logits/rejected": -2.8999171257019043,
-      "logps/chosen": -31.358158111572266,
-      "logps/rejected": -27.814193725585938,
-      "loss": 0.6308,
-      "rewards/accuracies": 0.737500011920929,
-      "rewards/chosen": 0.6051477789878845,
-      "rewards/margins": 0.7525756359100342,
-      "rewards/rejected": -0.14742788672447205,
+      "logits/chosen": -2.9053683280944824,
+      "logits/rejected": -2.9024596214294434,
+      "logps/chosen": -31.347707748413086,
+      "logps/rejected": -28.053564071655273,
+      "loss": 0.665,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": 0.34997984766960144,
+      "rewards/margins": 0.529971182346344,
+      "rewards/rejected": -0.17999130487442017,
       "step": 300
     },
     {
       "epoch": 0.78,
-      "eval_logits/chosen": -2.8207788467407227,
-      "eval_logits/rejected": -2.8184332847595215,
-      "eval_logps/chosen": -31.37415885925293,
-      "eval_logps/rejected": -34.95107650756836,
-      "eval_loss": 0.9011654257774353,
+      "eval_logits/chosen": -2.8241891860961914,
+      "eval_logits/rejected": -2.8221466541290283,
+      "eval_logps/chosen": -31.476619720458984,
+      "eval_logps/rejected": -35.105430603027344,
+      "eval_loss": 0.9201633334159851,
       "eval_rewards/accuracies": 0.565614640712738,
-      "eval_rewards/chosen": -0.06419765949249268,
-      "eval_rewards/margins": 0.11284750699996948,
-      "eval_rewards/rejected": -0.17704515159130096,
-      "eval_runtime": 113.1332,
+      "eval_rewards/chosen": -0.07766681164503098,
+      "eval_rewards/margins": 0.08524217456579208,
+      "eval_rewards/rejected": -0.16290898621082306,
+      "eval_runtime": 113.1247,
       "eval_samples_per_second": 3.032,
       "eval_steps_per_second": 0.38,
       "step": 300
     },
     {
       "epoch": 0.81,
-      "grad_norm": 52.0,
+      "grad_norm": 32.5,
       "learning_rate": 5.576113578589035e-07,
-      "logits/chosen": -2.7831432819366455,
-      "logits/rejected": -2.801016092300415,
-      "logps/chosen": -28.50387191772461,
-      "logps/rejected": -31.167369842529297,
-      "loss": 0.4507,
-      "rewards/accuracies": 0.862500011920929,
-      "rewards/chosen": 0.5078839659690857,
-      "rewards/margins": 0.8126392364501953,
-      "rewards/rejected": -0.30475518107414246,
+      "logits/chosen": -2.7875466346740723,
+      "logits/rejected": -2.8046317100524902,
+      "logps/chosen": -28.552978515625,
+      "logps/rejected": -31.452463150024414,
+      "loss": 0.5777,
+      "rewards/accuracies": 0.8500000238418579,
+      "rewards/chosen": 0.27057594060897827,
+      "rewards/margins": 0.5587579011917114,
+      "rewards/rejected": -0.28818196058273315,
       "step": 310
     },
     {
       "epoch": 0.83,
-      "grad_norm": 31.625,
+      "grad_norm": 26.75,
       "learning_rate": 4.229036944380913e-07,
-      "logits/chosen": -3.029501438140869,
-      "logits/rejected": -3.0155110359191895,
-      "logps/chosen": -29.30743408203125,
-      "logps/rejected": -28.680965423583984,
-      "loss": 0.4592,
-      "rewards/accuracies": 0.8374999761581421,
-      "rewards/chosen": 0.6329048871994019,
-      "rewards/margins": 0.8864381909370422,
-      "rewards/rejected": -0.25353336334228516,
+      "logits/chosen": -3.0314865112304688,
+      "logits/rejected": -3.0174806118011475,
+      "logps/chosen": -29.399667739868164,
+      "logps/rejected": -28.924419403076172,
+      "loss": 0.5701,
+      "rewards/accuracies": 0.7749999761581421,
+      "rewards/chosen": 0.32476717233657837,
+      "rewards/margins": 0.567025363445282,
+      "rewards/rejected": -0.24225828051567078,
       "step": 320
     },
     {
       "epoch": 0.86,
-      "grad_norm": 17.0,
+      "grad_norm": 18.25,
       "learning_rate": 3.053082288996112e-07,
-      "logits/chosen": -2.943554401397705,
-      "logits/rejected": -2.925654888153076,
-      "logps/chosen": -27.233322143554688,
-      "logps/rejected": -30.869421005249023,
-      "loss": 0.4325,
-      "rewards/accuracies": 0.7875000238418579,
-      "rewards/chosen": 0.7901242971420288,
-      "rewards/margins": 1.1245664358139038,
-      "rewards/rejected": -0.33444222807884216,
+      "logits/chosen": -2.946519374847412,
+      "logits/rejected": -2.9290707111358643,
+      "logps/chosen": -27.334081649780273,
+      "logps/rejected": -31.206268310546875,
+      "loss": 0.4948,
+      "rewards/accuracies": 0.8374999761581421,
+      "rewards/chosen": 0.4111965298652649,
+      "rewards/margins": 0.7370456457138062,
+      "rewards/rejected": -0.32584911584854126,
       "step": 330
     },
     {
       "epoch": 0.88,
-      "grad_norm": 53.5,
+      "grad_norm": 35.0,
       "learning_rate": 2.0579377374915805e-07,
-      "logits/chosen": -3.1611580848693848,
-      "logits/rejected": -3.167454242706299,
-      "logps/chosen": -30.857864379882812,
-      "logps/rejected": -33.212623596191406,
-      "loss": 0.4854,
-      "rewards/accuracies": 0.8500000238418579,
-      "rewards/chosen": 0.546938955783844,
-      "rewards/margins": 0.908575177192688,
-      "rewards/rejected": -0.3616361618041992,
+      "logits/chosen": -3.16451096534729,
+      "logits/rejected": -3.1703600883483887,
+      "logps/chosen": -30.927570343017578,
+      "logps/rejected": -33.580810546875,
+      "loss": 0.5586,
+      "rewards/accuracies": 0.7875000238418579,
+      "rewards/chosen": 0.2846537232398987,
+      "rewards/margins": 0.6385756731033325,
+      "rewards/rejected": -0.35392194986343384,
       "step": 340
     },
     {
       "epoch": 0.91,
-      "grad_norm": 40.0,
+      "grad_norm": 23.875,
       "learning_rate": 1.2518018074041684e-07,
-      "logits/chosen": -3.037867546081543,
-      "logits/rejected": -3.040760040283203,
-      "logps/chosen": -29.823516845703125,
-      "logps/rejected": -31.82476806640625,
-      "loss": 0.5086,
+      "logits/chosen": -3.040112018585205,
+      "logits/rejected": -3.0432353019714355,
+      "logps/chosen": -29.85060691833496,
+      "logps/rejected": -32.070159912109375,
+      "loss": 0.5751,
       "rewards/accuracies": 0.762499988079071,
-      "rewards/chosen": 0.6883540153503418,
-      "rewards/margins": 0.9063750505447388,
-      "rewards/rejected": -0.21802091598510742,
+      "rewards/chosen": 0.3825104236602783,
+      "rewards/margins": 0.6052514314651489,
+      "rewards/rejected": -0.22274093329906464,
       "step": 350
     },
     {
       "epoch": 0.94,
-      "grad_norm": 41.5,
+      "grad_norm": 35.0,
       "learning_rate": 6.41315865106129e-08,
-      "logits/chosen": -2.877352476119995,
-      "logits/rejected": -2.878804922103882,
-      "logps/chosen": -27.442968368530273,
-      "logps/rejected": -30.033910751342773,
-      "loss": 0.4586,
-      "rewards/accuracies": 0.824999988079071,
-      "rewards/chosen": 0.8447319865226746,
-      "rewards/margins": 0.986809253692627,
-      "rewards/rejected": -0.1420772522687912,
+      "logits/chosen": -2.8801891803741455,
+      "logits/rejected": -2.8814785480499268,
+      "logps/chosen": -27.602718353271484,
+      "logps/rejected": -30.264202117919922,
+      "loss": 0.5555,
+      "rewards/accuracies": 0.762499988079071,
+      "rewards/chosen": 0.41880473494529724,
+      "rewards/margins": 0.5921100974082947,
+      "rewards/rejected": -0.17330540716648102,
       "step": 360
     },
     {
       "epoch": 0.96,
-      "grad_norm": 66.5,
+      "grad_norm": 41.0,
       "learning_rate": 2.3150941078050325e-08,
-      "logits/chosen": -2.9520726203918457,
-      "logits/rejected": -2.949944257736206,
-      "logps/chosen": -29.80544090270996,
-      "logps/rejected": -32.19409942626953,
-      "loss": 0.5126,
+      "logits/chosen": -2.9562087059020996,
+      "logits/rejected": -2.9534800052642822,
+      "logps/chosen": -29.909082412719727,
+      "logps/rejected": -32.4835205078125,
+      "loss": 0.5884,
       "rewards/accuracies": 0.7875000238418579,
-      "rewards/chosen": 0.6868453025817871,
-      "rewards/margins": 0.873086154460907,
-      "rewards/rejected": -0.1862407922744751,
+      "rewards/chosen": 0.3510281443595886,
+      "rewards/margins": 0.5732189416885376,
+      "rewards/rejected": -0.22219082713127136,
       "step": 370
     },
     {
       "epoch": 0.99,
-      "grad_norm": 49.5,
+      "grad_norm": 30.75,
       "learning_rate": 2.575864278703266e-09,
-      "logits/chosen": -2.908616542816162,
-      "logits/rejected": -2.890885829925537,
-      "logps/chosen": -28.092891693115234,
-      "logps/rejected": -28.3051700592041,
-      "loss": 0.4503,
-      "rewards/accuracies": 0.8500000238418579,
-      "rewards/chosen": 0.5965021252632141,
-      "rewards/margins": 0.8873867988586426,
-      "rewards/rejected": -0.290884792804718,
+      "logits/chosen": -2.9113378524780273,
+      "logits/rejected": -2.894592523574829,
+      "logps/chosen": -28.256174087524414,
+      "logps/rejected": -28.505691528320312,
+      "loss": 0.5793,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": 0.275543749332428,
+      "rewards/margins": 0.521973192691803,
+      "rewards/rejected": -0.24642948806285858,
       "step": 380
     },
     {
       "epoch": 1.0,
       "step": 385,
       "total_flos": 0.0,
-      "train_loss": 0.6684874992866021,
-      "train_runtime": 2718.5428,
-      "train_samples_per_second": 1.133,
+      "train_loss": 0.7203964704042906,
+      "train_runtime": 2720.5927,
+      "train_samples_per_second": 1.132,
       "train_steps_per_second": 0.142
     }
   ],