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: 0.9008
-- Rewards/chosen: -0.1005
-- Rewards/rejected: -0.2216
-- Rewards/accuracies: 0.5860
-- Rewards/margins: 0.1211
-- Logps/rejected: -34.9752
-- Logps/chosen: -31.4081
-- Logits/rejected: -2.8074
-- Logits/chosen: -2.8099
+- Loss: 0.9216
+- Rewards/chosen: -0.0864
+- Rewards/rejected: -0.1686
+- Rewards/accuracies: 0.5573
+- Rewards/margins: 0.0822
+- Logps/rejected: -35.0354
+- Logps/chosen: -31.4553
+- Logits/rejected: -2.8133
+- Logits/chosen: -2.8155
 
 ## 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 |
 |:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
-| 0.8401        | 0.26  | 100  | 0.8873          | 0.0047         | -0.1245          | 0.6067             | 0.1292          | -34.8538       | -31.2766     | -2.8093         | -2.8118       |
-| 0.724         | 0.52  | 200  | 0.9140          | -0.0857        | -0.1841          | 0.5656             | 0.0983          | -34.9282       | -31.3896     | -2.8156         | -2.8179       |
-| 0.6408        | 0.78  | 300  | 0.9091          | -0.0880        | -0.2014          | 0.5627             | 0.1135          | -34.9500       | -31.3924     | -2.8077         | -2.8102       |
+| 0.841         | 0.26  | 100  | 0.9188          | -0.0118        | -0.0997          | 0.6100             | 0.0879          | -34.8976       | -31.3062     | -2.8098         | -2.8128       |
+| 0.7163        | 0.52  | 200  | 0.9221          | -0.0953        | -0.1781          | 0.5515             | 0.0829          | -35.0544       | -31.4730     | -2.8173         | -2.8196       |
+| 0.6439        | 0.78  | 300  | 0.9216          | -0.0864        | -0.1686          | 0.5573             | 0.0822          | -35.0354       | -31.4553     | -2.8133         | -2.8155       |
 
 
 ### Framework versions

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b8de37cc5ff8fac8bd9cd818356b7563ae3bd14b8eb454198d8b961c116a4cf0
+oid sha256:e06bc74492e4ee77017513e8c2dee98bb3a45c695e337a2ca2f1dd93222f3d99
 size 671150064

all_results.json

@@ -1,21 +1,8 @@
 {
     "epoch": 1.0,
-    "eval_logits/chosen": -2.8098936080932617,
-    "eval_logits/rejected": -2.807415723800659,
-    "eval_logps/chosen": -31.408082962036133,
-    "eval_logps/rejected": -34.975181579589844,
-    "eval_loss": 0.9007851481437683,
-    "eval_rewards/accuracies": 0.5859634280204773,
-    "eval_rewards/chosen": -0.1005057543516159,
-    "eval_rewards/margins": 0.12111683934926987,
-    "eval_rewards/rejected": -0.22162257134914398,
-    "eval_runtime": 112.7349,
-    "eval_samples": 343,
-    "eval_samples_per_second": 3.043,
-    "eval_steps_per_second": 0.381,
-    "train_loss": 0.6588526527602951,
-    "train_runtime": 2719.293,
+    "train_loss": 0.694122297732861,
+    "train_runtime": 2717.9333,
     "train_samples": 3079,
-    "train_samples_per_second": 1.132,
+    "train_samples_per_second": 1.133,
     "train_steps_per_second": 0.142
 }

train_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 1.0,
-    "train_loss": 0.6588526527602951,
-    "train_runtime": 2719.293,
+    "train_loss": 0.694122297732861,
+    "train_runtime": 2717.9333,
     "train_samples": 3079,
-    "train_samples_per_second": 1.132,
+    "train_samples_per_second": 1.133,
     "train_steps_per_second": 0.142
 }

trainer_state.json

@@ -10,7 +10,7 @@
   "log_history": [
     {
       "epoch": 0.0,
-      "grad_norm": 63.25,
+      "grad_norm": 39.5,
       "learning_rate": 1.282051282051282e-07,
       "logits/chosen": -2.7358343601226807,
       "logits/rejected": -2.7480404376983643,
@@ -25,629 +25,629 @@
     },
     {
       "epoch": 0.03,
-      "grad_norm": 108.0,
+      "grad_norm": 67.5,
       "learning_rate": 1.282051282051282e-06,
-      "logits/chosen": -3.009650945663452,
-      "logits/rejected": -2.998239040374756,
-      "logps/chosen": -33.192203521728516,
-      "logps/rejected": -31.957557678222656,
-      "loss": 1.0141,
-      "rewards/accuracies": 0.4166666567325592,
-      "rewards/chosen": -0.009009478613734245,
-      "rewards/margins": -0.014087951742112637,
-      "rewards/rejected": 0.005078474525362253,
+      "logits/chosen": -3.0102524757385254,
+      "logits/rejected": -2.9992215633392334,
+      "logps/chosen": -33.198204040527344,
+      "logps/rejected": -31.97161865234375,
+      "loss": 1.0048,
+      "rewards/accuracies": 0.3888888955116272,
+      "rewards/chosen": -0.008629633113741875,
+      "rewards/margins": -0.00477217324078083,
+      "rewards/rejected": -0.0038574603386223316,
       "step": 10
     },
     {
       "epoch": 0.05,
-      "grad_norm": 73.5,
+      "grad_norm": 45.5,
       "learning_rate": 2.564102564102564e-06,
-      "logits/chosen": -2.899263381958008,
-      "logits/rejected": -2.894313335418701,
-      "logps/chosen": -32.45400619506836,
-      "logps/rejected": -28.9648494720459,
-      "loss": 0.9967,
-      "rewards/accuracies": 0.48750001192092896,
-      "rewards/chosen": 0.016872350126504898,
-      "rewards/margins": 0.003259001299738884,
-      "rewards/rejected": 0.01361334603279829,
+      "logits/chosen": -2.8995842933654785,
+      "logits/rejected": -2.8944945335388184,
+      "logps/chosen": -32.47186279296875,
+      "logps/rejected": -28.95808982849121,
+      "loss": 1.0103,
+      "rewards/accuracies": 0.44999998807907104,
+      "rewards/chosen": 0.0016174853080883622,
+      "rewards/margins": -0.01026993989944458,
+      "rewards/rejected": 0.011887425556778908,
       "step": 20
     },
     {
       "epoch": 0.08,
-      "grad_norm": 71.0,
+      "grad_norm": 44.25,
       "learning_rate": 3.846153846153847e-06,
-      "logits/chosen": -3.0970497131347656,
-      "logits/rejected": -3.108996868133545,
-      "logps/chosen": -32.78731918334961,
-      "logps/rejected": -30.140506744384766,
-      "loss": 0.9473,
-      "rewards/accuracies": 0.574999988079071,
-      "rewards/chosen": 0.09709431231021881,
-      "rewards/margins": 0.05265679210424423,
-      "rewards/rejected": 0.044437527656555176,
+      "logits/chosen": -3.0970377922058105,
+      "logits/rejected": -3.1085267066955566,
+      "logps/chosen": -32.78118133544922,
+      "logps/rejected": -30.125417709350586,
+      "loss": 0.9716,
+      "rewards/accuracies": 0.5375000238418579,
+      "rewards/chosen": 0.06375427544116974,
+      "rewards/margins": 0.02843797765672207,
+      "rewards/rejected": 0.03531629964709282,
       "step": 30
     },
     {
       "epoch": 0.1,
-      "grad_norm": 57.75,
+      "grad_norm": 36.0,
       "learning_rate": 4.999896948438434e-06,
-      "logits/chosen": -2.8630309104919434,
-      "logits/rejected": -2.8540406227111816,
-      "logps/chosen": -31.542278289794922,
-      "logps/rejected": -32.394432067871094,
-      "loss": 0.8414,
-      "rewards/accuracies": 0.6499999761581421,
-      "rewards/chosen": 0.23043569922447205,
-      "rewards/margins": 0.260015070438385,
-      "rewards/rejected": -0.02957936003804207,
+      "logits/chosen": -2.8625752925872803,
+      "logits/rejected": -2.853350877761841,
+      "logps/chosen": -31.52760887145996,
+      "logps/rejected": -32.38980484008789,
+      "loss": 0.8635,
+      "rewards/accuracies": 0.574999988079071,
+      "rewards/chosen": 0.15135782957077026,
+      "rewards/margins": 0.16753093898296356,
+      "rewards/rejected": -0.016173111274838448,
       "step": 40
     },
     {
       "epoch": 0.13,
-      "grad_norm": 44.75,
+      "grad_norm": 28.0,
       "learning_rate": 4.987541037542187e-06,
-      "logits/chosen": -2.8809738159179688,
-      "logits/rejected": -2.8790669441223145,
-      "logps/chosen": -29.41156005859375,
-      "logps/rejected": -30.12240219116211,
-      "loss": 0.8761,
-      "rewards/accuracies": 0.550000011920929,
-      "rewards/chosen": 0.26288196444511414,
-      "rewards/margins": 0.2896483242511749,
-      "rewards/rejected": -0.02676635980606079,
+      "logits/chosen": -2.880800724029541,
+      "logits/rejected": -2.8784618377685547,
+      "logps/chosen": -29.383716583251953,
+      "logps/rejected": -30.082469940185547,
+      "loss": 0.8801,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": 0.1782219111919403,
+      "rewards/margins": 0.17498424649238586,
+      "rewards/rejected": 0.00323770334944129,
       "step": 50
     },
     {
       "epoch": 0.16,
-      "grad_norm": 63.75,
+      "grad_norm": 41.5,
       "learning_rate": 4.954691471941119e-06,
-      "logits/chosen": -2.9108948707580566,
-      "logits/rejected": -2.912576675415039,
-      "logps/chosen": -29.901845932006836,
-      "logps/rejected": -28.0941219329834,
-      "loss": 0.7968,
-      "rewards/accuracies": 0.6499999761581421,
-      "rewards/chosen": 0.18233974277973175,
-      "rewards/margins": 0.2930926978588104,
-      "rewards/rejected": -0.11075299978256226,
+      "logits/chosen": -2.9095096588134766,
+      "logits/rejected": -2.9110829830169678,
+      "logps/chosen": -29.854755401611328,
+      "logps/rejected": -28.0576114654541,
+      "loss": 0.8396,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.13750770688056946,
+      "rewards/margins": 0.18847230076789856,
+      "rewards/rejected": -0.0509646013379097,
       "step": 60
     },
     {
       "epoch": 0.18,
-      "grad_norm": 72.5,
+      "grad_norm": 54.75,
       "learning_rate": 4.901618883413549e-06,
-      "logits/chosen": -2.9933552742004395,
-      "logits/rejected": -2.9995028972625732,
-      "logps/chosen": -29.255428314208984,
-      "logps/rejected": -31.047006607055664,
-      "loss": 0.9015,
-      "rewards/accuracies": 0.574999988079071,
-      "rewards/chosen": 0.020431842654943466,
-      "rewards/margins": 0.14459456503391266,
-      "rewards/rejected": -0.1241627112030983,
+      "logits/chosen": -2.9918501377105713,
+      "logits/rejected": -2.998004913330078,
+      "logps/chosen": -29.19577407836914,
+      "logps/rejected": -31.001514434814453,
+      "loss": 0.9064,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.042597077786922455,
+      "rewards/margins": 0.0974539965391159,
+      "rewards/rejected": -0.05485691875219345,
       "step": 70
     },
     {
       "epoch": 0.21,
-      "grad_norm": 70.5,
+      "grad_norm": 47.25,
       "learning_rate": 4.828760511501322e-06,
-      "logits/chosen": -2.808861255645752,
-      "logits/rejected": -2.8243188858032227,
-      "logps/chosen": -29.406871795654297,
-      "logps/rejected": -29.915807723999023,
-      "loss": 0.7973,
-      "rewards/accuracies": 0.6875,
-      "rewards/chosen": 0.10708501189947128,
-      "rewards/margins": 0.30030542612075806,
-      "rewards/rejected": -0.19322039186954498,
+      "logits/chosen": -2.8070733547210693,
+      "logits/rejected": -2.8222880363464355,
+      "logps/chosen": -29.365554809570312,
+      "logps/rejected": -29.91384506225586,
+      "loss": 0.8182,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.08758720755577087,
+      "rewards/margins": 0.20736832916736603,
+      "rewards/rejected": -0.11978113651275635,
       "step": 80
     },
     {
       "epoch": 0.23,
-      "grad_norm": 56.25,
+      "grad_norm": 35.25,
       "learning_rate": 4.7367166013034295e-06,
-      "logits/chosen": -2.8983397483825684,
-      "logits/rejected": -2.880967617034912,
-      "logps/chosen": -32.71396255493164,
-      "logps/rejected": -30.347427368164062,
-      "loss": 0.8646,
-      "rewards/accuracies": 0.675000011920929,
-      "rewards/chosen": 0.09797407686710358,
-      "rewards/margins": 0.35859915614128113,
-      "rewards/rejected": -0.26062512397766113,
+      "logits/chosen": -2.8982958793640137,
+      "logits/rejected": -2.880619525909424,
+      "logps/chosen": -32.76674270629883,
+      "logps/rejected": -30.40105628967285,
+      "loss": 0.8554,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": 0.03484407439827919,
+      "rewards/margins": 0.2245481014251709,
+      "rewards/rejected": -0.1897040456533432,
       "step": 90
     },
     {
       "epoch": 0.26,
-      "grad_norm": 54.75,
+      "grad_norm": 34.5,
       "learning_rate": 4.626245458345211e-06,
-      "logits/chosen": -3.004662275314331,
-      "logits/rejected": -3.005678653717041,
-      "logps/chosen": -31.851581573486328,
-      "logps/rejected": -30.93560791015625,
-      "loss": 0.8401,
+      "logits/chosen": -3.005671501159668,
+      "logits/rejected": -3.0055477619171143,
+      "logps/chosen": -31.862071990966797,
+      "logps/rejected": -30.99982261657715,
+      "loss": 0.841,
       "rewards/accuracies": 0.6000000238418579,
-      "rewards/chosen": 0.1642352044582367,
-      "rewards/margins": 0.25978168845176697,
-      "rewards/rejected": -0.09554646909236908,
+      "rewards/chosen": 0.09740196168422699,
+      "rewards/margins": 0.18922677636146545,
+      "rewards/rejected": -0.09182481467723846,
       "step": 100
     },
     {
       "epoch": 0.26,
-      "eval_logits/chosen": -2.811776876449585,
-      "eval_logits/rejected": -2.809250831604004,
-      "eval_logps/chosen": -31.276582717895508,
-      "eval_logps/rejected": -34.853797912597656,
-      "eval_loss": 0.8872909545898438,
-      "eval_rewards/accuracies": 0.6067276000976562,
-      "eval_rewards/chosen": 0.0046949307434260845,
-      "eval_rewards/margins": 0.12921129167079926,
-      "eval_rewards/rejected": -0.12451635301113129,
-      "eval_runtime": 113.4101,
-      "eval_samples_per_second": 3.024,
+      "eval_logits/chosen": -2.812790870666504,
+      "eval_logits/rejected": -2.8098013401031494,
+      "eval_logps/chosen": -31.30615234375,
+      "eval_logps/rejected": -34.89762496948242,
+      "eval_loss": 0.9188198447227478,
+      "eval_rewards/accuracies": 0.610049843788147,
+      "eval_rewards/chosen": -0.01184946671128273,
+      "eval_rewards/margins": 0.08788519352674484,
+      "eval_rewards/rejected": -0.09973466396331787,
+      "eval_runtime": 113.3423,
+      "eval_samples_per_second": 3.026,
       "eval_steps_per_second": 0.379,
       "step": 100
     },
     {
       "epoch": 0.29,
-      "grad_norm": 83.0,
+      "grad_norm": 52.75,
       "learning_rate": 4.498257201263691e-06,
-      "logits/chosen": -2.9626810550689697,
-      "logits/rejected": -2.9382669925689697,
-      "logps/chosen": -31.831439971923828,
-      "logps/rejected": -31.40035057067871,
-      "loss": 0.6952,
-      "rewards/accuracies": 0.762499988079071,
-      "rewards/chosen": 0.3082864582538605,
-      "rewards/margins": 0.5014885663986206,
-      "rewards/rejected": -0.1932021528482437,
+      "logits/chosen": -2.962007999420166,
+      "logits/rejected": -2.937967538833618,
+      "logps/chosen": -31.860488891601562,
+      "logps/rejected": -31.523914337158203,
+      "loss": 0.724,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": 0.17815347015857697,
+      "rewards/margins": 0.3606840968132019,
+      "rewards/rejected": -0.18253062665462494,
       "step": 110
     },
     {
       "epoch": 0.31,
-      "grad_norm": 65.0,
+      "grad_norm": 44.5,
       "learning_rate": 4.353806263777678e-06,
-      "logits/chosen": -3.0443854331970215,
-      "logits/rejected": -3.073098659515381,
-      "logps/chosen": -28.707149505615234,
-      "logps/rejected": -34.32903289794922,
-      "loss": 0.7028,
-      "rewards/accuracies": 0.637499988079071,
-      "rewards/chosen": 0.3654031455516815,
-      "rewards/margins": 0.5105921030044556,
-      "rewards/rejected": -0.14518897235393524,
+      "logits/chosen": -3.046175241470337,
+      "logits/rejected": -3.074599027633667,
+      "logps/chosen": -28.74391746520996,
+      "logps/rejected": -34.43209457397461,
+      "loss": 0.7306,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": 0.2099936455488205,
+      "rewards/margins": 0.3522658348083496,
+      "rewards/rejected": -0.14227214455604553,
       "step": 120
     },
     {
       "epoch": 0.34,
-      "grad_norm": 44.5,
+      "grad_norm": 33.75,
       "learning_rate": 4.1940827077152755e-06,
-      "logits/chosen": -2.7482428550720215,
-      "logits/rejected": -2.743565082550049,
-      "logps/chosen": -28.566293716430664,
-      "logps/rejected": -30.31746482849121,
-      "loss": 0.7405,
-      "rewards/accuracies": 0.6625000238418579,
-      "rewards/chosen": 0.3267093598842621,
-      "rewards/margins": 0.47363653779029846,
-      "rewards/rejected": -0.14692717790603638,
+      "logits/chosen": -2.750256061553955,
+      "logits/rejected": -2.7464325428009033,
+      "logps/chosen": -28.576923370361328,
+      "logps/rejected": -30.405838012695312,
+      "loss": 0.7558,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": 0.19887787103652954,
+      "rewards/margins": 0.3348928391933441,
+      "rewards/rejected": -0.13601499795913696,
       "step": 130
     },
     {
       "epoch": 0.36,
-      "grad_norm": 46.25,
+      "grad_norm": 29.625,
       "learning_rate": 4.0204024186666215e-06,
-      "logits/chosen": -3.022853136062622,
-      "logits/rejected": -3.0205140113830566,
-      "logps/chosen": -27.139602661132812,
-      "logps/rejected": -31.89032554626465,
-      "loss": 0.7769,
-      "rewards/accuracies": 0.625,
-      "rewards/chosen": 0.27729907631874084,
-      "rewards/margins": 0.46414414048194885,
-      "rewards/rejected": -0.1868450939655304,
+      "logits/chosen": -3.026742458343506,
+      "logits/rejected": -3.02447247505188,
+      "logps/chosen": -27.27823829650879,
+      "logps/rejected": -32.02156066894531,
+      "loss": 0.805,
+      "rewards/accuracies": 0.574999988079071,
+      "rewards/chosen": 0.10399170964956284,
+      "rewards/margins": 0.28638824820518494,
+      "rewards/rejected": -0.1823965609073639,
       "step": 140
     },
     {
       "epoch": 0.39,
-      "grad_norm": 48.0,
+      "grad_norm": 34.25,
       "learning_rate": 3.834196265035119e-06,
-      "logits/chosen": -2.816462516784668,
-      "logits/rejected": -2.810920000076294,
-      "logps/chosen": -27.263103485107422,
-      "logps/rejected": -31.520715713500977,
-      "loss": 0.6276,
-      "rewards/accuracies": 0.7124999761581421,
-      "rewards/chosen": 0.4499203562736511,
-      "rewards/margins": 0.6875794529914856,
-      "rewards/rejected": -0.2376590520143509,
+      "logits/chosen": -2.8219058513641357,
+      "logits/rejected": -2.816573143005371,
+      "logps/chosen": -27.304168701171875,
+      "logps/rejected": -31.6512451171875,
+      "loss": 0.6938,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.26066717505455017,
+      "rewards/margins": 0.4744674563407898,
+      "rewards/rejected": -0.21380028128623962,
       "step": 150
     },
     {
       "epoch": 0.42,
-      "grad_norm": 68.5,
+      "grad_norm": 47.0,
       "learning_rate": 3.636998309800573e-06,
-      "logits/chosen": -3.133582592010498,
-      "logits/rejected": -3.1159331798553467,
-      "logps/chosen": -31.7524471282959,
-      "logps/rejected": -29.40524673461914,
-      "loss": 0.547,
-      "rewards/accuracies": 0.7124999761581421,
-      "rewards/chosen": 0.5914198756217957,
-      "rewards/margins": 0.9317981600761414,
-      "rewards/rejected": -0.3403782546520233,
+      "logits/chosen": -3.1381523609161377,
+      "logits/rejected": -3.121229648590088,
+      "logps/chosen": -31.75846290588379,
+      "logps/rejected": -29.593923568725586,
+      "loss": 0.5556,
+      "rewards/accuracies": 0.7250000238418579,
+      "rewards/chosen": 0.36662742495536804,
+      "rewards/margins": 0.6737016439437866,
+      "rewards/rejected": -0.30707424879074097,
       "step": 160
     },
     {
       "epoch": 0.44,
-      "grad_norm": 58.25,
+      "grad_norm": 41.75,
       "learning_rate": 3.4304331721118078e-06,
-      "logits/chosen": -2.944953203201294,
-      "logits/rejected": -2.952117681503296,
-      "logps/chosen": -29.341304779052734,
-      "logps/rejected": -31.5146427154541,
-      "loss": 0.6062,
-      "rewards/accuracies": 0.699999988079071,
-      "rewards/chosen": 0.4426456391811371,
-      "rewards/margins": 0.7747727632522583,
-      "rewards/rejected": -0.33212706446647644,
+      "logits/chosen": -2.9501750469207764,
+      "logits/rejected": -2.957343578338623,
+      "logps/chosen": -29.471771240234375,
+      "logps/rejected": -31.817256927490234,
+      "loss": 0.6244,
+      "rewards/accuracies": 0.7749999761581421,
+      "rewards/chosen": 0.21142145991325378,
+      "rewards/margins": 0.570308268070221,
+      "rewards/rejected": -0.35888683795928955,
       "step": 170
     },
     {
       "epoch": 0.47,
-      "grad_norm": 57.25,
+      "grad_norm": 37.25,
       "learning_rate": 3.2162026428305436e-06,
-      "logits/chosen": -2.795180082321167,
-      "logits/rejected": -2.792935848236084,
-      "logps/chosen": -29.07159423828125,
-      "logps/rejected": -30.085384368896484,
-      "loss": 0.5999,
-      "rewards/accuracies": 0.7124999761581421,
-      "rewards/chosen": 0.47860392928123474,
-      "rewards/margins": 0.7323731184005737,
-      "rewards/rejected": -0.2537691593170166,
+      "logits/chosen": -2.802713394165039,
+      "logits/rejected": -2.8004953861236572,
+      "logps/chosen": -29.250396728515625,
+      "logps/rejected": -30.246570587158203,
+      "loss": 0.6894,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": 0.20972609519958496,
+      "rewards/margins": 0.44892430305480957,
+      "rewards/rejected": -0.2391982078552246,
       "step": 180
     },
     {
       "epoch": 0.49,
-      "grad_norm": 35.5,
+      "grad_norm": 26.25,
       "learning_rate": 2.996071664294641e-06,
-      "logits/chosen": -2.9086050987243652,
-      "logits/rejected": -2.9050517082214355,
-      "logps/chosen": -29.76633644104004,
-      "logps/rejected": -28.544025421142578,
-      "loss": 0.778,
-      "rewards/accuracies": 0.7124999761581421,
-      "rewards/chosen": 0.3363017439842224,
-      "rewards/margins": 0.4854954779148102,
-      "rewards/rejected": -0.14919371902942657,
+      "logits/chosen": -2.9141011238098145,
+      "logits/rejected": -2.9099700450897217,
+      "logps/chosen": -29.81064796447754,
+      "logps/rejected": -28.773019790649414,
+      "loss": 0.7547,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": 0.18803250789642334,
+      "rewards/margins": 0.3957784175872803,
+      "rewards/rejected": -0.20774583518505096,
       "step": 190
     },
     {
       "epoch": 0.52,
-      "grad_norm": 51.0,
+      "grad_norm": 27.625,
       "learning_rate": 2.7718537898066833e-06,
-      "logits/chosen": -2.9782276153564453,
-      "logits/rejected": -2.9666411876678467,
-      "logps/chosen": -32.82664108276367,
-      "logps/rejected": -30.458984375,
-      "loss": 0.724,
+      "logits/chosen": -2.9808688163757324,
+      "logits/rejected": -2.9687676429748535,
+      "logps/chosen": -32.98746871948242,
+      "logps/rejected": -30.666748046875,
+      "loss": 0.7163,
       "rewards/accuracies": 0.6875,
-      "rewards/chosen": 0.6979023218154907,
-      "rewards/margins": 0.7178188562393188,
-      "rewards/rejected": -0.01991647481918335,
+      "rewards/chosen": 0.3557746112346649,
+      "rewards/margins": 0.4721054136753082,
+      "rewards/rejected": -0.11633072048425674,
       "step": 200
     },
     {
       "epoch": 0.52,
-      "eval_logits/chosen": -2.8178980350494385,
-      "eval_logits/rejected": -2.815643548965454,
-      "eval_logps/chosen": -31.38960838317871,
-      "eval_logps/rejected": -34.92823791503906,
-      "eval_loss": 0.9140273332595825,
-      "eval_rewards/accuracies": 0.565614640712738,
-      "eval_rewards/chosen": -0.08572381734848022,
-      "eval_rewards/margins": 0.09834489226341248,
-      "eval_rewards/rejected": -0.1840687096118927,
-      "eval_runtime": 113.15,
-      "eval_samples_per_second": 3.031,
+      "eval_logits/chosen": -2.819563388824463,
+      "eval_logits/rejected": -2.817270278930664,
+      "eval_logps/chosen": -31.472963333129883,
+      "eval_logps/rejected": -35.05440902709961,
+      "eval_loss": 0.9220677018165588,
+      "eval_rewards/accuracies": 0.5514950156211853,
+      "eval_rewards/chosen": -0.09525591880083084,
+      "eval_rewards/margins": 0.0828702449798584,
+      "eval_rewards/rejected": -0.17812614142894745,
+      "eval_runtime": 113.0963,
+      "eval_samples_per_second": 3.033,
       "eval_steps_per_second": 0.38,
       "step": 200
     },
     {
       "epoch": 0.55,
-      "grad_norm": 52.5,
+      "grad_norm": 37.0,
       "learning_rate": 2.5453962426402006e-06,
-      "logits/chosen": -2.9144248962402344,
-      "logits/rejected": -2.914703845977783,
-      "logps/chosen": -32.25563049316406,
-      "logps/rejected": -34.11750793457031,
-      "loss": 0.5504,
-      "rewards/accuracies": 0.8125,
-      "rewards/chosen": 0.6222246289253235,
-      "rewards/margins": 0.8270590901374817,
-      "rewards/rejected": -0.2048344612121582,
+      "logits/chosen": -2.9180774688720703,
+      "logits/rejected": -2.91807222366333,
+      "logps/chosen": -32.39319610595703,
+      "logps/rejected": -34.32632064819336,
+      "loss": 0.6144,
+      "rewards/accuracies": 0.7250000238418579,
+      "rewards/chosen": 0.3201075494289398,
+      "rewards/margins": 0.552533745765686,
+      "rewards/rejected": -0.23242616653442383,
       "step": 210
     },
     {
       "epoch": 0.57,
-      "grad_norm": 47.5,
+      "grad_norm": 30.25,
       "learning_rate": 2.3185646976551794e-06,
-      "logits/chosen": -2.8935537338256836,
-      "logits/rejected": -2.909308671951294,
-      "logps/chosen": -29.393646240234375,
-      "logps/rejected": -28.79998779296875,
-      "loss": 0.6005,
-      "rewards/accuracies": 0.6875,
-      "rewards/chosen": 0.6275160908699036,
-      "rewards/margins": 0.8448917269706726,
-      "rewards/rejected": -0.21737566590309143,
+      "logits/chosen": -2.8982133865356445,
+      "logits/rejected": -2.9134175777435303,
+      "logps/chosen": -29.346981048583984,
+      "logps/rejected": -28.9754581451416,
+      "loss": 0.6305,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": 0.415529727935791,
+      "rewards/margins": 0.6391232013702393,
+      "rewards/rejected": -0.22359342873096466,
       "step": 220
     },
     {
       "epoch": 0.6,
-      "grad_norm": 46.5,
+      "grad_norm": 35.0,
       "learning_rate": 2.0932279108998323e-06,
-      "logits/chosen": -2.9393625259399414,
-      "logits/rejected": -2.9434802532196045,
-      "logps/chosen": -30.659032821655273,
-      "logps/rejected": -31.946239471435547,
-      "loss": 0.7369,
+      "logits/chosen": -2.9441702365875244,
+      "logits/rejected": -2.948596477508545,
+      "logps/chosen": -30.654144287109375,
+      "logps/rejected": -32.061561584472656,
+      "loss": 0.7505,
       "rewards/accuracies": 0.637499988079071,
-      "rewards/chosen": 0.39426764845848083,
-      "rewards/margins": 0.5447386503219604,
-      "rewards/rejected": -0.15047098696231842,
+      "rewards/chosen": 0.2488628625869751,
+      "rewards/margins": 0.4005666673183441,
+      "rewards/rejected": -0.15170380473136902,
       "step": 230
     },
     {
       "epoch": 0.62,
-      "grad_norm": 46.75,
+      "grad_norm": 41.5,
       "learning_rate": 1.8712423238279358e-06,
-      "logits/chosen": -2.99426531791687,
-      "logits/rejected": -3.00126314163208,
-      "logps/chosen": -29.958850860595703,
-      "logps/rejected": -30.447010040283203,
-      "loss": 0.4564,
-      "rewards/accuracies": 0.800000011920929,
-      "rewards/chosen": 0.7568598985671997,
-      "rewards/margins": 0.9262750744819641,
-      "rewards/rejected": -0.16941508650779724,
+      "logits/chosen": -2.9988656044006348,
+      "logits/rejected": -3.006716251373291,
+      "logps/chosen": -30.050405502319336,
+      "logps/rejected": -30.539688110351562,
+      "loss": 0.5427,
+      "rewards/accuracies": 0.8125,
+      "rewards/chosen": 0.42725998163223267,
+      "rewards/margins": 0.5794823169708252,
+      "rewards/rejected": -0.1522223949432373,
       "step": 240
     },
     {
       "epoch": 0.65,
-      "grad_norm": 55.0,
+      "grad_norm": 39.75,
       "learning_rate": 1.6544367689701824e-06,
-      "logits/chosen": -2.8251967430114746,
-      "logits/rejected": -2.815450429916382,
-      "logps/chosen": -26.474361419677734,
-      "logps/rejected": -29.600570678710938,
-      "loss": 0.721,
-      "rewards/accuracies": 0.762499988079071,
-      "rewards/chosen": 0.3838837444782257,
-      "rewards/margins": 0.43902724981307983,
-      "rewards/rejected": -0.05514346435666084,
+      "logits/chosen": -2.8314945697784424,
+      "logits/rejected": -2.8215582370758057,
+      "logps/chosen": -26.297893524169922,
+      "logps/rejected": -29.683481216430664,
+      "loss": 0.737,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": 0.32816022634506226,
+      "rewards/margins": 0.4040789008140564,
+      "rewards/rejected": -0.07591867446899414,
       "step": 250
     },
     {
       "epoch": 0.68,
-      "grad_norm": 21.75,
+      "grad_norm": 24.875,
       "learning_rate": 1.4445974030621963e-06,
-      "logits/chosen": -2.8065786361694336,
-      "logits/rejected": -2.8274922370910645,
-      "logps/chosen": -28.88662338256836,
-      "logps/rejected": -34.49494934082031,
-      "loss": 0.4422,
-      "rewards/accuracies": 0.862500011920929,
-      "rewards/chosen": 0.7546705603599548,
-      "rewards/margins": 1.0113604068756104,
-      "rewards/rejected": -0.25668981671333313,
+      "logits/chosen": -2.8124470710754395,
+      "logits/rejected": -2.83256196975708,
+      "logps/chosen": -28.941558837890625,
+      "logps/rejected": -34.62192916870117,
+      "loss": 0.5683,
+      "rewards/accuracies": 0.8125,
+      "rewards/chosen": 0.44420257210731506,
+      "rewards/margins": 0.6681241989135742,
+      "rewards/rejected": -0.22392161190509796,
       "step": 260
     },
     {
       "epoch": 0.7,
-      "grad_norm": 60.0,
+      "grad_norm": 44.25,
       "learning_rate": 1.243452991757889e-06,
-      "logits/chosen": -2.948761463165283,
-      "logits/rejected": -2.955237627029419,
-      "logps/chosen": -30.040813446044922,
-      "logps/rejected": -30.221237182617188,
-      "loss": 0.5322,
-      "rewards/accuracies": 0.7749999761581421,
-      "rewards/chosen": 0.5924339890480042,
-      "rewards/margins": 0.8772123456001282,
-      "rewards/rejected": -0.2847784161567688,
+      "logits/chosen": -2.954413890838623,
+      "logits/rejected": -2.9607253074645996,
+      "logps/chosen": -30.094287872314453,
+      "logps/rejected": -30.396358489990234,
+      "loss": 0.5618,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": 0.3435356020927429,
+      "rewards/margins": 0.6090848445892334,
+      "rewards/rejected": -0.2655491232872009,
       "step": 270
     },
     {
       "epoch": 0.73,
-      "grad_norm": 29.375,
+      "grad_norm": 30.75,
       "learning_rate": 1.0526606671603523e-06,
-      "logits/chosen": -2.9630208015441895,
-      "logits/rejected": -2.949868679046631,
-      "logps/chosen": -30.0323429107666,
-      "logps/rejected": -28.719945907592773,
-      "loss": 0.7384,
+      "logits/chosen": -2.9698939323425293,
+      "logits/rejected": -2.956036329269409,
+      "logps/chosen": -30.114521026611328,
+      "logps/rejected": -28.823684692382812,
+      "loss": 0.7337,
       "rewards/accuracies": 0.7749999761581421,
-      "rewards/chosen": 0.5514736771583557,
-      "rewards/margins": 0.6223500967025757,
-      "rewards/rejected": -0.07087641209363937,
+      "rewards/chosen": 0.30358150601387024,
+      "rewards/margins": 0.399749219417572,
+      "rewards/rejected": -0.09616778045892715,
       "step": 280
     },
     {
       "epoch": 0.75,
-      "grad_norm": 36.5,
+      "grad_norm": 29.375,
       "learning_rate": 8.737922755071455e-07,
-      "logits/chosen": -2.890263080596924,
-      "logits/rejected": -2.872467279434204,
-      "logps/chosen": -31.614023208618164,
-      "logps/rejected": -30.969629287719727,
-      "loss": 0.4081,
+      "logits/chosen": -2.8976500034332275,
+      "logits/rejected": -2.8795270919799805,
+      "logps/chosen": -31.536123275756836,
+      "logps/rejected": -31.159521102905273,
+      "loss": 0.4628,
       "rewards/accuracies": 0.8374999761581421,
-      "rewards/chosen": 0.8178254961967468,
-      "rewards/margins": 1.1748807430267334,
-      "rewards/rejected": -0.3570552468299866,
+      "rewards/chosen": 0.5500896573066711,
+      "rewards/margins": 0.8681961297988892,
+      "rewards/rejected": -0.31810641288757324,
       "step": 290
     },
     {
       "epoch": 0.78,
-      "grad_norm": 47.0,
+      "grad_norm": 30.25,
       "learning_rate": 7.08321427484816e-07,
-      "logits/chosen": -2.892866611480713,
-      "logits/rejected": -2.8899810314178467,
-      "logps/chosen": -31.352060317993164,
-      "logps/rejected": -27.79092025756836,
-      "loss": 0.6408,
+      "logits/chosen": -2.89839506149292,
+      "logits/rejected": -2.8952107429504395,
+      "logps/chosen": -31.317790985107422,
+      "logps/rejected": -27.9440860748291,
+      "loss": 0.6439,
       "rewards/accuracies": 0.737500011920929,
-      "rewards/chosen": 0.6964761018753052,
-      "rewards/margins": 0.8463441729545593,
-      "rewards/rejected": -0.14986807107925415,
+      "rewards/chosen": 0.45242971181869507,
+      "rewards/margins": 0.6226793527603149,
+      "rewards/rejected": -0.17024962604045868,
       "step": 300
     },
     {
       "epoch": 0.78,
-      "eval_logits/chosen": -2.810232400894165,
-      "eval_logits/rejected": -2.807687520980835,
-      "eval_logps/chosen": -31.392431259155273,
-      "eval_logps/rejected": -34.949954986572266,
-      "eval_loss": 0.9091227650642395,
-      "eval_rewards/accuracies": 0.5627076625823975,
-      "eval_rewards/chosen": -0.08798420429229736,
-      "eval_rewards/margins": 0.11345873028039932,
-      "eval_rewards/rejected": -0.20144294202327728,
-      "eval_runtime": 113.1296,
-      "eval_samples_per_second": 3.032,
+      "eval_logits/chosen": -2.8154501914978027,
+      "eval_logits/rejected": -2.8133084774017334,
+      "eval_logps/chosen": -31.455341339111328,
+      "eval_logps/rejected": -35.035438537597656,
+      "eval_loss": 0.9215981364250183,
+      "eval_rewards/accuracies": 0.5573089718818665,
+      "eval_rewards/chosen": -0.0864456370472908,
+      "eval_rewards/margins": 0.08219624310731888,
+      "eval_rewards/rejected": -0.16864188015460968,
+      "eval_runtime": 113.1056,
+      "eval_samples_per_second": 3.033,
       "eval_steps_per_second": 0.38,
       "step": 300
     },
     {
       "epoch": 0.81,
-      "grad_norm": 57.5,
+      "grad_norm": 39.5,
       "learning_rate": 5.576113578589035e-07,
-      "logits/chosen": -2.771563768386841,
-      "logits/rejected": -2.790158748626709,
-      "logps/chosen": -28.552204132080078,
-      "logps/rejected": -31.123676300048828,
-      "loss": 0.4625,
-      "rewards/accuracies": 0.8374999761581421,
-      "rewards/chosen": 0.5417758226394653,
-      "rewards/margins": 0.855111300945282,
-      "rewards/rejected": -0.31333547830581665,
+      "logits/chosen": -2.779205560684204,
+      "logits/rejected": -2.7966835498809814,
+      "logps/chosen": -28.479604721069336,
+      "logps/rejected": -31.288341522216797,
+      "loss": 0.5317,
+      "rewards/accuracies": 0.875,
+      "rewards/chosen": 0.37490805983543396,
+      "rewards/margins": 0.6530753374099731,
+      "rewards/rejected": -0.2781672477722168,
       "step": 310
     },
     {
       "epoch": 0.83,
-      "grad_norm": 50.75,
+      "grad_norm": 32.25,
       "learning_rate": 4.229036944380913e-07,
-      "logits/chosen": -3.01896333694458,
-      "logits/rejected": -3.0036964416503906,
-      "logps/chosen": -29.334665298461914,
-      "logps/rejected": -28.647085189819336,
-      "loss": 0.4327,
-      "rewards/accuracies": 0.824999988079071,
-      "rewards/chosen": 0.7015351057052612,
-      "rewards/margins": 0.9641841650009155,
-      "rewards/rejected": -0.2626491189002991,
+      "logits/chosen": -3.0243513584136963,
+      "logits/rejected": -3.0100343227386475,
+      "logps/chosen": -29.37557601928711,
+      "logps/rejected": -28.844478607177734,
+      "loss": 0.5359,
+      "rewards/accuracies": 0.7875000238418579,
+      "rewards/chosen": 0.41800540685653687,
+      "rewards/margins": 0.6808558106422424,
+      "rewards/rejected": -0.262850284576416,
       "step": 320
     },
     {
       "epoch": 0.86,
-      "grad_norm": 19.25,
+      "grad_norm": 22.875,
       "learning_rate": 3.053082288996112e-07,
-      "logits/chosen": -2.9340624809265137,
-      "logits/rejected": -2.916344404220581,
-      "logps/chosen": -27.29937744140625,
-      "logps/rejected": -30.881011962890625,
-      "loss": 0.3934,
-      "rewards/accuracies": 0.8500000238418579,
-      "rewards/chosen": 0.8501566052436829,
-      "rewards/margins": 1.2416493892669678,
-      "rewards/rejected": -0.3914927840232849,
+      "logits/chosen": -2.9381539821624756,
+      "logits/rejected": -2.920642852783203,
+      "logps/chosen": -27.252010345458984,
+      "logps/rejected": -31.086780548095703,
+      "loss": 0.4505,
+      "rewards/accuracies": 0.8374999761581421,
+      "rewards/chosen": 0.5550299882888794,
+      "rewards/margins": 0.9025982022285461,
+      "rewards/rejected": -0.3475683331489563,
       "step": 330
     },
     {
       "epoch": 0.88,
-      "grad_norm": 54.0,
+      "grad_norm": 42.75,
       "learning_rate": 2.0579377374915805e-07,
-      "logits/chosen": -3.1503987312316895,
-      "logits/rejected": -3.156454563140869,
-      "logps/chosen": -30.917760848999023,
-      "logps/rejected": -33.217864990234375,
-      "loss": 0.4822,
-      "rewards/accuracies": 0.7749999761581421,
-      "rewards/chosen": 0.5771540403366089,
-      "rewards/margins": 0.994644820690155,
-      "rewards/rejected": -0.4174906611442566,
+      "logits/chosen": -3.1556878089904785,
+      "logits/rejected": -3.1616902351379395,
+      "logps/chosen": -30.897449493408203,
+      "logps/rejected": -33.362003326416016,
+      "loss": 0.532,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": 0.3708776533603668,
+      "rewards/margins": 0.7038747072219849,
+      "rewards/rejected": -0.33299699425697327,
       "step": 340
     },
     {
       "epoch": 0.91,
-      "grad_norm": 27.0,
+      "grad_norm": 36.25,
       "learning_rate": 1.2518018074041684e-07,
-      "logits/chosen": -3.026543140411377,
-      "logits/rejected": -3.0300345420837402,
-      "logps/chosen": -29.901714324951172,
-      "logps/rejected": -31.784900665283203,
-      "loss": 0.5385,
-      "rewards/accuracies": 0.762499988079071,
-      "rewards/chosen": 0.7241344451904297,
-      "rewards/margins": 0.9414092302322388,
-      "rewards/rejected": -0.2172747403383255,
+      "logits/chosen": -3.032553195953369,
+      "logits/rejected": -3.0359413623809814,
+      "logps/chosen": -29.842823028564453,
+      "logps/rejected": -31.924610137939453,
+      "loss": 0.5238,
+      "rewards/accuracies": 0.7749999761581421,
+      "rewards/chosen": 0.4820283055305481,
+      "rewards/margins": 0.6876784563064575,
+      "rewards/rejected": -0.2056502103805542,
       "step": 350
     },
     {
       "epoch": 0.94,
-      "grad_norm": 47.25,
+      "grad_norm": 36.25,
       "learning_rate": 6.41315865106129e-08,
-      "logits/chosen": -2.8658933639526367,
-      "logits/rejected": -2.8671889305114746,
-      "logps/chosen": -27.54721450805664,
-      "logps/rejected": -30.053890228271484,
-      "loss": 0.4298,
-      "rewards/accuracies": 0.800000011920929,
-      "rewards/chosen": 0.8820101022720337,
-      "rewards/margins": 1.0603699684143066,
-      "rewards/rejected": -0.17835985124111176,
+      "logits/chosen": -2.871373414993286,
+      "logits/rejected": -2.8733484745025635,
+      "logps/chosen": -27.545650482177734,
+      "logps/rejected": -30.173965454101562,
+      "loss": 0.5016,
+      "rewards/accuracies": 0.7875000238418579,
+      "rewards/chosen": 0.5520386099815369,
+      "rewards/margins": 0.7235507965087891,
+      "rewards/rejected": -0.1715122014284134,
       "step": 360
     },
     {
       "epoch": 0.96,
-      "grad_norm": 62.75,
+      "grad_norm": 50.75,
       "learning_rate": 2.3150941078050325e-08,
-      "logits/chosen": -2.9409708976745605,
-      "logits/rejected": -2.9394469261169434,
-      "logps/chosen": -29.869876861572266,
-      "logps/rejected": -32.248016357421875,
-      "loss": 0.486,
+      "logits/chosen": -2.948784589767456,
+      "logits/rejected": -2.9460253715515137,
+      "logps/chosen": -29.852313995361328,
+      "logps/rejected": -32.32441329956055,
+      "loss": 0.5601,
       "rewards/accuracies": 0.800000011920929,
-      "rewards/chosen": 0.7334206700325012,
-      "rewards/margins": 0.9893990755081177,
-      "rewards/rejected": -0.25597840547561646,
+      "rewards/chosen": 0.46716976165771484,
+      "rewards/margins": 0.6653568148612976,
+      "rewards/rejected": -0.19818702340126038,
       "step": 370
     },
     {
       "epoch": 0.99,
-      "grad_norm": 56.5,
+      "grad_norm": 37.5,
       "learning_rate": 2.575864278703266e-09,
-      "logits/chosen": -2.8988537788391113,
-      "logits/rejected": -2.8815865516662598,
-      "logps/chosen": -28.09465980529785,
-      "logps/rejected": -28.32416343688965,
-      "loss": 0.4149,
-      "rewards/accuracies": 0.875,
-      "rewards/chosen": 0.6803036332130432,
-      "rewards/margins": 1.027940273284912,
-      "rewards/rejected": -0.3476366698741913,
+      "logits/chosen": -2.9041707515716553,
+      "logits/rejected": -2.886718273162842,
+      "logps/chosen": -28.160015106201172,
+      "logps/rejected": -28.389995574951172,
+      "loss": 0.5382,
+      "rewards/accuracies": 0.8125,
+      "rewards/chosen": 0.3925102651119232,
+      "rewards/margins": 0.6426987648010254,
+      "rewards/rejected": -0.25018855929374695,
       "step": 380
     },
     {
       "epoch": 1.0,
       "step": 385,
       "total_flos": 0.0,
-      "train_loss": 0.6588526527602951,
-      "train_runtime": 2719.293,
-      "train_samples_per_second": 1.132,
+      "train_loss": 0.694122297732861,
+      "train_runtime": 2717.9333,
+      "train_samples_per_second": 1.133,
       "train_steps_per_second": 0.142
     }
   ],