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: 0.6292
-- Rewards/chosen: 0.0553
-- Rewards/rejected: 0.0226
-- Rewards/accuracies: 0.5166
-- Rewards/margins: 0.0327
-- Logps/rejected: -34.6659
-- Logps/chosen: -31.2035
-- Logits/rejected: -2.8063
-- Logits/chosen: -2.8088
+- Loss: 1.6904
+- Rewards/chosen: 0.0126
+- Rewards/rejected: -0.0056
+- Rewards/accuracies: 0.5573
+- Rewards/margins: 0.0181
+- Logps/rejected: -34.7121
+- Logps/chosen: -31.2510
+- Logits/rejected: -2.8041
+- Logits/chosen: -2.8065
 
 ## 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 |
 |:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
-| 1.1043        | 0.26  | 100  | 0.6260          | 0.0189         | -0.0071          | 0.5370             | 0.0260          | -34.7083       | -31.2554     | -2.8132         | -2.8159       |
-| 3.0672        | 0.52  | 200  | 0.6356          | 0.0453         | 0.0067           | 0.5627             | 0.0386          | -34.6886       | -31.2177     | -2.8107         | -2.8128       |
-| 1.2353        | 0.78  | 300  | 0.6302          | 0.0524         | 0.0270           | 0.5311             | 0.0254          | -34.6596       | -31.2076     | -2.8071         | -2.8097       |
+| 1.7948        | 0.26  | 100  | 1.6241          | 0.0102         | -0.0177          | 0.5685             | 0.0278          | -34.7424       | -31.2571     | -2.8147         | -2.8176       |
+| 2.6629        | 0.52  | 200  | 1.5649          | 0.0182         | -0.0143          | 0.5457             | 0.0325          | -34.7338       | -31.2369     | -2.8095         | -2.8121       |
+| 1.5067        | 0.78  | 300  | 1.6904          | 0.0126         | -0.0056          | 0.5573             | 0.0181          | -34.7121       | -31.2510     | -2.8041         | -2.8065       |
 
 
 ### Framework versions

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:03e35a31a2592142bc2c19ac2824b77733933c82f1d6ec64c539c85894c4da86
+oid sha256:c5ca5d5e4b54d37e89a137afb5f481e02b1f61d6b3efb3c4429b6de7314642dd
 size 671150064

all_results.json

@@ -1,21 +1,8 @@
 {
     "epoch": 1.0,
-    "eval_logits/chosen": -2.808791399002075,
-    "eval_logits/rejected": -2.8062777519226074,
-    "eval_logps/chosen": -31.203502655029297,
-    "eval_logps/rejected": -34.66593933105469,
-    "eval_loss": 0.629190981388092,
-    "eval_rewards/accuracies": 0.5166113376617432,
-    "eval_rewards/chosen": 0.05526338145136833,
-    "eval_rewards/margins": 0.03271109238266945,
-    "eval_rewards/rejected": 0.022552290931344032,
-    "eval_runtime": 112.7274,
-    "eval_samples": 343,
-    "eval_samples_per_second": 3.043,
-    "eval_steps_per_second": 0.381,
-    "train_loss": 1.2163990609057538,
-    "train_runtime": 2720.9807,
+    "train_loss": 1.7377146052075672,
+    "train_runtime": 2719.6001,
     "train_samples": 3079,
     "train_samples_per_second": 1.132,
-    "train_steps_per_second": 0.141
+    "train_steps_per_second": 0.142
 }

train_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 1.0,
-    "train_loss": 1.2163990609057538,
-    "train_runtime": 2720.9807,
+    "train_loss": 1.7377146052075672,
+    "train_runtime": 2719.6001,
     "train_samples": 3079,
     "train_samples_per_second": 1.132,
-    "train_steps_per_second": 0.141
+    "train_steps_per_second": 0.142
 }

trainer_state.json

@@ -10,13 +10,13 @@
   "log_history": [
     {
       "epoch": 0.0,
-      "grad_norm": 113.0,
+      "grad_norm": 198.0,
       "learning_rate": 1.282051282051282e-07,
       "logits/chosen": -2.7358343601226807,
       "logits/rejected": -2.7480404376983643,
       "logps/chosen": -27.35565757751465,
       "logps/rejected": -21.06114387512207,
-      "loss": 0.5102,
+      "loss": 1.5625,
       "rewards/accuracies": 0.0,
       "rewards/chosen": 0.0,
       "rewards/margins": 0.0,
@@ -25,630 +25,630 @@
     },
     {
       "epoch": 0.03,
-      "grad_norm": 146.0,
+      "grad_norm": 306.0,
       "learning_rate": 1.282051282051282e-06,
-      "logits/chosen": -3.0090532302856445,
-      "logits/rejected": -2.998255729675293,
-      "logps/chosen": -33.17539596557617,
-      "logps/rejected": -31.967647552490234,
-      "loss": 0.5411,
-      "rewards/accuracies": 0.4722222089767456,
-      "rewards/chosen": 0.0038858253974467516,
-      "rewards/margins": 0.006502790376543999,
-      "rewards/rejected": -0.0026169654447585344,
+      "logits/chosen": -3.009331703186035,
+      "logits/rejected": -2.9980063438415527,
+      "logps/chosen": -33.193904876708984,
+      "logps/rejected": -32.01276779174805,
+      "loss": 1.5179,
+      "rewards/accuracies": 0.5416666865348816,
+      "rewards/chosen": -0.005185974761843681,
+      "rewards/margins": 0.01435802225023508,
+      "rewards/rejected": -0.019543997943401337,
       "step": 10
     },
     {
       "epoch": 0.05,
-      "grad_norm": 144.0,
+      "grad_norm": 268.0,
       "learning_rate": 2.564102564102564e-06,
-      "logits/chosen": -2.89949369430542,
-      "logits/rejected": -2.8941283226013184,
-      "logps/chosen": -32.47248077392578,
-      "logps/rejected": -28.952869415283203,
-      "loss": 0.6026,
-      "rewards/accuracies": 0.5,
-      "rewards/chosen": 0.0018292926251888275,
-      "rewards/margins": -0.018470000475645065,
-      "rewards/rejected": 0.02029929682612419,
+      "logits/chosen": -2.8995871543884277,
+      "logits/rejected": -2.894530773162842,
+      "logps/chosen": -32.46534729003906,
+      "logps/rejected": -28.974437713623047,
+      "loss": 1.6084,
+      "rewards/accuracies": 0.5249999761581421,
+      "rewards/chosen": 0.0038995600771158934,
+      "rewards/margins": 0.0009288033470511436,
+      "rewards/rejected": 0.002970754634588957,
       "step": 20
     },
     {
       "epoch": 0.08,
-      "grad_norm": 159.0,
+      "grad_norm": 221.0,
       "learning_rate": 3.846153846153847e-06,
-      "logits/chosen": -3.096463680267334,
-      "logits/rejected": -3.1078667640686035,
-      "logps/chosen": -32.81493377685547,
-      "logps/rejected": -30.128747940063477,
-      "loss": 0.5658,
-      "rewards/accuracies": 0.4375,
-      "rewards/chosen": 0.06562719494104385,
-      "rewards/margins": 0.018515314906835556,
-      "rewards/rejected": 0.047111887484788895,
+      "logits/chosen": -3.097817897796631,
+      "logits/rejected": -3.1089720726013184,
+      "logps/chosen": -32.821983337402344,
+      "logps/rejected": -30.1259708404541,
+      "loss": 1.6124,
+      "rewards/accuracies": 0.4625000059604645,
+      "rewards/chosen": 0.0346810482442379,
+      "rewards/margins": 0.006648649461567402,
+      "rewards/rejected": 0.028032397851347923,
       "step": 30
     },
     {
       "epoch": 0.1,
-      "grad_norm": 223.0,
+      "grad_norm": 225.0,
       "learning_rate": 4.999896948438434e-06,
-      "logits/chosen": -2.8619472980499268,
-      "logits/rejected": -2.8528523445129395,
-      "logps/chosen": -31.569528579711914,
-      "logps/rejected": -32.359840393066406,
-      "loss": 0.8775,
-      "rewards/accuracies": 0.6000000238418579,
-      "rewards/chosen": 0.1825559139251709,
-      "rewards/margins": 0.18422597646713257,
-      "rewards/rejected": -0.0016700520645827055,
+      "logits/chosen": -2.8635919094085693,
+      "logits/rejected": -2.8542821407318115,
+      "logps/chosen": -31.561412811279297,
+      "logps/rejected": -32.368797302246094,
+      "loss": 1.5659,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": 0.10756368935108185,
+      "rewards/margins": 0.11209961026906967,
+      "rewards/rejected": -0.004535915795713663,
       "step": 40
     },
     {
       "epoch": 0.13,
-      "grad_norm": 156.0,
+      "grad_norm": 268.0,
       "learning_rate": 4.987541037542187e-06,
-      "logits/chosen": -2.8859169483184814,
-      "logits/rejected": -2.8837780952453613,
-      "logps/chosen": -29.52107810974121,
-      "logps/rejected": -30.1494083404541,
-      "loss": 0.7059,
-      "rewards/accuracies": 0.6625000238418579,
-      "rewards/chosen": 0.1533588469028473,
-      "rewards/margins": 0.1956842839717865,
-      "rewards/rejected": -0.042325448244810104,
+      "logits/chosen": -2.886489152908325,
+      "logits/rejected": -2.884337902069092,
+      "logps/chosen": -29.50502586364746,
+      "logps/rejected": -30.12764549255371,
+      "loss": 1.6791,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": 0.0940544530749321,
+      "rewards/margins": 0.10953446477651596,
+      "rewards/rejected": -0.015479998663067818,
       "step": 50
     },
     {
       "epoch": 0.16,
-      "grad_norm": 150.0,
+      "grad_norm": 217.0,
       "learning_rate": 4.954691471941119e-06,
-      "logits/chosen": -2.9192371368408203,
-      "logits/rejected": -2.9210586547851562,
-      "logps/chosen": -30.085376739501953,
-      "logps/rejected": -28.13214111328125,
-      "loss": 0.5449,
-      "rewards/accuracies": 0.637499988079071,
-      "rewards/chosen": 0.031074851751327515,
-      "rewards/margins": 0.1545926332473755,
-      "rewards/rejected": -0.12351777404546738,
+      "logits/chosen": -2.9178414344787598,
+      "logits/rejected": -2.9194931983947754,
+      "logps/chosen": -30.02047348022461,
+      "logps/rejected": -28.11154556274414,
+      "loss": 1.3051,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": 0.04371771961450577,
+      "rewards/margins": 0.10606314986944199,
+      "rewards/rejected": -0.06234544515609741,
       "step": 60
     },
     {
       "epoch": 0.18,
-      "grad_norm": 972.0,
+      "grad_norm": 1032.0,
       "learning_rate": 4.901618883413549e-06,
-      "logits/chosen": -3.001981496810913,
-      "logits/rejected": -3.009333610534668,
-      "logps/chosen": -29.32364273071289,
-      "logps/rejected": -31.037628173828125,
-      "loss": 1.1987,
-      "rewards/accuracies": 0.550000011920929,
-      "rewards/chosen": -0.029870102182030678,
-      "rewards/margins": 0.07220745086669922,
-      "rewards/rejected": -0.10207755863666534,
+      "logits/chosen": -3.0002479553222656,
+      "logits/rejected": -3.007493495941162,
+      "logps/chosen": -29.312774658203125,
+      "logps/rejected": -31.054119110107422,
+      "loss": 2.1422,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": -0.012723004445433617,
+      "rewards/margins": 0.05220501497387886,
+      "rewards/rejected": -0.06492801755666733,
       "step": 70
     },
     {
       "epoch": 0.21,
-      "grad_norm": 192.0,
+      "grad_norm": 326.0,
       "learning_rate": 4.828760511501322e-06,
-      "logits/chosen": -2.816981792449951,
-      "logits/rejected": -2.8327014446258545,
-      "logps/chosen": -29.412933349609375,
-      "logps/rejected": -29.788555145263672,
-      "loss": 0.6831,
+      "logits/chosen": -2.814807415008545,
+      "logits/rejected": -2.8305888175964355,
+      "logps/chosen": -29.402179718017578,
+      "logps/rejected": -29.830846786499023,
+      "loss": 1.5698,
       "rewards/accuracies": 0.612500011920929,
-      "rewards/chosen": 0.08945836126804352,
-      "rewards/margins": 0.16944995522499084,
-      "rewards/rejected": -0.07999160140752792,
+      "rewards/chosen": 0.05541967228055,
+      "rewards/margins": 0.11804421246051788,
+      "rewards/rejected": -0.06262453645467758,
       "step": 80
     },
     {
       "epoch": 0.23,
-      "grad_norm": 362.0,
+      "grad_norm": 322.0,
       "learning_rate": 4.7367166013034295e-06,
-      "logits/chosen": -2.9031848907470703,
-      "logits/rejected": -2.8844356536865234,
-      "logps/chosen": -32.63589859008789,
-      "logps/rejected": -30.085968017578125,
-      "loss": 4.9199,
-      "rewards/accuracies": 0.6499999761581421,
-      "rewards/chosen": 0.14037345349788666,
-      "rewards/margins": 0.18539710342884064,
-      "rewards/rejected": -0.04502364248037338,
+      "logits/chosen": -2.9036896228790283,
+      "logits/rejected": -2.885866641998291,
+      "logps/chosen": -32.615840911865234,
+      "logps/rejected": -30.156829833984375,
+      "loss": 5.9533,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": 0.08823554217815399,
+      "rewards/margins": 0.1423090249300003,
+      "rewards/rejected": -0.054073482751846313,
       "step": 90
     },
     {
       "epoch": 0.26,
-      "grad_norm": 157.0,
+      "grad_norm": 177.0,
       "learning_rate": 4.626245458345211e-06,
-      "logits/chosen": -3.0054373741149902,
-      "logits/rejected": -3.006031036376953,
-      "logps/chosen": -31.7724552154541,
-      "logps/rejected": -30.76922035217285,
-      "loss": 1.1043,
-      "rewards/accuracies": 0.5625,
-      "rewards/chosen": 0.19909459352493286,
-      "rewards/margins": 0.16622690856456757,
-      "rewards/rejected": 0.03286769241094589,
+      "logits/chosen": -3.007761001586914,
+      "logits/rejected": -3.0078930854797363,
+      "logps/chosen": -31.708675384521484,
+      "logps/rejected": -30.79708480834961,
+      "loss": 1.7948,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.1392797976732254,
+      "rewards/margins": 0.13164444267749786,
+      "rewards/rejected": 0.007635355927050114,
       "step": 100
     },
     {
       "epoch": 0.26,
-      "eval_logits/chosen": -2.815906524658203,
-      "eval_logits/rejected": -2.813185930252075,
-      "eval_logps/chosen": -31.25544548034668,
-      "eval_logps/rejected": -34.70828628540039,
-      "eval_loss": 0.6260043978691101,
-      "eval_rewards/accuracies": 0.5369601845741272,
-      "eval_rewards/chosen": 0.018902836367487907,
-      "eval_rewards/margins": 0.02599383145570755,
-      "eval_rewards/rejected": -0.00709099555388093,
-      "eval_runtime": 113.2921,
-      "eval_samples_per_second": 3.028,
-      "eval_steps_per_second": 0.38,
+      "eval_logits/chosen": -2.8176231384277344,
+      "eval_logits/rejected": -2.814725399017334,
+      "eval_logps/chosen": -31.257057189941406,
+      "eval_logps/rejected": -34.74238204956055,
+      "eval_loss": 1.6240766048431396,
+      "eval_rewards/accuracies": 0.5685215592384338,
+      "eval_rewards/chosen": 0.010158155113458633,
+      "eval_rewards/margins": 0.027848951518535614,
+      "eval_rewards/rejected": -0.01769079640507698,
+      "eval_runtime": 112.9872,
+      "eval_samples_per_second": 3.036,
+      "eval_steps_per_second": 0.381,
       "step": 100
     },
     {
       "epoch": 0.29,
-      "grad_norm": 157.0,
+      "grad_norm": 251.0,
       "learning_rate": 4.498257201263691e-06,
-      "logits/chosen": -2.9622676372528076,
-      "logits/rejected": -2.937613010406494,
-      "logps/chosen": -31.749774932861328,
-      "logps/rejected": -31.12813949584961,
-      "loss": 1.4773,
-      "rewards/accuracies": 0.699999988079071,
-      "rewards/chosen": 0.3269161581993103,
-      "rewards/margins": 0.30541908740997314,
-      "rewards/rejected": 0.02149704284965992,
+      "logits/chosen": -2.964388847351074,
+      "logits/rejected": -2.9403655529022217,
+      "logps/chosen": -31.717798233032227,
+      "logps/rejected": -31.185800552368164,
+      "loss": 1.9523,
+      "rewards/accuracies": 0.762499988079071,
+      "rewards/chosen": 0.19960005581378937,
+      "rewards/margins": 0.21038112044334412,
+      "rewards/rejected": -0.010781067423522472,
       "step": 110
     },
     {
       "epoch": 0.31,
-      "grad_norm": 109.0,
+      "grad_norm": 182.0,
       "learning_rate": 4.353806263777678e-06,
-      "logits/chosen": -3.043015718460083,
-      "logits/rejected": -3.0729401111602783,
-      "logps/chosen": -28.837779998779297,
-      "logps/rejected": -34.10157012939453,
-      "loss": 1.2172,
-      "rewards/accuracies": 0.6499999761581421,
-      "rewards/chosen": 0.22828666865825653,
-      "rewards/margins": 0.19610069692134857,
-      "rewards/rejected": 0.032185956835746765,
+      "logits/chosen": -3.0449271202087402,
+      "logits/rejected": -3.074241876602173,
+      "logps/chosen": -28.759485244750977,
+      "logps/rejected": -34.1177864074707,
+      "loss": 1.6223,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.1617673933506012,
+      "rewards/margins": 0.1498604714870453,
+      "rewards/rejected": 0.01190691627562046,
       "step": 120
     },
     {
       "epoch": 0.34,
-      "grad_norm": 95.0,
+      "grad_norm": 153.0,
       "learning_rate": 4.1940827077152755e-06,
-      "logits/chosen": -2.7426066398620605,
-      "logits/rejected": -2.738373279571533,
-      "logps/chosen": -28.734365463256836,
-      "logps/rejected": -30.13399887084961,
-      "loss": 0.9835,
-      "rewards/accuracies": 0.6499999761581421,
-      "rewards/chosen": 0.16822054982185364,
-      "rewards/margins": 0.16835634410381317,
-      "rewards/rejected": -0.00013580024824477732,
+      "logits/chosen": -2.745648145675659,
+      "logits/rejected": -2.7412946224212646,
+      "logps/chosen": -28.58340835571289,
+      "logps/rejected": -30.125900268554688,
+      "loss": 1.9846,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": 0.15650880336761475,
+      "rewards/margins": 0.1533457338809967,
+      "rewards/rejected": 0.0031630918383598328,
       "step": 130
     },
     {
       "epoch": 0.36,
-      "grad_norm": 192.0,
+      "grad_norm": 128.0,
       "learning_rate": 4.0204024186666215e-06,
-      "logits/chosen": -3.015483856201172,
-      "logits/rejected": -3.012770652770996,
-      "logps/chosen": -27.254077911376953,
-      "logps/rejected": -31.771997451782227,
-      "loss": 0.9001,
-      "rewards/accuracies": 0.6499999761581421,
-      "rewards/chosen": 0.1625039130449295,
-      "rewards/margins": 0.2431638240814209,
-      "rewards/rejected": -0.0806598886847496,
+      "logits/chosen": -3.016637086868286,
+      "logits/rejected": -3.014425277709961,
+      "logps/chosen": -27.189367294311523,
+      "logps/rejected": -31.768428802490234,
+      "loss": 1.6015,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.1187439113855362,
+      "rewards/margins": 0.16340722143650055,
+      "rewards/rejected": -0.04466330260038376,
       "step": 140
     },
     {
       "epoch": 0.39,
-      "grad_norm": 147.0,
+      "grad_norm": 191.0,
       "learning_rate": 3.834196265035119e-06,
-      "logits/chosen": -2.8096275329589844,
-      "logits/rejected": -2.804391860961914,
-      "logps/chosen": -27.50994300842285,
-      "logps/rejected": -31.280899047851562,
-      "loss": 0.7365,
-      "rewards/accuracies": 0.6625000238418579,
-      "rewards/chosen": 0.22089700400829315,
-      "rewards/margins": 0.26097819209098816,
-      "rewards/rejected": -0.04008117690682411,
+      "logits/chosen": -2.812203884124756,
+      "logits/rejected": -2.8067917823791504,
+      "logps/chosen": -27.404495239257812,
+      "logps/rejected": -31.38730239868164,
+      "loss": 1.2187,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": 0.16840532422065735,
+      "rewards/margins": 0.2338680922985077,
+      "rewards/rejected": -0.06546276807785034,
       "step": 150
     },
     {
       "epoch": 0.42,
-      "grad_norm": 122.5,
+      "grad_norm": 223.0,
       "learning_rate": 3.636998309800573e-06,
-      "logits/chosen": -3.127908229827881,
-      "logits/rejected": -3.1092982292175293,
-      "logps/chosen": -31.936166763305664,
-      "logps/rejected": -29.042150497436523,
-      "loss": 1.5699,
-      "rewards/accuracies": 0.6625000238418579,
-      "rewards/chosen": 0.38888686895370483,
-      "rewards/margins": 0.43254947662353516,
-      "rewards/rejected": -0.043662626296281815,
+      "logits/chosen": -3.127948522567749,
+      "logits/rejected": -3.1103968620300293,
+      "logps/chosen": -31.860748291015625,
+      "logps/rejected": -29.207168579101562,
+      "loss": 1.9596,
+      "rewards/accuracies": 0.762499988079071,
+      "rewards/chosen": 0.2523886263370514,
+      "rewards/margins": 0.3433467745780945,
+      "rewards/rejected": -0.09095814824104309,
       "step": 160
     },
     {
       "epoch": 0.44,
-      "grad_norm": 116.0,
+      "grad_norm": 112.0,
       "learning_rate": 3.4304331721118078e-06,
-      "logits/chosen": -2.9442355632781982,
-      "logits/rejected": -2.9510912895202637,
-      "logps/chosen": -29.370159149169922,
-      "logps/rejected": -31.104833602905273,
-      "loss": 1.1729,
-      "rewards/accuracies": 0.737500011920929,
-      "rewards/chosen": 0.367119163274765,
-      "rewards/margins": 0.37086552381515503,
-      "rewards/rejected": -0.003746363567188382,
+      "logits/chosen": -2.9414854049682617,
+      "logits/rejected": -2.9481499195098877,
+      "logps/chosen": -29.37969970703125,
+      "logps/rejected": -31.226505279541016,
+      "loss": 1.5997,
+      "rewards/accuracies": 0.762499988079071,
+      "rewards/chosen": 0.205966979265213,
+      "rewards/margins": 0.25677576661109924,
+      "rewards/rejected": -0.05080880969762802,
       "step": 170
     },
     {
       "epoch": 0.47,
-      "grad_norm": 206.0,
+      "grad_norm": 118.5,
       "learning_rate": 3.2162026428305436e-06,
-      "logits/chosen": -2.7904200553894043,
-      "logits/rejected": -2.7880074977874756,
-      "logps/chosen": -29.214679718017578,
-      "logps/rejected": -29.80947494506836,
-      "loss": 0.8255,
-      "rewards/accuracies": 0.7124999761581421,
-      "rewards/chosen": 0.3186172544956207,
-      "rewards/margins": 0.3475292921066284,
-      "rewards/rejected": -0.028912032023072243,
+      "logits/chosen": -2.789107322692871,
+      "logits/rejected": -2.7866973876953125,
+      "logps/chosen": -29.164358139038086,
+      "logps/rejected": -29.882699966430664,
+      "loss": 1.5371,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.2021961212158203,
+      "rewards/margins": 0.24800649285316467,
+      "rewards/rejected": -0.045810386538505554,
       "step": 180
     },
     {
       "epoch": 0.49,
-      "grad_norm": 264.0,
+      "grad_norm": 143.0,
       "learning_rate": 2.996071664294641e-06,
-      "logits/chosen": -2.9064321517944336,
-      "logits/rejected": -2.9026076793670654,
-      "logps/chosen": -29.777074813842773,
-      "logps/rejected": -28.358760833740234,
-      "loss": 1.8338,
-      "rewards/accuracies": 0.625,
-      "rewards/chosen": 0.2867472767829895,
-      "rewards/margins": 0.28760695457458496,
-      "rewards/rejected": -0.0008596793049946427,
+      "logits/chosen": -2.9042458534240723,
+      "logits/rejected": -2.900311231613159,
+      "logps/chosen": -29.820514678955078,
+      "logps/rejected": -28.412799835205078,
+      "loss": 2.882,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": 0.14647993445396423,
+      "rewards/margins": 0.1685885637998581,
+      "rewards/rejected": -0.022108638659119606,
       "step": 190
     },
     {
       "epoch": 0.52,
-      "grad_norm": 344.0,
+      "grad_norm": 200.0,
       "learning_rate": 2.7718537898066833e-06,
-      "logits/chosen": -2.9745757579803467,
-      "logits/rejected": -2.9624178409576416,
-      "logps/chosen": -33.12746810913086,
-      "logps/rejected": -30.224645614624023,
-      "loss": 3.0672,
-      "rewards/accuracies": 0.612500011920929,
-      "rewards/chosen": 0.4000841975212097,
-      "rewards/margins": 0.25347504019737244,
-      "rewards/rejected": 0.14660920202732086,
+      "logits/chosen": -2.9734153747558594,
+      "logits/rejected": -2.9612646102905273,
+      "logps/chosen": -32.977813720703125,
+      "logps/rejected": -30.30489730834961,
+      "loss": 2.6629,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.288481742143631,
+      "rewards/margins": 0.23680701851844788,
+      "rewards/rejected": 0.05167470499873161,
       "step": 200
     },
     {
       "epoch": 0.52,
-      "eval_logits/chosen": -2.8127942085266113,
-      "eval_logits/rejected": -2.8107194900512695,
-      "eval_logps/chosen": -31.217693328857422,
-      "eval_logps/rejected": -34.68858337402344,
-      "eval_loss": 0.6355926990509033,
-      "eval_rewards/accuracies": 0.5627076625823975,
-      "eval_rewards/chosen": 0.04533065855503082,
-      "eval_rewards/margins": 0.03862937539815903,
-      "eval_rewards/rejected": 0.006701283622533083,
-      "eval_runtime": 112.8257,
-      "eval_samples_per_second": 3.04,
+      "eval_logits/chosen": -2.812067985534668,
+      "eval_logits/rejected": -2.8095240592956543,
+      "eval_logps/chosen": -31.236892700195312,
+      "eval_logps/rejected": -34.73383712768555,
+      "eval_loss": 1.5648894309997559,
+      "eval_rewards/accuracies": 0.5456810593605042,
+      "eval_rewards/chosen": 0.018223287537693977,
+      "eval_rewards/margins": 0.032496776431798935,
+      "eval_rewards/rejected": -0.014273490756750107,
+      "eval_runtime": 112.7253,
+      "eval_samples_per_second": 3.043,
       "eval_steps_per_second": 0.381,
       "step": 200
     },
     {
       "epoch": 0.55,
-      "grad_norm": 272.0,
+      "grad_norm": 214.0,
       "learning_rate": 2.5453962426402006e-06,
-      "logits/chosen": -2.909015655517578,
-      "logits/rejected": -2.909053325653076,
-      "logps/chosen": -32.618431091308594,
-      "logps/rejected": -33.84246063232422,
-      "loss": 1.7638,
-      "rewards/accuracies": 0.762499988079071,
-      "rewards/chosen": 0.2904837131500244,
-      "rewards/margins": 0.2771795690059662,
-      "rewards/rejected": 0.013304118998348713,
+      "logits/chosen": -2.9090185165405273,
+      "logits/rejected": -2.910029888153076,
+      "logps/chosen": -32.432430267333984,
+      "logps/rejected": -33.925514221191406,
+      "loss": 1.49,
+      "rewards/accuracies": 0.7875000238418579,
+      "rewards/chosen": 0.2403920441865921,
+      "rewards/margins": 0.26600998640060425,
+      "rewards/rejected": -0.025617951527237892,
       "step": 210
     },
     {
       "epoch": 0.57,
-      "grad_norm": 195.0,
+      "grad_norm": 202.0,
       "learning_rate": 2.3185646976551794e-06,
-      "logits/chosen": -2.8854432106018066,
-      "logits/rejected": -2.901151657104492,
-      "logps/chosen": -29.67086410522461,
-      "logps/rejected": -28.4765567779541,
-      "loss": 1.6395,
-      "rewards/accuracies": 0.675000011920929,
-      "rewards/chosen": 0.3550260066986084,
-      "rewards/margins": 0.3188283145427704,
-      "rewards/rejected": 0.03619767725467682,
+      "logits/chosen": -2.8856301307678223,
+      "logits/rejected": -2.9012038707733154,
+      "logps/chosen": -29.458789825439453,
+      "logps/rejected": -28.523406982421875,
+      "loss": 2.1585,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": 0.2876998782157898,
+      "rewards/margins": 0.28575605154037476,
+      "rewards/rejected": 0.001943854964338243,
       "step": 220
     },
     {
       "epoch": 0.6,
-      "grad_norm": 102.0,
+      "grad_norm": 150.0,
       "learning_rate": 2.0932279108998323e-06,
-      "logits/chosen": -2.9346377849578857,
-      "logits/rejected": -2.9390294551849365,
-      "logps/chosen": -30.749608993530273,
-      "logps/rejected": -31.755859375,
-      "loss": 0.8936,
-      "rewards/accuracies": 0.6499999761581421,
-      "rewards/chosen": 0.28158265352249146,
-      "rewards/margins": 0.279979407787323,
-      "rewards/rejected": 0.0016032479470595717,
+      "logits/chosen": -2.9351718425750732,
+      "logits/rejected": -2.939762592315674,
+      "logps/chosen": -30.686193466186523,
+      "logps/rejected": -31.73343276977539,
+      "loss": 1.6516,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.18626943230628967,
+      "rewards/margins": 0.1763811856508255,
+      "rewards/rejected": 0.00988826435059309,
       "step": 230
     },
     {
       "epoch": 0.62,
-      "grad_norm": 77.5,
+      "grad_norm": 117.0,
       "learning_rate": 1.8712423238279358e-06,
-      "logits/chosen": -2.983455181121826,
-      "logits/rejected": -2.9928596019744873,
-      "logps/chosen": -30.32903480529785,
-      "logps/rejected": -30.288427352905273,
-      "loss": 0.9824,
-      "rewards/accuracies": 0.824999988079071,
-      "rewards/chosen": 0.4031239151954651,
-      "rewards/margins": 0.4403551518917084,
-      "rewards/rejected": -0.03723124787211418,
+      "logits/chosen": -2.98549485206604,
+      "logits/rejected": -2.9937961101531982,
+      "logps/chosen": -30.101831436157227,
+      "logps/rejected": -30.255285263061523,
+      "loss": 0.8256,
+      "rewards/accuracies": 0.862500011920929,
+      "rewards/chosen": 0.32123786211013794,
+      "rewards/margins": 0.329254686832428,
+      "rewards/rejected": -0.008016781881451607,
       "step": 240
     },
     {
       "epoch": 0.65,
-      "grad_norm": 282.0,
+      "grad_norm": 350.0,
       "learning_rate": 1.6544367689701824e-06,
-      "logits/chosen": -2.817836046218872,
-      "logits/rejected": -2.808262348175049,
-      "logps/chosen": -26.585535049438477,
-      "logps/rejected": -29.320880889892578,
-      "loss": 1.3431,
-      "rewards/accuracies": 0.675000011920929,
-      "rewards/chosen": 0.25807633996009827,
-      "rewards/margins": 0.11054243892431259,
-      "rewards/rejected": 0.14753387868404388,
+      "logits/chosen": -2.817945718765259,
+      "logits/rejected": -2.808492660522461,
+      "logps/chosen": -26.3896427154541,
+      "logps/rejected": -29.31142807006836,
+      "loss": 1.6788,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.22583059966564178,
+      "rewards/margins": 0.13774575293064117,
+      "rewards/rejected": 0.0880848616361618,
       "step": 250
     },
     {
       "epoch": 0.68,
-      "grad_norm": 89.0,
+      "grad_norm": 105.5,
       "learning_rate": 1.4445974030621963e-06,
-      "logits/chosen": -2.8003945350646973,
-      "logits/rejected": -2.8208835124969482,
-      "logps/chosen": -29.268306732177734,
-      "logps/rejected": -34.10762405395508,
-      "loss": 1.0424,
-      "rewards/accuracies": 0.675000011920929,
-      "rewards/chosen": 0.39315709471702576,
-      "rewards/margins": 0.34663277864456177,
-      "rewards/rejected": 0.04652435705065727,
+      "logits/chosen": -2.7999472618103027,
+      "logits/rejected": -2.8214704990386963,
+      "logps/chosen": -29.094181060791016,
+      "logps/rejected": -34.21991729736328,
+      "loss": 1.111,
+      "rewards/accuracies": 0.8374999761581421,
+      "rewards/chosen": 0.294312059879303,
+      "rewards/margins": 0.3126428425312042,
+      "rewards/rejected": -0.01833074726164341,
       "step": 260
     },
     {
       "epoch": 0.7,
-      "grad_norm": 54.25,
+      "grad_norm": 124.0,
       "learning_rate": 1.243452991757889e-06,
-      "logits/chosen": -2.9433441162109375,
-      "logits/rejected": -2.9482321739196777,
-      "logps/chosen": -30.294042587280273,
-      "logps/rejected": -29.78550148010254,
-      "loss": 1.0077,
-      "rewards/accuracies": 0.737500011920929,
-      "rewards/chosen": 0.341118723154068,
-      "rewards/margins": 0.2852852940559387,
-      "rewards/rejected": 0.05583342909812927,
+      "logits/chosen": -2.942542791366577,
+      "logits/rejected": -2.9476306438446045,
+      "logps/chosen": -30.103805541992188,
+      "logps/rejected": -29.8814640045166,
+      "loss": 1.3753,
+      "rewards/accuracies": 0.7875000238418579,
+      "rewards/chosen": 0.27102118730545044,
+      "rewards/margins": 0.27750203013420105,
+      "rewards/rejected": -0.0064808703027665615,
       "step": 270
     },
     {
       "epoch": 0.73,
-      "grad_norm": 78.0,
+      "grad_norm": 96.0,
       "learning_rate": 1.0526606671603523e-06,
-      "logits/chosen": -2.9577131271362305,
-      "logits/rejected": -2.944226026535034,
-      "logps/chosen": -30.35019874572754,
-      "logps/rejected": -28.367935180664062,
-      "loss": 1.9189,
-      "rewards/accuracies": 0.6000000238418579,
-      "rewards/chosen": 0.2600388824939728,
-      "rewards/margins": 0.0756484642624855,
-      "rewards/rejected": 0.18439041078090668,
+      "logits/chosen": -2.9563868045806885,
+      "logits/rejected": -2.9434051513671875,
+      "logps/chosen": -30.231460571289062,
+      "logps/rejected": -28.510696411132812,
+      "loss": 1.8713,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": 0.196088969707489,
+      "rewards/margins": 0.1478276252746582,
+      "rewards/rejected": 0.0482613667845726,
       "step": 280
     },
     {
       "epoch": 0.75,
-      "grad_norm": 227.0,
+      "grad_norm": 176.0,
       "learning_rate": 8.737922755071455e-07,
-      "logits/chosen": -2.885504722595215,
-      "logits/rejected": -2.8676083087921143,
-      "logps/chosen": -32.159873962402344,
-      "logps/rejected": -30.42794418334961,
-      "loss": 1.5952,
-      "rewards/accuracies": 0.675000011920929,
-      "rewards/chosen": 0.33349719643592834,
-      "rewards/margins": 0.26674339175224304,
-      "rewards/rejected": 0.06675383448600769,
+      "logits/chosen": -2.884552478790283,
+      "logits/rejected": -2.8662493228912354,
+      "logps/chosen": -31.870113372802734,
+      "logps/rejected": -30.56744384765625,
+      "loss": 1.5277,
+      "rewards/accuracies": 0.7749999761581421,
+      "rewards/chosen": 0.3064747452735901,
+      "rewards/margins": 0.32412776350975037,
+      "rewards/rejected": -0.017653033137321472,
       "step": 290
     },
     {
       "epoch": 0.78,
-      "grad_norm": 340.0,
+      "grad_norm": 372.0,
       "learning_rate": 7.08321427484816e-07,
-      "logits/chosen": -2.889909267425537,
-      "logits/rejected": -2.8869872093200684,
-      "logps/chosen": -31.75823402404785,
-      "logps/rejected": -27.508203506469727,
-      "loss": 1.2353,
-      "rewards/accuracies": 0.625,
-      "rewards/chosen": 0.32509008049964905,
-      "rewards/margins": 0.2583235204219818,
-      "rewards/rejected": 0.06676653772592545,
+      "logits/chosen": -2.889061689376831,
+      "logits/rejected": -2.885744571685791,
+      "logps/chosen": -31.58562660217285,
+      "logps/rejected": -27.505157470703125,
+      "loss": 1.5067,
+      "rewards/accuracies": 0.7250000238418579,
+      "rewards/chosen": 0.254812091588974,
+      "rewards/margins": 0.21544210612773895,
+      "rewards/rejected": 0.039369963109493256,
       "step": 300
     },
     {
       "epoch": 0.78,
-      "eval_logits/chosen": -2.809739589691162,
-      "eval_logits/rejected": -2.807137966156006,
-      "eval_logps/chosen": -31.20755958557129,
-      "eval_logps/rejected": -34.65956115722656,
-      "eval_loss": 0.6301615834236145,
-      "eval_rewards/accuracies": 0.531146228313446,
-      "eval_rewards/chosen": 0.052424702793359756,
-      "eval_rewards/margins": 0.025408506393432617,
-      "eval_rewards/rejected": 0.02701619826257229,
-      "eval_runtime": 113.0311,
-      "eval_samples_per_second": 3.035,
+      "eval_logits/chosen": -2.80647873878479,
+      "eval_logits/rejected": -2.8040761947631836,
+      "eval_logps/chosen": -31.251041412353516,
+      "eval_logps/rejected": -34.71211624145508,
+      "eval_loss": 1.6903845071792603,
+      "eval_rewards/accuracies": 0.5573089718818665,
+      "eval_rewards/chosen": 0.012563018128275871,
+      "eval_rewards/margins": 0.018146870657801628,
+      "eval_rewards/rejected": -0.005583854392170906,
+      "eval_runtime": 113.035,
+      "eval_samples_per_second": 3.034,
       "eval_steps_per_second": 0.38,
       "step": 300
     },
     {
       "epoch": 0.81,
-      "grad_norm": 251.0,
+      "grad_norm": 129.0,
       "learning_rate": 5.576113578589035e-07,
-      "logits/chosen": -2.7691893577575684,
-      "logits/rejected": -2.787349224090576,
-      "logps/chosen": -28.8099422454834,
-      "logps/rejected": -30.58675765991211,
-      "loss": 1.3555,
-      "rewards/accuracies": 0.637499988079071,
-      "rewards/chosen": 0.2936348617076874,
-      "rewards/margins": 0.19195988774299622,
-      "rewards/rejected": 0.10167495906352997,
+      "logits/chosen": -2.7671875953674316,
+      "logits/rejected": -2.785642147064209,
+      "logps/chosen": -28.604686737060547,
+      "logps/rejected": -30.709314346313477,
+      "loss": 1.2664,
+      "rewards/accuracies": 0.7875000238418579,
+      "rewards/chosen": 0.24989323318004608,
+      "rewards/margins": 0.24081555008888245,
+      "rewards/rejected": 0.009077723138034344,
       "step": 310
     },
     {
       "epoch": 0.83,
-      "grad_norm": 93.5,
+      "grad_norm": 92.5,
       "learning_rate": 4.229036944380913e-07,
-      "logits/chosen": -3.0164008140563965,
-      "logits/rejected": -3.0019757747650146,
-      "logps/chosen": -29.883535385131836,
-      "logps/rejected": -28.36444664001465,
-      "loss": 0.9933,
-      "rewards/accuracies": 0.7124999761581421,
-      "rewards/chosen": 0.22963504493236542,
-      "rewards/margins": 0.26160377264022827,
-      "rewards/rejected": -0.03196870535612106,
+      "logits/chosen": -3.014265537261963,
+      "logits/rejected": -2.9993736743927,
+      "logps/chosen": -29.781946182250977,
+      "logps/rejected": -28.38212013244629,
+      "loss": 1.4826,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": 0.17185598611831665,
+      "rewards/margins": 0.19719351828098297,
+      "rewards/rejected": -0.025337522849440575,
       "step": 320
     },
     {
       "epoch": 0.86,
-      "grad_norm": 119.5,
+      "grad_norm": 163.0,
       "learning_rate": 3.053082288996112e-07,
-      "logits/chosen": -2.931257724761963,
-      "logits/rejected": -2.9136319160461426,
-      "logps/chosen": -27.954730987548828,
-      "logps/rejected": -30.42913246154785,
-      "loss": 0.922,
-      "rewards/accuracies": 0.737500011920929,
-      "rewards/chosen": 0.285138338804245,
-      "rewards/margins": 0.3113783299922943,
-      "rewards/rejected": -0.026239976286888123,
+      "logits/chosen": -2.9294309616088867,
+      "logits/rejected": -2.9114794731140137,
+      "logps/chosen": -27.724899291992188,
+      "logps/rejected": -30.527271270751953,
+      "loss": 1.2966,
+      "rewards/accuracies": 0.8125,
+      "rewards/chosen": 0.2548690140247345,
+      "rewards/margins": 0.30911877751350403,
+      "rewards/rejected": -0.054249756038188934,
       "step": 330
     },
     {
       "epoch": 0.88,
-      "grad_norm": 131.0,
+      "grad_norm": 191.0,
       "learning_rate": 2.0579377374915805e-07,
-      "logits/chosen": -3.146080493927002,
-      "logits/rejected": -3.152937650680542,
-      "logps/chosen": -31.1461181640625,
-      "logps/rejected": -32.72796630859375,
-      "loss": 0.797,
+      "logits/chosen": -3.144012928009033,
+      "logits/rejected": -3.1506927013397217,
+      "logps/chosen": -31.059814453125,
+      "logps/rejected": -32.82073211669922,
+      "loss": 1.5694,
       "rewards/accuracies": 0.762499988079071,
-      "rewards/chosen": 0.3451611399650574,
-      "rewards/margins": 0.3675331473350525,
-      "rewards/rejected": -0.022372011095285416,
+      "rewards/chosen": 0.23175635933876038,
+      "rewards/margins": 0.2816471457481384,
+      "rewards/rejected": -0.04989078640937805,
       "step": 340
     },
     {
       "epoch": 0.91,
-      "grad_norm": 117.0,
+      "grad_norm": 112.5,
       "learning_rate": 1.2518018074041684e-07,
-      "logits/chosen": -3.0231246948242188,
-      "logits/rejected": -3.0266261100769043,
-      "logps/chosen": -30.33087158203125,
-      "logps/rejected": -31.515050888061523,
-      "loss": 0.8954,
-      "rewards/accuracies": 0.75,
-      "rewards/chosen": 0.33320528268814087,
-      "rewards/margins": 0.334422767162323,
-      "rewards/rejected": -0.001217484474182129,
+      "logits/chosen": -3.021515130996704,
+      "logits/rejected": -3.02475643157959,
+      "logps/chosen": -30.08322525024414,
+      "logps/rejected": -31.63114356994629,
+      "loss": 1.3735,
+      "rewards/accuracies": 0.7875000238418579,
+      "rewards/chosen": 0.2894637882709503,
+      "rewards/margins": 0.3365970253944397,
+      "rewards/rejected": -0.04713326320052147,
       "step": 350
     },
     {
       "epoch": 0.94,
-      "grad_norm": 223.0,
+      "grad_norm": 215.0,
       "learning_rate": 6.41315865106129e-08,
-      "logits/chosen": -2.86474871635437,
-      "logits/rejected": -2.866323947906494,
-      "logps/chosen": -27.935626983642578,
-      "logps/rejected": -29.81571388244629,
-      "loss": 0.7841,
-      "rewards/accuracies": 0.800000011920929,
-      "rewards/chosen": 0.49987268447875977,
-      "rewards/margins": 0.4892123341560364,
-      "rewards/rejected": 0.010660367086529732,
+      "logits/chosen": -2.8619635105133057,
+      "logits/rejected": -2.8640754222869873,
+      "logps/chosen": -27.785686492919922,
+      "logps/rejected": -29.807357788085938,
+      "loss": 1.4796,
+      "rewards/accuracies": 0.8125,
+      "rewards/chosen": 0.34561628103256226,
+      "rewards/margins": 0.33618348836898804,
+      "rewards/rejected": 0.00943276472389698,
       "step": 360
     },
     {
       "epoch": 0.96,
       "grad_norm": 127.5,
       "learning_rate": 2.3150941078050325e-08,
-      "logits/chosen": -2.9371769428253174,
-      "logits/rejected": -2.936225414276123,
-      "logps/chosen": -30.32781410217285,
-      "logps/rejected": -31.941625595092773,
-      "loss": 0.7687,
-      "rewards/accuracies": 0.7250000238418579,
-      "rewards/chosen": 0.3211846947669983,
-      "rewards/margins": 0.3306921422481537,
-      "rewards/rejected": -0.00950746051967144,
+      "logits/chosen": -2.9357876777648926,
+      "logits/rejected": -2.9344675540924072,
+      "logps/chosen": -30.16611671447754,
+      "logps/rejected": -32.088932037353516,
+      "loss": 1.1067,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": 0.24821257591247559,
+      "rewards/margins": 0.31256765127182007,
+      "rewards/rejected": -0.06435510516166687,
       "step": 370
     },
     {
       "epoch": 0.99,
-      "grad_norm": 484.0,
+      "grad_norm": 552.0,
       "learning_rate": 2.575864278703266e-09,
-      "logits/chosen": -2.8952066898345947,
-      "logits/rejected": -2.8765997886657715,
-      "logps/chosen": -28.4578857421875,
-      "logps/rejected": -27.789112091064453,
-      "loss": 0.8783,
-      "rewards/accuracies": 0.737500011920929,
-      "rewards/chosen": 0.3410065770149231,
-      "rewards/margins": 0.2706519067287445,
-      "rewards/rejected": 0.07035474479198456,
+      "logits/chosen": -2.89326548576355,
+      "logits/rejected": -2.8758132457733154,
+      "logps/chosen": -28.290267944335938,
+      "logps/rejected": -27.925792694091797,
+      "loss": 1.6812,
+      "rewards/accuracies": 0.762499988079071,
+      "rewards/chosen": 0.2619072496891022,
+      "rewards/margins": 0.27637776732444763,
+      "rewards/rejected": -0.01447053998708725,
       "step": 380
     },
     {
       "epoch": 1.0,
       "step": 385,
       "total_flos": 0.0,
-      "train_loss": 1.2163990609057538,
-      "train_runtime": 2720.9807,
+      "train_loss": 1.7377146052075672,
+      "train_runtime": 2719.6001,
       "train_samples_per_second": 1.132,
-      "train_steps_per_second": 0.141
+      "train_steps_per_second": 0.142
     }
   ],
   "logging_steps": 10,