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: 0.4908
-- Rewards/chosen: 0.0412
-- Rewards/rejected: -0.0039
-- Rewards/accuracies: 0.5868
-- Rewards/margins: 0.0450
-- Logps/rejected: -34.7175
-- Logps/chosen: -31.0766
-- Logits/rejected: -2.8191
-- Logits/chosen: -2.8217
+- Loss: 0.6302
+- Rewards/chosen: 0.0524
+- Rewards/rejected: 0.0270
+- Rewards/accuracies: 0.5311
+- Rewards/margins: 0.0254
+- Logps/rejected: -34.6596
+- Logps/chosen: -31.2076
+- Logits/rejected: -2.8071
+- Logits/chosen: -2.8097
 
 ## 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 |
 |:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
-| 0.4829        | 0.26  | 100  | 0.4945          | 0.0390         | 0.0134           | 0.5635             | 0.0256          | -34.6310       | -31.0874     | -2.8002         | -2.8030       |
-| 0.4376        | 0.52  | 200  | 0.4911          | 0.0473         | 0.0060           | 0.5577             | 0.0413          | -34.6682       | -31.0459     | -2.8137         | -2.8163       |
-| 0.4326        | 0.78  | 300  | 0.4909          | 0.0420         | -0.0034          | 0.6013             | 0.0454          | -34.7152       | -31.0725     | -2.8183         | -2.8209       |
+| 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       |
 
 
 ### Framework versions

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5281a74206c349eb3e1681bafca7b3c95e98bff78c27375b38dc42b4bcf5ae64
+oid sha256:f59a5a3009fd0d692a9e79a2e5db218bebcace35d25a55969a96b406565794b8
 size 671150064

all_results.json

@@ -1,21 +1,8 @@
 {
     "epoch": 1.0,
-    "eval_logits/chosen": -2.821664571762085,
-    "eval_logits/rejected": -2.819110155105591,
-    "eval_logps/chosen": -31.076623916625977,
-    "eval_logps/rejected": -34.71748352050781,
-    "eval_loss": 0.49079883098602295,
-    "eval_rewards/accuracies": 0.5867940187454224,
-    "eval_rewards/chosen": 0.04116562753915787,
-    "eval_rewards/margins": 0.045030828565359116,
-    "eval_rewards/rejected": -0.003865197068080306,
-    "eval_runtime": 112.8844,
-    "eval_samples": 343,
-    "eval_samples_per_second": 3.039,
-    "eval_steps_per_second": 0.381,
-    "train_loss": 0.4552104888024268,
-    "train_runtime": 2721.944,
+    "train_loss": 1.2163990609057538,
+    "train_runtime": 2720.9807,
     "train_samples": 3079,
-    "train_samples_per_second": 1.131,
+    "train_samples_per_second": 1.132,
     "train_steps_per_second": 0.141
 }

train_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 1.0,
-    "train_loss": 0.4552104888024268,
-    "train_runtime": 2721.944,
+    "train_loss": 1.2163990609057538,
+    "train_runtime": 2720.9807,
     "train_samples": 3079,
-    "train_samples_per_second": 1.131,
+    "train_samples_per_second": 1.132,
     "train_steps_per_second": 0.141
 }

trainer_state.json

@@ -10,13 +10,13 @@
   "log_history": [
     {
       "epoch": 0.0,
-      "grad_norm": 3.953125,
+      "grad_norm": 113.0,
       "learning_rate": 1.282051282051282e-07,
       "logits/chosen": -2.7358343601226807,
       "logits/rejected": -2.7480404376983643,
       "logps/chosen": -27.35565757751465,
       "logps/rejected": -21.06114387512207,
-      "loss": 0.5,
+      "loss": 0.5102,
       "rewards/accuracies": 0.0,
       "rewards/chosen": 0.0,
       "rewards/margins": 0.0,
@@ -25,629 +25,629 @@
     },
     {
       "epoch": 0.03,
-      "grad_norm": 3.75,
+      "grad_norm": 146.0,
       "learning_rate": 1.282051282051282e-06,
-      "logits/chosen": -3.0089564323425293,
-      "logits/rejected": -2.998065710067749,
-      "logps/chosen": -33.19136428833008,
-      "logps/rejected": -31.97561264038086,
-      "loss": 0.4998,
-      "rewards/accuracies": 0.4305555522441864,
-      "rewards/chosen": -0.0020843851380050182,
-      "rewards/margins": 0.0002575106918811798,
-      "rewards/rejected": -0.002341895829886198,
+      "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,
       "step": 10
     },
     {
       "epoch": 0.05,
-      "grad_norm": 3.359375,
+      "grad_norm": 144.0,
       "learning_rate": 2.564102564102564e-06,
-      "logits/chosen": -2.899214267730713,
-      "logits/rejected": -2.8942105770111084,
-      "logps/chosen": -32.48638916015625,
-      "logps/rejected": -28.965356826782227,
-      "loss": 0.5009,
-      "rewards/accuracies": 0.4749999940395355,
-      "rewards/chosen": -0.0022583019454032183,
-      "rewards/margins": -0.0055608078837394714,
-      "rewards/rejected": 0.0033025064039975405,
+      "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,
       "step": 20
     },
     {
       "epoch": 0.08,
-      "grad_norm": 4.0625,
+      "grad_norm": 159.0,
       "learning_rate": 3.846153846153847e-06,
-      "logits/chosen": -3.096067190170288,
-      "logits/rejected": -3.1081583499908447,
-      "logps/chosen": -32.83055877685547,
-      "logps/rejected": -30.159122467041016,
-      "loss": 0.4979,
-      "rewards/accuracies": 0.574999988079071,
-      "rewards/chosen": 0.0156254880130291,
-      "rewards/margins": 0.008240064606070518,
-      "rewards/rejected": 0.007385422941297293,
+      "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,
       "step": 30
     },
     {
       "epoch": 0.1,
-      "grad_norm": 4.0625,
+      "grad_norm": 223.0,
       "learning_rate": 4.999896948438434e-06,
-      "logits/chosen": -2.8633229732513428,
-      "logits/rejected": -2.8539462089538574,
-      "logps/chosen": -31.576160430908203,
-      "logps/rejected": -32.43328857421875,
-      "loss": 0.4858,
-      "rewards/accuracies": 0.675000011920929,
-      "rewards/chosen": 0.05083252862095833,
-      "rewards/margins": 0.06599889695644379,
-      "rewards/rejected": -0.01516636274755001,
+      "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,
       "step": 40
     },
     {
       "epoch": 0.13,
-      "grad_norm": 3.8125,
+      "grad_norm": 156.0,
       "learning_rate": 4.987541037542187e-06,
-      "logits/chosen": -2.880373954772949,
-      "logits/rejected": -2.87811541557312,
-      "logps/chosen": -29.325586318969727,
-      "logps/rejected": -30.077322006225586,
-      "loss": 0.4808,
-      "rewards/accuracies": 0.5,
-      "rewards/chosen": 0.08291508257389069,
-      "rewards/margins": 0.08059060573577881,
-      "rewards/rejected": 0.002324472414329648,
+      "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,
       "step": 50
     },
     {
       "epoch": 0.16,
-      "grad_norm": 3.59375,
+      "grad_norm": 150.0,
       "learning_rate": 4.954691471941119e-06,
-      "logits/chosen": -2.9070420265197754,
-      "logits/rejected": -2.908172369003296,
-      "logps/chosen": -29.754262924194336,
-      "logps/rejected": -27.948543548583984,
-      "loss": 0.4839,
-      "rewards/accuracies": 0.6499999761581421,
-      "rewards/chosen": 0.07510136812925339,
-      "rewards/margins": 0.0736737847328186,
-      "rewards/rejected": 0.0014275781577453017,
+      "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,
       "step": 60
     },
     {
       "epoch": 0.18,
-      "grad_norm": 4.3125,
+      "grad_norm": 972.0,
       "learning_rate": 4.901618883413549e-06,
-      "logits/chosen": -2.9913430213928223,
-      "logits/rejected": -2.9976165294647217,
-      "logps/chosen": -29.0341796875,
-      "logps/rejected": -30.7557430267334,
-      "loss": 0.4938,
-      "rewards/accuracies": 0.5375000238418579,
-      "rewards/chosen": 0.04935770481824875,
-      "rewards/margins": 0.022146493196487427,
-      "rewards/rejected": 0.027211207896471024,
+      "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,
       "step": 70
     },
     {
       "epoch": 0.21,
-      "grad_norm": 4.125,
+      "grad_norm": 192.0,
       "learning_rate": 4.828760511501322e-06,
-      "logits/chosen": -2.803948402404785,
-      "logits/rejected": -2.8194332122802734,
-      "logps/chosen": -29.147689819335938,
-      "logps/rejected": -29.726749420166016,
-      "loss": 0.4828,
-      "rewards/accuracies": 0.6625000238418579,
-      "rewards/chosen": 0.07860848307609558,
-      "rewards/margins": 0.08910086750984192,
-      "rewards/rejected": -0.010492382571101189,
+      "logits/chosen": -2.816981792449951,
+      "logits/rejected": -2.8327014446258545,
+      "logps/chosen": -29.412933349609375,
+      "logps/rejected": -29.788555145263672,
+      "loss": 0.6831,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": 0.08945836126804352,
+      "rewards/margins": 0.16944995522499084,
+      "rewards/rejected": -0.07999160140752792,
       "step": 80
     },
     {
       "epoch": 0.23,
-      "grad_norm": 4.4375,
+      "grad_norm": 362.0,
       "learning_rate": 4.7367166013034295e-06,
-      "logits/chosen": -2.892725706100464,
-      "logits/rejected": -2.874434232711792,
-      "logps/chosen": -32.414127349853516,
-      "logps/rejected": -30.058170318603516,
-      "loss": 0.4789,
-      "rewards/accuracies": 0.637499988079071,
-      "rewards/chosen": 0.08446021378040314,
-      "rewards/margins": 0.09176470339298248,
-      "rewards/rejected": -0.007304488681256771,
+      "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,
       "step": 90
     },
     {
       "epoch": 0.26,
-      "grad_norm": 3.5,
+      "grad_norm": 157.0,
       "learning_rate": 4.626245458345211e-06,
-      "logits/chosen": -2.9959635734558105,
-      "logits/rejected": -2.9973392486572266,
-      "logps/chosen": -31.55927085876465,
-      "logps/rejected": -30.724218368530273,
-      "loss": 0.4829,
-      "rewards/accuracies": 0.6625000238418579,
-      "rewards/chosen": 0.0995207279920578,
-      "rewards/margins": 0.0811297670006752,
-      "rewards/rejected": 0.0183909572660923,
+      "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,
       "step": 100
     },
     {
       "epoch": 0.26,
-      "eval_logits/chosen": -2.802985906600952,
-      "eval_logits/rejected": -2.8001766204833984,
-      "eval_logps/chosen": -31.0873966217041,
-      "eval_logps/rejected": -34.63096618652344,
-      "eval_loss": 0.49451369047164917,
-      "eval_rewards/accuracies": 0.5635381937026978,
-      "eval_rewards/chosen": 0.03901098668575287,
-      "eval_rewards/margins": 0.02557329833507538,
-      "eval_rewards/rejected": 0.013437685556709766,
-      "eval_runtime": 113.153,
-      "eval_samples_per_second": 3.031,
+      "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,
       "step": 100
     },
     {
       "epoch": 0.29,
-      "grad_norm": 4.09375,
+      "grad_norm": 157.0,
       "learning_rate": 4.498257201263691e-06,
-      "logits/chosen": -2.949113130569458,
-      "logits/rejected": -2.925306797027588,
-      "logps/chosen": -31.45932388305664,
-      "logps/rejected": -31.128564834594727,
-      "loss": 0.4692,
+      "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": 0.1514938771724701,
-      "rewards/margins": 0.14543716609477997,
-      "rewards/rejected": 0.006056725047528744,
+      "rewards/chosen": 0.3269161581993103,
+      "rewards/margins": 0.30541908740997314,
+      "rewards/rejected": 0.02149704284965992,
       "step": 110
     },
     {
       "epoch": 0.31,
-      "grad_norm": 4.125,
+      "grad_norm": 109.0,
       "learning_rate": 4.353806263777678e-06,
-      "logits/chosen": -3.031067371368408,
-      "logits/rejected": -3.060089588165283,
-      "logps/chosen": -28.312602996826172,
-      "logps/rejected": -34.01371765136719,
-      "loss": 0.4667,
-      "rewards/accuracies": 0.637499988079071,
-      "rewards/chosen": 0.17025941610336304,
-      "rewards/margins": 0.14349231123924255,
-      "rewards/rejected": 0.026767095550894737,
+      "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,
       "step": 120
     },
     {
       "epoch": 0.34,
-      "grad_norm": 2.515625,
+      "grad_norm": 95.0,
       "learning_rate": 4.1940827077152755e-06,
-      "logits/chosen": -2.733306884765625,
-      "logits/rejected": -2.729750871658325,
-      "logps/chosen": -28.17226219177246,
-      "logps/rejected": -30.136898040771484,
-      "loss": 0.4656,
-      "rewards/accuracies": 0.675000011920929,
-      "rewards/chosen": 0.1604831963777542,
-      "rewards/margins": 0.16110166907310486,
-      "rewards/rejected": -0.00061851367354393,
+      "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,
       "step": 130
     },
     {
       "epoch": 0.36,
-      "grad_norm": 3.03125,
+      "grad_norm": 192.0,
       "learning_rate": 4.0204024186666215e-06,
-      "logits/chosen": -3.0078999996185303,
-      "logits/rejected": -3.005507230758667,
-      "logps/chosen": -26.812694549560547,
-      "logps/rejected": -31.561386108398438,
-      "loss": 0.4732,
-      "rewards/accuracies": 0.5249999761581421,
-      "rewards/chosen": 0.13470572233200073,
-      "rewards/margins": 0.11562897264957428,
-      "rewards/rejected": 0.01907675340771675,
+      "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,
       "step": 140
     },
     {
       "epoch": 0.39,
-      "grad_norm": 3.671875,
+      "grad_norm": 147.0,
       "learning_rate": 3.834196265035119e-06,
-      "logits/chosen": -2.8041739463806152,
-      "logits/rejected": -2.7990527153015137,
-      "logps/chosen": -26.89498519897461,
-      "logps/rejected": -31.248632431030273,
-      "loss": 0.4572,
-      "rewards/accuracies": 0.699999988079071,
-      "rewards/chosen": 0.18610434234142303,
-      "rewards/margins": 0.1911022961139679,
-      "rewards/rejected": -0.004997962154448032,
+      "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,
       "step": 150
     },
     {
       "epoch": 0.42,
-      "grad_norm": 3.828125,
+      "grad_norm": 122.5,
       "learning_rate": 3.636998309800573e-06,
-      "logits/chosen": -3.122868299484253,
-      "logits/rejected": -3.1052639484405518,
-      "logps/chosen": -31.118968963623047,
-      "logps/rejected": -29.14169692993164,
-      "loss": 0.4349,
-      "rewards/accuracies": 0.7749999761581421,
-      "rewards/chosen": 0.2745501399040222,
-      "rewards/margins": 0.30693456530570984,
-      "rewards/rejected": -0.0323844812810421,
+      "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,
       "step": 160
     },
     {
       "epoch": 0.44,
-      "grad_norm": 4.0,
+      "grad_norm": 116.0,
       "learning_rate": 3.4304331721118078e-06,
-      "logits/chosen": -2.9357101917266846,
-      "logits/rejected": -2.9437410831451416,
-      "logps/chosen": -28.84354591369629,
-      "logps/rejected": -31.365503311157227,
-      "loss": 0.4446,
-      "rewards/accuracies": 0.6875,
-      "rewards/chosen": 0.21021375060081482,
-      "rewards/margins": 0.26341748237609863,
-      "rewards/rejected": -0.05320371314883232,
+      "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,
       "step": 170
     },
     {
       "epoch": 0.47,
-      "grad_norm": 2.9375,
+      "grad_norm": 206.0,
       "learning_rate": 3.2162026428305436e-06,
-      "logits/chosen": -2.7853760719299316,
-      "logits/rejected": -2.782726287841797,
-      "logps/chosen": -28.669147491455078,
-      "logps/rejected": -29.730758666992188,
-      "loss": 0.4565,
-      "rewards/accuracies": 0.6625000238418579,
-      "rewards/chosen": 0.2001400738954544,
-      "rewards/margins": 0.19265684485435486,
-      "rewards/rejected": 0.00748323742300272,
+      "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,
       "step": 180
     },
     {
       "epoch": 0.49,
-      "grad_norm": 2.390625,
+      "grad_norm": 264.0,
       "learning_rate": 2.996071664294641e-06,
-      "logits/chosen": -2.905327320098877,
-      "logits/rejected": -2.9015116691589355,
-      "logps/chosen": -29.04599380493164,
-      "logps/rejected": -28.085041046142578,
-      "loss": 0.4628,
+      "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": 0.2281440794467926,
-      "rewards/margins": 0.17364613711833954,
-      "rewards/rejected": 0.05449794605374336,
+      "rewards/chosen": 0.2867472767829895,
+      "rewards/margins": 0.28760695457458496,
+      "rewards/rejected": -0.0008596793049946427,
       "step": 190
     },
     {
       "epoch": 0.52,
-      "grad_norm": 2.84375,
+      "grad_norm": 344.0,
       "learning_rate": 2.7718537898066833e-06,
-      "logits/chosen": -2.9772303104400635,
-      "logits/rejected": -2.9655442237854004,
-      "logps/chosen": -31.91948890686035,
-      "logps/rejected": -30.082332611083984,
-      "loss": 0.4376,
-      "rewards/accuracies": 0.637499988079071,
-      "rewards/chosen": 0.3559049963951111,
-      "rewards/margins": 0.28555411100387573,
-      "rewards/rejected": 0.07035084813833237,
+      "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,
       "step": 200
     },
     {
       "epoch": 0.52,
-      "eval_logits/chosen": -2.8163414001464844,
-      "eval_logits/rejected": -2.8137245178222656,
-      "eval_logps/chosen": -31.0458927154541,
-      "eval_logps/rejected": -34.6682014465332,
-      "eval_loss": 0.49114447832107544,
-      "eval_rewards/accuracies": 0.5577242374420166,
-      "eval_rewards/chosen": 0.04731180891394615,
-      "eval_rewards/margins": 0.041320886462926865,
-      "eval_rewards/rejected": 0.005990919191390276,
-      "eval_runtime": 113.006,
-      "eval_samples_per_second": 3.035,
+      "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_steps_per_second": 0.381,
       "step": 200
     },
     {
       "epoch": 0.55,
-      "grad_norm": 4.28125,
+      "grad_norm": 272.0,
       "learning_rate": 2.5453962426402006e-06,
-      "logits/chosen": -2.914222240447998,
-      "logits/rejected": -2.916249990463257,
-      "logps/chosen": -31.570764541625977,
-      "logps/rejected": -33.81729507446289,
-      "loss": 0.4391,
-      "rewards/accuracies": 0.7124999761581421,
-      "rewards/chosen": 0.2925291359424591,
-      "rewards/margins": 0.2836949825286865,
-      "rewards/rejected": 0.008834179490804672,
+      "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,
       "step": 210
     },
     {
       "epoch": 0.57,
-      "grad_norm": 3.1875,
+      "grad_norm": 195.0,
       "learning_rate": 2.3185646976551794e-06,
-      "logits/chosen": -2.8961727619171143,
-      "logits/rejected": -2.9133689403533936,
-      "logps/chosen": -28.757190704345703,
-      "logps/rejected": -28.49643325805664,
-      "loss": 0.4382,
-      "rewards/accuracies": 0.7124999761581421,
-      "rewards/chosen": 0.28417009115219116,
-      "rewards/margins": 0.2778027653694153,
-      "rewards/rejected": 0.006367350462824106,
+      "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,
       "step": 220
     },
     {
       "epoch": 0.6,
-      "grad_norm": 3.75,
+      "grad_norm": 102.0,
       "learning_rate": 2.0932279108998323e-06,
-      "logits/chosen": -2.948033571243286,
-      "logits/rejected": -2.952341079711914,
-      "logps/chosen": -30.005146026611328,
-      "logps/rejected": -31.54683494567871,
-      "loss": 0.4579,
-      "rewards/accuracies": 0.699999988079071,
-      "rewards/chosen": 0.22934475541114807,
-      "rewards/margins": 0.18708109855651855,
-      "rewards/rejected": 0.042263638228178024,
+      "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,
       "step": 230
     },
     {
       "epoch": 0.62,
-      "grad_norm": 3.84375,
+      "grad_norm": 77.5,
       "learning_rate": 1.8712423238279358e-06,
-      "logits/chosen": -3.0030248165130615,
-      "logits/rejected": -3.009697675704956,
-      "logps/chosen": -29.56500816345215,
-      "logps/rejected": -30.203866958618164,
-      "loss": 0.4421,
-      "rewards/accuracies": 0.7749999761581421,
-      "rewards/chosen": 0.2679831385612488,
-      "rewards/margins": 0.2617082893848419,
-      "rewards/rejected": 0.00627488736063242,
+      "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,
       "step": 240
     },
     {
       "epoch": 0.65,
-      "grad_norm": 3.796875,
+      "grad_norm": 282.0,
       "learning_rate": 1.6544367689701824e-06,
-      "logits/chosen": -2.8312201499938965,
-      "logits/rejected": -2.821024179458618,
-      "logps/chosen": -25.73629379272461,
-      "logps/rejected": -29.133464813232422,
-      "loss": 0.4609,
-      "rewards/accuracies": 0.6000000238418579,
-      "rewards/chosen": 0.24358424544334412,
-      "rewards/margins": 0.1639489382505417,
-      "rewards/rejected": 0.07963528484106064,
+      "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,
       "step": 250
     },
     {
       "epoch": 0.68,
-      "grad_norm": 3.625,
+      "grad_norm": 89.0,
       "learning_rate": 1.4445974030621963e-06,
-      "logits/chosen": -2.8129467964172363,
-      "logits/rejected": -2.8341376781463623,
-      "logps/chosen": -28.19793701171875,
-      "logps/rejected": -33.99031448364258,
-      "loss": 0.436,
-      "rewards/accuracies": 0.7124999761581421,
-      "rewards/chosen": 0.3264053463935852,
-      "rewards/margins": 0.2896498441696167,
-      "rewards/rejected": 0.036755502223968506,
+      "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,
       "step": 260
     },
     {
       "epoch": 0.7,
-      "grad_norm": 3.484375,
+      "grad_norm": 54.25,
       "learning_rate": 1.243452991757889e-06,
-      "logits/chosen": -2.9577534198760986,
-      "logits/rejected": -2.9646975994110107,
-      "logps/chosen": -29.623525619506836,
-      "logps/rejected": -29.91342544555664,
-      "loss": 0.4449,
-      "rewards/accuracies": 0.762499988079071,
-      "rewards/chosen": 0.23156669735908508,
-      "rewards/margins": 0.2411995679140091,
-      "rewards/rejected": -0.009632894769310951,
+      "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,
       "step": 270
     },
     {
       "epoch": 0.73,
-      "grad_norm": 3.53125,
+      "grad_norm": 78.0,
       "learning_rate": 1.0526606671603523e-06,
-      "logits/chosen": -2.973743200302124,
-      "logits/rejected": -2.960118532180786,
-      "logps/chosen": -29.57967185974121,
-      "logps/rejected": -28.558263778686523,
-      "loss": 0.4498,
-      "rewards/accuracies": 0.6625000238418579,
-      "rewards/chosen": 0.22840175032615662,
-      "rewards/margins": 0.2137848436832428,
-      "rewards/rejected": 0.014616942033171654,
+      "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,
       "step": 280
     },
     {
       "epoch": 0.75,
-      "grad_norm": 2.9375,
+      "grad_norm": 227.0,
       "learning_rate": 8.737922755071455e-07,
-      "logits/chosen": -2.901043653488159,
-      "logits/rejected": -2.883784294128418,
-      "logps/chosen": -30.675678253173828,
-      "logps/rejected": -30.71207618713379,
-      "loss": 0.4063,
-      "rewards/accuracies": 0.800000011920929,
-      "rewards/chosen": 0.3921253979206085,
-      "rewards/margins": 0.4298785626888275,
-      "rewards/rejected": -0.03775321692228317,
+      "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,
       "step": 290
     },
     {
       "epoch": 0.78,
-      "grad_norm": 3.703125,
+      "grad_norm": 340.0,
       "learning_rate": 7.08321427484816e-07,
-      "logits/chosen": -2.905961513519287,
-      "logits/rejected": -2.9014759063720703,
-      "logps/chosen": -30.50136947631836,
-      "logps/rejected": -27.470678329467773,
-      "loss": 0.4326,
-      "rewards/accuracies": 0.637499988079071,
-      "rewards/chosen": 0.3442566692829132,
-      "rewards/margins": 0.3176761269569397,
-      "rewards/rejected": 0.02658059261739254,
+      "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,
       "step": 300
     },
     {
       "epoch": 0.78,
-      "eval_logits/chosen": -2.820904493331909,
-      "eval_logits/rejected": -2.8182966709136963,
-      "eval_logps/chosen": -31.072538375854492,
-      "eval_logps/rejected": -34.71522521972656,
-      "eval_loss": 0.49085888266563416,
-      "eval_rewards/accuracies": 0.6013289093971252,
-      "eval_rewards/chosen": 0.04198317229747772,
-      "eval_rewards/margins": 0.045396942645311356,
-      "eval_rewards/rejected": -0.0034137717448174953,
-      "eval_runtime": 112.9358,
-      "eval_samples_per_second": 3.037,
-      "eval_steps_per_second": 0.381,
+      "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_steps_per_second": 0.38,
       "step": 300
     },
     {
       "epoch": 0.81,
-      "grad_norm": 3.234375,
+      "grad_norm": 251.0,
       "learning_rate": 5.576113578589035e-07,
-      "logits/chosen": -2.781768321990967,
-      "logits/rejected": -2.799870014190674,
-      "logps/chosen": -28.066579818725586,
-      "logps/rejected": -30.81374740600586,
-      "loss": 0.4473,
-      "rewards/accuracies": 0.75,
-      "rewards/chosen": 0.23256821930408478,
-      "rewards/margins": 0.24891658127307892,
-      "rewards/rejected": -0.01634838618338108,
+      "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,
       "step": 310
     },
     {
       "epoch": 0.83,
-      "grad_norm": 3.5625,
+      "grad_norm": 93.5,
       "learning_rate": 4.229036944380913e-07,
-      "logits/chosen": -3.0310263633728027,
-      "logits/rejected": -3.016596794128418,
-      "logps/chosen": -28.514297485351562,
-      "logps/rejected": -28.385051727294922,
-      "loss": 0.424,
-      "rewards/accuracies": 0.737500011920929,
-      "rewards/chosen": 0.3394576907157898,
-      "rewards/margins": 0.35271310806274414,
-      "rewards/rejected": -0.013255435042083263,
+      "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,
       "step": 320
     },
     {
       "epoch": 0.86,
-      "grad_norm": 3.234375,
+      "grad_norm": 119.5,
       "learning_rate": 3.053082288996112e-07,
-      "logits/chosen": -2.94514799118042,
-      "logits/rejected": -2.927401065826416,
-      "logps/chosen": -26.721960067749023,
-      "logps/rejected": -30.596094131469727,
-      "loss": 0.4178,
-      "rewards/accuracies": 0.7749999761581421,
-      "rewards/chosen": 0.3280225396156311,
-      "rewards/margins": 0.3689119517803192,
-      "rewards/rejected": -0.0408894307911396,
+      "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,
       "step": 330
     },
     {
       "epoch": 0.88,
-      "grad_norm": 2.71875,
+      "grad_norm": 131.0,
       "learning_rate": 2.0579377374915805e-07,
-      "logits/chosen": -3.162492275238037,
-      "logits/rejected": -3.1681036949157715,
-      "logps/chosen": -30.14105796813965,
-      "logps/rejected": -32.993080139160156,
-      "loss": 0.4252,
+      "logits/chosen": -3.146080493927002,
+      "logits/rejected": -3.152937650680542,
+      "logps/chosen": -31.1461181640625,
+      "logps/rejected": -32.72796630859375,
+      "loss": 0.797,
       "rewards/accuracies": 0.762499988079071,
-      "rewards/chosen": 0.2996302545070648,
-      "rewards/margins": 0.3590452969074249,
-      "rewards/rejected": -0.05941504240036011,
+      "rewards/chosen": 0.3451611399650574,
+      "rewards/margins": 0.3675331473350525,
+      "rewards/rejected": -0.022372011095285416,
       "step": 340
     },
     {
       "epoch": 0.91,
-      "grad_norm": 3.3125,
+      "grad_norm": 117.0,
       "learning_rate": 1.2518018074041684e-07,
-      "logits/chosen": -3.0381181240081787,
-      "logits/rejected": -3.0421128273010254,
-      "logps/chosen": -29.018634796142578,
-      "logps/rejected": -31.29150390625,
-      "loss": 0.432,
+      "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": 0.3576488494873047,
-      "rewards/margins": 0.3132873773574829,
-      "rewards/rejected": 0.044361427426338196,
+      "rewards/chosen": 0.33320528268814087,
+      "rewards/margins": 0.334422767162323,
+      "rewards/rejected": -0.001217484474182129,
       "step": 350
     },
     {
       "epoch": 0.94,
-      "grad_norm": 4.09375,
+      "grad_norm": 223.0,
       "learning_rate": 6.41315865106129e-08,
-      "logits/chosen": -2.8777647018432617,
-      "logits/rejected": -2.879361152648926,
-      "logps/chosen": -26.779287338256836,
-      "logps/rejected": -29.69747543334961,
-      "loss": 0.4224,
-      "rewards/accuracies": 0.737500011920929,
-      "rewards/chosen": 0.374088317155838,
-      "rewards/margins": 0.34739503264427185,
-      "rewards/rejected": 0.026693273335695267,
+      "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,
       "step": 360
     },
     {
       "epoch": 0.96,
-      "grad_norm": 5.34375,
+      "grad_norm": 127.5,
       "learning_rate": 2.3150941078050325e-08,
-      "logits/chosen": -2.953718900680542,
-      "logits/rejected": -2.9520201683044434,
-      "logps/chosen": -29.355998992919922,
-      "logps/rejected": -32.019020080566406,
-      "loss": 0.4352,
-      "rewards/accuracies": 0.7124999761581421,
-      "rewards/chosen": 0.28613001108169556,
-      "rewards/margins": 0.30432528257369995,
-      "rewards/rejected": -0.01819526217877865,
+      "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,
       "step": 370
     },
     {
       "epoch": 0.99,
-      "grad_norm": 3.390625,
+      "grad_norm": 484.0,
       "learning_rate": 2.575864278703266e-09,
-      "logits/chosen": -2.9095873832702637,
-      "logits/rejected": -2.892209053039551,
-      "logps/chosen": -27.843591690063477,
-      "logps/rejected": -28.02511215209961,
-      "loss": 0.4472,
-      "rewards/accuracies": 0.7250000238418579,
-      "rewards/chosen": 0.22028858959674835,
-      "rewards/margins": 0.24738721549510956,
-      "rewards/rejected": -0.02709861472249031,
+      "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,
       "step": 380
     },
     {
       "epoch": 1.0,
       "step": 385,
       "total_flos": 0.0,
-      "train_loss": 0.4552104888024268,
-      "train_runtime": 2721.944,
-      "train_samples_per_second": 1.131,
+      "train_loss": 1.2163990609057538,
+      "train_runtime": 2720.9807,
+      "train_samples_per_second": 1.132,
       "train_steps_per_second": 0.141
     }
   ],