Model save

README.md

@@ -0,0 +1,82 @@
+---
+license: apache-2.0
+library_name: peft
+tags:
+- trl
+- dpo
+- generated_from_trainer
+base_model: mistralai/Mistral-7B-v0.1
+model-index:
+- name: zephyr-7b-gpo-iter0
+  results: []
+---
+
+<!-- This model card has been generated automatically according to the information the Trainer had access to. You
+should probably proofread and complete it, then remove this comment. -->
+
+# zephyr-7b-gpo-iter0
+
+This model is a fine-tuned version of [mistralai/Mistral-7B-v0.1](https://huggingface.co/mistralai/Mistral-7B-v0.1) on the None dataset.
+It achieves the following results on the evaluation set:
+- Loss: 0.0258
+- Rewards/chosen: -0.0580
+- Rewards/rejected: -0.0061
+- Rewards/accuracies: 0.3380
+- Rewards/margins: -0.0519
+- Logps/rejected: -249.4468
+- Logps/chosen: -274.3866
+- Logits/rejected: -2.2108
+- Logits/chosen: -2.4070
+
+## Model description
+
+More information needed
+
+## Intended uses & limitations
+
+More information needed
+
+## Training and evaluation data
+
+More information needed
+
+## Training procedure
+
+### Training hyperparameters
+
+The following hyperparameters were used during training:
+- learning_rate: 5e-06
+- train_batch_size: 1
+- eval_batch_size: 2
+- seed: 42
+- distributed_type: multi-GPU
+- gradient_accumulation_steps: 2
+- total_train_batch_size: 2
+- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
+- lr_scheduler_type: cosine
+- lr_scheduler_warmup_ratio: 0.1
+- num_epochs: 2
+
+### Training results
+
+| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
+|:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
+| 0.0008        | 0.2   | 100  | 0.0019          | -0.0111        | -0.0138          | 0.5300             | 0.0027          | -250.2170      | -269.6990    | -2.2026         | -2.4007       |
+| 0.0006        | 0.4   | 200  | 0.0029          | -0.0237        | -0.0230          | 0.4910             | -0.0007         | -251.1392      | -270.9541    | -2.2051         | -2.4034       |
+| 0.001         | 0.6   | 300  | 0.0019          | -0.0120        | -0.0142          | 0.5310             | 0.0022          | -250.2602      | -269.7912    | -2.2008         | -2.3984       |
+| 0.0011        | 0.8   | 400  | 0.0023          | -0.0201        | -0.0211          | 0.5010             | 0.0011          | -250.9541      | -270.5950    | -2.1993         | -2.3968       |
+| 0.0008        | 1.0   | 500  | 0.0021          | -0.0170        | -0.0189          | 0.5065             | 0.0019          | -250.7260      | -270.2850    | -2.1982         | -2.3960       |
+| 0.044         | 1.2   | 600  | 0.0091          | -0.0053        | 0.0198           | 0.3600             | -0.0252         | -246.8548      | -269.1194    | -2.1940         | -2.3899       |
+| 0.0682        | 1.4   | 700  | 0.0191          | -0.0345        | 0.0086           | 0.3450             | -0.0431         | -247.9818      | -272.0423    | -2.2035         | -2.3992       |
+| 0.0505        | 1.6   | 800  | 0.0237          | -0.0497        | -0.0001          | 0.3405             | -0.0496         | -248.8542      | -273.5587    | -2.2094         | -2.4056       |
+| 0.0243        | 1.8   | 900  | 0.0259          | -0.0581        | -0.0062          | 0.3340             | -0.0519         | -249.4570      | -274.3967    | -2.2117         | -2.4081       |
+| 0.0697        | 2.0   | 1000 | 0.0258          | -0.0580        | -0.0061          | 0.3380             | -0.0519         | -249.4468      | -274.3866    | -2.2108         | -2.4070       |
+
+
+### Framework versions
+
+- PEFT 0.7.1
+- Transformers 4.36.2
+- Pytorch 2.1.2+cu118
+- Datasets 2.14.6
+- Tokenizers 0.15.2

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d62e9aa2c8e98056d8063360fc5b8f7d2dbab7db13276c8eaa780ed05df0fb43
+oid sha256:bf14c9e09a2e0c9af1b622451b19e1881cb4891019a3c54abfb5de9de084f54d
 size 671150064

all_results.json

@@ -0,0 +1,21 @@
+{
+    "epoch": 2.0,
+    "eval_logits/chosen": -2.4070470333099365,
+    "eval_logits/rejected": -2.210773468017578,
+    "eval_logps/chosen": -274.3865966796875,
+    "eval_logps/rejected": -249.44677734375,
+    "eval_loss": 0.0258334930986166,
+    "eval_rewards/accuracies": 0.33799999952316284,
+    "eval_rewards/chosen": -0.05798804759979248,
+    "eval_rewards/margins": -0.05191566422581673,
+    "eval_rewards/rejected": -0.006072388496249914,
+    "eval_runtime": 1420.6371,
+    "eval_samples": 2000,
+    "eval_samples_per_second": 1.408,
+    "eval_steps_per_second": 0.704,
+    "train_loss": 0.019690872263745406,
+    "train_runtime": 18916.4989,
+    "train_samples": 61135,
+    "train_samples_per_second": 0.106,
+    "train_steps_per_second": 0.053
+}

eval_results.json

@@ -0,0 +1,16 @@
+{
+    "epoch": 2.0,
+    "eval_logits/chosen": -2.4070470333099365,
+    "eval_logits/rejected": -2.210773468017578,
+    "eval_logps/chosen": -274.3865966796875,
+    "eval_logps/rejected": -249.44677734375,
+    "eval_loss": 0.0258334930986166,
+    "eval_rewards/accuracies": 0.33799999952316284,
+    "eval_rewards/chosen": -0.05798804759979248,
+    "eval_rewards/margins": -0.05191566422581673,
+    "eval_rewards/rejected": -0.006072388496249914,
+    "eval_runtime": 1420.6371,
+    "eval_samples": 2000,
+    "eval_samples_per_second": 1.408,
+    "eval_steps_per_second": 0.704
+}

runs/Mar26_14-44-48_gpu4-119-4/events.out.tfevents.1711424947.gpu4-119-4.803138.0

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:04489168647bc6186b6a0502343e73ed185707fa750fe7006d25395ec574f725
-size 68763
+oid sha256:1061ee19f4cc392c0238618372bdde5251cb8ce7cf6d3912ea305f9cb85fb4ad
+size 76197

runs/Mar26_14-44-48_gpu4-119-4/events.out.tfevents.1711445328.gpu4-119-4.803138.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0fdbc205643a8074faab447db30e4dace56b9dc9ced5521f0bad7811bcace628
+size 828

train_results.json

@@ -0,0 +1,8 @@
+{
+    "epoch": 2.0,
+    "train_loss": 0.019690872263745406,
+    "train_runtime": 18916.4989,
+    "train_samples": 61135,
+    "train_samples_per_second": 0.106,
+    "train_steps_per_second": 0.053
+}

trainer_state.json

@@ -0,0 +1,1604 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.0,
+  "eval_steps": 100,
+  "global_step": 1000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.0000000000000004e-08,
+      "logits/chosen": -3.1812663078308105,
+      "logits/rejected": -2.810722827911377,
+      "logps/chosen": -118.07083129882812,
+      "logps/rejected": -100.80718994140625,
+      "loss": 0.0011,
+      "rewards/accuracies": 0.0,
+      "rewards/chosen": 0.0,
+      "rewards/margins": 0.0,
+      "rewards/rejected": 0.0,
+      "step": 1
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.000000000000001e-07,
+      "logits/chosen": -2.9193410873413086,
+      "logits/rejected": -2.7373409271240234,
+      "logps/chosen": -116.89999389648438,
+      "logps/rejected": -116.49290466308594,
+      "loss": 0.0011,
+      "rewards/accuracies": 0.3888888955116272,
+      "rewards/chosen": -0.00038699465221725404,
+      "rewards/margins": -0.0005435158964246511,
+      "rewards/rejected": 0.00015652130241505802,
+      "step": 10
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.0000000000000002e-06,
+      "logits/chosen": -2.981767416000366,
+      "logits/rejected": -2.6092238426208496,
+      "logps/chosen": -164.24819946289062,
+      "logps/rejected": -164.7591552734375,
+      "loss": 0.0011,
+      "rewards/accuracies": 0.4000000059604645,
+      "rewards/chosen": -6.044005203875713e-05,
+      "rewards/margins": -0.0003916244604624808,
+      "rewards/rejected": 0.0003311844193376601,
+      "step": 20
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.5e-06,
+      "logits/chosen": -2.913992404937744,
+      "logits/rejected": -2.6422486305236816,
+      "logps/chosen": -153.99203491210938,
+      "logps/rejected": -141.27601623535156,
+      "loss": 0.0011,
+      "rewards/accuracies": 0.44999998807907104,
+      "rewards/chosen": -0.00013318247511051595,
+      "rewards/margins": 0.0008146190084517002,
+      "rewards/rejected": -0.0009478016872890294,
+      "step": 30
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.0000000000000003e-06,
+      "logits/chosen": -2.9979870319366455,
+      "logits/rejected": -2.5024120807647705,
+      "logps/chosen": -180.6337890625,
+      "logps/rejected": -168.78204345703125,
+      "loss": 0.001,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": 0.002415223978459835,
+      "rewards/margins": 0.002099759876728058,
+      "rewards/rejected": 0.0003154640144202858,
+      "step": 40
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.5e-06,
+      "logits/chosen": -2.7826578617095947,
+      "logits/rejected": -2.437479019165039,
+      "logps/chosen": -167.60964965820312,
+      "logps/rejected": -154.45205688476562,
+      "loss": 0.001,
+      "rewards/accuracies": 0.550000011920929,
+      "rewards/chosen": 0.0028098905459046364,
+      "rewards/margins": 0.0014571474166586995,
+      "rewards/rejected": 0.0013527432456612587,
+      "step": 50
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3e-06,
+      "logits/chosen": -2.715552806854248,
+      "logits/rejected": -2.5350234508514404,
+      "logps/chosen": -147.0255126953125,
+      "logps/rejected": -136.99874877929688,
+      "loss": 0.0011,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.001564659527502954,
+      "rewards/margins": 0.0012098618317395449,
+      "rewards/rejected": 0.0003547971718944609,
+      "step": 60
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 3.5e-06,
+      "logits/chosen": -3.064145088195801,
+      "logits/rejected": -2.715378522872925,
+      "logps/chosen": -163.35397338867188,
+      "logps/rejected": -152.8401336669922,
+      "loss": 0.001,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.011319306679069996,
+      "rewards/margins": 0.002820510882884264,
+      "rewards/rejected": 0.00849879626184702,
+      "step": 70
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.000000000000001e-06,
+      "logits/chosen": -3.031562089920044,
+      "logits/rejected": -2.762894868850708,
+      "logps/chosen": -154.64096069335938,
+      "logps/rejected": -142.45066833496094,
+      "loss": 0.001,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": 0.015446094796061516,
+      "rewards/margins": 0.0030269301496446133,
+      "rewards/rejected": 0.012419164180755615,
+      "step": 80
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.5e-06,
+      "logits/chosen": -3.0730316638946533,
+      "logits/rejected": -2.5636696815490723,
+      "logps/chosen": -126.04837799072266,
+      "logps/rejected": -110.87501525878906,
+      "loss": 0.0009,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.018308930099010468,
+      "rewards/margins": 0.008363587781786919,
+      "rewards/rejected": 0.009945342317223549,
+      "step": 90
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 5e-06,
+      "logits/chosen": -2.9546635150909424,
+      "logits/rejected": -2.5928895473480225,
+      "logps/chosen": -133.506103515625,
+      "logps/rejected": -123.97562408447266,
+      "loss": 0.0008,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": 0.014314673840999603,
+      "rewards/margins": 0.005265203304588795,
+      "rewards/rejected": 0.009049469605088234,
+      "step": 100
+    },
+    {
+      "epoch": 0.2,
+      "eval_logits/chosen": -2.4007108211517334,
+      "eval_logits/rejected": -2.2025678157806396,
+      "eval_logps/chosen": -269.69903564453125,
+      "eval_logps/rejected": -250.2169952392578,
+      "eval_loss": 0.0019292256329208612,
+      "eval_rewards/accuracies": 0.5299999713897705,
+      "eval_rewards/chosen": -0.01111243013292551,
+      "eval_rewards/margins": 0.002662122482433915,
+      "eval_rewards/rejected": -0.013774552382528782,
+      "eval_runtime": 1423.4882,
+      "eval_samples_per_second": 1.405,
+      "eval_steps_per_second": 0.702,
+      "step": 100
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.99847706754774e-06,
+      "logits/chosen": -3.0832509994506836,
+      "logits/rejected": -2.6540307998657227,
+      "logps/chosen": -158.02037048339844,
+      "logps/rejected": -128.09632873535156,
+      "loss": 0.0012,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": 0.03201522305607796,
+      "rewards/margins": 0.014926651492714882,
+      "rewards/rejected": 0.017088569700717926,
+      "step": 110
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.993910125649561e-06,
+      "logits/chosen": -3.000650644302368,
+      "logits/rejected": -2.5093066692352295,
+      "logps/chosen": -111.5718765258789,
+      "logps/rejected": -91.67008209228516,
+      "loss": 0.0007,
+      "rewards/accuracies": 0.8500000238418579,
+      "rewards/chosen": 0.017279163002967834,
+      "rewards/margins": 0.016767729073762894,
+      "rewards/rejected": 0.0005114328814670444,
+      "step": 120
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.986304738420684e-06,
+      "logits/chosen": -2.8899765014648438,
+      "logits/rejected": -2.545956611633301,
+      "logps/chosen": -157.5283660888672,
+      "logps/rejected": -135.06536865234375,
+      "loss": 0.001,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": 0.006769341416656971,
+      "rewards/margins": 0.013469241559505463,
+      "rewards/rejected": -0.006699901074171066,
+      "step": 130
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.975670171853926e-06,
+      "logits/chosen": -3.0038022994995117,
+      "logits/rejected": -2.8060638904571533,
+      "logps/chosen": -172.43551635742188,
+      "logps/rejected": -161.0656280517578,
+      "loss": 0.001,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.006354503333568573,
+      "rewards/margins": 0.004105663392692804,
+      "rewards/rejected": -0.01046016626060009,
+      "step": 140
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.962019382530521e-06,
+      "logits/chosen": -2.938009023666382,
+      "logits/rejected": -2.3462884426116943,
+      "logps/chosen": -113.1685791015625,
+      "logps/rejected": -100.8033676147461,
+      "loss": 0.001,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.00343899498693645,
+      "rewards/margins": 0.007120263762772083,
+      "rewards/rejected": -0.0036812685430049896,
+      "step": 150
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.9453690018345144e-06,
+      "logits/chosen": -3.1557974815368652,
+      "logits/rejected": -2.805929183959961,
+      "logps/chosen": -212.8404998779297,
+      "logps/rejected": -204.03390502929688,
+      "loss": 0.0013,
+      "rewards/accuracies": 0.550000011920929,
+      "rewards/chosen": -0.006425648927688599,
+      "rewards/margins": -0.0002856804057955742,
+      "rewards/rejected": -0.0061399685218930244,
+      "step": 160
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 4.925739315689991e-06,
+      "logits/chosen": -2.8449888229370117,
+      "logits/rejected": -2.6668992042541504,
+      "logps/chosen": -205.7440643310547,
+      "logps/rejected": -196.88563537597656,
+      "loss": 0.0012,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.005873252637684345,
+      "rewards/margins": 0.0015904292231425643,
+      "rewards/rejected": -0.007463681511580944,
+      "step": 170
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 4.903154239845798e-06,
+      "logits/chosen": -3.059779167175293,
+      "logits/rejected": -2.7152631282806396,
+      "logps/chosen": -178.5559539794922,
+      "logps/rejected": -174.60049438476562,
+      "loss": 0.001,
+      "rewards/accuracies": 0.4000000059604645,
+      "rewards/chosen": 0.004108003806322813,
+      "rewards/margins": 0.0049694306217134,
+      "rewards/rejected": -0.0008614275720901787,
+      "step": 180
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 4.8776412907378845e-06,
+      "logits/chosen": -2.9436910152435303,
+      "logits/rejected": -2.4861669540405273,
+      "logps/chosen": -154.61508178710938,
+      "logps/rejected": -129.8065643310547,
+      "loss": 0.0008,
+      "rewards/accuracies": 0.8500000238418579,
+      "rewards/chosen": 0.02211841568350792,
+      "rewards/margins": 0.012492652982473373,
+      "rewards/rejected": 0.00962576363235712,
+      "step": 190
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.849231551964771e-06,
+      "logits/chosen": -2.934830904006958,
+      "logits/rejected": -2.599217653274536,
+      "logps/chosen": -148.74435424804688,
+      "logps/rejected": -139.92453002929688,
+      "loss": 0.0006,
+      "rewards/accuracies": 0.8500000238418579,
+      "rewards/chosen": 0.02799961529672146,
+      "rewards/margins": 0.012480301782488823,
+      "rewards/rejected": 0.015519311651587486,
+      "step": 200
+    },
+    {
+      "epoch": 0.4,
+      "eval_logits/chosen": -2.4034218788146973,
+      "eval_logits/rejected": -2.205068349838257,
+      "eval_logps/chosen": -270.9541015625,
+      "eval_logps/rejected": -251.13922119140625,
+      "eval_loss": 0.002935634693130851,
+      "eval_rewards/accuracies": 0.4909999966621399,
+      "eval_rewards/chosen": -0.0236628670245409,
+      "eval_rewards/margins": -0.0006660502986051142,
+      "eval_rewards/rejected": -0.022996816784143448,
+      "eval_runtime": 1423.7196,
+      "eval_samples_per_second": 1.405,
+      "eval_steps_per_second": 0.702,
+      "step": 200
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 4.817959636416969e-06,
+      "logits/chosen": -2.7869691848754883,
+      "logits/rejected": -2.3660902976989746,
+      "logps/chosen": -188.44235229492188,
+      "logps/rejected": -165.0704345703125,
+      "loss": 0.0007,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.023742977529764175,
+      "rewards/margins": 0.013938216492533684,
+      "rewards/rejected": 0.009804759174585342,
+      "step": 210
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 4.783863644106502e-06,
+      "logits/chosen": -2.7817561626434326,
+      "logits/rejected": -2.4513843059539795,
+      "logps/chosen": -129.5194854736328,
+      "logps/rejected": -119.9756088256836,
+      "loss": 0.0011,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.014000030234456062,
+      "rewards/margins": 0.006808738224208355,
+      "rewards/rejected": 0.007191292010247707,
+      "step": 220
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 4.746985115747918e-06,
+      "logits/chosen": -2.997894525527954,
+      "logits/rejected": -2.7682430744171143,
+      "logps/chosen": -206.5100860595703,
+      "logps/rejected": -196.3065185546875,
+      "loss": 0.0007,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": 0.01694752275943756,
+      "rewards/margins": 0.008601350709795952,
+      "rewards/rejected": 0.00834617204964161,
+      "step": 230
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 4.707368982147318e-06,
+      "logits/chosen": -2.9324448108673096,
+      "logits/rejected": -2.6252548694610596,
+      "logps/chosen": -140.8667449951172,
+      "logps/rejected": -124.31671142578125,
+      "loss": 0.001,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.010156641714274883,
+      "rewards/margins": 0.011887853033840656,
+      "rewards/rejected": -0.0017312124837189913,
+      "step": 240
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.665063509461098e-06,
+      "logits/chosen": -3.0423521995544434,
+      "logits/rejected": -2.578011989593506,
+      "logps/chosen": -136.1217498779297,
+      "logps/rejected": -116.2592544555664,
+      "loss": 0.0006,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": 0.022909339517354965,
+      "rewards/margins": 0.01127886213362217,
+      "rewards/rejected": 0.011630477383732796,
+      "step": 250
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.620120240391065e-06,
+      "logits/chosen": -2.8443424701690674,
+      "logits/rejected": -2.6343538761138916,
+      "logps/chosen": -100.57234191894531,
+      "logps/rejected": -98.87772369384766,
+      "loss": 0.0011,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": 0.010612092912197113,
+      "rewards/margins": 0.005750484298914671,
+      "rewards/rejected": 0.004861608147621155,
+      "step": 260
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.572593931387604e-06,
+      "logits/chosen": -3.0539159774780273,
+      "logits/rejected": -2.753909111022949,
+      "logps/chosen": -126.12535095214844,
+      "logps/rejected": -113.8405990600586,
+      "loss": 0.0008,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.00659831753000617,
+      "rewards/margins": 0.010872049257159233,
+      "rewards/rejected": -0.004273730795830488,
+      "step": 270
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.522542485937369e-06,
+      "logits/chosen": -3.0358662605285645,
+      "logits/rejected": -2.4368128776550293,
+      "logps/chosen": -150.52206420898438,
+      "logps/rejected": -120.9325942993164,
+      "loss": 0.0008,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": 0.015789732336997986,
+      "rewards/margins": 0.017957117408514023,
+      "rewards/rejected": -0.002167386468499899,
+      "step": 280
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.470026884016805e-06,
+      "logits/chosen": -3.0055549144744873,
+      "logits/rejected": -2.710265636444092,
+      "logps/chosen": -143.09898376464844,
+      "logps/rejected": -144.5852508544922,
+      "loss": 0.0014,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": 0.013310904614627361,
+      "rewards/margins": 0.008920473977923393,
+      "rewards/rejected": 0.004390430636703968,
+      "step": 290
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.415111107797445e-06,
+      "logits/chosen": -2.88038969039917,
+      "logits/rejected": -2.524245500564575,
+      "logps/chosen": -116.92478942871094,
+      "logps/rejected": -106.48399353027344,
+      "loss": 0.001,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": -0.016026504337787628,
+      "rewards/margins": 0.011352911591529846,
+      "rewards/rejected": -0.027379417791962624,
+      "step": 300
+    },
+    {
+      "epoch": 0.6,
+      "eval_logits/chosen": -2.3983988761901855,
+      "eval_logits/rejected": -2.200815200805664,
+      "eval_logps/chosen": -269.79119873046875,
+      "eval_logps/rejected": -250.26023864746094,
+      "eval_loss": 0.0018744752742350101,
+      "eval_rewards/accuracies": 0.531000018119812,
+      "eval_rewards/chosen": -0.012033737264573574,
+      "eval_rewards/margins": 0.0021730433218181133,
+      "eval_rewards/rejected": -0.01420677825808525,
+      "eval_runtime": 1423.8246,
+      "eval_samples_per_second": 1.405,
+      "eval_steps_per_second": 0.702,
+      "step": 300
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 4.357862063693486e-06,
+      "logits/chosen": -3.0442090034484863,
+      "logits/rejected": -2.7478137016296387,
+      "logps/chosen": -200.55148315429688,
+      "logps/rejected": -198.62619018554688,
+      "loss": 0.001,
+      "rewards/accuracies": 0.550000011920929,
+      "rewards/chosen": -0.01667696237564087,
+      "rewards/margins": 0.007582807447761297,
+      "rewards/rejected": -0.024259772151708603,
+      "step": 310
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.2983495008466285e-06,
+      "logits/chosen": -2.8576464653015137,
+      "logits/rejected": -2.5876190662384033,
+      "logps/chosen": -129.97561645507812,
+      "logps/rejected": -118.2795181274414,
+      "loss": 0.0013,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.01680707558989525,
+      "rewards/margins": 0.00610563438385725,
+      "rewards/rejected": -0.022912709042429924,
+      "step": 320
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 4.236645926147493e-06,
+      "logits/chosen": -3.026426315307617,
+      "logits/rejected": -2.7503247261047363,
+      "logps/chosen": -146.84666442871094,
+      "logps/rejected": -139.24612426757812,
+      "loss": 0.001,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.014562124386429787,
+      "rewards/margins": 0.005175650119781494,
+      "rewards/rejected": 0.009386474266648293,
+      "step": 330
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.172826515897146e-06,
+      "logits/chosen": -3.144963502883911,
+      "logits/rejected": -2.831359624862671,
+      "logps/chosen": -169.22183227539062,
+      "logps/rejected": -168.03089904785156,
+      "loss": 0.0012,
+      "rewards/accuracies": 0.44999998807907104,
+      "rewards/chosen": 0.004123937338590622,
+      "rewards/margins": 0.0042907195165753365,
+      "rewards/rejected": -0.00016678198880981654,
+      "step": 340
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.106969024216348e-06,
+      "logits/chosen": -3.0851778984069824,
+      "logits/rejected": -2.769751787185669,
+      "logps/chosen": -130.99282836914062,
+      "logps/rejected": -126.33982849121094,
+      "loss": 0.0012,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": 0.007414456456899643,
+      "rewards/margins": 0.006272236816585064,
+      "rewards/rejected": 0.0011422175448387861,
+      "step": 350
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.039153688314146e-06,
+      "logits/chosen": -3.0202767848968506,
+      "logits/rejected": -2.817288637161255,
+      "logps/chosen": -201.23435974121094,
+      "logps/rejected": -194.98385620117188,
+      "loss": 0.0021,
+      "rewards/accuracies": 0.4000000059604645,
+      "rewards/chosen": 0.04826142266392708,
+      "rewards/margins": -0.002473982283845544,
+      "rewards/rejected": 0.05073540285229683,
+      "step": 360
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.969463130731183e-06,
+      "logits/chosen": -2.9788031578063965,
+      "logits/rejected": -2.5181713104248047,
+      "logps/chosen": -172.3706817626953,
+      "logps/rejected": -149.59950256347656,
+      "loss": 0.0009,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": 0.009602314792573452,
+      "rewards/margins": 0.012693755328655243,
+      "rewards/rejected": -0.00309143983758986,
+      "step": 370
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.897982258676867e-06,
+      "logits/chosen": -3.0599074363708496,
+      "logits/rejected": -2.749642848968506,
+      "logps/chosen": -216.9517822265625,
+      "logps/rejected": -202.60659790039062,
+      "loss": 0.0012,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.008673595264554024,
+      "rewards/margins": 0.00854658242315054,
+      "rewards/rejected": -0.01722017675638199,
+      "step": 380
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.824798160583012e-06,
+      "logits/chosen": -2.6857285499572754,
+      "logits/rejected": -2.262728214263916,
+      "logps/chosen": -133.64395141601562,
+      "logps/rejected": -119.55937194824219,
+      "loss": 0.0009,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.020837822929024696,
+      "rewards/margins": 0.01239033229649067,
+      "rewards/rejected": 0.008447489701211452,
+      "step": 390
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.7500000000000005e-06,
+      "logits/chosen": -2.9022040367126465,
+      "logits/rejected": -2.7616379261016846,
+      "logps/chosen": -185.07240295410156,
+      "logps/rejected": -179.7260284423828,
+      "loss": 0.0011,
+      "rewards/accuracies": 0.44999998807907104,
+      "rewards/chosen": 0.003294271184131503,
+      "rewards/margins": 0.0035428921692073345,
+      "rewards/rejected": -0.0002486211305949837,
+      "step": 400
+    },
+    {
+      "epoch": 0.8,
+      "eval_logits/chosen": -2.396838426589966,
+      "eval_logits/rejected": -2.1993439197540283,
+      "eval_logps/chosen": -270.5950012207031,
+      "eval_logps/rejected": -250.95413208007812,
+      "eval_loss": 0.0023364874068647623,
+      "eval_rewards/accuracies": 0.5009999871253967,
+      "eval_rewards/chosen": -0.02007202059030533,
+      "eval_rewards/margins": 0.0010739399585872889,
+      "eval_rewards/rejected": -0.021145964041352272,
+      "eval_runtime": 1423.743,
+      "eval_samples_per_second": 1.405,
+      "eval_steps_per_second": 0.702,
+      "step": 400
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.6736789069647273e-06,
+      "logits/chosen": -3.0739264488220215,
+      "logits/rejected": -2.8628082275390625,
+      "logps/chosen": -149.10903930664062,
+      "logps/rejected": -134.99038696289062,
+      "loss": 0.0011,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.008919402956962585,
+      "rewards/margins": 0.0073748803697526455,
+      "rewards/rejected": 0.0015445235185325146,
+      "step": 410
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.595927866972694e-06,
+      "logits/chosen": -3.0844318866729736,
+      "logits/rejected": -2.8257293701171875,
+      "logps/chosen": -252.5592498779297,
+      "logps/rejected": -248.02474975585938,
+      "loss": 0.001,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.011812982149422169,
+      "rewards/margins": 0.007170848548412323,
+      "rewards/rejected": -0.018983829766511917,
+      "step": 420
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 3.516841607689501e-06,
+      "logits/chosen": -3.055979013442993,
+      "logits/rejected": -2.7850348949432373,
+      "logps/chosen": -147.57447814941406,
+      "logps/rejected": -138.53738403320312,
+      "loss": 0.0009,
+      "rewards/accuracies": 0.550000011920929,
+      "rewards/chosen": -0.005650115665048361,
+      "rewards/margins": 0.012325585819780827,
+      "rewards/rejected": -0.0179757010191679,
+      "step": 430
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.436516483539781e-06,
+      "logits/chosen": -2.7994961738586426,
+      "logits/rejected": -2.700761079788208,
+      "logps/chosen": -189.2479248046875,
+      "logps/rejected": -180.1698760986328,
+      "loss": 0.0016,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.003194189164787531,
+      "rewards/margins": -0.00013344390026759356,
+      "rewards/rejected": -0.0030607457738369703,
+      "step": 440
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 3.3550503583141726e-06,
+      "logits/chosen": -3.072540521621704,
+      "logits/rejected": -2.8204236030578613,
+      "logps/chosen": -133.08212280273438,
+      "logps/rejected": -119.2376480102539,
+      "loss": 0.0009,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.00826738215982914,
+      "rewards/margins": 0.011625150218605995,
+      "rewards/rejected": -0.019892532378435135,
+      "step": 450
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.272542485937369e-06,
+      "logits/chosen": -2.9859511852264404,
+      "logits/rejected": -2.7228169441223145,
+      "logps/chosen": -211.18002319335938,
+      "logps/rejected": -202.37867736816406,
+      "loss": 0.0009,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.00889207050204277,
+      "rewards/margins": 0.014142923057079315,
+      "rewards/rejected": -0.023034993559122086,
+      "step": 460
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.189093389542498e-06,
+      "logits/chosen": -2.9243226051330566,
+      "logits/rejected": -2.708249092102051,
+      "logps/chosen": -178.67819213867188,
+      "logps/rejected": -162.6269073486328,
+      "loss": 0.0013,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.008272857405245304,
+      "rewards/margins": 0.005242692772299051,
+      "rewards/rejected": 0.0030301641672849655,
+      "step": 470
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.1048047389991693e-06,
+      "logits/chosen": -3.1769165992736816,
+      "logits/rejected": -2.587887763977051,
+      "logps/chosen": -163.45852661132812,
+      "logps/rejected": -134.19691467285156,
+      "loss": 0.0012,
+      "rewards/accuracies": 0.550000011920929,
+      "rewards/chosen": -0.0025352207012474537,
+      "rewards/margins": 0.014497722499072552,
+      "rewards/rejected": -0.017032943665981293,
+      "step": 480
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.019779227044398e-06,
+      "logits/chosen": -3.0145325660705566,
+      "logits/rejected": -2.7123026847839355,
+      "logps/chosen": -167.49745178222656,
+      "logps/rejected": -152.8248748779297,
+      "loss": 0.0012,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.01072392426431179,
+      "rewards/margins": 0.01687874272465706,
+      "rewards/rejected": -0.006154821254312992,
+      "step": 490
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.9341204441673267e-06,
+      "logits/chosen": -3.132664918899536,
+      "logits/rejected": -2.911961317062378,
+      "logps/chosen": -108.4159927368164,
+      "logps/rejected": -94.57585144042969,
+      "loss": 0.0008,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.0020372807048261166,
+      "rewards/margins": 0.008723934181034565,
+      "rewards/rejected": -0.006686653010547161,
+      "step": 500
+    },
+    {
+      "epoch": 1.0,
+      "eval_logits/chosen": -2.395993947982788,
+      "eval_logits/rejected": -2.1981606483459473,
+      "eval_logps/chosen": -270.28497314453125,
+      "eval_logps/rejected": -250.72601318359375,
+      "eval_loss": 0.002102552680298686,
+      "eval_rewards/accuracies": 0.5065000057220459,
+      "eval_rewards/chosen": -0.016971532255411148,
+      "eval_rewards/margins": 0.0018933486426249146,
+      "eval_rewards/rejected": -0.018864883109927177,
+      "eval_runtime": 1423.5769,
+      "eval_samples_per_second": 1.405,
+      "eval_steps_per_second": 0.702,
+      "step": 500
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.847932752400164e-06,
+      "logits/chosen": -3.008751392364502,
+      "logits/rejected": -2.6613450050354004,
+      "logps/chosen": -123.0440444946289,
+      "logps/rejected": -122.49776458740234,
+      "loss": 0.024,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": 0.06933443248271942,
+      "rewards/margins": 0.12685301899909973,
+      "rewards/rejected": -0.05751859024167061,
+      "step": 510
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.761321158169134e-06,
+      "logits/chosen": -2.982330322265625,
+      "logits/rejected": -2.6290860176086426,
+      "logps/chosen": -207.28955078125,
+      "logps/rejected": -207.44711303710938,
+      "loss": 0.0169,
+      "rewards/accuracies": 0.8500000238418579,
+      "rewards/chosen": 0.023494327440857887,
+      "rewards/margins": 0.059116560965776443,
+      "rewards/rejected": -0.035622235387563705,
+      "step": 520
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.6743911843603134e-06,
+      "logits/chosen": -3.140193223953247,
+      "logits/rejected": -2.7758877277374268,
+      "logps/chosen": -161.1981658935547,
+      "logps/rejected": -160.78762817382812,
+      "loss": 0.0854,
+      "rewards/accuracies": 0.8500000238418579,
+      "rewards/chosen": 0.08391492813825607,
+      "rewards/margins": 0.21972373127937317,
+      "rewards/rejected": -0.1358087956905365,
+      "step": 530
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.587248741756253e-06,
+      "logits/chosen": -2.866288661956787,
+      "logits/rejected": -2.557806968688965,
+      "logps/chosen": -229.82077026367188,
+      "logps/rejected": -227.31967163085938,
+      "loss": 0.0241,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": -0.0034355625975877047,
+      "rewards/margins": 0.05683975666761398,
+      "rewards/rejected": -0.06027532368898392,
+      "step": 540
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.5e-06,
+      "logits/chosen": -2.7409634590148926,
+      "logits/rejected": -2.412743330001831,
+      "logps/chosen": -110.87519836425781,
+      "logps/rejected": -106.71611022949219,
+      "loss": 0.0099,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": 0.004647939465939999,
+      "rewards/margins": 0.07160676270723343,
+      "rewards/rejected": -0.06695882230997086,
+      "step": 550
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.4127512582437486e-06,
+      "logits/chosen": -3.080806255340576,
+      "logits/rejected": -2.819638729095459,
+      "logps/chosen": -138.67153930664062,
+      "logps/rejected": -144.43199157714844,
+      "loss": 0.0542,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": 0.025073865428566933,
+      "rewards/margins": 0.14305388927459717,
+      "rewards/rejected": -0.11798002570867538,
+      "step": 560
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.325608815639687e-06,
+      "logits/chosen": -3.0271854400634766,
+      "logits/rejected": -2.7436680793762207,
+      "logps/chosen": -174.3506317138672,
+      "logps/rejected": -171.3226776123047,
+      "loss": 0.0444,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": 0.034181032329797745,
+      "rewards/margins": 0.1548537313938141,
+      "rewards/rejected": -0.12067268788814545,
+      "step": 570
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.238678841830867e-06,
+      "logits/chosen": -2.6729888916015625,
+      "logits/rejected": -2.449924945831299,
+      "logps/chosen": -143.39756774902344,
+      "logps/rejected": -147.48681640625,
+      "loss": 0.0234,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": 0.00795525498688221,
+      "rewards/margins": 0.11490567773580551,
+      "rewards/rejected": -0.10695041716098785,
+      "step": 580
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.1520672475998374e-06,
+      "logits/chosen": -2.940767526626587,
+      "logits/rejected": -2.584683895111084,
+      "logps/chosen": -128.87655639648438,
+      "logps/rejected": -120.95220947265625,
+      "loss": 0.0663,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.025388438254594803,
+      "rewards/margins": 0.09438282996416092,
+      "rewards/rejected": -0.11977125704288483,
+      "step": 590
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.0658795558326745e-06,
+      "logits/chosen": -2.8861563205718994,
+      "logits/rejected": -2.5959839820861816,
+      "logps/chosen": -124.5409927368164,
+      "logps/rejected": -133.5659637451172,
+      "loss": 0.044,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": -0.0006943264743313193,
+      "rewards/margins": 0.09414149820804596,
+      "rewards/rejected": -0.0948358103632927,
+      "step": 600
+    },
+    {
+      "epoch": 1.2,
+      "eval_logits/chosen": -2.389890670776367,
+      "eval_logits/rejected": -2.1939876079559326,
+      "eval_logps/chosen": -269.1193542480469,
+      "eval_logps/rejected": -246.85476684570312,
+      "eval_loss": 0.009110072627663612,
+      "eval_rewards/accuracies": 0.36000001430511475,
+      "eval_rewards/chosen": -0.005315269809216261,
+      "eval_rewards/margins": -0.025163182988762856,
+      "eval_rewards/rejected": 0.019847916439175606,
+      "eval_runtime": 1422.5682,
+      "eval_samples_per_second": 1.406,
+      "eval_steps_per_second": 0.703,
+      "step": 600
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.9802207729556023e-06,
+      "logits/chosen": -2.90655779838562,
+      "logits/rejected": -2.578254222869873,
+      "logps/chosen": -182.38720703125,
+      "logps/rejected": -175.3732147216797,
+      "loss": 0.0322,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": -0.02949387952685356,
+      "rewards/margins": 0.04020578786730766,
+      "rewards/rejected": -0.06969965994358063,
+      "step": 610
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.895195261000831e-06,
+      "logits/chosen": -3.0672218799591064,
+      "logits/rejected": -2.651693105697632,
+      "logps/chosen": -170.82894897460938,
+      "logps/rejected": -157.90194702148438,
+      "loss": 0.0163,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": -0.03491836413741112,
+      "rewards/margins": 0.04328788444399834,
+      "rewards/rejected": -0.07820625603199005,
+      "step": 620
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.8109066104575023e-06,
+      "logits/chosen": -3.076021432876587,
+      "logits/rejected": -2.759376049041748,
+      "logps/chosen": -130.21498107910156,
+      "logps/rejected": -134.99252319335938,
+      "loss": 0.0842,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": 0.011367540806531906,
+      "rewards/margins": 0.15639443695545197,
+      "rewards/rejected": -0.14502687752246857,
+      "step": 630
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.7274575140626318e-06,
+      "logits/chosen": -2.9677953720092773,
+      "logits/rejected": -2.5879857540130615,
+      "logps/chosen": -160.40704345703125,
+      "logps/rejected": -148.29469299316406,
+      "loss": 0.0238,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.04639770835638046,
+      "rewards/margins": 0.028317932039499283,
+      "rewards/rejected": -0.07471564412117004,
+      "step": 640
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.6449496416858285e-06,
+      "logits/chosen": -2.873720169067383,
+      "logits/rejected": -2.657524347305298,
+      "logps/chosen": -211.8774871826172,
+      "logps/rejected": -208.0607452392578,
+      "loss": 0.052,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": -0.05117432028055191,
+      "rewards/margins": 0.10337792336940765,
+      "rewards/rejected": -0.15455225110054016,
+      "step": 650
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.56348351646022e-06,
+      "logits/chosen": -3.1277270317077637,
+      "logits/rejected": -2.781031370162964,
+      "logps/chosen": -159.2963104248047,
+      "logps/rejected": -157.16714477539062,
+      "loss": 0.0255,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.08234690129756927,
+      "rewards/margins": 0.07028084993362427,
+      "rewards/rejected": -0.15262776613235474,
+      "step": 660
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.4831583923105e-06,
+      "logits/chosen": -2.9012064933776855,
+      "logits/rejected": -2.5057435035705566,
+      "logps/chosen": -133.7654571533203,
+      "logps/rejected": -123.15315246582031,
+      "loss": 0.0204,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.06615938246250153,
+      "rewards/margins": -0.006010846234858036,
+      "rewards/rejected": -0.060148537158966064,
+      "step": 670
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.4040721330273063e-06,
+      "logits/chosen": -3.0350639820098877,
+      "logits/rejected": -2.688361167907715,
+      "logps/chosen": -161.60037231445312,
+      "logps/rejected": -148.4673614501953,
+      "loss": 0.018,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": -0.012395807541906834,
+      "rewards/margins": 0.07525575160980225,
+      "rewards/rejected": -0.08765155076980591,
+      "step": 680
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.3263210930352737e-06,
+      "logits/chosen": -2.987074613571167,
+      "logits/rejected": -2.41349458694458,
+      "logps/chosen": -155.8392333984375,
+      "logps/rejected": -120.13862609863281,
+      "loss": 0.1358,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": -0.06855552643537521,
+      "rewards/margins": 0.0165361650288105,
+      "rewards/rejected": -0.08509168773889542,
+      "step": 690
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.2500000000000007e-06,
+      "logits/chosen": -3.142533302307129,
+      "logits/rejected": -2.7605295181274414,
+      "logps/chosen": -196.124267578125,
+      "logps/rejected": -175.75848388671875,
+      "loss": 0.0682,
+      "rewards/accuracies": 0.550000011920929,
+      "rewards/chosen": -0.08456405252218246,
+      "rewards/margins": 0.07064096629619598,
+      "rewards/rejected": -0.15520504117012024,
+      "step": 700
+    },
+    {
+      "epoch": 1.4,
+      "eval_logits/chosen": -2.399232864379883,
+      "eval_logits/rejected": -2.203479528427124,
+      "eval_logps/chosen": -272.0422668457031,
+      "eval_logps/rejected": -247.98184204101562,
+      "eval_loss": 0.019056277349591255,
+      "eval_rewards/accuracies": 0.3449999988079071,
+      "eval_rewards/chosen": -0.03454471752047539,
+      "eval_rewards/margins": -0.043121837079524994,
+      "eval_rewards/rejected": 0.008577119559049606,
+      "eval_runtime": 1422.244,
+      "eval_samples_per_second": 1.406,
+      "eval_steps_per_second": 0.703,
+      "step": 700
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.1752018394169882e-06,
+      "logits/chosen": -2.923161029815674,
+      "logits/rejected": -2.7423832416534424,
+      "logps/chosen": -232.46353149414062,
+      "logps/rejected": -236.95712280273438,
+      "loss": 0.1005,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": -0.058026909828186035,
+      "rewards/margins": 0.11247865855693817,
+      "rewards/rejected": -0.170505553483963,
+      "step": 710
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.1020177413231334e-06,
+      "logits/chosen": -2.9437034130096436,
+      "logits/rejected": -2.7746634483337402,
+      "logps/chosen": -182.21336364746094,
+      "logps/rejected": -170.1109619140625,
+      "loss": 0.0119,
+      "rewards/accuracies": 0.550000011920929,
+      "rewards/chosen": -0.009687254205346107,
+      "rewards/margins": 0.010108504444360733,
+      "rewards/rejected": -0.01979575864970684,
+      "step": 720
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0305368692688175e-06,
+      "logits/chosen": -2.91925048828125,
+      "logits/rejected": -2.4575612545013428,
+      "logps/chosen": -141.11546325683594,
+      "logps/rejected": -132.57675170898438,
+      "loss": 0.0074,
+      "rewards/accuracies": 0.550000011920929,
+      "rewards/chosen": -0.058561988174915314,
+      "rewards/margins": 0.03072230890393257,
+      "rewards/rejected": -0.08928428590297699,
+      "step": 730
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 9.608463116858544e-07,
+      "logits/chosen": -2.8368327617645264,
+      "logits/rejected": -2.523725986480713,
+      "logps/chosen": -102.6483383178711,
+      "logps/rejected": -94.72925567626953,
+      "loss": 0.0083,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.0365123450756073,
+      "rewards/margins": 0.021143099293112755,
+      "rewards/rejected": -0.057655446231365204,
+      "step": 740
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 8.930309757836517e-07,
+      "logits/chosen": -2.9483554363250732,
+      "logits/rejected": -2.4691848754882812,
+      "logps/chosen": -170.67236328125,
+      "logps/rejected": -154.7274932861328,
+      "loss": 0.0595,
+      "rewards/accuracies": 0.550000011920929,
+      "rewards/chosen": -0.09041354060173035,
+      "rewards/margins": 0.019873833283782005,
+      "rewards/rejected": -0.1102873831987381,
+      "step": 750
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 8.271734841028553e-07,
+      "logits/chosen": -2.998615264892578,
+      "logits/rejected": -2.765198230743408,
+      "logps/chosen": -178.57168579101562,
+      "logps/rejected": -165.0284423828125,
+      "loss": 0.0603,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": -0.07312550395727158,
+      "rewards/margins": 0.03924999386072159,
+      "rewards/rejected": -0.11237549781799316,
+      "step": 760
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 7.633540738525066e-07,
+      "logits/chosen": -2.9713294506073,
+      "logits/rejected": -2.581371784210205,
+      "logps/chosen": -188.65562438964844,
+      "logps/rejected": -198.20057678222656,
+      "loss": 0.0376,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.08031792938709259,
+      "rewards/margins": 0.0691676214337349,
+      "rewards/rejected": -0.1494855433702469,
+      "step": 770
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 7.016504991533727e-07,
+      "logits/chosen": -3.094494342803955,
+      "logits/rejected": -2.8429107666015625,
+      "logps/chosen": -172.08921813964844,
+      "logps/rejected": -168.11550903320312,
+      "loss": 0.0146,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.03326564282178879,
+      "rewards/margins": 0.049937326461076736,
+      "rewards/rejected": -0.08320296555757523,
+      "step": 780
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.421379363065142e-07,
+      "logits/chosen": -2.8603293895721436,
+      "logits/rejected": -2.629847288131714,
+      "logps/chosen": -187.458251953125,
+      "logps/rejected": -181.56585693359375,
+      "loss": 0.0346,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.026567768305540085,
+      "rewards/margins": 0.08442084491252899,
+      "rewards/rejected": -0.11098861694335938,
+      "step": 790
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 5.848888922025553e-07,
+      "logits/chosen": -3.0436525344848633,
+      "logits/rejected": -2.855353355407715,
+      "logps/chosen": -116.41280364990234,
+      "logps/rejected": -118.14847564697266,
+      "loss": 0.0505,
+      "rewards/accuracies": 0.8500000238418579,
+      "rewards/chosen": 0.013444220647215843,
+      "rewards/margins": 0.11258585751056671,
+      "rewards/rejected": -0.09914163500070572,
+      "step": 800
+    },
+    {
+      "epoch": 1.6,
+      "eval_logits/chosen": -2.4056026935577393,
+      "eval_logits/rejected": -2.209423780441284,
+      "eval_logps/chosen": -273.5587158203125,
+      "eval_logps/rejected": -248.85421752929688,
+      "eval_loss": 0.023677825927734375,
+      "eval_rewards/accuracies": 0.34049999713897705,
+      "eval_rewards/chosen": -0.049709003418684006,
+      "eval_rewards/margins": -0.04956228658556938,
+      "eval_rewards/rejected": -0.0001467149268137291,
+      "eval_runtime": 1422.3272,
+      "eval_samples_per_second": 1.406,
+      "eval_steps_per_second": 0.703,
+      "step": 800
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 5.299731159831953e-07,
+      "logits/chosen": -2.867088794708252,
+      "logits/rejected": -2.544645309448242,
+      "logps/chosen": -144.49082946777344,
+      "logps/rejected": -137.3140106201172,
+      "loss": 0.0169,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": -0.05922747775912285,
+      "rewards/margins": 0.03063639998435974,
+      "rewards/rejected": -0.08986388146877289,
+      "step": 810
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 4.774575140626317e-07,
+      "logits/chosen": -2.855769634246826,
+      "logits/rejected": -2.6003212928771973,
+      "logps/chosen": -149.01148986816406,
+      "logps/rejected": -150.924072265625,
+      "loss": 0.0045,
+      "rewards/accuracies": 0.44999998807907104,
+      "rewards/chosen": -0.0036942525766789913,
+      "rewards/margins": 0.02296125330030918,
+      "rewards/rejected": -0.02665550634264946,
+      "step": 820
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 4.27406068612396e-07,
+      "logits/chosen": -3.0886871814727783,
+      "logits/rejected": -2.7754549980163574,
+      "logps/chosen": -159.57974243164062,
+      "logps/rejected": -160.63307189941406,
+      "loss": 0.0227,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.05405544117093086,
+      "rewards/margins": 0.0963703840970993,
+      "rewards/rejected": -0.15042582154273987,
+      "step": 830
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.798797596089351e-07,
+      "logits/chosen": -3.102863073348999,
+      "logits/rejected": -2.912782669067383,
+      "logps/chosen": -148.99937438964844,
+      "logps/rejected": -143.40872192382812,
+      "loss": 0.0379,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.06271068751811981,
+      "rewards/margins": 0.025210902094841003,
+      "rewards/rejected": -0.08792158961296082,
+      "step": 840
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.3493649053890325e-07,
+      "logits/chosen": -2.9544389247894287,
+      "logits/rejected": -2.6804440021514893,
+      "logps/chosen": -142.18478393554688,
+      "logps/rejected": -136.8746795654297,
+      "loss": 0.0061,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.008848746307194233,
+      "rewards/margins": 0.052127204835414886,
+      "rewards/rejected": -0.06097595766186714,
+      "step": 850
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9263101785268253e-07,
+      "logits/chosen": -2.970134735107422,
+      "logits/rejected": -2.784672737121582,
+      "logps/chosen": -122.1519775390625,
+      "logps/rejected": -113.82493591308594,
+      "loss": 0.0044,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.058860231190919876,
+      "rewards/margins": 0.03174171224236488,
+      "rewards/rejected": -0.09060193598270416,
+      "step": 860
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.53014884252083e-07,
+      "logits/chosen": -2.949235200881958,
+      "logits/rejected": -2.5656838417053223,
+      "logps/chosen": -258.15875244140625,
+      "logps/rejected": -249.7416534423828,
+      "loss": 0.091,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.05386580899357796,
+      "rewards/margins": 0.09710750728845596,
+      "rewards/rejected": -0.15097330510616302,
+      "step": 870
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.1613635589349756e-07,
+      "logits/chosen": -3.1699745655059814,
+      "logits/rejected": -2.638118028640747,
+      "logps/chosen": -151.49009704589844,
+      "logps/rejected": -145.30239868164062,
+      "loss": 0.0532,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": -0.06000872328877449,
+      "rewards/margins": 0.11070142686367035,
+      "rewards/rejected": -0.17071016132831573,
+      "step": 880
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 1.8204036358303173e-07,
+      "logits/chosen": -2.985215663909912,
+      "logits/rejected": -2.570612668991089,
+      "logps/chosen": -109.09466552734375,
+      "logps/rejected": -98.76487731933594,
+      "loss": 0.0108,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.0210666935890913,
+      "rewards/margins": 0.044059909880161285,
+      "rewards/rejected": -0.06512660533189774,
+      "step": 890
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 1.507684480352292e-07,
+      "logits/chosen": -3.0888664722442627,
+      "logits/rejected": -2.6895575523376465,
+      "logps/chosen": -163.73812866210938,
+      "logps/rejected": -161.14007568359375,
+      "loss": 0.0243,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.08492692559957504,
+      "rewards/margins": 0.025812974199652672,
+      "rewards/rejected": -0.11073990166187286,
+      "step": 900
+    },
+    {
+      "epoch": 1.8,
+      "eval_logits/chosen": -2.4080586433410645,
+      "eval_logits/rejected": -2.2116518020629883,
+      "eval_logps/chosen": -274.3966979980469,
+      "eval_logps/rejected": -249.45701599121094,
+      "eval_loss": 0.025878531858325005,
+      "eval_rewards/accuracies": 0.33399999141693115,
+      "eval_rewards/chosen": -0.058088988065719604,
+      "eval_rewards/margins": -0.05191420391201973,
+      "eval_rewards/rejected": -0.00617477810010314,
+      "eval_runtime": 1421.9219,
+      "eval_samples_per_second": 1.407,
+      "eval_steps_per_second": 0.703,
+      "step": 900
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.223587092621162e-07,
+      "logits/chosen": -3.060760259628296,
+      "logits/rejected": -2.724635124206543,
+      "logps/chosen": -126.0656509399414,
+      "logps/rejected": -112.5301742553711,
+      "loss": 0.0271,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.014755621552467346,
+      "rewards/margins": 0.10490630567073822,
+      "rewards/rejected": -0.11966194212436676,
+      "step": 910
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 9.684576015420277e-08,
+      "logits/chosen": -2.515842914581299,
+      "logits/rejected": -1.9269781112670898,
+      "logps/chosen": -214.9114532470703,
+      "logps/rejected": -191.04974365234375,
+      "loss": 0.0234,
+      "rewards/accuracies": 0.44999998807907104,
+      "rewards/chosen": -0.11193618923425674,
+      "rewards/margins": 0.01120082102715969,
+      "rewards/rejected": -0.12313701212406158,
+      "step": 920
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.426068431000883e-08,
+      "logits/chosen": -3.096656322479248,
+      "logits/rejected": -2.6982362270355225,
+      "logps/chosen": -156.7047576904297,
+      "logps/rejected": -140.3162841796875,
+      "loss": 0.0558,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.1117779016494751,
+      "rewards/margins": 0.010117399506270885,
+      "rewards/rejected": -0.12189529836177826,
+      "step": 930
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.463099816548578e-08,
+      "logits/chosen": -2.985032558441162,
+      "logits/rejected": -2.7575299739837646,
+      "logps/chosen": -217.82192993164062,
+      "logps/rejected": -225.7892303466797,
+      "loss": 0.0701,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.16152246296405792,
+      "rewards/margins": 0.10506312549114227,
+      "rewards/rejected": -0.2665855586528778,
+      "step": 940
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 3.798061746947995e-08,
+      "logits/chosen": -2.925285816192627,
+      "logits/rejected": -2.7017900943756104,
+      "logps/chosen": -161.64578247070312,
+      "logps/rejected": -155.53384399414062,
+      "loss": 0.0125,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": -0.04954836145043373,
+      "rewards/margins": 0.03911132365465164,
+      "rewards/rejected": -0.08865968883037567,
+      "step": 950
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.4329828146074096e-08,
+      "logits/chosen": -3.1424098014831543,
+      "logits/rejected": -2.879060745239258,
+      "logps/chosen": -189.52865600585938,
+      "logps/rejected": -195.4296875,
+      "loss": 0.0599,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.06265055388212204,
+      "rewards/margins": 0.1324789822101593,
+      "rewards/rejected": -0.19512954354286194,
+      "step": 960
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 1.3695261579316776e-08,
+      "logits/chosen": -3.0380070209503174,
+      "logits/rejected": -2.844536304473877,
+      "logps/chosen": -164.58413696289062,
+      "logps/rejected": -158.5533447265625,
+      "loss": 0.027,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": -0.0411117784678936,
+      "rewards/margins": 0.07761015743017197,
+      "rewards/rejected": -0.11872193962335587,
+      "step": 970
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.089874350439507e-09,
+      "logits/chosen": -3.0895726680755615,
+      "logits/rejected": -2.8259506225585938,
+      "logps/chosen": -196.0420684814453,
+      "logps/rejected": -190.8524169921875,
+      "loss": 0.022,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.060807038098573685,
+      "rewards/margins": 0.03220720216631889,
+      "rewards/rejected": -0.09301424026489258,
+      "step": 980
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 1.5229324522605949e-09,
+      "logits/chosen": -3.0721163749694824,
+      "logits/rejected": -2.770005941390991,
+      "logps/chosen": -117.59437561035156,
+      "logps/rejected": -115.3232421875,
+      "loss": 0.0234,
+      "rewards/accuracies": 0.8500000238418579,
+      "rewards/chosen": -0.033795468509197235,
+      "rewards/margins": 0.1025773137807846,
+      "rewards/rejected": -0.13637278974056244,
+      "step": 990
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 0.0,
+      "logits/chosen": -2.9728710651397705,
+      "logits/rejected": -2.4575047492980957,
+      "logps/chosen": -203.40164184570312,
+      "logps/rejected": -187.337890625,
+      "loss": 0.0697,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.11433009803295135,
+      "rewards/margins": 0.06476924568414688,
+      "rewards/rejected": -0.17909933626651764,
+      "step": 1000
+    },
+    {
+      "epoch": 2.0,
+      "eval_logits/chosen": -2.4070470333099365,
+      "eval_logits/rejected": -2.210773468017578,
+      "eval_logps/chosen": -274.3865966796875,
+      "eval_logps/rejected": -249.44677734375,
+      "eval_loss": 0.0258334930986166,
+      "eval_rewards/accuracies": 0.33799999952316284,
+      "eval_rewards/chosen": -0.05798804759979248,
+      "eval_rewards/margins": -0.05191566422581673,
+      "eval_rewards/rejected": -0.006072388496249914,
+      "eval_runtime": 1422.3182,
+      "eval_samples_per_second": 1.406,
+      "eval_steps_per_second": 0.703,
+      "step": 1000
+    },
+    {
+      "epoch": 2.0,
+      "step": 1000,
+      "total_flos": 0.0,
+      "train_loss": 0.019690872263745406,
+      "train_runtime": 18916.4989,
+      "train_samples_per_second": 0.106,
+      "train_steps_per_second": 0.053
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 1000,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 2,
+  "save_steps": 100,
+  "total_flos": 0.0,
+  "train_batch_size": 1,
+  "trial_name": null,
+  "trial_params": null
+}