shenxq commited on
Commit
ac637be
1 Parent(s): 1496b58

Model save

Browse files
README.md ADDED
@@ -0,0 +1,84 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ library_name: peft
4
+ tags:
5
+ - trl
6
+ - dpo
7
+ - generated_from_trainer
8
+ base_model: mistralai/Mistral-7B-v0.1
9
+ model-index:
10
+ - name: zephyr-7b-dpo-qlora
11
+ results: []
12
+ ---
13
+
14
+ <!-- This model card has been generated automatically according to the information the Trainer had access to. You
15
+ should probably proofread and complete it, then remove this comment. -->
16
+
17
+ # zephyr-7b-dpo-qlora
18
+
19
+ 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.
20
+ It achieves the following results on the evaluation set:
21
+ - Loss: 0.6707
22
+ - Rewards/chosen: -0.2860
23
+ - Rewards/rejected: -0.3548
24
+ - Rewards/accuracies: 0.5983
25
+ - Rewards/margins: 0.0687
26
+ - Logps/rejected: -367.6676
27
+ - Logps/chosen: -351.0971
28
+ - Logits/rejected: -2.5801
29
+ - Logits/chosen: -2.5726
30
+
31
+ ## Model description
32
+
33
+ More information needed
34
+
35
+ ## Intended uses & limitations
36
+
37
+ More information needed
38
+
39
+ ## Training and evaluation data
40
+
41
+ More information needed
42
+
43
+ ## Training procedure
44
+
45
+ ### Training hyperparameters
46
+
47
+ The following hyperparameters were used during training:
48
+ - learning_rate: 5e-06
49
+ - train_batch_size: 4
50
+ - eval_batch_size: 8
51
+ - seed: 42
52
+ - distributed_type: multi-GPU
53
+ - gradient_accumulation_steps: 4
54
+ - total_train_batch_size: 16
55
+ - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
56
+ - lr_scheduler_type: cosine
57
+ - lr_scheduler_warmup_ratio: 0.1
58
+ - num_epochs: 1
59
+
60
+ ### Training results
61
+
62
+ | Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
63
+ |:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
64
+ | 0.6932 | 0.08 | 100 | 0.6930 | -0.0030 | -0.0033 | 0.5220 | 0.0003 | -332.5208 | -322.7949 | -2.4978 | -2.4908 |
65
+ | 0.6921 | 0.16 | 200 | 0.6927 | -0.0232 | -0.0243 | 0.5183 | 0.0011 | -334.6197 | -324.8167 | -2.4970 | -2.4900 |
66
+ | 0.6913 | 0.24 | 300 | 0.6919 | -0.0414 | -0.0441 | 0.5340 | 0.0027 | -336.6059 | -326.6393 | -2.4967 | -2.4895 |
67
+ | 0.6893 | 0.32 | 400 | 0.6891 | -0.0791 | -0.0883 | 0.5547 | 0.0093 | -341.0244 | -330.4017 | -2.5023 | -2.4953 |
68
+ | 0.6724 | 0.4 | 500 | 0.6844 | -0.2018 | -0.2253 | 0.5530 | 0.0235 | -354.7256 | -342.6785 | -2.5100 | -2.5029 |
69
+ | 0.6849 | 0.48 | 600 | 0.6805 | -0.3366 | -0.3770 | 0.5597 | 0.0404 | -369.8958 | -356.1591 | -2.5412 | -2.5347 |
70
+ | 0.6503 | 0.56 | 700 | 0.6774 | -0.4376 | -0.4919 | 0.5630 | 0.0543 | -381.3843 | -366.2523 | -2.5492 | -2.5431 |
71
+ | 0.6841 | 0.64 | 800 | 0.6735 | -0.3183 | -0.3788 | 0.5913 | 0.0605 | -370.0676 | -354.3206 | -2.5662 | -2.5592 |
72
+ | 0.6773 | 0.72 | 900 | 0.6724 | -0.3986 | -0.4678 | 0.5887 | 0.0692 | -378.9693 | -362.3546 | -2.5774 | -2.5706 |
73
+ | 0.657 | 0.8 | 1000 | 0.6711 | -0.2774 | -0.3440 | 0.5997 | 0.0666 | -366.5909 | -350.2372 | -2.5784 | -2.5708 |
74
+ | 0.6577 | 0.88 | 1100 | 0.6706 | -0.2934 | -0.3628 | 0.5993 | 0.0693 | -368.4680 | -351.8376 | -2.5805 | -2.5729 |
75
+ | 0.6444 | 0.96 | 1200 | 0.6708 | -0.2860 | -0.3547 | 0.5993 | 0.0687 | -367.6592 | -351.0949 | -2.5801 | -2.5725 |
76
+
77
+
78
+ ### Framework versions
79
+
80
+ - PEFT 0.7.1
81
+ - Transformers 4.36.2
82
+ - Pytorch 2.1.2
83
+ - Datasets 2.14.6
84
+ - Tokenizers 0.15.0
adapter_config.json CHANGED
@@ -19,12 +19,12 @@
19
  "rank_pattern": {},
20
  "revision": null,
21
  "target_modules": [
22
- "v_proj",
23
- "k_proj",
24
- "o_proj",
25
  "gate_proj",
26
- "down_proj",
 
27
  "q_proj",
 
 
28
  "up_proj"
29
  ],
30
  "task_type": "CAUSAL_LM"
 
19
  "rank_pattern": {},
20
  "revision": null,
21
  "target_modules": [
 
 
 
22
  "gate_proj",
23
+ "k_proj",
24
+ "v_proj",
25
  "q_proj",
26
+ "down_proj",
27
+ "o_proj",
28
  "up_proj"
29
  ],
30
  "task_type": "CAUSAL_LM"
adapter_model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:6c2dd2e91042c567a32a1c3c899b935d5f131e1c96db23724c115eaa43d29e62
3
- size 167832240
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:aa34178bd80f8d3a5f7293547ed95d1e9a509a94b2df87e75b782f3843a5f42c
3
+ size 83945744
all_results.json ADDED
@@ -0,0 +1,21 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 1.0,
3
+ "eval_logits/chosen": -2.5725526809692383,
4
+ "eval_logits/rejected": -2.5801451206207275,
5
+ "eval_logps/chosen": -351.0971374511719,
6
+ "eval_logps/rejected": -367.6675720214844,
7
+ "eval_loss": 0.6707386374473572,
8
+ "eval_rewards/accuracies": 0.5983333587646484,
9
+ "eval_rewards/chosen": -0.28601789474487305,
10
+ "eval_rewards/margins": 0.0687330961227417,
11
+ "eval_rewards/rejected": -0.35475102066993713,
12
+ "eval_runtime": 1581.2259,
13
+ "eval_samples": 2994,
14
+ "eval_samples_per_second": 1.893,
15
+ "eval_steps_per_second": 0.237,
16
+ "train_loss": 0.67484232365751,
17
+ "train_runtime": 40604.4701,
18
+ "train_samples": 19996,
19
+ "train_samples_per_second": 0.492,
20
+ "train_steps_per_second": 0.031
21
+ }
eval_results.json ADDED
@@ -0,0 +1,16 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 1.0,
3
+ "eval_logits/chosen": -2.5725526809692383,
4
+ "eval_logits/rejected": -2.5801451206207275,
5
+ "eval_logps/chosen": -351.0971374511719,
6
+ "eval_logps/rejected": -367.6675720214844,
7
+ "eval_loss": 0.6707386374473572,
8
+ "eval_rewards/accuracies": 0.5983333587646484,
9
+ "eval_rewards/chosen": -0.28601789474487305,
10
+ "eval_rewards/margins": 0.0687330961227417,
11
+ "eval_rewards/rejected": -0.35475102066993713,
12
+ "eval_runtime": 1581.2259,
13
+ "eval_samples": 2994,
14
+ "eval_samples_per_second": 1.893,
15
+ "eval_steps_per_second": 0.237
16
+ }
train_results.json ADDED
@@ -0,0 +1,8 @@
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 1.0,
3
+ "train_loss": 0.67484232365751,
4
+ "train_runtime": 40604.4701,
5
+ "train_samples": 19996,
6
+ "train_samples_per_second": 0.492,
7
+ "train_steps_per_second": 0.031
8
+ }
trainer_state.json ADDED
@@ -0,0 +1,1972 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 0.9993998799759952,
5
+ "eval_steps": 100,
6
+ "global_step": 1249,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.0,
13
+ "learning_rate": 4e-08,
14
+ "logits/chosen": -2.5970396995544434,
15
+ "logits/rejected": -2.6002578735351562,
16
+ "logps/chosen": -391.7848205566406,
17
+ "logps/rejected": -379.88592529296875,
18
+ "loss": 0.6931,
19
+ "rewards/accuracies": 0.0,
20
+ "rewards/chosen": 0.0,
21
+ "rewards/margins": 0.0,
22
+ "rewards/rejected": 0.0,
23
+ "step": 1
24
+ },
25
+ {
26
+ "epoch": 0.01,
27
+ "learning_rate": 4.0000000000000003e-07,
28
+ "logits/chosen": -2.6321280002593994,
29
+ "logits/rejected": -2.654905319213867,
30
+ "logps/chosen": -415.5525207519531,
31
+ "logps/rejected": -451.3063049316406,
32
+ "loss": 0.6934,
33
+ "rewards/accuracies": 0.4166666567325592,
34
+ "rewards/chosen": 3.593924702727236e-05,
35
+ "rewards/margins": -0.0004077497578691691,
36
+ "rewards/rejected": 0.00044368900125846267,
37
+ "step": 10
38
+ },
39
+ {
40
+ "epoch": 0.02,
41
+ "learning_rate": 8.000000000000001e-07,
42
+ "logits/chosen": -2.5803229808807373,
43
+ "logits/rejected": -2.6041855812072754,
44
+ "logps/chosen": -378.57830810546875,
45
+ "logps/rejected": -406.0473937988281,
46
+ "loss": 0.6938,
47
+ "rewards/accuracies": 0.41874998807907104,
48
+ "rewards/chosen": -0.0005379965296015143,
49
+ "rewards/margins": -0.0012370327021926641,
50
+ "rewards/rejected": 0.0006990364054217935,
51
+ "step": 20
52
+ },
53
+ {
54
+ "epoch": 0.02,
55
+ "learning_rate": 1.2000000000000002e-06,
56
+ "logits/chosen": -2.5524168014526367,
57
+ "logits/rejected": -2.610682249069214,
58
+ "logps/chosen": -405.12225341796875,
59
+ "logps/rejected": -437.0355529785156,
60
+ "loss": 0.6931,
61
+ "rewards/accuracies": 0.5062500238418579,
62
+ "rewards/chosen": -2.3546338070445927e-06,
63
+ "rewards/margins": 4.741999146062881e-05,
64
+ "rewards/rejected": -4.97746623295825e-05,
65
+ "step": 30
66
+ },
67
+ {
68
+ "epoch": 0.03,
69
+ "learning_rate": 1.6000000000000001e-06,
70
+ "logits/chosen": -2.551060199737549,
71
+ "logits/rejected": -2.5709078311920166,
72
+ "logps/chosen": -374.8092346191406,
73
+ "logps/rejected": -387.48931884765625,
74
+ "loss": 0.6934,
75
+ "rewards/accuracies": 0.46875,
76
+ "rewards/chosen": -0.0006756076472811401,
77
+ "rewards/margins": -0.00045229861279949546,
78
+ "rewards/rejected": -0.00022330903448164463,
79
+ "step": 40
80
+ },
81
+ {
82
+ "epoch": 0.04,
83
+ "learning_rate": 2.0000000000000003e-06,
84
+ "logits/chosen": -2.5436787605285645,
85
+ "logits/rejected": -2.540389060974121,
86
+ "logps/chosen": -366.8270568847656,
87
+ "logps/rejected": -374.41497802734375,
88
+ "loss": 0.6934,
89
+ "rewards/accuracies": 0.4937500059604645,
90
+ "rewards/chosen": -0.0006743489648215473,
91
+ "rewards/margins": -0.00049586221575737,
92
+ "rewards/rejected": -0.0001784868072718382,
93
+ "step": 50
94
+ },
95
+ {
96
+ "epoch": 0.05,
97
+ "learning_rate": 2.4000000000000003e-06,
98
+ "logits/chosen": -2.5643861293792725,
99
+ "logits/rejected": -2.5933163166046143,
100
+ "logps/chosen": -389.86444091796875,
101
+ "logps/rejected": -415.7872619628906,
102
+ "loss": 0.6932,
103
+ "rewards/accuracies": 0.5249999761581421,
104
+ "rewards/chosen": -0.0006498451111838222,
105
+ "rewards/margins": -0.00017824865062721074,
106
+ "rewards/rejected": -0.0004715964605566114,
107
+ "step": 60
108
+ },
109
+ {
110
+ "epoch": 0.06,
111
+ "learning_rate": 2.8000000000000003e-06,
112
+ "logits/chosen": -2.608799457550049,
113
+ "logits/rejected": -2.6101136207580566,
114
+ "logps/chosen": -390.1094970703125,
115
+ "logps/rejected": -396.507568359375,
116
+ "loss": 0.693,
117
+ "rewards/accuracies": 0.5062500238418579,
118
+ "rewards/chosen": -0.0011631803354248405,
119
+ "rewards/margins": 0.00039177463622763753,
120
+ "rewards/rejected": -0.0015549547970294952,
121
+ "step": 70
122
+ },
123
+ {
124
+ "epoch": 0.06,
125
+ "learning_rate": 3.2000000000000003e-06,
126
+ "logits/chosen": -2.6286911964416504,
127
+ "logits/rejected": -2.6467819213867188,
128
+ "logps/chosen": -383.10284423828125,
129
+ "logps/rejected": -406.47393798828125,
130
+ "loss": 0.6933,
131
+ "rewards/accuracies": 0.4937500059604645,
132
+ "rewards/chosen": -0.0017519788816571236,
133
+ "rewards/margins": -0.0002883758279494941,
134
+ "rewards/rejected": -0.0014636029954999685,
135
+ "step": 80
136
+ },
137
+ {
138
+ "epoch": 0.07,
139
+ "learning_rate": 3.6000000000000003e-06,
140
+ "logits/chosen": -2.6428580284118652,
141
+ "logits/rejected": -2.6388907432556152,
142
+ "logps/chosen": -431.7904357910156,
143
+ "logps/rejected": -446.31732177734375,
144
+ "loss": 0.6927,
145
+ "rewards/accuracies": 0.5625,
146
+ "rewards/chosen": -0.0028546121902763844,
147
+ "rewards/margins": 0.0008213006658479571,
148
+ "rewards/rejected": -0.0036759127397090197,
149
+ "step": 90
150
+ },
151
+ {
152
+ "epoch": 0.08,
153
+ "learning_rate": 4.000000000000001e-06,
154
+ "logits/chosen": -2.6136410236358643,
155
+ "logits/rejected": -2.6054491996765137,
156
+ "logps/chosen": -382.89886474609375,
157
+ "logps/rejected": -399.26068115234375,
158
+ "loss": 0.6932,
159
+ "rewards/accuracies": 0.5062500238418579,
160
+ "rewards/chosen": -0.003379733068868518,
161
+ "rewards/margins": -0.00018416713282931596,
162
+ "rewards/rejected": -0.0031955656595528126,
163
+ "step": 100
164
+ },
165
+ {
166
+ "epoch": 0.08,
167
+ "eval_logits/chosen": -2.490821361541748,
168
+ "eval_logits/rejected": -2.497763156890869,
169
+ "eval_logps/chosen": -322.7948913574219,
170
+ "eval_logps/rejected": -332.520751953125,
171
+ "eval_loss": 0.6930087208747864,
172
+ "eval_rewards/accuracies": 0.5220000147819519,
173
+ "eval_rewards/chosen": -0.002995692892000079,
174
+ "eval_rewards/margins": 0.0002869492454919964,
175
+ "eval_rewards/rejected": -0.0032826424576342106,
176
+ "eval_runtime": 1592.8954,
177
+ "eval_samples_per_second": 1.88,
178
+ "eval_steps_per_second": 0.235,
179
+ "step": 100
180
+ },
181
+ {
182
+ "epoch": 0.09,
183
+ "learning_rate": 4.4e-06,
184
+ "logits/chosen": -2.6191463470458984,
185
+ "logits/rejected": -2.626410722732544,
186
+ "logps/chosen": -418.57061767578125,
187
+ "logps/rejected": -431.57568359375,
188
+ "loss": 0.6929,
189
+ "rewards/accuracies": 0.5249999761581421,
190
+ "rewards/chosen": -0.004728841129690409,
191
+ "rewards/margins": 0.00041932202293537557,
192
+ "rewards/rejected": -0.005148162599653006,
193
+ "step": 110
194
+ },
195
+ {
196
+ "epoch": 0.1,
197
+ "learning_rate": 4.800000000000001e-06,
198
+ "logits/chosen": -2.616450309753418,
199
+ "logits/rejected": -2.6316654682159424,
200
+ "logps/chosen": -387.65936279296875,
201
+ "logps/rejected": -406.0279235839844,
202
+ "loss": 0.6924,
203
+ "rewards/accuracies": 0.53125,
204
+ "rewards/chosen": -0.006956800818443298,
205
+ "rewards/margins": 0.00143801374360919,
206
+ "rewards/rejected": -0.008394813165068626,
207
+ "step": 120
208
+ },
209
+ {
210
+ "epoch": 0.1,
211
+ "learning_rate": 4.999755876225375e-06,
212
+ "logits/chosen": -2.5832996368408203,
213
+ "logits/rejected": -2.5880000591278076,
214
+ "logps/chosen": -411.5313415527344,
215
+ "logps/rejected": -434.4432678222656,
216
+ "loss": 0.6925,
217
+ "rewards/accuracies": 0.59375,
218
+ "rewards/chosen": -0.009584048762917519,
219
+ "rewards/margins": 0.0013888883404433727,
220
+ "rewards/rejected": -0.010972936637699604,
221
+ "step": 130
222
+ },
223
+ {
224
+ "epoch": 0.11,
225
+ "learning_rate": 4.997803172081864e-06,
226
+ "logits/chosen": -2.5645198822021484,
227
+ "logits/rejected": -2.5772173404693604,
228
+ "logps/chosen": -400.5196533203125,
229
+ "logps/rejected": -419.6172790527344,
230
+ "loss": 0.6928,
231
+ "rewards/accuracies": 0.5062500238418579,
232
+ "rewards/chosen": -0.013501152396202087,
233
+ "rewards/margins": 0.0006207471014931798,
234
+ "rewards/rejected": -0.01412190031260252,
235
+ "step": 140
236
+ },
237
+ {
238
+ "epoch": 0.12,
239
+ "learning_rate": 4.9938992891651825e-06,
240
+ "logits/chosen": -2.581132411956787,
241
+ "logits/rejected": -2.5872106552124023,
242
+ "logps/chosen": -382.4533386230469,
243
+ "logps/rejected": -414.8580627441406,
244
+ "loss": 0.6922,
245
+ "rewards/accuracies": 0.5562499761581421,
246
+ "rewards/chosen": -0.016398200765252113,
247
+ "rewards/margins": 0.0018568082014098763,
248
+ "rewards/rejected": -0.01825500652194023,
249
+ "step": 150
250
+ },
251
+ {
252
+ "epoch": 0.13,
253
+ "learning_rate": 4.988047277024456e-06,
254
+ "logits/chosen": -2.6221516132354736,
255
+ "logits/rejected": -2.6419806480407715,
256
+ "logps/chosen": -396.94757080078125,
257
+ "logps/rejected": -416.6105041503906,
258
+ "loss": 0.6922,
259
+ "rewards/accuracies": 0.5687500238418579,
260
+ "rewards/chosen": -0.019227243959903717,
261
+ "rewards/margins": 0.0020019258372485638,
262
+ "rewards/rejected": -0.021229172125458717,
263
+ "step": 160
264
+ },
265
+ {
266
+ "epoch": 0.14,
267
+ "learning_rate": 4.980251707005417e-06,
268
+ "logits/chosen": -2.618068218231201,
269
+ "logits/rejected": -2.6144068241119385,
270
+ "logps/chosen": -425.07244873046875,
271
+ "logps/rejected": -435.1886291503906,
272
+ "loss": 0.6935,
273
+ "rewards/accuracies": 0.4625000059604645,
274
+ "rewards/chosen": -0.02663121186196804,
275
+ "rewards/margins": -0.0007048315601423383,
276
+ "rewards/rejected": -0.0259263813495636,
277
+ "step": 170
278
+ },
279
+ {
280
+ "epoch": 0.14,
281
+ "learning_rate": 4.970518668679459e-06,
282
+ "logits/chosen": -2.5808920860290527,
283
+ "logits/rejected": -2.5735676288604736,
284
+ "logps/chosen": -417.1087341308594,
285
+ "logps/rejected": -421.8807678222656,
286
+ "loss": 0.6932,
287
+ "rewards/accuracies": 0.543749988079071,
288
+ "rewards/chosen": -0.026894057169556618,
289
+ "rewards/margins": -6.247097917366773e-05,
290
+ "rewards/rejected": -0.02683158591389656,
291
+ "step": 180
292
+ },
293
+ {
294
+ "epoch": 0.15,
295
+ "learning_rate": 4.958855765086722e-06,
296
+ "logits/chosen": -2.606273651123047,
297
+ "logits/rejected": -2.608921527862549,
298
+ "logps/chosen": -391.02362060546875,
299
+ "logps/rejected": -399.1781311035156,
300
+ "loss": 0.6927,
301
+ "rewards/accuracies": 0.5562499761581421,
302
+ "rewards/chosen": -0.02517121471464634,
303
+ "rewards/margins": 0.0010087827686220407,
304
+ "rewards/rejected": -0.026179995387792587,
305
+ "step": 190
306
+ },
307
+ {
308
+ "epoch": 0.16,
309
+ "learning_rate": 4.945272106796919e-06,
310
+ "logits/chosen": -2.578331708908081,
311
+ "logits/rejected": -2.594592332839966,
312
+ "logps/chosen": -388.1728515625,
313
+ "logps/rejected": -412.2106018066406,
314
+ "loss": 0.6921,
315
+ "rewards/accuracies": 0.6312500238418579,
316
+ "rewards/chosen": -0.022485502064228058,
317
+ "rewards/margins": 0.0022102731745690107,
318
+ "rewards/rejected": -0.024695774540305138,
319
+ "step": 200
320
+ },
321
+ {
322
+ "epoch": 0.16,
323
+ "eval_logits/chosen": -2.4899628162384033,
324
+ "eval_logits/rejected": -2.497047185897827,
325
+ "eval_logps/chosen": -324.8166809082031,
326
+ "eval_logps/rejected": -334.6196594238281,
327
+ "eval_loss": 0.6926525831222534,
328
+ "eval_rewards/accuracies": 0.5183333158493042,
329
+ "eval_rewards/chosen": -0.023213516920804977,
330
+ "eval_rewards/margins": 0.0010577173670753837,
331
+ "eval_rewards/rejected": -0.02427123300731182,
332
+ "eval_runtime": 1586.0043,
333
+ "eval_samples_per_second": 1.888,
334
+ "eval_steps_per_second": 0.236,
335
+ "step": 200
336
+ },
337
+ {
338
+ "epoch": 0.17,
339
+ "learning_rate": 4.929778304792537e-06,
340
+ "logits/chosen": -2.5928235054016113,
341
+ "logits/rejected": -2.601378917694092,
342
+ "logps/chosen": -424.74066162109375,
343
+ "logps/rejected": -424.7421875,
344
+ "loss": 0.6914,
345
+ "rewards/accuracies": 0.581250011920929,
346
+ "rewards/chosen": -0.029381360858678818,
347
+ "rewards/margins": 0.003531560767441988,
348
+ "rewards/rejected": -0.03291292116045952,
349
+ "step": 210
350
+ },
351
+ {
352
+ "epoch": 0.18,
353
+ "learning_rate": 4.912386462179987e-06,
354
+ "logits/chosen": -2.630770206451416,
355
+ "logits/rejected": -2.637812376022339,
356
+ "logps/chosen": -404.63275146484375,
357
+ "logps/rejected": -440.338134765625,
358
+ "loss": 0.6902,
359
+ "rewards/accuracies": 0.5687500238418579,
360
+ "rewards/chosen": -0.035039953887462616,
361
+ "rewards/margins": 0.005983405746519566,
362
+ "rewards/rejected": -0.041023362427949905,
363
+ "step": 220
364
+ },
365
+ {
366
+ "epoch": 0.18,
367
+ "learning_rate": 4.893110164735167e-06,
368
+ "logits/chosen": -2.6500868797302246,
369
+ "logits/rejected": -2.6668245792388916,
370
+ "logps/chosen": -419.52508544921875,
371
+ "logps/rejected": -435.5750427246094,
372
+ "loss": 0.6928,
373
+ "rewards/accuracies": 0.518750011920929,
374
+ "rewards/chosen": -0.04090961441397667,
375
+ "rewards/margins": 0.000937508069910109,
376
+ "rewards/rejected": -0.04184712469577789,
377
+ "step": 230
378
+ },
379
+ {
380
+ "epoch": 0.19,
381
+ "learning_rate": 4.871964470290823e-06,
382
+ "logits/chosen": -2.5765552520751953,
383
+ "logits/rejected": -2.603816509246826,
384
+ "logps/chosen": -419.22332763671875,
385
+ "logps/rejected": -441.0245666503906,
386
+ "loss": 0.692,
387
+ "rewards/accuracies": 0.53125,
388
+ "rewards/chosen": -0.03845367208123207,
389
+ "rewards/margins": 0.0023850365541875362,
390
+ "rewards/rejected": -0.040838707238435745,
391
+ "step": 240
392
+ },
393
+ {
394
+ "epoch": 0.2,
395
+ "learning_rate": 4.848965896974006e-06,
396
+ "logits/chosen": -2.62872052192688,
397
+ "logits/rejected": -2.652649402618408,
398
+ "logps/chosen": -402.7193603515625,
399
+ "logps/rejected": -426.7840270996094,
400
+ "loss": 0.6924,
401
+ "rewards/accuracies": 0.543749988079071,
402
+ "rewards/chosen": -0.038372453302145004,
403
+ "rewards/margins": 0.0015891485381871462,
404
+ "rewards/rejected": -0.03996160253882408,
405
+ "step": 250
406
+ },
407
+ {
408
+ "epoch": 0.21,
409
+ "learning_rate": 4.8241324103028055e-06,
410
+ "logits/chosen": -2.6180949211120605,
411
+ "logits/rejected": -2.5828635692596436,
412
+ "logps/chosen": -408.49200439453125,
413
+ "logps/rejected": -419.191650390625,
414
+ "loss": 0.6918,
415
+ "rewards/accuracies": 0.581250011920929,
416
+ "rewards/chosen": -0.037816666066646576,
417
+ "rewards/margins": 0.002941995393484831,
418
+ "rewards/rejected": -0.040758661925792694,
419
+ "step": 260
420
+ },
421
+ {
422
+ "epoch": 0.22,
423
+ "learning_rate": 4.797483409152438e-06,
424
+ "logits/chosen": -2.617037296295166,
425
+ "logits/rejected": -2.636782169342041,
426
+ "logps/chosen": -391.20391845703125,
427
+ "logps/rejected": -412.9513244628906,
428
+ "loss": 0.691,
429
+ "rewards/accuracies": 0.543749988079071,
430
+ "rewards/chosen": -0.0356525257229805,
431
+ "rewards/margins": 0.004444011487066746,
432
+ "rewards/rejected": -0.04009654000401497,
433
+ "step": 270
434
+ },
435
+ {
436
+ "epoch": 0.22,
437
+ "learning_rate": 4.769039710601669e-06,
438
+ "logits/chosen": -2.6231722831726074,
439
+ "logits/rejected": -2.6144392490386963,
440
+ "logps/chosen": -387.4334411621094,
441
+ "logps/rejected": -403.40057373046875,
442
+ "loss": 0.6898,
443
+ "rewards/accuracies": 0.581250011920929,
444
+ "rewards/chosen": -0.03822268173098564,
445
+ "rewards/margins": 0.00696157943457365,
446
+ "rewards/rejected": -0.045184262096881866,
447
+ "step": 280
448
+ },
449
+ {
450
+ "epoch": 0.23,
451
+ "learning_rate": 4.738823533671383e-06,
452
+ "logits/chosen": -2.614532709121704,
453
+ "logits/rejected": -2.6345345973968506,
454
+ "logps/chosen": -421.5420837402344,
455
+ "logps/rejected": -442.87994384765625,
456
+ "loss": 0.6916,
457
+ "rewards/accuracies": 0.6000000238418579,
458
+ "rewards/chosen": -0.04416601359844208,
459
+ "rewards/margins": 0.003424243303015828,
460
+ "rewards/rejected": -0.047590263187885284,
461
+ "step": 290
462
+ },
463
+ {
464
+ "epoch": 0.24,
465
+ "learning_rate": 4.706858481968017e-06,
466
+ "logits/chosen": -2.5896575450897217,
467
+ "logits/rejected": -2.5868849754333496,
468
+ "logps/chosen": -402.4905700683594,
469
+ "logps/rejected": -409.7896423339844,
470
+ "loss": 0.6913,
471
+ "rewards/accuracies": 0.53125,
472
+ "rewards/chosen": -0.04842864349484444,
473
+ "rewards/margins": 0.003919907845556736,
474
+ "rewards/rejected": -0.052348554134368896,
475
+ "step": 300
476
+ },
477
+ {
478
+ "epoch": 0.24,
479
+ "eval_logits/chosen": -2.4894840717315674,
480
+ "eval_logits/rejected": -2.4966704845428467,
481
+ "eval_logps/chosen": -326.6393127441406,
482
+ "eval_logps/rejected": -336.60589599609375,
483
+ "eval_loss": 0.6919258236885071,
484
+ "eval_rewards/accuracies": 0.5339999794960022,
485
+ "eval_rewards/chosen": -0.0414397232234478,
486
+ "eval_rewards/margins": 0.0026943066623061895,
487
+ "eval_rewards/rejected": -0.04413403198122978,
488
+ "eval_runtime": 1585.7818,
489
+ "eval_samples_per_second": 1.888,
490
+ "eval_steps_per_second": 0.236,
491
+ "step": 300
492
+ },
493
+ {
494
+ "epoch": 0.25,
495
+ "learning_rate": 4.673169525245416e-06,
496
+ "logits/chosen": -2.5968692302703857,
497
+ "logits/rejected": -2.585181474685669,
498
+ "logps/chosen": -404.8775939941406,
499
+ "logps/rejected": -433.09954833984375,
500
+ "loss": 0.6902,
501
+ "rewards/accuracies": 0.6000000238418579,
502
+ "rewards/chosen": -0.04641267657279968,
503
+ "rewards/margins": 0.00627540796995163,
504
+ "rewards/rejected": -0.052688080817461014,
505
+ "step": 310
506
+ },
507
+ {
508
+ "epoch": 0.26,
509
+ "learning_rate": 4.63778297989952e-06,
510
+ "logits/chosen": -2.604461193084717,
511
+ "logits/rejected": -2.610708475112915,
512
+ "logps/chosen": -403.58184814453125,
513
+ "logps/rejected": -424.60772705078125,
514
+ "loss": 0.691,
515
+ "rewards/accuracies": 0.574999988079071,
516
+ "rewards/chosen": -0.05133752152323723,
517
+ "rewards/margins": 0.004545481409877539,
518
+ "rewards/rejected": -0.0558830089867115,
519
+ "step": 320
520
+ },
521
+ {
522
+ "epoch": 0.26,
523
+ "learning_rate": 4.60072648841109e-06,
524
+ "logits/chosen": -2.589761734008789,
525
+ "logits/rejected": -2.5855839252471924,
526
+ "logps/chosen": -411.21026611328125,
527
+ "logps/rejected": -430.5553283691406,
528
+ "loss": 0.6872,
529
+ "rewards/accuracies": 0.574999988079071,
530
+ "rewards/chosen": -0.0565360002219677,
531
+ "rewards/margins": 0.012383319437503815,
532
+ "rewards/rejected": -0.06891932338476181,
533
+ "step": 330
534
+ },
535
+ {
536
+ "epoch": 0.27,
537
+ "learning_rate": 4.562028997752574e-06,
538
+ "logits/chosen": -2.5468997955322266,
539
+ "logits/rejected": -2.5858261585235596,
540
+ "logps/chosen": -409.3692321777344,
541
+ "logps/rejected": -441.8675842285156,
542
+ "loss": 0.6876,
543
+ "rewards/accuracies": 0.5687500238418579,
544
+ "rewards/chosen": -0.04863084852695465,
545
+ "rewards/margins": 0.01160926092416048,
546
+ "rewards/rejected": -0.06024010106921196,
547
+ "step": 340
548
+ },
549
+ {
550
+ "epoch": 0.28,
551
+ "learning_rate": 4.521720736775947e-06,
552
+ "logits/chosen": -2.604952335357666,
553
+ "logits/rejected": -2.596958637237549,
554
+ "logps/chosen": -426.7057189941406,
555
+ "logps/rejected": -427.89971923828125,
556
+ "loss": 0.6907,
557
+ "rewards/accuracies": 0.5625,
558
+ "rewards/chosen": -0.06284764409065247,
559
+ "rewards/margins": 0.005581230856478214,
560
+ "rewards/rejected": -0.0684288740158081,
561
+ "step": 350
562
+ },
563
+ {
564
+ "epoch": 0.29,
565
+ "learning_rate": 4.479833192599198e-06,
566
+ "logits/chosen": -2.6308584213256836,
567
+ "logits/rejected": -2.627711057662964,
568
+ "logps/chosen": -435.3694763183594,
569
+ "logps/rejected": -454.00775146484375,
570
+ "loss": 0.6867,
571
+ "rewards/accuracies": 0.612500011920929,
572
+ "rewards/chosen": -0.06149282306432724,
573
+ "rewards/margins": 0.013573038391768932,
574
+ "rewards/rejected": -0.07506586611270905,
575
+ "step": 360
576
+ },
577
+ {
578
+ "epoch": 0.3,
579
+ "learning_rate": 4.436399086009928e-06,
580
+ "logits/chosen": -2.6563782691955566,
581
+ "logits/rejected": -2.6729133129119873,
582
+ "logps/chosen": -409.5260925292969,
583
+ "logps/rejected": -424.39208984375,
584
+ "loss": 0.6874,
585
+ "rewards/accuracies": 0.574999988079071,
586
+ "rewards/chosen": -0.07003921270370483,
587
+ "rewards/margins": 0.012296813540160656,
588
+ "rewards/rejected": -0.08233602344989777,
589
+ "step": 370
590
+ },
591
+ {
592
+ "epoch": 0.3,
593
+ "learning_rate": 4.391452345905239e-06,
594
+ "logits/chosen": -2.5911147594451904,
595
+ "logits/rejected": -2.6013450622558594,
596
+ "logps/chosen": -420.21124267578125,
597
+ "logps/rejected": -429.5535583496094,
598
+ "loss": 0.6907,
599
+ "rewards/accuracies": 0.550000011920929,
600
+ "rewards/chosen": -0.10565753281116486,
601
+ "rewards/margins": 0.006660384126007557,
602
+ "rewards/rejected": -0.11231790482997894,
603
+ "step": 380
604
+ },
605
+ {
606
+ "epoch": 0.31,
607
+ "learning_rate": 4.3450280827879125e-06,
608
+ "logits/chosen": -2.6037206649780273,
609
+ "logits/rejected": -2.6106367111206055,
610
+ "logps/chosen": -413.44122314453125,
611
+ "logps/rejected": -427.99951171875,
612
+ "loss": 0.6882,
613
+ "rewards/accuracies": 0.574999988079071,
614
+ "rewards/chosen": -0.10301616042852402,
615
+ "rewards/margins": 0.011568797752261162,
616
+ "rewards/rejected": -0.11458496004343033,
617
+ "step": 390
618
+ },
619
+ {
620
+ "epoch": 0.32,
621
+ "learning_rate": 4.297162561339554e-06,
622
+ "logits/chosen": -2.522566556930542,
623
+ "logits/rejected": -2.563248634338379,
624
+ "logps/chosen": -421.3636169433594,
625
+ "logps/rejected": -440.02276611328125,
626
+ "loss": 0.6893,
627
+ "rewards/accuracies": 0.518750011920929,
628
+ "rewards/chosen": -0.09894236922264099,
629
+ "rewards/margins": 0.009308888576924801,
630
+ "rewards/rejected": -0.10825125873088837,
631
+ "step": 400
632
+ },
633
+ {
634
+ "epoch": 0.32,
635
+ "eval_logits/chosen": -2.4952824115753174,
636
+ "eval_logits/rejected": -2.5022895336151123,
637
+ "eval_logps/chosen": -330.40167236328125,
638
+ "eval_logps/rejected": -341.0243835449219,
639
+ "eval_loss": 0.6891305446624756,
640
+ "eval_rewards/accuracies": 0.5546666383743286,
641
+ "eval_rewards/chosen": -0.0790637731552124,
642
+ "eval_rewards/margins": 0.009254684671759605,
643
+ "eval_rewards/rejected": -0.08831847459077835,
644
+ "eval_runtime": 1591.081,
645
+ "eval_samples_per_second": 1.882,
646
+ "eval_steps_per_second": 0.236,
647
+ "step": 400
648
+ },
649
+ {
650
+ "epoch": 0.33,
651
+ "learning_rate": 4.247893172092157e-06,
652
+ "logits/chosen": -2.5598597526550293,
653
+ "logits/rejected": -2.5833513736724854,
654
+ "logps/chosen": -404.2002868652344,
655
+ "logps/rejected": -435.16357421875,
656
+ "loss": 0.6857,
657
+ "rewards/accuracies": 0.6000000238418579,
658
+ "rewards/chosen": -0.10219142585992813,
659
+ "rewards/margins": 0.016902286559343338,
660
+ "rewards/rejected": -0.11909371614456177,
661
+ "step": 410
662
+ },
663
+ {
664
+ "epoch": 0.34,
665
+ "learning_rate": 4.197258402220187e-06,
666
+ "logits/chosen": -2.5842721462249756,
667
+ "logits/rejected": -2.5888819694519043,
668
+ "logps/chosen": -414.70166015625,
669
+ "logps/rejected": -425.90667724609375,
670
+ "loss": 0.6896,
671
+ "rewards/accuracies": 0.53125,
672
+ "rewards/chosen": -0.12604860961437225,
673
+ "rewards/margins": 0.009105135686695576,
674
+ "rewards/rejected": -0.13515374064445496,
675
+ "step": 420
676
+ },
677
+ {
678
+ "epoch": 0.34,
679
+ "learning_rate": 4.145297805476023e-06,
680
+ "logits/chosen": -2.639338493347168,
681
+ "logits/rejected": -2.6249961853027344,
682
+ "logps/chosen": -412.44232177734375,
683
+ "logps/rejected": -427.6927185058594,
684
+ "loss": 0.6859,
685
+ "rewards/accuracies": 0.606249988079071,
686
+ "rewards/chosen": -0.11504076421260834,
687
+ "rewards/margins": 0.016550499945878983,
688
+ "rewards/rejected": -0.13159126043319702,
689
+ "step": 430
690
+ },
691
+ {
692
+ "epoch": 0.35,
693
+ "learning_rate": 4.092051971292228e-06,
694
+ "logits/chosen": -2.598680019378662,
695
+ "logits/rejected": -2.617053508758545,
696
+ "logps/chosen": -404.762939453125,
697
+ "logps/rejected": -424.2640686035156,
698
+ "loss": 0.6839,
699
+ "rewards/accuracies": 0.5625,
700
+ "rewards/chosen": -0.1352802962064743,
701
+ "rewards/margins": 0.02169608511030674,
702
+ "rewards/rejected": -0.1569763869047165,
703
+ "step": 440
704
+ },
705
+ {
706
+ "epoch": 0.36,
707
+ "learning_rate": 4.037562493074792e-06,
708
+ "logits/chosen": -2.5920491218566895,
709
+ "logits/rejected": -2.59037446975708,
710
+ "logps/chosen": -434.03118896484375,
711
+ "logps/rejected": -439.4048767089844,
712
+ "loss": 0.6832,
713
+ "rewards/accuracies": 0.5687500238418579,
714
+ "rewards/chosen": -0.1662554293870926,
715
+ "rewards/margins": 0.024613002315163612,
716
+ "rewards/rejected": -0.19086843729019165,
717
+ "step": 450
718
+ },
719
+ {
720
+ "epoch": 0.37,
721
+ "learning_rate": 3.981871935712112e-06,
722
+ "logits/chosen": -2.6342484951019287,
723
+ "logits/rejected": -2.6563720703125,
724
+ "logps/chosen": -411.9359436035156,
725
+ "logps/rejected": -435.57586669921875,
726
+ "loss": 0.6833,
727
+ "rewards/accuracies": 0.5562499761581421,
728
+ "rewards/chosen": -0.16083207726478577,
729
+ "rewards/margins": 0.023298023268580437,
730
+ "rewards/rejected": -0.18413010239601135,
731
+ "step": 460
732
+ },
733
+ {
734
+ "epoch": 0.38,
735
+ "learning_rate": 3.925023802325094e-06,
736
+ "logits/chosen": -2.664499044418335,
737
+ "logits/rejected": -2.6798224449157715,
738
+ "logps/chosen": -434.82354736328125,
739
+ "logps/rejected": -453.67938232421875,
740
+ "loss": 0.6776,
741
+ "rewards/accuracies": 0.6187499761581421,
742
+ "rewards/chosen": -0.18204860389232635,
743
+ "rewards/margins": 0.03722093254327774,
744
+ "rewards/rejected": -0.2192695438861847,
745
+ "step": 470
746
+ },
747
+ {
748
+ "epoch": 0.38,
749
+ "learning_rate": 3.867062500284342e-06,
750
+ "logits/chosen": -2.6116766929626465,
751
+ "logits/rejected": -2.627462863922119,
752
+ "logps/chosen": -422.6748962402344,
753
+ "logps/rejected": -454.072509765625,
754
+ "loss": 0.6764,
755
+ "rewards/accuracies": 0.6187499761581421,
756
+ "rewards/chosen": -0.23436252772808075,
757
+ "rewards/margins": 0.041565194725990295,
758
+ "rewards/rejected": -0.27592775225639343,
759
+ "step": 480
760
+ },
761
+ {
762
+ "epoch": 0.39,
763
+ "learning_rate": 3.8080333065209885e-06,
764
+ "logits/chosen": -2.6146063804626465,
765
+ "logits/rejected": -2.6115143299102783,
766
+ "logps/chosen": -428.1953125,
767
+ "logps/rejected": -423.606201171875,
768
+ "loss": 0.7007,
769
+ "rewards/accuracies": 0.4749999940395355,
770
+ "rewards/chosen": -0.31299567222595215,
771
+ "rewards/margins": -0.0009805113077163696,
772
+ "rewards/rejected": -0.3120151460170746,
773
+ "step": 490
774
+ },
775
+ {
776
+ "epoch": 0.4,
777
+ "learning_rate": 3.7479823321582624e-06,
778
+ "logits/chosen": -2.6379127502441406,
779
+ "logits/rejected": -2.66550874710083,
780
+ "logps/chosen": -430.17572021484375,
781
+ "logps/rejected": -461.09942626953125,
782
+ "loss": 0.6724,
783
+ "rewards/accuracies": 0.612500011920929,
784
+ "rewards/chosen": -0.24496836960315704,
785
+ "rewards/margins": 0.04783773794770241,
786
+ "rewards/rejected": -0.29280608892440796,
787
+ "step": 500
788
+ },
789
+ {
790
+ "epoch": 0.4,
791
+ "eval_logits/chosen": -2.5028529167175293,
792
+ "eval_logits/rejected": -2.5099613666534424,
793
+ "eval_logps/chosen": -342.6784973144531,
794
+ "eval_logps/rejected": -354.7255859375,
795
+ "eval_loss": 0.684412956237793,
796
+ "eval_rewards/accuracies": 0.5529999732971191,
797
+ "eval_rewards/chosen": -0.20183168351650238,
798
+ "eval_rewards/margins": 0.023498954251408577,
799
+ "eval_rewards/rejected": -0.2253306657075882,
800
+ "eval_runtime": 1585.7151,
801
+ "eval_samples_per_second": 1.888,
802
+ "eval_steps_per_second": 0.236,
803
+ "step": 500
804
+ },
805
+ {
806
+ "epoch": 0.41,
807
+ "learning_rate": 3.686956486491419e-06,
808
+ "logits/chosen": -2.59777569770813,
809
+ "logits/rejected": -2.600053071975708,
810
+ "logps/chosen": -440.66217041015625,
811
+ "logps/rejected": -454.6631774902344,
812
+ "loss": 0.6921,
813
+ "rewards/accuracies": 0.550000011920929,
814
+ "rewards/chosen": -0.2590482234954834,
815
+ "rewards/margins": 0.0103686498478055,
816
+ "rewards/rejected": -0.269416868686676,
817
+ "step": 510
818
+ },
819
+ {
820
+ "epoch": 0.42,
821
+ "learning_rate": 3.625003440344166e-06,
822
+ "logits/chosen": -2.6117706298828125,
823
+ "logits/rejected": -2.625525712966919,
824
+ "logps/chosen": -416.2491760253906,
825
+ "logps/rejected": -427.0350646972656,
826
+ "loss": 0.6781,
827
+ "rewards/accuracies": 0.5687500238418579,
828
+ "rewards/chosen": -0.228049635887146,
829
+ "rewards/margins": 0.037227652966976166,
830
+ "rewards/rejected": -0.26527729630470276,
831
+ "step": 520
832
+ },
833
+ {
834
+ "epoch": 0.42,
835
+ "learning_rate": 3.562171588830231e-06,
836
+ "logits/chosen": -2.670912265777588,
837
+ "logits/rejected": -2.6928374767303467,
838
+ "logps/chosen": -428.692626953125,
839
+ "logps/rejected": -462.62274169921875,
840
+ "loss": 0.677,
841
+ "rewards/accuracies": 0.6499999761581421,
842
+ "rewards/chosen": -0.22094754874706268,
843
+ "rewards/margins": 0.041822899132966995,
844
+ "rewards/rejected": -0.26277047395706177,
845
+ "step": 530
846
+ },
847
+ {
848
+ "epoch": 0.43,
849
+ "learning_rate": 3.4985100135491245e-06,
850
+ "logits/chosen": -2.6906635761260986,
851
+ "logits/rejected": -2.708909511566162,
852
+ "logps/chosen": -444.14483642578125,
853
+ "logps/rejected": -480.6028747558594,
854
+ "loss": 0.6691,
855
+ "rewards/accuracies": 0.5562499761581421,
856
+ "rewards/chosen": -0.2567470967769623,
857
+ "rewards/margins": 0.058775972574949265,
858
+ "rewards/rejected": -0.31552308797836304,
859
+ "step": 540
860
+ },
861
+ {
862
+ "epoch": 0.44,
863
+ "learning_rate": 3.4340684442456673e-06,
864
+ "logits/chosen": -2.5944106578826904,
865
+ "logits/rejected": -2.6143410205841064,
866
+ "logps/chosen": -438.25738525390625,
867
+ "logps/rejected": -459.01763916015625,
868
+ "loss": 0.6759,
869
+ "rewards/accuracies": 0.6000000238418579,
870
+ "rewards/chosen": -0.3285837173461914,
871
+ "rewards/margins": 0.04724721983075142,
872
+ "rewards/rejected": -0.37583091855049133,
873
+ "step": 550
874
+ },
875
+ {
876
+ "epoch": 0.45,
877
+ "learning_rate": 3.3688972199631974e-06,
878
+ "logits/chosen": -2.6361145973205566,
879
+ "logits/rejected": -2.613219976425171,
880
+ "logps/chosen": -441.1446838378906,
881
+ "logps/rejected": -451.8172912597656,
882
+ "loss": 0.6769,
883
+ "rewards/accuracies": 0.606249988079071,
884
+ "rewards/chosen": -0.3550431728363037,
885
+ "rewards/margins": 0.05103546380996704,
886
+ "rewards/rejected": -0.40607863664627075,
887
+ "step": 560
888
+ },
889
+ {
890
+ "epoch": 0.46,
891
+ "learning_rate": 3.3030472497208354e-06,
892
+ "logits/chosen": -2.6290247440338135,
893
+ "logits/rejected": -2.652219295501709,
894
+ "logps/chosen": -426.34234619140625,
895
+ "logps/rejected": -484.0997009277344,
896
+ "loss": 0.6598,
897
+ "rewards/accuracies": 0.6312500238418579,
898
+ "rewards/chosen": -0.3327481150627136,
899
+ "rewards/margins": 0.08233290165662766,
900
+ "rewards/rejected": -0.4150810241699219,
901
+ "step": 570
902
+ },
903
+ {
904
+ "epoch": 0.46,
905
+ "learning_rate": 3.236569972745492e-06,
906
+ "logits/chosen": -2.640798807144165,
907
+ "logits/rejected": -2.6551499366760254,
908
+ "logps/chosen": -419.3091735839844,
909
+ "logps/rejected": -434.71527099609375,
910
+ "loss": 0.6828,
911
+ "rewards/accuracies": 0.5625,
912
+ "rewards/chosen": -0.31533199548721313,
913
+ "rewards/margins": 0.03718234971165657,
914
+ "rewards/rejected": -0.3525143265724182,
915
+ "step": 580
916
+ },
917
+ {
918
+ "epoch": 0.47,
919
+ "learning_rate": 3.1695173182897126e-06,
920
+ "logits/chosen": -2.5991015434265137,
921
+ "logits/rejected": -2.6085472106933594,
922
+ "logps/chosen": -435.3345642089844,
923
+ "logps/rejected": -467.74078369140625,
924
+ "loss": 0.6698,
925
+ "rewards/accuracies": 0.606249988079071,
926
+ "rewards/chosen": -0.3784990906715393,
927
+ "rewards/margins": 0.060858141630887985,
928
+ "rewards/rejected": -0.4393572211265564,
929
+ "step": 590
930
+ },
931
+ {
932
+ "epoch": 0.48,
933
+ "learning_rate": 3.10194166506673e-06,
934
+ "logits/chosen": -2.624173164367676,
935
+ "logits/rejected": -2.649747848510742,
936
+ "logps/chosen": -443.6612243652344,
937
+ "logps/rejected": -470.52447509765625,
938
+ "loss": 0.6849,
939
+ "rewards/accuracies": 0.581250011920929,
940
+ "rewards/chosen": -0.3967738747596741,
941
+ "rewards/margins": 0.03684181347489357,
942
+ "rewards/rejected": -0.43361568450927734,
943
+ "step": 600
944
+ },
945
+ {
946
+ "epoch": 0.48,
947
+ "eval_logits/chosen": -2.534677505493164,
948
+ "eval_logits/rejected": -2.5411627292633057,
949
+ "eval_logps/chosen": -356.15911865234375,
950
+ "eval_logps/rejected": -369.895751953125,
951
+ "eval_loss": 0.6804664731025696,
952
+ "eval_rewards/accuracies": 0.5596666932106018,
953
+ "eval_rewards/chosen": -0.33663830161094666,
954
+ "eval_rewards/margins": 0.04039384424686432,
955
+ "eval_rewards/rejected": -0.3770321309566498,
956
+ "eval_runtime": 1591.4153,
957
+ "eval_samples_per_second": 1.881,
958
+ "eval_steps_per_second": 0.236,
959
+ "step": 600
960
+ },
961
+ {
962
+ "epoch": 0.49,
963
+ "learning_rate": 3.0338958003344115e-06,
964
+ "logits/chosen": -2.628021717071533,
965
+ "logits/rejected": -2.654170036315918,
966
+ "logps/chosen": -423.56768798828125,
967
+ "logps/rejected": -451.36041259765625,
968
+ "loss": 0.6603,
969
+ "rewards/accuracies": 0.6000000238418579,
970
+ "rewards/chosen": -0.39979976415634155,
971
+ "rewards/margins": 0.08663686364889145,
972
+ "rewards/rejected": -0.486436665058136,
973
+ "step": 610
974
+ },
975
+ {
976
+ "epoch": 0.5,
977
+ "learning_rate": 2.9654328786600823e-06,
978
+ "logits/chosen": -2.6157093048095703,
979
+ "logits/rejected": -2.6425511837005615,
980
+ "logps/chosen": -437.9432678222656,
981
+ "logps/rejected": -482.203125,
982
+ "loss": 0.6558,
983
+ "rewards/accuracies": 0.5625,
984
+ "rewards/chosen": -0.4792413115501404,
985
+ "rewards/margins": 0.10463404655456543,
986
+ "rewards/rejected": -0.5838753581047058,
987
+ "step": 620
988
+ },
989
+ {
990
+ "epoch": 0.5,
991
+ "learning_rate": 2.896606380398402e-06,
992
+ "logits/chosen": -2.6413679122924805,
993
+ "logits/rejected": -2.6552627086639404,
994
+ "logps/chosen": -455.11260986328125,
995
+ "logps/rejected": -468.00885009765625,
996
+ "loss": 0.6878,
997
+ "rewards/accuracies": 0.59375,
998
+ "rewards/chosen": -0.5618987083435059,
999
+ "rewards/margins": 0.03895387798547745,
1000
+ "rewards/rejected": -0.6008526086807251,
1001
+ "step": 630
1002
+ },
1003
+ {
1004
+ "epoch": 0.51,
1005
+ "learning_rate": 2.827470069914772e-06,
1006
+ "logits/chosen": -2.681443214416504,
1007
+ "logits/rejected": -2.6886324882507324,
1008
+ "logps/chosen": -459.51953125,
1009
+ "logps/rejected": -491.306396484375,
1010
+ "loss": 0.6753,
1011
+ "rewards/accuracies": 0.59375,
1012
+ "rewards/chosen": -0.5284784436225891,
1013
+ "rewards/margins": 0.06967002153396606,
1014
+ "rewards/rejected": -0.5981484651565552,
1015
+ "step": 640
1016
+ },
1017
+ {
1018
+ "epoch": 0.52,
1019
+ "learning_rate": 2.7580779535868675e-06,
1020
+ "logits/chosen": -2.630887508392334,
1021
+ "logits/rejected": -2.645677328109741,
1022
+ "logps/chosen": -457.000732421875,
1023
+ "logps/rejected": -472.52545166015625,
1024
+ "loss": 0.6748,
1025
+ "rewards/accuracies": 0.550000011920929,
1026
+ "rewards/chosen": -0.47293543815612793,
1027
+ "rewards/margins": 0.06035799905657768,
1028
+ "rewards/rejected": -0.5332934260368347,
1029
+ "step": 650
1030
+ },
1031
+ {
1032
+ "epoch": 0.53,
1033
+ "learning_rate": 2.688484237617129e-06,
1034
+ "logits/chosen": -2.6493170261383057,
1035
+ "logits/rejected": -2.68170166015625,
1036
+ "logps/chosen": -453.5931091308594,
1037
+ "logps/rejected": -479.94818115234375,
1038
+ "loss": 0.6716,
1039
+ "rewards/accuracies": 0.6187499761581421,
1040
+ "rewards/chosen": -0.4414174556732178,
1041
+ "rewards/margins": 0.06612064689397812,
1042
+ "rewards/rejected": -0.5075381398200989,
1043
+ "step": 660
1044
+ },
1045
+ {
1046
+ "epoch": 0.54,
1047
+ "learning_rate": 2.6187432856891585e-06,
1048
+ "logits/chosen": -2.6594104766845703,
1049
+ "logits/rejected": -2.6713552474975586,
1050
+ "logps/chosen": -456.5660095214844,
1051
+ "logps/rejected": -488.9071350097656,
1052
+ "loss": 0.6665,
1053
+ "rewards/accuracies": 0.6187499761581421,
1054
+ "rewards/chosen": -0.44480568170547485,
1055
+ "rewards/margins": 0.07257355749607086,
1056
+ "rewards/rejected": -0.5173791646957397,
1057
+ "step": 670
1058
+ },
1059
+ {
1060
+ "epoch": 0.54,
1061
+ "learning_rate": 2.548909576501096e-06,
1062
+ "logits/chosen": -2.7302539348602295,
1063
+ "logits/rejected": -2.7218852043151855,
1064
+ "logps/chosen": -458.1416931152344,
1065
+ "logps/rejected": -482.7391662597656,
1066
+ "loss": 0.6629,
1067
+ "rewards/accuracies": 0.612500011920929,
1068
+ "rewards/chosen": -0.4737418591976166,
1069
+ "rewards/margins": 0.08715833723545074,
1070
+ "rewards/rejected": -0.5609001517295837,
1071
+ "step": 680
1072
+ },
1073
+ {
1074
+ "epoch": 0.55,
1075
+ "learning_rate": 2.4790376612091503e-06,
1076
+ "logits/chosen": -2.7039058208465576,
1077
+ "logits/rejected": -2.7273173332214355,
1078
+ "logps/chosen": -473.97064208984375,
1079
+ "logps/rejected": -498.3961486816406,
1080
+ "loss": 0.6715,
1081
+ "rewards/accuracies": 0.550000011920929,
1082
+ "rewards/chosen": -0.5430811643600464,
1083
+ "rewards/margins": 0.08151289075613022,
1084
+ "rewards/rejected": -0.6245940327644348,
1085
+ "step": 690
1086
+ },
1087
+ {
1088
+ "epoch": 0.56,
1089
+ "learning_rate": 2.40918212081453e-06,
1090
+ "logits/chosen": -2.6450188159942627,
1091
+ "logits/rejected": -2.667860507965088,
1092
+ "logps/chosen": -424.064208984375,
1093
+ "logps/rejected": -463.056640625,
1094
+ "loss": 0.6503,
1095
+ "rewards/accuracies": 0.574999988079071,
1096
+ "rewards/chosen": -0.4541637897491455,
1097
+ "rewards/margins": 0.10934968292713165,
1098
+ "rewards/rejected": -0.5635135173797607,
1099
+ "step": 700
1100
+ },
1101
+ {
1102
+ "epoch": 0.56,
1103
+ "eval_logits/chosen": -2.543064832687378,
1104
+ "eval_logits/rejected": -2.5492172241210938,
1105
+ "eval_logps/chosen": -366.25225830078125,
1106
+ "eval_logps/rejected": -381.38427734375,
1107
+ "eval_loss": 0.677395224571228,
1108
+ "eval_rewards/accuracies": 0.5630000233650208,
1109
+ "eval_rewards/chosen": -0.4375694692134857,
1110
+ "eval_rewards/margins": 0.054348159581422806,
1111
+ "eval_rewards/rejected": -0.4919176399707794,
1112
+ "eval_runtime": 1584.8471,
1113
+ "eval_samples_per_second": 1.889,
1114
+ "eval_steps_per_second": 0.237,
1115
+ "step": 700
1116
+ },
1117
+ {
1118
+ "epoch": 0.57,
1119
+ "learning_rate": 2.3393975235270654e-06,
1120
+ "logits/chosen": -2.6429007053375244,
1121
+ "logits/rejected": -2.635617256164551,
1122
+ "logps/chosen": -460.6632385253906,
1123
+ "logps/rejected": -495.57098388671875,
1124
+ "loss": 0.6491,
1125
+ "rewards/accuracies": 0.6000000238418579,
1126
+ "rewards/chosen": -0.5247341990470886,
1127
+ "rewards/margins": 0.12096454948186874,
1128
+ "rewards/rejected": -0.6456986665725708,
1129
+ "step": 710
1130
+ },
1131
+ {
1132
+ "epoch": 0.58,
1133
+ "learning_rate": 2.2697383821388153e-06,
1134
+ "logits/chosen": -2.665684700012207,
1135
+ "logits/rejected": -2.63784122467041,
1136
+ "logps/chosen": -454.69110107421875,
1137
+ "logps/rejected": -466.15020751953125,
1138
+ "loss": 0.6856,
1139
+ "rewards/accuracies": 0.5562499761581421,
1140
+ "rewards/chosen": -0.49990567564964294,
1141
+ "rewards/margins": 0.04791254550218582,
1142
+ "rewards/rejected": -0.547818124294281,
1143
+ "step": 720
1144
+ },
1145
+ {
1146
+ "epoch": 0.58,
1147
+ "learning_rate": 2.2002591114409657e-06,
1148
+ "logits/chosen": -2.672154426574707,
1149
+ "logits/rejected": -2.6631722450256348,
1150
+ "logps/chosen": -453.4434509277344,
1151
+ "logps/rejected": -476.5977478027344,
1152
+ "loss": 0.6716,
1153
+ "rewards/accuracies": 0.543749988079071,
1154
+ "rewards/chosen": -0.4569157660007477,
1155
+ "rewards/margins": 0.07301469147205353,
1156
+ "rewards/rejected": -0.5299304723739624,
1157
+ "step": 730
1158
+ },
1159
+ {
1160
+ "epoch": 0.59,
1161
+ "learning_rate": 2.131013985717285e-06,
1162
+ "logits/chosen": -2.6686697006225586,
1163
+ "logits/rejected": -2.662213087081909,
1164
+ "logps/chosen": -456.8734436035156,
1165
+ "logps/rejected": -498.776123046875,
1166
+ "loss": 0.6499,
1167
+ "rewards/accuracies": 0.5874999761581421,
1168
+ "rewards/chosen": -0.38933560252189636,
1169
+ "rewards/margins": 0.1104939803481102,
1170
+ "rewards/rejected": -0.49982959032058716,
1171
+ "step": 740
1172
+ },
1173
+ {
1174
+ "epoch": 0.6,
1175
+ "learning_rate": 2.062057096347338e-06,
1176
+ "logits/chosen": -2.6757967472076416,
1177
+ "logits/rejected": -2.683659553527832,
1178
+ "logps/chosen": -449.1112365722656,
1179
+ "logps/rejected": -460.0345764160156,
1180
+ "loss": 0.6687,
1181
+ "rewards/accuracies": 0.5562499761581421,
1182
+ "rewards/chosen": -0.40121787786483765,
1183
+ "rewards/margins": 0.07739736139774323,
1184
+ "rewards/rejected": -0.47861528396606445,
1185
+ "step": 750
1186
+ },
1187
+ {
1188
+ "epoch": 0.61,
1189
+ "learning_rate": 1.9934423095525733e-06,
1190
+ "logits/chosen": -2.660757541656494,
1191
+ "logits/rejected": -2.641571521759033,
1192
+ "logps/chosen": -453.0924377441406,
1193
+ "logps/rejected": -462.0196228027344,
1194
+ "loss": 0.6656,
1195
+ "rewards/accuracies": 0.612500011920929,
1196
+ "rewards/chosen": -0.3819692134857178,
1197
+ "rewards/margins": 0.08219519257545471,
1198
+ "rewards/rejected": -0.4641643464565277,
1199
+ "step": 760
1200
+ },
1201
+ {
1202
+ "epoch": 0.62,
1203
+ "learning_rate": 1.9252232243182986e-06,
1204
+ "logits/chosen": -2.6494193077087402,
1205
+ "logits/rejected": -2.6644535064697266,
1206
+ "logps/chosen": -420.4537658691406,
1207
+ "logps/rejected": -452.49700927734375,
1208
+ "loss": 0.6659,
1209
+ "rewards/accuracies": 0.612500011920929,
1210
+ "rewards/chosen": -0.4124923348426819,
1211
+ "rewards/margins": 0.08118386566638947,
1212
+ "rewards/rejected": -0.49367618560791016,
1213
+ "step": 770
1214
+ },
1215
+ {
1216
+ "epoch": 0.62,
1217
+ "learning_rate": 1.8574531305244043e-06,
1218
+ "logits/chosen": -2.674877882003784,
1219
+ "logits/rejected": -2.698631763458252,
1220
+ "logps/chosen": -459.5841369628906,
1221
+ "logps/rejected": -493.99969482421875,
1222
+ "loss": 0.6453,
1223
+ "rewards/accuracies": 0.606249988079071,
1224
+ "rewards/chosen": -0.4147106111049652,
1225
+ "rewards/margins": 0.12960824370384216,
1226
+ "rewards/rejected": -0.5443187952041626,
1227
+ "step": 780
1228
+ },
1229
+ {
1230
+ "epoch": 0.63,
1231
+ "learning_rate": 1.7901849673175559e-06,
1232
+ "logits/chosen": -2.69256329536438,
1233
+ "logits/rejected": -2.7218129634857178,
1234
+ "logps/chosen": -455.54736328125,
1235
+ "logps/rejected": -486.2936096191406,
1236
+ "loss": 0.66,
1237
+ "rewards/accuracies": 0.65625,
1238
+ "rewards/chosen": -0.38235577940940857,
1239
+ "rewards/margins": 0.09029395878314972,
1240
+ "rewards/rejected": -0.4726497530937195,
1241
+ "step": 790
1242
+ },
1243
+ {
1244
+ "epoch": 0.64,
1245
+ "learning_rate": 1.7234712817573555e-06,
1246
+ "logits/chosen": -2.670511484146118,
1247
+ "logits/rejected": -2.6233620643615723,
1248
+ "logps/chosen": -479.3063049316406,
1249
+ "logps/rejected": -484.30718994140625,
1250
+ "loss": 0.6841,
1251
+ "rewards/accuracies": 0.6000000238418579,
1252
+ "rewards/chosen": -0.45734840631484985,
1253
+ "rewards/margins": 0.05288088321685791,
1254
+ "rewards/rejected": -0.510229229927063,
1255
+ "step": 800
1256
+ },
1257
+ {
1258
+ "epoch": 0.64,
1259
+ "eval_logits/chosen": -2.559168577194214,
1260
+ "eval_logits/rejected": -2.56618070602417,
1261
+ "eval_logps/chosen": -354.3205871582031,
1262
+ "eval_logps/rejected": -370.067626953125,
1263
+ "eval_loss": 0.6735012531280518,
1264
+ "eval_rewards/accuracies": 0.5913333296775818,
1265
+ "eval_rewards/chosen": -0.31825292110443115,
1266
+ "eval_rewards/margins": 0.06049799174070358,
1267
+ "eval_rewards/rejected": -0.37875089049339294,
1268
+ "eval_runtime": 1585.6649,
1269
+ "eval_samples_per_second": 1.888,
1270
+ "eval_steps_per_second": 0.236,
1271
+ "step": 800
1272
+ },
1273
+ {
1274
+ "epoch": 0.65,
1275
+ "learning_rate": 1.6573641877687936e-06,
1276
+ "logits/chosen": -2.688758373260498,
1277
+ "logits/rejected": -2.7048513889312744,
1278
+ "logps/chosen": -433.21484375,
1279
+ "logps/rejected": -472.264892578125,
1280
+ "loss": 0.6559,
1281
+ "rewards/accuracies": 0.625,
1282
+ "rewards/chosen": -0.3647409975528717,
1283
+ "rewards/margins": 0.10189852863550186,
1284
+ "rewards/rejected": -0.46663951873779297,
1285
+ "step": 810
1286
+ },
1287
+ {
1288
+ "epoch": 0.66,
1289
+ "learning_rate": 1.591915325433034e-06,
1290
+ "logits/chosen": -2.705111265182495,
1291
+ "logits/rejected": -2.690777540206909,
1292
+ "logps/chosen": -436.6507263183594,
1293
+ "logps/rejected": -449.94891357421875,
1294
+ "loss": 0.6732,
1295
+ "rewards/accuracies": 0.637499988079071,
1296
+ "rewards/chosen": -0.39074164628982544,
1297
+ "rewards/margins": 0.06702588498592377,
1298
+ "rewards/rejected": -0.4577675461769104,
1299
+ "step": 820
1300
+ },
1301
+ {
1302
+ "epoch": 0.66,
1303
+ "learning_rate": 1.5271758206483664e-06,
1304
+ "logits/chosen": -2.7012417316436768,
1305
+ "logits/rejected": -2.692389726638794,
1306
+ "logps/chosen": -457.00238037109375,
1307
+ "logps/rejected": -475.9202575683594,
1308
+ "loss": 0.6669,
1309
+ "rewards/accuracies": 0.625,
1310
+ "rewards/chosen": -0.4308633804321289,
1311
+ "rewards/margins": 0.09242836385965347,
1312
+ "rewards/rejected": -0.5232917666435242,
1313
+ "step": 830
1314
+ },
1315
+ {
1316
+ "epoch": 0.67,
1317
+ "learning_rate": 1.4631962451927966e-06,
1318
+ "logits/chosen": -2.675565719604492,
1319
+ "logits/rejected": -2.708937168121338,
1320
+ "logps/chosen": -453.7381286621094,
1321
+ "logps/rejected": -485.38360595703125,
1322
+ "loss": 0.6646,
1323
+ "rewards/accuracies": 0.6312500238418579,
1324
+ "rewards/chosen": -0.42438727617263794,
1325
+ "rewards/margins": 0.09076879173517227,
1326
+ "rewards/rejected": -0.515156090259552,
1327
+ "step": 840
1328
+ },
1329
+ {
1330
+ "epoch": 0.68,
1331
+ "learning_rate": 1.4000265772195032e-06,
1332
+ "logits/chosen": -2.706509590148926,
1333
+ "logits/rejected": -2.7089927196502686,
1334
+ "logps/chosen": -460.16326904296875,
1335
+ "logps/rejected": -492.61163330078125,
1336
+ "loss": 0.6616,
1337
+ "rewards/accuracies": 0.606249988079071,
1338
+ "rewards/chosen": -0.4706672132015228,
1339
+ "rewards/margins": 0.09530263394117355,
1340
+ "rewards/rejected": -0.5659698247909546,
1341
+ "step": 850
1342
+ },
1343
+ {
1344
+ "epoch": 0.69,
1345
+ "learning_rate": 1.3377161622160077e-06,
1346
+ "logits/chosen": -2.688765048980713,
1347
+ "logits/rejected": -2.686649799346924,
1348
+ "logps/chosen": -452.43035888671875,
1349
+ "logps/rejected": -478.95855712890625,
1350
+ "loss": 0.6575,
1351
+ "rewards/accuracies": 0.6312500238418579,
1352
+ "rewards/chosen": -0.4564797878265381,
1353
+ "rewards/margins": 0.09994712471961975,
1354
+ "rewards/rejected": -0.5564268827438354,
1355
+ "step": 860
1356
+ },
1357
+ {
1358
+ "epoch": 0.7,
1359
+ "learning_rate": 1.276313674457553e-06,
1360
+ "logits/chosen": -2.687105417251587,
1361
+ "logits/rejected": -2.6988847255706787,
1362
+ "logps/chosen": -431.041259765625,
1363
+ "logps/rejected": -468.3299865722656,
1364
+ "loss": 0.644,
1365
+ "rewards/accuracies": 0.675000011920929,
1366
+ "rewards/chosen": -0.4324752390384674,
1367
+ "rewards/margins": 0.13809339702129364,
1368
+ "rewards/rejected": -0.5705686211585999,
1369
+ "step": 870
1370
+ },
1371
+ {
1372
+ "epoch": 0.7,
1373
+ "learning_rate": 1.2158670789848095e-06,
1374
+ "logits/chosen": -2.6659626960754395,
1375
+ "logits/rejected": -2.6399731636047363,
1376
+ "logps/chosen": -455.6881408691406,
1377
+ "logps/rejected": -471.60479736328125,
1378
+ "loss": 0.6906,
1379
+ "rewards/accuracies": 0.574999988079071,
1380
+ "rewards/chosen": -0.4899720549583435,
1381
+ "rewards/margins": 0.03855130821466446,
1382
+ "rewards/rejected": -0.528523325920105,
1383
+ "step": 880
1384
+ },
1385
+ {
1386
+ "epoch": 0.71,
1387
+ "learning_rate": 1.1564235941356016e-06,
1388
+ "logits/chosen": -2.636784791946411,
1389
+ "logits/rejected": -2.665184497833252,
1390
+ "logps/chosen": -443.228759765625,
1391
+ "logps/rejected": -486.3818359375,
1392
+ "loss": 0.6375,
1393
+ "rewards/accuracies": 0.643750011920929,
1394
+ "rewards/chosen": -0.49858155846595764,
1395
+ "rewards/margins": 0.15248027443885803,
1396
+ "rewards/rejected": -0.6510618329048157,
1397
+ "step": 890
1398
+ },
1399
+ {
1400
+ "epoch": 0.72,
1401
+ "learning_rate": 1.0980296546599254e-06,
1402
+ "logits/chosen": -2.6810688972473145,
1403
+ "logits/rejected": -2.66475248336792,
1404
+ "logps/chosen": -478.31512451171875,
1405
+ "logps/rejected": -483.4214782714844,
1406
+ "loss": 0.6773,
1407
+ "rewards/accuracies": 0.550000011920929,
1408
+ "rewards/chosen": -0.5605891942977905,
1409
+ "rewards/margins": 0.07103095203638077,
1410
+ "rewards/rejected": -0.6316200494766235,
1411
+ "step": 900
1412
+ },
1413
+ {
1414
+ "epoch": 0.72,
1415
+ "eval_logits/chosen": -2.5706326961517334,
1416
+ "eval_logits/rejected": -2.577380418777466,
1417
+ "eval_logps/chosen": -362.3546142578125,
1418
+ "eval_logps/rejected": -378.96929931640625,
1419
+ "eval_loss": 0.6724240183830261,
1420
+ "eval_rewards/accuracies": 0.5886666774749756,
1421
+ "eval_rewards/chosen": -0.3985931873321533,
1422
+ "eval_rewards/margins": 0.06917466968297958,
1423
+ "eval_rewards/rejected": -0.4677678346633911,
1424
+ "eval_runtime": 1591.0139,
1425
+ "eval_samples_per_second": 1.882,
1426
+ "eval_steps_per_second": 0.236,
1427
+ "step": 900
1428
+ },
1429
+ {
1430
+ "epoch": 0.73,
1431
+ "learning_rate": 1.040730875447083e-06,
1432
+ "logits/chosen": -2.6869008541107178,
1433
+ "logits/rejected": -2.6931540966033936,
1434
+ "logps/chosen": -458.8529357910156,
1435
+ "logps/rejected": -475.8218688964844,
1436
+ "loss": 0.6623,
1437
+ "rewards/accuracies": 0.637499988079071,
1438
+ "rewards/chosen": -0.47533708810806274,
1439
+ "rewards/margins": 0.09188753366470337,
1440
+ "rewards/rejected": -0.5672245621681213,
1441
+ "step": 910
1442
+ },
1443
+ {
1444
+ "epoch": 0.74,
1445
+ "learning_rate": 9.845720158932414e-07,
1446
+ "logits/chosen": -2.655848741531372,
1447
+ "logits/rejected": -2.650817632675171,
1448
+ "logps/chosen": -427.18896484375,
1449
+ "logps/rejected": -438.2098693847656,
1450
+ "loss": 0.6659,
1451
+ "rewards/accuracies": 0.5687500238418579,
1452
+ "rewards/chosen": -0.4379213750362396,
1453
+ "rewards/margins": 0.079404316842556,
1454
+ "rewards/rejected": -0.5173256993293762,
1455
+ "step": 920
1456
+ },
1457
+ {
1458
+ "epoch": 0.74,
1459
+ "learning_rate": 9.295969449372796e-07,
1460
+ "logits/chosen": -2.682236433029175,
1461
+ "logits/rejected": -2.706988573074341,
1462
+ "logps/chosen": -434.9564514160156,
1463
+ "logps/rejected": -458.70831298828125,
1464
+ "loss": 0.6677,
1465
+ "rewards/accuracies": 0.5625,
1466
+ "rewards/chosen": -0.38662606477737427,
1467
+ "rewards/margins": 0.08299224078655243,
1468
+ "rewards/rejected": -0.4696182608604431,
1469
+ "step": 930
1470
+ },
1471
+ {
1472
+ "epoch": 0.75,
1473
+ "learning_rate": 8.758486067922176e-07,
1474
+ "logits/chosen": -2.68632435798645,
1475
+ "logits/rejected": -2.700981616973877,
1476
+ "logps/chosen": -425.8504943847656,
1477
+ "logps/rejected": -465.1412048339844,
1478
+ "loss": 0.651,
1479
+ "rewards/accuracies": 0.6000000238418579,
1480
+ "rewards/chosen": -0.372600257396698,
1481
+ "rewards/margins": 0.1192459687590599,
1482
+ "rewards/rejected": -0.4918461740016937,
1483
+ "step": 940
1484
+ },
1485
+ {
1486
+ "epoch": 0.76,
1487
+ "learning_rate": 8.233689873990006e-07,
1488
+ "logits/chosen": -2.661367654800415,
1489
+ "logits/rejected": -2.663963794708252,
1490
+ "logps/chosen": -430.2080993652344,
1491
+ "logps/rejected": -466.76092529296875,
1492
+ "loss": 0.6451,
1493
+ "rewards/accuracies": 0.643750011920929,
1494
+ "rewards/chosen": -0.3554816246032715,
1495
+ "rewards/margins": 0.12667140364646912,
1496
+ "rewards/rejected": -0.4821530282497406,
1497
+ "step": 950
1498
+ },
1499
+ {
1500
+ "epoch": 0.77,
1501
+ "learning_rate": 7.721990816288555e-07,
1502
+ "logits/chosen": -2.7383949756622314,
1503
+ "logits/rejected": -2.7589166164398193,
1504
+ "logps/chosen": -413.57135009765625,
1505
+ "logps/rejected": -441.5545349121094,
1506
+ "loss": 0.6724,
1507
+ "rewards/accuracies": 0.59375,
1508
+ "rewards/chosen": -0.371616005897522,
1509
+ "rewards/margins": 0.06721927225589752,
1510
+ "rewards/rejected": -0.4388352930545807,
1511
+ "step": 960
1512
+ },
1513
+ {
1514
+ "epoch": 0.78,
1515
+ "learning_rate": 7.223788612598148e-07,
1516
+ "logits/chosen": -2.7295823097229004,
1517
+ "logits/rejected": -2.740691661834717,
1518
+ "logps/chosen": -443.89776611328125,
1519
+ "logps/rejected": -474.42254638671875,
1520
+ "loss": 0.6495,
1521
+ "rewards/accuracies": 0.637499988079071,
1522
+ "rewards/chosen": -0.332157701253891,
1523
+ "rewards/margins": 0.1200638860464096,
1524
+ "rewards/rejected": -0.4522215723991394,
1525
+ "step": 970
1526
+ },
1527
+ {
1528
+ "epoch": 0.78,
1529
+ "learning_rate": 6.73947243752448e-07,
1530
+ "logits/chosen": -2.70650053024292,
1531
+ "logits/rejected": -2.6953940391540527,
1532
+ "logps/chosen": -460.09918212890625,
1533
+ "logps/rejected": -455.58489990234375,
1534
+ "loss": 0.6806,
1535
+ "rewards/accuracies": 0.5249999761581421,
1536
+ "rewards/chosen": -0.4144526422023773,
1537
+ "rewards/margins": 0.05592719465494156,
1538
+ "rewards/rejected": -0.4703798294067383,
1539
+ "step": 980
1540
+ },
1541
+ {
1542
+ "epoch": 0.79,
1543
+ "learning_rate": 6.269420618491759e-07,
1544
+ "logits/chosen": -2.7356605529785156,
1545
+ "logits/rejected": -2.737133026123047,
1546
+ "logps/chosen": -423.57537841796875,
1547
+ "logps/rejected": -452.10260009765625,
1548
+ "loss": 0.6665,
1549
+ "rewards/accuracies": 0.6000000238418579,
1550
+ "rewards/chosen": -0.3517349660396576,
1551
+ "rewards/margins": 0.08071886003017426,
1552
+ "rewards/rejected": -0.43245387077331543,
1553
+ "step": 990
1554
+ },
1555
+ {
1556
+ "epoch": 0.8,
1557
+ "learning_rate": 5.814000340209267e-07,
1558
+ "logits/chosen": -2.7007267475128174,
1559
+ "logits/rejected": -2.700061798095703,
1560
+ "logps/chosen": -442.41705322265625,
1561
+ "logps/rejected": -477.9020080566406,
1562
+ "loss": 0.657,
1563
+ "rewards/accuracies": 0.606249988079071,
1564
+ "rewards/chosen": -0.34119993448257446,
1565
+ "rewards/margins": 0.10114644467830658,
1566
+ "rewards/rejected": -0.44234639406204224,
1567
+ "step": 1000
1568
+ },
1569
+ {
1570
+ "epoch": 0.8,
1571
+ "eval_logits/chosen": -2.570801258087158,
1572
+ "eval_logits/rejected": -2.578376293182373,
1573
+ "eval_logps/chosen": -350.23724365234375,
1574
+ "eval_logps/rejected": -366.5909118652344,
1575
+ "eval_loss": 0.6711118221282959,
1576
+ "eval_rewards/accuracies": 0.5996666550636292,
1577
+ "eval_rewards/chosen": -0.277418851852417,
1578
+ "eval_rewards/margins": 0.06656476855278015,
1579
+ "eval_rewards/rejected": -0.34398359060287476,
1580
+ "eval_runtime": 1581.2525,
1581
+ "eval_samples_per_second": 1.893,
1582
+ "eval_steps_per_second": 0.237,
1583
+ "step": 1000
1584
+ },
1585
+ {
1586
+ "epoch": 0.81,
1587
+ "learning_rate": 5.373567357842111e-07,
1588
+ "logits/chosen": -2.6614608764648438,
1589
+ "logits/rejected": -2.694235324859619,
1590
+ "logps/chosen": -432.1556091308594,
1591
+ "logps/rejected": -458.6222229003906,
1592
+ "loss": 0.6625,
1593
+ "rewards/accuracies": 0.6312500238418579,
1594
+ "rewards/chosen": -0.2784090042114258,
1595
+ "rewards/margins": 0.0895313173532486,
1596
+ "rewards/rejected": -0.36794033646583557,
1597
+ "step": 1010
1598
+ },
1599
+ {
1600
+ "epoch": 0.82,
1601
+ "learning_rate": 4.948465719110226e-07,
1602
+ "logits/chosen": -2.6836206912994385,
1603
+ "logits/rejected": -2.697563409805298,
1604
+ "logps/chosen": -415.11041259765625,
1605
+ "logps/rejected": -438.9033203125,
1606
+ "loss": 0.6496,
1607
+ "rewards/accuracies": 0.637499988079071,
1608
+ "rewards/chosen": -0.31501251459121704,
1609
+ "rewards/margins": 0.11396019160747528,
1610
+ "rewards/rejected": -0.4289727210998535,
1611
+ "step": 1020
1612
+ },
1613
+ {
1614
+ "epoch": 0.82,
1615
+ "learning_rate": 4.539027495532766e-07,
1616
+ "logits/chosen": -2.6660382747650146,
1617
+ "logits/rejected": -2.6685073375701904,
1618
+ "logps/chosen": -425.77490234375,
1619
+ "logps/rejected": -446.0552673339844,
1620
+ "loss": 0.6661,
1621
+ "rewards/accuracies": 0.6187499761581421,
1622
+ "rewards/chosen": -0.35333576798439026,
1623
+ "rewards/margins": 0.08899492025375366,
1624
+ "rewards/rejected": -0.4423306882381439,
1625
+ "step": 1030
1626
+ },
1627
+ {
1628
+ "epoch": 0.83,
1629
+ "learning_rate": 4.14557252302783e-07,
1630
+ "logits/chosen": -2.6789193153381348,
1631
+ "logits/rejected": -2.6913952827453613,
1632
+ "logps/chosen": -429.2127990722656,
1633
+ "logps/rejected": -461.7294006347656,
1634
+ "loss": 0.6592,
1635
+ "rewards/accuracies": 0.65625,
1636
+ "rewards/chosen": -0.3878576159477234,
1637
+ "rewards/margins": 0.0975080356001854,
1638
+ "rewards/rejected": -0.48536568880081177,
1639
+ "step": 1040
1640
+ },
1641
+ {
1642
+ "epoch": 0.84,
1643
+ "learning_rate": 3.7684081520700884e-07,
1644
+ "logits/chosen": -2.6964402198791504,
1645
+ "logits/rejected": -2.70263934135437,
1646
+ "logps/chosen": -462.135986328125,
1647
+ "logps/rejected": -464.80035400390625,
1648
+ "loss": 0.6855,
1649
+ "rewards/accuracies": 0.612500011920929,
1650
+ "rewards/chosen": -0.437302827835083,
1651
+ "rewards/margins": 0.041490085422992706,
1652
+ "rewards/rejected": -0.4787929654121399,
1653
+ "step": 1050
1654
+ },
1655
+ {
1656
+ "epoch": 0.85,
1657
+ "learning_rate": 3.407829007601507e-07,
1658
+ "logits/chosen": -2.6650023460388184,
1659
+ "logits/rejected": -2.6826632022857666,
1660
+ "logps/chosen": -443.31982421875,
1661
+ "logps/rejected": -482.0733947753906,
1662
+ "loss": 0.6466,
1663
+ "rewards/accuracies": 0.606249988079071,
1664
+ "rewards/chosen": -0.37219929695129395,
1665
+ "rewards/margins": 0.13468633592128754,
1666
+ "rewards/rejected": -0.5068856477737427,
1667
+ "step": 1060
1668
+ },
1669
+ {
1670
+ "epoch": 0.86,
1671
+ "learning_rate": 3.064116758882724e-07,
1672
+ "logits/chosen": -2.704041004180908,
1673
+ "logits/rejected": -2.712378978729248,
1674
+ "logps/chosen": -449.221435546875,
1675
+ "logps/rejected": -488.82733154296875,
1676
+ "loss": 0.6569,
1677
+ "rewards/accuracies": 0.6312500238418579,
1678
+ "rewards/chosen": -0.38919639587402344,
1679
+ "rewards/margins": 0.10973678529262543,
1680
+ "rewards/rejected": -0.49893322587013245,
1681
+ "step": 1070
1682
+ },
1683
+ {
1684
+ "epoch": 0.86,
1685
+ "learning_rate": 2.737539899464908e-07,
1686
+ "logits/chosen": -2.66728138923645,
1687
+ "logits/rejected": -2.6824071407318115,
1688
+ "logps/chosen": -408.9554138183594,
1689
+ "logps/rejected": -435.3258361816406,
1690
+ "loss": 0.6621,
1691
+ "rewards/accuracies": 0.59375,
1692
+ "rewards/chosen": -0.34276992082595825,
1693
+ "rewards/margins": 0.08587387949228287,
1694
+ "rewards/rejected": -0.4286437928676605,
1695
+ "step": 1080
1696
+ },
1697
+ {
1698
+ "epoch": 0.87,
1699
+ "learning_rate": 2.4283535374538645e-07,
1700
+ "logits/chosen": -2.731504440307617,
1701
+ "logits/rejected": -2.733990430831909,
1702
+ "logps/chosen": -450.702392578125,
1703
+ "logps/rejected": -473.677490234375,
1704
+ "loss": 0.6486,
1705
+ "rewards/accuracies": 0.637499988079071,
1706
+ "rewards/chosen": -0.3203676640987396,
1707
+ "rewards/margins": 0.12553098797798157,
1708
+ "rewards/rejected": -0.4458986818790436,
1709
+ "step": 1090
1710
+ },
1711
+ {
1712
+ "epoch": 0.88,
1713
+ "learning_rate": 2.1367991962303298e-07,
1714
+ "logits/chosen": -2.6853950023651123,
1715
+ "logits/rejected": -2.702267646789551,
1716
+ "logps/chosen": -415.25048828125,
1717
+ "logps/rejected": -443.6421813964844,
1718
+ "loss": 0.6577,
1719
+ "rewards/accuracies": 0.6312500238418579,
1720
+ "rewards/chosen": -0.361422061920166,
1721
+ "rewards/margins": 0.10248668491840363,
1722
+ "rewards/rejected": -0.46390876173973083,
1723
+ "step": 1100
1724
+ },
1725
+ {
1726
+ "epoch": 0.88,
1727
+ "eval_logits/chosen": -2.5729336738586426,
1728
+ "eval_logits/rejected": -2.580476999282837,
1729
+ "eval_logps/chosen": -351.83758544921875,
1730
+ "eval_logps/rejected": -368.46795654296875,
1731
+ "eval_loss": 0.6706418395042419,
1732
+ "eval_rewards/accuracies": 0.5993333458900452,
1733
+ "eval_rewards/chosen": -0.29342252016067505,
1734
+ "eval_rewards/margins": 0.06933192163705826,
1735
+ "eval_rewards/rejected": -0.3627544641494751,
1736
+ "eval_runtime": 1589.982,
1737
+ "eval_samples_per_second": 1.883,
1738
+ "eval_steps_per_second": 0.236,
1739
+ "step": 1100
1740
+ },
1741
+ {
1742
+ "epoch": 0.89,
1743
+ "learning_rate": 1.8631046257820278e-07,
1744
+ "logits/chosen": -2.713442802429199,
1745
+ "logits/rejected": -2.7153563499450684,
1746
+ "logps/chosen": -448.9380798339844,
1747
+ "logps/rejected": -459.6236267089844,
1748
+ "loss": 0.6822,
1749
+ "rewards/accuracies": 0.606249988079071,
1750
+ "rewards/chosen": -0.40070003271102905,
1751
+ "rewards/margins": 0.04853721335530281,
1752
+ "rewards/rejected": -0.44923728704452515,
1753
+ "step": 1110
1754
+ },
1755
+ {
1756
+ "epoch": 0.9,
1757
+ "learning_rate": 1.6074836247950143e-07,
1758
+ "logits/chosen": -2.7305519580841064,
1759
+ "logits/rejected": -2.727928638458252,
1760
+ "logps/chosen": -458.64208984375,
1761
+ "logps/rejected": -479.4064025878906,
1762
+ "loss": 0.6573,
1763
+ "rewards/accuracies": 0.625,
1764
+ "rewards/chosen": -0.35037961602211,
1765
+ "rewards/margins": 0.09948080033063889,
1766
+ "rewards/rejected": -0.4498603940010071,
1767
+ "step": 1120
1768
+ },
1769
+ {
1770
+ "epoch": 0.9,
1771
+ "learning_rate": 1.370135873643097e-07,
1772
+ "logits/chosen": -2.6915557384490967,
1773
+ "logits/rejected": -2.675395965576172,
1774
+ "logps/chosen": -444.5713806152344,
1775
+ "logps/rejected": -442.8504943847656,
1776
+ "loss": 0.6738,
1777
+ "rewards/accuracies": 0.581250011920929,
1778
+ "rewards/chosen": -0.3383481502532959,
1779
+ "rewards/margins": 0.06244407966732979,
1780
+ "rewards/rejected": -0.4007922112941742,
1781
+ "step": 1130
1782
+ },
1783
+ {
1784
+ "epoch": 0.91,
1785
+ "learning_rate": 1.1512467784059372e-07,
1786
+ "logits/chosen": -2.697908639907837,
1787
+ "logits/rejected": -2.7043871879577637,
1788
+ "logps/chosen": -409.3360595703125,
1789
+ "logps/rejected": -430.56939697265625,
1790
+ "loss": 0.6764,
1791
+ "rewards/accuracies": 0.5562499761581421,
1792
+ "rewards/chosen": -0.353982150554657,
1793
+ "rewards/margins": 0.07203380763530731,
1794
+ "rewards/rejected": -0.4260159432888031,
1795
+ "step": 1140
1796
+ },
1797
+ {
1798
+ "epoch": 0.92,
1799
+ "learning_rate": 9.509873260376251e-08,
1800
+ "logits/chosen": -2.6704964637756348,
1801
+ "logits/rejected": -2.680117130279541,
1802
+ "logps/chosen": -426.01568603515625,
1803
+ "logps/rejected": -481.14056396484375,
1804
+ "loss": 0.6301,
1805
+ "rewards/accuracies": 0.6937500238418579,
1806
+ "rewards/chosen": -0.292033851146698,
1807
+ "rewards/margins": 0.16145673394203186,
1808
+ "rewards/rejected": -0.45349058508872986,
1809
+ "step": 1150
1810
+ },
1811
+ {
1812
+ "epoch": 0.93,
1813
+ "learning_rate": 7.695139507988559e-08,
1814
+ "logits/chosen": -2.690162181854248,
1815
+ "logits/rejected": -2.6840763092041016,
1816
+ "logps/chosen": -455.21429443359375,
1817
+ "logps/rejected": -471.9693908691406,
1818
+ "loss": 0.6911,
1819
+ "rewards/accuracies": 0.574999988079071,
1820
+ "rewards/chosen": -0.39844316244125366,
1821
+ "rewards/margins": 0.04045991227030754,
1822
+ "rewards/rejected": -0.4389030337333679,
1823
+ "step": 1160
1824
+ },
1825
+ {
1826
+ "epoch": 0.94,
1827
+ "learning_rate": 6.069684120570684e-08,
1828
+ "logits/chosen": -2.6355466842651367,
1829
+ "logits/rejected": -2.6722214221954346,
1830
+ "logps/chosen": -428.15887451171875,
1831
+ "logps/rejected": -469.9241638183594,
1832
+ "loss": 0.6556,
1833
+ "rewards/accuracies": 0.65625,
1834
+ "rewards/chosen": -0.31940728425979614,
1835
+ "rewards/margins": 0.10891290009021759,
1836
+ "rewards/rejected": -0.42832016944885254,
1837
+ "step": 1170
1838
+ },
1839
+ {
1840
+ "epoch": 0.94,
1841
+ "learning_rate": 4.634776835499871e-08,
1842
+ "logits/chosen": -2.648555278778076,
1843
+ "logits/rejected": -2.648024082183838,
1844
+ "logps/chosen": -414.77166748046875,
1845
+ "logps/rejected": -438.7622985839844,
1846
+ "loss": 0.6786,
1847
+ "rewards/accuracies": 0.574999988079071,
1848
+ "rewards/chosen": -0.33630624413490295,
1849
+ "rewards/margins": 0.059269875288009644,
1850
+ "rewards/rejected": -0.3955760896205902,
1851
+ "step": 1180
1852
+ },
1853
+ {
1854
+ "epoch": 0.95,
1855
+ "learning_rate": 3.3915385419908964e-08,
1856
+ "logits/chosen": -2.724853992462158,
1857
+ "logits/rejected": -2.7020461559295654,
1858
+ "logps/chosen": -443.90399169921875,
1859
+ "logps/rejected": -462.37384033203125,
1860
+ "loss": 0.667,
1861
+ "rewards/accuracies": 0.625,
1862
+ "rewards/chosen": -0.35228732228279114,
1863
+ "rewards/margins": 0.08694271743297577,
1864
+ "rewards/rejected": -0.4392300546169281,
1865
+ "step": 1190
1866
+ },
1867
+ {
1868
+ "epoch": 0.96,
1869
+ "learning_rate": 2.3409404055043938e-08,
1870
+ "logits/chosen": -2.6520884037017822,
1871
+ "logits/rejected": -2.677919387817383,
1872
+ "logps/chosen": -431.74053955078125,
1873
+ "logps/rejected": -462.65179443359375,
1874
+ "loss": 0.6444,
1875
+ "rewards/accuracies": 0.6625000238418579,
1876
+ "rewards/chosen": -0.3235710561275482,
1877
+ "rewards/margins": 0.12747570872306824,
1878
+ "rewards/rejected": -0.45104676485061646,
1879
+ "step": 1200
1880
+ },
1881
+ {
1882
+ "epoch": 0.96,
1883
+ "eval_logits/chosen": -2.5725319385528564,
1884
+ "eval_logits/rejected": -2.580101490020752,
1885
+ "eval_logps/chosen": -351.09490966796875,
1886
+ "eval_logps/rejected": -367.6591796875,
1887
+ "eval_loss": 0.670768678188324,
1888
+ "eval_rewards/accuracies": 0.5993333458900452,
1889
+ "eval_rewards/chosen": -0.2859960198402405,
1890
+ "eval_rewards/margins": 0.06867069751024246,
1891
+ "eval_rewards/rejected": -0.35466670989990234,
1892
+ "eval_runtime": 1585.5273,
1893
+ "eval_samples_per_second": 1.888,
1894
+ "eval_steps_per_second": 0.237,
1895
+ "step": 1200
1896
+ },
1897
+ {
1898
+ "epoch": 0.97,
1899
+ "learning_rate": 1.4838031091134186e-08,
1900
+ "logits/chosen": -2.6522107124328613,
1901
+ "logits/rejected": -2.6851000785827637,
1902
+ "logps/chosen": -404.271484375,
1903
+ "logps/rejected": -455.322998046875,
1904
+ "loss": 0.6397,
1905
+ "rewards/accuracies": 0.668749988079071,
1906
+ "rewards/chosen": -0.3024134933948517,
1907
+ "rewards/margins": 0.13375119864940643,
1908
+ "rewards/rejected": -0.4361646771430969,
1909
+ "step": 1210
1910
+ },
1911
+ {
1912
+ "epoch": 0.98,
1913
+ "learning_rate": 8.207962124201774e-09,
1914
+ "logits/chosen": -2.687406063079834,
1915
+ "logits/rejected": -2.6915080547332764,
1916
+ "logps/chosen": -446.8517150878906,
1917
+ "logps/rejected": -463.1326599121094,
1918
+ "loss": 0.6692,
1919
+ "rewards/accuracies": 0.612500011920929,
1920
+ "rewards/chosen": -0.33363211154937744,
1921
+ "rewards/margins": 0.08175511658191681,
1922
+ "rewards/rejected": -0.41538724303245544,
1923
+ "step": 1220
1924
+ },
1925
+ {
1926
+ "epoch": 0.98,
1927
+ "learning_rate": 3.5243762852441023e-09,
1928
+ "logits/chosen": -2.68792462348938,
1929
+ "logits/rejected": -2.705341339111328,
1930
+ "logps/chosen": -433.271484375,
1931
+ "logps/rejected": -466.77362060546875,
1932
+ "loss": 0.6609,
1933
+ "rewards/accuracies": 0.581250011920929,
1934
+ "rewards/chosen": -0.34372276067733765,
1935
+ "rewards/margins": 0.1005091518163681,
1936
+ "rewards/rejected": -0.44423192739486694,
1937
+ "step": 1230
1938
+ },
1939
+ {
1940
+ "epoch": 0.99,
1941
+ "learning_rate": 7.909321945129278e-10,
1942
+ "logits/chosen": -2.704862117767334,
1943
+ "logits/rejected": -2.7122840881347656,
1944
+ "logps/chosen": -461.7477111816406,
1945
+ "logps/rejected": -496.6314392089844,
1946
+ "loss": 0.6514,
1947
+ "rewards/accuracies": 0.625,
1948
+ "rewards/chosen": -0.3258623480796814,
1949
+ "rewards/margins": 0.12274013459682465,
1950
+ "rewards/rejected": -0.4486025273799896,
1951
+ "step": 1240
1952
+ },
1953
+ {
1954
+ "epoch": 1.0,
1955
+ "step": 1249,
1956
+ "total_flos": 0.0,
1957
+ "train_loss": 0.67484232365751,
1958
+ "train_runtime": 40604.4701,
1959
+ "train_samples_per_second": 0.492,
1960
+ "train_steps_per_second": 0.031
1961
+ }
1962
+ ],
1963
+ "logging_steps": 10,
1964
+ "max_steps": 1249,
1965
+ "num_input_tokens_seen": 0,
1966
+ "num_train_epochs": 1,
1967
+ "save_steps": 100,
1968
+ "total_flos": 0.0,
1969
+ "train_batch_size": 4,
1970
+ "trial_name": null,
1971
+ "trial_params": null
1972
+ }
training_args.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:fee41ec1b45b86b78f6aa2bc1bd14d3332831518f258d69fa9acb1393bd5b3ed
3
  size 5176
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:387786aa767b34abf9b422cc0cdb3b0676810f6854c743acf6d68e01b38091c3
3
  size 5176