yiran-wang3 commited on
Commit
e30f802
1 Parent(s): 660e4f3

End of training

Browse files
Files changed (6) hide show
  1. README.md +64 -0
  2. all_results.json +9 -0
  3. config.json +1 -1
  4. generation_config.json +14 -0
  5. train_results.json +9 -0
  6. trainer_state.json +1806 -0
README.md ADDED
@@ -0,0 +1,64 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ library_name: transformers
3
+ license: apache-2.0
4
+ base_model: yiran-wang3/qwen2_coder_reflct_adamw_iter1
5
+ tags:
6
+ - alignment-handbook
7
+ - generated_from_trainer
8
+ - trl
9
+ - dpo
10
+ datasets:
11
+ - self-generate/qwcoder2_reflct_sppo_hard_new_cn_mining_oj_iter1-binarized-reflection-scored
12
+ model-index:
13
+ - name: qwen2_coder_reflct_adamw_iter2
14
+ results: []
15
+ ---
16
+
17
+ <!-- This model card has been generated automatically according to the information the Trainer had access to. You
18
+ should probably proofread and complete it, then remove this comment. -->
19
+
20
+ # qwen2_coder_reflct_adamw_iter2
21
+
22
+ This model is a fine-tuned version of [yiran-wang3/qwen2_coder_reflct_adamw_iter1](https://huggingface.co/yiran-wang3/qwen2_coder_reflct_adamw_iter1) on the self-generate/qwcoder2_reflct_sppo_hard_new_cn_mining_oj_iter1-binarized-reflection-scored dataset.
23
+
24
+ ## Model description
25
+
26
+ More information needed
27
+
28
+ ## Intended uses & limitations
29
+
30
+ More information needed
31
+
32
+ ## Training and evaluation data
33
+
34
+ More information needed
35
+
36
+ ## Training procedure
37
+
38
+ ### Training hyperparameters
39
+
40
+ The following hyperparameters were used during training:
41
+ - learning_rate: 1e-06
42
+ - train_batch_size: 8
43
+ - eval_batch_size: 4
44
+ - seed: 42
45
+ - distributed_type: multi-GPU
46
+ - num_devices: 8
47
+ - total_train_batch_size: 64
48
+ - total_eval_batch_size: 32
49
+ - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
50
+ - lr_scheduler_type: constant
51
+ - lr_scheduler_warmup_ratio: 0.1
52
+ - lr_scheduler_warmup_steps: 100
53
+ - num_epochs: 1.0
54
+
55
+ ### Training results
56
+
57
+
58
+
59
+ ### Framework versions
60
+
61
+ - Transformers 4.45.0
62
+ - Pytorch 2.4.0+cu121
63
+ - Datasets 2.14.6
64
+ - Tokenizers 0.20.1
all_results.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 1.0,
3
+ "total_flos": 0.0,
4
+ "train_loss": 0.33750108753641445,
5
+ "train_runtime": 246.5194,
6
+ "train_samples": 5328,
7
+ "train_samples_per_second": 21.613,
8
+ "train_steps_per_second": 0.341
9
+ }
config.json CHANGED
@@ -23,7 +23,7 @@
23
  "tie_word_embeddings": true,
24
  "torch_dtype": "bfloat16",
25
  "transformers_version": "4.45.0",
26
- "use_cache": false,
27
  "use_sliding_window": false,
28
  "vocab_size": 151936
29
  }
 
23
  "tie_word_embeddings": true,
24
  "torch_dtype": "bfloat16",
25
  "transformers_version": "4.45.0",
26
+ "use_cache": true,
27
  "use_sliding_window": false,
28
  "vocab_size": 151936
29
  }
generation_config.json ADDED
@@ -0,0 +1,14 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token_id": 151643,
3
+ "do_sample": true,
4
+ "eos_token_id": [
5
+ 151645,
6
+ 151643
7
+ ],
8
+ "pad_token_id": 151643,
9
+ "repetition_penalty": 1.1,
10
+ "temperature": 0.7,
11
+ "top_k": 20,
12
+ "top_p": 0.8,
13
+ "transformers_version": "4.45.0"
14
+ }
train_results.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 1.0,
3
+ "total_flos": 0.0,
4
+ "train_loss": 0.33750108753641445,
5
+ "train_runtime": 246.5194,
6
+ "train_samples": 5328,
7
+ "train_samples_per_second": 21.613,
8
+ "train_steps_per_second": 0.341
9
+ }
trainer_state.json ADDED
@@ -0,0 +1,1806 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 1.0,
5
+ "eval_steps": 100,
6
+ "global_step": 84,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "debug/policy_chosen_logits": -3.2206931114196777,
13
+ "debug/policy_chosen_logps": -200.86012268066406,
14
+ "debug/policy_rejected_logits": -2.898437738418579,
15
+ "debug/policy_rejected_logps": -208.0646514892578,
16
+ "debug/reference_chosen_logps": -200.86012268066406,
17
+ "debug/reference_rejected_logps": -208.0646514892578,
18
+ "epoch": 0.011904761904761904,
19
+ "grad_norm": 7.909108829449386,
20
+ "learning_rate": 1e-06,
21
+ "logits/chosen": -3.2206931114196777,
22
+ "logits/rejected": -2.898437738418579,
23
+ "logps/chosen": -200.86012268066406,
24
+ "logps/rejected": -208.0646514892578,
25
+ "loss": 0.5,
26
+ "rewards/accuracies": 0.0,
27
+ "rewards/chosen": 0.0,
28
+ "rewards/margins": 0.0,
29
+ "rewards/rejected": 0.0,
30
+ "step": 1
31
+ },
32
+ {
33
+ "debug/policy_chosen_logits": -3.1326141357421875,
34
+ "debug/policy_chosen_logps": -219.65463256835938,
35
+ "debug/policy_rejected_logits": -3.1044466495513916,
36
+ "debug/policy_rejected_logps": -218.29165649414062,
37
+ "debug/reference_chosen_logps": -219.87649536132812,
38
+ "debug/reference_rejected_logps": -218.21566772460938,
39
+ "epoch": 0.023809523809523808,
40
+ "grad_norm": 7.063786345467847,
41
+ "learning_rate": 1e-06,
42
+ "logits/chosen": -3.1326141357421875,
43
+ "logits/rejected": -3.1044466495513916,
44
+ "logps/chosen": -219.65463256835938,
45
+ "logps/rejected": -218.29165649414062,
46
+ "loss": 0.4983,
47
+ "rewards/accuracies": 0.625,
48
+ "rewards/chosen": 0.0022186278365552425,
49
+ "rewards/margins": 0.0029784394428133965,
50
+ "rewards/rejected": -0.0007598113734275103,
51
+ "step": 2
52
+ },
53
+ {
54
+ "debug/policy_chosen_logits": -3.135671377182007,
55
+ "debug/policy_chosen_logps": -208.30648803710938,
56
+ "debug/policy_rejected_logits": -3.0028326511383057,
57
+ "debug/policy_rejected_logps": -239.4796142578125,
58
+ "debug/reference_chosen_logps": -208.16616821289062,
59
+ "debug/reference_rejected_logps": -239.21315002441406,
60
+ "epoch": 0.03571428571428571,
61
+ "grad_norm": 7.814141071934522,
62
+ "learning_rate": 1e-06,
63
+ "logits/chosen": -3.135671377182007,
64
+ "logits/rejected": -3.0028326511383057,
65
+ "logps/chosen": -208.30648803710938,
66
+ "logps/rejected": -239.4796142578125,
67
+ "loss": 0.4933,
68
+ "rewards/accuracies": 0.5,
69
+ "rewards/chosen": -0.0014033315237611532,
70
+ "rewards/margins": 0.0012612915597856045,
71
+ "rewards/rejected": -0.0026646230835467577,
72
+ "step": 3
73
+ },
74
+ {
75
+ "debug/policy_chosen_logits": -3.187422275543213,
76
+ "debug/policy_chosen_logps": -217.25140380859375,
77
+ "debug/policy_rejected_logits": -3.1444270610809326,
78
+ "debug/policy_rejected_logps": -219.7812957763672,
79
+ "debug/reference_chosen_logps": -217.35531616210938,
80
+ "debug/reference_rejected_logps": -218.98512268066406,
81
+ "epoch": 0.047619047619047616,
82
+ "grad_norm": 7.09702591230904,
83
+ "learning_rate": 1e-06,
84
+ "logits/chosen": -3.187422275543213,
85
+ "logits/rejected": -3.1444270610809326,
86
+ "logps/chosen": -217.25140380859375,
87
+ "logps/rejected": -219.7812957763672,
88
+ "loss": 0.4896,
89
+ "rewards/accuracies": 0.625,
90
+ "rewards/chosen": 0.001039218856021762,
91
+ "rewards/margins": 0.009000949561595917,
92
+ "rewards/rejected": -0.007961731404066086,
93
+ "step": 4
94
+ },
95
+ {
96
+ "debug/policy_chosen_logits": -3.0982069969177246,
97
+ "debug/policy_chosen_logps": -216.79022216796875,
98
+ "debug/policy_rejected_logits": -2.9893712997436523,
99
+ "debug/policy_rejected_logps": -236.919189453125,
100
+ "debug/reference_chosen_logps": -217.1226806640625,
101
+ "debug/reference_rejected_logps": -235.52293395996094,
102
+ "epoch": 0.05952380952380952,
103
+ "grad_norm": 5.630512315088139,
104
+ "learning_rate": 1e-06,
105
+ "logits/chosen": -3.0982069969177246,
106
+ "logits/rejected": -2.9893712997436523,
107
+ "logps/chosen": -216.79022216796875,
108
+ "logps/rejected": -236.919189453125,
109
+ "loss": 0.4832,
110
+ "rewards/accuracies": 0.875,
111
+ "rewards/chosen": 0.0033246802631765604,
112
+ "rewards/margins": 0.01728721708059311,
113
+ "rewards/rejected": -0.013962535187602043,
114
+ "step": 5
115
+ },
116
+ {
117
+ "debug/policy_chosen_logits": -3.1938157081604004,
118
+ "debug/policy_chosen_logps": -207.0302734375,
119
+ "debug/policy_rejected_logits": -3.085257053375244,
120
+ "debug/policy_rejected_logps": -217.99813842773438,
121
+ "debug/reference_chosen_logps": -207.5093536376953,
122
+ "debug/reference_rejected_logps": -216.325927734375,
123
+ "epoch": 0.07142857142857142,
124
+ "grad_norm": 5.70424011476702,
125
+ "learning_rate": 1e-06,
126
+ "logits/chosen": -3.1938157081604004,
127
+ "logits/rejected": -3.085257053375244,
128
+ "logps/chosen": -207.0302734375,
129
+ "logps/rejected": -217.99813842773438,
130
+ "loss": 0.4773,
131
+ "rewards/accuracies": 0.625,
132
+ "rewards/chosen": 0.004790782928466797,
133
+ "rewards/margins": 0.02151281200349331,
134
+ "rewards/rejected": -0.01672203093767166,
135
+ "step": 6
136
+ },
137
+ {
138
+ "debug/policy_chosen_logits": -3.163257122039795,
139
+ "debug/policy_chosen_logps": -211.92942810058594,
140
+ "debug/policy_rejected_logits": -3.0449893474578857,
141
+ "debug/policy_rejected_logps": -214.0583953857422,
142
+ "debug/reference_chosen_logps": -212.87844848632812,
143
+ "debug/reference_rejected_logps": -212.81723022460938,
144
+ "epoch": 0.08333333333333333,
145
+ "grad_norm": 6.547599704469646,
146
+ "learning_rate": 1e-06,
147
+ "logits/chosen": -3.163257122039795,
148
+ "logits/rejected": -3.0449893474578857,
149
+ "logps/chosen": -211.92942810058594,
150
+ "logps/rejected": -214.0583953857422,
151
+ "loss": 0.4684,
152
+ "rewards/accuracies": 0.75,
153
+ "rewards/chosen": 0.00949014537036419,
154
+ "rewards/margins": 0.021901872009038925,
155
+ "rewards/rejected": -0.012411728501319885,
156
+ "step": 7
157
+ },
158
+ {
159
+ "debug/policy_chosen_logits": -3.2458667755126953,
160
+ "debug/policy_chosen_logps": -197.6811065673828,
161
+ "debug/policy_rejected_logits": -3.0591673851013184,
162
+ "debug/policy_rejected_logps": -231.7470245361328,
163
+ "debug/reference_chosen_logps": -199.1888427734375,
164
+ "debug/reference_rejected_logps": -228.80226135253906,
165
+ "epoch": 0.09523809523809523,
166
+ "grad_norm": 7.248047931654466,
167
+ "learning_rate": 1e-06,
168
+ "logits/chosen": -3.2458667755126953,
169
+ "logits/rejected": -3.0591673851013184,
170
+ "logps/chosen": -197.6811065673828,
171
+ "logps/rejected": -231.7470245361328,
172
+ "loss": 0.4588,
173
+ "rewards/accuracies": 0.75,
174
+ "rewards/chosen": 0.015077304095029831,
175
+ "rewards/margins": 0.04452499374747276,
176
+ "rewards/rejected": -0.029447689652442932,
177
+ "step": 8
178
+ },
179
+ {
180
+ "debug/policy_chosen_logits": -3.212696075439453,
181
+ "debug/policy_chosen_logps": -196.501708984375,
182
+ "debug/policy_rejected_logits": -3.0544850826263428,
183
+ "debug/policy_rejected_logps": -217.50942993164062,
184
+ "debug/reference_chosen_logps": -198.6233673095703,
185
+ "debug/reference_rejected_logps": -214.6073760986328,
186
+ "epoch": 0.10714285714285714,
187
+ "grad_norm": 6.028275040076952,
188
+ "learning_rate": 1e-06,
189
+ "logits/chosen": -3.212696075439453,
190
+ "logits/rejected": -3.0544850826263428,
191
+ "logps/chosen": -196.501708984375,
192
+ "logps/rejected": -217.50942993164062,
193
+ "loss": 0.4341,
194
+ "rewards/accuracies": 0.75,
195
+ "rewards/chosen": 0.021216563880443573,
196
+ "rewards/margins": 0.050237104296684265,
197
+ "rewards/rejected": -0.029020538553595543,
198
+ "step": 9
199
+ },
200
+ {
201
+ "debug/policy_chosen_logits": -3.1316659450531006,
202
+ "debug/policy_chosen_logps": -216.9590301513672,
203
+ "debug/policy_rejected_logits": -3.0151987075805664,
204
+ "debug/policy_rejected_logps": -225.83355712890625,
205
+ "debug/reference_chosen_logps": -221.61790466308594,
206
+ "debug/reference_rejected_logps": -219.59786987304688,
207
+ "epoch": 0.11904761904761904,
208
+ "grad_norm": 5.4813634886972515,
209
+ "learning_rate": 1e-06,
210
+ "logits/chosen": -3.1316659450531006,
211
+ "logits/rejected": -3.0151987075805664,
212
+ "logps/chosen": -216.9590301513672,
213
+ "logps/rejected": -225.83355712890625,
214
+ "loss": 0.4064,
215
+ "rewards/accuracies": 0.625,
216
+ "rewards/chosen": 0.046588800847530365,
217
+ "rewards/margins": 0.10894560813903809,
218
+ "rewards/rejected": -0.06235681474208832,
219
+ "step": 10
220
+ },
221
+ {
222
+ "debug/policy_chosen_logits": -3.2891530990600586,
223
+ "debug/policy_chosen_logps": -196.72274780273438,
224
+ "debug/policy_rejected_logits": -3.048377513885498,
225
+ "debug/policy_rejected_logps": -233.62039184570312,
226
+ "debug/reference_chosen_logps": -202.43775939941406,
227
+ "debug/reference_rejected_logps": -216.876708984375,
228
+ "epoch": 0.13095238095238096,
229
+ "grad_norm": 4.876535878960536,
230
+ "learning_rate": 1e-06,
231
+ "logits/chosen": -3.2891530990600586,
232
+ "logits/rejected": -3.048377513885498,
233
+ "logps/chosen": -196.72274780273438,
234
+ "logps/rejected": -233.62039184570312,
235
+ "loss": 0.4083,
236
+ "rewards/accuracies": 0.875,
237
+ "rewards/chosen": 0.05715004354715347,
238
+ "rewards/margins": 0.22458696365356445,
239
+ "rewards/rejected": -0.16743692755699158,
240
+ "step": 11
241
+ },
242
+ {
243
+ "debug/policy_chosen_logits": -3.1717939376831055,
244
+ "debug/policy_chosen_logps": -210.72622680664062,
245
+ "debug/policy_rejected_logits": -3.060811758041382,
246
+ "debug/policy_rejected_logps": -230.44659423828125,
247
+ "debug/reference_chosen_logps": -219.1645050048828,
248
+ "debug/reference_rejected_logps": -223.36727905273438,
249
+ "epoch": 0.14285714285714285,
250
+ "grad_norm": 5.9567820612609585,
251
+ "learning_rate": 1e-06,
252
+ "logits/chosen": -3.1717939376831055,
253
+ "logits/rejected": -3.060811758041382,
254
+ "logps/chosen": -210.72622680664062,
255
+ "logps/rejected": -230.44659423828125,
256
+ "loss": 0.3617,
257
+ "rewards/accuracies": 1.0,
258
+ "rewards/chosen": 0.08438283205032349,
259
+ "rewards/margins": 0.155176043510437,
260
+ "rewards/rejected": -0.07079321146011353,
261
+ "step": 12
262
+ },
263
+ {
264
+ "debug/policy_chosen_logits": -3.175966739654541,
265
+ "debug/policy_chosen_logps": -198.75677490234375,
266
+ "debug/policy_rejected_logits": -3.0943973064422607,
267
+ "debug/policy_rejected_logps": -209.6800994873047,
268
+ "debug/reference_chosen_logps": -206.38829040527344,
269
+ "debug/reference_rejected_logps": -198.82118225097656,
270
+ "epoch": 0.15476190476190477,
271
+ "grad_norm": 4.372108578101748,
272
+ "learning_rate": 1e-06,
273
+ "logits/chosen": -3.175966739654541,
274
+ "logits/rejected": -3.0943973064422607,
275
+ "logps/chosen": -198.75677490234375,
276
+ "logps/rejected": -209.6800994873047,
277
+ "loss": 0.378,
278
+ "rewards/accuracies": 0.75,
279
+ "rewards/chosen": 0.07631513476371765,
280
+ "rewards/margins": 0.1849043220281601,
281
+ "rewards/rejected": -0.10858918726444244,
282
+ "step": 13
283
+ },
284
+ {
285
+ "debug/policy_chosen_logits": -3.17907977104187,
286
+ "debug/policy_chosen_logps": -204.30484008789062,
287
+ "debug/policy_rejected_logits": -3.0033490657806396,
288
+ "debug/policy_rejected_logps": -224.59976196289062,
289
+ "debug/reference_chosen_logps": -215.46783447265625,
290
+ "debug/reference_rejected_logps": -213.1477813720703,
291
+ "epoch": 0.16666666666666666,
292
+ "grad_norm": 4.951824369660157,
293
+ "learning_rate": 1e-06,
294
+ "logits/chosen": -3.17907977104187,
295
+ "logits/rejected": -3.0033490657806396,
296
+ "logps/chosen": -204.30484008789062,
297
+ "logps/rejected": -224.59976196289062,
298
+ "loss": 0.354,
299
+ "rewards/accuracies": 1.0,
300
+ "rewards/chosen": 0.1116299033164978,
301
+ "rewards/margins": 0.22614973783493042,
302
+ "rewards/rejected": -0.11451983451843262,
303
+ "step": 14
304
+ },
305
+ {
306
+ "debug/policy_chosen_logits": -3.0961408615112305,
307
+ "debug/policy_chosen_logps": -209.14535522460938,
308
+ "debug/policy_rejected_logits": -3.00311279296875,
309
+ "debug/policy_rejected_logps": -224.08718872070312,
310
+ "debug/reference_chosen_logps": -220.22549438476562,
311
+ "debug/reference_rejected_logps": -222.09124755859375,
312
+ "epoch": 0.17857142857142858,
313
+ "grad_norm": 4.835613329592286,
314
+ "learning_rate": 1e-06,
315
+ "logits/chosen": -3.0961408615112305,
316
+ "logits/rejected": -3.00311279296875,
317
+ "logps/chosen": -209.14535522460938,
318
+ "logps/rejected": -224.08718872070312,
319
+ "loss": 0.386,
320
+ "rewards/accuracies": 0.75,
321
+ "rewards/chosen": 0.11080135405063629,
322
+ "rewards/margins": 0.13076072931289673,
323
+ "rewards/rejected": -0.019959375262260437,
324
+ "step": 15
325
+ },
326
+ {
327
+ "debug/policy_chosen_logits": -3.2088327407836914,
328
+ "debug/policy_chosen_logps": -206.8312530517578,
329
+ "debug/policy_rejected_logits": -3.0719218254089355,
330
+ "debug/policy_rejected_logps": -226.76593017578125,
331
+ "debug/reference_chosen_logps": -218.99668884277344,
332
+ "debug/reference_rejected_logps": -211.85366821289062,
333
+ "epoch": 0.19047619047619047,
334
+ "grad_norm": 4.544057800793572,
335
+ "learning_rate": 1e-06,
336
+ "logits/chosen": -3.2088327407836914,
337
+ "logits/rejected": -3.0719218254089355,
338
+ "logps/chosen": -206.8312530517578,
339
+ "logps/rejected": -226.76593017578125,
340
+ "loss": 0.3723,
341
+ "rewards/accuracies": 0.875,
342
+ "rewards/chosen": 0.12165433168411255,
343
+ "rewards/margins": 0.27077698707580566,
344
+ "rewards/rejected": -0.1491226851940155,
345
+ "step": 16
346
+ },
347
+ {
348
+ "debug/policy_chosen_logits": -3.1967031955718994,
349
+ "debug/policy_chosen_logps": -206.11988830566406,
350
+ "debug/policy_rejected_logits": -3.192483425140381,
351
+ "debug/policy_rejected_logps": -230.90736389160156,
352
+ "debug/reference_chosen_logps": -221.86349487304688,
353
+ "debug/reference_rejected_logps": -223.01187133789062,
354
+ "epoch": 0.20238095238095238,
355
+ "grad_norm": 4.376140545216712,
356
+ "learning_rate": 1e-06,
357
+ "logits/chosen": -3.1967031955718994,
358
+ "logits/rejected": -3.192483425140381,
359
+ "logps/chosen": -206.11988830566406,
360
+ "logps/rejected": -230.90736389160156,
361
+ "loss": 0.3998,
362
+ "rewards/accuracies": 0.75,
363
+ "rewards/chosen": 0.15743595361709595,
364
+ "rewards/margins": 0.23639078438282013,
365
+ "rewards/rejected": -0.07895481586456299,
366
+ "step": 17
367
+ },
368
+ {
369
+ "debug/policy_chosen_logits": -3.1152353286743164,
370
+ "debug/policy_chosen_logps": -193.357177734375,
371
+ "debug/policy_rejected_logits": -2.959299087524414,
372
+ "debug/policy_rejected_logps": -263.7359313964844,
373
+ "debug/reference_chosen_logps": -210.5402069091797,
374
+ "debug/reference_rejected_logps": -232.1613311767578,
375
+ "epoch": 0.21428571428571427,
376
+ "grad_norm": 7.866472140538224,
377
+ "learning_rate": 1e-06,
378
+ "logits/chosen": -3.1152353286743164,
379
+ "logits/rejected": -2.959299087524414,
380
+ "logps/chosen": -193.357177734375,
381
+ "logps/rejected": -263.7359313964844,
382
+ "loss": 0.4077,
383
+ "rewards/accuracies": 1.0,
384
+ "rewards/chosen": 0.17183025181293488,
385
+ "rewards/margins": 0.4875761568546295,
386
+ "rewards/rejected": -0.31574589014053345,
387
+ "step": 18
388
+ },
389
+ {
390
+ "debug/policy_chosen_logits": -3.2761292457580566,
391
+ "debug/policy_chosen_logps": -185.4181365966797,
392
+ "debug/policy_rejected_logits": -3.269240379333496,
393
+ "debug/policy_rejected_logps": -225.70327758789062,
394
+ "debug/reference_chosen_logps": -202.86187744140625,
395
+ "debug/reference_rejected_logps": -228.69943237304688,
396
+ "epoch": 0.2261904761904762,
397
+ "grad_norm": 7.929981237103111,
398
+ "learning_rate": 1e-06,
399
+ "logits/chosen": -3.2761292457580566,
400
+ "logits/rejected": -3.269240379333496,
401
+ "logps/chosen": -185.4181365966797,
402
+ "logps/rejected": -225.70327758789062,
403
+ "loss": 0.3953,
404
+ "rewards/accuracies": 0.625,
405
+ "rewards/chosen": 0.17443738877773285,
406
+ "rewards/margins": 0.14447586238384247,
407
+ "rewards/rejected": 0.02996152639389038,
408
+ "step": 19
409
+ },
410
+ {
411
+ "debug/policy_chosen_logits": -3.2340097427368164,
412
+ "debug/policy_chosen_logps": -193.43096923828125,
413
+ "debug/policy_rejected_logits": -3.099465847015381,
414
+ "debug/policy_rejected_logps": -230.0125732421875,
415
+ "debug/reference_chosen_logps": -216.61485290527344,
416
+ "debug/reference_rejected_logps": -211.0814208984375,
417
+ "epoch": 0.23809523809523808,
418
+ "grad_norm": 8.640304701773179,
419
+ "learning_rate": 1e-06,
420
+ "logits/chosen": -3.2340097427368164,
421
+ "logits/rejected": -3.099465847015381,
422
+ "logps/chosen": -193.43096923828125,
423
+ "logps/rejected": -230.0125732421875,
424
+ "loss": 0.3575,
425
+ "rewards/accuracies": 1.0,
426
+ "rewards/chosen": 0.23183873295783997,
427
+ "rewards/margins": 0.42115041613578796,
428
+ "rewards/rejected": -0.189311683177948,
429
+ "step": 20
430
+ },
431
+ {
432
+ "debug/policy_chosen_logits": -3.162797689437866,
433
+ "debug/policy_chosen_logps": -191.775146484375,
434
+ "debug/policy_rejected_logits": -3.2479846477508545,
435
+ "debug/policy_rejected_logps": -217.1043243408203,
436
+ "debug/reference_chosen_logps": -215.61062622070312,
437
+ "debug/reference_rejected_logps": -209.42953491210938,
438
+ "epoch": 0.25,
439
+ "grad_norm": 7.114249349006281,
440
+ "learning_rate": 1e-06,
441
+ "logits/chosen": -3.162797689437866,
442
+ "logits/rejected": -3.2479846477508545,
443
+ "logps/chosen": -191.775146484375,
444
+ "logps/rejected": -217.1043243408203,
445
+ "loss": 0.4064,
446
+ "rewards/accuracies": 0.875,
447
+ "rewards/chosen": 0.23835478723049164,
448
+ "rewards/margins": 0.3151026964187622,
449
+ "rewards/rejected": -0.07674790918827057,
450
+ "step": 21
451
+ },
452
+ {
453
+ "debug/policy_chosen_logits": -3.3014657497406006,
454
+ "debug/policy_chosen_logps": -185.117919921875,
455
+ "debug/policy_rejected_logits": -3.2064030170440674,
456
+ "debug/policy_rejected_logps": -231.311279296875,
457
+ "debug/reference_chosen_logps": -208.03945922851562,
458
+ "debug/reference_rejected_logps": -204.46234130859375,
459
+ "epoch": 0.2619047619047619,
460
+ "grad_norm": 6.717051925734186,
461
+ "learning_rate": 1e-06,
462
+ "logits/chosen": -3.3014657497406006,
463
+ "logits/rejected": -3.2064030170440674,
464
+ "logps/chosen": -185.117919921875,
465
+ "logps/rejected": -231.311279296875,
466
+ "loss": 0.4124,
467
+ "rewards/accuracies": 0.875,
468
+ "rewards/chosen": 0.22921523451805115,
469
+ "rewards/margins": 0.4977045953273773,
470
+ "rewards/rejected": -0.26848936080932617,
471
+ "step": 22
472
+ },
473
+ {
474
+ "debug/policy_chosen_logits": -3.2503294944763184,
475
+ "debug/policy_chosen_logps": -193.4159698486328,
476
+ "debug/policy_rejected_logits": -3.1396243572235107,
477
+ "debug/policy_rejected_logps": -254.65135192871094,
478
+ "debug/reference_chosen_logps": -217.694580078125,
479
+ "debug/reference_rejected_logps": -212.81295776367188,
480
+ "epoch": 0.27380952380952384,
481
+ "grad_norm": 7.339797017549244,
482
+ "learning_rate": 1e-06,
483
+ "logits/chosen": -3.2503294944763184,
484
+ "logits/rejected": -3.1396243572235107,
485
+ "logps/chosen": -193.4159698486328,
486
+ "logps/rejected": -254.65135192871094,
487
+ "loss": 0.3197,
488
+ "rewards/accuracies": 1.0,
489
+ "rewards/chosen": 0.24278610944747925,
490
+ "rewards/margins": 0.6611701250076294,
491
+ "rewards/rejected": -0.41838398575782776,
492
+ "step": 23
493
+ },
494
+ {
495
+ "debug/policy_chosen_logits": -3.299981117248535,
496
+ "debug/policy_chosen_logps": -185.9925537109375,
497
+ "debug/policy_rejected_logits": -3.0819525718688965,
498
+ "debug/policy_rejected_logps": -258.153076171875,
499
+ "debug/reference_chosen_logps": -212.69705200195312,
500
+ "debug/reference_rejected_logps": -221.9197235107422,
501
+ "epoch": 0.2857142857142857,
502
+ "grad_norm": 6.4780070021425145,
503
+ "learning_rate": 1e-06,
504
+ "logits/chosen": -3.299981117248535,
505
+ "logits/rejected": -3.0819525718688965,
506
+ "logps/chosen": -185.9925537109375,
507
+ "logps/rejected": -258.153076171875,
508
+ "loss": 0.3351,
509
+ "rewards/accuracies": 0.875,
510
+ "rewards/chosen": 0.2670450806617737,
511
+ "rewards/margins": 0.629378616809845,
512
+ "rewards/rejected": -0.3623335361480713,
513
+ "step": 24
514
+ },
515
+ {
516
+ "debug/policy_chosen_logits": -3.3384597301483154,
517
+ "debug/policy_chosen_logps": -179.01553344726562,
518
+ "debug/policy_rejected_logits": -3.2596333026885986,
519
+ "debug/policy_rejected_logps": -230.0491180419922,
520
+ "debug/reference_chosen_logps": -201.91806030273438,
521
+ "debug/reference_rejected_logps": -212.782958984375,
522
+ "epoch": 0.2976190476190476,
523
+ "grad_norm": 5.646156232636996,
524
+ "learning_rate": 1e-06,
525
+ "logits/chosen": -3.3384597301483154,
526
+ "logits/rejected": -3.2596333026885986,
527
+ "logps/chosen": -179.01553344726562,
528
+ "logps/rejected": -230.0491180419922,
529
+ "loss": 0.3597,
530
+ "rewards/accuracies": 0.875,
531
+ "rewards/chosen": 0.2290252447128296,
532
+ "rewards/margins": 0.40168675780296326,
533
+ "rewards/rejected": -0.17266148328781128,
534
+ "step": 25
535
+ },
536
+ {
537
+ "debug/policy_chosen_logits": -3.2951643466949463,
538
+ "debug/policy_chosen_logps": -187.70423889160156,
539
+ "debug/policy_rejected_logits": -3.064873218536377,
540
+ "debug/policy_rejected_logps": -259.9974365234375,
541
+ "debug/reference_chosen_logps": -209.17364501953125,
542
+ "debug/reference_rejected_logps": -230.99514770507812,
543
+ "epoch": 0.30952380952380953,
544
+ "grad_norm": 5.004804914940331,
545
+ "learning_rate": 1e-06,
546
+ "logits/chosen": -3.2951643466949463,
547
+ "logits/rejected": -3.064873218536377,
548
+ "logps/chosen": -187.70423889160156,
549
+ "logps/rejected": -259.9974365234375,
550
+ "loss": 0.3607,
551
+ "rewards/accuracies": 0.75,
552
+ "rewards/chosen": 0.21469400823116302,
553
+ "rewards/margins": 0.5047171115875244,
554
+ "rewards/rejected": -0.2900230884552002,
555
+ "step": 26
556
+ },
557
+ {
558
+ "debug/policy_chosen_logits": -3.205540895462036,
559
+ "debug/policy_chosen_logps": -202.82135009765625,
560
+ "debug/policy_rejected_logits": -3.1168112754821777,
561
+ "debug/policy_rejected_logps": -220.36105346679688,
562
+ "debug/reference_chosen_logps": -219.64862060546875,
563
+ "debug/reference_rejected_logps": -217.84548950195312,
564
+ "epoch": 0.32142857142857145,
565
+ "grad_norm": 7.861811148122209,
566
+ "learning_rate": 1e-06,
567
+ "logits/chosen": -3.205540895462036,
568
+ "logits/rejected": -3.1168112754821777,
569
+ "logps/chosen": -202.82135009765625,
570
+ "logps/rejected": -220.36105346679688,
571
+ "loss": 0.3706,
572
+ "rewards/accuracies": 0.75,
573
+ "rewards/chosen": 0.16827276349067688,
574
+ "rewards/margins": 0.19342824816703796,
575
+ "rewards/rejected": -0.02515549585223198,
576
+ "step": 27
577
+ },
578
+ {
579
+ "debug/policy_chosen_logits": -3.266860246658325,
580
+ "debug/policy_chosen_logps": -190.13174438476562,
581
+ "debug/policy_rejected_logits": -3.2241365909576416,
582
+ "debug/policy_rejected_logps": -217.80401611328125,
583
+ "debug/reference_chosen_logps": -207.9426727294922,
584
+ "debug/reference_rejected_logps": -213.35520935058594,
585
+ "epoch": 0.3333333333333333,
586
+ "grad_norm": 5.17181036610514,
587
+ "learning_rate": 1e-06,
588
+ "logits/chosen": -3.266860246658325,
589
+ "logits/rejected": -3.2241365909576416,
590
+ "logps/chosen": -190.13174438476562,
591
+ "logps/rejected": -217.80401611328125,
592
+ "loss": 0.381,
593
+ "rewards/accuracies": 0.875,
594
+ "rewards/chosen": 0.17810919880867004,
595
+ "rewards/margins": 0.22259722650051117,
596
+ "rewards/rejected": -0.044488027691841125,
597
+ "step": 28
598
+ },
599
+ {
600
+ "debug/policy_chosen_logits": -3.138282060623169,
601
+ "debug/policy_chosen_logps": -194.2283935546875,
602
+ "debug/policy_rejected_logits": -3.242144823074341,
603
+ "debug/policy_rejected_logps": -240.74034118652344,
604
+ "debug/reference_chosen_logps": -205.63931274414062,
605
+ "debug/reference_rejected_logps": -223.89479064941406,
606
+ "epoch": 0.34523809523809523,
607
+ "grad_norm": 13.702732402282919,
608
+ "learning_rate": 1e-06,
609
+ "logits/chosen": -3.138282060623169,
610
+ "logits/rejected": -3.242144823074341,
611
+ "logps/chosen": -194.2283935546875,
612
+ "logps/rejected": -240.74034118652344,
613
+ "loss": 0.3319,
614
+ "rewards/accuracies": 0.875,
615
+ "rewards/chosen": 0.11410927027463913,
616
+ "rewards/margins": 0.2825648784637451,
617
+ "rewards/rejected": -0.1684555858373642,
618
+ "step": 29
619
+ },
620
+ {
621
+ "debug/policy_chosen_logits": -3.2786245346069336,
622
+ "debug/policy_chosen_logps": -196.5978546142578,
623
+ "debug/policy_rejected_logits": -3.1386148929595947,
624
+ "debug/policy_rejected_logps": -242.57998657226562,
625
+ "debug/reference_chosen_logps": -202.89393615722656,
626
+ "debug/reference_rejected_logps": -212.0622100830078,
627
+ "epoch": 0.35714285714285715,
628
+ "grad_norm": 18.8411217134156,
629
+ "learning_rate": 1e-06,
630
+ "logits/chosen": -3.2786245346069336,
631
+ "logits/rejected": -3.1386148929595947,
632
+ "logps/chosen": -196.5978546142578,
633
+ "logps/rejected": -242.57998657226562,
634
+ "loss": 0.3392,
635
+ "rewards/accuracies": 0.875,
636
+ "rewards/chosen": 0.0629609078168869,
637
+ "rewards/margins": 0.36813876032829285,
638
+ "rewards/rejected": -0.30517783761024475,
639
+ "step": 30
640
+ },
641
+ {
642
+ "debug/policy_chosen_logits": -3.220453977584839,
643
+ "debug/policy_chosen_logps": -210.29718017578125,
644
+ "debug/policy_rejected_logits": -3.0406980514526367,
645
+ "debug/policy_rejected_logps": -254.49639892578125,
646
+ "debug/reference_chosen_logps": -217.5068359375,
647
+ "debug/reference_rejected_logps": -222.9973907470703,
648
+ "epoch": 0.36904761904761907,
649
+ "grad_norm": 17.63556570482219,
650
+ "learning_rate": 1e-06,
651
+ "logits/chosen": -3.220453977584839,
652
+ "logits/rejected": -3.0406980514526367,
653
+ "logps/chosen": -210.29718017578125,
654
+ "logps/rejected": -254.49639892578125,
655
+ "loss": 0.3337,
656
+ "rewards/accuracies": 0.875,
657
+ "rewards/chosen": 0.0720965787768364,
658
+ "rewards/margins": 0.38708674907684326,
659
+ "rewards/rejected": -0.31499019265174866,
660
+ "step": 31
661
+ },
662
+ {
663
+ "debug/policy_chosen_logits": -3.283477544784546,
664
+ "debug/policy_chosen_logps": -195.7086944580078,
665
+ "debug/policy_rejected_logits": -3.2998650074005127,
666
+ "debug/policy_rejected_logps": -248.77716064453125,
667
+ "debug/reference_chosen_logps": -205.69100952148438,
668
+ "debug/reference_rejected_logps": -238.83279418945312,
669
+ "epoch": 0.38095238095238093,
670
+ "grad_norm": 3.545596888211082,
671
+ "learning_rate": 1e-06,
672
+ "logits/chosen": -3.283477544784546,
673
+ "logits/rejected": -3.2998650074005127,
674
+ "logps/chosen": -195.7086944580078,
675
+ "logps/rejected": -248.77716064453125,
676
+ "loss": 0.3367,
677
+ "rewards/accuracies": 0.875,
678
+ "rewards/chosen": 0.09982330352067947,
679
+ "rewards/margins": 0.19926708936691284,
680
+ "rewards/rejected": -0.09944379329681396,
681
+ "step": 32
682
+ },
683
+ {
684
+ "debug/policy_chosen_logits": -3.2302284240722656,
685
+ "debug/policy_chosen_logps": -189.14857482910156,
686
+ "debug/policy_rejected_logits": -3.1394155025482178,
687
+ "debug/policy_rejected_logps": -231.4539031982422,
688
+ "debug/reference_chosen_logps": -204.17739868164062,
689
+ "debug/reference_rejected_logps": -209.3110809326172,
690
+ "epoch": 0.39285714285714285,
691
+ "grad_norm": 3.835026591129013,
692
+ "learning_rate": 1e-06,
693
+ "logits/chosen": -3.2302284240722656,
694
+ "logits/rejected": -3.1394155025482178,
695
+ "logps/chosen": -189.14857482910156,
696
+ "logps/rejected": -231.4539031982422,
697
+ "loss": 0.337,
698
+ "rewards/accuracies": 0.875,
699
+ "rewards/chosen": 0.15028832852840424,
700
+ "rewards/margins": 0.3717164993286133,
701
+ "rewards/rejected": -0.22142818570137024,
702
+ "step": 33
703
+ },
704
+ {
705
+ "debug/policy_chosen_logits": -3.2129104137420654,
706
+ "debug/policy_chosen_logps": -199.6995849609375,
707
+ "debug/policy_rejected_logits": -3.0475525856018066,
708
+ "debug/policy_rejected_logps": -242.47573852539062,
709
+ "debug/reference_chosen_logps": -211.38697814941406,
710
+ "debug/reference_rejected_logps": -218.95986938476562,
711
+ "epoch": 0.40476190476190477,
712
+ "grad_norm": 3.6065017129358563,
713
+ "learning_rate": 1e-06,
714
+ "logits/chosen": -3.2129104137420654,
715
+ "logits/rejected": -3.0475525856018066,
716
+ "logps/chosen": -199.6995849609375,
717
+ "logps/rejected": -242.47573852539062,
718
+ "loss": 0.3252,
719
+ "rewards/accuracies": 0.75,
720
+ "rewards/chosen": 0.11687390506267548,
721
+ "rewards/margins": 0.35203248262405396,
722
+ "rewards/rejected": -0.23515859246253967,
723
+ "step": 34
724
+ },
725
+ {
726
+ "debug/policy_chosen_logits": -3.0805389881134033,
727
+ "debug/policy_chosen_logps": -200.73974609375,
728
+ "debug/policy_rejected_logits": -3.04288649559021,
729
+ "debug/policy_rejected_logps": -218.5482940673828,
730
+ "debug/reference_chosen_logps": -213.4292449951172,
731
+ "debug/reference_rejected_logps": -223.4329071044922,
732
+ "epoch": 0.4166666666666667,
733
+ "grad_norm": 5.765726016207518,
734
+ "learning_rate": 1e-06,
735
+ "logits/chosen": -3.0805389881134033,
736
+ "logits/rejected": -3.04288649559021,
737
+ "logps/chosen": -200.73974609375,
738
+ "logps/rejected": -218.5482940673828,
739
+ "loss": 0.3783,
740
+ "rewards/accuracies": 0.75,
741
+ "rewards/chosen": 0.12689509987831116,
742
+ "rewards/margins": 0.07804906368255615,
743
+ "rewards/rejected": 0.048846036195755005,
744
+ "step": 35
745
+ },
746
+ {
747
+ "debug/policy_chosen_logits": -3.339388370513916,
748
+ "debug/policy_chosen_logps": -175.0019073486328,
749
+ "debug/policy_rejected_logits": -3.125450611114502,
750
+ "debug/policy_rejected_logps": -234.52488708496094,
751
+ "debug/reference_chosen_logps": -191.13693237304688,
752
+ "debug/reference_rejected_logps": -211.32481384277344,
753
+ "epoch": 0.42857142857142855,
754
+ "grad_norm": 3.4118668773069563,
755
+ "learning_rate": 1e-06,
756
+ "logits/chosen": -3.339388370513916,
757
+ "logits/rejected": -3.125450611114502,
758
+ "logps/chosen": -175.0019073486328,
759
+ "logps/rejected": -234.52488708496094,
760
+ "loss": 0.324,
761
+ "rewards/accuracies": 0.875,
762
+ "rewards/chosen": 0.16135042905807495,
763
+ "rewards/margins": 0.3933510482311249,
764
+ "rewards/rejected": -0.23200063407421112,
765
+ "step": 36
766
+ },
767
+ {
768
+ "debug/policy_chosen_logits": -3.1920435428619385,
769
+ "debug/policy_chosen_logps": -200.70831298828125,
770
+ "debug/policy_rejected_logits": -3.0675861835479736,
771
+ "debug/policy_rejected_logps": -240.75643920898438,
772
+ "debug/reference_chosen_logps": -220.32891845703125,
773
+ "debug/reference_rejected_logps": -224.6131591796875,
774
+ "epoch": 0.44047619047619047,
775
+ "grad_norm": 3.775080381030047,
776
+ "learning_rate": 1e-06,
777
+ "logits/chosen": -3.1920435428619385,
778
+ "logits/rejected": -3.0675861835479736,
779
+ "logps/chosen": -200.70831298828125,
780
+ "logps/rejected": -240.75643920898438,
781
+ "loss": 0.3004,
782
+ "rewards/accuracies": 1.0,
783
+ "rewards/chosen": 0.19620609283447266,
784
+ "rewards/margins": 0.357638955116272,
785
+ "rewards/rejected": -0.16143286228179932,
786
+ "step": 37
787
+ },
788
+ {
789
+ "debug/policy_chosen_logits": -3.173511505126953,
790
+ "debug/policy_chosen_logps": -203.63047790527344,
791
+ "debug/policy_rejected_logits": -3.0617382526397705,
792
+ "debug/policy_rejected_logps": -252.37109375,
793
+ "debug/reference_chosen_logps": -214.03150939941406,
794
+ "debug/reference_rejected_logps": -230.29051208496094,
795
+ "epoch": 0.4523809523809524,
796
+ "grad_norm": 4.456940286801414,
797
+ "learning_rate": 1e-06,
798
+ "logits/chosen": -3.173511505126953,
799
+ "logits/rejected": -3.0617382526397705,
800
+ "logps/chosen": -203.63047790527344,
801
+ "logps/rejected": -252.37109375,
802
+ "loss": 0.3421,
803
+ "rewards/accuracies": 1.0,
804
+ "rewards/chosen": 0.10401026904582977,
805
+ "rewards/margins": 0.3248162269592285,
806
+ "rewards/rejected": -0.22080595791339874,
807
+ "step": 38
808
+ },
809
+ {
810
+ "debug/policy_chosen_logits": -3.153968095779419,
811
+ "debug/policy_chosen_logps": -215.73043823242188,
812
+ "debug/policy_rejected_logits": -3.1119275093078613,
813
+ "debug/policy_rejected_logps": -230.99270629882812,
814
+ "debug/reference_chosen_logps": -224.58657836914062,
815
+ "debug/reference_rejected_logps": -230.7893829345703,
816
+ "epoch": 0.4642857142857143,
817
+ "grad_norm": 3.7503647290224547,
818
+ "learning_rate": 1e-06,
819
+ "logits/chosen": -3.153968095779419,
820
+ "logits/rejected": -3.1119275093078613,
821
+ "logps/chosen": -215.73043823242188,
822
+ "logps/rejected": -230.99270629882812,
823
+ "loss": 0.3678,
824
+ "rewards/accuracies": 0.625,
825
+ "rewards/chosen": 0.08856132626533508,
826
+ "rewards/margins": 0.09059463441371918,
827
+ "rewards/rejected": -0.0020333081483840942,
828
+ "step": 39
829
+ },
830
+ {
831
+ "debug/policy_chosen_logits": -3.0754432678222656,
832
+ "debug/policy_chosen_logps": -196.66781616210938,
833
+ "debug/policy_rejected_logits": -2.9984235763549805,
834
+ "debug/policy_rejected_logps": -238.45590209960938,
835
+ "debug/reference_chosen_logps": -210.78643798828125,
836
+ "debug/reference_rejected_logps": -208.17205810546875,
837
+ "epoch": 0.47619047619047616,
838
+ "grad_norm": 4.009273704463349,
839
+ "learning_rate": 1e-06,
840
+ "logits/chosen": -3.0754432678222656,
841
+ "logits/rejected": -2.9984235763549805,
842
+ "logps/chosen": -196.66781616210938,
843
+ "logps/rejected": -238.45590209960938,
844
+ "loss": 0.327,
845
+ "rewards/accuracies": 0.875,
846
+ "rewards/chosen": 0.14118614792823792,
847
+ "rewards/margins": 0.4440246522426605,
848
+ "rewards/rejected": -0.3028385043144226,
849
+ "step": 40
850
+ },
851
+ {
852
+ "debug/policy_chosen_logits": -3.180971622467041,
853
+ "debug/policy_chosen_logps": -211.52041625976562,
854
+ "debug/policy_rejected_logits": -3.1294796466827393,
855
+ "debug/policy_rejected_logps": -226.37899780273438,
856
+ "debug/reference_chosen_logps": -221.50608825683594,
857
+ "debug/reference_rejected_logps": -220.98367309570312,
858
+ "epoch": 0.4880952380952381,
859
+ "grad_norm": 4.554216689284069,
860
+ "learning_rate": 1e-06,
861
+ "logits/chosen": -3.180971622467041,
862
+ "logits/rejected": -3.1294796466827393,
863
+ "logps/chosen": -211.52041625976562,
864
+ "logps/rejected": -226.37899780273438,
865
+ "loss": 0.3403,
866
+ "rewards/accuracies": 0.875,
867
+ "rewards/chosen": 0.09985677897930145,
868
+ "rewards/margins": 0.15380997955799103,
869
+ "rewards/rejected": -0.05395320802927017,
870
+ "step": 41
871
+ },
872
+ {
873
+ "debug/policy_chosen_logits": -3.1375789642333984,
874
+ "debug/policy_chosen_logps": -213.60757446289062,
875
+ "debug/policy_rejected_logits": -3.1117618083953857,
876
+ "debug/policy_rejected_logps": -256.28436279296875,
877
+ "debug/reference_chosen_logps": -226.36074829101562,
878
+ "debug/reference_rejected_logps": -242.12509155273438,
879
+ "epoch": 0.5,
880
+ "grad_norm": 4.775957985916952,
881
+ "learning_rate": 1e-06,
882
+ "logits/chosen": -3.1375789642333984,
883
+ "logits/rejected": -3.1117618083953857,
884
+ "logps/chosen": -213.60757446289062,
885
+ "logps/rejected": -256.28436279296875,
886
+ "loss": 0.3186,
887
+ "rewards/accuracies": 0.625,
888
+ "rewards/chosen": 0.1275317668914795,
889
+ "rewards/margins": 0.26912426948547363,
890
+ "rewards/rejected": -0.14159251749515533,
891
+ "step": 42
892
+ },
893
+ {
894
+ "debug/policy_chosen_logits": -3.1112849712371826,
895
+ "debug/policy_chosen_logps": -208.876220703125,
896
+ "debug/policy_rejected_logits": -3.0731041431427,
897
+ "debug/policy_rejected_logps": -213.07131958007812,
898
+ "debug/reference_chosen_logps": -226.13137817382812,
899
+ "debug/reference_rejected_logps": -202.86383056640625,
900
+ "epoch": 0.5119047619047619,
901
+ "grad_norm": 3.169314540293903,
902
+ "learning_rate": 1e-06,
903
+ "logits/chosen": -3.1112849712371826,
904
+ "logits/rejected": -3.0731041431427,
905
+ "logps/chosen": -208.876220703125,
906
+ "logps/rejected": -213.07131958007812,
907
+ "loss": 0.3156,
908
+ "rewards/accuracies": 0.75,
909
+ "rewards/chosen": 0.17255139350891113,
910
+ "rewards/margins": 0.2746262848377228,
911
+ "rewards/rejected": -0.10207486897706985,
912
+ "step": 43
913
+ },
914
+ {
915
+ "debug/policy_chosen_logits": -3.1239426136016846,
916
+ "debug/policy_chosen_logps": -201.20697021484375,
917
+ "debug/policy_rejected_logits": -3.087494373321533,
918
+ "debug/policy_rejected_logps": -259.4544372558594,
919
+ "debug/reference_chosen_logps": -221.05288696289062,
920
+ "debug/reference_rejected_logps": -240.31405639648438,
921
+ "epoch": 0.5238095238095238,
922
+ "grad_norm": 3.436483143809452,
923
+ "learning_rate": 1e-06,
924
+ "logits/chosen": -3.1239426136016846,
925
+ "logits/rejected": -3.087494373321533,
926
+ "logps/chosen": -201.20697021484375,
927
+ "logps/rejected": -259.4544372558594,
928
+ "loss": 0.3123,
929
+ "rewards/accuracies": 1.0,
930
+ "rewards/chosen": 0.19845911860466003,
931
+ "rewards/margins": 0.3898632526397705,
932
+ "rewards/rejected": -0.19140410423278809,
933
+ "step": 44
934
+ },
935
+ {
936
+ "debug/policy_chosen_logits": -3.138341188430786,
937
+ "debug/policy_chosen_logps": -175.82582092285156,
938
+ "debug/policy_rejected_logits": -2.981558084487915,
939
+ "debug/policy_rejected_logps": -231.8880615234375,
940
+ "debug/reference_chosen_logps": -196.50595092773438,
941
+ "debug/reference_rejected_logps": -216.26844787597656,
942
+ "epoch": 0.5357142857142857,
943
+ "grad_norm": 3.953231116742769,
944
+ "learning_rate": 1e-06,
945
+ "logits/chosen": -3.138341188430786,
946
+ "logits/rejected": -2.981558084487915,
947
+ "logps/chosen": -175.82582092285156,
948
+ "logps/rejected": -231.8880615234375,
949
+ "loss": 0.3425,
950
+ "rewards/accuracies": 0.875,
951
+ "rewards/chosen": 0.2068011611700058,
952
+ "rewards/margins": 0.3629972040653229,
953
+ "rewards/rejected": -0.15619604289531708,
954
+ "step": 45
955
+ },
956
+ {
957
+ "debug/policy_chosen_logits": -3.097372531890869,
958
+ "debug/policy_chosen_logps": -190.10743713378906,
959
+ "debug/policy_rejected_logits": -2.9607269763946533,
960
+ "debug/policy_rejected_logps": -255.23684692382812,
961
+ "debug/reference_chosen_logps": -207.03024291992188,
962
+ "debug/reference_rejected_logps": -224.69744873046875,
963
+ "epoch": 0.5476190476190477,
964
+ "grad_norm": 4.130780539545194,
965
+ "learning_rate": 1e-06,
966
+ "logits/chosen": -3.097372531890869,
967
+ "logits/rejected": -2.9607269763946533,
968
+ "logps/chosen": -190.10743713378906,
969
+ "logps/rejected": -255.23684692382812,
970
+ "loss": 0.334,
971
+ "rewards/accuracies": 0.875,
972
+ "rewards/chosen": 0.16922807693481445,
973
+ "rewards/margins": 0.4746219217777252,
974
+ "rewards/rejected": -0.30539384484291077,
975
+ "step": 46
976
+ },
977
+ {
978
+ "debug/policy_chosen_logits": -3.1764347553253174,
979
+ "debug/policy_chosen_logps": -204.2672119140625,
980
+ "debug/policy_rejected_logits": -3.070920467376709,
981
+ "debug/policy_rejected_logps": -236.72357177734375,
982
+ "debug/reference_chosen_logps": -222.78765869140625,
983
+ "debug/reference_rejected_logps": -211.966552734375,
984
+ "epoch": 0.5595238095238095,
985
+ "grad_norm": 2.9027881991083926,
986
+ "learning_rate": 1e-06,
987
+ "logits/chosen": -3.1764347553253174,
988
+ "logits/rejected": -3.070920467376709,
989
+ "logps/chosen": -204.2672119140625,
990
+ "logps/rejected": -236.72357177734375,
991
+ "loss": 0.3134,
992
+ "rewards/accuracies": 0.875,
993
+ "rewards/chosen": 0.18520456552505493,
994
+ "rewards/margins": 0.4327746033668518,
995
+ "rewards/rejected": -0.24757003784179688,
996
+ "step": 47
997
+ },
998
+ {
999
+ "debug/policy_chosen_logits": -3.0339467525482178,
1000
+ "debug/policy_chosen_logps": -214.1638641357422,
1001
+ "debug/policy_rejected_logits": -3.061377763748169,
1002
+ "debug/policy_rejected_logps": -234.35415649414062,
1003
+ "debug/reference_chosen_logps": -219.6826171875,
1004
+ "debug/reference_rejected_logps": -225.13951110839844,
1005
+ "epoch": 0.5714285714285714,
1006
+ "grad_norm": 3.492138091854823,
1007
+ "learning_rate": 1e-06,
1008
+ "logits/chosen": -3.0339467525482178,
1009
+ "logits/rejected": -3.061377763748169,
1010
+ "logps/chosen": -214.1638641357422,
1011
+ "logps/rejected": -234.35415649414062,
1012
+ "loss": 0.3227,
1013
+ "rewards/accuracies": 0.75,
1014
+ "rewards/chosen": 0.05518750846385956,
1015
+ "rewards/margins": 0.14733393490314484,
1016
+ "rewards/rejected": -0.09214641898870468,
1017
+ "step": 48
1018
+ },
1019
+ {
1020
+ "debug/policy_chosen_logits": -3.1581661701202393,
1021
+ "debug/policy_chosen_logps": -185.55007934570312,
1022
+ "debug/policy_rejected_logits": -3.0453264713287354,
1023
+ "debug/policy_rejected_logps": -262.7301025390625,
1024
+ "debug/reference_chosen_logps": -206.8365478515625,
1025
+ "debug/reference_rejected_logps": -242.17652893066406,
1026
+ "epoch": 0.5833333333333334,
1027
+ "grad_norm": 3.1731020410352615,
1028
+ "learning_rate": 1e-06,
1029
+ "logits/chosen": -3.1581661701202393,
1030
+ "logits/rejected": -3.0453264713287354,
1031
+ "logps/chosen": -185.55007934570312,
1032
+ "logps/rejected": -262.7301025390625,
1033
+ "loss": 0.3134,
1034
+ "rewards/accuracies": 0.875,
1035
+ "rewards/chosen": 0.21286450326442719,
1036
+ "rewards/margins": 0.4184000492095947,
1037
+ "rewards/rejected": -0.20553553104400635,
1038
+ "step": 49
1039
+ },
1040
+ {
1041
+ "debug/policy_chosen_logits": -3.206784725189209,
1042
+ "debug/policy_chosen_logps": -170.53402709960938,
1043
+ "debug/policy_rejected_logits": -2.9109795093536377,
1044
+ "debug/policy_rejected_logps": -258.2296142578125,
1045
+ "debug/reference_chosen_logps": -196.68124389648438,
1046
+ "debug/reference_rejected_logps": -230.39051818847656,
1047
+ "epoch": 0.5952380952380952,
1048
+ "grad_norm": 2.934981229097639,
1049
+ "learning_rate": 1e-06,
1050
+ "logits/chosen": -3.206784725189209,
1051
+ "logits/rejected": -2.9109795093536377,
1052
+ "logps/chosen": -170.53402709960938,
1053
+ "logps/rejected": -258.2296142578125,
1054
+ "loss": 0.2659,
1055
+ "rewards/accuracies": 1.0,
1056
+ "rewards/chosen": 0.26147204637527466,
1057
+ "rewards/margins": 0.5398629307746887,
1058
+ "rewards/rejected": -0.27839091420173645,
1059
+ "step": 50
1060
+ },
1061
+ {
1062
+ "debug/policy_chosen_logits": -3.0622358322143555,
1063
+ "debug/policy_chosen_logps": -200.63052368164062,
1064
+ "debug/policy_rejected_logits": -2.947862148284912,
1065
+ "debug/policy_rejected_logps": -252.6195068359375,
1066
+ "debug/reference_chosen_logps": -221.46884155273438,
1067
+ "debug/reference_rejected_logps": -228.0500030517578,
1068
+ "epoch": 0.6071428571428571,
1069
+ "grad_norm": 3.5442225737215387,
1070
+ "learning_rate": 1e-06,
1071
+ "logits/chosen": -3.0622358322143555,
1072
+ "logits/rejected": -2.947862148284912,
1073
+ "logps/chosen": -200.63052368164062,
1074
+ "logps/rejected": -252.6195068359375,
1075
+ "loss": 0.3124,
1076
+ "rewards/accuracies": 1.0,
1077
+ "rewards/chosen": 0.20838311314582825,
1078
+ "rewards/margins": 0.4540780782699585,
1079
+ "rewards/rejected": -0.24569493532180786,
1080
+ "step": 51
1081
+ },
1082
+ {
1083
+ "debug/policy_chosen_logits": -3.1601505279541016,
1084
+ "debug/policy_chosen_logps": -182.13150024414062,
1085
+ "debug/policy_rejected_logits": -2.988621711730957,
1086
+ "debug/policy_rejected_logps": -228.06826782226562,
1087
+ "debug/reference_chosen_logps": -201.15792846679688,
1088
+ "debug/reference_rejected_logps": -213.58132934570312,
1089
+ "epoch": 0.6190476190476191,
1090
+ "grad_norm": 4.536944665819507,
1091
+ "learning_rate": 1e-06,
1092
+ "logits/chosen": -3.1601505279541016,
1093
+ "logits/rejected": -2.988621711730957,
1094
+ "logps/chosen": -182.13150024414062,
1095
+ "logps/rejected": -228.06826782226562,
1096
+ "loss": 0.3305,
1097
+ "rewards/accuracies": 1.0,
1098
+ "rewards/chosen": 0.19026409089565277,
1099
+ "rewards/margins": 0.33513346314430237,
1100
+ "rewards/rejected": -0.1448693871498108,
1101
+ "step": 52
1102
+ },
1103
+ {
1104
+ "debug/policy_chosen_logits": -3.198798894882202,
1105
+ "debug/policy_chosen_logps": -189.36521911621094,
1106
+ "debug/policy_rejected_logits": -3.0948612689971924,
1107
+ "debug/policy_rejected_logps": -224.48573303222656,
1108
+ "debug/reference_chosen_logps": -206.97372436523438,
1109
+ "debug/reference_rejected_logps": -216.28659057617188,
1110
+ "epoch": 0.6309523809523809,
1111
+ "grad_norm": 3.1233076322027067,
1112
+ "learning_rate": 1e-06,
1113
+ "logits/chosen": -3.198798894882202,
1114
+ "logits/rejected": -3.0948612689971924,
1115
+ "logps/chosen": -189.36521911621094,
1116
+ "logps/rejected": -224.48573303222656,
1117
+ "loss": 0.2912,
1118
+ "rewards/accuracies": 0.625,
1119
+ "rewards/chosen": 0.17608505487442017,
1120
+ "rewards/margins": 0.25807660818099976,
1121
+ "rewards/rejected": -0.08199156075716019,
1122
+ "step": 53
1123
+ },
1124
+ {
1125
+ "debug/policy_chosen_logits": -3.0483901500701904,
1126
+ "debug/policy_chosen_logps": -220.47686767578125,
1127
+ "debug/policy_rejected_logits": -3.0346791744232178,
1128
+ "debug/policy_rejected_logps": -233.89801025390625,
1129
+ "debug/reference_chosen_logps": -234.3529052734375,
1130
+ "debug/reference_rejected_logps": -234.83953857421875,
1131
+ "epoch": 0.6428571428571429,
1132
+ "grad_norm": 4.209474032104615,
1133
+ "learning_rate": 1e-06,
1134
+ "logits/chosen": -3.0483901500701904,
1135
+ "logits/rejected": -3.0346791744232178,
1136
+ "logps/chosen": -220.47686767578125,
1137
+ "logps/rejected": -233.89801025390625,
1138
+ "loss": 0.3198,
1139
+ "rewards/accuracies": 0.875,
1140
+ "rewards/chosen": 0.1387602984905243,
1141
+ "rewards/margins": 0.12934501469135284,
1142
+ "rewards/rejected": 0.00941528007388115,
1143
+ "step": 54
1144
+ },
1145
+ {
1146
+ "debug/policy_chosen_logits": -3.171710968017578,
1147
+ "debug/policy_chosen_logps": -177.91751098632812,
1148
+ "debug/policy_rejected_logits": -3.0502405166625977,
1149
+ "debug/policy_rejected_logps": -286.4560852050781,
1150
+ "debug/reference_chosen_logps": -200.41464233398438,
1151
+ "debug/reference_rejected_logps": -257.74066162109375,
1152
+ "epoch": 0.6547619047619048,
1153
+ "grad_norm": 4.118989628914695,
1154
+ "learning_rate": 1e-06,
1155
+ "logits/chosen": -3.171710968017578,
1156
+ "logits/rejected": -3.0502405166625977,
1157
+ "logps/chosen": -177.91751098632812,
1158
+ "logps/rejected": -286.4560852050781,
1159
+ "loss": 0.2885,
1160
+ "rewards/accuracies": 1.0,
1161
+ "rewards/chosen": 0.22497142851352692,
1162
+ "rewards/margins": 0.5121252536773682,
1163
+ "rewards/rejected": -0.28715386986732483,
1164
+ "step": 55
1165
+ },
1166
+ {
1167
+ "debug/policy_chosen_logits": -3.09352445602417,
1168
+ "debug/policy_chosen_logps": -201.69338989257812,
1169
+ "debug/policy_rejected_logits": -2.858083486557007,
1170
+ "debug/policy_rejected_logps": -237.42831420898438,
1171
+ "debug/reference_chosen_logps": -221.5948944091797,
1172
+ "debug/reference_rejected_logps": -233.11854553222656,
1173
+ "epoch": 0.6666666666666666,
1174
+ "grad_norm": 5.4156894530313835,
1175
+ "learning_rate": 1e-06,
1176
+ "logits/chosen": -3.09352445602417,
1177
+ "logits/rejected": -2.858083486557007,
1178
+ "logps/chosen": -201.69338989257812,
1179
+ "logps/rejected": -237.42831420898438,
1180
+ "loss": 0.3487,
1181
+ "rewards/accuracies": 1.0,
1182
+ "rewards/chosen": 0.19901515543460846,
1183
+ "rewards/margins": 0.24211297929286957,
1184
+ "rewards/rejected": -0.04309781268239021,
1185
+ "step": 56
1186
+ },
1187
+ {
1188
+ "debug/policy_chosen_logits": -3.0208303928375244,
1189
+ "debug/policy_chosen_logps": -204.16375732421875,
1190
+ "debug/policy_rejected_logits": -2.9121055603027344,
1191
+ "debug/policy_rejected_logps": -266.00836181640625,
1192
+ "debug/reference_chosen_logps": -221.49114990234375,
1193
+ "debug/reference_rejected_logps": -238.82809448242188,
1194
+ "epoch": 0.6785714285714286,
1195
+ "grad_norm": 3.4258355626960837,
1196
+ "learning_rate": 1e-06,
1197
+ "logits/chosen": -3.0208303928375244,
1198
+ "logits/rejected": -2.9121055603027344,
1199
+ "logps/chosen": -204.16375732421875,
1200
+ "logps/rejected": -266.00836181640625,
1201
+ "loss": 0.3158,
1202
+ "rewards/accuracies": 0.875,
1203
+ "rewards/chosen": 0.17327386140823364,
1204
+ "rewards/margins": 0.44507646560668945,
1205
+ "rewards/rejected": -0.2718026041984558,
1206
+ "step": 57
1207
+ },
1208
+ {
1209
+ "debug/policy_chosen_logits": -3.2102866172790527,
1210
+ "debug/policy_chosen_logps": -218.61669921875,
1211
+ "debug/policy_rejected_logits": -3.0547966957092285,
1212
+ "debug/policy_rejected_logps": -265.02374267578125,
1213
+ "debug/reference_chosen_logps": -230.63229370117188,
1214
+ "debug/reference_rejected_logps": -222.15362548828125,
1215
+ "epoch": 0.6904761904761905,
1216
+ "grad_norm": 6.73796058848574,
1217
+ "learning_rate": 1e-06,
1218
+ "logits/chosen": -3.2102866172790527,
1219
+ "logits/rejected": -3.0547966957092285,
1220
+ "logps/chosen": -218.61669921875,
1221
+ "logps/rejected": -265.02374267578125,
1222
+ "loss": 0.2995,
1223
+ "rewards/accuracies": 0.875,
1224
+ "rewards/chosen": 0.12015601992607117,
1225
+ "rewards/margins": 0.5488572120666504,
1226
+ "rewards/rejected": -0.42870116233825684,
1227
+ "step": 58
1228
+ },
1229
+ {
1230
+ "debug/policy_chosen_logits": -3.1211562156677246,
1231
+ "debug/policy_chosen_logps": -186.32528686523438,
1232
+ "debug/policy_rejected_logits": -3.046443462371826,
1233
+ "debug/policy_rejected_logps": -230.16293334960938,
1234
+ "debug/reference_chosen_logps": -207.93145751953125,
1235
+ "debug/reference_rejected_logps": -210.10281372070312,
1236
+ "epoch": 0.7023809523809523,
1237
+ "grad_norm": 3.5427381092079506,
1238
+ "learning_rate": 1e-06,
1239
+ "logits/chosen": -3.1211562156677246,
1240
+ "logits/rejected": -3.046443462371826,
1241
+ "logps/chosen": -186.32528686523438,
1242
+ "logps/rejected": -230.16293334960938,
1243
+ "loss": 0.3172,
1244
+ "rewards/accuracies": 0.875,
1245
+ "rewards/chosen": 0.21606168150901794,
1246
+ "rewards/margins": 0.4166628122329712,
1247
+ "rewards/rejected": -0.20060113072395325,
1248
+ "step": 59
1249
+ },
1250
+ {
1251
+ "debug/policy_chosen_logits": -3.099186420440674,
1252
+ "debug/policy_chosen_logps": -179.74508666992188,
1253
+ "debug/policy_rejected_logits": -3.098525047302246,
1254
+ "debug/policy_rejected_logps": -218.37451171875,
1255
+ "debug/reference_chosen_logps": -204.20477294921875,
1256
+ "debug/reference_rejected_logps": -208.37628173828125,
1257
+ "epoch": 0.7142857142857143,
1258
+ "grad_norm": 3.822141253561726,
1259
+ "learning_rate": 1e-06,
1260
+ "logits/chosen": -3.099186420440674,
1261
+ "logits/rejected": -3.098525047302246,
1262
+ "logps/chosen": -179.74508666992188,
1263
+ "logps/rejected": -218.37451171875,
1264
+ "loss": 0.2797,
1265
+ "rewards/accuracies": 0.875,
1266
+ "rewards/chosen": 0.2445967197418213,
1267
+ "rewards/margins": 0.3445791006088257,
1268
+ "rewards/rejected": -0.0999823734164238,
1269
+ "step": 60
1270
+ },
1271
+ {
1272
+ "debug/policy_chosen_logits": -3.1759068965911865,
1273
+ "debug/policy_chosen_logps": -189.11534118652344,
1274
+ "debug/policy_rejected_logits": -3.130244255065918,
1275
+ "debug/policy_rejected_logps": -233.64544677734375,
1276
+ "debug/reference_chosen_logps": -214.55841064453125,
1277
+ "debug/reference_rejected_logps": -213.53517150878906,
1278
+ "epoch": 0.7261904761904762,
1279
+ "grad_norm": 3.7708296837344957,
1280
+ "learning_rate": 1e-06,
1281
+ "logits/chosen": -3.1759068965911865,
1282
+ "logits/rejected": -3.130244255065918,
1283
+ "logps/chosen": -189.11534118652344,
1284
+ "logps/rejected": -233.64544677734375,
1285
+ "loss": 0.2868,
1286
+ "rewards/accuracies": 1.0,
1287
+ "rewards/chosen": 0.25443071126937866,
1288
+ "rewards/margins": 0.45553332567214966,
1289
+ "rewards/rejected": -0.201102614402771,
1290
+ "step": 61
1291
+ },
1292
+ {
1293
+ "debug/policy_chosen_logits": -3.1119725704193115,
1294
+ "debug/policy_chosen_logps": -178.9134063720703,
1295
+ "debug/policy_rejected_logits": -3.0561046600341797,
1296
+ "debug/policy_rejected_logps": -235.5465545654297,
1297
+ "debug/reference_chosen_logps": -201.8980712890625,
1298
+ "debug/reference_rejected_logps": -204.71258544921875,
1299
+ "epoch": 0.7380952380952381,
1300
+ "grad_norm": 3.6489245962660175,
1301
+ "learning_rate": 1e-06,
1302
+ "logits/chosen": -3.1119725704193115,
1303
+ "logits/rejected": -3.0561046600341797,
1304
+ "logps/chosen": -178.9134063720703,
1305
+ "logps/rejected": -235.5465545654297,
1306
+ "loss": 0.2636,
1307
+ "rewards/accuracies": 1.0,
1308
+ "rewards/chosen": 0.22984656691551208,
1309
+ "rewards/margins": 0.5381861925125122,
1310
+ "rewards/rejected": -0.3083396553993225,
1311
+ "step": 62
1312
+ },
1313
+ {
1314
+ "debug/policy_chosen_logits": -3.1469156742095947,
1315
+ "debug/policy_chosen_logps": -201.1051025390625,
1316
+ "debug/policy_rejected_logits": -3.169464111328125,
1317
+ "debug/policy_rejected_logps": -232.58746337890625,
1318
+ "debug/reference_chosen_logps": -221.8560028076172,
1319
+ "debug/reference_rejected_logps": -219.1409149169922,
1320
+ "epoch": 0.75,
1321
+ "grad_norm": 4.929161871830666,
1322
+ "learning_rate": 1e-06,
1323
+ "logits/chosen": -3.1469156742095947,
1324
+ "logits/rejected": -3.169464111328125,
1325
+ "logps/chosen": -201.1051025390625,
1326
+ "logps/rejected": -232.58746337890625,
1327
+ "loss": 0.2553,
1328
+ "rewards/accuracies": 0.875,
1329
+ "rewards/chosen": 0.20750907063484192,
1330
+ "rewards/margins": 0.34197473526000977,
1331
+ "rewards/rejected": -0.13446564972400665,
1332
+ "step": 63
1333
+ },
1334
+ {
1335
+ "debug/policy_chosen_logits": -3.264002799987793,
1336
+ "debug/policy_chosen_logps": -176.3603515625,
1337
+ "debug/policy_rejected_logits": -3.2041733264923096,
1338
+ "debug/policy_rejected_logps": -237.74607849121094,
1339
+ "debug/reference_chosen_logps": -194.56317138671875,
1340
+ "debug/reference_rejected_logps": -220.99493408203125,
1341
+ "epoch": 0.7619047619047619,
1342
+ "grad_norm": 3.818382564446503,
1343
+ "learning_rate": 1e-06,
1344
+ "logits/chosen": -3.264002799987793,
1345
+ "logits/rejected": -3.2041733264923096,
1346
+ "logps/chosen": -176.3603515625,
1347
+ "logps/rejected": -237.74607849121094,
1348
+ "loss": 0.2587,
1349
+ "rewards/accuracies": 0.875,
1350
+ "rewards/chosen": 0.18202824890613556,
1351
+ "rewards/margins": 0.3495399057865143,
1352
+ "rewards/rejected": -0.16751165688037872,
1353
+ "step": 64
1354
+ },
1355
+ {
1356
+ "debug/policy_chosen_logits": -3.119412899017334,
1357
+ "debug/policy_chosen_logps": -203.42031860351562,
1358
+ "debug/policy_rejected_logits": -3.1157796382904053,
1359
+ "debug/policy_rejected_logps": -233.96710205078125,
1360
+ "debug/reference_chosen_logps": -222.92288208007812,
1361
+ "debug/reference_rejected_logps": -212.2803955078125,
1362
+ "epoch": 0.7738095238095238,
1363
+ "grad_norm": 3.2445663415245227,
1364
+ "learning_rate": 1e-06,
1365
+ "logits/chosen": -3.119412899017334,
1366
+ "logits/rejected": -3.1157796382904053,
1367
+ "logps/chosen": -203.42031860351562,
1368
+ "logps/rejected": -233.96710205078125,
1369
+ "loss": 0.2657,
1370
+ "rewards/accuracies": 1.0,
1371
+ "rewards/chosen": 0.19502560794353485,
1372
+ "rewards/margins": 0.41189298033714294,
1373
+ "rewards/rejected": -0.2168673723936081,
1374
+ "step": 65
1375
+ },
1376
+ {
1377
+ "debug/policy_chosen_logits": -3.1231942176818848,
1378
+ "debug/policy_chosen_logps": -187.53953552246094,
1379
+ "debug/policy_rejected_logits": -2.986668109893799,
1380
+ "debug/policy_rejected_logps": -240.6594696044922,
1381
+ "debug/reference_chosen_logps": -203.42880249023438,
1382
+ "debug/reference_rejected_logps": -214.08140563964844,
1383
+ "epoch": 0.7857142857142857,
1384
+ "grad_norm": 3.993146507021109,
1385
+ "learning_rate": 1e-06,
1386
+ "logits/chosen": -3.1231942176818848,
1387
+ "logits/rejected": -2.986668109893799,
1388
+ "logps/chosen": -187.53953552246094,
1389
+ "logps/rejected": -240.6594696044922,
1390
+ "loss": 0.2497,
1391
+ "rewards/accuracies": 0.875,
1392
+ "rewards/chosen": 0.15889252722263336,
1393
+ "rewards/margins": 0.4246731102466583,
1394
+ "rewards/rejected": -0.26578059792518616,
1395
+ "step": 66
1396
+ },
1397
+ {
1398
+ "debug/policy_chosen_logits": -3.1682167053222656,
1399
+ "debug/policy_chosen_logps": -205.82066345214844,
1400
+ "debug/policy_rejected_logits": -3.2192001342773438,
1401
+ "debug/policy_rejected_logps": -218.02267456054688,
1402
+ "debug/reference_chosen_logps": -218.94798278808594,
1403
+ "debug/reference_rejected_logps": -207.19180297851562,
1404
+ "epoch": 0.7976190476190477,
1405
+ "grad_norm": 4.631060327525655,
1406
+ "learning_rate": 1e-06,
1407
+ "logits/chosen": -3.1682167053222656,
1408
+ "logits/rejected": -3.2192001342773438,
1409
+ "logps/chosen": -205.82066345214844,
1410
+ "logps/rejected": -218.02267456054688,
1411
+ "loss": 0.256,
1412
+ "rewards/accuracies": 0.625,
1413
+ "rewards/chosen": 0.1312730759382248,
1414
+ "rewards/margins": 0.23958177864551544,
1415
+ "rewards/rejected": -0.10830870270729065,
1416
+ "step": 67
1417
+ },
1418
+ {
1419
+ "debug/policy_chosen_logits": -3.203059196472168,
1420
+ "debug/policy_chosen_logps": -178.87045288085938,
1421
+ "debug/policy_rejected_logits": -2.987100839614868,
1422
+ "debug/policy_rejected_logps": -232.19949340820312,
1423
+ "debug/reference_chosen_logps": -203.59388732910156,
1424
+ "debug/reference_rejected_logps": -210.4298553466797,
1425
+ "epoch": 0.8095238095238095,
1426
+ "grad_norm": 3.521359819189684,
1427
+ "learning_rate": 1e-06,
1428
+ "logits/chosen": -3.203059196472168,
1429
+ "logits/rejected": -2.987100839614868,
1430
+ "logps/chosen": -178.87045288085938,
1431
+ "logps/rejected": -232.19949340820312,
1432
+ "loss": 0.3033,
1433
+ "rewards/accuracies": 1.0,
1434
+ "rewards/chosen": 0.24723434448242188,
1435
+ "rewards/margins": 0.4649306535720825,
1436
+ "rewards/rejected": -0.21769630908966064,
1437
+ "step": 68
1438
+ },
1439
+ {
1440
+ "debug/policy_chosen_logits": -3.0755350589752197,
1441
+ "debug/policy_chosen_logps": -187.50741577148438,
1442
+ "debug/policy_rejected_logits": -3.0124213695526123,
1443
+ "debug/policy_rejected_logps": -249.99533081054688,
1444
+ "debug/reference_chosen_logps": -211.29078674316406,
1445
+ "debug/reference_rejected_logps": -215.66058349609375,
1446
+ "epoch": 0.8214285714285714,
1447
+ "grad_norm": 4.192433976871342,
1448
+ "learning_rate": 1e-06,
1449
+ "logits/chosen": -3.0755350589752197,
1450
+ "logits/rejected": -3.0124213695526123,
1451
+ "logps/chosen": -187.50741577148438,
1452
+ "logps/rejected": -249.99533081054688,
1453
+ "loss": 0.2695,
1454
+ "rewards/accuracies": 1.0,
1455
+ "rewards/chosen": 0.23783370852470398,
1456
+ "rewards/margins": 0.5811812281608582,
1457
+ "rewards/rejected": -0.3433475196361542,
1458
+ "step": 69
1459
+ },
1460
+ {
1461
+ "debug/policy_chosen_logits": -3.1850805282592773,
1462
+ "debug/policy_chosen_logps": -188.49639892578125,
1463
+ "debug/policy_rejected_logits": -3.084040880203247,
1464
+ "debug/policy_rejected_logps": -220.804931640625,
1465
+ "debug/reference_chosen_logps": -212.1209716796875,
1466
+ "debug/reference_rejected_logps": -211.52203369140625,
1467
+ "epoch": 0.8333333333333334,
1468
+ "grad_norm": 3.2739721314248973,
1469
+ "learning_rate": 1e-06,
1470
+ "logits/chosen": -3.1850805282592773,
1471
+ "logits/rejected": -3.084040880203247,
1472
+ "logps/chosen": -188.49639892578125,
1473
+ "logps/rejected": -220.804931640625,
1474
+ "loss": 0.267,
1475
+ "rewards/accuracies": 0.875,
1476
+ "rewards/chosen": 0.23624567687511444,
1477
+ "rewards/margins": 0.32907477021217346,
1478
+ "rewards/rejected": -0.09282909333705902,
1479
+ "step": 70
1480
+ },
1481
+ {
1482
+ "debug/policy_chosen_logits": -3.096564292907715,
1483
+ "debug/policy_chosen_logps": -207.81924438476562,
1484
+ "debug/policy_rejected_logits": -3.0242645740509033,
1485
+ "debug/policy_rejected_logps": -229.8334503173828,
1486
+ "debug/reference_chosen_logps": -227.1575164794922,
1487
+ "debug/reference_rejected_logps": -214.90184020996094,
1488
+ "epoch": 0.8452380952380952,
1489
+ "grad_norm": 3.0819588522165686,
1490
+ "learning_rate": 1e-06,
1491
+ "logits/chosen": -3.096564292907715,
1492
+ "logits/rejected": -3.0242645740509033,
1493
+ "logps/chosen": -207.81924438476562,
1494
+ "logps/rejected": -229.8334503173828,
1495
+ "loss": 0.2744,
1496
+ "rewards/accuracies": 0.75,
1497
+ "rewards/chosen": 0.1933828443288803,
1498
+ "rewards/margins": 0.34269896149635315,
1499
+ "rewards/rejected": -0.14931611716747284,
1500
+ "step": 71
1501
+ },
1502
+ {
1503
+ "debug/policy_chosen_logits": -3.1674182415008545,
1504
+ "debug/policy_chosen_logps": -192.69192504882812,
1505
+ "debug/policy_rejected_logits": -3.0628364086151123,
1506
+ "debug/policy_rejected_logps": -225.1174774169922,
1507
+ "debug/reference_chosen_logps": -215.3536376953125,
1508
+ "debug/reference_rejected_logps": -209.8455352783203,
1509
+ "epoch": 0.8571428571428571,
1510
+ "grad_norm": 3.773640551428593,
1511
+ "learning_rate": 1e-06,
1512
+ "logits/chosen": -3.1674182415008545,
1513
+ "logits/rejected": -3.0628364086151123,
1514
+ "logps/chosen": -192.69192504882812,
1515
+ "logps/rejected": -225.1174774169922,
1516
+ "loss": 0.2596,
1517
+ "rewards/accuracies": 0.75,
1518
+ "rewards/chosen": 0.2266169786453247,
1519
+ "rewards/margins": 0.37933632731437683,
1520
+ "rewards/rejected": -0.15271936357021332,
1521
+ "step": 72
1522
+ },
1523
+ {
1524
+ "debug/policy_chosen_logits": -3.077094793319702,
1525
+ "debug/policy_chosen_logps": -187.55560302734375,
1526
+ "debug/policy_rejected_logits": -3.0116915702819824,
1527
+ "debug/policy_rejected_logps": -248.090087890625,
1528
+ "debug/reference_chosen_logps": -208.9130096435547,
1529
+ "debug/reference_rejected_logps": -221.7311248779297,
1530
+ "epoch": 0.8690476190476191,
1531
+ "grad_norm": 3.2776627025025333,
1532
+ "learning_rate": 1e-06,
1533
+ "logits/chosen": -3.077094793319702,
1534
+ "logits/rejected": -3.0116915702819824,
1535
+ "logps/chosen": -187.55560302734375,
1536
+ "logps/rejected": -248.090087890625,
1537
+ "loss": 0.2596,
1538
+ "rewards/accuracies": 1.0,
1539
+ "rewards/chosen": 0.21357415616512299,
1540
+ "rewards/margins": 0.47716373205184937,
1541
+ "rewards/rejected": -0.2635895609855652,
1542
+ "step": 73
1543
+ },
1544
+ {
1545
+ "debug/policy_chosen_logits": -3.182755470275879,
1546
+ "debug/policy_chosen_logps": -205.19509887695312,
1547
+ "debug/policy_rejected_logits": -2.9780564308166504,
1548
+ "debug/policy_rejected_logps": -253.59747314453125,
1549
+ "debug/reference_chosen_logps": -220.27328491210938,
1550
+ "debug/reference_rejected_logps": -215.6810302734375,
1551
+ "epoch": 0.8809523809523809,
1552
+ "grad_norm": 3.5105144029616446,
1553
+ "learning_rate": 1e-06,
1554
+ "logits/chosen": -3.182755470275879,
1555
+ "logits/rejected": -2.9780564308166504,
1556
+ "logps/chosen": -205.19509887695312,
1557
+ "logps/rejected": -253.59747314453125,
1558
+ "loss": 0.3096,
1559
+ "rewards/accuracies": 0.875,
1560
+ "rewards/chosen": 0.1507818102836609,
1561
+ "rewards/margins": 0.5299463272094727,
1562
+ "rewards/rejected": -0.37916454672813416,
1563
+ "step": 74
1564
+ },
1565
+ {
1566
+ "debug/policy_chosen_logits": -3.0349855422973633,
1567
+ "debug/policy_chosen_logps": -197.20550537109375,
1568
+ "debug/policy_rejected_logits": -2.968355178833008,
1569
+ "debug/policy_rejected_logps": -260.9197998046875,
1570
+ "debug/reference_chosen_logps": -217.93182373046875,
1571
+ "debug/reference_rejected_logps": -250.27139282226562,
1572
+ "epoch": 0.8928571428571429,
1573
+ "grad_norm": 5.618897079839302,
1574
+ "learning_rate": 1e-06,
1575
+ "logits/chosen": -3.0349855422973633,
1576
+ "logits/rejected": -2.968355178833008,
1577
+ "logps/chosen": -197.20550537109375,
1578
+ "logps/rejected": -260.9197998046875,
1579
+ "loss": 0.3018,
1580
+ "rewards/accuracies": 0.75,
1581
+ "rewards/chosen": 0.20726338028907776,
1582
+ "rewards/margins": 0.3137475848197937,
1583
+ "rewards/rejected": -0.10648422688245773,
1584
+ "step": 75
1585
+ },
1586
+ {
1587
+ "debug/policy_chosen_logits": -3.1911251544952393,
1588
+ "debug/policy_chosen_logps": -184.67440795898438,
1589
+ "debug/policy_rejected_logits": -3.182987928390503,
1590
+ "debug/policy_rejected_logps": -210.40711975097656,
1591
+ "debug/reference_chosen_logps": -203.4073486328125,
1592
+ "debug/reference_rejected_logps": -204.89013671875,
1593
+ "epoch": 0.9047619047619048,
1594
+ "grad_norm": 3.696876592657335,
1595
+ "learning_rate": 1e-06,
1596
+ "logits/chosen": -3.1911251544952393,
1597
+ "logits/rejected": -3.182987928390503,
1598
+ "logps/chosen": -184.67440795898438,
1599
+ "logps/rejected": -210.40711975097656,
1600
+ "loss": 0.3278,
1601
+ "rewards/accuracies": 0.75,
1602
+ "rewards/chosen": 0.1873292624950409,
1603
+ "rewards/margins": 0.24249891936779022,
1604
+ "rewards/rejected": -0.05516962707042694,
1605
+ "step": 76
1606
+ },
1607
+ {
1608
+ "debug/policy_chosen_logits": -3.178257465362549,
1609
+ "debug/policy_chosen_logps": -187.74302673339844,
1610
+ "debug/policy_rejected_logits": -3.0253536701202393,
1611
+ "debug/policy_rejected_logps": -255.27618408203125,
1612
+ "debug/reference_chosen_logps": -204.26242065429688,
1613
+ "debug/reference_rejected_logps": -219.57601928710938,
1614
+ "epoch": 0.9166666666666666,
1615
+ "grad_norm": 6.676177800896766,
1616
+ "learning_rate": 1e-06,
1617
+ "logits/chosen": -3.178257465362549,
1618
+ "logits/rejected": -3.0253536701202393,
1619
+ "logps/chosen": -187.74302673339844,
1620
+ "logps/rejected": -255.27618408203125,
1621
+ "loss": 0.2808,
1622
+ "rewards/accuracies": 1.0,
1623
+ "rewards/chosen": 0.16519393026828766,
1624
+ "rewards/margins": 0.5221953988075256,
1625
+ "rewards/rejected": -0.35700148344039917,
1626
+ "step": 77
1627
+ },
1628
+ {
1629
+ "debug/policy_chosen_logits": -3.172184705734253,
1630
+ "debug/policy_chosen_logps": -197.2301788330078,
1631
+ "debug/policy_rejected_logits": -3.122185230255127,
1632
+ "debug/policy_rejected_logps": -227.55764770507812,
1633
+ "debug/reference_chosen_logps": -208.9789276123047,
1634
+ "debug/reference_rejected_logps": -220.4404296875,
1635
+ "epoch": 0.9285714285714286,
1636
+ "grad_norm": 3.121642412627353,
1637
+ "learning_rate": 1e-06,
1638
+ "logits/chosen": -3.172184705734253,
1639
+ "logits/rejected": -3.122185230255127,
1640
+ "logps/chosen": -197.2301788330078,
1641
+ "logps/rejected": -227.55764770507812,
1642
+ "loss": 0.2861,
1643
+ "rewards/accuracies": 0.75,
1644
+ "rewards/chosen": 0.11748749017715454,
1645
+ "rewards/margins": 0.18865975737571716,
1646
+ "rewards/rejected": -0.07117227464914322,
1647
+ "step": 78
1648
+ },
1649
+ {
1650
+ "debug/policy_chosen_logits": -3.149040937423706,
1651
+ "debug/policy_chosen_logps": -180.52720642089844,
1652
+ "debug/policy_rejected_logits": -3.1451313495635986,
1653
+ "debug/policy_rejected_logps": -242.54115295410156,
1654
+ "debug/reference_chosen_logps": -205.92019653320312,
1655
+ "debug/reference_rejected_logps": -219.41590881347656,
1656
+ "epoch": 0.9404761904761905,
1657
+ "grad_norm": 3.1134708691492845,
1658
+ "learning_rate": 1e-06,
1659
+ "logits/chosen": -3.149040937423706,
1660
+ "logits/rejected": -3.1451313495635986,
1661
+ "logps/chosen": -180.52720642089844,
1662
+ "logps/rejected": -242.54115295410156,
1663
+ "loss": 0.2408,
1664
+ "rewards/accuracies": 1.0,
1665
+ "rewards/chosen": 0.25392991304397583,
1666
+ "rewards/margins": 0.4851823151111603,
1667
+ "rewards/rejected": -0.23125241696834564,
1668
+ "step": 79
1669
+ },
1670
+ {
1671
+ "debug/policy_chosen_logits": -3.215009927749634,
1672
+ "debug/policy_chosen_logps": -161.9908447265625,
1673
+ "debug/policy_rejected_logits": -2.9049410820007324,
1674
+ "debug/policy_rejected_logps": -237.18609619140625,
1675
+ "debug/reference_chosen_logps": -198.1100311279297,
1676
+ "debug/reference_rejected_logps": -205.93499755859375,
1677
+ "epoch": 0.9523809523809523,
1678
+ "grad_norm": 6.27678156865478,
1679
+ "learning_rate": 1e-06,
1680
+ "logits/chosen": -3.215009927749634,
1681
+ "logits/rejected": -2.9049410820007324,
1682
+ "logps/chosen": -161.9908447265625,
1683
+ "logps/rejected": -237.18609619140625,
1684
+ "loss": 0.3087,
1685
+ "rewards/accuracies": 1.0,
1686
+ "rewards/chosen": 0.361191987991333,
1687
+ "rewards/margins": 0.6737030744552612,
1688
+ "rewards/rejected": -0.3125110864639282,
1689
+ "step": 80
1690
+ },
1691
+ {
1692
+ "debug/policy_chosen_logits": -3.0824358463287354,
1693
+ "debug/policy_chosen_logps": -185.10894775390625,
1694
+ "debug/policy_rejected_logits": -2.936793804168701,
1695
+ "debug/policy_rejected_logps": -268.28875732421875,
1696
+ "debug/reference_chosen_logps": -200.82766723632812,
1697
+ "debug/reference_rejected_logps": -233.51441955566406,
1698
+ "epoch": 0.9642857142857143,
1699
+ "grad_norm": 3.3522970674177937,
1700
+ "learning_rate": 1e-06,
1701
+ "logits/chosen": -3.0824358463287354,
1702
+ "logits/rejected": -2.936793804168701,
1703
+ "logps/chosen": -185.10894775390625,
1704
+ "logps/rejected": -268.28875732421875,
1705
+ "loss": 0.2521,
1706
+ "rewards/accuracies": 1.0,
1707
+ "rewards/chosen": 0.15718725323677063,
1708
+ "rewards/margins": 0.5049305558204651,
1709
+ "rewards/rejected": -0.34774333238601685,
1710
+ "step": 81
1711
+ },
1712
+ {
1713
+ "debug/policy_chosen_logits": -3.135251760482788,
1714
+ "debug/policy_chosen_logps": -170.79705810546875,
1715
+ "debug/policy_rejected_logits": -3.0056710243225098,
1716
+ "debug/policy_rejected_logps": -227.4779052734375,
1717
+ "debug/reference_chosen_logps": -200.08804321289062,
1718
+ "debug/reference_rejected_logps": -197.99270629882812,
1719
+ "epoch": 0.9761904761904762,
1720
+ "grad_norm": 3.284253852501879,
1721
+ "learning_rate": 1e-06,
1722
+ "logits/chosen": -3.135251760482788,
1723
+ "logits/rejected": -3.0056710243225098,
1724
+ "logps/chosen": -170.79705810546875,
1725
+ "logps/rejected": -227.4779052734375,
1726
+ "loss": 0.2355,
1727
+ "rewards/accuracies": 0.875,
1728
+ "rewards/chosen": 0.2929098308086395,
1729
+ "rewards/margins": 0.5877617597579956,
1730
+ "rewards/rejected": -0.2948519289493561,
1731
+ "step": 82
1732
+ },
1733
+ {
1734
+ "debug/policy_chosen_logits": -3.0462417602539062,
1735
+ "debug/policy_chosen_logps": -195.46688842773438,
1736
+ "debug/policy_rejected_logits": -2.9659104347229004,
1737
+ "debug/policy_rejected_logps": -253.25881958007812,
1738
+ "debug/reference_chosen_logps": -216.95620727539062,
1739
+ "debug/reference_rejected_logps": -213.9208984375,
1740
+ "epoch": 0.9880952380952381,
1741
+ "grad_norm": 3.3816029123776024,
1742
+ "learning_rate": 1e-06,
1743
+ "logits/chosen": -3.0462417602539062,
1744
+ "logits/rejected": -2.9659104347229004,
1745
+ "logps/chosen": -195.46688842773438,
1746
+ "logps/rejected": -253.25881958007812,
1747
+ "loss": 0.2725,
1748
+ "rewards/accuracies": 1.0,
1749
+ "rewards/chosen": 0.21489323675632477,
1750
+ "rewards/margins": 0.6082723140716553,
1751
+ "rewards/rejected": -0.3933790922164917,
1752
+ "step": 83
1753
+ },
1754
+ {
1755
+ "debug/policy_chosen_logits": -3.181661605834961,
1756
+ "debug/policy_chosen_logps": -201.0180206298828,
1757
+ "debug/policy_rejected_logits": -3.114567518234253,
1758
+ "debug/policy_rejected_logps": -248.83258056640625,
1759
+ "debug/reference_chosen_logps": -217.26596069335938,
1760
+ "debug/reference_rejected_logps": -230.24380493164062,
1761
+ "epoch": 1.0,
1762
+ "grad_norm": 3.6750878689595683,
1763
+ "learning_rate": 1e-06,
1764
+ "logits/chosen": -3.181661605834961,
1765
+ "logits/rejected": -3.114567518234253,
1766
+ "logps/chosen": -201.0180206298828,
1767
+ "logps/rejected": -248.83258056640625,
1768
+ "loss": 0.2642,
1769
+ "rewards/accuracies": 0.75,
1770
+ "rewards/chosen": 0.16247960925102234,
1771
+ "rewards/margins": 0.3483673632144928,
1772
+ "rewards/rejected": -0.18588775396347046,
1773
+ "step": 84
1774
+ },
1775
+ {
1776
+ "epoch": 1.0,
1777
+ "step": 84,
1778
+ "total_flos": 0.0,
1779
+ "train_loss": 0.33750108753641445,
1780
+ "train_runtime": 246.5194,
1781
+ "train_samples_per_second": 21.613,
1782
+ "train_steps_per_second": 0.341
1783
+ }
1784
+ ],
1785
+ "logging_steps": 1,
1786
+ "max_steps": 84,
1787
+ "num_input_tokens_seen": 0,
1788
+ "num_train_epochs": 1,
1789
+ "save_steps": 500,
1790
+ "stateful_callbacks": {
1791
+ "TrainerControl": {
1792
+ "args": {
1793
+ "should_epoch_stop": false,
1794
+ "should_evaluate": false,
1795
+ "should_log": false,
1796
+ "should_save": true,
1797
+ "should_training_stop": true
1798
+ },
1799
+ "attributes": {}
1800
+ }
1801
+ },
1802
+ "total_flos": 0.0,
1803
+ "train_batch_size": 8,
1804
+ "trial_name": null,
1805
+ "trial_params": null
1806
+ }