yiran-wang3 commited on
Commit
ca1dbc2
1 Parent(s): 431b502

End of training

Browse files
README.md ADDED
@@ -0,0 +1,64 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ library_name: transformers
3
+ license: other
4
+ base_model: deepseek-ai/deepseek-coder-7b-instruct-v1.5
5
+ tags:
6
+ - alignment-handbook
7
+ - generated_from_trainer
8
+ - trl
9
+ - dpo
10
+ datasets:
11
+ - self-generate/ds_coder_reflct_original_cn_mining_oj_iter0-binarized-reflection-scored
12
+ model-index:
13
+ - name: ds_coder_reflct_adamw_iter1
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
+ # ds_coder_reflct_adamw_iter1
21
+
22
+ This model is a fine-tuned version of [deepseek-ai/deepseek-coder-7b-instruct-v1.5](https://huggingface.co/deepseek-ai/deepseek-coder-7b-instruct-v1.5) on the self-generate/ds_coder_reflct_original_cn_mining_oj_iter0-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 0.20.3
all_results.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 1.0,
3
+ "total_flos": 0.0,
4
+ "train_loss": 0.32403188310563563,
5
+ "train_runtime": 674.2321,
6
+ "train_samples": 5119,
7
+ "train_samples_per_second": 7.592,
8
+ "train_steps_per_second": 0.119
9
+ }
config.json CHANGED
@@ -25,6 +25,6 @@
25
  "tie_word_embeddings": false,
26
  "torch_dtype": "bfloat16",
27
  "transformers_version": "4.45.0",
28
- "use_cache": false,
29
  "vocab_size": 102400
30
  }
 
25
  "tie_word_embeddings": false,
26
  "torch_dtype": "bfloat16",
27
  "transformers_version": "4.45.0",
28
+ "use_cache": true,
29
  "vocab_size": 102400
30
  }
generation_config.json ADDED
@@ -0,0 +1,6 @@
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 100000,
4
+ "eos_token_id": 100015,
5
+ "transformers_version": "4.45.0"
6
+ }
train_results.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 1.0,
3
+ "total_flos": 0.0,
4
+ "train_loss": 0.32403188310563563,
5
+ "train_runtime": 674.2321,
6
+ "train_samples": 5119,
7
+ "train_samples_per_second": 7.592,
8
+ "train_steps_per_second": 0.119
9
+ }
trainer_state.json ADDED
@@ -0,0 +1,1722 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 1.0,
5
+ "eval_steps": 100,
6
+ "global_step": 80,
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": 20.44719123840332,
13
+ "debug/policy_chosen_logps": -408.70501708984375,
14
+ "debug/policy_rejected_logits": 27.257293701171875,
15
+ "debug/policy_rejected_logps": -446.4876708984375,
16
+ "debug/reference_chosen_logps": -408.70501708984375,
17
+ "debug/reference_rejected_logps": -446.4876708984375,
18
+ "epoch": 0.0125,
19
+ "grad_norm": 7.8166135524175155,
20
+ "learning_rate": 1e-06,
21
+ "logits/chosen": 20.44719123840332,
22
+ "logits/rejected": 27.257293701171875,
23
+ "logps/chosen": -408.70501708984375,
24
+ "logps/rejected": -446.4876708984375,
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": 19.666156768798828,
34
+ "debug/policy_chosen_logps": -409.9366760253906,
35
+ "debug/policy_rejected_logits": 24.532733917236328,
36
+ "debug/policy_rejected_logps": -436.7401428222656,
37
+ "debug/reference_chosen_logps": -409.74560546875,
38
+ "debug/reference_rejected_logps": -435.607421875,
39
+ "epoch": 0.025,
40
+ "grad_norm": 7.95235642646064,
41
+ "learning_rate": 1e-06,
42
+ "logits/chosen": 19.666156768798828,
43
+ "logits/rejected": 24.532733917236328,
44
+ "logps/chosen": -409.9366760253906,
45
+ "logps/rejected": -436.7401428222656,
46
+ "loss": 0.4941,
47
+ "rewards/accuracies": 0.875,
48
+ "rewards/chosen": -0.0019105530809611082,
49
+ "rewards/margins": 0.009416657499969006,
50
+ "rewards/rejected": -0.011327208951115608,
51
+ "step": 2
52
+ },
53
+ {
54
+ "debug/policy_chosen_logits": 20.438295364379883,
55
+ "debug/policy_chosen_logps": -425.41778564453125,
56
+ "debug/policy_rejected_logits": 21.396846771240234,
57
+ "debug/policy_rejected_logps": -436.84344482421875,
58
+ "debug/reference_chosen_logps": -423.95123291015625,
59
+ "debug/reference_rejected_logps": -434.63232421875,
60
+ "epoch": 0.0375,
61
+ "grad_norm": 9.487672784484882,
62
+ "learning_rate": 1e-06,
63
+ "logits/chosen": 20.438295364379883,
64
+ "logits/rejected": 21.396846771240234,
65
+ "logps/chosen": -425.41778564453125,
66
+ "logps/rejected": -436.84344482421875,
67
+ "loss": 0.494,
68
+ "rewards/accuracies": 0.875,
69
+ "rewards/chosen": -0.014665374532341957,
70
+ "rewards/margins": 0.007445754483342171,
71
+ "rewards/rejected": -0.022111129015684128,
72
+ "step": 3
73
+ },
74
+ {
75
+ "debug/policy_chosen_logits": 18.89822769165039,
76
+ "debug/policy_chosen_logps": -431.1507568359375,
77
+ "debug/policy_rejected_logits": 22.7398624420166,
78
+ "debug/policy_rejected_logps": -449.3642578125,
79
+ "debug/reference_chosen_logps": -431.5155944824219,
80
+ "debug/reference_rejected_logps": -447.38824462890625,
81
+ "epoch": 0.05,
82
+ "grad_norm": 8.935635898408727,
83
+ "learning_rate": 1e-06,
84
+ "logits/chosen": 18.89822769165039,
85
+ "logits/rejected": 22.7398624420166,
86
+ "logps/chosen": -431.1507568359375,
87
+ "logps/rejected": -449.3642578125,
88
+ "loss": 0.4922,
89
+ "rewards/accuracies": 1.0,
90
+ "rewards/chosen": 0.0036485292948782444,
91
+ "rewards/margins": 0.023408813402056694,
92
+ "rewards/rejected": -0.019760284572839737,
93
+ "step": 4
94
+ },
95
+ {
96
+ "debug/policy_chosen_logits": 19.336538314819336,
97
+ "debug/policy_chosen_logps": -420.7198486328125,
98
+ "debug/policy_rejected_logits": 28.9335880279541,
99
+ "debug/policy_rejected_logps": -456.2585144042969,
100
+ "debug/reference_chosen_logps": -420.73626708984375,
101
+ "debug/reference_rejected_logps": -454.71124267578125,
102
+ "epoch": 0.0625,
103
+ "grad_norm": 9.253028260363507,
104
+ "learning_rate": 1e-06,
105
+ "logits/chosen": 19.336538314819336,
106
+ "logits/rejected": 28.9335880279541,
107
+ "logps/chosen": -420.7198486328125,
108
+ "logps/rejected": -456.2585144042969,
109
+ "loss": 0.4849,
110
+ "rewards/accuracies": 0.875,
111
+ "rewards/chosen": 0.00016406993381679058,
112
+ "rewards/margins": 0.015636596828699112,
113
+ "rewards/rejected": -0.015472527593374252,
114
+ "step": 5
115
+ },
116
+ {
117
+ "debug/policy_chosen_logits": 22.817670822143555,
118
+ "debug/policy_chosen_logps": -433.24945068359375,
119
+ "debug/policy_rejected_logits": 23.68866729736328,
120
+ "debug/policy_rejected_logps": -431.10113525390625,
121
+ "debug/reference_chosen_logps": -432.390869140625,
122
+ "debug/reference_rejected_logps": -430.1065368652344,
123
+ "epoch": 0.075,
124
+ "grad_norm": 8.448473864866386,
125
+ "learning_rate": 1e-06,
126
+ "logits/chosen": 22.817670822143555,
127
+ "logits/rejected": 23.68866729736328,
128
+ "logps/chosen": -433.24945068359375,
129
+ "logps/rejected": -431.10113525390625,
130
+ "loss": 0.4789,
131
+ "rewards/accuracies": 0.5,
132
+ "rewards/chosen": -0.008586043491959572,
133
+ "rewards/margins": 0.0013600160600617528,
134
+ "rewards/rejected": -0.009946059435606003,
135
+ "step": 6
136
+ },
137
+ {
138
+ "debug/policy_chosen_logits": 20.37796401977539,
139
+ "debug/policy_chosen_logps": -423.56182861328125,
140
+ "debug/policy_rejected_logits": 27.41672134399414,
141
+ "debug/policy_rejected_logps": -450.90185546875,
142
+ "debug/reference_chosen_logps": -423.9954833984375,
143
+ "debug/reference_rejected_logps": -446.689208984375,
144
+ "epoch": 0.0875,
145
+ "grad_norm": 8.300945379199838,
146
+ "learning_rate": 1e-06,
147
+ "logits/chosen": 20.37796401977539,
148
+ "logits/rejected": 27.41672134399414,
149
+ "logps/chosen": -423.56182861328125,
150
+ "logps/rejected": -450.90185546875,
151
+ "loss": 0.4751,
152
+ "rewards/accuracies": 0.875,
153
+ "rewards/chosen": 0.004336662590503693,
154
+ "rewards/margins": 0.0464630126953125,
155
+ "rewards/rejected": -0.04212634637951851,
156
+ "step": 7
157
+ },
158
+ {
159
+ "debug/policy_chosen_logits": 18.481666564941406,
160
+ "debug/policy_chosen_logps": -427.36419677734375,
161
+ "debug/policy_rejected_logits": 29.171558380126953,
162
+ "debug/policy_rejected_logps": -466.4520263671875,
163
+ "debug/reference_chosen_logps": -429.41650390625,
164
+ "debug/reference_rejected_logps": -462.8001708984375,
165
+ "epoch": 0.1,
166
+ "grad_norm": 11.20281145898035,
167
+ "learning_rate": 1e-06,
168
+ "logits/chosen": 18.481666564941406,
169
+ "logits/rejected": 29.171558380126953,
170
+ "logps/chosen": -427.36419677734375,
171
+ "logps/rejected": -466.4520263671875,
172
+ "loss": 0.4584,
173
+ "rewards/accuracies": 1.0,
174
+ "rewards/chosen": 0.02052280306816101,
175
+ "rewards/margins": 0.05704108998179436,
176
+ "rewards/rejected": -0.03651828691363335,
177
+ "step": 8
178
+ },
179
+ {
180
+ "debug/policy_chosen_logits": 22.923797607421875,
181
+ "debug/policy_chosen_logps": -439.3218688964844,
182
+ "debug/policy_rejected_logits": 27.558902740478516,
183
+ "debug/policy_rejected_logps": -443.62939453125,
184
+ "debug/reference_chosen_logps": -441.331298828125,
185
+ "debug/reference_rejected_logps": -439.750244140625,
186
+ "epoch": 0.1125,
187
+ "grad_norm": 8.404029783243205,
188
+ "learning_rate": 1e-06,
189
+ "logits/chosen": 22.923797607421875,
190
+ "logits/rejected": 27.558902740478516,
191
+ "logps/chosen": -439.3218688964844,
192
+ "logps/rejected": -443.62939453125,
193
+ "loss": 0.4508,
194
+ "rewards/accuracies": 0.875,
195
+ "rewards/chosen": 0.020094260573387146,
196
+ "rewards/margins": 0.05888580530881882,
197
+ "rewards/rejected": -0.03879154101014137,
198
+ "step": 9
199
+ },
200
+ {
201
+ "debug/policy_chosen_logits": 18.430870056152344,
202
+ "debug/policy_chosen_logps": -405.11029052734375,
203
+ "debug/policy_rejected_logits": 24.46626091003418,
204
+ "debug/policy_rejected_logps": -480.96319580078125,
205
+ "debug/reference_chosen_logps": -408.2779235839844,
206
+ "debug/reference_rejected_logps": -473.3382263183594,
207
+ "epoch": 0.125,
208
+ "grad_norm": 7.719942039636563,
209
+ "learning_rate": 1e-06,
210
+ "logits/chosen": 18.430870056152344,
211
+ "logits/rejected": 24.46626091003418,
212
+ "logps/chosen": -405.11029052734375,
213
+ "logps/rejected": -480.96319580078125,
214
+ "loss": 0.421,
215
+ "rewards/accuracies": 1.0,
216
+ "rewards/chosen": 0.03167613595724106,
217
+ "rewards/margins": 0.10792595148086548,
218
+ "rewards/rejected": -0.07624980807304382,
219
+ "step": 10
220
+ },
221
+ {
222
+ "debug/policy_chosen_logits": 18.422691345214844,
223
+ "debug/policy_chosen_logps": -430.35479736328125,
224
+ "debug/policy_rejected_logits": 19.293363571166992,
225
+ "debug/policy_rejected_logps": -440.55755615234375,
226
+ "debug/reference_chosen_logps": -431.34063720703125,
227
+ "debug/reference_rejected_logps": -434.83477783203125,
228
+ "epoch": 0.1375,
229
+ "grad_norm": 7.022895536807182,
230
+ "learning_rate": 1e-06,
231
+ "logits/chosen": 18.422691345214844,
232
+ "logits/rejected": 19.293363571166992,
233
+ "logps/chosen": -430.35479736328125,
234
+ "logps/rejected": -440.55755615234375,
235
+ "loss": 0.4078,
236
+ "rewards/accuracies": 0.875,
237
+ "rewards/chosen": 0.009858207777142525,
238
+ "rewards/margins": 0.06708598881959915,
239
+ "rewards/rejected": -0.05722777917981148,
240
+ "step": 11
241
+ },
242
+ {
243
+ "debug/policy_chosen_logits": 19.392061233520508,
244
+ "debug/policy_chosen_logps": -428.3876037597656,
245
+ "debug/policy_rejected_logits": 27.25312042236328,
246
+ "debug/policy_rejected_logps": -461.404296875,
247
+ "debug/reference_chosen_logps": -431.02984619140625,
248
+ "debug/reference_rejected_logps": -451.54437255859375,
249
+ "epoch": 0.15,
250
+ "grad_norm": 6.6130326271590585,
251
+ "learning_rate": 1e-06,
252
+ "logits/chosen": 19.392061233520508,
253
+ "logits/rejected": 27.25312042236328,
254
+ "logps/chosen": -428.3876037597656,
255
+ "logps/rejected": -461.404296875,
256
+ "loss": 0.416,
257
+ "rewards/accuracies": 0.875,
258
+ "rewards/chosen": 0.026422196999192238,
259
+ "rewards/margins": 0.12502101063728333,
260
+ "rewards/rejected": -0.09859882295131683,
261
+ "step": 12
262
+ },
263
+ {
264
+ "debug/policy_chosen_logits": 20.876129150390625,
265
+ "debug/policy_chosen_logps": -419.8938293457031,
266
+ "debug/policy_rejected_logits": 24.144548416137695,
267
+ "debug/policy_rejected_logps": -437.48370361328125,
268
+ "debug/reference_chosen_logps": -426.415771484375,
269
+ "debug/reference_rejected_logps": -429.47802734375,
270
+ "epoch": 0.1625,
271
+ "grad_norm": 7.606735140301725,
272
+ "learning_rate": 1e-06,
273
+ "logits/chosen": 20.876129150390625,
274
+ "logits/rejected": 24.144548416137695,
275
+ "logps/chosen": -419.8938293457031,
276
+ "logps/rejected": -437.48370361328125,
277
+ "loss": 0.3942,
278
+ "rewards/accuracies": 0.875,
279
+ "rewards/chosen": 0.0652196854352951,
280
+ "rewards/margins": 0.1452765166759491,
281
+ "rewards/rejected": -0.08005683869123459,
282
+ "step": 13
283
+ },
284
+ {
285
+ "debug/policy_chosen_logits": 17.269620895385742,
286
+ "debug/policy_chosen_logps": -398.87115478515625,
287
+ "debug/policy_rejected_logits": 25.764076232910156,
288
+ "debug/policy_rejected_logps": -460.26678466796875,
289
+ "debug/reference_chosen_logps": -408.37481689453125,
290
+ "debug/reference_rejected_logps": -455.5852966308594,
291
+ "epoch": 0.175,
292
+ "grad_norm": 5.932839484754313,
293
+ "learning_rate": 1e-06,
294
+ "logits/chosen": 17.269620895385742,
295
+ "logits/rejected": 25.764076232910156,
296
+ "logps/chosen": -398.87115478515625,
297
+ "logps/rejected": -460.26678466796875,
298
+ "loss": 0.4258,
299
+ "rewards/accuracies": 0.875,
300
+ "rewards/chosen": 0.095036581158638,
301
+ "rewards/margins": 0.1418514996767044,
302
+ "rewards/rejected": -0.046814918518066406,
303
+ "step": 14
304
+ },
305
+ {
306
+ "debug/policy_chosen_logits": 18.60376739501953,
307
+ "debug/policy_chosen_logps": -411.95623779296875,
308
+ "debug/policy_rejected_logits": 23.55546760559082,
309
+ "debug/policy_rejected_logps": -423.6629638671875,
310
+ "debug/reference_chosen_logps": -427.5350341796875,
311
+ "debug/reference_rejected_logps": -418.62518310546875,
312
+ "epoch": 0.1875,
313
+ "grad_norm": 5.826719160562423,
314
+ "learning_rate": 1e-06,
315
+ "logits/chosen": 18.60376739501953,
316
+ "logits/rejected": 23.55546760559082,
317
+ "logps/chosen": -411.95623779296875,
318
+ "logps/rejected": -423.6629638671875,
319
+ "loss": 0.4082,
320
+ "rewards/accuracies": 1.0,
321
+ "rewards/chosen": 0.15578752756118774,
322
+ "rewards/margins": 0.20616531372070312,
323
+ "rewards/rejected": -0.05037776753306389,
324
+ "step": 15
325
+ },
326
+ {
327
+ "debug/policy_chosen_logits": 19.708065032958984,
328
+ "debug/policy_chosen_logps": -421.86419677734375,
329
+ "debug/policy_rejected_logits": 23.621994018554688,
330
+ "debug/policy_rejected_logps": -431.23486328125,
331
+ "debug/reference_chosen_logps": -426.4266662597656,
332
+ "debug/reference_rejected_logps": -436.3722229003906,
333
+ "epoch": 0.2,
334
+ "grad_norm": 6.112736963468344,
335
+ "learning_rate": 1e-06,
336
+ "logits/chosen": 19.708065032958984,
337
+ "logits/rejected": 23.621994018554688,
338
+ "logps/chosen": -421.86419677734375,
339
+ "logps/rejected": -431.23486328125,
340
+ "loss": 0.3901,
341
+ "rewards/accuracies": 0.5,
342
+ "rewards/chosen": 0.04562458395957947,
343
+ "rewards/margins": -0.005748823285102844,
344
+ "rewards/rejected": 0.051373403519392014,
345
+ "step": 16
346
+ },
347
+ {
348
+ "debug/policy_chosen_logits": 20.457584381103516,
349
+ "debug/policy_chosen_logps": -421.99639892578125,
350
+ "debug/policy_rejected_logits": 27.524091720581055,
351
+ "debug/policy_rejected_logps": -444.53289794921875,
352
+ "debug/reference_chosen_logps": -430.9820556640625,
353
+ "debug/reference_rejected_logps": -434.97113037109375,
354
+ "epoch": 0.2125,
355
+ "grad_norm": 5.201415641144627,
356
+ "learning_rate": 1e-06,
357
+ "logits/chosen": 20.457584381103516,
358
+ "logits/rejected": 27.524091720581055,
359
+ "logps/chosen": -421.99639892578125,
360
+ "logps/rejected": -444.53289794921875,
361
+ "loss": 0.3627,
362
+ "rewards/accuracies": 0.875,
363
+ "rewards/chosen": 0.08985641598701477,
364
+ "rewards/margins": 0.18547385931015015,
365
+ "rewards/rejected": -0.09561744332313538,
366
+ "step": 17
367
+ },
368
+ {
369
+ "debug/policy_chosen_logits": 19.323848724365234,
370
+ "debug/policy_chosen_logps": -410.0373840332031,
371
+ "debug/policy_rejected_logits": 30.60475730895996,
372
+ "debug/policy_rejected_logps": -429.958740234375,
373
+ "debug/reference_chosen_logps": -416.14520263671875,
374
+ "debug/reference_rejected_logps": -423.42071533203125,
375
+ "epoch": 0.225,
376
+ "grad_norm": 5.252127969055532,
377
+ "learning_rate": 1e-06,
378
+ "logits/chosen": 19.323848724365234,
379
+ "logits/rejected": 30.60475730895996,
380
+ "logps/chosen": -410.0373840332031,
381
+ "logps/rejected": -429.958740234375,
382
+ "loss": 0.3815,
383
+ "rewards/accuracies": 0.625,
384
+ "rewards/chosen": 0.061078451573848724,
385
+ "rewards/margins": 0.1264585554599762,
386
+ "rewards/rejected": -0.06538009643554688,
387
+ "step": 18
388
+ },
389
+ {
390
+ "debug/policy_chosen_logits": 16.97679328918457,
391
+ "debug/policy_chosen_logps": -392.15478515625,
392
+ "debug/policy_rejected_logits": 27.920000076293945,
393
+ "debug/policy_rejected_logps": -447.72747802734375,
394
+ "debug/reference_chosen_logps": -410.5232238769531,
395
+ "debug/reference_rejected_logps": -432.1972961425781,
396
+ "epoch": 0.2375,
397
+ "grad_norm": 4.748490487188559,
398
+ "learning_rate": 1e-06,
399
+ "logits/chosen": 16.97679328918457,
400
+ "logits/rejected": 27.920000076293945,
401
+ "logps/chosen": -392.15478515625,
402
+ "logps/rejected": -447.72747802734375,
403
+ "loss": 0.338,
404
+ "rewards/accuracies": 1.0,
405
+ "rewards/chosen": 0.1836845427751541,
406
+ "rewards/margins": 0.33898642659187317,
407
+ "rewards/rejected": -0.15530189871788025,
408
+ "step": 19
409
+ },
410
+ {
411
+ "debug/policy_chosen_logits": 21.503040313720703,
412
+ "debug/policy_chosen_logps": -425.363037109375,
413
+ "debug/policy_rejected_logits": 31.72104263305664,
414
+ "debug/policy_rejected_logps": -472.585693359375,
415
+ "debug/reference_chosen_logps": -439.15594482421875,
416
+ "debug/reference_rejected_logps": -448.0031433105469,
417
+ "epoch": 0.25,
418
+ "grad_norm": 4.598515072265396,
419
+ "learning_rate": 1e-06,
420
+ "logits/chosen": 21.503040313720703,
421
+ "logits/rejected": 31.72104263305664,
422
+ "logps/chosen": -425.363037109375,
423
+ "logps/rejected": -472.585693359375,
424
+ "loss": 0.2996,
425
+ "rewards/accuracies": 1.0,
426
+ "rewards/chosen": 0.1379288136959076,
427
+ "rewards/margins": 0.3837544322013855,
428
+ "rewards/rejected": -0.2458256185054779,
429
+ "step": 20
430
+ },
431
+ {
432
+ "debug/policy_chosen_logits": 19.64223289489746,
433
+ "debug/policy_chosen_logps": -411.8244323730469,
434
+ "debug/policy_rejected_logits": 25.16058349609375,
435
+ "debug/policy_rejected_logps": -479.46990966796875,
436
+ "debug/reference_chosen_logps": -419.84149169921875,
437
+ "debug/reference_rejected_logps": -448.9364013671875,
438
+ "epoch": 0.2625,
439
+ "grad_norm": 4.506314070543816,
440
+ "learning_rate": 1e-06,
441
+ "logits/chosen": 19.64223289489746,
442
+ "logits/rejected": 25.16058349609375,
443
+ "logps/chosen": -411.8244323730469,
444
+ "logps/rejected": -479.46990966796875,
445
+ "loss": 0.3326,
446
+ "rewards/accuracies": 0.875,
447
+ "rewards/chosen": 0.08017054945230484,
448
+ "rewards/margins": 0.38550513982772827,
449
+ "rewards/rejected": -0.30533456802368164,
450
+ "step": 21
451
+ },
452
+ {
453
+ "debug/policy_chosen_logits": 20.455036163330078,
454
+ "debug/policy_chosen_logps": -401.1058044433594,
455
+ "debug/policy_rejected_logits": 27.924205780029297,
456
+ "debug/policy_rejected_logps": -479.4624938964844,
457
+ "debug/reference_chosen_logps": -415.33001708984375,
458
+ "debug/reference_rejected_logps": -445.6070861816406,
459
+ "epoch": 0.275,
460
+ "grad_norm": 8.086589389208452,
461
+ "learning_rate": 1e-06,
462
+ "logits/chosen": 20.455036163330078,
463
+ "logits/rejected": 27.924205780029297,
464
+ "logps/chosen": -401.1058044433594,
465
+ "logps/rejected": -479.4624938964844,
466
+ "loss": 0.284,
467
+ "rewards/accuracies": 0.875,
468
+ "rewards/chosen": 0.14224204421043396,
469
+ "rewards/margins": 0.4807964563369751,
470
+ "rewards/rejected": -0.3385544419288635,
471
+ "step": 22
472
+ },
473
+ {
474
+ "debug/policy_chosen_logits": 22.28139305114746,
475
+ "debug/policy_chosen_logps": -428.8080139160156,
476
+ "debug/policy_rejected_logits": 29.031978607177734,
477
+ "debug/policy_rejected_logps": -496.9381103515625,
478
+ "debug/reference_chosen_logps": -425.82586669921875,
479
+ "debug/reference_rejected_logps": -460.11199951171875,
480
+ "epoch": 0.2875,
481
+ "grad_norm": 7.117765571472066,
482
+ "learning_rate": 1e-06,
483
+ "logits/chosen": 22.28139305114746,
484
+ "logits/rejected": 29.031978607177734,
485
+ "logps/chosen": -428.8080139160156,
486
+ "logps/rejected": -496.9381103515625,
487
+ "loss": 0.3124,
488
+ "rewards/accuracies": 0.875,
489
+ "rewards/chosen": -0.02982150949537754,
490
+ "rewards/margins": 0.33843994140625,
491
+ "rewards/rejected": -0.3682614266872406,
492
+ "step": 23
493
+ },
494
+ {
495
+ "debug/policy_chosen_logits": 16.489967346191406,
496
+ "debug/policy_chosen_logps": -386.43011474609375,
497
+ "debug/policy_rejected_logits": 27.82581329345703,
498
+ "debug/policy_rejected_logps": -452.87548828125,
499
+ "debug/reference_chosen_logps": -407.1941223144531,
500
+ "debug/reference_rejected_logps": -445.4601745605469,
501
+ "epoch": 0.3,
502
+ "grad_norm": 5.311518244442526,
503
+ "learning_rate": 1e-06,
504
+ "logits/chosen": 16.489967346191406,
505
+ "logits/rejected": 27.82581329345703,
506
+ "logps/chosen": -386.43011474609375,
507
+ "logps/rejected": -452.87548828125,
508
+ "loss": 0.3473,
509
+ "rewards/accuracies": 1.0,
510
+ "rewards/chosen": 0.20764002203941345,
511
+ "rewards/margins": 0.2817932963371277,
512
+ "rewards/rejected": -0.07415324449539185,
513
+ "step": 24
514
+ },
515
+ {
516
+ "debug/policy_chosen_logits": 20.677587509155273,
517
+ "debug/policy_chosen_logps": -401.9738464355469,
518
+ "debug/policy_rejected_logits": 26.294002532958984,
519
+ "debug/policy_rejected_logps": -462.0292663574219,
520
+ "debug/reference_chosen_logps": -419.8895568847656,
521
+ "debug/reference_rejected_logps": -468.2488708496094,
522
+ "epoch": 0.3125,
523
+ "grad_norm": 6.577603302459432,
524
+ "learning_rate": 1e-06,
525
+ "logits/chosen": 20.677587509155273,
526
+ "logits/rejected": 26.294002532958984,
527
+ "logps/chosen": -401.9738464355469,
528
+ "logps/rejected": -462.0292663574219,
529
+ "loss": 0.3873,
530
+ "rewards/accuracies": 0.625,
531
+ "rewards/chosen": 0.1791572868824005,
532
+ "rewards/margins": 0.11696115881204605,
533
+ "rewards/rejected": 0.06219611316919327,
534
+ "step": 25
535
+ },
536
+ {
537
+ "debug/policy_chosen_logits": 21.27174186706543,
538
+ "debug/policy_chosen_logps": -403.5057678222656,
539
+ "debug/policy_rejected_logits": 28.13071060180664,
540
+ "debug/policy_rejected_logps": -465.28216552734375,
541
+ "debug/reference_chosen_logps": -426.16949462890625,
542
+ "debug/reference_rejected_logps": -428.91204833984375,
543
+ "epoch": 0.325,
544
+ "grad_norm": 9.280568242676463,
545
+ "learning_rate": 1e-06,
546
+ "logits/chosen": 21.27174186706543,
547
+ "logits/rejected": 28.13071060180664,
548
+ "logps/chosen": -403.5057678222656,
549
+ "logps/rejected": -465.28216552734375,
550
+ "loss": 0.3524,
551
+ "rewards/accuracies": 1.0,
552
+ "rewards/chosen": 0.2266373336315155,
553
+ "rewards/margins": 0.5903383493423462,
554
+ "rewards/rejected": -0.3637010157108307,
555
+ "step": 26
556
+ },
557
+ {
558
+ "debug/policy_chosen_logits": 19.095767974853516,
559
+ "debug/policy_chosen_logps": -382.9574890136719,
560
+ "debug/policy_rejected_logits": 31.333406448364258,
561
+ "debug/policy_rejected_logps": -468.0421447753906,
562
+ "debug/reference_chosen_logps": -405.85858154296875,
563
+ "debug/reference_rejected_logps": -444.3897705078125,
564
+ "epoch": 0.3375,
565
+ "grad_norm": 4.305140313967255,
566
+ "learning_rate": 1e-06,
567
+ "logits/chosen": 19.095767974853516,
568
+ "logits/rejected": 31.333406448364258,
569
+ "logps/chosen": -382.9574890136719,
570
+ "logps/rejected": -468.0421447753906,
571
+ "loss": 0.3094,
572
+ "rewards/accuracies": 0.75,
573
+ "rewards/chosen": 0.2290109544992447,
574
+ "rewards/margins": 0.46553459763526917,
575
+ "rewards/rejected": -0.23652365803718567,
576
+ "step": 27
577
+ },
578
+ {
579
+ "debug/policy_chosen_logits": 20.944793701171875,
580
+ "debug/policy_chosen_logps": -403.42010498046875,
581
+ "debug/policy_rejected_logits": 25.954708099365234,
582
+ "debug/policy_rejected_logps": -493.7495422363281,
583
+ "debug/reference_chosen_logps": -419.398681640625,
584
+ "debug/reference_rejected_logps": -464.70562744140625,
585
+ "epoch": 0.35,
586
+ "grad_norm": 5.091666773488247,
587
+ "learning_rate": 1e-06,
588
+ "logits/chosen": 20.944793701171875,
589
+ "logits/rejected": 25.954708099365234,
590
+ "logps/chosen": -403.42010498046875,
591
+ "logps/rejected": -493.7495422363281,
592
+ "loss": 0.3118,
593
+ "rewards/accuracies": 0.875,
594
+ "rewards/chosen": 0.15978576242923737,
595
+ "rewards/margins": 0.45022472739219666,
596
+ "rewards/rejected": -0.29043900966644287,
597
+ "step": 28
598
+ },
599
+ {
600
+ "debug/policy_chosen_logits": 18.986360549926758,
601
+ "debug/policy_chosen_logps": -396.65667724609375,
602
+ "debug/policy_rejected_logits": 26.656171798706055,
603
+ "debug/policy_rejected_logps": -464.8284912109375,
604
+ "debug/reference_chosen_logps": -413.4096984863281,
605
+ "debug/reference_rejected_logps": -439.12982177734375,
606
+ "epoch": 0.3625,
607
+ "grad_norm": 7.528653365476016,
608
+ "learning_rate": 1e-06,
609
+ "logits/chosen": 18.986360549926758,
610
+ "logits/rejected": 26.656171798706055,
611
+ "logps/chosen": -396.65667724609375,
612
+ "logps/rejected": -464.8284912109375,
613
+ "loss": 0.3019,
614
+ "rewards/accuracies": 0.875,
615
+ "rewards/chosen": 0.16753005981445312,
616
+ "rewards/margins": 0.4245164692401886,
617
+ "rewards/rejected": -0.2569863796234131,
618
+ "step": 29
619
+ },
620
+ {
621
+ "debug/policy_chosen_logits": 19.743337631225586,
622
+ "debug/policy_chosen_logps": -390.146484375,
623
+ "debug/policy_rejected_logits": 31.34699058532715,
624
+ "debug/policy_rejected_logps": -483.0184631347656,
625
+ "debug/reference_chosen_logps": -425.2975158691406,
626
+ "debug/reference_rejected_logps": -459.4593811035156,
627
+ "epoch": 0.375,
628
+ "grad_norm": 6.576815284840866,
629
+ "learning_rate": 1e-06,
630
+ "logits/chosen": 19.743337631225586,
631
+ "logits/rejected": 31.34699058532715,
632
+ "logps/chosen": -390.146484375,
633
+ "logps/rejected": -483.0184631347656,
634
+ "loss": 0.3144,
635
+ "rewards/accuracies": 1.0,
636
+ "rewards/chosen": 0.35151058435440063,
637
+ "rewards/margins": 0.5871013402938843,
638
+ "rewards/rejected": -0.23559081554412842,
639
+ "step": 30
640
+ },
641
+ {
642
+ "debug/policy_chosen_logits": 23.170053482055664,
643
+ "debug/policy_chosen_logps": -412.2349548339844,
644
+ "debug/policy_rejected_logits": 29.11348533630371,
645
+ "debug/policy_rejected_logps": -455.674072265625,
646
+ "debug/reference_chosen_logps": -428.70257568359375,
647
+ "debug/reference_rejected_logps": -441.6961975097656,
648
+ "epoch": 0.3875,
649
+ "grad_norm": 5.728033865834093,
650
+ "learning_rate": 1e-06,
651
+ "logits/chosen": 23.170053482055664,
652
+ "logits/rejected": 29.11348533630371,
653
+ "logps/chosen": -412.2349548339844,
654
+ "logps/rejected": -455.674072265625,
655
+ "loss": 0.3214,
656
+ "rewards/accuracies": 0.75,
657
+ "rewards/chosen": 0.16467612981796265,
658
+ "rewards/margins": 0.3044551610946655,
659
+ "rewards/rejected": -0.1397790014743805,
660
+ "step": 31
661
+ },
662
+ {
663
+ "debug/policy_chosen_logits": 16.38572883605957,
664
+ "debug/policy_chosen_logps": -390.042724609375,
665
+ "debug/policy_rejected_logits": 22.589452743530273,
666
+ "debug/policy_rejected_logps": -453.4793701171875,
667
+ "debug/reference_chosen_logps": -415.5836181640625,
668
+ "debug/reference_rejected_logps": -437.25,
669
+ "epoch": 0.4,
670
+ "grad_norm": 5.438026921069897,
671
+ "learning_rate": 1e-06,
672
+ "logits/chosen": 16.38572883605957,
673
+ "logits/rejected": 22.589452743530273,
674
+ "logps/chosen": -390.042724609375,
675
+ "logps/rejected": -453.4793701171875,
676
+ "loss": 0.3274,
677
+ "rewards/accuracies": 0.875,
678
+ "rewards/chosen": 0.2554091513156891,
679
+ "rewards/margins": 0.41770267486572266,
680
+ "rewards/rejected": -0.16229349374771118,
681
+ "step": 32
682
+ },
683
+ {
684
+ "debug/policy_chosen_logits": 19.873872756958008,
685
+ "debug/policy_chosen_logps": -417.18951416015625,
686
+ "debug/policy_rejected_logits": 26.914785385131836,
687
+ "debug/policy_rejected_logps": -478.3274841308594,
688
+ "debug/reference_chosen_logps": -428.84234619140625,
689
+ "debug/reference_rejected_logps": -451.3583068847656,
690
+ "epoch": 0.4125,
691
+ "grad_norm": 5.039379106966324,
692
+ "learning_rate": 1e-06,
693
+ "logits/chosen": 19.873872756958008,
694
+ "logits/rejected": 26.914785385131836,
695
+ "logps/chosen": -417.18951416015625,
696
+ "logps/rejected": -478.3274841308594,
697
+ "loss": 0.2881,
698
+ "rewards/accuracies": 0.75,
699
+ "rewards/chosen": 0.11652831733226776,
700
+ "rewards/margins": 0.38622018694877625,
701
+ "rewards/rejected": -0.2696918845176697,
702
+ "step": 33
703
+ },
704
+ {
705
+ "debug/policy_chosen_logits": 21.201560974121094,
706
+ "debug/policy_chosen_logps": -414.26641845703125,
707
+ "debug/policy_rejected_logits": 23.623708724975586,
708
+ "debug/policy_rejected_logps": -460.96356201171875,
709
+ "debug/reference_chosen_logps": -434.948974609375,
710
+ "debug/reference_rejected_logps": -444.27191162109375,
711
+ "epoch": 0.425,
712
+ "grad_norm": 4.310717910219745,
713
+ "learning_rate": 1e-06,
714
+ "logits/chosen": 21.201560974121094,
715
+ "logits/rejected": 23.623708724975586,
716
+ "logps/chosen": -414.26641845703125,
717
+ "logps/rejected": -460.96356201171875,
718
+ "loss": 0.2987,
719
+ "rewards/accuracies": 0.75,
720
+ "rewards/chosen": 0.20682545006275177,
721
+ "rewards/margins": 0.3737417459487915,
722
+ "rewards/rejected": -0.16691631078720093,
723
+ "step": 34
724
+ },
725
+ {
726
+ "debug/policy_chosen_logits": 21.479965209960938,
727
+ "debug/policy_chosen_logps": -378.7373352050781,
728
+ "debug/policy_rejected_logits": 28.288991928100586,
729
+ "debug/policy_rejected_logps": -499.9162292480469,
730
+ "debug/reference_chosen_logps": -399.58843994140625,
731
+ "debug/reference_rejected_logps": -472.7664489746094,
732
+ "epoch": 0.4375,
733
+ "grad_norm": 4.1957614688185645,
734
+ "learning_rate": 1e-06,
735
+ "logits/chosen": 21.479965209960938,
736
+ "logits/rejected": 28.288991928100586,
737
+ "logps/chosen": -378.7373352050781,
738
+ "logps/rejected": -499.9162292480469,
739
+ "loss": 0.2693,
740
+ "rewards/accuracies": 0.875,
741
+ "rewards/chosen": 0.20851120352745056,
742
+ "rewards/margins": 0.48000913858413696,
743
+ "rewards/rejected": -0.2714979648590088,
744
+ "step": 35
745
+ },
746
+ {
747
+ "debug/policy_chosen_logits": 20.284452438354492,
748
+ "debug/policy_chosen_logps": -394.7576904296875,
749
+ "debug/policy_rejected_logits": 24.526214599609375,
750
+ "debug/policy_rejected_logps": -462.82427978515625,
751
+ "debug/reference_chosen_logps": -414.8307800292969,
752
+ "debug/reference_rejected_logps": -430.0937805175781,
753
+ "epoch": 0.45,
754
+ "grad_norm": 4.529884106044187,
755
+ "learning_rate": 1e-06,
756
+ "logits/chosen": 20.284452438354492,
757
+ "logits/rejected": 24.526214599609375,
758
+ "logps/chosen": -394.7576904296875,
759
+ "logps/rejected": -462.82427978515625,
760
+ "loss": 0.2508,
761
+ "rewards/accuracies": 1.0,
762
+ "rewards/chosen": 0.2007310390472412,
763
+ "rewards/margins": 0.5280359387397766,
764
+ "rewards/rejected": -0.3273049294948578,
765
+ "step": 36
766
+ },
767
+ {
768
+ "debug/policy_chosen_logits": 19.54199981689453,
769
+ "debug/policy_chosen_logps": -400.54559326171875,
770
+ "debug/policy_rejected_logits": 32.27294158935547,
771
+ "debug/policy_rejected_logps": -503.24749755859375,
772
+ "debug/reference_chosen_logps": -424.2836608886719,
773
+ "debug/reference_rejected_logps": -475.5150146484375,
774
+ "epoch": 0.4625,
775
+ "grad_norm": 4.291241072360298,
776
+ "learning_rate": 1e-06,
777
+ "logits/chosen": 19.54199981689453,
778
+ "logits/rejected": 32.27294158935547,
779
+ "logps/chosen": -400.54559326171875,
780
+ "logps/rejected": -503.24749755859375,
781
+ "loss": 0.2755,
782
+ "rewards/accuracies": 1.0,
783
+ "rewards/chosen": 0.23738029599189758,
784
+ "rewards/margins": 0.51470547914505,
785
+ "rewards/rejected": -0.27732518315315247,
786
+ "step": 37
787
+ },
788
+ {
789
+ "debug/policy_chosen_logits": 24.72857093811035,
790
+ "debug/policy_chosen_logps": -426.8404541015625,
791
+ "debug/policy_rejected_logits": 27.871118545532227,
792
+ "debug/policy_rejected_logps": -462.07275390625,
793
+ "debug/reference_chosen_logps": -438.54156494140625,
794
+ "debug/reference_rejected_logps": -446.54925537109375,
795
+ "epoch": 0.475,
796
+ "grad_norm": 4.049213448224943,
797
+ "learning_rate": 1e-06,
798
+ "logits/chosen": 24.72857093811035,
799
+ "logits/rejected": 27.871118545532227,
800
+ "logps/chosen": -426.8404541015625,
801
+ "logps/rejected": -462.07275390625,
802
+ "loss": 0.3102,
803
+ "rewards/accuracies": 0.875,
804
+ "rewards/chosen": 0.11701113730669022,
805
+ "rewards/margins": 0.27224594354629517,
806
+ "rewards/rejected": -0.15523478388786316,
807
+ "step": 38
808
+ },
809
+ {
810
+ "debug/policy_chosen_logits": 21.91419792175293,
811
+ "debug/policy_chosen_logps": -408.05224609375,
812
+ "debug/policy_rejected_logits": 21.785812377929688,
813
+ "debug/policy_rejected_logps": -427.7870178222656,
814
+ "debug/reference_chosen_logps": -421.9055480957031,
815
+ "debug/reference_rejected_logps": -422.02252197265625,
816
+ "epoch": 0.4875,
817
+ "grad_norm": 5.051022630915263,
818
+ "learning_rate": 1e-06,
819
+ "logits/chosen": 21.91419792175293,
820
+ "logits/rejected": 21.785812377929688,
821
+ "logps/chosen": -408.05224609375,
822
+ "logps/rejected": -427.7870178222656,
823
+ "loss": 0.3337,
824
+ "rewards/accuracies": 0.75,
825
+ "rewards/chosen": 0.13853324949741364,
826
+ "rewards/margins": 0.19617804884910583,
827
+ "rewards/rejected": -0.0576448068022728,
828
+ "step": 39
829
+ },
830
+ {
831
+ "debug/policy_chosen_logits": 21.50871467590332,
832
+ "debug/policy_chosen_logps": -404.84783935546875,
833
+ "debug/policy_rejected_logits": 26.14636993408203,
834
+ "debug/policy_rejected_logps": -463.62982177734375,
835
+ "debug/reference_chosen_logps": -417.2750244140625,
836
+ "debug/reference_rejected_logps": -446.09521484375,
837
+ "epoch": 0.5,
838
+ "grad_norm": 4.15379359930677,
839
+ "learning_rate": 1e-06,
840
+ "logits/chosen": 21.50871467590332,
841
+ "logits/rejected": 26.14636993408203,
842
+ "logps/chosen": -404.84783935546875,
843
+ "logps/rejected": -463.62982177734375,
844
+ "loss": 0.3061,
845
+ "rewards/accuracies": 0.75,
846
+ "rewards/chosen": 0.12427165359258652,
847
+ "rewards/margins": 0.2996178865432739,
848
+ "rewards/rejected": -0.1753462255001068,
849
+ "step": 40
850
+ },
851
+ {
852
+ "debug/policy_chosen_logits": 19.015485763549805,
853
+ "debug/policy_chosen_logps": -403.9427185058594,
854
+ "debug/policy_rejected_logits": 25.147703170776367,
855
+ "debug/policy_rejected_logps": -482.85748291015625,
856
+ "debug/reference_chosen_logps": -431.6396179199219,
857
+ "debug/reference_rejected_logps": -454.42474365234375,
858
+ "epoch": 0.5125,
859
+ "grad_norm": 4.030629767467611,
860
+ "learning_rate": 1e-06,
861
+ "logits/chosen": 19.015485763549805,
862
+ "logits/rejected": 25.147703170776367,
863
+ "logps/chosen": -403.9427185058594,
864
+ "logps/rejected": -482.85748291015625,
865
+ "loss": 0.2803,
866
+ "rewards/accuracies": 0.875,
867
+ "rewards/chosen": 0.27696919441223145,
868
+ "rewards/margins": 0.5612967014312744,
869
+ "rewards/rejected": -0.28432750701904297,
870
+ "step": 41
871
+ },
872
+ {
873
+ "debug/policy_chosen_logits": 20.53624153137207,
874
+ "debug/policy_chosen_logps": -390.5385437011719,
875
+ "debug/policy_rejected_logits": 29.287269592285156,
876
+ "debug/policy_rejected_logps": -497.9327392578125,
877
+ "debug/reference_chosen_logps": -414.4248962402344,
878
+ "debug/reference_rejected_logps": -457.1654357910156,
879
+ "epoch": 0.525,
880
+ "grad_norm": 5.170853621888292,
881
+ "learning_rate": 1e-06,
882
+ "logits/chosen": 20.53624153137207,
883
+ "logits/rejected": 29.287269592285156,
884
+ "logps/chosen": -390.5385437011719,
885
+ "logps/rejected": -497.9327392578125,
886
+ "loss": 0.2504,
887
+ "rewards/accuracies": 1.0,
888
+ "rewards/chosen": 0.2388634830713272,
889
+ "rewards/margins": 0.6465361714363098,
890
+ "rewards/rejected": -0.4076727032661438,
891
+ "step": 42
892
+ },
893
+ {
894
+ "debug/policy_chosen_logits": 20.2578125,
895
+ "debug/policy_chosen_logps": -405.1622314453125,
896
+ "debug/policy_rejected_logits": 28.01288414001465,
897
+ "debug/policy_rejected_logps": -465.55914306640625,
898
+ "debug/reference_chosen_logps": -429.2195129394531,
899
+ "debug/reference_rejected_logps": -439.44915771484375,
900
+ "epoch": 0.5375,
901
+ "grad_norm": 3.882500417686335,
902
+ "learning_rate": 1e-06,
903
+ "logits/chosen": 20.2578125,
904
+ "logits/rejected": 28.01288414001465,
905
+ "logps/chosen": -405.1622314453125,
906
+ "logps/rejected": -465.55914306640625,
907
+ "loss": 0.2292,
908
+ "rewards/accuracies": 0.75,
909
+ "rewards/chosen": 0.2405730038881302,
910
+ "rewards/margins": 0.5016728639602661,
911
+ "rewards/rejected": -0.2610998749732971,
912
+ "step": 43
913
+ },
914
+ {
915
+ "debug/policy_chosen_logits": 17.483173370361328,
916
+ "debug/policy_chosen_logps": -364.91351318359375,
917
+ "debug/policy_rejected_logits": 29.26898765563965,
918
+ "debug/policy_rejected_logps": -483.2657165527344,
919
+ "debug/reference_chosen_logps": -391.24664306640625,
920
+ "debug/reference_rejected_logps": -457.18780517578125,
921
+ "epoch": 0.55,
922
+ "grad_norm": 4.7605947943975675,
923
+ "learning_rate": 1e-06,
924
+ "logits/chosen": 17.483173370361328,
925
+ "logits/rejected": 29.26898765563965,
926
+ "logps/chosen": -364.91351318359375,
927
+ "logps/rejected": -483.2657165527344,
928
+ "loss": 0.285,
929
+ "rewards/accuracies": 1.0,
930
+ "rewards/chosen": 0.26333099603652954,
931
+ "rewards/margins": 0.5241098403930664,
932
+ "rewards/rejected": -0.26077884435653687,
933
+ "step": 44
934
+ },
935
+ {
936
+ "debug/policy_chosen_logits": 19.23508071899414,
937
+ "debug/policy_chosen_logps": -387.4403076171875,
938
+ "debug/policy_rejected_logits": 27.274477005004883,
939
+ "debug/policy_rejected_logps": -444.69635009765625,
940
+ "debug/reference_chosen_logps": -412.97003173828125,
941
+ "debug/reference_rejected_logps": -437.71649169921875,
942
+ "epoch": 0.5625,
943
+ "grad_norm": 7.535235898290158,
944
+ "learning_rate": 1e-06,
945
+ "logits/chosen": 19.23508071899414,
946
+ "logits/rejected": 27.274477005004883,
947
+ "logps/chosen": -387.4403076171875,
948
+ "logps/rejected": -444.69635009765625,
949
+ "loss": 0.3417,
950
+ "rewards/accuracies": 0.75,
951
+ "rewards/chosen": 0.25529688596725464,
952
+ "rewards/margins": 0.32509559392929077,
953
+ "rewards/rejected": -0.06979873776435852,
954
+ "step": 45
955
+ },
956
+ {
957
+ "debug/policy_chosen_logits": 20.150699615478516,
958
+ "debug/policy_chosen_logps": -383.12786865234375,
959
+ "debug/policy_rejected_logits": 23.752517700195312,
960
+ "debug/policy_rejected_logps": -429.2421875,
961
+ "debug/reference_chosen_logps": -409.3837585449219,
962
+ "debug/reference_rejected_logps": -413.81390380859375,
963
+ "epoch": 0.575,
964
+ "grad_norm": 6.267381407493171,
965
+ "learning_rate": 1e-06,
966
+ "logits/chosen": 20.150699615478516,
967
+ "logits/rejected": 23.752517700195312,
968
+ "logps/chosen": -383.12786865234375,
969
+ "logps/rejected": -429.2421875,
970
+ "loss": 0.3195,
971
+ "rewards/accuracies": 0.875,
972
+ "rewards/chosen": 0.26255881786346436,
973
+ "rewards/margins": 0.41684138774871826,
974
+ "rewards/rejected": -0.1542825698852539,
975
+ "step": 46
976
+ },
977
+ {
978
+ "debug/policy_chosen_logits": 19.68763542175293,
979
+ "debug/policy_chosen_logps": -385.29144287109375,
980
+ "debug/policy_rejected_logits": 28.113876342773438,
981
+ "debug/policy_rejected_logps": -451.77642822265625,
982
+ "debug/reference_chosen_logps": -412.47412109375,
983
+ "debug/reference_rejected_logps": -438.6710205078125,
984
+ "epoch": 0.5875,
985
+ "grad_norm": 5.633066278680009,
986
+ "learning_rate": 1e-06,
987
+ "logits/chosen": 19.68763542175293,
988
+ "logits/rejected": 28.113876342773438,
989
+ "logps/chosen": -385.29144287109375,
990
+ "logps/rejected": -451.77642822265625,
991
+ "loss": 0.2709,
992
+ "rewards/accuracies": 0.875,
993
+ "rewards/chosen": 0.2718266546726227,
994
+ "rewards/margins": 0.4028809666633606,
995
+ "rewards/rejected": -0.13105429708957672,
996
+ "step": 47
997
+ },
998
+ {
999
+ "debug/policy_chosen_logits": 21.418119430541992,
1000
+ "debug/policy_chosen_logps": -418.92596435546875,
1001
+ "debug/policy_rejected_logits": 28.619443893432617,
1002
+ "debug/policy_rejected_logps": -447.74090576171875,
1003
+ "debug/reference_chosen_logps": -431.38714599609375,
1004
+ "debug/reference_rejected_logps": -444.2582092285156,
1005
+ "epoch": 0.6,
1006
+ "grad_norm": 4.297385322384447,
1007
+ "learning_rate": 1e-06,
1008
+ "logits/chosen": 21.418119430541992,
1009
+ "logits/rejected": 28.619443893432617,
1010
+ "logps/chosen": -418.92596435546875,
1011
+ "logps/rejected": -447.74090576171875,
1012
+ "loss": 0.302,
1013
+ "rewards/accuracies": 0.625,
1014
+ "rewards/chosen": 0.12461188435554504,
1015
+ "rewards/margins": 0.15943893790245056,
1016
+ "rewards/rejected": -0.034827034920454025,
1017
+ "step": 48
1018
+ },
1019
+ {
1020
+ "debug/policy_chosen_logits": 21.347043991088867,
1021
+ "debug/policy_chosen_logps": -409.0511474609375,
1022
+ "debug/policy_rejected_logits": 29.213226318359375,
1023
+ "debug/policy_rejected_logps": -479.66387939453125,
1024
+ "debug/reference_chosen_logps": -424.4683837890625,
1025
+ "debug/reference_rejected_logps": -453.17791748046875,
1026
+ "epoch": 0.6125,
1027
+ "grad_norm": 3.600194804734596,
1028
+ "learning_rate": 1e-06,
1029
+ "logits/chosen": 21.347043991088867,
1030
+ "logits/rejected": 29.213226318359375,
1031
+ "logps/chosen": -409.0511474609375,
1032
+ "logps/rejected": -479.66387939453125,
1033
+ "loss": 0.3015,
1034
+ "rewards/accuracies": 1.0,
1035
+ "rewards/chosen": 0.15417221188545227,
1036
+ "rewards/margins": 0.41903185844421387,
1037
+ "rewards/rejected": -0.2648596465587616,
1038
+ "step": 49
1039
+ },
1040
+ {
1041
+ "debug/policy_chosen_logits": 19.539743423461914,
1042
+ "debug/policy_chosen_logps": -406.68109130859375,
1043
+ "debug/policy_rejected_logits": 23.604896545410156,
1044
+ "debug/policy_rejected_logps": -447.7806701660156,
1045
+ "debug/reference_chosen_logps": -427.583740234375,
1046
+ "debug/reference_rejected_logps": -432.884765625,
1047
+ "epoch": 0.625,
1048
+ "grad_norm": 4.128916174651822,
1049
+ "learning_rate": 1e-06,
1050
+ "logits/chosen": 19.539743423461914,
1051
+ "logits/rejected": 23.604896545410156,
1052
+ "logps/chosen": -406.68109130859375,
1053
+ "logps/rejected": -447.7806701660156,
1054
+ "loss": 0.3165,
1055
+ "rewards/accuracies": 1.0,
1056
+ "rewards/chosen": 0.2090263068675995,
1057
+ "rewards/margins": 0.35798513889312744,
1058
+ "rewards/rejected": -0.14895884692668915,
1059
+ "step": 50
1060
+ },
1061
+ {
1062
+ "debug/policy_chosen_logits": 23.078472137451172,
1063
+ "debug/policy_chosen_logps": -413.09417724609375,
1064
+ "debug/policy_rejected_logits": 27.335500717163086,
1065
+ "debug/policy_rejected_logps": -472.38079833984375,
1066
+ "debug/reference_chosen_logps": -425.72039794921875,
1067
+ "debug/reference_rejected_logps": -457.5447082519531,
1068
+ "epoch": 0.6375,
1069
+ "grad_norm": 4.640148404581812,
1070
+ "learning_rate": 1e-06,
1071
+ "logits/chosen": 23.078472137451172,
1072
+ "logits/rejected": 27.335500717163086,
1073
+ "logps/chosen": -413.09417724609375,
1074
+ "logps/rejected": -472.38079833984375,
1075
+ "loss": 0.339,
1076
+ "rewards/accuracies": 0.75,
1077
+ "rewards/chosen": 0.12626220285892487,
1078
+ "rewards/margins": 0.27462291717529297,
1079
+ "rewards/rejected": -0.1483606994152069,
1080
+ "step": 51
1081
+ },
1082
+ {
1083
+ "debug/policy_chosen_logits": 18.798938751220703,
1084
+ "debug/policy_chosen_logps": -392.9048767089844,
1085
+ "debug/policy_rejected_logits": 31.566543579101562,
1086
+ "debug/policy_rejected_logps": -530.4522094726562,
1087
+ "debug/reference_chosen_logps": -410.1727294921875,
1088
+ "debug/reference_rejected_logps": -481.0628967285156,
1089
+ "epoch": 0.65,
1090
+ "grad_norm": 5.586177810089015,
1091
+ "learning_rate": 1e-06,
1092
+ "logits/chosen": 18.798938751220703,
1093
+ "logits/rejected": 31.566543579101562,
1094
+ "logps/chosen": -392.9048767089844,
1095
+ "logps/rejected": -530.4522094726562,
1096
+ "loss": 0.2993,
1097
+ "rewards/accuracies": 1.0,
1098
+ "rewards/chosen": 0.17267844080924988,
1099
+ "rewards/margins": 0.6665716171264648,
1100
+ "rewards/rejected": -0.49389317631721497,
1101
+ "step": 52
1102
+ },
1103
+ {
1104
+ "debug/policy_chosen_logits": 22.037538528442383,
1105
+ "debug/policy_chosen_logps": -415.8323669433594,
1106
+ "debug/policy_rejected_logits": 27.0804386138916,
1107
+ "debug/policy_rejected_logps": -468.87591552734375,
1108
+ "debug/reference_chosen_logps": -431.1224365234375,
1109
+ "debug/reference_rejected_logps": -426.37371826171875,
1110
+ "epoch": 0.6625,
1111
+ "grad_norm": 6.488572834551389,
1112
+ "learning_rate": 1e-06,
1113
+ "logits/chosen": 22.037538528442383,
1114
+ "logits/rejected": 27.0804386138916,
1115
+ "logps/chosen": -415.8323669433594,
1116
+ "logps/rejected": -468.87591552734375,
1117
+ "loss": 0.3265,
1118
+ "rewards/accuracies": 0.75,
1119
+ "rewards/chosen": 0.1529008448123932,
1120
+ "rewards/margins": 0.5779224634170532,
1121
+ "rewards/rejected": -0.42502161860466003,
1122
+ "step": 53
1123
+ },
1124
+ {
1125
+ "debug/policy_chosen_logits": 18.0117130279541,
1126
+ "debug/policy_chosen_logps": -398.383056640625,
1127
+ "debug/policy_rejected_logits": 30.224939346313477,
1128
+ "debug/policy_rejected_logps": -452.33502197265625,
1129
+ "debug/reference_chosen_logps": -418.76708984375,
1130
+ "debug/reference_rejected_logps": -439.8731384277344,
1131
+ "epoch": 0.675,
1132
+ "grad_norm": 6.398923285824906,
1133
+ "learning_rate": 1e-06,
1134
+ "logits/chosen": 18.0117130279541,
1135
+ "logits/rejected": 30.224939346313477,
1136
+ "logps/chosen": -398.383056640625,
1137
+ "logps/rejected": -452.33502197265625,
1138
+ "loss": 0.2931,
1139
+ "rewards/accuracies": 0.875,
1140
+ "rewards/chosen": 0.20384037494659424,
1141
+ "rewards/margins": 0.32845935225486755,
1142
+ "rewards/rejected": -0.12461898475885391,
1143
+ "step": 54
1144
+ },
1145
+ {
1146
+ "debug/policy_chosen_logits": 19.977191925048828,
1147
+ "debug/policy_chosen_logps": -387.9195556640625,
1148
+ "debug/policy_rejected_logits": 27.902311325073242,
1149
+ "debug/policy_rejected_logps": -465.3125,
1150
+ "debug/reference_chosen_logps": -413.6059265136719,
1151
+ "debug/reference_rejected_logps": -443.1114196777344,
1152
+ "epoch": 0.6875,
1153
+ "grad_norm": 3.685551814432106,
1154
+ "learning_rate": 1e-06,
1155
+ "logits/chosen": 19.977191925048828,
1156
+ "logits/rejected": 27.902311325073242,
1157
+ "logps/chosen": -387.9195556640625,
1158
+ "logps/rejected": -465.3125,
1159
+ "loss": 0.2927,
1160
+ "rewards/accuracies": 0.875,
1161
+ "rewards/chosen": 0.25686389207839966,
1162
+ "rewards/margins": 0.4788748323917389,
1163
+ "rewards/rejected": -0.22201094031333923,
1164
+ "step": 55
1165
+ },
1166
+ {
1167
+ "debug/policy_chosen_logits": 19.4679012298584,
1168
+ "debug/policy_chosen_logps": -387.61016845703125,
1169
+ "debug/policy_rejected_logits": 24.729759216308594,
1170
+ "debug/policy_rejected_logps": -484.8683776855469,
1171
+ "debug/reference_chosen_logps": -407.5805969238281,
1172
+ "debug/reference_rejected_logps": -456.95703125,
1173
+ "epoch": 0.7,
1174
+ "grad_norm": 3.5458155693246467,
1175
+ "learning_rate": 1e-06,
1176
+ "logits/chosen": 19.4679012298584,
1177
+ "logits/rejected": 24.729759216308594,
1178
+ "logps/chosen": -387.61016845703125,
1179
+ "logps/rejected": -484.8683776855469,
1180
+ "loss": 0.2736,
1181
+ "rewards/accuracies": 0.75,
1182
+ "rewards/chosen": 0.19970446825027466,
1183
+ "rewards/margins": 0.4788181483745575,
1184
+ "rewards/rejected": -0.27911368012428284,
1185
+ "step": 56
1186
+ },
1187
+ {
1188
+ "debug/policy_chosen_logits": 19.905282974243164,
1189
+ "debug/policy_chosen_logps": -404.56280517578125,
1190
+ "debug/policy_rejected_logits": 23.55043601989746,
1191
+ "debug/policy_rejected_logps": -455.38018798828125,
1192
+ "debug/reference_chosen_logps": -425.48101806640625,
1193
+ "debug/reference_rejected_logps": -444.97747802734375,
1194
+ "epoch": 0.7125,
1195
+ "grad_norm": 3.2623990350257848,
1196
+ "learning_rate": 1e-06,
1197
+ "logits/chosen": 19.905282974243164,
1198
+ "logits/rejected": 23.55043601989746,
1199
+ "logps/chosen": -404.56280517578125,
1200
+ "logps/rejected": -455.38018798828125,
1201
+ "loss": 0.2562,
1202
+ "rewards/accuracies": 0.75,
1203
+ "rewards/chosen": 0.20918196439743042,
1204
+ "rewards/margins": 0.31320905685424805,
1205
+ "rewards/rejected": -0.10402709245681763,
1206
+ "step": 57
1207
+ },
1208
+ {
1209
+ "debug/policy_chosen_logits": 23.94160270690918,
1210
+ "debug/policy_chosen_logps": -399.81439208984375,
1211
+ "debug/policy_rejected_logits": 31.01986312866211,
1212
+ "debug/policy_rejected_logps": -480.4059753417969,
1213
+ "debug/reference_chosen_logps": -429.8614501953125,
1214
+ "debug/reference_rejected_logps": -448.6155700683594,
1215
+ "epoch": 0.725,
1216
+ "grad_norm": 5.094791792265245,
1217
+ "learning_rate": 1e-06,
1218
+ "logits/chosen": 23.94160270690918,
1219
+ "logits/rejected": 31.01986312866211,
1220
+ "logps/chosen": -399.81439208984375,
1221
+ "logps/rejected": -480.4059753417969,
1222
+ "loss": 0.3252,
1223
+ "rewards/accuracies": 1.0,
1224
+ "rewards/chosen": 0.30047041177749634,
1225
+ "rewards/margins": 0.6183742880821228,
1226
+ "rewards/rejected": -0.31790387630462646,
1227
+ "step": 58
1228
+ },
1229
+ {
1230
+ "debug/policy_chosen_logits": 22.9466495513916,
1231
+ "debug/policy_chosen_logps": -385.3292541503906,
1232
+ "debug/policy_rejected_logits": 36.00619888305664,
1233
+ "debug/policy_rejected_logps": -493.22198486328125,
1234
+ "debug/reference_chosen_logps": -421.3563232421875,
1235
+ "debug/reference_rejected_logps": -477.6571350097656,
1236
+ "epoch": 0.7375,
1237
+ "grad_norm": 5.69752289722826,
1238
+ "learning_rate": 1e-06,
1239
+ "logits/chosen": 22.9466495513916,
1240
+ "logits/rejected": 36.00619888305664,
1241
+ "logps/chosen": -385.3292541503906,
1242
+ "logps/rejected": -493.22198486328125,
1243
+ "loss": 0.2982,
1244
+ "rewards/accuracies": 1.0,
1245
+ "rewards/chosen": 0.360270619392395,
1246
+ "rewards/margins": 0.515919029712677,
1247
+ "rewards/rejected": -0.1556483805179596,
1248
+ "step": 59
1249
+ },
1250
+ {
1251
+ "debug/policy_chosen_logits": 23.001218795776367,
1252
+ "debug/policy_chosen_logps": -398.92132568359375,
1253
+ "debug/policy_rejected_logits": 30.06394386291504,
1254
+ "debug/policy_rejected_logps": -459.4337463378906,
1255
+ "debug/reference_chosen_logps": -427.70135498046875,
1256
+ "debug/reference_rejected_logps": -427.77716064453125,
1257
+ "epoch": 0.75,
1258
+ "grad_norm": 3.651921352682424,
1259
+ "learning_rate": 1e-06,
1260
+ "logits/chosen": 23.001218795776367,
1261
+ "logits/rejected": 30.06394386291504,
1262
+ "logps/chosen": -398.92132568359375,
1263
+ "logps/rejected": -459.4337463378906,
1264
+ "loss": 0.26,
1265
+ "rewards/accuracies": 1.0,
1266
+ "rewards/chosen": 0.2878003716468811,
1267
+ "rewards/margins": 0.6043664216995239,
1268
+ "rewards/rejected": -0.3165660500526428,
1269
+ "step": 60
1270
+ },
1271
+ {
1272
+ "debug/policy_chosen_logits": 19.97127914428711,
1273
+ "debug/policy_chosen_logps": -396.8033447265625,
1274
+ "debug/policy_rejected_logits": 27.122671127319336,
1275
+ "debug/policy_rejected_logps": -485.0689697265625,
1276
+ "debug/reference_chosen_logps": -416.9521484375,
1277
+ "debug/reference_rejected_logps": -456.4365539550781,
1278
+ "epoch": 0.7625,
1279
+ "grad_norm": 3.768202921022212,
1280
+ "learning_rate": 1e-06,
1281
+ "logits/chosen": 19.97127914428711,
1282
+ "logits/rejected": 27.122671127319336,
1283
+ "logps/chosen": -396.8033447265625,
1284
+ "logps/rejected": -485.0689697265625,
1285
+ "loss": 0.2754,
1286
+ "rewards/accuracies": 1.0,
1287
+ "rewards/chosen": 0.20148837566375732,
1288
+ "rewards/margins": 0.487812340259552,
1289
+ "rewards/rejected": -0.28632399439811707,
1290
+ "step": 61
1291
+ },
1292
+ {
1293
+ "debug/policy_chosen_logits": 25.002408981323242,
1294
+ "debug/policy_chosen_logps": -440.4338073730469,
1295
+ "debug/policy_rejected_logits": 29.166627883911133,
1296
+ "debug/policy_rejected_logps": -472.5847473144531,
1297
+ "debug/reference_chosen_logps": -458.51873779296875,
1298
+ "debug/reference_rejected_logps": -445.6849365234375,
1299
+ "epoch": 0.775,
1300
+ "grad_norm": 3.8018379677299614,
1301
+ "learning_rate": 1e-06,
1302
+ "logits/chosen": 25.002408981323242,
1303
+ "logits/rejected": 29.166627883911133,
1304
+ "logps/chosen": -440.4338073730469,
1305
+ "logps/rejected": -472.5847473144531,
1306
+ "loss": 0.2766,
1307
+ "rewards/accuracies": 0.875,
1308
+ "rewards/chosen": 0.18084929883480072,
1309
+ "rewards/margins": 0.4498472809791565,
1310
+ "rewards/rejected": -0.26899799704551697,
1311
+ "step": 62
1312
+ },
1313
+ {
1314
+ "debug/policy_chosen_logits": 19.132322311401367,
1315
+ "debug/policy_chosen_logps": -405.4966735839844,
1316
+ "debug/policy_rejected_logits": 28.601638793945312,
1317
+ "debug/policy_rejected_logps": -506.548828125,
1318
+ "debug/reference_chosen_logps": -430.78936767578125,
1319
+ "debug/reference_rejected_logps": -464.448486328125,
1320
+ "epoch": 0.7875,
1321
+ "grad_norm": 3.9013050377479486,
1322
+ "learning_rate": 1e-06,
1323
+ "logits/chosen": 19.132322311401367,
1324
+ "logits/rejected": 28.601638793945312,
1325
+ "logps/chosen": -405.4966735839844,
1326
+ "logps/rejected": -506.548828125,
1327
+ "loss": 0.2766,
1328
+ "rewards/accuracies": 1.0,
1329
+ "rewards/chosen": 0.2529268264770508,
1330
+ "rewards/margins": 0.6739299297332764,
1331
+ "rewards/rejected": -0.42100316286087036,
1332
+ "step": 63
1333
+ },
1334
+ {
1335
+ "debug/policy_chosen_logits": 19.04224395751953,
1336
+ "debug/policy_chosen_logps": -399.923095703125,
1337
+ "debug/policy_rejected_logits": 29.6719970703125,
1338
+ "debug/policy_rejected_logps": -487.8897399902344,
1339
+ "debug/reference_chosen_logps": -422.9737243652344,
1340
+ "debug/reference_rejected_logps": -463.2634582519531,
1341
+ "epoch": 0.8,
1342
+ "grad_norm": 4.331208338864034,
1343
+ "learning_rate": 1e-06,
1344
+ "logits/chosen": 19.04224395751953,
1345
+ "logits/rejected": 29.6719970703125,
1346
+ "logps/chosen": -399.923095703125,
1347
+ "logps/rejected": -487.8897399902344,
1348
+ "loss": 0.2082,
1349
+ "rewards/accuracies": 0.875,
1350
+ "rewards/chosen": 0.2305060774087906,
1351
+ "rewards/margins": 0.4767689108848572,
1352
+ "rewards/rejected": -0.2462628036737442,
1353
+ "step": 64
1354
+ },
1355
+ {
1356
+ "debug/policy_chosen_logits": 21.41056251525879,
1357
+ "debug/policy_chosen_logps": -421.7527770996094,
1358
+ "debug/policy_rejected_logits": 26.84547233581543,
1359
+ "debug/policy_rejected_logps": -474.83660888671875,
1360
+ "debug/reference_chosen_logps": -435.3292236328125,
1361
+ "debug/reference_rejected_logps": -446.2524108886719,
1362
+ "epoch": 0.8125,
1363
+ "grad_norm": 4.552945704427676,
1364
+ "learning_rate": 1e-06,
1365
+ "logits/chosen": 21.41056251525879,
1366
+ "logits/rejected": 26.84547233581543,
1367
+ "logps/chosen": -421.7527770996094,
1368
+ "logps/rejected": -474.83660888671875,
1369
+ "loss": 0.3366,
1370
+ "rewards/accuracies": 0.875,
1371
+ "rewards/chosen": 0.13576430082321167,
1372
+ "rewards/margins": 0.421606183052063,
1373
+ "rewards/rejected": -0.2858418822288513,
1374
+ "step": 65
1375
+ },
1376
+ {
1377
+ "debug/policy_chosen_logits": 20.119802474975586,
1378
+ "debug/policy_chosen_logps": -381.8564147949219,
1379
+ "debug/policy_rejected_logits": 28.411861419677734,
1380
+ "debug/policy_rejected_logps": -466.48577880859375,
1381
+ "debug/reference_chosen_logps": -406.800048828125,
1382
+ "debug/reference_rejected_logps": -452.9040222167969,
1383
+ "epoch": 0.825,
1384
+ "grad_norm": 3.389829713086373,
1385
+ "learning_rate": 1e-06,
1386
+ "logits/chosen": 20.119802474975586,
1387
+ "logits/rejected": 28.411861419677734,
1388
+ "logps/chosen": -381.8564147949219,
1389
+ "logps/rejected": -466.48577880859375,
1390
+ "loss": 0.2748,
1391
+ "rewards/accuracies": 0.625,
1392
+ "rewards/chosen": 0.24943621456623077,
1393
+ "rewards/margins": 0.385253369808197,
1394
+ "rewards/rejected": -0.13581714034080505,
1395
+ "step": 66
1396
+ },
1397
+ {
1398
+ "debug/policy_chosen_logits": 18.72797203063965,
1399
+ "debug/policy_chosen_logps": -390.49090576171875,
1400
+ "debug/policy_rejected_logits": 26.687557220458984,
1401
+ "debug/policy_rejected_logps": -512.882568359375,
1402
+ "debug/reference_chosen_logps": -418.24884033203125,
1403
+ "debug/reference_rejected_logps": -474.900390625,
1404
+ "epoch": 0.8375,
1405
+ "grad_norm": 3.623515504007148,
1406
+ "learning_rate": 1e-06,
1407
+ "logits/chosen": 18.72797203063965,
1408
+ "logits/rejected": 26.687557220458984,
1409
+ "logps/chosen": -390.49090576171875,
1410
+ "logps/rejected": -512.882568359375,
1411
+ "loss": 0.2272,
1412
+ "rewards/accuracies": 1.0,
1413
+ "rewards/chosen": 0.2775792181491852,
1414
+ "rewards/margins": 0.6574012637138367,
1415
+ "rewards/rejected": -0.3798220455646515,
1416
+ "step": 67
1417
+ },
1418
+ {
1419
+ "debug/policy_chosen_logits": 23.56059455871582,
1420
+ "debug/policy_chosen_logps": -410.64044189453125,
1421
+ "debug/policy_rejected_logits": 34.364986419677734,
1422
+ "debug/policy_rejected_logps": -478.304931640625,
1423
+ "debug/reference_chosen_logps": -428.95367431640625,
1424
+ "debug/reference_rejected_logps": -446.1668395996094,
1425
+ "epoch": 0.85,
1426
+ "grad_norm": 3.8118009157857644,
1427
+ "learning_rate": 1e-06,
1428
+ "logits/chosen": 23.56059455871582,
1429
+ "logits/rejected": 34.364986419677734,
1430
+ "logps/chosen": -410.64044189453125,
1431
+ "logps/rejected": -478.304931640625,
1432
+ "loss": 0.2653,
1433
+ "rewards/accuracies": 1.0,
1434
+ "rewards/chosen": 0.18313205242156982,
1435
+ "rewards/margins": 0.5045130848884583,
1436
+ "rewards/rejected": -0.3213810324668884,
1437
+ "step": 68
1438
+ },
1439
+ {
1440
+ "debug/policy_chosen_logits": 20.668880462646484,
1441
+ "debug/policy_chosen_logps": -408.55419921875,
1442
+ "debug/policy_rejected_logits": 27.681875228881836,
1443
+ "debug/policy_rejected_logps": -442.37799072265625,
1444
+ "debug/reference_chosen_logps": -434.24542236328125,
1445
+ "debug/reference_rejected_logps": -441.95709228515625,
1446
+ "epoch": 0.8625,
1447
+ "grad_norm": 3.7604869397831107,
1448
+ "learning_rate": 1e-06,
1449
+ "logits/chosen": 20.668880462646484,
1450
+ "logits/rejected": 27.681875228881836,
1451
+ "logps/chosen": -408.55419921875,
1452
+ "logps/rejected": -442.37799072265625,
1453
+ "loss": 0.2795,
1454
+ "rewards/accuracies": 0.75,
1455
+ "rewards/chosen": 0.2569122612476349,
1456
+ "rewards/margins": 0.2611211836338043,
1457
+ "rewards/rejected": -0.004208952188491821,
1458
+ "step": 69
1459
+ },
1460
+ {
1461
+ "debug/policy_chosen_logits": 25.379365921020508,
1462
+ "debug/policy_chosen_logps": -395.3825988769531,
1463
+ "debug/policy_rejected_logits": 31.113773345947266,
1464
+ "debug/policy_rejected_logps": -473.5229797363281,
1465
+ "debug/reference_chosen_logps": -419.5213623046875,
1466
+ "debug/reference_rejected_logps": -455.77142333984375,
1467
+ "epoch": 0.875,
1468
+ "grad_norm": 4.527244920289224,
1469
+ "learning_rate": 1e-06,
1470
+ "logits/chosen": 25.379365921020508,
1471
+ "logits/rejected": 31.113773345947266,
1472
+ "logps/chosen": -395.3825988769531,
1473
+ "logps/rejected": -473.5229797363281,
1474
+ "loss": 0.2967,
1475
+ "rewards/accuracies": 1.0,
1476
+ "rewards/chosen": 0.24138759076595306,
1477
+ "rewards/margins": 0.41890305280685425,
1478
+ "rewards/rejected": -0.17751547694206238,
1479
+ "step": 70
1480
+ },
1481
+ {
1482
+ "debug/policy_chosen_logits": 19.842777252197266,
1483
+ "debug/policy_chosen_logps": -397.84136962890625,
1484
+ "debug/policy_rejected_logits": 28.35538101196289,
1485
+ "debug/policy_rejected_logps": -484.2126770019531,
1486
+ "debug/reference_chosen_logps": -415.0436096191406,
1487
+ "debug/reference_rejected_logps": -459.60400390625,
1488
+ "epoch": 0.8875,
1489
+ "grad_norm": 5.4667643203198075,
1490
+ "learning_rate": 1e-06,
1491
+ "logits/chosen": 19.842777252197266,
1492
+ "logits/rejected": 28.35538101196289,
1493
+ "logps/chosen": -397.84136962890625,
1494
+ "logps/rejected": -484.2126770019531,
1495
+ "loss": 0.2524,
1496
+ "rewards/accuracies": 0.875,
1497
+ "rewards/chosen": 0.17202238738536835,
1498
+ "rewards/margins": 0.41810929775238037,
1499
+ "rewards/rejected": -0.24608686566352844,
1500
+ "step": 71
1501
+ },
1502
+ {
1503
+ "debug/policy_chosen_logits": 23.183486938476562,
1504
+ "debug/policy_chosen_logps": -400.0027160644531,
1505
+ "debug/policy_rejected_logits": 27.23314094543457,
1506
+ "debug/policy_rejected_logps": -473.27593994140625,
1507
+ "debug/reference_chosen_logps": -427.227783203125,
1508
+ "debug/reference_rejected_logps": -448.28363037109375,
1509
+ "epoch": 0.9,
1510
+ "grad_norm": 3.5365215033749213,
1511
+ "learning_rate": 1e-06,
1512
+ "logits/chosen": 23.183486938476562,
1513
+ "logits/rejected": 27.23314094543457,
1514
+ "logps/chosen": -400.0027160644531,
1515
+ "logps/rejected": -473.27593994140625,
1516
+ "loss": 0.2537,
1517
+ "rewards/accuracies": 1.0,
1518
+ "rewards/chosen": 0.272250771522522,
1519
+ "rewards/margins": 0.5221738815307617,
1520
+ "rewards/rejected": -0.24992308020591736,
1521
+ "step": 72
1522
+ },
1523
+ {
1524
+ "debug/policy_chosen_logits": 20.710277557373047,
1525
+ "debug/policy_chosen_logps": -398.7235107421875,
1526
+ "debug/policy_rejected_logits": 24.029035568237305,
1527
+ "debug/policy_rejected_logps": -449.9461364746094,
1528
+ "debug/reference_chosen_logps": -419.6963195800781,
1529
+ "debug/reference_rejected_logps": -447.973876953125,
1530
+ "epoch": 0.9125,
1531
+ "grad_norm": 3.7346432080690812,
1532
+ "learning_rate": 1e-06,
1533
+ "logits/chosen": 20.710277557373047,
1534
+ "logits/rejected": 24.029035568237305,
1535
+ "logps/chosen": -398.7235107421875,
1536
+ "logps/rejected": -449.9461364746094,
1537
+ "loss": 0.2855,
1538
+ "rewards/accuracies": 0.625,
1539
+ "rewards/chosen": 0.20972807705402374,
1540
+ "rewards/margins": 0.22945068776607513,
1541
+ "rewards/rejected": -0.019722595810890198,
1542
+ "step": 73
1543
+ },
1544
+ {
1545
+ "debug/policy_chosen_logits": 22.077816009521484,
1546
+ "debug/policy_chosen_logps": -401.99786376953125,
1547
+ "debug/policy_rejected_logits": 30.93896484375,
1548
+ "debug/policy_rejected_logps": -486.0145568847656,
1549
+ "debug/reference_chosen_logps": -420.620361328125,
1550
+ "debug/reference_rejected_logps": -452.76373291015625,
1551
+ "epoch": 0.925,
1552
+ "grad_norm": 4.320209693666537,
1553
+ "learning_rate": 1e-06,
1554
+ "logits/chosen": 22.077816009521484,
1555
+ "logits/rejected": 30.93896484375,
1556
+ "logps/chosen": -401.99786376953125,
1557
+ "logps/rejected": -486.0145568847656,
1558
+ "loss": 0.2651,
1559
+ "rewards/accuracies": 0.875,
1560
+ "rewards/chosen": 0.18622498214244843,
1561
+ "rewards/margins": 0.5187329649925232,
1562
+ "rewards/rejected": -0.33250799775123596,
1563
+ "step": 74
1564
+ },
1565
+ {
1566
+ "debug/policy_chosen_logits": 16.68404197692871,
1567
+ "debug/policy_chosen_logps": -377.2335510253906,
1568
+ "debug/policy_rejected_logits": 31.94536781311035,
1569
+ "debug/policy_rejected_logps": -488.6978454589844,
1570
+ "debug/reference_chosen_logps": -400.86431884765625,
1571
+ "debug/reference_rejected_logps": -439.6689758300781,
1572
+ "epoch": 0.9375,
1573
+ "grad_norm": 3.3831819183050054,
1574
+ "learning_rate": 1e-06,
1575
+ "logits/chosen": 16.68404197692871,
1576
+ "logits/rejected": 31.94536781311035,
1577
+ "logps/chosen": -377.2335510253906,
1578
+ "logps/rejected": -488.6978454589844,
1579
+ "loss": 0.2509,
1580
+ "rewards/accuracies": 1.0,
1581
+ "rewards/chosen": 0.23630763590335846,
1582
+ "rewards/margins": 0.7265963554382324,
1583
+ "rewards/rejected": -0.49028879404067993,
1584
+ "step": 75
1585
+ },
1586
+ {
1587
+ "debug/policy_chosen_logits": 24.752717971801758,
1588
+ "debug/policy_chosen_logps": -414.84014892578125,
1589
+ "debug/policy_rejected_logits": 25.71200942993164,
1590
+ "debug/policy_rejected_logps": -501.4398498535156,
1591
+ "debug/reference_chosen_logps": -441.64886474609375,
1592
+ "debug/reference_rejected_logps": -463.5169372558594,
1593
+ "epoch": 0.95,
1594
+ "grad_norm": 4.517327070548167,
1595
+ "learning_rate": 1e-06,
1596
+ "logits/chosen": 24.752717971801758,
1597
+ "logits/rejected": 25.71200942993164,
1598
+ "logps/chosen": -414.84014892578125,
1599
+ "logps/rejected": -501.4398498535156,
1600
+ "loss": 0.2746,
1601
+ "rewards/accuracies": 1.0,
1602
+ "rewards/chosen": 0.26808685064315796,
1603
+ "rewards/margins": 0.6473158597946167,
1604
+ "rewards/rejected": -0.37922903895378113,
1605
+ "step": 76
1606
+ },
1607
+ {
1608
+ "debug/policy_chosen_logits": 17.914918899536133,
1609
+ "debug/policy_chosen_logps": -399.0069274902344,
1610
+ "debug/policy_rejected_logits": 22.21956443786621,
1611
+ "debug/policy_rejected_logps": -473.9609375,
1612
+ "debug/reference_chosen_logps": -425.5627136230469,
1613
+ "debug/reference_rejected_logps": -450.96319580078125,
1614
+ "epoch": 0.9625,
1615
+ "grad_norm": 3.1357429365118166,
1616
+ "learning_rate": 1e-06,
1617
+ "logits/chosen": 17.914918899536133,
1618
+ "logits/rejected": 22.21956443786621,
1619
+ "logps/chosen": -399.0069274902344,
1620
+ "logps/rejected": -473.9609375,
1621
+ "loss": 0.1999,
1622
+ "rewards/accuracies": 1.0,
1623
+ "rewards/chosen": 0.26555800437927246,
1624
+ "rewards/margins": 0.4955354630947113,
1625
+ "rewards/rejected": -0.22997744381427765,
1626
+ "step": 77
1627
+ },
1628
+ {
1629
+ "debug/policy_chosen_logits": 18.327871322631836,
1630
+ "debug/policy_chosen_logps": -370.01171875,
1631
+ "debug/policy_rejected_logits": 28.05528450012207,
1632
+ "debug/policy_rejected_logps": -427.2958984375,
1633
+ "debug/reference_chosen_logps": -400.8182373046875,
1634
+ "debug/reference_rejected_logps": -420.2340393066406,
1635
+ "epoch": 0.975,
1636
+ "grad_norm": 4.140874548230489,
1637
+ "learning_rate": 1e-06,
1638
+ "logits/chosen": 18.327871322631836,
1639
+ "logits/rejected": 28.05528450012207,
1640
+ "logps/chosen": -370.01171875,
1641
+ "logps/rejected": -427.2958984375,
1642
+ "loss": 0.2728,
1643
+ "rewards/accuracies": 1.0,
1644
+ "rewards/chosen": 0.3080649673938751,
1645
+ "rewards/margins": 0.37868380546569824,
1646
+ "rewards/rejected": -0.07061880826950073,
1647
+ "step": 78
1648
+ },
1649
+ {
1650
+ "debug/policy_chosen_logits": 22.553178787231445,
1651
+ "debug/policy_chosen_logps": -408.61163330078125,
1652
+ "debug/policy_rejected_logits": 23.832901000976562,
1653
+ "debug/policy_rejected_logps": -502.69970703125,
1654
+ "debug/reference_chosen_logps": -414.4495544433594,
1655
+ "debug/reference_rejected_logps": -477.4999694824219,
1656
+ "epoch": 0.9875,
1657
+ "grad_norm": 3.8978087275053612,
1658
+ "learning_rate": 1e-06,
1659
+ "logits/chosen": 22.553178787231445,
1660
+ "logits/rejected": 23.832901000976562,
1661
+ "logps/chosen": -408.61163330078125,
1662
+ "logps/rejected": -502.69970703125,
1663
+ "loss": 0.271,
1664
+ "rewards/accuracies": 0.75,
1665
+ "rewards/chosen": 0.05837925150990486,
1666
+ "rewards/margins": 0.31037673354148865,
1667
+ "rewards/rejected": -0.2519974708557129,
1668
+ "step": 79
1669
+ },
1670
+ {
1671
+ "debug/policy_chosen_logits": 20.476898193359375,
1672
+ "debug/policy_chosen_logps": -401.72320556640625,
1673
+ "debug/policy_rejected_logits": 29.332805633544922,
1674
+ "debug/policy_rejected_logps": -478.8862609863281,
1675
+ "debug/reference_chosen_logps": -428.4986572265625,
1676
+ "debug/reference_rejected_logps": -465.5113220214844,
1677
+ "epoch": 1.0,
1678
+ "grad_norm": 4.733733434133006,
1679
+ "learning_rate": 1e-06,
1680
+ "logits/chosen": 20.476898193359375,
1681
+ "logits/rejected": 29.332805633544922,
1682
+ "logps/chosen": -401.72320556640625,
1683
+ "logps/rejected": -478.8862609863281,
1684
+ "loss": 0.2109,
1685
+ "rewards/accuracies": 0.875,
1686
+ "rewards/chosen": 0.267754465341568,
1687
+ "rewards/margins": 0.40150386095046997,
1688
+ "rewards/rejected": -0.13374939560890198,
1689
+ "step": 80
1690
+ },
1691
+ {
1692
+ "epoch": 1.0,
1693
+ "step": 80,
1694
+ "total_flos": 0.0,
1695
+ "train_loss": 0.32403188310563563,
1696
+ "train_runtime": 674.2321,
1697
+ "train_samples_per_second": 7.592,
1698
+ "train_steps_per_second": 0.119
1699
+ }
1700
+ ],
1701
+ "logging_steps": 1,
1702
+ "max_steps": 80,
1703
+ "num_input_tokens_seen": 0,
1704
+ "num_train_epochs": 1,
1705
+ "save_steps": 500,
1706
+ "stateful_callbacks": {
1707
+ "TrainerControl": {
1708
+ "args": {
1709
+ "should_epoch_stop": false,
1710
+ "should_evaluate": false,
1711
+ "should_log": false,
1712
+ "should_save": true,
1713
+ "should_training_stop": true
1714
+ },
1715
+ "attributes": {}
1716
+ }
1717
+ },
1718
+ "total_flos": 0.0,
1719
+ "train_batch_size": 8,
1720
+ "trial_name": null,
1721
+ "trial_params": null
1722
+ }