yiran-wang3 commited on
Commit
9e1bbb4
1 Parent(s): d4ea494

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 +1386 -0
README.md ADDED
@@ -0,0 +1,64 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ library_name: transformers
3
+ license: apache-2.0
4
+ base_model: yiran-wang3/qwen1_chat_adamw_iter1
5
+ tags:
6
+ - alignment-handbook
7
+ - generated_from_trainer
8
+ - trl
9
+ - dpo
10
+ datasets:
11
+ - self-generate/qw1_sppo_hard_new_cn_mining_oj_iter1-binarized
12
+ model-index:
13
+ - name: qwen1_chat_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
+ # qwen1_chat_adamw_iter2
21
+
22
+ This model is a fine-tuned version of [yiran-wang3/qwen1_chat_adamw_iter1](https://huggingface.co/yiran-wang3/qwen1_chat_adamw_iter1) on the self-generate/qw1_sppo_hard_new_cn_mining_oj_iter1-binarized 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": 0.46988651156425476,
5
+ "train_runtime": 197.5097,
6
+ "train_samples": 4074,
7
+ "train_samples_per_second": 20.627,
8
+ "train_steps_per_second": 0.324
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": 0.46988651156425476,
5
+ "train_runtime": 197.5097,
6
+ "train_samples": 4074,
7
+ "train_samples_per_second": 20.627,
8
+ "train_steps_per_second": 0.324
9
+ }
trainer_state.json ADDED
@@ -0,0 +1,1386 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 1.0,
5
+ "eval_steps": 100,
6
+ "global_step": 64,
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": -0.9758062958717346,
13
+ "debug/policy_chosen_logps": -151.53146362304688,
14
+ "debug/policy_rejected_logits": -0.9790539741516113,
15
+ "debug/policy_rejected_logps": -147.30738830566406,
16
+ "debug/reference_chosen_logps": -151.53146362304688,
17
+ "debug/reference_rejected_logps": -147.30738830566406,
18
+ "epoch": 0.015625,
19
+ "grad_norm": 7.077985294284826,
20
+ "learning_rate": 1e-06,
21
+ "logits/chosen": -0.9758062958717346,
22
+ "logits/rejected": -0.9790539741516113,
23
+ "logps/chosen": -151.53146362304688,
24
+ "logps/rejected": -147.30738830566406,
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": -0.9311444759368896,
34
+ "debug/policy_chosen_logps": -157.56246948242188,
35
+ "debug/policy_rejected_logits": -0.6852308511734009,
36
+ "debug/policy_rejected_logps": -181.45333862304688,
37
+ "debug/reference_chosen_logps": -157.81744384765625,
38
+ "debug/reference_rejected_logps": -181.56724548339844,
39
+ "epoch": 0.03125,
40
+ "grad_norm": 5.504247352373829,
41
+ "learning_rate": 1e-06,
42
+ "logits/chosen": -0.9311444759368896,
43
+ "logits/rejected": -0.6852308511734009,
44
+ "logps/chosen": -157.56246948242188,
45
+ "logps/rejected": -181.45333862304688,
46
+ "loss": 0.4999,
47
+ "rewards/accuracies": 0.75,
48
+ "rewards/chosen": 0.002549667377024889,
49
+ "rewards/margins": 0.0014108085306361318,
50
+ "rewards/rejected": 0.0011388587299734354,
51
+ "step": 2
52
+ },
53
+ {
54
+ "debug/policy_chosen_logits": -0.8895314335823059,
55
+ "debug/policy_chosen_logps": -195.25637817382812,
56
+ "debug/policy_rejected_logits": -0.575265109539032,
57
+ "debug/policy_rejected_logps": -194.01516723632812,
58
+ "debug/reference_chosen_logps": -195.29818725585938,
59
+ "debug/reference_rejected_logps": -194.30865478515625,
60
+ "epoch": 0.046875,
61
+ "grad_norm": 5.929427478772651,
62
+ "learning_rate": 1e-06,
63
+ "logits/chosen": -0.8895314335823059,
64
+ "logits/rejected": -0.575265109539032,
65
+ "logps/chosen": -195.25637817382812,
66
+ "logps/rejected": -194.01516723632812,
67
+ "loss": 0.5,
68
+ "rewards/accuracies": 0.5,
69
+ "rewards/chosen": 0.00041794776916503906,
70
+ "rewards/margins": -0.00251706107519567,
71
+ "rewards/rejected": 0.002935009077191353,
72
+ "step": 3
73
+ },
74
+ {
75
+ "debug/policy_chosen_logits": -1.1189931631088257,
76
+ "debug/policy_chosen_logps": -143.90708923339844,
77
+ "debug/policy_rejected_logits": -0.9135940670967102,
78
+ "debug/policy_rejected_logps": -157.21446228027344,
79
+ "debug/reference_chosen_logps": -144.0689697265625,
80
+ "debug/reference_rejected_logps": -156.4416046142578,
81
+ "epoch": 0.0625,
82
+ "grad_norm": 6.2120667679725585,
83
+ "learning_rate": 1e-06,
84
+ "logits/chosen": -1.1189931631088257,
85
+ "logits/rejected": -0.9135940670967102,
86
+ "logps/chosen": -143.90708923339844,
87
+ "logps/rejected": -157.21446228027344,
88
+ "loss": 0.4959,
89
+ "rewards/accuracies": 0.875,
90
+ "rewards/chosen": 0.0016189860180020332,
91
+ "rewards/margins": 0.009347562678158283,
92
+ "rewards/rejected": -0.00772857666015625,
93
+ "step": 4
94
+ },
95
+ {
96
+ "debug/policy_chosen_logits": -0.8267368674278259,
97
+ "debug/policy_chosen_logps": -183.93475341796875,
98
+ "debug/policy_rejected_logits": -0.761789858341217,
99
+ "debug/policy_rejected_logps": -177.76959228515625,
100
+ "debug/reference_chosen_logps": -183.55718994140625,
101
+ "debug/reference_rejected_logps": -176.53091430664062,
102
+ "epoch": 0.078125,
103
+ "grad_norm": 6.227908126221475,
104
+ "learning_rate": 1e-06,
105
+ "logits/chosen": -0.8267368674278259,
106
+ "logits/rejected": -0.761789858341217,
107
+ "logps/chosen": -183.93475341796875,
108
+ "logps/rejected": -177.76959228515625,
109
+ "loss": 0.4983,
110
+ "rewards/accuracies": 0.625,
111
+ "rewards/chosen": -0.0037756823003292084,
112
+ "rewards/margins": 0.008610926568508148,
113
+ "rewards/rejected": -0.012386607937514782,
114
+ "step": 5
115
+ },
116
+ {
117
+ "debug/policy_chosen_logits": -1.1983323097229004,
118
+ "debug/policy_chosen_logps": -134.5792236328125,
119
+ "debug/policy_rejected_logits": -1.1904683113098145,
120
+ "debug/policy_rejected_logps": -152.11712646484375,
121
+ "debug/reference_chosen_logps": -135.0609130859375,
122
+ "debug/reference_rejected_logps": -151.68475341796875,
123
+ "epoch": 0.09375,
124
+ "grad_norm": 6.006130436499243,
125
+ "learning_rate": 1e-06,
126
+ "logits/chosen": -1.1983323097229004,
127
+ "logits/rejected": -1.1904683113098145,
128
+ "logps/chosen": -134.5792236328125,
129
+ "logps/rejected": -152.11712646484375,
130
+ "loss": 0.4928,
131
+ "rewards/accuracies": 0.5,
132
+ "rewards/chosen": 0.004816893953830004,
133
+ "rewards/margins": 0.009140652604401112,
134
+ "rewards/rejected": -0.004323759116232395,
135
+ "step": 6
136
+ },
137
+ {
138
+ "debug/policy_chosen_logits": -1.0346052646636963,
139
+ "debug/policy_chosen_logps": -134.0786895751953,
140
+ "debug/policy_rejected_logits": -0.8808152079582214,
141
+ "debug/policy_rejected_logps": -155.0458984375,
142
+ "debug/reference_chosen_logps": -135.0550537109375,
143
+ "debug/reference_rejected_logps": -154.17298889160156,
144
+ "epoch": 0.109375,
145
+ "grad_norm": 5.440760637907165,
146
+ "learning_rate": 1e-06,
147
+ "logits/chosen": -1.0346052646636963,
148
+ "logits/rejected": -0.8808152079582214,
149
+ "logps/chosen": -134.0786895751953,
150
+ "logps/rejected": -155.0458984375,
151
+ "loss": 0.4961,
152
+ "rewards/accuracies": 0.5,
153
+ "rewards/chosen": 0.009763631038367748,
154
+ "rewards/margins": 0.018492689356207848,
155
+ "rewards/rejected": -0.008729057386517525,
156
+ "step": 7
157
+ },
158
+ {
159
+ "debug/policy_chosen_logits": -0.9342319369316101,
160
+ "debug/policy_chosen_logps": -137.67352294921875,
161
+ "debug/policy_rejected_logits": -0.9422162771224976,
162
+ "debug/policy_rejected_logps": -163.11819458007812,
163
+ "debug/reference_chosen_logps": -138.39669799804688,
164
+ "debug/reference_rejected_logps": -162.8925018310547,
165
+ "epoch": 0.125,
166
+ "grad_norm": 5.917503926417565,
167
+ "learning_rate": 1e-06,
168
+ "logits/chosen": -0.9342319369316101,
169
+ "logits/rejected": -0.9422162771224976,
170
+ "logps/chosen": -137.67352294921875,
171
+ "logps/rejected": -163.11819458007812,
172
+ "loss": 0.4951,
173
+ "rewards/accuracies": 0.75,
174
+ "rewards/chosen": 0.007231750525534153,
175
+ "rewards/margins": 0.009488582611083984,
176
+ "rewards/rejected": -0.002256832318380475,
177
+ "step": 8
178
+ },
179
+ {
180
+ "debug/policy_chosen_logits": -0.7332696914672852,
181
+ "debug/policy_chosen_logps": -188.66183471679688,
182
+ "debug/policy_rejected_logits": -0.5404833555221558,
183
+ "debug/policy_rejected_logps": -211.7197723388672,
184
+ "debug/reference_chosen_logps": -187.5482635498047,
185
+ "debug/reference_rejected_logps": -209.54415893554688,
186
+ "epoch": 0.140625,
187
+ "grad_norm": 5.9550182446271185,
188
+ "learning_rate": 1e-06,
189
+ "logits/chosen": -0.7332696914672852,
190
+ "logits/rejected": -0.5404833555221558,
191
+ "logps/chosen": -188.66183471679688,
192
+ "logps/rejected": -211.7197723388672,
193
+ "loss": 0.4949,
194
+ "rewards/accuracies": 0.625,
195
+ "rewards/chosen": -0.011135692708194256,
196
+ "rewards/margins": 0.010620327666401863,
197
+ "rewards/rejected": -0.021756019443273544,
198
+ "step": 9
199
+ },
200
+ {
201
+ "debug/policy_chosen_logits": -0.9110831022262573,
202
+ "debug/policy_chosen_logps": -148.4445343017578,
203
+ "debug/policy_rejected_logits": -0.8464857339859009,
204
+ "debug/policy_rejected_logps": -157.1292724609375,
205
+ "debug/reference_chosen_logps": -150.76031494140625,
206
+ "debug/reference_rejected_logps": -156.712890625,
207
+ "epoch": 0.15625,
208
+ "grad_norm": 5.99386820607358,
209
+ "learning_rate": 1e-06,
210
+ "logits/chosen": -0.9110831022262573,
211
+ "logits/rejected": -0.8464857339859009,
212
+ "logps/chosen": -148.4445343017578,
213
+ "logps/rejected": -157.1292724609375,
214
+ "loss": 0.492,
215
+ "rewards/accuracies": 0.625,
216
+ "rewards/chosen": 0.023157909512519836,
217
+ "rewards/margins": 0.027321862056851387,
218
+ "rewards/rejected": -0.004163951613008976,
219
+ "step": 10
220
+ },
221
+ {
222
+ "debug/policy_chosen_logits": -0.7407766580581665,
223
+ "debug/policy_chosen_logps": -179.63021850585938,
224
+ "debug/policy_rejected_logits": -0.8468393683433533,
225
+ "debug/policy_rejected_logps": -164.4591522216797,
226
+ "debug/reference_chosen_logps": -178.74200439453125,
227
+ "debug/reference_rejected_logps": -164.0178680419922,
228
+ "epoch": 0.171875,
229
+ "grad_norm": 6.56332815982213,
230
+ "learning_rate": 1e-06,
231
+ "logits/chosen": -0.7407766580581665,
232
+ "logits/rejected": -0.8468393683433533,
233
+ "logps/chosen": -179.63021850585938,
234
+ "logps/rejected": -164.4591522216797,
235
+ "loss": 0.4958,
236
+ "rewards/accuracies": 0.375,
237
+ "rewards/chosen": -0.008881940506398678,
238
+ "rewards/margins": -0.0044690510258078575,
239
+ "rewards/rejected": -0.004412889014929533,
240
+ "step": 11
241
+ },
242
+ {
243
+ "debug/policy_chosen_logits": -0.8543327450752258,
244
+ "debug/policy_chosen_logps": -173.95263671875,
245
+ "debug/policy_rejected_logits": -0.7469156980514526,
246
+ "debug/policy_rejected_logps": -187.60519409179688,
247
+ "debug/reference_chosen_logps": -174.4696502685547,
248
+ "debug/reference_rejected_logps": -184.88877868652344,
249
+ "epoch": 0.1875,
250
+ "grad_norm": 5.8252129211333825,
251
+ "learning_rate": 1e-06,
252
+ "logits/chosen": -0.8543327450752258,
253
+ "logits/rejected": -0.7469156980514526,
254
+ "logps/chosen": -173.95263671875,
255
+ "logps/rejected": -187.60519409179688,
256
+ "loss": 0.497,
257
+ "rewards/accuracies": 1.0,
258
+ "rewards/chosen": 0.005170030519366264,
259
+ "rewards/margins": 0.03233422338962555,
260
+ "rewards/rejected": -0.027164191007614136,
261
+ "step": 12
262
+ },
263
+ {
264
+ "debug/policy_chosen_logits": -0.8461030125617981,
265
+ "debug/policy_chosen_logps": -143.32421875,
266
+ "debug/policy_rejected_logits": -0.8543170690536499,
267
+ "debug/policy_rejected_logps": -186.77857971191406,
268
+ "debug/reference_chosen_logps": -146.84774780273438,
269
+ "debug/reference_rejected_logps": -186.40374755859375,
270
+ "epoch": 0.203125,
271
+ "grad_norm": 5.76637069916284,
272
+ "learning_rate": 1e-06,
273
+ "logits/chosen": -0.8461030125617981,
274
+ "logits/rejected": -0.8543170690536499,
275
+ "logps/chosen": -143.32421875,
276
+ "logps/rejected": -186.77857971191406,
277
+ "loss": 0.4832,
278
+ "rewards/accuracies": 0.75,
279
+ "rewards/chosen": 0.03523515537381172,
280
+ "rewards/margins": 0.03898348659276962,
281
+ "rewards/rejected": -0.0037483316846191883,
282
+ "step": 13
283
+ },
284
+ {
285
+ "debug/policy_chosen_logits": -0.9689863920211792,
286
+ "debug/policy_chosen_logps": -154.18991088867188,
287
+ "debug/policy_rejected_logits": -0.6996111869812012,
288
+ "debug/policy_rejected_logps": -168.9903564453125,
289
+ "debug/reference_chosen_logps": -157.5633544921875,
290
+ "debug/reference_rejected_logps": -168.59417724609375,
291
+ "epoch": 0.21875,
292
+ "grad_norm": 5.591725227630854,
293
+ "learning_rate": 1e-06,
294
+ "logits/chosen": -0.9689863920211792,
295
+ "logits/rejected": -0.6996111869812012,
296
+ "logps/chosen": -154.18991088867188,
297
+ "logps/rejected": -168.9903564453125,
298
+ "loss": 0.4893,
299
+ "rewards/accuracies": 1.0,
300
+ "rewards/chosen": 0.0337345227599144,
301
+ "rewards/margins": 0.03769625723361969,
302
+ "rewards/rejected": -0.003961734939366579,
303
+ "step": 14
304
+ },
305
+ {
306
+ "debug/policy_chosen_logits": -0.9054354429244995,
307
+ "debug/policy_chosen_logps": -145.04635620117188,
308
+ "debug/policy_rejected_logits": -0.8977835774421692,
309
+ "debug/policy_rejected_logps": -163.13455200195312,
310
+ "debug/reference_chosen_logps": -147.23922729492188,
311
+ "debug/reference_rejected_logps": -162.36553955078125,
312
+ "epoch": 0.234375,
313
+ "grad_norm": 6.04155713949948,
314
+ "learning_rate": 1e-06,
315
+ "logits/chosen": -0.9054354429244995,
316
+ "logits/rejected": -0.8977835774421692,
317
+ "logps/chosen": -145.04635620117188,
318
+ "logps/rejected": -163.13455200195312,
319
+ "loss": 0.4855,
320
+ "rewards/accuracies": 0.625,
321
+ "rewards/chosen": 0.0219286996871233,
322
+ "rewards/margins": 0.02961883321404457,
323
+ "rewards/rejected": -0.00769013399258256,
324
+ "step": 15
325
+ },
326
+ {
327
+ "debug/policy_chosen_logits": -0.8554494380950928,
328
+ "debug/policy_chosen_logps": -161.84022521972656,
329
+ "debug/policy_rejected_logits": -0.8830718994140625,
330
+ "debug/policy_rejected_logps": -173.75918579101562,
331
+ "debug/reference_chosen_logps": -164.03298950195312,
332
+ "debug/reference_rejected_logps": -174.45535278320312,
333
+ "epoch": 0.25,
334
+ "grad_norm": 5.970121851694702,
335
+ "learning_rate": 1e-06,
336
+ "logits/chosen": -0.8554494380950928,
337
+ "logits/rejected": -0.8830718994140625,
338
+ "logps/chosen": -161.84022521972656,
339
+ "logps/rejected": -173.75918579101562,
340
+ "loss": 0.4842,
341
+ "rewards/accuracies": 0.75,
342
+ "rewards/chosen": 0.021927593275904655,
343
+ "rewards/margins": 0.014965922571718693,
344
+ "rewards/rejected": 0.006961670238524675,
345
+ "step": 16
346
+ },
347
+ {
348
+ "debug/policy_chosen_logits": -0.9481765627861023,
349
+ "debug/policy_chosen_logps": -162.32388305664062,
350
+ "debug/policy_rejected_logits": -0.7944669127464294,
351
+ "debug/policy_rejected_logps": -188.93453979492188,
352
+ "debug/reference_chosen_logps": -163.7265625,
353
+ "debug/reference_rejected_logps": -185.91177368164062,
354
+ "epoch": 0.265625,
355
+ "grad_norm": 5.71178003145209,
356
+ "learning_rate": 1e-06,
357
+ "logits/chosen": -0.9481765627861023,
358
+ "logits/rejected": -0.7944669127464294,
359
+ "logps/chosen": -162.32388305664062,
360
+ "logps/rejected": -188.93453979492188,
361
+ "loss": 0.4899,
362
+ "rewards/accuracies": 1.0,
363
+ "rewards/chosen": 0.014026792719960213,
364
+ "rewards/margins": 0.044254451990127563,
365
+ "rewards/rejected": -0.0302276611328125,
366
+ "step": 17
367
+ },
368
+ {
369
+ "debug/policy_chosen_logits": -0.6716693043708801,
370
+ "debug/policy_chosen_logps": -187.95364379882812,
371
+ "debug/policy_rejected_logits": -0.8150543570518494,
372
+ "debug/policy_rejected_logps": -169.115478515625,
373
+ "debug/reference_chosen_logps": -187.42266845703125,
374
+ "debug/reference_rejected_logps": -166.06893920898438,
375
+ "epoch": 0.28125,
376
+ "grad_norm": 6.343140969572851,
377
+ "learning_rate": 1e-06,
378
+ "logits/chosen": -0.6716693043708801,
379
+ "logits/rejected": -0.8150543570518494,
380
+ "logps/chosen": -187.95364379882812,
381
+ "logps/rejected": -169.115478515625,
382
+ "loss": 0.4942,
383
+ "rewards/accuracies": 0.75,
384
+ "rewards/chosen": -0.005309591069817543,
385
+ "rewards/margins": 0.025155849754810333,
386
+ "rewards/rejected": -0.030465438961982727,
387
+ "step": 18
388
+ },
389
+ {
390
+ "debug/policy_chosen_logits": -0.9492964744567871,
391
+ "debug/policy_chosen_logps": -158.28807067871094,
392
+ "debug/policy_rejected_logits": -0.8160263895988464,
393
+ "debug/policy_rejected_logps": -167.84140014648438,
394
+ "debug/reference_chosen_logps": -159.18081665039062,
395
+ "debug/reference_rejected_logps": -166.4847869873047,
396
+ "epoch": 0.296875,
397
+ "grad_norm": 5.7981298310331795,
398
+ "learning_rate": 1e-06,
399
+ "logits/chosen": -0.9492964744567871,
400
+ "logits/rejected": -0.8160263895988464,
401
+ "logps/chosen": -158.28807067871094,
402
+ "logps/rejected": -167.84140014648438,
403
+ "loss": 0.4863,
404
+ "rewards/accuracies": 0.75,
405
+ "rewards/chosen": 0.008927486836910248,
406
+ "rewards/margins": 0.022493677213788033,
407
+ "rewards/rejected": -0.013566188514232635,
408
+ "step": 19
409
+ },
410
+ {
411
+ "debug/policy_chosen_logits": -0.9160640239715576,
412
+ "debug/policy_chosen_logps": -165.8433837890625,
413
+ "debug/policy_rejected_logits": -0.8921107053756714,
414
+ "debug/policy_rejected_logps": -172.75804138183594,
415
+ "debug/reference_chosen_logps": -164.21383666992188,
416
+ "debug/reference_rejected_logps": -168.41940307617188,
417
+ "epoch": 0.3125,
418
+ "grad_norm": 6.672971626135043,
419
+ "learning_rate": 1e-06,
420
+ "logits/chosen": -0.9160640239715576,
421
+ "logits/rejected": -0.8921107053756714,
422
+ "logps/chosen": -165.8433837890625,
423
+ "logps/rejected": -172.75804138183594,
424
+ "loss": 0.4778,
425
+ "rewards/accuracies": 0.75,
426
+ "rewards/chosen": -0.016295330598950386,
427
+ "rewards/margins": 0.027090895920991898,
428
+ "rewards/rejected": -0.04338622838258743,
429
+ "step": 20
430
+ },
431
+ {
432
+ "debug/policy_chosen_logits": -0.6745712161064148,
433
+ "debug/policy_chosen_logps": -169.40414428710938,
434
+ "debug/policy_rejected_logits": -0.8499138355255127,
435
+ "debug/policy_rejected_logps": -174.95440673828125,
436
+ "debug/reference_chosen_logps": -169.4923095703125,
437
+ "debug/reference_rejected_logps": -175.6483154296875,
438
+ "epoch": 0.328125,
439
+ "grad_norm": 5.673305428723763,
440
+ "learning_rate": 1e-06,
441
+ "logits/chosen": -0.6745712161064148,
442
+ "logits/rejected": -0.8499138355255127,
443
+ "logps/chosen": -169.40414428710938,
444
+ "logps/rejected": -174.95440673828125,
445
+ "loss": 0.4998,
446
+ "rewards/accuracies": 0.375,
447
+ "rewards/chosen": 0.0008817007765173912,
448
+ "rewards/margins": -0.006057358346879482,
449
+ "rewards/rejected": 0.0069390591233968735,
450
+ "step": 21
451
+ },
452
+ {
453
+ "debug/policy_chosen_logits": -0.7490401268005371,
454
+ "debug/policy_chosen_logps": -180.66146850585938,
455
+ "debug/policy_rejected_logits": -0.6564124226570129,
456
+ "debug/policy_rejected_logps": -195.39569091796875,
457
+ "debug/reference_chosen_logps": -184.63441467285156,
458
+ "debug/reference_rejected_logps": -196.98995971679688,
459
+ "epoch": 0.34375,
460
+ "grad_norm": 6.449944674945484,
461
+ "learning_rate": 1e-06,
462
+ "logits/chosen": -0.7490401268005371,
463
+ "logits/rejected": -0.6564124226570129,
464
+ "logps/chosen": -180.66146850585938,
465
+ "logps/rejected": -195.39569091796875,
466
+ "loss": 0.4655,
467
+ "rewards/accuracies": 0.625,
468
+ "rewards/chosen": 0.03972943127155304,
469
+ "rewards/margins": 0.02378678321838379,
470
+ "rewards/rejected": 0.0159426499158144,
471
+ "step": 22
472
+ },
473
+ {
474
+ "debug/policy_chosen_logits": -0.6730453372001648,
475
+ "debug/policy_chosen_logps": -178.2505340576172,
476
+ "debug/policy_rejected_logits": -0.6955782175064087,
477
+ "debug/policy_rejected_logps": -174.2504119873047,
478
+ "debug/reference_chosen_logps": -182.0872802734375,
479
+ "debug/reference_rejected_logps": -175.9362030029297,
480
+ "epoch": 0.359375,
481
+ "grad_norm": 6.173388994966317,
482
+ "learning_rate": 1e-06,
483
+ "logits/chosen": -0.6730453372001648,
484
+ "logits/rejected": -0.6955782175064087,
485
+ "logps/chosen": -178.2505340576172,
486
+ "logps/rejected": -174.2504119873047,
487
+ "loss": 0.4747,
488
+ "rewards/accuracies": 0.5,
489
+ "rewards/chosen": 0.038367509841918945,
490
+ "rewards/margins": 0.021509580314159393,
491
+ "rewards/rejected": 0.016857929527759552,
492
+ "step": 23
493
+ },
494
+ {
495
+ "debug/policy_chosen_logits": -0.8440623879432678,
496
+ "debug/policy_chosen_logps": -158.67828369140625,
497
+ "debug/policy_rejected_logits": -0.8467646837234497,
498
+ "debug/policy_rejected_logps": -173.41604614257812,
499
+ "debug/reference_chosen_logps": -160.15481567382812,
500
+ "debug/reference_rejected_logps": -172.91629028320312,
501
+ "epoch": 0.375,
502
+ "grad_norm": 6.519435131661764,
503
+ "learning_rate": 1e-06,
504
+ "logits/chosen": -0.8440623879432678,
505
+ "logits/rejected": -0.8467646837234497,
506
+ "logps/chosen": -158.67828369140625,
507
+ "logps/rejected": -173.41604614257812,
508
+ "loss": 0.4892,
509
+ "rewards/accuracies": 0.75,
510
+ "rewards/chosen": 0.0147654814645648,
511
+ "rewards/margins": 0.01976308785378933,
512
+ "rewards/rejected": -0.004997607786208391,
513
+ "step": 24
514
+ },
515
+ {
516
+ "debug/policy_chosen_logits": -0.6752190589904785,
517
+ "debug/policy_chosen_logps": -187.1298065185547,
518
+ "debug/policy_rejected_logits": -0.7732603549957275,
519
+ "debug/policy_rejected_logps": -158.61585998535156,
520
+ "debug/reference_chosen_logps": -190.5185089111328,
521
+ "debug/reference_rejected_logps": -159.48568725585938,
522
+ "epoch": 0.390625,
523
+ "grad_norm": 5.353977426256108,
524
+ "learning_rate": 1e-06,
525
+ "logits/chosen": -0.6752190589904785,
526
+ "logits/rejected": -0.7732603549957275,
527
+ "logps/chosen": -187.1298065185547,
528
+ "logps/rejected": -158.61585998535156,
529
+ "loss": 0.4656,
530
+ "rewards/accuracies": 0.75,
531
+ "rewards/chosen": 0.03388698399066925,
532
+ "rewards/margins": 0.025188742205500603,
533
+ "rewards/rejected": 0.008698243647813797,
534
+ "step": 25
535
+ },
536
+ {
537
+ "debug/policy_chosen_logits": -0.932550311088562,
538
+ "debug/policy_chosen_logps": -157.26089477539062,
539
+ "debug/policy_rejected_logits": -0.8474171161651611,
540
+ "debug/policy_rejected_logps": -163.1019744873047,
541
+ "debug/reference_chosen_logps": -154.3992919921875,
542
+ "debug/reference_rejected_logps": -162.41476440429688,
543
+ "epoch": 0.40625,
544
+ "grad_norm": 6.6664666902082494,
545
+ "learning_rate": 1e-06,
546
+ "logits/chosen": -0.932550311088562,
547
+ "logits/rejected": -0.8474171161651611,
548
+ "logps/chosen": -157.26089477539062,
549
+ "logps/rejected": -163.1019744873047,
550
+ "loss": 0.4883,
551
+ "rewards/accuracies": 0.5,
552
+ "rewards/chosen": -0.02861594967544079,
553
+ "rewards/margins": -0.021743860095739365,
554
+ "rewards/rejected": -0.006872090511023998,
555
+ "step": 26
556
+ },
557
+ {
558
+ "debug/policy_chosen_logits": -0.9092602133750916,
559
+ "debug/policy_chosen_logps": -178.76705932617188,
560
+ "debug/policy_rejected_logits": -0.8668628334999084,
561
+ "debug/policy_rejected_logps": -176.48455810546875,
562
+ "debug/reference_chosen_logps": -178.7393798828125,
563
+ "debug/reference_rejected_logps": -174.53854370117188,
564
+ "epoch": 0.421875,
565
+ "grad_norm": 6.321703911878928,
566
+ "learning_rate": 1e-06,
567
+ "logits/chosen": -0.9092602133750916,
568
+ "logits/rejected": -0.8668628334999084,
569
+ "logps/chosen": -178.76705932617188,
570
+ "logps/rejected": -176.48455810546875,
571
+ "loss": 0.4628,
572
+ "rewards/accuracies": 0.75,
573
+ "rewards/chosen": -0.00027678441256284714,
574
+ "rewards/margins": 0.01918334886431694,
575
+ "rewards/rejected": -0.019460134208202362,
576
+ "step": 27
577
+ },
578
+ {
579
+ "debug/policy_chosen_logits": -1.1956357955932617,
580
+ "debug/policy_chosen_logps": -122.40371704101562,
581
+ "debug/policy_rejected_logits": -0.9366591572761536,
582
+ "debug/policy_rejected_logps": -168.29766845703125,
583
+ "debug/reference_chosen_logps": -127.96609497070312,
584
+ "debug/reference_rejected_logps": -166.14305114746094,
585
+ "epoch": 0.4375,
586
+ "grad_norm": 5.594609963042074,
587
+ "learning_rate": 1e-06,
588
+ "logits/chosen": -1.1956357955932617,
589
+ "logits/rejected": -0.9366591572761536,
590
+ "logps/chosen": -122.40371704101562,
591
+ "logps/rejected": -168.29766845703125,
592
+ "loss": 0.4683,
593
+ "rewards/accuracies": 0.625,
594
+ "rewards/chosen": 0.05562380701303482,
595
+ "rewards/margins": 0.07716996222734451,
596
+ "rewards/rejected": -0.021546155214309692,
597
+ "step": 28
598
+ },
599
+ {
600
+ "debug/policy_chosen_logits": -0.8461283445358276,
601
+ "debug/policy_chosen_logps": -166.58126831054688,
602
+ "debug/policy_rejected_logits": -0.854517936706543,
603
+ "debug/policy_rejected_logps": -160.43394470214844,
604
+ "debug/reference_chosen_logps": -167.19557189941406,
605
+ "debug/reference_rejected_logps": -156.71856689453125,
606
+ "epoch": 0.453125,
607
+ "grad_norm": 5.913333766169721,
608
+ "learning_rate": 1e-06,
609
+ "logits/chosen": -0.8461283445358276,
610
+ "logits/rejected": -0.854517936706543,
611
+ "logps/chosen": -166.58126831054688,
612
+ "logps/rejected": -160.43394470214844,
613
+ "loss": 0.4895,
614
+ "rewards/accuracies": 0.625,
615
+ "rewards/chosen": 0.006143084727227688,
616
+ "rewards/margins": 0.04329695552587509,
617
+ "rewards/rejected": -0.03715386986732483,
618
+ "step": 29
619
+ },
620
+ {
621
+ "debug/policy_chosen_logits": -0.9404685497283936,
622
+ "debug/policy_chosen_logps": -169.5876922607422,
623
+ "debug/policy_rejected_logits": -0.7541916966438293,
624
+ "debug/policy_rejected_logps": -193.78857421875,
625
+ "debug/reference_chosen_logps": -177.3969268798828,
626
+ "debug/reference_rejected_logps": -190.34381103515625,
627
+ "epoch": 0.46875,
628
+ "grad_norm": 5.307805768669272,
629
+ "learning_rate": 1e-06,
630
+ "logits/chosen": -0.9404685497283936,
631
+ "logits/rejected": -0.7541916966438293,
632
+ "logps/chosen": -169.5876922607422,
633
+ "logps/rejected": -193.78857421875,
634
+ "loss": 0.4504,
635
+ "rewards/accuracies": 0.875,
636
+ "rewards/chosen": 0.07809236645698547,
637
+ "rewards/margins": 0.11254002153873444,
638
+ "rewards/rejected": -0.034447651356458664,
639
+ "step": 30
640
+ },
641
+ {
642
+ "debug/policy_chosen_logits": -0.9712099432945251,
643
+ "debug/policy_chosen_logps": -148.6423797607422,
644
+ "debug/policy_rejected_logits": -0.7749654650688171,
645
+ "debug/policy_rejected_logps": -186.98977661132812,
646
+ "debug/reference_chosen_logps": -149.51290893554688,
647
+ "debug/reference_rejected_logps": -178.90377807617188,
648
+ "epoch": 0.484375,
649
+ "grad_norm": 6.072069194174168,
650
+ "learning_rate": 1e-06,
651
+ "logits/chosen": -0.9712099432945251,
652
+ "logits/rejected": -0.7749654650688171,
653
+ "logps/chosen": -148.6423797607422,
654
+ "logps/rejected": -186.98977661132812,
655
+ "loss": 0.4551,
656
+ "rewards/accuracies": 0.875,
657
+ "rewards/chosen": 0.008705183863639832,
658
+ "rewards/margins": 0.08956518769264221,
659
+ "rewards/rejected": -0.08086000382900238,
660
+ "step": 31
661
+ },
662
+ {
663
+ "debug/policy_chosen_logits": -0.7779644131660461,
664
+ "debug/policy_chosen_logps": -181.0400390625,
665
+ "debug/policy_rejected_logits": -0.6311701536178589,
666
+ "debug/policy_rejected_logps": -197.38998413085938,
667
+ "debug/reference_chosen_logps": -182.41529846191406,
668
+ "debug/reference_rejected_logps": -193.42015075683594,
669
+ "epoch": 0.5,
670
+ "grad_norm": 5.6393792651737495,
671
+ "learning_rate": 1e-06,
672
+ "logits/chosen": -0.7779644131660461,
673
+ "logits/rejected": -0.6311701536178589,
674
+ "logps/chosen": -181.0400390625,
675
+ "logps/rejected": -197.38998413085938,
676
+ "loss": 0.4724,
677
+ "rewards/accuracies": 0.75,
678
+ "rewards/chosen": 0.013752726837992668,
679
+ "rewards/margins": 0.05345122888684273,
680
+ "rewards/rejected": -0.03969850391149521,
681
+ "step": 32
682
+ },
683
+ {
684
+ "debug/policy_chosen_logits": -0.9899529218673706,
685
+ "debug/policy_chosen_logps": -140.02195739746094,
686
+ "debug/policy_rejected_logits": -0.805591344833374,
687
+ "debug/policy_rejected_logps": -171.04238891601562,
688
+ "debug/reference_chosen_logps": -142.90939331054688,
689
+ "debug/reference_rejected_logps": -164.79971313476562,
690
+ "epoch": 0.515625,
691
+ "grad_norm": 6.65298054287674,
692
+ "learning_rate": 1e-06,
693
+ "logits/chosen": -0.9899529218673706,
694
+ "logits/rejected": -0.805591344833374,
695
+ "logps/chosen": -140.02195739746094,
696
+ "logps/rejected": -171.04238891601562,
697
+ "loss": 0.4449,
698
+ "rewards/accuracies": 0.625,
699
+ "rewards/chosen": 0.028874464333057404,
700
+ "rewards/margins": 0.09130106121301651,
701
+ "rewards/rejected": -0.06242658942937851,
702
+ "step": 33
703
+ },
704
+ {
705
+ "debug/policy_chosen_logits": -0.8487688302993774,
706
+ "debug/policy_chosen_logps": -150.60659790039062,
707
+ "debug/policy_rejected_logits": -0.7607054710388184,
708
+ "debug/policy_rejected_logps": -206.38253784179688,
709
+ "debug/reference_chosen_logps": -152.25167846679688,
710
+ "debug/reference_rejected_logps": -200.65679931640625,
711
+ "epoch": 0.53125,
712
+ "grad_norm": 6.500705317924212,
713
+ "learning_rate": 1e-06,
714
+ "logits/chosen": -0.8487688302993774,
715
+ "logits/rejected": -0.7607054710388184,
716
+ "logps/chosen": -150.60659790039062,
717
+ "logps/rejected": -206.38253784179688,
718
+ "loss": 0.4744,
719
+ "rewards/accuracies": 0.5,
720
+ "rewards/chosen": 0.01645086146891117,
721
+ "rewards/margins": 0.07370824366807938,
722
+ "rewards/rejected": -0.057257384061813354,
723
+ "step": 34
724
+ },
725
+ {
726
+ "debug/policy_chosen_logits": -0.7768784165382385,
727
+ "debug/policy_chosen_logps": -152.84442138671875,
728
+ "debug/policy_rejected_logits": -0.8615243434906006,
729
+ "debug/policy_rejected_logps": -178.1951904296875,
730
+ "debug/reference_chosen_logps": -156.7847900390625,
731
+ "debug/reference_rejected_logps": -175.36306762695312,
732
+ "epoch": 0.546875,
733
+ "grad_norm": 6.2342972742178615,
734
+ "learning_rate": 1e-06,
735
+ "logits/chosen": -0.7768784165382385,
736
+ "logits/rejected": -0.8615243434906006,
737
+ "logps/chosen": -152.84442138671875,
738
+ "logps/rejected": -178.1951904296875,
739
+ "loss": 0.4689,
740
+ "rewards/accuracies": 0.625,
741
+ "rewards/chosen": 0.03940363600850105,
742
+ "rewards/margins": 0.06772496551275253,
743
+ "rewards/rejected": -0.02832133322954178,
744
+ "step": 35
745
+ },
746
+ {
747
+ "debug/policy_chosen_logits": -1.0841803550720215,
748
+ "debug/policy_chosen_logps": -132.16390991210938,
749
+ "debug/policy_rejected_logits": -0.8767221570014954,
750
+ "debug/policy_rejected_logps": -177.02395629882812,
751
+ "debug/reference_chosen_logps": -141.05548095703125,
752
+ "debug/reference_rejected_logps": -180.10845947265625,
753
+ "epoch": 0.5625,
754
+ "grad_norm": 6.118058488612106,
755
+ "learning_rate": 1e-06,
756
+ "logits/chosen": -1.0841803550720215,
757
+ "logits/rejected": -0.8767221570014954,
758
+ "logps/chosen": -132.16390991210938,
759
+ "logps/rejected": -177.02395629882812,
760
+ "loss": 0.4562,
761
+ "rewards/accuracies": 0.75,
762
+ "rewards/chosen": 0.08891567587852478,
763
+ "rewards/margins": 0.05807068198919296,
764
+ "rewards/rejected": 0.030844993889331818,
765
+ "step": 36
766
+ },
767
+ {
768
+ "debug/policy_chosen_logits": -0.8277906179428101,
769
+ "debug/policy_chosen_logps": -160.42657470703125,
770
+ "debug/policy_rejected_logits": -0.9817237257957458,
771
+ "debug/policy_rejected_logps": -175.60308837890625,
772
+ "debug/reference_chosen_logps": -165.00286865234375,
773
+ "debug/reference_rejected_logps": -175.93499755859375,
774
+ "epoch": 0.578125,
775
+ "grad_norm": 8.456361293746621,
776
+ "learning_rate": 1e-06,
777
+ "logits/chosen": -0.8277906179428101,
778
+ "logits/rejected": -0.9817237257957458,
779
+ "logps/chosen": -160.42657470703125,
780
+ "logps/rejected": -175.60308837890625,
781
+ "loss": 0.4951,
782
+ "rewards/accuracies": 0.5,
783
+ "rewards/chosen": 0.0457630455493927,
784
+ "rewards/margins": 0.042444050312042236,
785
+ "rewards/rejected": 0.003318987786769867,
786
+ "step": 37
787
+ },
788
+ {
789
+ "debug/policy_chosen_logits": -0.7899513244628906,
790
+ "debug/policy_chosen_logps": -169.28268432617188,
791
+ "debug/policy_rejected_logits": -0.8575571179389954,
792
+ "debug/policy_rejected_logps": -171.0587921142578,
793
+ "debug/reference_chosen_logps": -172.85174560546875,
794
+ "debug/reference_rejected_logps": -171.73057556152344,
795
+ "epoch": 0.59375,
796
+ "grad_norm": 7.396146328689038,
797
+ "learning_rate": 1e-06,
798
+ "logits/chosen": -0.7899513244628906,
799
+ "logits/rejected": -0.8575571179389954,
800
+ "logps/chosen": -169.28268432617188,
801
+ "logps/rejected": -171.0587921142578,
802
+ "loss": 0.4858,
803
+ "rewards/accuracies": 0.625,
804
+ "rewards/chosen": 0.0356905460357666,
805
+ "rewards/margins": 0.028972698375582695,
806
+ "rewards/rejected": 0.006717845797538757,
807
+ "step": 38
808
+ },
809
+ {
810
+ "debug/policy_chosen_logits": -0.9996007680892944,
811
+ "debug/policy_chosen_logps": -147.13336181640625,
812
+ "debug/policy_rejected_logits": -0.7932425737380981,
813
+ "debug/policy_rejected_logps": -184.53997802734375,
814
+ "debug/reference_chosen_logps": -149.77064514160156,
815
+ "debug/reference_rejected_logps": -183.35354614257812,
816
+ "epoch": 0.609375,
817
+ "grad_norm": 5.981284123668312,
818
+ "learning_rate": 1e-06,
819
+ "logits/chosen": -0.9996007680892944,
820
+ "logits/rejected": -0.7932425737380981,
821
+ "logps/chosen": -147.13336181640625,
822
+ "logps/rejected": -184.53997802734375,
823
+ "loss": 0.4553,
824
+ "rewards/accuracies": 0.75,
825
+ "rewards/chosen": 0.02637273073196411,
826
+ "rewards/margins": 0.03823715075850487,
827
+ "rewards/rejected": -0.011864423751831055,
828
+ "step": 39
829
+ },
830
+ {
831
+ "debug/policy_chosen_logits": -0.7280028462409973,
832
+ "debug/policy_chosen_logps": -176.14102172851562,
833
+ "debug/policy_rejected_logits": -0.6034502387046814,
834
+ "debug/policy_rejected_logps": -213.32142639160156,
835
+ "debug/reference_chosen_logps": -169.86448669433594,
836
+ "debug/reference_rejected_logps": -208.47547912597656,
837
+ "epoch": 0.625,
838
+ "grad_norm": 8.108956880586513,
839
+ "learning_rate": 1e-06,
840
+ "logits/chosen": -0.7280028462409973,
841
+ "logits/rejected": -0.6034502387046814,
842
+ "logps/chosen": -176.14102172851562,
843
+ "logps/rejected": -213.32142639160156,
844
+ "loss": 0.4908,
845
+ "rewards/accuracies": 0.5,
846
+ "rewards/chosen": -0.06276529282331467,
847
+ "rewards/margins": -0.014305687509477139,
848
+ "rewards/rejected": -0.048459604382514954,
849
+ "step": 40
850
+ },
851
+ {
852
+ "debug/policy_chosen_logits": -1.0434032678604126,
853
+ "debug/policy_chosen_logps": -155.97015380859375,
854
+ "debug/policy_rejected_logits": -0.8284803032875061,
855
+ "debug/policy_rejected_logps": -191.99710083007812,
856
+ "debug/reference_chosen_logps": -157.0880584716797,
857
+ "debug/reference_rejected_logps": -182.62484741210938,
858
+ "epoch": 0.640625,
859
+ "grad_norm": 7.874523792109847,
860
+ "learning_rate": 1e-06,
861
+ "logits/chosen": -1.0434032678604126,
862
+ "logits/rejected": -0.8284803032875061,
863
+ "logps/chosen": -155.97015380859375,
864
+ "logps/rejected": -191.99710083007812,
865
+ "loss": 0.4365,
866
+ "rewards/accuracies": 0.75,
867
+ "rewards/chosen": 0.011179141700267792,
868
+ "rewards/margins": 0.10490170121192932,
869
+ "rewards/rejected": -0.09372256696224213,
870
+ "step": 41
871
+ },
872
+ {
873
+ "debug/policy_chosen_logits": -0.588237464427948,
874
+ "debug/policy_chosen_logps": -166.4864959716797,
875
+ "debug/policy_rejected_logits": -0.7331523299217224,
876
+ "debug/policy_rejected_logps": -181.28952026367188,
877
+ "debug/reference_chosen_logps": -166.72225952148438,
878
+ "debug/reference_rejected_logps": -174.8398895263672,
879
+ "epoch": 0.65625,
880
+ "grad_norm": 6.8274046949424045,
881
+ "learning_rate": 1e-06,
882
+ "logits/chosen": -0.588237464427948,
883
+ "logits/rejected": -0.7331523299217224,
884
+ "logps/chosen": -166.4864959716797,
885
+ "logps/rejected": -181.28952026367188,
886
+ "loss": 0.4523,
887
+ "rewards/accuracies": 0.625,
888
+ "rewards/chosen": 0.00235767406411469,
889
+ "rewards/margins": 0.06685390323400497,
890
+ "rewards/rejected": -0.0644962340593338,
891
+ "step": 42
892
+ },
893
+ {
894
+ "debug/policy_chosen_logits": -0.8734938502311707,
895
+ "debug/policy_chosen_logps": -129.81631469726562,
896
+ "debug/policy_rejected_logits": -0.879021942615509,
897
+ "debug/policy_rejected_logps": -151.5695037841797,
898
+ "debug/reference_chosen_logps": -132.09263610839844,
899
+ "debug/reference_rejected_logps": -146.77114868164062,
900
+ "epoch": 0.671875,
901
+ "grad_norm": 6.18097446051221,
902
+ "learning_rate": 1e-06,
903
+ "logits/chosen": -0.8734938502311707,
904
+ "logits/rejected": -0.879021942615509,
905
+ "logps/chosen": -129.81631469726562,
906
+ "logps/rejected": -151.5695037841797,
907
+ "loss": 0.4407,
908
+ "rewards/accuracies": 0.625,
909
+ "rewards/chosen": 0.022763298824429512,
910
+ "rewards/margins": 0.07074688374996185,
911
+ "rewards/rejected": -0.04798358678817749,
912
+ "step": 43
913
+ },
914
+ {
915
+ "debug/policy_chosen_logits": -0.90626060962677,
916
+ "debug/policy_chosen_logps": -139.66921997070312,
917
+ "debug/policy_rejected_logits": -0.8777171969413757,
918
+ "debug/policy_rejected_logps": -187.69277954101562,
919
+ "debug/reference_chosen_logps": -145.80397033691406,
920
+ "debug/reference_rejected_logps": -181.13963317871094,
921
+ "epoch": 0.6875,
922
+ "grad_norm": 7.207738479298261,
923
+ "learning_rate": 1e-06,
924
+ "logits/chosen": -0.90626060962677,
925
+ "logits/rejected": -0.8777171969413757,
926
+ "logps/chosen": -139.66921997070312,
927
+ "logps/rejected": -187.69277954101562,
928
+ "loss": 0.4527,
929
+ "rewards/accuracies": 0.875,
930
+ "rewards/chosen": 0.06134761869907379,
931
+ "rewards/margins": 0.1268792748451233,
932
+ "rewards/rejected": -0.0655316635966301,
933
+ "step": 44
934
+ },
935
+ {
936
+ "debug/policy_chosen_logits": -0.7800077199935913,
937
+ "debug/policy_chosen_logps": -153.02247619628906,
938
+ "debug/policy_rejected_logits": -0.8804305791854858,
939
+ "debug/policy_rejected_logps": -175.3987274169922,
940
+ "debug/reference_chosen_logps": -160.71377563476562,
941
+ "debug/reference_rejected_logps": -164.52713012695312,
942
+ "epoch": 0.703125,
943
+ "grad_norm": 6.76588354529725,
944
+ "learning_rate": 1e-06,
945
+ "logits/chosen": -0.7800077199935913,
946
+ "logits/rejected": -0.8804305791854858,
947
+ "logps/chosen": -153.02247619628906,
948
+ "logps/rejected": -175.3987274169922,
949
+ "loss": 0.4337,
950
+ "rewards/accuracies": 0.75,
951
+ "rewards/chosen": 0.07691291719675064,
952
+ "rewards/margins": 0.18562886118888855,
953
+ "rewards/rejected": -0.1087159514427185,
954
+ "step": 45
955
+ },
956
+ {
957
+ "debug/policy_chosen_logits": -0.6891363859176636,
958
+ "debug/policy_chosen_logps": -174.951416015625,
959
+ "debug/policy_rejected_logits": -0.7365065217018127,
960
+ "debug/policy_rejected_logps": -210.58712768554688,
961
+ "debug/reference_chosen_logps": -179.22348022460938,
962
+ "debug/reference_rejected_logps": -203.9578399658203,
963
+ "epoch": 0.71875,
964
+ "grad_norm": 6.33536574920728,
965
+ "learning_rate": 1e-06,
966
+ "logits/chosen": -0.6891363859176636,
967
+ "logits/rejected": -0.7365065217018127,
968
+ "logps/chosen": -174.951416015625,
969
+ "logps/rejected": -210.58712768554688,
970
+ "loss": 0.4355,
971
+ "rewards/accuracies": 0.625,
972
+ "rewards/chosen": 0.04272085428237915,
973
+ "rewards/margins": 0.109013631939888,
974
+ "rewards/rejected": -0.06629277765750885,
975
+ "step": 46
976
+ },
977
+ {
978
+ "debug/policy_chosen_logits": -1.0616976022720337,
979
+ "debug/policy_chosen_logps": -140.75953674316406,
980
+ "debug/policy_rejected_logits": -0.8193640112876892,
981
+ "debug/policy_rejected_logps": -166.1212158203125,
982
+ "debug/reference_chosen_logps": -145.49392700195312,
983
+ "debug/reference_rejected_logps": -169.64987182617188,
984
+ "epoch": 0.734375,
985
+ "grad_norm": 7.072020108503254,
986
+ "learning_rate": 1e-06,
987
+ "logits/chosen": -1.0616976022720337,
988
+ "logits/rejected": -0.8193640112876892,
989
+ "logps/chosen": -140.75953674316406,
990
+ "logps/rejected": -166.1212158203125,
991
+ "loss": 0.4804,
992
+ "rewards/accuracies": 0.5,
993
+ "rewards/chosen": 0.04734383523464203,
994
+ "rewards/margins": 0.012057198211550713,
995
+ "rewards/rejected": 0.03528663516044617,
996
+ "step": 47
997
+ },
998
+ {
999
+ "debug/policy_chosen_logits": -0.8684768080711365,
1000
+ "debug/policy_chosen_logps": -168.0657958984375,
1001
+ "debug/policy_rejected_logits": -0.9659979939460754,
1002
+ "debug/policy_rejected_logps": -164.24932861328125,
1003
+ "debug/reference_chosen_logps": -175.15855407714844,
1004
+ "debug/reference_rejected_logps": -156.83241271972656,
1005
+ "epoch": 0.75,
1006
+ "grad_norm": 6.146768918478943,
1007
+ "learning_rate": 1e-06,
1008
+ "logits/chosen": -0.8684768080711365,
1009
+ "logits/rejected": -0.9659979939460754,
1010
+ "logps/chosen": -168.0657958984375,
1011
+ "logps/rejected": -164.24932861328125,
1012
+ "loss": 0.4465,
1013
+ "rewards/accuracies": 0.625,
1014
+ "rewards/chosen": 0.0709274560213089,
1015
+ "rewards/margins": 0.14509651064872742,
1016
+ "rewards/rejected": -0.07416905462741852,
1017
+ "step": 48
1018
+ },
1019
+ {
1020
+ "debug/policy_chosen_logits": -0.7563624382019043,
1021
+ "debug/policy_chosen_logps": -178.76950073242188,
1022
+ "debug/policy_rejected_logits": -0.639800488948822,
1023
+ "debug/policy_rejected_logps": -168.65484619140625,
1024
+ "debug/reference_chosen_logps": -178.57791137695312,
1025
+ "debug/reference_rejected_logps": -171.87460327148438,
1026
+ "epoch": 0.765625,
1027
+ "grad_norm": 5.930093840031159,
1028
+ "learning_rate": 1e-06,
1029
+ "logits/chosen": -0.7563624382019043,
1030
+ "logits/rejected": -0.639800488948822,
1031
+ "logps/chosen": -178.76950073242188,
1032
+ "logps/rejected": -168.65484619140625,
1033
+ "loss": 0.4697,
1034
+ "rewards/accuracies": 0.375,
1035
+ "rewards/chosen": -0.001915864646434784,
1036
+ "rewards/margins": -0.034113530069589615,
1037
+ "rewards/rejected": 0.03219766542315483,
1038
+ "step": 49
1039
+ },
1040
+ {
1041
+ "debug/policy_chosen_logits": -0.7761915922164917,
1042
+ "debug/policy_chosen_logps": -167.15562438964844,
1043
+ "debug/policy_rejected_logits": -0.8122952580451965,
1044
+ "debug/policy_rejected_logps": -167.90469360351562,
1045
+ "debug/reference_chosen_logps": -169.98074340820312,
1046
+ "debug/reference_rejected_logps": -171.24453735351562,
1047
+ "epoch": 0.78125,
1048
+ "grad_norm": 6.402012147266644,
1049
+ "learning_rate": 1e-06,
1050
+ "logits/chosen": -0.7761915922164917,
1051
+ "logits/rejected": -0.8122952580451965,
1052
+ "logps/chosen": -167.15562438964844,
1053
+ "logps/rejected": -167.90469360351562,
1054
+ "loss": 0.4469,
1055
+ "rewards/accuracies": 0.375,
1056
+ "rewards/chosen": 0.028251150622963905,
1057
+ "rewards/margins": -0.0051473043859004974,
1058
+ "rewards/rejected": 0.03339845687150955,
1059
+ "step": 50
1060
+ },
1061
+ {
1062
+ "debug/policy_chosen_logits": -1.0047523975372314,
1063
+ "debug/policy_chosen_logps": -161.81784057617188,
1064
+ "debug/policy_rejected_logits": -0.6831015944480896,
1065
+ "debug/policy_rejected_logps": -189.47030639648438,
1066
+ "debug/reference_chosen_logps": -159.06884765625,
1067
+ "debug/reference_rejected_logps": -185.1800537109375,
1068
+ "epoch": 0.796875,
1069
+ "grad_norm": 6.034796225510894,
1070
+ "learning_rate": 1e-06,
1071
+ "logits/chosen": -1.0047523975372314,
1072
+ "logits/rejected": -0.6831015944480896,
1073
+ "logps/chosen": -161.81784057617188,
1074
+ "logps/rejected": -189.47030639648438,
1075
+ "loss": 0.45,
1076
+ "rewards/accuracies": 0.5,
1077
+ "rewards/chosen": -0.027489986270666122,
1078
+ "rewards/margins": 0.015412424691021442,
1079
+ "rewards/rejected": -0.04290241375565529,
1080
+ "step": 51
1081
+ },
1082
+ {
1083
+ "debug/policy_chosen_logits": -0.9873104691505432,
1084
+ "debug/policy_chosen_logps": -147.3292236328125,
1085
+ "debug/policy_rejected_logits": -0.9074857234954834,
1086
+ "debug/policy_rejected_logps": -160.4976806640625,
1087
+ "debug/reference_chosen_logps": -150.1821746826172,
1088
+ "debug/reference_rejected_logps": -164.38450622558594,
1089
+ "epoch": 0.8125,
1090
+ "grad_norm": 5.995554347552576,
1091
+ "learning_rate": 1e-06,
1092
+ "logits/chosen": -0.9873104691505432,
1093
+ "logits/rejected": -0.9074857234954834,
1094
+ "logps/chosen": -147.3292236328125,
1095
+ "logps/rejected": -160.4976806640625,
1096
+ "loss": 0.4478,
1097
+ "rewards/accuracies": 0.375,
1098
+ "rewards/chosen": 0.028529377654194832,
1099
+ "rewards/margins": -0.010338811203837395,
1100
+ "rewards/rejected": 0.03886818885803223,
1101
+ "step": 52
1102
+ },
1103
+ {
1104
+ "debug/policy_chosen_logits": -0.8061836957931519,
1105
+ "debug/policy_chosen_logps": -154.01083374023438,
1106
+ "debug/policy_rejected_logits": -0.9387882947921753,
1107
+ "debug/policy_rejected_logps": -162.59048461914062,
1108
+ "debug/reference_chosen_logps": -165.92892456054688,
1109
+ "debug/reference_rejected_logps": -159.9312744140625,
1110
+ "epoch": 0.828125,
1111
+ "grad_norm": 6.573209394571419,
1112
+ "learning_rate": 1e-06,
1113
+ "logits/chosen": -0.8061836957931519,
1114
+ "logits/rejected": -0.9387882947921753,
1115
+ "logps/chosen": -154.01083374023438,
1116
+ "logps/rejected": -162.59048461914062,
1117
+ "loss": 0.4394,
1118
+ "rewards/accuracies": 0.875,
1119
+ "rewards/chosen": 0.11918096244335175,
1120
+ "rewards/margins": 0.14577314257621765,
1121
+ "rewards/rejected": -0.02659217081964016,
1122
+ "step": 53
1123
+ },
1124
+ {
1125
+ "debug/policy_chosen_logits": -0.8860620260238647,
1126
+ "debug/policy_chosen_logps": -171.57965087890625,
1127
+ "debug/policy_rejected_logits": -0.8173072934150696,
1128
+ "debug/policy_rejected_logps": -172.11398315429688,
1129
+ "debug/reference_chosen_logps": -175.11090087890625,
1130
+ "debug/reference_rejected_logps": -177.931396484375,
1131
+ "epoch": 0.84375,
1132
+ "grad_norm": 6.3343088344275875,
1133
+ "learning_rate": 1e-06,
1134
+ "logits/chosen": -0.8860620260238647,
1135
+ "logits/rejected": -0.8173072934150696,
1136
+ "logps/chosen": -171.57965087890625,
1137
+ "logps/rejected": -172.11398315429688,
1138
+ "loss": 0.4577,
1139
+ "rewards/accuracies": 0.75,
1140
+ "rewards/chosen": 0.03531248867511749,
1141
+ "rewards/margins": -0.02286178432404995,
1142
+ "rewards/rejected": 0.05817427486181259,
1143
+ "step": 54
1144
+ },
1145
+ {
1146
+ "debug/policy_chosen_logits": -0.9360355734825134,
1147
+ "debug/policy_chosen_logps": -159.60911560058594,
1148
+ "debug/policy_rejected_logits": -0.8322389125823975,
1149
+ "debug/policy_rejected_logps": -184.9361572265625,
1150
+ "debug/reference_chosen_logps": -161.784912109375,
1151
+ "debug/reference_rejected_logps": -181.3787841796875,
1152
+ "epoch": 0.859375,
1153
+ "grad_norm": 6.940576793005815,
1154
+ "learning_rate": 1e-06,
1155
+ "logits/chosen": -0.9360355734825134,
1156
+ "logits/rejected": -0.8322389125823975,
1157
+ "logps/chosen": -159.60911560058594,
1158
+ "logps/rejected": -184.9361572265625,
1159
+ "loss": 0.4589,
1160
+ "rewards/accuracies": 0.375,
1161
+ "rewards/chosen": 0.02175775356590748,
1162
+ "rewards/margins": 0.05733121931552887,
1163
+ "rewards/rejected": -0.03557346388697624,
1164
+ "step": 55
1165
+ },
1166
+ {
1167
+ "debug/policy_chosen_logits": -0.7893481850624084,
1168
+ "debug/policy_chosen_logps": -171.56735229492188,
1169
+ "debug/policy_rejected_logits": -0.7332755327224731,
1170
+ "debug/policy_rejected_logps": -196.43365478515625,
1171
+ "debug/reference_chosen_logps": -171.05889892578125,
1172
+ "debug/reference_rejected_logps": -182.0035400390625,
1173
+ "epoch": 0.875,
1174
+ "grad_norm": 6.971775270738982,
1175
+ "learning_rate": 1e-06,
1176
+ "logits/chosen": -0.7893481850624084,
1177
+ "logits/rejected": -0.7332755327224731,
1178
+ "logps/chosen": -171.56735229492188,
1179
+ "logps/rejected": -196.43365478515625,
1180
+ "loss": 0.4405,
1181
+ "rewards/accuracies": 0.875,
1182
+ "rewards/chosen": -0.00508442847058177,
1183
+ "rewards/margins": 0.13921663165092468,
1184
+ "rewards/rejected": -0.14430105686187744,
1185
+ "step": 56
1186
+ },
1187
+ {
1188
+ "debug/policy_chosen_logits": -1.2259751558303833,
1189
+ "debug/policy_chosen_logps": -143.96656799316406,
1190
+ "debug/policy_rejected_logits": -0.825638473033905,
1191
+ "debug/policy_rejected_logps": -187.6162567138672,
1192
+ "debug/reference_chosen_logps": -142.93975830078125,
1193
+ "debug/reference_rejected_logps": -183.60317993164062,
1194
+ "epoch": 0.890625,
1195
+ "grad_norm": 7.861463588461739,
1196
+ "learning_rate": 1e-06,
1197
+ "logits/chosen": -1.2259751558303833,
1198
+ "logits/rejected": -0.825638473033905,
1199
+ "logps/chosen": -143.96656799316406,
1200
+ "logps/rejected": -187.6162567138672,
1201
+ "loss": 0.4669,
1202
+ "rewards/accuracies": 0.625,
1203
+ "rewards/chosen": -0.010268086567521095,
1204
+ "rewards/margins": 0.029862696304917336,
1205
+ "rewards/rejected": -0.04013078659772873,
1206
+ "step": 57
1207
+ },
1208
+ {
1209
+ "debug/policy_chosen_logits": -0.6283476948738098,
1210
+ "debug/policy_chosen_logps": -179.42198181152344,
1211
+ "debug/policy_rejected_logits": -0.7192294001579285,
1212
+ "debug/policy_rejected_logps": -160.84140014648438,
1213
+ "debug/reference_chosen_logps": -179.8939666748047,
1214
+ "debug/reference_rejected_logps": -157.6416778564453,
1215
+ "epoch": 0.90625,
1216
+ "grad_norm": 5.979721246123928,
1217
+ "learning_rate": 1e-06,
1218
+ "logits/chosen": -0.6283476948738098,
1219
+ "logits/rejected": -0.7192294001579285,
1220
+ "logps/chosen": -179.42198181152344,
1221
+ "logps/rejected": -160.84140014648438,
1222
+ "loss": 0.4573,
1223
+ "rewards/accuracies": 0.5,
1224
+ "rewards/chosen": 0.004719886463135481,
1225
+ "rewards/margins": 0.03671710193157196,
1226
+ "rewards/rejected": -0.03199721872806549,
1227
+ "step": 58
1228
+ },
1229
+ {
1230
+ "debug/policy_chosen_logits": -1.0611419677734375,
1231
+ "debug/policy_chosen_logps": -162.58331298828125,
1232
+ "debug/policy_rejected_logits": -0.9486455917358398,
1233
+ "debug/policy_rejected_logps": -164.91607666015625,
1234
+ "debug/reference_chosen_logps": -168.58851623535156,
1235
+ "debug/reference_rejected_logps": -160.13070678710938,
1236
+ "epoch": 0.921875,
1237
+ "grad_norm": 6.255365355958018,
1238
+ "learning_rate": 1e-06,
1239
+ "logits/chosen": -1.0611419677734375,
1240
+ "logits/rejected": -0.9486455917358398,
1241
+ "logps/chosen": -162.58331298828125,
1242
+ "logps/rejected": -164.91607666015625,
1243
+ "loss": 0.4572,
1244
+ "rewards/accuracies": 0.75,
1245
+ "rewards/chosen": 0.06005190685391426,
1246
+ "rewards/margins": 0.10790553689002991,
1247
+ "rewards/rejected": -0.04785362631082535,
1248
+ "step": 59
1249
+ },
1250
+ {
1251
+ "debug/policy_chosen_logits": -0.8236188292503357,
1252
+ "debug/policy_chosen_logps": -162.90415954589844,
1253
+ "debug/policy_rejected_logits": -0.8510185480117798,
1254
+ "debug/policy_rejected_logps": -187.468994140625,
1255
+ "debug/reference_chosen_logps": -171.4542236328125,
1256
+ "debug/reference_rejected_logps": -171.10787963867188,
1257
+ "epoch": 0.9375,
1258
+ "grad_norm": 7.036142220314558,
1259
+ "learning_rate": 1e-06,
1260
+ "logits/chosen": -0.8236188292503357,
1261
+ "logits/rejected": -0.8510185480117798,
1262
+ "logps/chosen": -162.90415954589844,
1263
+ "logps/rejected": -187.468994140625,
1264
+ "loss": 0.4524,
1265
+ "rewards/accuracies": 0.75,
1266
+ "rewards/chosen": 0.08550070226192474,
1267
+ "rewards/margins": 0.24911174178123474,
1268
+ "rewards/rejected": -0.1636110544204712,
1269
+ "step": 60
1270
+ },
1271
+ {
1272
+ "debug/policy_chosen_logits": -0.9428707361221313,
1273
+ "debug/policy_chosen_logps": -148.73089599609375,
1274
+ "debug/policy_rejected_logits": -0.8415032029151917,
1275
+ "debug/policy_rejected_logps": -179.90115356445312,
1276
+ "debug/reference_chosen_logps": -155.75228881835938,
1277
+ "debug/reference_rejected_logps": -173.25796508789062,
1278
+ "epoch": 0.953125,
1279
+ "grad_norm": 7.382144233072753,
1280
+ "learning_rate": 1e-06,
1281
+ "logits/chosen": -0.9428707361221313,
1282
+ "logits/rejected": -0.8415032029151917,
1283
+ "logps/chosen": -148.73089599609375,
1284
+ "logps/rejected": -179.90115356445312,
1285
+ "loss": 0.4308,
1286
+ "rewards/accuracies": 0.75,
1287
+ "rewards/chosen": 0.07021382451057434,
1288
+ "rewards/margins": 0.1366458535194397,
1289
+ "rewards/rejected": -0.06643202900886536,
1290
+ "step": 61
1291
+ },
1292
+ {
1293
+ "debug/policy_chosen_logits": -0.7142590880393982,
1294
+ "debug/policy_chosen_logps": -183.39328002929688,
1295
+ "debug/policy_rejected_logits": -0.8318662643432617,
1296
+ "debug/policy_rejected_logps": -185.28347778320312,
1297
+ "debug/reference_chosen_logps": -185.95765686035156,
1298
+ "debug/reference_rejected_logps": -188.14785766601562,
1299
+ "epoch": 0.96875,
1300
+ "grad_norm": 7.772884707461087,
1301
+ "learning_rate": 1e-06,
1302
+ "logits/chosen": -0.7142590880393982,
1303
+ "logits/rejected": -0.8318662643432617,
1304
+ "logps/chosen": -183.39328002929688,
1305
+ "logps/rejected": -185.28347778320312,
1306
+ "loss": 0.4731,
1307
+ "rewards/accuracies": 0.5,
1308
+ "rewards/chosen": 0.025643818080425262,
1309
+ "rewards/margins": -0.0030001532286405563,
1310
+ "rewards/rejected": 0.02864396944642067,
1311
+ "step": 62
1312
+ },
1313
+ {
1314
+ "debug/policy_chosen_logits": -0.8611500859260559,
1315
+ "debug/policy_chosen_logps": -148.0809326171875,
1316
+ "debug/policy_rejected_logits": -0.9426791667938232,
1317
+ "debug/policy_rejected_logps": -159.23069763183594,
1318
+ "debug/reference_chosen_logps": -153.9005126953125,
1319
+ "debug/reference_rejected_logps": -147.47735595703125,
1320
+ "epoch": 0.984375,
1321
+ "grad_norm": 6.5837351291708375,
1322
+ "learning_rate": 1e-06,
1323
+ "logits/chosen": -0.8611500859260559,
1324
+ "logits/rejected": -0.9426791667938232,
1325
+ "logps/chosen": -148.0809326171875,
1326
+ "logps/rejected": -159.23069763183594,
1327
+ "loss": 0.4355,
1328
+ "rewards/accuracies": 1.0,
1329
+ "rewards/chosen": 0.058195747435092926,
1330
+ "rewards/margins": 0.1757291853427887,
1331
+ "rewards/rejected": -0.11753343790769577,
1332
+ "step": 63
1333
+ },
1334
+ {
1335
+ "debug/policy_chosen_logits": -0.8686034083366394,
1336
+ "debug/policy_chosen_logps": -163.016357421875,
1337
+ "debug/policy_rejected_logits": -0.8933126926422119,
1338
+ "debug/policy_rejected_logps": -178.46234130859375,
1339
+ "debug/reference_chosen_logps": -166.87405395507812,
1340
+ "debug/reference_rejected_logps": -176.6748046875,
1341
+ "epoch": 1.0,
1342
+ "grad_norm": 8.108425274874563,
1343
+ "learning_rate": 1e-06,
1344
+ "logits/chosen": -0.8686034083366394,
1345
+ "logits/rejected": -0.8933126926422119,
1346
+ "logps/chosen": -163.016357421875,
1347
+ "logps/rejected": -178.46234130859375,
1348
+ "loss": 0.4087,
1349
+ "rewards/accuracies": 0.5,
1350
+ "rewards/chosen": 0.03857698291540146,
1351
+ "rewards/margins": 0.05645231530070305,
1352
+ "rewards/rejected": -0.017875326797366142,
1353
+ "step": 64
1354
+ },
1355
+ {
1356
+ "epoch": 1.0,
1357
+ "step": 64,
1358
+ "total_flos": 0.0,
1359
+ "train_loss": 0.46988651156425476,
1360
+ "train_runtime": 197.5097,
1361
+ "train_samples_per_second": 20.627,
1362
+ "train_steps_per_second": 0.324
1363
+ }
1364
+ ],
1365
+ "logging_steps": 1,
1366
+ "max_steps": 64,
1367
+ "num_input_tokens_seen": 0,
1368
+ "num_train_epochs": 1,
1369
+ "save_steps": 500,
1370
+ "stateful_callbacks": {
1371
+ "TrainerControl": {
1372
+ "args": {
1373
+ "should_epoch_stop": false,
1374
+ "should_evaluate": false,
1375
+ "should_log": false,
1376
+ "should_save": true,
1377
+ "should_training_stop": true
1378
+ },
1379
+ "attributes": {}
1380
+ }
1381
+ },
1382
+ "total_flos": 0.0,
1383
+ "train_batch_size": 8,
1384
+ "trial_name": null,
1385
+ "trial_params": null
1386
+ }