yiran-wang3 commited on
Commit
ee38257
1 Parent(s): e2c2041

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 +1008 -0
README.md ADDED
@@ -0,0 +1,64 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ library_name: transformers
3
+ license: apache-2.0
4
+ base_model: Qwen/Qwen2.5-Coder-7B-Instruct
5
+ tags:
6
+ - alignment-handbook
7
+ - generated_from_trainer
8
+ - trl
9
+ - dpo
10
+ datasets:
11
+ - self-generate/qw2coder_reflct7b_original_cn_mining_oj_iter0-binarized-reflection-scored
12
+ model-index:
13
+ - name: qwen2_coder7b_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
+ # qwen2_coder7b_reflct_adamw_iter1
21
+
22
+ This model is a fine-tuned version of [Qwen/Qwen2.5-Coder-7B-Instruct](https://huggingface.co/Qwen/Qwen2.5-Coder-7B-Instruct) on the self-generate/qw2coder_reflct7b_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": 0.4573696670324906,
5
+ "train_runtime": 475.0336,
6
+ "train_samples": 2882,
7
+ "train_samples_per_second": 6.067,
8
+ "train_steps_per_second": 0.097
9
+ }
config.json CHANGED
@@ -23,7 +23,7 @@
23
  "tie_word_embeddings": false,
24
  "torch_dtype": "bfloat16",
25
  "transformers_version": "4.45.0",
26
- "use_cache": false,
27
  "use_sliding_window": false,
28
  "vocab_size": 152064
29
  }
 
23
  "tie_word_embeddings": false,
24
  "torch_dtype": "bfloat16",
25
  "transformers_version": "4.45.0",
26
+ "use_cache": true,
27
  "use_sliding_window": false,
28
  "vocab_size": 152064
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": 0.4573696670324906,
5
+ "train_runtime": 475.0336,
6
+ "train_samples": 2882,
7
+ "train_samples_per_second": 6.067,
8
+ "train_steps_per_second": 0.097
9
+ }
trainer_state.json ADDED
@@ -0,0 +1,1008 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 1.0,
5
+ "eval_steps": 100,
6
+ "global_step": 46,
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": -2.5406789779663086,
13
+ "debug/policy_chosen_logps": -174.41754150390625,
14
+ "debug/policy_rejected_logits": -2.521660566329956,
15
+ "debug/policy_rejected_logps": -159.5323486328125,
16
+ "debug/reference_chosen_logps": -174.41754150390625,
17
+ "debug/reference_rejected_logps": -159.5323486328125,
18
+ "epoch": 0.021739130434782608,
19
+ "grad_norm": 2.9138556694482802,
20
+ "learning_rate": 1e-06,
21
+ "logits/chosen": -2.5406789779663086,
22
+ "logits/rejected": -2.521660566329956,
23
+ "logps/chosen": -174.41754150390625,
24
+ "logps/rejected": -159.5323486328125,
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": -2.3866376876831055,
34
+ "debug/policy_chosen_logps": -191.08587646484375,
35
+ "debug/policy_rejected_logits": -2.4478790760040283,
36
+ "debug/policy_rejected_logps": -191.51112365722656,
37
+ "debug/reference_chosen_logps": -191.0208740234375,
38
+ "debug/reference_rejected_logps": -191.41552734375,
39
+ "epoch": 0.043478260869565216,
40
+ "grad_norm": 2.73946049555914,
41
+ "learning_rate": 1e-06,
42
+ "logits/chosen": -2.3866376876831055,
43
+ "logits/rejected": -2.4478790760040283,
44
+ "logps/chosen": -191.08587646484375,
45
+ "logps/rejected": -191.51112365722656,
46
+ "loss": 0.5002,
47
+ "rewards/accuracies": 0.5,
48
+ "rewards/chosen": -0.0006500434246845543,
49
+ "rewards/margins": 0.0003061485185753554,
50
+ "rewards/rejected": -0.0009561920305714011,
51
+ "step": 2
52
+ },
53
+ {
54
+ "debug/policy_chosen_logits": -2.465501308441162,
55
+ "debug/policy_chosen_logps": -206.74551391601562,
56
+ "debug/policy_rejected_logits": -2.5732452869415283,
57
+ "debug/policy_rejected_logps": -183.27645874023438,
58
+ "debug/reference_chosen_logps": -206.72006225585938,
59
+ "debug/reference_rejected_logps": -183.27920532226562,
60
+ "epoch": 0.06521739130434782,
61
+ "grad_norm": 2.8074234755722296,
62
+ "learning_rate": 1e-06,
63
+ "logits/chosen": -2.465501308441162,
64
+ "logits/rejected": -2.5732452869415283,
65
+ "logps/chosen": -206.74551391601562,
66
+ "logps/rejected": -183.27645874023438,
67
+ "loss": 0.4997,
68
+ "rewards/accuracies": 0.375,
69
+ "rewards/chosen": -0.00025447848020121455,
70
+ "rewards/margins": -0.00028175354236736894,
71
+ "rewards/rejected": 2.7275120373815298e-05,
72
+ "step": 3
73
+ },
74
+ {
75
+ "debug/policy_chosen_logits": -2.4564504623413086,
76
+ "debug/policy_chosen_logps": -196.91964721679688,
77
+ "debug/policy_rejected_logits": -2.4699106216430664,
78
+ "debug/policy_rejected_logps": -166.32009887695312,
79
+ "debug/reference_chosen_logps": -197.21090698242188,
80
+ "debug/reference_rejected_logps": -166.31948852539062,
81
+ "epoch": 0.08695652173913043,
82
+ "grad_norm": 3.206369897884046,
83
+ "learning_rate": 1e-06,
84
+ "logits/chosen": -2.4564504623413086,
85
+ "logits/rejected": -2.4699106216430664,
86
+ "logps/chosen": -196.91964721679688,
87
+ "logps/rejected": -166.32009887695312,
88
+ "loss": 0.4981,
89
+ "rewards/accuracies": 0.75,
90
+ "rewards/chosen": 0.002912597730755806,
91
+ "rewards/margins": 0.0029187584295868874,
92
+ "rewards/rejected": -6.160815246403217e-06,
93
+ "step": 4
94
+ },
95
+ {
96
+ "debug/policy_chosen_logits": -2.444270610809326,
97
+ "debug/policy_chosen_logps": -189.8358154296875,
98
+ "debug/policy_rejected_logits": -2.319845676422119,
99
+ "debug/policy_rejected_logps": -174.91769409179688,
100
+ "debug/reference_chosen_logps": -189.87257385253906,
101
+ "debug/reference_rejected_logps": -174.7454071044922,
102
+ "epoch": 0.10869565217391304,
103
+ "grad_norm": 2.9622178910820045,
104
+ "learning_rate": 1e-06,
105
+ "logits/chosen": -2.444270610809326,
106
+ "logits/rejected": -2.319845676422119,
107
+ "logps/chosen": -189.8358154296875,
108
+ "logps/rejected": -174.91769409179688,
109
+ "loss": 0.4967,
110
+ "rewards/accuracies": 0.625,
111
+ "rewards/chosen": 0.0003675461048260331,
112
+ "rewards/margins": 0.002090301364660263,
113
+ "rewards/rejected": -0.0017227554926648736,
114
+ "step": 5
115
+ },
116
+ {
117
+ "debug/policy_chosen_logits": -2.393871307373047,
118
+ "debug/policy_chosen_logps": -202.0091094970703,
119
+ "debug/policy_rejected_logits": -2.4609367847442627,
120
+ "debug/policy_rejected_logps": -192.79449462890625,
121
+ "debug/reference_chosen_logps": -202.7803955078125,
122
+ "debug/reference_rejected_logps": -192.9990997314453,
123
+ "epoch": 0.13043478260869565,
124
+ "grad_norm": 2.890923079465906,
125
+ "learning_rate": 1e-06,
126
+ "logits/chosen": -2.393871307373047,
127
+ "logits/rejected": -2.4609367847442627,
128
+ "logps/chosen": -202.0091094970703,
129
+ "logps/rejected": -192.79449462890625,
130
+ "loss": 0.4933,
131
+ "rewards/accuracies": 0.625,
132
+ "rewards/chosen": 0.007712879218161106,
133
+ "rewards/margins": 0.005666866432875395,
134
+ "rewards/rejected": 0.0020460127852857113,
135
+ "step": 6
136
+ },
137
+ {
138
+ "debug/policy_chosen_logits": -2.584399700164795,
139
+ "debug/policy_chosen_logps": -201.59007263183594,
140
+ "debug/policy_rejected_logits": -2.624255895614624,
141
+ "debug/policy_rejected_logps": -177.23953247070312,
142
+ "debug/reference_chosen_logps": -202.87220764160156,
143
+ "debug/reference_rejected_logps": -176.957763671875,
144
+ "epoch": 0.15217391304347827,
145
+ "grad_norm": 2.957447005375319,
146
+ "learning_rate": 1e-06,
147
+ "logits/chosen": -2.584399700164795,
148
+ "logits/rejected": -2.624255895614624,
149
+ "logps/chosen": -201.59007263183594,
150
+ "logps/rejected": -177.23953247070312,
151
+ "loss": 0.495,
152
+ "rewards/accuracies": 0.75,
153
+ "rewards/chosen": 0.012821445241570473,
154
+ "rewards/margins": 0.01563918963074684,
155
+ "rewards/rejected": -0.0028177450876682997,
156
+ "step": 7
157
+ },
158
+ {
159
+ "debug/policy_chosen_logits": -2.5293426513671875,
160
+ "debug/policy_chosen_logps": -191.92535400390625,
161
+ "debug/policy_rejected_logits": -2.4522714614868164,
162
+ "debug/policy_rejected_logps": -190.64605712890625,
163
+ "debug/reference_chosen_logps": -192.4668426513672,
164
+ "debug/reference_rejected_logps": -191.4180145263672,
165
+ "epoch": 0.17391304347826086,
166
+ "grad_norm": 2.857224539567903,
167
+ "learning_rate": 1e-06,
168
+ "logits/chosen": -2.5293426513671875,
169
+ "logits/rejected": -2.4522714614868164,
170
+ "logps/chosen": -191.92535400390625,
171
+ "logps/rejected": -190.64605712890625,
172
+ "loss": 0.4894,
173
+ "rewards/accuracies": 0.375,
174
+ "rewards/chosen": 0.005414981860667467,
175
+ "rewards/margins": -0.002304553519934416,
176
+ "rewards/rejected": 0.00771953584626317,
177
+ "step": 8
178
+ },
179
+ {
180
+ "debug/policy_chosen_logits": -2.524057149887085,
181
+ "debug/policy_chosen_logps": -181.09461975097656,
182
+ "debug/policy_rejected_logits": -2.53678035736084,
183
+ "debug/policy_rejected_logps": -184.49664306640625,
184
+ "debug/reference_chosen_logps": -181.84410095214844,
185
+ "debug/reference_rejected_logps": -183.90103149414062,
186
+ "epoch": 0.1956521739130435,
187
+ "grad_norm": 2.659967589766961,
188
+ "learning_rate": 1e-06,
189
+ "logits/chosen": -2.524057149887085,
190
+ "logits/rejected": -2.53678035736084,
191
+ "logps/chosen": -181.09461975097656,
192
+ "logps/rejected": -184.49664306640625,
193
+ "loss": 0.4895,
194
+ "rewards/accuracies": 0.625,
195
+ "rewards/chosen": 0.00749471690505743,
196
+ "rewards/margins": 0.013450756669044495,
197
+ "rewards/rejected": -0.005956039298325777,
198
+ "step": 9
199
+ },
200
+ {
201
+ "debug/policy_chosen_logits": -2.3921713829040527,
202
+ "debug/policy_chosen_logps": -179.6142578125,
203
+ "debug/policy_rejected_logits": -2.3711440563201904,
204
+ "debug/policy_rejected_logps": -162.54669189453125,
205
+ "debug/reference_chosen_logps": -180.62747192382812,
206
+ "debug/reference_rejected_logps": -160.78297424316406,
207
+ "epoch": 0.21739130434782608,
208
+ "grad_norm": 2.907825352825365,
209
+ "learning_rate": 1e-06,
210
+ "logits/chosen": -2.3921713829040527,
211
+ "logits/rejected": -2.3711440563201904,
212
+ "logps/chosen": -179.6142578125,
213
+ "logps/rejected": -162.54669189453125,
214
+ "loss": 0.4802,
215
+ "rewards/accuracies": 0.875,
216
+ "rewards/chosen": 0.010132160037755966,
217
+ "rewards/margins": 0.027769355103373528,
218
+ "rewards/rejected": -0.01763719692826271,
219
+ "step": 10
220
+ },
221
+ {
222
+ "debug/policy_chosen_logits": -2.521888494491577,
223
+ "debug/policy_chosen_logps": -184.33038330078125,
224
+ "debug/policy_rejected_logits": -2.344513416290283,
225
+ "debug/policy_rejected_logps": -178.75074768066406,
226
+ "debug/reference_chosen_logps": -184.63455200195312,
227
+ "debug/reference_rejected_logps": -175.36129760742188,
228
+ "epoch": 0.2391304347826087,
229
+ "grad_norm": 2.4616167539116005,
230
+ "learning_rate": 1e-06,
231
+ "logits/chosen": -2.521888494491577,
232
+ "logits/rejected": -2.344513416290283,
233
+ "logps/chosen": -184.33038330078125,
234
+ "logps/rejected": -178.75074768066406,
235
+ "loss": 0.4843,
236
+ "rewards/accuracies": 1.0,
237
+ "rewards/chosen": 0.003041705582290888,
238
+ "rewards/margins": 0.03693626448512077,
239
+ "rewards/rejected": -0.033894557505846024,
240
+ "step": 11
241
+ },
242
+ {
243
+ "debug/policy_chosen_logits": -2.405729055404663,
244
+ "debug/policy_chosen_logps": -175.63497924804688,
245
+ "debug/policy_rejected_logits": -2.4264020919799805,
246
+ "debug/policy_rejected_logps": -179.73696899414062,
247
+ "debug/reference_chosen_logps": -174.1973876953125,
248
+ "debug/reference_rejected_logps": -177.64727783203125,
249
+ "epoch": 0.2608695652173913,
250
+ "grad_norm": 2.736323742946957,
251
+ "learning_rate": 1e-06,
252
+ "logits/chosen": -2.405729055404663,
253
+ "logits/rejected": -2.4264020919799805,
254
+ "logps/chosen": -175.63497924804688,
255
+ "logps/rejected": -179.73696899414062,
256
+ "loss": 0.4859,
257
+ "rewards/accuracies": 0.5,
258
+ "rewards/chosen": -0.014375820755958557,
259
+ "rewards/margins": 0.006521034054458141,
260
+ "rewards/rejected": -0.020896855741739273,
261
+ "step": 12
262
+ },
263
+ {
264
+ "debug/policy_chosen_logits": -2.3890933990478516,
265
+ "debug/policy_chosen_logps": -189.14987182617188,
266
+ "debug/policy_rejected_logits": -2.305311918258667,
267
+ "debug/policy_rejected_logps": -180.1926727294922,
268
+ "debug/reference_chosen_logps": -191.95681762695312,
269
+ "debug/reference_rejected_logps": -178.28973388671875,
270
+ "epoch": 0.2826086956521739,
271
+ "grad_norm": 2.7383468500636767,
272
+ "learning_rate": 1e-06,
273
+ "logits/chosen": -2.3890933990478516,
274
+ "logits/rejected": -2.305311918258667,
275
+ "logps/chosen": -189.14987182617188,
276
+ "logps/rejected": -180.1926727294922,
277
+ "loss": 0.4688,
278
+ "rewards/accuracies": 1.0,
279
+ "rewards/chosen": 0.028069380670785904,
280
+ "rewards/margins": 0.047098807990550995,
281
+ "rewards/rejected": -0.01902942731976509,
282
+ "step": 13
283
+ },
284
+ {
285
+ "debug/policy_chosen_logits": -2.4275922775268555,
286
+ "debug/policy_chosen_logps": -181.49740600585938,
287
+ "debug/policy_rejected_logits": -2.43969464302063,
288
+ "debug/policy_rejected_logps": -199.9078369140625,
289
+ "debug/reference_chosen_logps": -180.03472900390625,
290
+ "debug/reference_rejected_logps": -198.68141174316406,
291
+ "epoch": 0.30434782608695654,
292
+ "grad_norm": 2.4272251174632857,
293
+ "learning_rate": 1e-06,
294
+ "logits/chosen": -2.4275922775268555,
295
+ "logits/rejected": -2.43969464302063,
296
+ "logps/chosen": -181.49740600585938,
297
+ "logps/rejected": -199.9078369140625,
298
+ "loss": 0.4772,
299
+ "rewards/accuracies": 0.5,
300
+ "rewards/chosen": -0.014626597985625267,
301
+ "rewards/margins": -0.002362403552979231,
302
+ "rewards/rejected": -0.012264193966984749,
303
+ "step": 14
304
+ },
305
+ {
306
+ "debug/policy_chosen_logits": -2.47592830657959,
307
+ "debug/policy_chosen_logps": -164.09986877441406,
308
+ "debug/policy_rejected_logits": -2.435157537460327,
309
+ "debug/policy_rejected_logps": -176.61294555664062,
310
+ "debug/reference_chosen_logps": -163.7836151123047,
311
+ "debug/reference_rejected_logps": -173.32427978515625,
312
+ "epoch": 0.32608695652173914,
313
+ "grad_norm": 2.6909888775385027,
314
+ "learning_rate": 1e-06,
315
+ "logits/chosen": -2.47592830657959,
316
+ "logits/rejected": -2.435157537460327,
317
+ "logps/chosen": -164.09986877441406,
318
+ "logps/rejected": -176.61294555664062,
319
+ "loss": 0.4752,
320
+ "rewards/accuracies": 0.75,
321
+ "rewards/chosen": -0.0031626131385564804,
322
+ "rewards/margins": 0.029723986983299255,
323
+ "rewards/rejected": -0.032886601984500885,
324
+ "step": 15
325
+ },
326
+ {
327
+ "debug/policy_chosen_logits": -2.377653121948242,
328
+ "debug/policy_chosen_logps": -173.7503662109375,
329
+ "debug/policy_rejected_logits": -2.386281967163086,
330
+ "debug/policy_rejected_logps": -194.0179443359375,
331
+ "debug/reference_chosen_logps": -173.58297729492188,
332
+ "debug/reference_rejected_logps": -192.9736328125,
333
+ "epoch": 0.34782608695652173,
334
+ "grad_norm": 2.5551392813174902,
335
+ "learning_rate": 1e-06,
336
+ "logits/chosen": -2.377653121948242,
337
+ "logits/rejected": -2.386281967163086,
338
+ "logps/chosen": -173.7503662109375,
339
+ "logps/rejected": -194.0179443359375,
340
+ "loss": 0.4761,
341
+ "rewards/accuracies": 0.5,
342
+ "rewards/chosen": -0.0016740784049034119,
343
+ "rewards/margins": 0.0087689608335495,
344
+ "rewards/rejected": -0.010443039238452911,
345
+ "step": 16
346
+ },
347
+ {
348
+ "debug/policy_chosen_logits": -2.529132843017578,
349
+ "debug/policy_chosen_logps": -172.70501708984375,
350
+ "debug/policy_rejected_logits": -2.4587604999542236,
351
+ "debug/policy_rejected_logps": -186.12283325195312,
352
+ "debug/reference_chosen_logps": -172.2464599609375,
353
+ "debug/reference_rejected_logps": -186.38314819335938,
354
+ "epoch": 0.3695652173913043,
355
+ "grad_norm": 2.4118823503893054,
356
+ "learning_rate": 1e-06,
357
+ "logits/chosen": -2.529132843017578,
358
+ "logits/rejected": -2.4587604999542236,
359
+ "logps/chosen": -172.70501708984375,
360
+ "logps/rejected": -186.12283325195312,
361
+ "loss": 0.487,
362
+ "rewards/accuracies": 0.75,
363
+ "rewards/chosen": -0.0045854183845222,
364
+ "rewards/margins": -0.007188738323748112,
365
+ "rewards/rejected": 0.002603319473564625,
366
+ "step": 17
367
+ },
368
+ {
369
+ "debug/policy_chosen_logits": -2.4017813205718994,
370
+ "debug/policy_chosen_logps": -166.11911010742188,
371
+ "debug/policy_rejected_logits": -2.3311731815338135,
372
+ "debug/policy_rejected_logps": -182.46627807617188,
373
+ "debug/reference_chosen_logps": -168.52517700195312,
374
+ "debug/reference_rejected_logps": -179.03460693359375,
375
+ "epoch": 0.391304347826087,
376
+ "grad_norm": 2.615012586102309,
377
+ "learning_rate": 1e-06,
378
+ "logits/chosen": -2.4017813205718994,
379
+ "logits/rejected": -2.3311731815338135,
380
+ "logps/chosen": -166.11911010742188,
381
+ "logps/rejected": -182.46627807617188,
382
+ "loss": 0.4631,
383
+ "rewards/accuracies": 0.875,
384
+ "rewards/chosen": 0.02406078390777111,
385
+ "rewards/margins": 0.05837753415107727,
386
+ "rewards/rejected": -0.03431674838066101,
387
+ "step": 18
388
+ },
389
+ {
390
+ "debug/policy_chosen_logits": -2.4313623905181885,
391
+ "debug/policy_chosen_logps": -174.13351440429688,
392
+ "debug/policy_rejected_logits": -2.3800158500671387,
393
+ "debug/policy_rejected_logps": -218.58282470703125,
394
+ "debug/reference_chosen_logps": -172.78656005859375,
395
+ "debug/reference_rejected_logps": -214.70162963867188,
396
+ "epoch": 0.41304347826086957,
397
+ "grad_norm": 2.801324694410347,
398
+ "learning_rate": 1e-06,
399
+ "logits/chosen": -2.4313623905181885,
400
+ "logits/rejected": -2.3800158500671387,
401
+ "logps/chosen": -174.13351440429688,
402
+ "logps/rejected": -218.58282470703125,
403
+ "loss": 0.4571,
404
+ "rewards/accuracies": 0.75,
405
+ "rewards/chosen": -0.013469582423567772,
406
+ "rewards/margins": 0.025342464447021484,
407
+ "rewards/rejected": -0.038812048733234406,
408
+ "step": 19
409
+ },
410
+ {
411
+ "debug/policy_chosen_logits": -2.4221510887145996,
412
+ "debug/policy_chosen_logps": -193.74337768554688,
413
+ "debug/policy_rejected_logits": -2.429517984390259,
414
+ "debug/policy_rejected_logps": -211.8312530517578,
415
+ "debug/reference_chosen_logps": -197.71206665039062,
416
+ "debug/reference_rejected_logps": -212.77467346191406,
417
+ "epoch": 0.43478260869565216,
418
+ "grad_norm": 2.452728910615928,
419
+ "learning_rate": 1e-06,
420
+ "logits/chosen": -2.4221510887145996,
421
+ "logits/rejected": -2.429517984390259,
422
+ "logps/chosen": -193.74337768554688,
423
+ "logps/rejected": -211.8312530517578,
424
+ "loss": 0.468,
425
+ "rewards/accuracies": 0.5,
426
+ "rewards/chosen": 0.039686813950538635,
427
+ "rewards/margins": 0.030252492055296898,
428
+ "rewards/rejected": 0.009434318169951439,
429
+ "step": 20
430
+ },
431
+ {
432
+ "debug/policy_chosen_logits": -2.3384790420532227,
433
+ "debug/policy_chosen_logps": -168.90838623046875,
434
+ "debug/policy_rejected_logits": -2.249765396118164,
435
+ "debug/policy_rejected_logps": -170.010009765625,
436
+ "debug/reference_chosen_logps": -168.66502380371094,
437
+ "debug/reference_rejected_logps": -162.436279296875,
438
+ "epoch": 0.45652173913043476,
439
+ "grad_norm": 2.2296634238737805,
440
+ "learning_rate": 1e-06,
441
+ "logits/chosen": -2.3384790420532227,
442
+ "logits/rejected": -2.249765396118164,
443
+ "logps/chosen": -168.90838623046875,
444
+ "logps/rejected": -170.010009765625,
445
+ "loss": 0.4555,
446
+ "rewards/accuracies": 0.875,
447
+ "rewards/chosen": -0.002433512359857559,
448
+ "rewards/margins": 0.07330392301082611,
449
+ "rewards/rejected": -0.07573743164539337,
450
+ "step": 21
451
+ },
452
+ {
453
+ "debug/policy_chosen_logits": -2.49155855178833,
454
+ "debug/policy_chosen_logps": -202.97955322265625,
455
+ "debug/policy_rejected_logits": -2.5280747413635254,
456
+ "debug/policy_rejected_logps": -187.58743286132812,
457
+ "debug/reference_chosen_logps": -203.92587280273438,
458
+ "debug/reference_rejected_logps": -185.87338256835938,
459
+ "epoch": 0.4782608695652174,
460
+ "grad_norm": 2.481538979707535,
461
+ "learning_rate": 1e-06,
462
+ "logits/chosen": -2.49155855178833,
463
+ "logits/rejected": -2.5280747413635254,
464
+ "logps/chosen": -202.97955322265625,
465
+ "logps/rejected": -187.58743286132812,
466
+ "loss": 0.4759,
467
+ "rewards/accuracies": 0.5,
468
+ "rewards/chosen": 0.009463195689022541,
469
+ "rewards/margins": 0.02660369873046875,
470
+ "rewards/rejected": -0.017140503972768784,
471
+ "step": 22
472
+ },
473
+ {
474
+ "debug/policy_chosen_logits": -2.5137152671813965,
475
+ "debug/policy_chosen_logps": -181.94056701660156,
476
+ "debug/policy_rejected_logits": -2.5435562133789062,
477
+ "debug/policy_rejected_logps": -169.93206787109375,
478
+ "debug/reference_chosen_logps": -188.2291717529297,
479
+ "debug/reference_rejected_logps": -167.34231567382812,
480
+ "epoch": 0.5,
481
+ "grad_norm": 2.221961851151498,
482
+ "learning_rate": 1e-06,
483
+ "logits/chosen": -2.5137152671813965,
484
+ "logits/rejected": -2.5435562133789062,
485
+ "logps/chosen": -181.94056701660156,
486
+ "logps/rejected": -169.93206787109375,
487
+ "loss": 0.4645,
488
+ "rewards/accuracies": 0.75,
489
+ "rewards/chosen": 0.06288604438304901,
490
+ "rewards/margins": 0.08878357708454132,
491
+ "rewards/rejected": -0.025897536426782608,
492
+ "step": 23
493
+ },
494
+ {
495
+ "debug/policy_chosen_logits": -2.343871593475342,
496
+ "debug/policy_chosen_logps": -172.3570556640625,
497
+ "debug/policy_rejected_logits": -2.3897156715393066,
498
+ "debug/policy_rejected_logps": -182.73715209960938,
499
+ "debug/reference_chosen_logps": -171.8414306640625,
500
+ "debug/reference_rejected_logps": -182.32119750976562,
501
+ "epoch": 0.5217391304347826,
502
+ "grad_norm": 2.3144852334581345,
503
+ "learning_rate": 1e-06,
504
+ "logits/chosen": -2.343871593475342,
505
+ "logits/rejected": -2.3897156715393066,
506
+ "logps/chosen": -172.3570556640625,
507
+ "logps/rejected": -182.73715209960938,
508
+ "loss": 0.4726,
509
+ "rewards/accuracies": 0.25,
510
+ "rewards/chosen": -0.005156230181455612,
511
+ "rewards/margins": -0.0009968150407075882,
512
+ "rewards/rejected": -0.0041594114154577255,
513
+ "step": 24
514
+ },
515
+ {
516
+ "debug/policy_chosen_logits": -2.4783878326416016,
517
+ "debug/policy_chosen_logps": -181.16600036621094,
518
+ "debug/policy_rejected_logits": -2.4538094997406006,
519
+ "debug/policy_rejected_logps": -192.53765869140625,
520
+ "debug/reference_chosen_logps": -185.05032348632812,
521
+ "debug/reference_rejected_logps": -186.58950805664062,
522
+ "epoch": 0.5434782608695652,
523
+ "grad_norm": 2.4281399717516448,
524
+ "learning_rate": 1e-06,
525
+ "logits/chosen": -2.4783878326416016,
526
+ "logits/rejected": -2.4538094997406006,
527
+ "logps/chosen": -181.16600036621094,
528
+ "logps/rejected": -192.53765869140625,
529
+ "loss": 0.4538,
530
+ "rewards/accuracies": 0.75,
531
+ "rewards/chosen": 0.03884328901767731,
532
+ "rewards/margins": 0.09832490980625153,
533
+ "rewards/rejected": -0.05948162078857422,
534
+ "step": 25
535
+ },
536
+ {
537
+ "debug/policy_chosen_logits": -2.4092724323272705,
538
+ "debug/policy_chosen_logps": -160.72732543945312,
539
+ "debug/policy_rejected_logits": -2.477595329284668,
540
+ "debug/policy_rejected_logps": -175.16604614257812,
541
+ "debug/reference_chosen_logps": -166.90414428710938,
542
+ "debug/reference_rejected_logps": -173.4667510986328,
543
+ "epoch": 0.5652173913043478,
544
+ "grad_norm": 2.4404949726132137,
545
+ "learning_rate": 1e-06,
546
+ "logits/chosen": -2.4092724323272705,
547
+ "logits/rejected": -2.477595329284668,
548
+ "logps/chosen": -160.72732543945312,
549
+ "logps/rejected": -175.16604614257812,
550
+ "loss": 0.4506,
551
+ "rewards/accuracies": 0.75,
552
+ "rewards/chosen": 0.06176815181970596,
553
+ "rewards/margins": 0.07876109331846237,
554
+ "rewards/rejected": -0.016992948949337006,
555
+ "step": 26
556
+ },
557
+ {
558
+ "debug/policy_chosen_logits": -2.4233715534210205,
559
+ "debug/policy_chosen_logps": -190.73126220703125,
560
+ "debug/policy_rejected_logits": -2.382371187210083,
561
+ "debug/policy_rejected_logps": -197.88619995117188,
562
+ "debug/reference_chosen_logps": -196.33099365234375,
563
+ "debug/reference_rejected_logps": -193.63540649414062,
564
+ "epoch": 0.5869565217391305,
565
+ "grad_norm": 2.33043065534125,
566
+ "learning_rate": 1e-06,
567
+ "logits/chosen": -2.4233715534210205,
568
+ "logits/rejected": -2.382371187210083,
569
+ "logps/chosen": -190.73126220703125,
570
+ "logps/rejected": -197.88619995117188,
571
+ "loss": 0.4187,
572
+ "rewards/accuracies": 0.875,
573
+ "rewards/chosen": 0.05599748343229294,
574
+ "rewards/margins": 0.0985054224729538,
575
+ "rewards/rejected": -0.04250793531537056,
576
+ "step": 27
577
+ },
578
+ {
579
+ "debug/policy_chosen_logits": -2.5104820728302,
580
+ "debug/policy_chosen_logps": -170.4730682373047,
581
+ "debug/policy_rejected_logits": -2.4717085361480713,
582
+ "debug/policy_rejected_logps": -189.20010375976562,
583
+ "debug/reference_chosen_logps": -168.71560668945312,
584
+ "debug/reference_rejected_logps": -185.61746215820312,
585
+ "epoch": 0.6086956521739131,
586
+ "grad_norm": 2.8275079552922446,
587
+ "learning_rate": 1e-06,
588
+ "logits/chosen": -2.5104820728302,
589
+ "logits/rejected": -2.4717085361480713,
590
+ "logps/chosen": -170.4730682373047,
591
+ "logps/rejected": -189.20010375976562,
592
+ "loss": 0.4666,
593
+ "rewards/accuracies": 0.375,
594
+ "rewards/chosen": -0.01757451891899109,
595
+ "rewards/margins": 0.018252111971378326,
596
+ "rewards/rejected": -0.035826630890369415,
597
+ "step": 28
598
+ },
599
+ {
600
+ "debug/policy_chosen_logits": -2.484487771987915,
601
+ "debug/policy_chosen_logps": -186.02122497558594,
602
+ "debug/policy_rejected_logits": -2.4840569496154785,
603
+ "debug/policy_rejected_logps": -174.24012756347656,
604
+ "debug/reference_chosen_logps": -190.0591278076172,
605
+ "debug/reference_rejected_logps": -166.46307373046875,
606
+ "epoch": 0.6304347826086957,
607
+ "grad_norm": 2.5260328428308236,
608
+ "learning_rate": 1e-06,
609
+ "logits/chosen": -2.484487771987915,
610
+ "logits/rejected": -2.4840569496154785,
611
+ "logps/chosen": -186.02122497558594,
612
+ "logps/rejected": -174.24012756347656,
613
+ "loss": 0.4554,
614
+ "rewards/accuracies": 0.75,
615
+ "rewards/chosen": 0.04037889465689659,
616
+ "rewards/margins": 0.11814939230680466,
617
+ "rewards/rejected": -0.07777049392461777,
618
+ "step": 29
619
+ },
620
+ {
621
+ "debug/policy_chosen_logits": -2.4572229385375977,
622
+ "debug/policy_chosen_logps": -177.25079345703125,
623
+ "debug/policy_rejected_logits": -2.335451364517212,
624
+ "debug/policy_rejected_logps": -211.37869262695312,
625
+ "debug/reference_chosen_logps": -184.99151611328125,
626
+ "debug/reference_rejected_logps": -207.14004516601562,
627
+ "epoch": 0.6521739130434783,
628
+ "grad_norm": 2.7172712356585476,
629
+ "learning_rate": 1e-06,
630
+ "logits/chosen": -2.4572229385375977,
631
+ "logits/rejected": -2.335451364517212,
632
+ "logps/chosen": -177.25079345703125,
633
+ "logps/rejected": -211.37869262695312,
634
+ "loss": 0.463,
635
+ "rewards/accuracies": 0.625,
636
+ "rewards/chosen": 0.07740737497806549,
637
+ "rewards/margins": 0.11979396641254425,
638
+ "rewards/rejected": -0.04238658770918846,
639
+ "step": 30
640
+ },
641
+ {
642
+ "debug/policy_chosen_logits": -2.4997236728668213,
643
+ "debug/policy_chosen_logps": -176.38267517089844,
644
+ "debug/policy_rejected_logits": -2.3669979572296143,
645
+ "debug/policy_rejected_logps": -196.77554321289062,
646
+ "debug/reference_chosen_logps": -178.01614379882812,
647
+ "debug/reference_rejected_logps": -191.33193969726562,
648
+ "epoch": 0.6739130434782609,
649
+ "grad_norm": 2.353661137242158,
650
+ "learning_rate": 1e-06,
651
+ "logits/chosen": -2.4997236728668213,
652
+ "logits/rejected": -2.3669979572296143,
653
+ "logps/chosen": -176.38267517089844,
654
+ "logps/rejected": -196.77554321289062,
655
+ "loss": 0.4335,
656
+ "rewards/accuracies": 0.625,
657
+ "rewards/chosen": 0.01633470691740513,
658
+ "rewards/margins": 0.07077069580554962,
659
+ "rewards/rejected": -0.05443599820137024,
660
+ "step": 31
661
+ },
662
+ {
663
+ "debug/policy_chosen_logits": -2.4377670288085938,
664
+ "debug/policy_chosen_logps": -173.71165466308594,
665
+ "debug/policy_rejected_logits": -2.3115673065185547,
666
+ "debug/policy_rejected_logps": -178.8059539794922,
667
+ "debug/reference_chosen_logps": -179.61834716796875,
668
+ "debug/reference_rejected_logps": -181.16517639160156,
669
+ "epoch": 0.6956521739130435,
670
+ "grad_norm": 2.3528189249640397,
671
+ "learning_rate": 1e-06,
672
+ "logits/chosen": -2.4377670288085938,
673
+ "logits/rejected": -2.3115673065185547,
674
+ "logps/chosen": -173.71165466308594,
675
+ "logps/rejected": -178.8059539794922,
676
+ "loss": 0.4606,
677
+ "rewards/accuracies": 0.5,
678
+ "rewards/chosen": 0.05906686931848526,
679
+ "rewards/margins": 0.0354747548699379,
680
+ "rewards/rejected": 0.023592106997966766,
681
+ "step": 32
682
+ },
683
+ {
684
+ "debug/policy_chosen_logits": -2.467334747314453,
685
+ "debug/policy_chosen_logps": -188.59512329101562,
686
+ "debug/policy_rejected_logits": -2.487856388092041,
687
+ "debug/policy_rejected_logps": -200.94195556640625,
688
+ "debug/reference_chosen_logps": -190.67750549316406,
689
+ "debug/reference_rejected_logps": -195.129638671875,
690
+ "epoch": 0.717391304347826,
691
+ "grad_norm": 2.5411257448652553,
692
+ "learning_rate": 1e-06,
693
+ "logits/chosen": -2.467334747314453,
694
+ "logits/rejected": -2.487856388092041,
695
+ "logps/chosen": -188.59512329101562,
696
+ "logps/rejected": -200.94195556640625,
697
+ "loss": 0.4339,
698
+ "rewards/accuracies": 0.75,
699
+ "rewards/chosen": 0.020823916420340538,
700
+ "rewards/margins": 0.07894693315029144,
701
+ "rewards/rejected": -0.058123014867305756,
702
+ "step": 33
703
+ },
704
+ {
705
+ "debug/policy_chosen_logits": -2.3515822887420654,
706
+ "debug/policy_chosen_logps": -172.7677001953125,
707
+ "debug/policy_rejected_logits": -2.395439386367798,
708
+ "debug/policy_rejected_logps": -188.29266357421875,
709
+ "debug/reference_chosen_logps": -181.23960876464844,
710
+ "debug/reference_rejected_logps": -179.37924194335938,
711
+ "epoch": 0.7391304347826086,
712
+ "grad_norm": 2.3779889701354375,
713
+ "learning_rate": 1e-06,
714
+ "logits/chosen": -2.3515822887420654,
715
+ "logits/rejected": -2.395439386367798,
716
+ "logps/chosen": -172.7677001953125,
717
+ "logps/rejected": -188.29266357421875,
718
+ "loss": 0.4141,
719
+ "rewards/accuracies": 0.875,
720
+ "rewards/chosen": 0.08471919596195221,
721
+ "rewards/margins": 0.17385347187519073,
722
+ "rewards/rejected": -0.08913427591323853,
723
+ "step": 34
724
+ },
725
+ {
726
+ "debug/policy_chosen_logits": -2.4782862663269043,
727
+ "debug/policy_chosen_logps": -178.69082641601562,
728
+ "debug/policy_rejected_logits": -2.456272840499878,
729
+ "debug/policy_rejected_logps": -187.01419067382812,
730
+ "debug/reference_chosen_logps": -177.35031127929688,
731
+ "debug/reference_rejected_logps": -177.96685791015625,
732
+ "epoch": 0.7608695652173914,
733
+ "grad_norm": 2.3565954950933694,
734
+ "learning_rate": 1e-06,
735
+ "logits/chosen": -2.4782862663269043,
736
+ "logits/rejected": -2.456272840499878,
737
+ "logps/chosen": -178.69082641601562,
738
+ "logps/rejected": -187.01419067382812,
739
+ "loss": 0.4367,
740
+ "rewards/accuracies": 0.625,
741
+ "rewards/chosen": -0.013405149802565575,
742
+ "rewards/margins": 0.07706813514232635,
743
+ "rewards/rejected": -0.09047328680753708,
744
+ "step": 35
745
+ },
746
+ {
747
+ "debug/policy_chosen_logits": -2.457199811935425,
748
+ "debug/policy_chosen_logps": -173.6107177734375,
749
+ "debug/policy_rejected_logits": -2.484218120574951,
750
+ "debug/policy_rejected_logps": -184.22190856933594,
751
+ "debug/reference_chosen_logps": -185.9441375732422,
752
+ "debug/reference_rejected_logps": -179.55136108398438,
753
+ "epoch": 0.782608695652174,
754
+ "grad_norm": 2.429486721658497,
755
+ "learning_rate": 1e-06,
756
+ "logits/chosen": -2.457199811935425,
757
+ "logits/rejected": -2.484218120574951,
758
+ "logps/chosen": -173.6107177734375,
759
+ "logps/rejected": -184.22190856933594,
760
+ "loss": 0.4355,
761
+ "rewards/accuracies": 0.875,
762
+ "rewards/chosen": 0.12333421409130096,
763
+ "rewards/margins": 0.17003962397575378,
764
+ "rewards/rejected": -0.04670540988445282,
765
+ "step": 36
766
+ },
767
+ {
768
+ "debug/policy_chosen_logits": -2.4880294799804688,
769
+ "debug/policy_chosen_logps": -174.43316650390625,
770
+ "debug/policy_rejected_logits": -2.4997189044952393,
771
+ "debug/policy_rejected_logps": -177.84881591796875,
772
+ "debug/reference_chosen_logps": -177.31240844726562,
773
+ "debug/reference_rejected_logps": -165.77145385742188,
774
+ "epoch": 0.8043478260869565,
775
+ "grad_norm": 2.441269723279697,
776
+ "learning_rate": 1e-06,
777
+ "logits/chosen": -2.4880294799804688,
778
+ "logits/rejected": -2.4997189044952393,
779
+ "logps/chosen": -174.43316650390625,
780
+ "logps/rejected": -177.84881591796875,
781
+ "loss": 0.4394,
782
+ "rewards/accuracies": 1.0,
783
+ "rewards/chosen": 0.02879244089126587,
784
+ "rewards/margins": 0.14956600964069366,
785
+ "rewards/rejected": -0.12077357620000839,
786
+ "step": 37
787
+ },
788
+ {
789
+ "debug/policy_chosen_logits": -2.5809593200683594,
790
+ "debug/policy_chosen_logps": -181.38897705078125,
791
+ "debug/policy_rejected_logits": -2.483285665512085,
792
+ "debug/policy_rejected_logps": -190.49742126464844,
793
+ "debug/reference_chosen_logps": -187.25927734375,
794
+ "debug/reference_rejected_logps": -184.10122680664062,
795
+ "epoch": 0.8260869565217391,
796
+ "grad_norm": 2.36617525939136,
797
+ "learning_rate": 1e-06,
798
+ "logits/chosen": -2.5809593200683594,
799
+ "logits/rejected": -2.483285665512085,
800
+ "logps/chosen": -181.38897705078125,
801
+ "logps/rejected": -190.49742126464844,
802
+ "loss": 0.4358,
803
+ "rewards/accuracies": 0.75,
804
+ "rewards/chosen": 0.05870307236909866,
805
+ "rewards/margins": 0.12266507744789124,
806
+ "rewards/rejected": -0.06396199762821198,
807
+ "step": 38
808
+ },
809
+ {
810
+ "debug/policy_chosen_logits": -2.449005126953125,
811
+ "debug/policy_chosen_logps": -163.90740966796875,
812
+ "debug/policy_rejected_logits": -2.4301159381866455,
813
+ "debug/policy_rejected_logps": -194.1475067138672,
814
+ "debug/reference_chosen_logps": -169.6490478515625,
815
+ "debug/reference_rejected_logps": -184.88339233398438,
816
+ "epoch": 0.8478260869565217,
817
+ "grad_norm": 2.4604115357660445,
818
+ "learning_rate": 1e-06,
819
+ "logits/chosen": -2.449005126953125,
820
+ "logits/rejected": -2.4301159381866455,
821
+ "logps/chosen": -163.90740966796875,
822
+ "logps/rejected": -194.1475067138672,
823
+ "loss": 0.4329,
824
+ "rewards/accuracies": 0.75,
825
+ "rewards/chosen": 0.05741637200117111,
826
+ "rewards/margins": 0.15005755424499512,
827
+ "rewards/rejected": -0.0926411971449852,
828
+ "step": 39
829
+ },
830
+ {
831
+ "debug/policy_chosen_logits": -2.479243040084839,
832
+ "debug/policy_chosen_logps": -177.88658142089844,
833
+ "debug/policy_rejected_logits": -2.521466016769409,
834
+ "debug/policy_rejected_logps": -174.86874389648438,
835
+ "debug/reference_chosen_logps": -189.6613311767578,
836
+ "debug/reference_rejected_logps": -172.8925323486328,
837
+ "epoch": 0.8695652173913043,
838
+ "grad_norm": 2.6005589820274,
839
+ "learning_rate": 1e-06,
840
+ "logits/chosen": -2.479243040084839,
841
+ "logits/rejected": -2.521466016769409,
842
+ "logps/chosen": -177.88658142089844,
843
+ "logps/rejected": -174.86874389648438,
844
+ "loss": 0.3689,
845
+ "rewards/accuracies": 0.75,
846
+ "rewards/chosen": 0.11774755269289017,
847
+ "rewards/margins": 0.13750958442687988,
848
+ "rewards/rejected": -0.01976201869547367,
849
+ "step": 40
850
+ },
851
+ {
852
+ "debug/policy_chosen_logits": -2.5032005310058594,
853
+ "debug/policy_chosen_logps": -167.452392578125,
854
+ "debug/policy_rejected_logits": -2.390470027923584,
855
+ "debug/policy_rejected_logps": -187.92933654785156,
856
+ "debug/reference_chosen_logps": -170.6841583251953,
857
+ "debug/reference_rejected_logps": -191.27334594726562,
858
+ "epoch": 0.8913043478260869,
859
+ "grad_norm": 2.784651354754412,
860
+ "learning_rate": 1e-06,
861
+ "logits/chosen": -2.5032005310058594,
862
+ "logits/rejected": -2.390470027923584,
863
+ "logps/chosen": -167.452392578125,
864
+ "logps/rejected": -187.92933654785156,
865
+ "loss": 0.4514,
866
+ "rewards/accuracies": 0.25,
867
+ "rewards/chosen": 0.03231758251786232,
868
+ "rewards/margins": -0.0011225640773773193,
869
+ "rewards/rejected": 0.03344014659523964,
870
+ "step": 41
871
+ },
872
+ {
873
+ "debug/policy_chosen_logits": -2.5493698120117188,
874
+ "debug/policy_chosen_logps": -172.9954071044922,
875
+ "debug/policy_rejected_logits": -2.476661205291748,
876
+ "debug/policy_rejected_logps": -179.5677947998047,
877
+ "debug/reference_chosen_logps": -183.045654296875,
878
+ "debug/reference_rejected_logps": -181.4873046875,
879
+ "epoch": 0.9130434782608695,
880
+ "grad_norm": 2.677440788734919,
881
+ "learning_rate": 1e-06,
882
+ "logits/chosen": -2.5493698120117188,
883
+ "logits/rejected": -2.476661205291748,
884
+ "logps/chosen": -172.9954071044922,
885
+ "logps/rejected": -179.5677947998047,
886
+ "loss": 0.4056,
887
+ "rewards/accuracies": 0.625,
888
+ "rewards/chosen": 0.10050234198570251,
889
+ "rewards/margins": 0.08130710572004318,
890
+ "rewards/rejected": 0.019195232540369034,
891
+ "step": 42
892
+ },
893
+ {
894
+ "debug/policy_chosen_logits": -2.5432963371276855,
895
+ "debug/policy_chosen_logps": -175.61265563964844,
896
+ "debug/policy_rejected_logits": -2.4852747917175293,
897
+ "debug/policy_rejected_logps": -201.0880584716797,
898
+ "debug/reference_chosen_logps": -176.60537719726562,
899
+ "debug/reference_rejected_logps": -192.08706665039062,
900
+ "epoch": 0.9347826086956522,
901
+ "grad_norm": 2.7145887961284068,
902
+ "learning_rate": 1e-06,
903
+ "logits/chosen": -2.5432963371276855,
904
+ "logits/rejected": -2.4852747917175293,
905
+ "logps/chosen": -175.61265563964844,
906
+ "logps/rejected": -201.0880584716797,
907
+ "loss": 0.4241,
908
+ "rewards/accuracies": 0.625,
909
+ "rewards/chosen": 0.009927215054631233,
910
+ "rewards/margins": 0.09993711113929749,
911
+ "rewards/rejected": -0.0900098979473114,
912
+ "step": 43
913
+ },
914
+ {
915
+ "debug/policy_chosen_logits": -2.497959852218628,
916
+ "debug/policy_chosen_logps": -174.22557067871094,
917
+ "debug/policy_rejected_logits": -2.427154541015625,
918
+ "debug/policy_rejected_logps": -183.83010864257812,
919
+ "debug/reference_chosen_logps": -186.29788208007812,
920
+ "debug/reference_rejected_logps": -183.1434326171875,
921
+ "epoch": 0.9565217391304348,
922
+ "grad_norm": 2.581573767785096,
923
+ "learning_rate": 1e-06,
924
+ "logits/chosen": -2.497959852218628,
925
+ "logits/rejected": -2.427154541015625,
926
+ "logps/chosen": -174.22557067871094,
927
+ "logps/rejected": -183.83010864257812,
928
+ "loss": 0.4177,
929
+ "rewards/accuracies": 0.75,
930
+ "rewards/chosen": 0.12072312831878662,
931
+ "rewards/margins": 0.1275898963212967,
932
+ "rewards/rejected": -0.006866758689284325,
933
+ "step": 44
934
+ },
935
+ {
936
+ "debug/policy_chosen_logits": -2.5587399005889893,
937
+ "debug/policy_chosen_logps": -194.60675048828125,
938
+ "debug/policy_rejected_logits": -2.4343531131744385,
939
+ "debug/policy_rejected_logps": -195.62185668945312,
940
+ "debug/reference_chosen_logps": -198.23883056640625,
941
+ "debug/reference_rejected_logps": -180.94686889648438,
942
+ "epoch": 0.9782608695652174,
943
+ "grad_norm": 2.815838392583519,
944
+ "learning_rate": 1e-06,
945
+ "logits/chosen": -2.5587399005889893,
946
+ "logits/rejected": -2.4343531131744385,
947
+ "logps/chosen": -194.60675048828125,
948
+ "logps/rejected": -195.62185668945312,
949
+ "loss": 0.4548,
950
+ "rewards/accuracies": 0.75,
951
+ "rewards/chosen": 0.036320820450782776,
952
+ "rewards/margins": 0.18307071924209595,
953
+ "rewards/rejected": -0.14674988389015198,
954
+ "step": 45
955
+ },
956
+ {
957
+ "debug/policy_chosen_logits": -2.663532257080078,
958
+ "debug/policy_chosen_logps": -181.77276611328125,
959
+ "debug/policy_rejected_logits": -2.5467002391815186,
960
+ "debug/policy_rejected_logps": -176.77719116210938,
961
+ "debug/reference_chosen_logps": -182.98736572265625,
962
+ "debug/reference_rejected_logps": -163.69749450683594,
963
+ "epoch": 1.0,
964
+ "grad_norm": 2.1986019501463825,
965
+ "learning_rate": 1e-06,
966
+ "logits/chosen": -2.663532257080078,
967
+ "logits/rejected": -2.5467002391815186,
968
+ "logps/chosen": -181.77276611328125,
969
+ "logps/rejected": -176.77719116210938,
970
+ "loss": 0.3328,
971
+ "rewards/accuracies": 0.875,
972
+ "rewards/chosen": 0.012146133929491043,
973
+ "rewards/margins": 0.14294317364692688,
974
+ "rewards/rejected": -0.13079702854156494,
975
+ "step": 46
976
+ },
977
+ {
978
+ "epoch": 1.0,
979
+ "step": 46,
980
+ "total_flos": 0.0,
981
+ "train_loss": 0.4573696670324906,
982
+ "train_runtime": 475.0336,
983
+ "train_samples_per_second": 6.067,
984
+ "train_steps_per_second": 0.097
985
+ }
986
+ ],
987
+ "logging_steps": 1,
988
+ "max_steps": 46,
989
+ "num_input_tokens_seen": 0,
990
+ "num_train_epochs": 1,
991
+ "save_steps": 500,
992
+ "stateful_callbacks": {
993
+ "TrainerControl": {
994
+ "args": {
995
+ "should_epoch_stop": false,
996
+ "should_evaluate": false,
997
+ "should_log": false,
998
+ "should_save": true,
999
+ "should_training_stop": true
1000
+ },
1001
+ "attributes": {}
1002
+ }
1003
+ },
1004
+ "total_flos": 0.0,
1005
+ "train_batch_size": 8,
1006
+ "trial_name": null,
1007
+ "trial_params": null
1008
+ }