yiran-wang3 commited on
Commit
bf10b05
1 Parent(s): 4346f08

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 +1029 -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_iter2
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_iter2-binarized
12
+ model-index:
13
+ - name: qwen1_chat_adamw_iter3
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_iter3
21
+
22
+ This model is a fine-tuned version of [yiran-wang3/qwen1_chat_adamw_iter2](https://huggingface.co/yiran-wang3/qwen1_chat_adamw_iter2) on the self-generate/qw1_sppo_hard_new_cn_mining_oj_iter2-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": 0.4919274278143619,
5
+ "train_runtime": 164.7654,
6
+ "train_samples": 2968,
7
+ "train_samples_per_second": 18.013,
8
+ "train_steps_per_second": 0.285
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": 0.4919274278143619,
5
+ "train_runtime": 164.7654,
6
+ "train_samples": 2968,
7
+ "train_samples_per_second": 18.013,
8
+ "train_steps_per_second": 0.285
9
+ }
trainer_state.json ADDED
@@ -0,0 +1,1029 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 1.0,
5
+ "eval_steps": 100,
6
+ "global_step": 47,
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": -0.8068546652793884,
13
+ "debug/policy_chosen_logps": -174.92620849609375,
14
+ "debug/policy_rejected_logits": -0.8490269184112549,
15
+ "debug/policy_rejected_logps": -157.37281799316406,
16
+ "debug/reference_chosen_logps": -174.92620849609375,
17
+ "debug/reference_rejected_logps": -157.37281799316406,
18
+ "epoch": 0.02127659574468085,
19
+ "grad_norm": 6.984644017712198,
20
+ "learning_rate": 1e-06,
21
+ "logits/chosen": -0.8068546652793884,
22
+ "logits/rejected": -0.8490269184112549,
23
+ "logps/chosen": -174.92620849609375,
24
+ "logps/rejected": -157.37281799316406,
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": -0.9022992253303528,
34
+ "debug/policy_chosen_logps": -143.34872436523438,
35
+ "debug/policy_rejected_logits": -0.8076929450035095,
36
+ "debug/policy_rejected_logps": -161.58590698242188,
37
+ "debug/reference_chosen_logps": -143.50845336914062,
38
+ "debug/reference_rejected_logps": -162.044921875,
39
+ "epoch": 0.0425531914893617,
40
+ "grad_norm": 6.185349544683876,
41
+ "learning_rate": 1e-06,
42
+ "logits/chosen": -0.9022992253303528,
43
+ "logits/rejected": -0.8076929450035095,
44
+ "logps/chosen": -143.34872436523438,
45
+ "logps/rejected": -161.58590698242188,
46
+ "loss": 0.5018,
47
+ "rewards/accuracies": 0.375,
48
+ "rewards/chosen": 0.0015972137916833162,
49
+ "rewards/margins": -0.0029931068420410156,
50
+ "rewards/rejected": 0.0045903208665549755,
51
+ "step": 2
52
+ },
53
+ {
54
+ "debug/policy_chosen_logits": -0.734718382358551,
55
+ "debug/policy_chosen_logps": -165.74978637695312,
56
+ "debug/policy_rejected_logits": -0.8039400577545166,
57
+ "debug/policy_rejected_logps": -155.00161743164062,
58
+ "debug/reference_chosen_logps": -166.1104736328125,
59
+ "debug/reference_rejected_logps": -155.65774536132812,
60
+ "epoch": 0.06382978723404255,
61
+ "grad_norm": 6.173007368590962,
62
+ "learning_rate": 1e-06,
63
+ "logits/chosen": -0.734718382358551,
64
+ "logits/rejected": -0.8039400577545166,
65
+ "logps/chosen": -165.74978637695312,
66
+ "logps/rejected": -155.00161743164062,
67
+ "loss": 0.5003,
68
+ "rewards/accuracies": 0.25,
69
+ "rewards/chosen": 0.0036070060450583696,
70
+ "rewards/margins": -0.002954263472929597,
71
+ "rewards/rejected": 0.006561269983649254,
72
+ "step": 3
73
+ },
74
+ {
75
+ "debug/policy_chosen_logits": -0.724987268447876,
76
+ "debug/policy_chosen_logps": -170.4888153076172,
77
+ "debug/policy_rejected_logits": -0.8433169722557068,
78
+ "debug/policy_rejected_logps": -181.96383666992188,
79
+ "debug/reference_chosen_logps": -171.01820373535156,
80
+ "debug/reference_rejected_logps": -182.69778442382812,
81
+ "epoch": 0.0851063829787234,
82
+ "grad_norm": 7.7885462092667535,
83
+ "learning_rate": 1e-06,
84
+ "logits/chosen": -0.724987268447876,
85
+ "logits/rejected": -0.8433169722557068,
86
+ "logps/chosen": -170.4888153076172,
87
+ "logps/rejected": -181.96383666992188,
88
+ "loss": 0.5001,
89
+ "rewards/accuracies": 0.25,
90
+ "rewards/chosen": 0.005293817259371281,
91
+ "rewards/margins": -0.0020457743667066097,
92
+ "rewards/rejected": 0.00733959162607789,
93
+ "step": 4
94
+ },
95
+ {
96
+ "debug/policy_chosen_logits": -1.0596671104431152,
97
+ "debug/policy_chosen_logps": -158.74615478515625,
98
+ "debug/policy_rejected_logits": -0.9138705730438232,
99
+ "debug/policy_rejected_logps": -172.4247589111328,
100
+ "debug/reference_chosen_logps": -159.16357421875,
101
+ "debug/reference_rejected_logps": -172.97216796875,
102
+ "epoch": 0.10638297872340426,
103
+ "grad_norm": 7.233824957283756,
104
+ "learning_rate": 1e-06,
105
+ "logits/chosen": -1.0596671104431152,
106
+ "logits/rejected": -0.9138705730438232,
107
+ "logps/chosen": -158.74615478515625,
108
+ "logps/rejected": -172.4247589111328,
109
+ "loss": 0.5014,
110
+ "rewards/accuracies": 0.5,
111
+ "rewards/chosen": 0.004174098838120699,
112
+ "rewards/margins": -0.0012998387683182955,
113
+ "rewards/rejected": 0.005473937373608351,
114
+ "step": 5
115
+ },
116
+ {
117
+ "debug/policy_chosen_logits": -0.8790372014045715,
118
+ "debug/policy_chosen_logps": -176.21961975097656,
119
+ "debug/policy_rejected_logits": -0.772028923034668,
120
+ "debug/policy_rejected_logps": -168.8019256591797,
121
+ "debug/reference_chosen_logps": -175.71527099609375,
122
+ "debug/reference_rejected_logps": -168.62828063964844,
123
+ "epoch": 0.1276595744680851,
124
+ "grad_norm": 6.529443105978694,
125
+ "learning_rate": 1e-06,
126
+ "logits/chosen": -0.8790372014045715,
127
+ "logits/rejected": -0.772028923034668,
128
+ "logps/chosen": -176.21961975097656,
129
+ "logps/rejected": -168.8019256591797,
130
+ "loss": 0.4993,
131
+ "rewards/accuracies": 0.5,
132
+ "rewards/chosen": -0.005043611396104097,
133
+ "rewards/margins": -0.003307218663394451,
134
+ "rewards/rejected": -0.0017363927327096462,
135
+ "step": 6
136
+ },
137
+ {
138
+ "debug/policy_chosen_logits": -0.8641963601112366,
139
+ "debug/policy_chosen_logps": -183.53302001953125,
140
+ "debug/policy_rejected_logits": -0.7855976223945618,
141
+ "debug/policy_rejected_logps": -169.0634765625,
142
+ "debug/reference_chosen_logps": -182.7043914794922,
143
+ "debug/reference_rejected_logps": -168.21730041503906,
144
+ "epoch": 0.14893617021276595,
145
+ "grad_norm": 6.8645227771260595,
146
+ "learning_rate": 1e-06,
147
+ "logits/chosen": -0.8641963601112366,
148
+ "logits/rejected": -0.7855976223945618,
149
+ "logps/chosen": -183.53302001953125,
150
+ "logps/rejected": -169.0634765625,
151
+ "loss": 0.4984,
152
+ "rewards/accuracies": 0.375,
153
+ "rewards/chosen": -0.008286199532449245,
154
+ "rewards/margins": 0.00017546641174703836,
155
+ "rewards/rejected": -0.00846166629344225,
156
+ "step": 7
157
+ },
158
+ {
159
+ "debug/policy_chosen_logits": -0.8728761076927185,
160
+ "debug/policy_chosen_logps": -161.760009765625,
161
+ "debug/policy_rejected_logits": -0.6517330408096313,
162
+ "debug/policy_rejected_logps": -192.8695068359375,
163
+ "debug/reference_chosen_logps": -159.8403778076172,
164
+ "debug/reference_rejected_logps": -191.3787841796875,
165
+ "epoch": 0.1702127659574468,
166
+ "grad_norm": 6.69812102779195,
167
+ "learning_rate": 1e-06,
168
+ "logits/chosen": -0.8728761076927185,
169
+ "logits/rejected": -0.6517330408096313,
170
+ "logps/chosen": -161.760009765625,
171
+ "logps/rejected": -192.8695068359375,
172
+ "loss": 0.5021,
173
+ "rewards/accuracies": 0.25,
174
+ "rewards/chosen": -0.019196156412363052,
175
+ "rewards/margins": -0.004288902040570974,
176
+ "rewards/rejected": -0.014907254837453365,
177
+ "step": 8
178
+ },
179
+ {
180
+ "debug/policy_chosen_logits": -0.9805322885513306,
181
+ "debug/policy_chosen_logps": -149.35302734375,
182
+ "debug/policy_rejected_logits": -0.6159168481826782,
183
+ "debug/policy_rejected_logps": -176.2978515625,
184
+ "debug/reference_chosen_logps": -149.08575439453125,
185
+ "debug/reference_rejected_logps": -175.6068572998047,
186
+ "epoch": 0.19148936170212766,
187
+ "grad_norm": 6.531738664013216,
188
+ "learning_rate": 1e-06,
189
+ "logits/chosen": -0.9805322885513306,
190
+ "logits/rejected": -0.6159168481826782,
191
+ "logps/chosen": -149.35302734375,
192
+ "logps/rejected": -176.2978515625,
193
+ "loss": 0.4995,
194
+ "rewards/accuracies": 0.75,
195
+ "rewards/chosen": -0.002672843635082245,
196
+ "rewards/margins": 0.004237032029777765,
197
+ "rewards/rejected": -0.0069098761305212975,
198
+ "step": 9
199
+ },
200
+ {
201
+ "debug/policy_chosen_logits": -0.8477765321731567,
202
+ "debug/policy_chosen_logps": -157.08721923828125,
203
+ "debug/policy_rejected_logits": -0.9879372119903564,
204
+ "debug/policy_rejected_logps": -152.79302978515625,
205
+ "debug/reference_chosen_logps": -155.71051025390625,
206
+ "debug/reference_rejected_logps": -152.11395263671875,
207
+ "epoch": 0.2127659574468085,
208
+ "grad_norm": 6.82652212657647,
209
+ "learning_rate": 1e-06,
210
+ "logits/chosen": -0.8477765321731567,
211
+ "logits/rejected": -0.9879372119903564,
212
+ "logps/chosen": -157.08721923828125,
213
+ "logps/rejected": -152.79302978515625,
214
+ "loss": 0.5003,
215
+ "rewards/accuracies": 0.5,
216
+ "rewards/chosen": -0.013767031952738762,
217
+ "rewards/margins": -0.006976260803639889,
218
+ "rewards/rejected": -0.006790771149098873,
219
+ "step": 10
220
+ },
221
+ {
222
+ "debug/policy_chosen_logits": -0.726215124130249,
223
+ "debug/policy_chosen_logps": -182.1802978515625,
224
+ "debug/policy_rejected_logits": -0.8878192901611328,
225
+ "debug/policy_rejected_logps": -140.88198852539062,
226
+ "debug/reference_chosen_logps": -181.7808380126953,
227
+ "debug/reference_rejected_logps": -140.300537109375,
228
+ "epoch": 0.23404255319148937,
229
+ "grad_norm": 5.656212124505056,
230
+ "learning_rate": 1e-06,
231
+ "logits/chosen": -0.726215124130249,
232
+ "logits/rejected": -0.8878192901611328,
233
+ "logps/chosen": -182.1802978515625,
234
+ "logps/rejected": -140.88198852539062,
235
+ "loss": 0.4998,
236
+ "rewards/accuracies": 0.625,
237
+ "rewards/chosen": -0.0039947605691850185,
238
+ "rewards/margins": 0.001819915371015668,
239
+ "rewards/rejected": -0.00581467617303133,
240
+ "step": 11
241
+ },
242
+ {
243
+ "debug/policy_chosen_logits": -1.11215341091156,
244
+ "debug/policy_chosen_logps": -134.15692138671875,
245
+ "debug/policy_rejected_logits": -0.9618105888366699,
246
+ "debug/policy_rejected_logps": -170.78240966796875,
247
+ "debug/reference_chosen_logps": -135.10552978515625,
248
+ "debug/reference_rejected_logps": -171.48892211914062,
249
+ "epoch": 0.2553191489361702,
250
+ "grad_norm": 6.646207906595571,
251
+ "learning_rate": 1e-06,
252
+ "logits/chosen": -1.11215341091156,
253
+ "logits/rejected": -0.9618105888366699,
254
+ "logps/chosen": -134.15692138671875,
255
+ "logps/rejected": -170.78240966796875,
256
+ "loss": 0.498,
257
+ "rewards/accuracies": 0.5,
258
+ "rewards/chosen": 0.009486007504165173,
259
+ "rewards/margins": 0.00242087384685874,
260
+ "rewards/rejected": 0.007065133657306433,
261
+ "step": 12
262
+ },
263
+ {
264
+ "debug/policy_chosen_logits": -0.7991014719009399,
265
+ "debug/policy_chosen_logps": -162.961181640625,
266
+ "debug/policy_rejected_logits": -0.8125512003898621,
267
+ "debug/policy_rejected_logps": -164.0631103515625,
268
+ "debug/reference_chosen_logps": -163.7264862060547,
269
+ "debug/reference_rejected_logps": -164.45962524414062,
270
+ "epoch": 0.2765957446808511,
271
+ "grad_norm": 6.71678781200528,
272
+ "learning_rate": 1e-06,
273
+ "logits/chosen": -0.7991014719009399,
274
+ "logits/rejected": -0.8125512003898621,
275
+ "logps/chosen": -162.961181640625,
276
+ "logps/rejected": -164.0631103515625,
277
+ "loss": 0.4968,
278
+ "rewards/accuracies": 0.5,
279
+ "rewards/chosen": 0.007653169799596071,
280
+ "rewards/margins": 0.0036880772095173597,
281
+ "rewards/rejected": 0.003965091425925493,
282
+ "step": 13
283
+ },
284
+ {
285
+ "debug/policy_chosen_logits": -0.7829986214637756,
286
+ "debug/policy_chosen_logps": -166.5230712890625,
287
+ "debug/policy_rejected_logits": -0.7261733412742615,
288
+ "debug/policy_rejected_logps": -163.46420288085938,
289
+ "debug/reference_chosen_logps": -167.57135009765625,
290
+ "debug/reference_rejected_logps": -164.37411499023438,
291
+ "epoch": 0.2978723404255319,
292
+ "grad_norm": 6.835923036694967,
293
+ "learning_rate": 1e-06,
294
+ "logits/chosen": -0.7829986214637756,
295
+ "logits/rejected": -0.7261733412742615,
296
+ "logps/chosen": -166.5230712890625,
297
+ "logps/rejected": -163.46420288085938,
298
+ "loss": 0.4935,
299
+ "rewards/accuracies": 0.625,
300
+ "rewards/chosen": 0.010482797399163246,
301
+ "rewards/margins": 0.0013835998252034187,
302
+ "rewards/rejected": 0.009099196642637253,
303
+ "step": 14
304
+ },
305
+ {
306
+ "debug/policy_chosen_logits": -0.8212059736251831,
307
+ "debug/policy_chosen_logps": -142.06138610839844,
308
+ "debug/policy_rejected_logits": -0.658382773399353,
309
+ "debug/policy_rejected_logps": -181.79806518554688,
310
+ "debug/reference_chosen_logps": -142.90821838378906,
311
+ "debug/reference_rejected_logps": -181.2681121826172,
312
+ "epoch": 0.3191489361702128,
313
+ "grad_norm": 5.973889273303634,
314
+ "learning_rate": 1e-06,
315
+ "logits/chosen": -0.8212059736251831,
316
+ "logits/rejected": -0.658382773399353,
317
+ "logps/chosen": -142.06138610839844,
318
+ "logps/rejected": -181.79806518554688,
319
+ "loss": 0.4974,
320
+ "rewards/accuracies": 0.75,
321
+ "rewards/chosen": 0.008468389511108398,
322
+ "rewards/margins": 0.013767938129603863,
323
+ "rewards/rejected": -0.005299548618495464,
324
+ "step": 15
325
+ },
326
+ {
327
+ "debug/policy_chosen_logits": -0.8922485113143921,
328
+ "debug/policy_chosen_logps": -171.1387176513672,
329
+ "debug/policy_rejected_logits": -0.8397103548049927,
330
+ "debug/policy_rejected_logps": -196.10498046875,
331
+ "debug/reference_chosen_logps": -172.7020721435547,
332
+ "debug/reference_rejected_logps": -194.37750244140625,
333
+ "epoch": 0.3404255319148936,
334
+ "grad_norm": 6.500565528945796,
335
+ "learning_rate": 1e-06,
336
+ "logits/chosen": -0.8922485113143921,
337
+ "logits/rejected": -0.8397103548049927,
338
+ "logps/chosen": -171.1387176513672,
339
+ "logps/rejected": -196.10498046875,
340
+ "loss": 0.4935,
341
+ "rewards/accuracies": 0.625,
342
+ "rewards/chosen": 0.01563352718949318,
343
+ "rewards/margins": 0.032908324152231216,
344
+ "rewards/rejected": -0.017274796962738037,
345
+ "step": 16
346
+ },
347
+ {
348
+ "debug/policy_chosen_logits": -0.5738427639007568,
349
+ "debug/policy_chosen_logps": -199.78443908691406,
350
+ "debug/policy_rejected_logits": -0.7579363584518433,
351
+ "debug/policy_rejected_logps": -186.72186279296875,
352
+ "debug/reference_chosen_logps": -198.720458984375,
353
+ "debug/reference_rejected_logps": -186.148193359375,
354
+ "epoch": 0.3617021276595745,
355
+ "grad_norm": 6.391107539842202,
356
+ "learning_rate": 1e-06,
357
+ "logits/chosen": -0.5738427639007568,
358
+ "logits/rejected": -0.7579363584518433,
359
+ "logps/chosen": -199.78443908691406,
360
+ "logps/rejected": -186.72186279296875,
361
+ "loss": 0.4969,
362
+ "rewards/accuracies": 0.5,
363
+ "rewards/chosen": -0.010639895685017109,
364
+ "rewards/margins": -0.0049033742398023605,
365
+ "rewards/rejected": -0.005736521910876036,
366
+ "step": 17
367
+ },
368
+ {
369
+ "debug/policy_chosen_logits": -0.6704298853874207,
370
+ "debug/policy_chosen_logps": -189.298095703125,
371
+ "debug/policy_rejected_logits": -0.8089768886566162,
372
+ "debug/policy_rejected_logps": -157.5814666748047,
373
+ "debug/reference_chosen_logps": -188.82077026367188,
374
+ "debug/reference_rejected_logps": -157.17251586914062,
375
+ "epoch": 0.3829787234042553,
376
+ "grad_norm": 6.255280365272668,
377
+ "learning_rate": 1e-06,
378
+ "logits/chosen": -0.6704298853874207,
379
+ "logits/rejected": -0.8089768886566162,
380
+ "logps/chosen": -189.298095703125,
381
+ "logps/rejected": -157.5814666748047,
382
+ "loss": 0.4995,
383
+ "rewards/accuracies": 0.5,
384
+ "rewards/chosen": -0.004773206077516079,
385
+ "rewards/margins": -0.0006837937980890274,
386
+ "rewards/rejected": -0.0040894122794270515,
387
+ "step": 18
388
+ },
389
+ {
390
+ "debug/policy_chosen_logits": -0.7190896272659302,
391
+ "debug/policy_chosen_logps": -192.3218231201172,
392
+ "debug/policy_rejected_logits": -0.7404218316078186,
393
+ "debug/policy_rejected_logps": -197.3118896484375,
394
+ "debug/reference_chosen_logps": -191.5443878173828,
395
+ "debug/reference_rejected_logps": -194.45379638671875,
396
+ "epoch": 0.40425531914893614,
397
+ "grad_norm": 6.623087284607447,
398
+ "learning_rate": 1e-06,
399
+ "logits/chosen": -0.7190896272659302,
400
+ "logits/rejected": -0.7404218316078186,
401
+ "logps/chosen": -192.3218231201172,
402
+ "logps/rejected": -197.3118896484375,
403
+ "loss": 0.494,
404
+ "rewards/accuracies": 0.5,
405
+ "rewards/chosen": -0.00777435302734375,
406
+ "rewards/margins": 0.020806580781936646,
407
+ "rewards/rejected": -0.028580933809280396,
408
+ "step": 19
409
+ },
410
+ {
411
+ "debug/policy_chosen_logits": -0.9601216912269592,
412
+ "debug/policy_chosen_logps": -163.33322143554688,
413
+ "debug/policy_rejected_logits": -0.9078224301338196,
414
+ "debug/policy_rejected_logps": -165.64710998535156,
415
+ "debug/reference_chosen_logps": -162.98707580566406,
416
+ "debug/reference_rejected_logps": -163.82925415039062,
417
+ "epoch": 0.425531914893617,
418
+ "grad_norm": 6.4425625725133555,
419
+ "learning_rate": 1e-06,
420
+ "logits/chosen": -0.9601216912269592,
421
+ "logits/rejected": -0.9078224301338196,
422
+ "logps/chosen": -163.33322143554688,
423
+ "logps/rejected": -165.64710998535156,
424
+ "loss": 0.4936,
425
+ "rewards/accuracies": 0.75,
426
+ "rewards/chosen": -0.0034613802563399076,
427
+ "rewards/margins": 0.014717292971909046,
428
+ "rewards/rejected": -0.018178671598434448,
429
+ "step": 20
430
+ },
431
+ {
432
+ "debug/policy_chosen_logits": -0.7237945199012756,
433
+ "debug/policy_chosen_logps": -172.45687866210938,
434
+ "debug/policy_rejected_logits": -0.920426070690155,
435
+ "debug/policy_rejected_logps": -157.43887329101562,
436
+ "debug/reference_chosen_logps": -171.19932556152344,
437
+ "debug/reference_rejected_logps": -158.15611267089844,
438
+ "epoch": 0.44680851063829785,
439
+ "grad_norm": 6.339140504334335,
440
+ "learning_rate": 1e-06,
441
+ "logits/chosen": -0.7237945199012756,
442
+ "logits/rejected": -0.920426070690155,
443
+ "logps/chosen": -172.45687866210938,
444
+ "logps/rejected": -157.43887329101562,
445
+ "loss": 0.5013,
446
+ "rewards/accuracies": 0.5,
447
+ "rewards/chosen": -0.012575510889291763,
448
+ "rewards/margins": -0.019747886806726456,
449
+ "rewards/rejected": 0.007172374054789543,
450
+ "step": 21
451
+ },
452
+ {
453
+ "debug/policy_chosen_logits": -0.5433034896850586,
454
+ "debug/policy_chosen_logps": -201.61782836914062,
455
+ "debug/policy_rejected_logits": -0.7539889216423035,
456
+ "debug/policy_rejected_logps": -197.580322265625,
457
+ "debug/reference_chosen_logps": -201.07705688476562,
458
+ "debug/reference_rejected_logps": -197.327880859375,
459
+ "epoch": 0.46808510638297873,
460
+ "grad_norm": 6.06193663791783,
461
+ "learning_rate": 1e-06,
462
+ "logits/chosen": -0.5433034896850586,
463
+ "logits/rejected": -0.7539889216423035,
464
+ "logps/chosen": -201.61782836914062,
465
+ "logps/rejected": -197.580322265625,
466
+ "loss": 0.4999,
467
+ "rewards/accuracies": 0.5,
468
+ "rewards/chosen": -0.005407695658504963,
469
+ "rewards/margins": -0.002883119275793433,
470
+ "rewards/rejected": -0.002524576149880886,
471
+ "step": 22
472
+ },
473
+ {
474
+ "debug/policy_chosen_logits": -0.7976532578468323,
475
+ "debug/policy_chosen_logps": -157.63087463378906,
476
+ "debug/policy_rejected_logits": -1.0295953750610352,
477
+ "debug/policy_rejected_logps": -135.0724639892578,
478
+ "debug/reference_chosen_logps": -158.51002502441406,
479
+ "debug/reference_rejected_logps": -134.96658325195312,
480
+ "epoch": 0.48936170212765956,
481
+ "grad_norm": 6.533766675168247,
482
+ "learning_rate": 1e-06,
483
+ "logits/chosen": -0.7976532578468323,
484
+ "logits/rejected": -1.0295953750610352,
485
+ "logps/chosen": -157.63087463378906,
486
+ "logps/rejected": -135.0724639892578,
487
+ "loss": 0.4975,
488
+ "rewards/accuracies": 0.75,
489
+ "rewards/chosen": 0.008791436441242695,
490
+ "rewards/margins": 0.009850319474935532,
491
+ "rewards/rejected": -0.001058883499354124,
492
+ "step": 23
493
+ },
494
+ {
495
+ "debug/policy_chosen_logits": -0.738879382610321,
496
+ "debug/policy_chosen_logps": -177.37928771972656,
497
+ "debug/policy_rejected_logits": -0.7938289642333984,
498
+ "debug/policy_rejected_logps": -177.46755981445312,
499
+ "debug/reference_chosen_logps": -179.34732055664062,
500
+ "debug/reference_rejected_logps": -176.83135986328125,
501
+ "epoch": 0.5106382978723404,
502
+ "grad_norm": 5.87105022701246,
503
+ "learning_rate": 1e-06,
504
+ "logits/chosen": -0.738879382610321,
505
+ "logits/rejected": -0.7938289642333984,
506
+ "logps/chosen": -177.37928771972656,
507
+ "logps/rejected": -177.46755981445312,
508
+ "loss": 0.492,
509
+ "rewards/accuracies": 0.625,
510
+ "rewards/chosen": 0.01968032866716385,
511
+ "rewards/margins": 0.026042308658361435,
512
+ "rewards/rejected": -0.006361980922520161,
513
+ "step": 24
514
+ },
515
+ {
516
+ "debug/policy_chosen_logits": -0.7241502404212952,
517
+ "debug/policy_chosen_logps": -170.52017211914062,
518
+ "debug/policy_rejected_logits": -0.7602882385253906,
519
+ "debug/policy_rejected_logps": -170.40660095214844,
520
+ "debug/reference_chosen_logps": -169.72085571289062,
521
+ "debug/reference_rejected_logps": -168.38609313964844,
522
+ "epoch": 0.5319148936170213,
523
+ "grad_norm": 6.266560430846777,
524
+ "learning_rate": 1e-06,
525
+ "logits/chosen": -0.7241502404212952,
526
+ "logits/rejected": -0.7602882385253906,
527
+ "logps/chosen": -170.52017211914062,
528
+ "logps/rejected": -170.40660095214844,
529
+ "loss": 0.4915,
530
+ "rewards/accuracies": 0.625,
531
+ "rewards/chosen": -0.007993297651410103,
532
+ "rewards/margins": 0.012211760506033897,
533
+ "rewards/rejected": -0.020205058157444,
534
+ "step": 25
535
+ },
536
+ {
537
+ "debug/policy_chosen_logits": -0.9610673189163208,
538
+ "debug/policy_chosen_logps": -136.63485717773438,
539
+ "debug/policy_rejected_logits": -0.8876134753227234,
540
+ "debug/policy_rejected_logps": -161.75608825683594,
541
+ "debug/reference_chosen_logps": -137.76214599609375,
542
+ "debug/reference_rejected_logps": -160.455078125,
543
+ "epoch": 0.5531914893617021,
544
+ "grad_norm": 6.516253702222995,
545
+ "learning_rate": 1e-06,
546
+ "logits/chosen": -0.9610673189163208,
547
+ "logits/rejected": -0.8876134753227234,
548
+ "logps/chosen": -136.63485717773438,
549
+ "logps/rejected": -161.75608825683594,
550
+ "loss": 0.4875,
551
+ "rewards/accuracies": 0.5,
552
+ "rewards/chosen": 0.011272869072854519,
553
+ "rewards/margins": 0.024282770231366158,
554
+ "rewards/rejected": -0.013009900227189064,
555
+ "step": 26
556
+ },
557
+ {
558
+ "debug/policy_chosen_logits": -0.7550271153450012,
559
+ "debug/policy_chosen_logps": -171.28823852539062,
560
+ "debug/policy_rejected_logits": -1.0050970315933228,
561
+ "debug/policy_rejected_logps": -144.0048370361328,
562
+ "debug/reference_chosen_logps": -171.15249633789062,
563
+ "debug/reference_rejected_logps": -143.9324951171875,
564
+ "epoch": 0.574468085106383,
565
+ "grad_norm": 6.752594434046414,
566
+ "learning_rate": 1e-06,
567
+ "logits/chosen": -0.7550271153450012,
568
+ "logits/rejected": -1.0050970315933228,
569
+ "logps/chosen": -171.28823852539062,
570
+ "logps/rejected": -144.0048370361328,
571
+ "loss": 0.4827,
572
+ "rewards/accuracies": 0.375,
573
+ "rewards/chosen": -0.0013572599273175001,
574
+ "rewards/margins": -0.000633773859590292,
575
+ "rewards/rejected": -0.0007234860677272081,
576
+ "step": 27
577
+ },
578
+ {
579
+ "debug/policy_chosen_logits": -0.6094688773155212,
580
+ "debug/policy_chosen_logps": -182.88189697265625,
581
+ "debug/policy_rejected_logits": -0.547568142414093,
582
+ "debug/policy_rejected_logps": -174.17201232910156,
583
+ "debug/reference_chosen_logps": -184.39437866210938,
584
+ "debug/reference_rejected_logps": -174.24937438964844,
585
+ "epoch": 0.5957446808510638,
586
+ "grad_norm": 6.7329002768958315,
587
+ "learning_rate": 1e-06,
588
+ "logits/chosen": -0.6094688773155212,
589
+ "logits/rejected": -0.547568142414093,
590
+ "logps/chosen": -182.88189697265625,
591
+ "logps/rejected": -174.17201232910156,
592
+ "loss": 0.4843,
593
+ "rewards/accuracies": 0.75,
594
+ "rewards/chosen": 0.015124987810850143,
595
+ "rewards/margins": 0.014351272024214268,
596
+ "rewards/rejected": 0.000773716252297163,
597
+ "step": 28
598
+ },
599
+ {
600
+ "debug/policy_chosen_logits": -1.0009883642196655,
601
+ "debug/policy_chosen_logps": -141.870849609375,
602
+ "debug/policy_rejected_logits": -0.822296679019928,
603
+ "debug/policy_rejected_logps": -160.79566955566406,
604
+ "debug/reference_chosen_logps": -143.44473266601562,
605
+ "debug/reference_rejected_logps": -161.58270263671875,
606
+ "epoch": 0.6170212765957447,
607
+ "grad_norm": 6.822157877935097,
608
+ "learning_rate": 1e-06,
609
+ "logits/chosen": -1.0009883642196655,
610
+ "logits/rejected": -0.822296679019928,
611
+ "logps/chosen": -141.870849609375,
612
+ "logps/rejected": -160.79566955566406,
613
+ "loss": 0.4965,
614
+ "rewards/accuracies": 0.625,
615
+ "rewards/chosen": 0.015738725662231445,
616
+ "rewards/margins": 0.007868318818509579,
617
+ "rewards/rejected": 0.007870407775044441,
618
+ "step": 29
619
+ },
620
+ {
621
+ "debug/policy_chosen_logits": -0.7499756813049316,
622
+ "debug/policy_chosen_logps": -175.52218627929688,
623
+ "debug/policy_rejected_logits": -0.7294086217880249,
624
+ "debug/policy_rejected_logps": -176.3539276123047,
625
+ "debug/reference_chosen_logps": -176.32379150390625,
626
+ "debug/reference_rejected_logps": -177.22299194335938,
627
+ "epoch": 0.6382978723404256,
628
+ "grad_norm": 6.748616605300787,
629
+ "learning_rate": 1e-06,
630
+ "logits/chosen": -0.7499756813049316,
631
+ "logits/rejected": -0.7294086217880249,
632
+ "logps/chosen": -175.52218627929688,
633
+ "logps/rejected": -176.3539276123047,
634
+ "loss": 0.4931,
635
+ "rewards/accuracies": 0.375,
636
+ "rewards/chosen": 0.008015966042876244,
637
+ "rewards/margins": -0.0006747061852365732,
638
+ "rewards/rejected": 0.008690671995282173,
639
+ "step": 30
640
+ },
641
+ {
642
+ "debug/policy_chosen_logits": -0.7985236644744873,
643
+ "debug/policy_chosen_logps": -162.06536865234375,
644
+ "debug/policy_rejected_logits": -0.7594453692436218,
645
+ "debug/policy_rejected_logps": -160.13999938964844,
646
+ "debug/reference_chosen_logps": -163.79864501953125,
647
+ "debug/reference_rejected_logps": -160.25460815429688,
648
+ "epoch": 0.6595744680851063,
649
+ "grad_norm": 6.49623881202892,
650
+ "learning_rate": 1e-06,
651
+ "logits/chosen": -0.7985236644744873,
652
+ "logits/rejected": -0.7594453692436218,
653
+ "logps/chosen": -162.06536865234375,
654
+ "logps/rejected": -160.13999938964844,
655
+ "loss": 0.4906,
656
+ "rewards/accuracies": 0.875,
657
+ "rewards/chosen": 0.01733267679810524,
658
+ "rewards/margins": 0.016186676919460297,
659
+ "rewards/rejected": 0.0011460019741207361,
660
+ "step": 31
661
+ },
662
+ {
663
+ "debug/policy_chosen_logits": -0.8090217709541321,
664
+ "debug/policy_chosen_logps": -158.298828125,
665
+ "debug/policy_rejected_logits": -0.8534717559814453,
666
+ "debug/policy_rejected_logps": -141.4454345703125,
667
+ "debug/reference_chosen_logps": -158.9809112548828,
668
+ "debug/reference_rejected_logps": -142.3428955078125,
669
+ "epoch": 0.6808510638297872,
670
+ "grad_norm": 5.950784192712806,
671
+ "learning_rate": 1e-06,
672
+ "logits/chosen": -0.8090217709541321,
673
+ "logits/rejected": -0.8534717559814453,
674
+ "logps/chosen": -158.298828125,
675
+ "logps/rejected": -141.4454345703125,
676
+ "loss": 0.4845,
677
+ "rewards/accuracies": 0.375,
678
+ "rewards/chosen": 0.006820859387516975,
679
+ "rewards/margins": -0.002153654582798481,
680
+ "rewards/rejected": 0.008974513970315456,
681
+ "step": 32
682
+ },
683
+ {
684
+ "debug/policy_chosen_logits": -0.8266898989677429,
685
+ "debug/policy_chosen_logps": -157.5637664794922,
686
+ "debug/policy_rejected_logits": -0.8254861831665039,
687
+ "debug/policy_rejected_logps": -165.0647735595703,
688
+ "debug/reference_chosen_logps": -157.45263671875,
689
+ "debug/reference_rejected_logps": -165.04684448242188,
690
+ "epoch": 0.7021276595744681,
691
+ "grad_norm": 6.730783083748976,
692
+ "learning_rate": 1e-06,
693
+ "logits/chosen": -0.8266898989677429,
694
+ "logits/rejected": -0.8254861831665039,
695
+ "logps/chosen": -157.5637664794922,
696
+ "logps/rejected": -165.0647735595703,
697
+ "loss": 0.4916,
698
+ "rewards/accuracies": 0.5,
699
+ "rewards/chosen": -0.0011113360524177551,
700
+ "rewards/margins": -0.0009320641402155161,
701
+ "rewards/rejected": -0.00017927167937159538,
702
+ "step": 33
703
+ },
704
+ {
705
+ "debug/policy_chosen_logits": -0.9408995509147644,
706
+ "debug/policy_chosen_logps": -143.58526611328125,
707
+ "debug/policy_rejected_logits": -0.9038028717041016,
708
+ "debug/policy_rejected_logps": -161.87374877929688,
709
+ "debug/reference_chosen_logps": -144.58607482910156,
710
+ "debug/reference_rejected_logps": -163.10000610351562,
711
+ "epoch": 0.723404255319149,
712
+ "grad_norm": 6.594448569198325,
713
+ "learning_rate": 1e-06,
714
+ "logits/chosen": -0.9408995509147644,
715
+ "logits/rejected": -0.9038028717041016,
716
+ "logps/chosen": -143.58526611328125,
717
+ "logps/rejected": -161.87374877929688,
718
+ "loss": 0.4874,
719
+ "rewards/accuracies": 0.375,
720
+ "rewards/chosen": 0.010008077137172222,
721
+ "rewards/margins": -0.002254362218081951,
722
+ "rewards/rejected": 0.012262439355254173,
723
+ "step": 34
724
+ },
725
+ {
726
+ "debug/policy_chosen_logits": -0.6247268319129944,
727
+ "debug/policy_chosen_logps": -191.8319091796875,
728
+ "debug/policy_rejected_logits": -0.602497935295105,
729
+ "debug/policy_rejected_logps": -206.44180297851562,
730
+ "debug/reference_chosen_logps": -192.3876953125,
731
+ "debug/reference_rejected_logps": -204.05772399902344,
732
+ "epoch": 0.7446808510638298,
733
+ "grad_norm": 6.771835020241646,
734
+ "learning_rate": 1e-06,
735
+ "logits/chosen": -0.6247268319129944,
736
+ "logits/rejected": -0.602497935295105,
737
+ "logps/chosen": -191.8319091796875,
738
+ "logps/rejected": -206.44180297851562,
739
+ "loss": 0.4889,
740
+ "rewards/accuracies": 0.75,
741
+ "rewards/chosen": 0.005557993892580271,
742
+ "rewards/margins": 0.029398726299405098,
743
+ "rewards/rejected": -0.023840732872486115,
744
+ "step": 35
745
+ },
746
+ {
747
+ "debug/policy_chosen_logits": -0.7129251956939697,
748
+ "debug/policy_chosen_logps": -177.34715270996094,
749
+ "debug/policy_rejected_logits": -0.6988980770111084,
750
+ "debug/policy_rejected_logps": -186.610595703125,
751
+ "debug/reference_chosen_logps": -177.86798095703125,
752
+ "debug/reference_rejected_logps": -184.3507843017578,
753
+ "epoch": 0.7659574468085106,
754
+ "grad_norm": 7.0724076064881265,
755
+ "learning_rate": 1e-06,
756
+ "logits/chosen": -0.7129251956939697,
757
+ "logits/rejected": -0.6988980770111084,
758
+ "logps/chosen": -177.34715270996094,
759
+ "logps/rejected": -186.610595703125,
760
+ "loss": 0.4848,
761
+ "rewards/accuracies": 1.0,
762
+ "rewards/chosen": 0.0052083395421504974,
763
+ "rewards/margins": 0.027806565165519714,
764
+ "rewards/rejected": -0.022598227486014366,
765
+ "step": 36
766
+ },
767
+ {
768
+ "debug/policy_chosen_logits": -0.7906146049499512,
769
+ "debug/policy_chosen_logps": -158.28594970703125,
770
+ "debug/policy_rejected_logits": -0.7429754137992859,
771
+ "debug/policy_rejected_logps": -170.31695556640625,
772
+ "debug/reference_chosen_logps": -160.74452209472656,
773
+ "debug/reference_rejected_logps": -172.1337890625,
774
+ "epoch": 0.7872340425531915,
775
+ "grad_norm": 6.839441812464104,
776
+ "learning_rate": 1e-06,
777
+ "logits/chosen": -0.7906146049499512,
778
+ "logits/rejected": -0.7429754137992859,
779
+ "logps/chosen": -158.28594970703125,
780
+ "logps/rejected": -170.31695556640625,
781
+ "loss": 0.4821,
782
+ "rewards/accuracies": 0.625,
783
+ "rewards/chosen": 0.024585731327533722,
784
+ "rewards/margins": 0.006417407654225826,
785
+ "rewards/rejected": 0.01816832460463047,
786
+ "step": 37
787
+ },
788
+ {
789
+ "debug/policy_chosen_logits": -0.9739854335784912,
790
+ "debug/policy_chosen_logps": -151.2429656982422,
791
+ "debug/policy_rejected_logits": -0.8670800924301147,
792
+ "debug/policy_rejected_logps": -158.70822143554688,
793
+ "debug/reference_chosen_logps": -151.56866455078125,
794
+ "debug/reference_rejected_logps": -156.908935546875,
795
+ "epoch": 0.8085106382978723,
796
+ "grad_norm": 6.484470121315807,
797
+ "learning_rate": 1e-06,
798
+ "logits/chosen": -0.9739854335784912,
799
+ "logits/rejected": -0.8670800924301147,
800
+ "logps/chosen": -151.2429656982422,
801
+ "logps/rejected": -158.70822143554688,
802
+ "loss": 0.4742,
803
+ "rewards/accuracies": 0.875,
804
+ "rewards/chosen": 0.0032570073381066322,
805
+ "rewards/margins": 0.021250000223517418,
806
+ "rewards/rejected": -0.01799299195408821,
807
+ "step": 38
808
+ },
809
+ {
810
+ "debug/policy_chosen_logits": -0.6604007482528687,
811
+ "debug/policy_chosen_logps": -177.76666259765625,
812
+ "debug/policy_rejected_logits": -0.7032350897789001,
813
+ "debug/policy_rejected_logps": -207.57077026367188,
814
+ "debug/reference_chosen_logps": -180.40255737304688,
815
+ "debug/reference_rejected_logps": -207.9987335205078,
816
+ "epoch": 0.8297872340425532,
817
+ "grad_norm": 6.580429098950178,
818
+ "learning_rate": 1e-06,
819
+ "logits/chosen": -0.6604007482528687,
820
+ "logits/rejected": -0.7032350897789001,
821
+ "logps/chosen": -177.76666259765625,
822
+ "logps/rejected": -207.57077026367188,
823
+ "loss": 0.4922,
824
+ "rewards/accuracies": 0.625,
825
+ "rewards/chosen": 0.026359042152762413,
826
+ "rewards/margins": 0.022079389542341232,
827
+ "rewards/rejected": 0.004279651213437319,
828
+ "step": 39
829
+ },
830
+ {
831
+ "debug/policy_chosen_logits": -0.823306143283844,
832
+ "debug/policy_chosen_logps": -175.26303100585938,
833
+ "debug/policy_rejected_logits": -0.8142601847648621,
834
+ "debug/policy_rejected_logps": -168.3768768310547,
835
+ "debug/reference_chosen_logps": -175.55160522460938,
836
+ "debug/reference_rejected_logps": -165.81240844726562,
837
+ "epoch": 0.851063829787234,
838
+ "grad_norm": 7.047550848745438,
839
+ "learning_rate": 1e-06,
840
+ "logits/chosen": -0.823306143283844,
841
+ "logits/rejected": -0.8142601847648621,
842
+ "logps/chosen": -175.26303100585938,
843
+ "logps/rejected": -168.3768768310547,
844
+ "loss": 0.4956,
845
+ "rewards/accuracies": 0.625,
846
+ "rewards/chosen": 0.002885790541768074,
847
+ "rewards/margins": 0.028530634939670563,
848
+ "rewards/rejected": -0.02564484439790249,
849
+ "step": 40
850
+ },
851
+ {
852
+ "debug/policy_chosen_logits": -0.9761400818824768,
853
+ "debug/policy_chosen_logps": -137.45208740234375,
854
+ "debug/policy_rejected_logits": -0.9314041137695312,
855
+ "debug/policy_rejected_logps": -152.2181396484375,
856
+ "debug/reference_chosen_logps": -137.8333740234375,
857
+ "debug/reference_rejected_logps": -152.38201904296875,
858
+ "epoch": 0.8723404255319149,
859
+ "grad_norm": 6.868410892362044,
860
+ "learning_rate": 1e-06,
861
+ "logits/chosen": -0.9761400818824768,
862
+ "logits/rejected": -0.9314041137695312,
863
+ "logps/chosen": -137.45208740234375,
864
+ "logps/rejected": -152.2181396484375,
865
+ "loss": 0.4882,
866
+ "rewards/accuracies": 0.625,
867
+ "rewards/chosen": 0.003812961047515273,
868
+ "rewards/margins": 0.002174043096601963,
869
+ "rewards/rejected": 0.0016389191150665283,
870
+ "step": 41
871
+ },
872
+ {
873
+ "debug/policy_chosen_logits": -0.9311011433601379,
874
+ "debug/policy_chosen_logps": -165.78811645507812,
875
+ "debug/policy_rejected_logits": -0.8658612370491028,
876
+ "debug/policy_rejected_logps": -168.62680053710938,
877
+ "debug/reference_chosen_logps": -166.7484130859375,
878
+ "debug/reference_rejected_logps": -165.18545532226562,
879
+ "epoch": 0.8936170212765957,
880
+ "grad_norm": 7.338986391808919,
881
+ "learning_rate": 1e-06,
882
+ "logits/chosen": -0.9311011433601379,
883
+ "logits/rejected": -0.8658612370491028,
884
+ "logps/chosen": -165.78811645507812,
885
+ "logps/rejected": -168.62680053710938,
886
+ "loss": 0.4771,
887
+ "rewards/accuracies": 0.75,
888
+ "rewards/chosen": 0.00960303284227848,
889
+ "rewards/margins": 0.04401634261012077,
890
+ "rewards/rejected": -0.034413307905197144,
891
+ "step": 42
892
+ },
893
+ {
894
+ "debug/policy_chosen_logits": -0.9136653542518616,
895
+ "debug/policy_chosen_logps": -161.53817749023438,
896
+ "debug/policy_rejected_logits": -0.7834804654121399,
897
+ "debug/policy_rejected_logps": -149.05377197265625,
898
+ "debug/reference_chosen_logps": -160.81060791015625,
899
+ "debug/reference_rejected_logps": -149.98220825195312,
900
+ "epoch": 0.9148936170212766,
901
+ "grad_norm": 7.374791139812744,
902
+ "learning_rate": 1e-06,
903
+ "logits/chosen": -0.9136653542518616,
904
+ "logits/rejected": -0.7834804654121399,
905
+ "logps/chosen": -161.53817749023438,
906
+ "logps/rejected": -149.05377197265625,
907
+ "loss": 0.4919,
908
+ "rewards/accuracies": 0.25,
909
+ "rewards/chosen": -0.007275791838765144,
910
+ "rewards/margins": -0.016560165211558342,
911
+ "rewards/rejected": 0.009284373372793198,
912
+ "step": 43
913
+ },
914
+ {
915
+ "debug/policy_chosen_logits": -0.8033220171928406,
916
+ "debug/policy_chosen_logps": -155.9744110107422,
917
+ "debug/policy_rejected_logits": -0.7281078696250916,
918
+ "debug/policy_rejected_logps": -163.78639221191406,
919
+ "debug/reference_chosen_logps": -156.5009765625,
920
+ "debug/reference_rejected_logps": -162.9461212158203,
921
+ "epoch": 0.9361702127659575,
922
+ "grad_norm": 6.903617537962546,
923
+ "learning_rate": 1e-06,
924
+ "logits/chosen": -0.8033220171928406,
925
+ "logits/rejected": -0.7281078696250916,
926
+ "logps/chosen": -155.9744110107422,
927
+ "logps/rejected": -163.78639221191406,
928
+ "loss": 0.4772,
929
+ "rewards/accuracies": 0.625,
930
+ "rewards/chosen": 0.005265532061457634,
931
+ "rewards/margins": 0.013668222352862358,
932
+ "rewards/rejected": -0.008402691222727299,
933
+ "step": 44
934
+ },
935
+ {
936
+ "debug/policy_chosen_logits": -0.8795375823974609,
937
+ "debug/policy_chosen_logps": -165.64361572265625,
938
+ "debug/policy_rejected_logits": -0.8522282838821411,
939
+ "debug/policy_rejected_logps": -176.4066162109375,
940
+ "debug/reference_chosen_logps": -169.339111328125,
941
+ "debug/reference_rejected_logps": -176.59585571289062,
942
+ "epoch": 0.9574468085106383,
943
+ "grad_norm": 6.742059230170566,
944
+ "learning_rate": 1e-06,
945
+ "logits/chosen": -0.8795375823974609,
946
+ "logits/rejected": -0.8522282838821411,
947
+ "logps/chosen": -165.64361572265625,
948
+ "logps/rejected": -176.4066162109375,
949
+ "loss": 0.4782,
950
+ "rewards/accuracies": 0.75,
951
+ "rewards/chosen": 0.0369548425078392,
952
+ "rewards/margins": 0.0350622832775116,
953
+ "rewards/rejected": 0.0018925576005131006,
954
+ "step": 45
955
+ },
956
+ {
957
+ "debug/policy_chosen_logits": -0.8723025918006897,
958
+ "debug/policy_chosen_logps": -157.41818237304688,
959
+ "debug/policy_rejected_logits": -0.7500877380371094,
960
+ "debug/policy_rejected_logps": -169.74964904785156,
961
+ "debug/reference_chosen_logps": -157.27471923828125,
962
+ "debug/reference_rejected_logps": -167.62408447265625,
963
+ "epoch": 0.9787234042553191,
964
+ "grad_norm": 9.147397357489162,
965
+ "learning_rate": 1e-06,
966
+ "logits/chosen": -0.8723025918006897,
967
+ "logits/rejected": -0.7500877380371094,
968
+ "logps/chosen": -157.41818237304688,
969
+ "logps/rejected": -169.74964904785156,
970
+ "loss": 0.495,
971
+ "rewards/accuracies": 0.375,
972
+ "rewards/chosen": -0.0014344798400998116,
973
+ "rewards/margins": 0.01982107199728489,
974
+ "rewards/rejected": -0.021255549043416977,
975
+ "step": 46
976
+ },
977
+ {
978
+ "debug/policy_chosen_logits": -0.9559898376464844,
979
+ "debug/policy_chosen_logps": -143.5883026123047,
980
+ "debug/policy_rejected_logits": -1.0301291942596436,
981
+ "debug/policy_rejected_logps": -131.79568481445312,
982
+ "debug/reference_chosen_logps": -141.41384887695312,
983
+ "debug/reference_rejected_logps": -129.98980712890625,
984
+ "epoch": 1.0,
985
+ "grad_norm": 6.477500179393295,
986
+ "learning_rate": 1e-06,
987
+ "logits/chosen": -0.9559898376464844,
988
+ "logits/rejected": -1.0301291942596436,
989
+ "logps/chosen": -143.5883026123047,
990
+ "logps/rejected": -131.79568481445312,
991
+ "loss": 0.4484,
992
+ "rewards/accuracies": 0.625,
993
+ "rewards/chosen": -0.021744633093476295,
994
+ "rewards/margins": -0.003685922361910343,
995
+ "rewards/rejected": -0.018058709800243378,
996
+ "step": 47
997
+ },
998
+ {
999
+ "epoch": 1.0,
1000
+ "step": 47,
1001
+ "total_flos": 0.0,
1002
+ "train_loss": 0.4919274278143619,
1003
+ "train_runtime": 164.7654,
1004
+ "train_samples_per_second": 18.013,
1005
+ "train_steps_per_second": 0.285
1006
+ }
1007
+ ],
1008
+ "logging_steps": 1,
1009
+ "max_steps": 47,
1010
+ "num_input_tokens_seen": 0,
1011
+ "num_train_epochs": 1,
1012
+ "save_steps": 500,
1013
+ "stateful_callbacks": {
1014
+ "TrainerControl": {
1015
+ "args": {
1016
+ "should_epoch_stop": false,
1017
+ "should_evaluate": false,
1018
+ "should_log": false,
1019
+ "should_save": true,
1020
+ "should_training_stop": true
1021
+ },
1022
+ "attributes": {}
1023
+ }
1024
+ },
1025
+ "total_flos": 0.0,
1026
+ "train_batch_size": 8,
1027
+ "trial_name": null,
1028
+ "trial_params": null
1029
+ }