yiran-wang3 commited on
Commit
e029483
1 Parent(s): 91fdad7

End of training

Browse files
README.md ADDED
@@ -0,0 +1,64 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ library_name: transformers
3
+ license: apache-2.0
4
+ base_model: yiran-wang3/qwen2_coder_reflct_adamw_iter5
5
+ tags:
6
+ - alignment-handbook
7
+ - generated_from_trainer
8
+ - trl
9
+ - dpo
10
+ datasets:
11
+ - self-generate/qwcoder2_reflct_sppo_hard_new_cn_mining_oj_iter5-binarized-reflection-scored
12
+ model-index:
13
+ - name: qwen2_coder_reflct_adamw_iter6
14
+ results: []
15
+ ---
16
+
17
+ <!-- This model card has been generated automatically according to the information the Trainer had access to. You
18
+ should probably proofread and complete it, then remove this comment. -->
19
+
20
+ # qwen2_coder_reflct_adamw_iter6
21
+
22
+ This model is a fine-tuned version of [yiran-wang3/qwen2_coder_reflct_adamw_iter5](https://huggingface.co/yiran-wang3/qwen2_coder_reflct_adamw_iter5) on the self-generate/qwcoder2_reflct_sppo_hard_new_cn_mining_oj_iter5-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 0.20.2
all_results.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 1.0,
3
+ "total_flos": 0.0,
4
+ "train_loss": 0.48688735404322225,
5
+ "train_runtime": 123.759,
6
+ "train_samples": 1967,
7
+ "train_samples_per_second": 15.894,
8
+ "train_steps_per_second": 0.25
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": 0.48688735404322225,
5
+ "train_runtime": 123.759,
6
+ "train_samples": 1967,
7
+ "train_samples_per_second": 15.894,
8
+ "train_steps_per_second": 0.25
9
+ }
trainer_state.json ADDED
@@ -0,0 +1,693 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 1.0,
5
+ "eval_steps": 100,
6
+ "global_step": 31,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "debug/policy_chosen_logits": -3.141004800796509,
13
+ "debug/policy_chosen_logps": -179.27029418945312,
14
+ "debug/policy_rejected_logits": -3.201063394546509,
15
+ "debug/policy_rejected_logps": -197.1785125732422,
16
+ "debug/reference_chosen_logps": -179.27029418945312,
17
+ "debug/reference_rejected_logps": -197.1785125732422,
18
+ "epoch": 0.03225806451612903,
19
+ "grad_norm": 9.90184688013874,
20
+ "learning_rate": 1e-06,
21
+ "logits/chosen": -3.141004800796509,
22
+ "logits/rejected": -3.201063394546509,
23
+ "logps/chosen": -179.27029418945312,
24
+ "logps/rejected": -197.1785125732422,
25
+ "loss": 0.5,
26
+ "rewards/accuracies": 0.0,
27
+ "rewards/chosen": 0.0,
28
+ "rewards/margins": 0.0,
29
+ "rewards/rejected": 0.0,
30
+ "step": 1
31
+ },
32
+ {
33
+ "debug/policy_chosen_logits": -3.175147771835327,
34
+ "debug/policy_chosen_logps": -170.20516967773438,
35
+ "debug/policy_rejected_logits": -3.1672799587249756,
36
+ "debug/policy_rejected_logps": -170.16004943847656,
37
+ "debug/reference_chosen_logps": -172.45477294921875,
38
+ "debug/reference_rejected_logps": -170.80857849121094,
39
+ "epoch": 0.06451612903225806,
40
+ "grad_norm": 12.759093734337377,
41
+ "learning_rate": 1e-06,
42
+ "logits/chosen": -3.175147771835327,
43
+ "logits/rejected": -3.1672799587249756,
44
+ "logps/chosen": -170.20516967773438,
45
+ "logps/rejected": -170.16004943847656,
46
+ "loss": 0.492,
47
+ "rewards/accuracies": 0.5,
48
+ "rewards/chosen": 0.02249586209654808,
49
+ "rewards/margins": 0.016010627150535583,
50
+ "rewards/rejected": 0.006485233083367348,
51
+ "step": 2
52
+ },
53
+ {
54
+ "debug/policy_chosen_logits": -3.182211399078369,
55
+ "debug/policy_chosen_logps": -137.9634246826172,
56
+ "debug/policy_rejected_logits": -3.2010762691497803,
57
+ "debug/policy_rejected_logps": -186.76589965820312,
58
+ "debug/reference_chosen_logps": -139.10812377929688,
59
+ "debug/reference_rejected_logps": -189.68460083007812,
60
+ "epoch": 0.0967741935483871,
61
+ "grad_norm": 39.234700405953184,
62
+ "learning_rate": 1e-06,
63
+ "logits/chosen": -3.182211399078369,
64
+ "logits/rejected": -3.2010762691497803,
65
+ "logps/chosen": -137.9634246826172,
66
+ "logps/rejected": -186.76589965820312,
67
+ "loss": 0.5175,
68
+ "rewards/accuracies": 0.375,
69
+ "rewards/chosen": 0.011446855962276459,
70
+ "rewards/margins": -0.017740031704306602,
71
+ "rewards/rejected": 0.02918688766658306,
72
+ "step": 3
73
+ },
74
+ {
75
+ "debug/policy_chosen_logits": -3.3540351390838623,
76
+ "debug/policy_chosen_logps": -164.7408447265625,
77
+ "debug/policy_rejected_logits": -3.372749090194702,
78
+ "debug/policy_rejected_logps": -192.99838256835938,
79
+ "debug/reference_chosen_logps": -167.7454833984375,
80
+ "debug/reference_rejected_logps": -194.29327392578125,
81
+ "epoch": 0.12903225806451613,
82
+ "grad_norm": 14.024887806474224,
83
+ "learning_rate": 1e-06,
84
+ "logits/chosen": -3.3540351390838623,
85
+ "logits/rejected": -3.372749090194702,
86
+ "logps/chosen": -164.7408447265625,
87
+ "logps/rejected": -192.99838256835938,
88
+ "loss": 0.5026,
89
+ "rewards/accuracies": 0.5,
90
+ "rewards/chosen": 0.03004644252359867,
91
+ "rewards/margins": 0.017097529023885727,
92
+ "rewards/rejected": 0.012948913499712944,
93
+ "step": 4
94
+ },
95
+ {
96
+ "debug/policy_chosen_logits": -3.0800650119781494,
97
+ "debug/policy_chosen_logps": -183.02537536621094,
98
+ "debug/policy_rejected_logits": -3.0585241317749023,
99
+ "debug/policy_rejected_logps": -228.36184692382812,
100
+ "debug/reference_chosen_logps": -183.33753967285156,
101
+ "debug/reference_rejected_logps": -228.81768798828125,
102
+ "epoch": 0.16129032258064516,
103
+ "grad_norm": 32.252102984713154,
104
+ "learning_rate": 1e-06,
105
+ "logits/chosen": -3.0800650119781494,
106
+ "logits/rejected": -3.0585241317749023,
107
+ "logps/chosen": -183.02537536621094,
108
+ "logps/rejected": -228.36184692382812,
109
+ "loss": 0.4924,
110
+ "rewards/accuracies": 0.25,
111
+ "rewards/chosen": 0.0031216423958539963,
112
+ "rewards/margins": -0.0014367960393428802,
113
+ "rewards/rejected": 0.004558438900858164,
114
+ "step": 5
115
+ },
116
+ {
117
+ "debug/policy_chosen_logits": -3.224591016769409,
118
+ "debug/policy_chosen_logps": -167.71742248535156,
119
+ "debug/policy_rejected_logits": -3.289069175720215,
120
+ "debug/policy_rejected_logps": -221.51846313476562,
121
+ "debug/reference_chosen_logps": -169.82650756835938,
122
+ "debug/reference_rejected_logps": -219.76731872558594,
123
+ "epoch": 0.1935483870967742,
124
+ "grad_norm": 11.2418551066032,
125
+ "learning_rate": 1e-06,
126
+ "logits/chosen": -3.224591016769409,
127
+ "logits/rejected": -3.289069175720215,
128
+ "logps/chosen": -167.71742248535156,
129
+ "logps/rejected": -221.51846313476562,
130
+ "loss": 0.4992,
131
+ "rewards/accuracies": 0.875,
132
+ "rewards/chosen": 0.021090947091579437,
133
+ "rewards/margins": 0.038602352142333984,
134
+ "rewards/rejected": -0.017511405050754547,
135
+ "step": 6
136
+ },
137
+ {
138
+ "debug/policy_chosen_logits": -3.296588182449341,
139
+ "debug/policy_chosen_logps": -149.22991943359375,
140
+ "debug/policy_rejected_logits": -3.2666971683502197,
141
+ "debug/policy_rejected_logps": -164.7155303955078,
142
+ "debug/reference_chosen_logps": -147.17678833007812,
143
+ "debug/reference_rejected_logps": -163.22897338867188,
144
+ "epoch": 0.22580645161290322,
145
+ "grad_norm": 26.842712029286442,
146
+ "learning_rate": 1e-06,
147
+ "logits/chosen": -3.296588182449341,
148
+ "logits/rejected": -3.2666971683502197,
149
+ "logps/chosen": -149.22991943359375,
150
+ "logps/rejected": -164.7155303955078,
151
+ "loss": 0.4918,
152
+ "rewards/accuracies": 0.875,
153
+ "rewards/chosen": -0.020531119778752327,
154
+ "rewards/margins": -0.0056656356900930405,
155
+ "rewards/rejected": -0.014865485019981861,
156
+ "step": 7
157
+ },
158
+ {
159
+ "debug/policy_chosen_logits": -3.1323249340057373,
160
+ "debug/policy_chosen_logps": -151.85377502441406,
161
+ "debug/policy_rejected_logits": -3.1660234928131104,
162
+ "debug/policy_rejected_logps": -176.71502685546875,
163
+ "debug/reference_chosen_logps": -153.5457000732422,
164
+ "debug/reference_rejected_logps": -177.84255981445312,
165
+ "epoch": 0.25806451612903225,
166
+ "grad_norm": 24.498644294188434,
167
+ "learning_rate": 1e-06,
168
+ "logits/chosen": -3.1323249340057373,
169
+ "logits/rejected": -3.1660234928131104,
170
+ "logps/chosen": -151.85377502441406,
171
+ "logps/rejected": -176.71502685546875,
172
+ "loss": 0.4997,
173
+ "rewards/accuracies": 0.75,
174
+ "rewards/chosen": 0.016919326037168503,
175
+ "rewards/margins": 0.005644083023071289,
176
+ "rewards/rejected": 0.011275244876742363,
177
+ "step": 8
178
+ },
179
+ {
180
+ "debug/policy_chosen_logits": -3.202829599380493,
181
+ "debug/policy_chosen_logps": -156.52719116210938,
182
+ "debug/policy_rejected_logits": -3.2741198539733887,
183
+ "debug/policy_rejected_logps": -178.1981201171875,
184
+ "debug/reference_chosen_logps": -165.70819091796875,
185
+ "debug/reference_rejected_logps": -176.71347045898438,
186
+ "epoch": 0.2903225806451613,
187
+ "grad_norm": 21.38108522550141,
188
+ "learning_rate": 1e-06,
189
+ "logits/chosen": -3.202829599380493,
190
+ "logits/rejected": -3.2741198539733887,
191
+ "logps/chosen": -156.52719116210938,
192
+ "logps/rejected": -178.1981201171875,
193
+ "loss": 0.4911,
194
+ "rewards/accuracies": 0.75,
195
+ "rewards/chosen": 0.0918099582195282,
196
+ "rewards/margins": 0.10665654391050339,
197
+ "rewards/rejected": -0.014846592210233212,
198
+ "step": 9
199
+ },
200
+ {
201
+ "debug/policy_chosen_logits": -3.072221040725708,
202
+ "debug/policy_chosen_logps": -158.25836181640625,
203
+ "debug/policy_rejected_logits": -3.057267427444458,
204
+ "debug/policy_rejected_logps": -192.45242309570312,
205
+ "debug/reference_chosen_logps": -164.49575805664062,
206
+ "debug/reference_rejected_logps": -190.25021362304688,
207
+ "epoch": 0.3225806451612903,
208
+ "grad_norm": 18.969223507012597,
209
+ "learning_rate": 1e-06,
210
+ "logits/chosen": -3.072221040725708,
211
+ "logits/rejected": -3.057267427444458,
212
+ "logps/chosen": -158.25836181640625,
213
+ "logps/rejected": -192.45242309570312,
214
+ "loss": 0.5235,
215
+ "rewards/accuracies": 0.875,
216
+ "rewards/chosen": 0.06237373501062393,
217
+ "rewards/margins": 0.08439576625823975,
218
+ "rewards/rejected": -0.022022036835551262,
219
+ "step": 10
220
+ },
221
+ {
222
+ "debug/policy_chosen_logits": -3.092670440673828,
223
+ "debug/policy_chosen_logps": -161.42913818359375,
224
+ "debug/policy_rejected_logits": -3.159600257873535,
225
+ "debug/policy_rejected_logps": -218.52291870117188,
226
+ "debug/reference_chosen_logps": -160.97142028808594,
227
+ "debug/reference_rejected_logps": -214.56564331054688,
228
+ "epoch": 0.3548387096774194,
229
+ "grad_norm": 13.28904793895601,
230
+ "learning_rate": 1e-06,
231
+ "logits/chosen": -3.092670440673828,
232
+ "logits/rejected": -3.159600257873535,
233
+ "logps/chosen": -161.42913818359375,
234
+ "logps/rejected": -218.52291870117188,
235
+ "loss": 0.4986,
236
+ "rewards/accuracies": 0.5,
237
+ "rewards/chosen": -0.004577284678816795,
238
+ "rewards/margins": 0.03499564155936241,
239
+ "rewards/rejected": -0.03957292437553406,
240
+ "step": 11
241
+ },
242
+ {
243
+ "debug/policy_chosen_logits": -3.211040735244751,
244
+ "debug/policy_chosen_logps": -191.30038452148438,
245
+ "debug/policy_rejected_logits": -3.1906237602233887,
246
+ "debug/policy_rejected_logps": -196.66650390625,
247
+ "debug/reference_chosen_logps": -189.11093139648438,
248
+ "debug/reference_rejected_logps": -193.2270965576172,
249
+ "epoch": 0.3870967741935484,
250
+ "grad_norm": 24.159637455477917,
251
+ "learning_rate": 1e-06,
252
+ "logits/chosen": -3.211040735244751,
253
+ "logits/rejected": -3.1906237602233887,
254
+ "logps/chosen": -191.30038452148438,
255
+ "logps/rejected": -196.66650390625,
256
+ "loss": 0.5118,
257
+ "rewards/accuracies": 0.5,
258
+ "rewards/chosen": -0.02189437672495842,
259
+ "rewards/margins": 0.012499636970460415,
260
+ "rewards/rejected": -0.03439401462674141,
261
+ "step": 12
262
+ },
263
+ {
264
+ "debug/policy_chosen_logits": -3.3043720722198486,
265
+ "debug/policy_chosen_logps": -161.9444580078125,
266
+ "debug/policy_rejected_logits": -3.3002195358276367,
267
+ "debug/policy_rejected_logps": -193.20816040039062,
268
+ "debug/reference_chosen_logps": -157.93389892578125,
269
+ "debug/reference_rejected_logps": -187.40017700195312,
270
+ "epoch": 0.41935483870967744,
271
+ "grad_norm": 18.263800856216903,
272
+ "learning_rate": 1e-06,
273
+ "logits/chosen": -3.3043720722198486,
274
+ "logits/rejected": -3.3002195358276367,
275
+ "logps/chosen": -161.9444580078125,
276
+ "logps/rejected": -193.20816040039062,
277
+ "loss": 0.4985,
278
+ "rewards/accuracies": 0.625,
279
+ "rewards/chosen": -0.0401054248213768,
280
+ "rewards/margins": 0.017974376678466797,
281
+ "rewards/rejected": -0.058079805225133896,
282
+ "step": 13
283
+ },
284
+ {
285
+ "debug/policy_chosen_logits": -3.25443696975708,
286
+ "debug/policy_chosen_logps": -192.36001586914062,
287
+ "debug/policy_rejected_logits": -3.1161751747131348,
288
+ "debug/policy_rejected_logps": -188.52992248535156,
289
+ "debug/reference_chosen_logps": -186.53701782226562,
290
+ "debug/reference_rejected_logps": -183.0063934326172,
291
+ "epoch": 0.45161290322580644,
292
+ "grad_norm": 17.08256681847481,
293
+ "learning_rate": 1e-06,
294
+ "logits/chosen": -3.25443696975708,
295
+ "logits/rejected": -3.1161751747131348,
296
+ "logps/chosen": -192.36001586914062,
297
+ "logps/rejected": -188.52992248535156,
298
+ "loss": 0.4921,
299
+ "rewards/accuracies": 0.25,
300
+ "rewards/chosen": -0.05822989344596863,
301
+ "rewards/margins": -0.002994718961417675,
302
+ "rewards/rejected": -0.05523517355322838,
303
+ "step": 14
304
+ },
305
+ {
306
+ "debug/policy_chosen_logits": -3.351620674133301,
307
+ "debug/policy_chosen_logps": -161.11260986328125,
308
+ "debug/policy_rejected_logits": -3.3511033058166504,
309
+ "debug/policy_rejected_logps": -182.30551147460938,
310
+ "debug/reference_chosen_logps": -156.1150665283203,
311
+ "debug/reference_rejected_logps": -177.539794921875,
312
+ "epoch": 0.4838709677419355,
313
+ "grad_norm": 8.229874549490967,
314
+ "learning_rate": 1e-06,
315
+ "logits/chosen": -3.351620674133301,
316
+ "logits/rejected": -3.3511033058166504,
317
+ "logps/chosen": -161.11260986328125,
318
+ "logps/rejected": -182.30551147460938,
319
+ "loss": 0.4926,
320
+ "rewards/accuracies": 0.5,
321
+ "rewards/chosen": -0.04997550696134567,
322
+ "rewards/margins": -0.00231829471886158,
323
+ "rewards/rejected": -0.04765721410512924,
324
+ "step": 15
325
+ },
326
+ {
327
+ "debug/policy_chosen_logits": -3.0169596672058105,
328
+ "debug/policy_chosen_logps": -177.72250366210938,
329
+ "debug/policy_rejected_logits": -3.0592868328094482,
330
+ "debug/policy_rejected_logps": -202.39126586914062,
331
+ "debug/reference_chosen_logps": -175.81646728515625,
332
+ "debug/reference_rejected_logps": -196.47059631347656,
333
+ "epoch": 0.5161290322580645,
334
+ "grad_norm": 18.547449632791665,
335
+ "learning_rate": 1e-06,
336
+ "logits/chosen": -3.0169596672058105,
337
+ "logits/rejected": -3.0592868328094482,
338
+ "logps/chosen": -177.72250366210938,
339
+ "logps/rejected": -202.39126586914062,
340
+ "loss": 0.4991,
341
+ "rewards/accuracies": 0.875,
342
+ "rewards/chosen": -0.01906036213040352,
343
+ "rewards/margins": 0.04014638811349869,
344
+ "rewards/rejected": -0.059206753969192505,
345
+ "step": 16
346
+ },
347
+ {
348
+ "debug/policy_chosen_logits": -3.0885531902313232,
349
+ "debug/policy_chosen_logps": -181.26922607421875,
350
+ "debug/policy_rejected_logits": -3.1282451152801514,
351
+ "debug/policy_rejected_logps": -192.98558044433594,
352
+ "debug/reference_chosen_logps": -179.6282958984375,
353
+ "debug/reference_rejected_logps": -190.09738159179688,
354
+ "epoch": 0.5483870967741935,
355
+ "grad_norm": 16.396288041866185,
356
+ "learning_rate": 1e-06,
357
+ "logits/chosen": -3.0885531902313232,
358
+ "logits/rejected": -3.1282451152801514,
359
+ "logps/chosen": -181.26922607421875,
360
+ "logps/rejected": -192.98558044433594,
361
+ "loss": 0.4811,
362
+ "rewards/accuracies": 0.375,
363
+ "rewards/chosen": -0.0164091307669878,
364
+ "rewards/margins": 0.01247288566082716,
365
+ "rewards/rejected": -0.028882015496492386,
366
+ "step": 17
367
+ },
368
+ {
369
+ "debug/policy_chosen_logits": -3.1581532955169678,
370
+ "debug/policy_chosen_logps": -159.01693725585938,
371
+ "debug/policy_rejected_logits": -3.2684059143066406,
372
+ "debug/policy_rejected_logps": -179.94430541992188,
373
+ "debug/reference_chosen_logps": -159.53839111328125,
374
+ "debug/reference_rejected_logps": -175.76220703125,
375
+ "epoch": 0.5806451612903226,
376
+ "grad_norm": 15.55451033065194,
377
+ "learning_rate": 1e-06,
378
+ "logits/chosen": -3.1581532955169678,
379
+ "logits/rejected": -3.2684059143066406,
380
+ "logps/chosen": -159.01693725585938,
381
+ "logps/rejected": -179.94430541992188,
382
+ "loss": 0.4987,
383
+ "rewards/accuracies": 1.0,
384
+ "rewards/chosen": 0.005214462522417307,
385
+ "rewards/margins": 0.04703548550605774,
386
+ "rewards/rejected": -0.04182102158665657,
387
+ "step": 18
388
+ },
389
+ {
390
+ "debug/policy_chosen_logits": -3.1134331226348877,
391
+ "debug/policy_chosen_logps": -139.32861328125,
392
+ "debug/policy_rejected_logits": -3.094226837158203,
393
+ "debug/policy_rejected_logps": -206.40151977539062,
394
+ "debug/reference_chosen_logps": -141.53060913085938,
395
+ "debug/reference_rejected_logps": -203.99859619140625,
396
+ "epoch": 0.6129032258064516,
397
+ "grad_norm": 11.669033627292357,
398
+ "learning_rate": 1e-06,
399
+ "logits/chosen": -3.1134331226348877,
400
+ "logits/rejected": -3.094226837158203,
401
+ "logps/chosen": -139.32861328125,
402
+ "logps/rejected": -206.40151977539062,
403
+ "loss": 0.4711,
404
+ "rewards/accuracies": 0.625,
405
+ "rewards/chosen": 0.02201991155743599,
406
+ "rewards/margins": 0.046049244701862335,
407
+ "rewards/rejected": -0.024029331281781197,
408
+ "step": 19
409
+ },
410
+ {
411
+ "debug/policy_chosen_logits": -3.1212384700775146,
412
+ "debug/policy_chosen_logps": -191.66444396972656,
413
+ "debug/policy_rejected_logits": -3.260685920715332,
414
+ "debug/policy_rejected_logps": -187.78424072265625,
415
+ "debug/reference_chosen_logps": -191.69525146484375,
416
+ "debug/reference_rejected_logps": -184.98468017578125,
417
+ "epoch": 0.6451612903225806,
418
+ "grad_norm": 25.191796111527317,
419
+ "learning_rate": 1e-06,
420
+ "logits/chosen": -3.1212384700775146,
421
+ "logits/rejected": -3.260685920715332,
422
+ "logps/chosen": -191.66444396972656,
423
+ "logps/rejected": -187.78424072265625,
424
+ "loss": 0.4729,
425
+ "rewards/accuracies": 0.875,
426
+ "rewards/chosen": 0.000308074988424778,
427
+ "rewards/margins": 0.028303585946559906,
428
+ "rewards/rejected": -0.027995511889457703,
429
+ "step": 20
430
+ },
431
+ {
432
+ "debug/policy_chosen_logits": -3.119765520095825,
433
+ "debug/policy_chosen_logps": -155.2023162841797,
434
+ "debug/policy_rejected_logits": -3.102405071258545,
435
+ "debug/policy_rejected_logps": -227.47677612304688,
436
+ "debug/reference_chosen_logps": -159.71484375,
437
+ "debug/reference_rejected_logps": -224.30484008789062,
438
+ "epoch": 0.6774193548387096,
439
+ "grad_norm": 29.084969849970683,
440
+ "learning_rate": 1e-06,
441
+ "logits/chosen": -3.119765520095825,
442
+ "logits/rejected": -3.102405071258545,
443
+ "logps/chosen": -155.2023162841797,
444
+ "logps/rejected": -227.47677612304688,
445
+ "loss": 0.4949,
446
+ "rewards/accuracies": 0.875,
447
+ "rewards/chosen": 0.04512513801455498,
448
+ "rewards/margins": 0.07684443891048431,
449
+ "rewards/rejected": -0.031719304621219635,
450
+ "step": 21
451
+ },
452
+ {
453
+ "debug/policy_chosen_logits": -3.1689579486846924,
454
+ "debug/policy_chosen_logps": -183.11013793945312,
455
+ "debug/policy_rejected_logits": -3.2236738204956055,
456
+ "debug/policy_rejected_logps": -208.21568298339844,
457
+ "debug/reference_chosen_logps": -185.74827575683594,
458
+ "debug/reference_rejected_logps": -208.32473754882812,
459
+ "epoch": 0.7096774193548387,
460
+ "grad_norm": 30.856887681162963,
461
+ "learning_rate": 1e-06,
462
+ "logits/chosen": -3.1689579486846924,
463
+ "logits/rejected": -3.2236738204956055,
464
+ "logps/chosen": -183.11013793945312,
465
+ "logps/rejected": -208.21568298339844,
466
+ "loss": 0.4693,
467
+ "rewards/accuracies": 0.75,
468
+ "rewards/chosen": 0.026381436735391617,
469
+ "rewards/margins": 0.025290966033935547,
470
+ "rewards/rejected": 0.00109047070145607,
471
+ "step": 22
472
+ },
473
+ {
474
+ "debug/policy_chosen_logits": -3.2187812328338623,
475
+ "debug/policy_chosen_logps": -149.282470703125,
476
+ "debug/policy_rejected_logits": -3.192440986633301,
477
+ "debug/policy_rejected_logps": -173.46737670898438,
478
+ "debug/reference_chosen_logps": -152.2470245361328,
479
+ "debug/reference_rejected_logps": -174.3741455078125,
480
+ "epoch": 0.7419354838709677,
481
+ "grad_norm": 7.948773470240676,
482
+ "learning_rate": 1e-06,
483
+ "logits/chosen": -3.2187812328338623,
484
+ "logits/rejected": -3.192440986633301,
485
+ "logps/chosen": -149.282470703125,
486
+ "logps/rejected": -173.46737670898438,
487
+ "loss": 0.4702,
488
+ "rewards/accuracies": 0.5,
489
+ "rewards/chosen": 0.02964554727077484,
490
+ "rewards/margins": 0.02057783305644989,
491
+ "rewards/rejected": 0.0090677160769701,
492
+ "step": 23
493
+ },
494
+ {
495
+ "debug/policy_chosen_logits": -3.232670783996582,
496
+ "debug/policy_chosen_logps": -183.32423400878906,
497
+ "debug/policy_rejected_logits": -3.289520025253296,
498
+ "debug/policy_rejected_logps": -165.54977416992188,
499
+ "debug/reference_chosen_logps": -180.52044677734375,
500
+ "debug/reference_rejected_logps": -166.45697021484375,
501
+ "epoch": 0.7741935483870968,
502
+ "grad_norm": 19.88569884373009,
503
+ "learning_rate": 1e-06,
504
+ "logits/chosen": -3.232670783996582,
505
+ "logits/rejected": -3.289520025253296,
506
+ "logps/chosen": -183.32423400878906,
507
+ "logps/rejected": -165.54977416992188,
508
+ "loss": 0.4722,
509
+ "rewards/accuracies": 0.375,
510
+ "rewards/chosen": -0.028037691488862038,
511
+ "rewards/margins": -0.037109650671482086,
512
+ "rewards/rejected": 0.009071960113942623,
513
+ "step": 24
514
+ },
515
+ {
516
+ "debug/policy_chosen_logits": -3.147738456726074,
517
+ "debug/policy_chosen_logps": -177.48394775390625,
518
+ "debug/policy_rejected_logits": -3.245333433151245,
519
+ "debug/policy_rejected_logps": -199.45582580566406,
520
+ "debug/reference_chosen_logps": -175.49041748046875,
521
+ "debug/reference_rejected_logps": -194.36691284179688,
522
+ "epoch": 0.8064516129032258,
523
+ "grad_norm": 28.982749008082763,
524
+ "learning_rate": 1e-06,
525
+ "logits/chosen": -3.147738456726074,
526
+ "logits/rejected": -3.245333433151245,
527
+ "logps/chosen": -177.48394775390625,
528
+ "logps/rejected": -199.45582580566406,
529
+ "loss": 0.4766,
530
+ "rewards/accuracies": 0.625,
531
+ "rewards/chosen": -0.01993529126048088,
532
+ "rewards/margins": 0.030953893437981606,
533
+ "rewards/rejected": -0.05088917911052704,
534
+ "step": 25
535
+ },
536
+ {
537
+ "debug/policy_chosen_logits": -3.2787725925445557,
538
+ "debug/policy_chosen_logps": -133.00405883789062,
539
+ "debug/policy_rejected_logits": -3.2454986572265625,
540
+ "debug/policy_rejected_logps": -200.76132202148438,
541
+ "debug/reference_chosen_logps": -135.29257202148438,
542
+ "debug/reference_rejected_logps": -194.49354553222656,
543
+ "epoch": 0.8387096774193549,
544
+ "grad_norm": 12.049647968845019,
545
+ "learning_rate": 1e-06,
546
+ "logits/chosen": -3.2787725925445557,
547
+ "logits/rejected": -3.2454986572265625,
548
+ "logps/chosen": -133.00405883789062,
549
+ "logps/rejected": -200.76132202148438,
550
+ "loss": 0.4624,
551
+ "rewards/accuracies": 0.75,
552
+ "rewards/chosen": 0.022885091602802277,
553
+ "rewards/margins": 0.08556278049945831,
554
+ "rewards/rejected": -0.06267768144607544,
555
+ "step": 26
556
+ },
557
+ {
558
+ "debug/policy_chosen_logits": -3.0659728050231934,
559
+ "debug/policy_chosen_logps": -146.81527709960938,
560
+ "debug/policy_rejected_logits": -3.2430717945098877,
561
+ "debug/policy_rejected_logps": -197.0322265625,
562
+ "debug/reference_chosen_logps": -154.13369750976562,
563
+ "debug/reference_rejected_logps": -196.10546875,
564
+ "epoch": 0.8709677419354839,
565
+ "grad_norm": 32.63477734447832,
566
+ "learning_rate": 1e-06,
567
+ "logits/chosen": -3.0659728050231934,
568
+ "logits/rejected": -3.2430717945098877,
569
+ "logps/chosen": -146.81527709960938,
570
+ "logps/rejected": -197.0322265625,
571
+ "loss": 0.4691,
572
+ "rewards/accuracies": 0.75,
573
+ "rewards/chosen": 0.07318423688411713,
574
+ "rewards/margins": 0.08245191723108292,
575
+ "rewards/rejected": -0.009267672896385193,
576
+ "step": 27
577
+ },
578
+ {
579
+ "debug/policy_chosen_logits": -3.2434401512145996,
580
+ "debug/policy_chosen_logps": -149.3644256591797,
581
+ "debug/policy_rejected_logits": -3.216662883758545,
582
+ "debug/policy_rejected_logps": -192.85025024414062,
583
+ "debug/reference_chosen_logps": -149.99478149414062,
584
+ "debug/reference_rejected_logps": -189.23365783691406,
585
+ "epoch": 0.9032258064516129,
586
+ "grad_norm": 10.799436750148159,
587
+ "learning_rate": 1e-06,
588
+ "logits/chosen": -3.2434401512145996,
589
+ "logits/rejected": -3.216662883758545,
590
+ "logps/chosen": -149.3644256591797,
591
+ "logps/rejected": -192.85025024414062,
592
+ "loss": 0.47,
593
+ "rewards/accuracies": 0.875,
594
+ "rewards/chosen": 0.006303424946963787,
595
+ "rewards/margins": 0.04246926307678223,
596
+ "rewards/rejected": -0.036165837198495865,
597
+ "step": 28
598
+ },
599
+ {
600
+ "debug/policy_chosen_logits": -3.089280128479004,
601
+ "debug/policy_chosen_logps": -214.14886474609375,
602
+ "debug/policy_rejected_logits": -3.208700656890869,
603
+ "debug/policy_rejected_logps": -192.32411193847656,
604
+ "debug/reference_chosen_logps": -214.442138671875,
605
+ "debug/reference_rejected_logps": -191.06576538085938,
606
+ "epoch": 0.9354838709677419,
607
+ "grad_norm": 17.0594378565657,
608
+ "learning_rate": 1e-06,
609
+ "logits/chosen": -3.089280128479004,
610
+ "logits/rejected": -3.208700656890869,
611
+ "logps/chosen": -214.14886474609375,
612
+ "logps/rejected": -192.32411193847656,
613
+ "loss": 0.4689,
614
+ "rewards/accuracies": 0.625,
615
+ "rewards/chosen": 0.0029327012598514557,
616
+ "rewards/margins": 0.015516050159931183,
617
+ "rewards/rejected": -0.012583350762724876,
618
+ "step": 29
619
+ },
620
+ {
621
+ "debug/policy_chosen_logits": -3.2566771507263184,
622
+ "debug/policy_chosen_logps": -147.951904296875,
623
+ "debug/policy_rejected_logits": -3.314493417739868,
624
+ "debug/policy_rejected_logps": -165.28897094726562,
625
+ "debug/reference_chosen_logps": -153.832275390625,
626
+ "debug/reference_rejected_logps": -167.7118377685547,
627
+ "epoch": 0.967741935483871,
628
+ "grad_norm": 10.478877864542989,
629
+ "learning_rate": 1e-06,
630
+ "logits/chosen": -3.2566771507263184,
631
+ "logits/rejected": -3.314493417739868,
632
+ "logps/chosen": -147.951904296875,
633
+ "logps/rejected": -165.28897094726562,
634
+ "loss": 0.4552,
635
+ "rewards/accuracies": 0.875,
636
+ "rewards/chosen": 0.05880369991064072,
637
+ "rewards/margins": 0.034574974328279495,
638
+ "rewards/rejected": 0.02422872558236122,
639
+ "step": 30
640
+ },
641
+ {
642
+ "debug/policy_chosen_logits": -3.231640338897705,
643
+ "debug/policy_chosen_logps": -153.23507690429688,
644
+ "debug/policy_rejected_logits": -3.283898115158081,
645
+ "debug/policy_rejected_logps": -196.86080932617188,
646
+ "debug/reference_chosen_logps": -165.0509796142578,
647
+ "debug/reference_rejected_logps": -195.94383239746094,
648
+ "epoch": 1.0,
649
+ "grad_norm": 38.712551042753965,
650
+ "learning_rate": 1e-06,
651
+ "logits/chosen": -3.231640338897705,
652
+ "logits/rejected": -3.283898115158081,
653
+ "logps/chosen": -153.23507690429688,
654
+ "logps/rejected": -196.86080932617188,
655
+ "loss": 0.4584,
656
+ "rewards/accuracies": 1.0,
657
+ "rewards/chosen": 0.11815895140171051,
658
+ "rewards/margins": 0.1273288130760193,
659
+ "rewards/rejected": -0.009169863536953926,
660
+ "step": 31
661
+ },
662
+ {
663
+ "epoch": 1.0,
664
+ "step": 31,
665
+ "total_flos": 0.0,
666
+ "train_loss": 0.48688735404322225,
667
+ "train_runtime": 123.759,
668
+ "train_samples_per_second": 15.894,
669
+ "train_steps_per_second": 0.25
670
+ }
671
+ ],
672
+ "logging_steps": 1,
673
+ "max_steps": 31,
674
+ "num_input_tokens_seen": 0,
675
+ "num_train_epochs": 1,
676
+ "save_steps": 500,
677
+ "stateful_callbacks": {
678
+ "TrainerControl": {
679
+ "args": {
680
+ "should_epoch_stop": false,
681
+ "should_evaluate": false,
682
+ "should_log": false,
683
+ "should_save": true,
684
+ "should_training_stop": true
685
+ },
686
+ "attributes": {}
687
+ }
688
+ },
689
+ "total_flos": 0.0,
690
+ "train_batch_size": 8,
691
+ "trial_name": null,
692
+ "trial_params": null
693
+ }