silviasapora commited on
Commit
47fdd84
·
verified ·
1 Parent(s): ae08532

Model save

Browse files
Files changed (4) hide show
  1. README.md +67 -0
  2. all_results.json +9 -0
  3. train_results.json +9 -0
  4. trainer_state.json +1113 -0
README.md ADDED
@@ -0,0 +1,67 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ base_model: google/gemma-7b
3
+ library_name: transformers
4
+ model_name: gemma-7b-silvia-basic-5e-5-05-vshp27
5
+ tags:
6
+ - generated_from_trainer
7
+ - trl
8
+ - orpo
9
+ licence: license
10
+ ---
11
+
12
+ # Model Card for gemma-7b-silvia-basic-5e-5-05-vshp27
13
+
14
+ This model is a fine-tuned version of [google/gemma-7b](https://huggingface.co/google/gemma-7b).
15
+ It has been trained using [TRL](https://github.com/huggingface/trl).
16
+
17
+ ## Quick start
18
+
19
+ ```python
20
+ from transformers import pipeline
21
+
22
+ question = "If you had a time machine, but could only go to the past or the future once and never return, which would you choose and why?"
23
+ generator = pipeline("text-generation", model="silviasapora/gemma-7b-silvia-basic-5e-5-05-vshp27", device="cuda")
24
+ output = generator([{"role": "user", "content": question}], max_new_tokens=128, return_full_text=False)[0]
25
+ print(output["generated_text"])
26
+ ```
27
+
28
+ ## Training procedure
29
+
30
+ [<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="150" height="24"/>](https://wandb.ai/silvias/huggingface/runs/lwgqa4x5)
31
+
32
+
33
+ This model was trained with ORPO, a method introduced in [ORPO: Monolithic Preference Optimization without Reference Model](https://huggingface.co/papers/2403.07691).
34
+
35
+ ### Framework versions
36
+
37
+ - TRL: 0.13.0
38
+ - Transformers: 4.48.1
39
+ - Pytorch: 2.5.1
40
+ - Datasets: 3.2.0
41
+ - Tokenizers: 0.21.0
42
+
43
+ ## Citations
44
+
45
+ Cite ORPO as:
46
+
47
+ ```bibtex
48
+ @article{hong2024orpo,
49
+ title = {{ORPO: Monolithic Preference Optimization without Reference Model}},
50
+ author = {Jiwoo Hong and Noah Lee and James Thorne},
51
+ year = 2024,
52
+ eprint = {arXiv:2403.07691}
53
+ }
54
+ ```
55
+
56
+ Cite TRL as:
57
+
58
+ ```bibtex
59
+ @misc{vonwerra2022trl,
60
+ title = {{TRL: Transformer Reinforcement Learning}},
61
+ author = {Leandro von Werra and Younes Belkada and Lewis Tunstall and Edward Beeching and Tristan Thrush and Nathan Lambert and Shengyi Huang and Kashif Rasul and Quentin Gallouédec},
62
+ year = 2020,
63
+ journal = {GitHub repository},
64
+ publisher = {GitHub},
65
+ howpublished = {\url{https://github.com/huggingface/trl}}
66
+ }
67
+ ```
all_results.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 2.9765925925925925,
3
+ "total_flos": 0.0,
4
+ "train_loss": 16.270644454350546,
5
+ "train_runtime": 9547.5166,
6
+ "train_samples": 6750,
7
+ "train_samples_per_second": 2.121,
8
+ "train_steps_per_second": 0.033
9
+ }
train_results.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 2.9765925925925925,
3
+ "total_flos": 0.0,
4
+ "train_loss": 16.270644454350546,
5
+ "train_runtime": 9547.5166,
6
+ "train_samples": 6750,
7
+ "train_samples_per_second": 2.121,
8
+ "train_steps_per_second": 0.033
9
+ }
trainer_state.json ADDED
@@ -0,0 +1,1113 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 2.9765925925925925,
5
+ "eval_steps": 500,
6
+ "global_step": 315,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.047407407407407405,
13
+ "grad_norm": 1064.504150390625,
14
+ "learning_rate": 7.8125e-06,
15
+ "log_odds_chosen": -1.0592396259307861,
16
+ "log_odds_ratio": -11.0020170211792,
17
+ "logps/chosen": -22.70431900024414,
18
+ "logps/rejected": -21.645305633544922,
19
+ "loss": 450.5827,
20
+ "nll_loss": 9.83934497833252,
21
+ "po_loss": 4.241364479064941,
22
+ "rewards/accuracies": 0.4781250059604645,
23
+ "rewards/chosen": -11.35215950012207,
24
+ "rewards/margins": -0.5295066237449646,
25
+ "rewards/rejected": -10.822652816772461,
26
+ "step": 5
27
+ },
28
+ {
29
+ "epoch": 0.09481481481481481,
30
+ "grad_norm": 610.734619140625,
31
+ "learning_rate": 1.5625e-05,
32
+ "log_odds_chosen": -3.1254055500030518,
33
+ "log_odds_ratio": -11.053281784057617,
34
+ "logps/chosen": -22.00815200805664,
35
+ "logps/rejected": -18.883081436157227,
36
+ "loss": 422.0994,
37
+ "nll_loss": 8.754120826721191,
38
+ "po_loss": 4.43648624420166,
39
+ "rewards/accuracies": 0.44999998807907104,
40
+ "rewards/chosen": -11.00407600402832,
41
+ "rewards/margins": -1.5625368356704712,
42
+ "rewards/rejected": -9.441540718078613,
43
+ "step": 10
44
+ },
45
+ {
46
+ "epoch": 0.14222222222222222,
47
+ "grad_norm": 600.33056640625,
48
+ "learning_rate": 2.34375e-05,
49
+ "log_odds_chosen": -1.30233633518219,
50
+ "log_odds_ratio": -11.293824195861816,
51
+ "logps/chosen": -21.929264068603516,
52
+ "logps/rejected": -20.62543487548828,
53
+ "loss": 392.0485,
54
+ "nll_loss": 7.978579044342041,
55
+ "po_loss": 4.2729363441467285,
56
+ "rewards/accuracies": 0.515625,
57
+ "rewards/chosen": -10.964632034301758,
58
+ "rewards/margins": -0.6519135236740112,
59
+ "rewards/rejected": -10.31271743774414,
60
+ "step": 15
61
+ },
62
+ {
63
+ "epoch": 0.18962962962962962,
64
+ "grad_norm": 3273.544677734375,
65
+ "learning_rate": 3.125e-05,
66
+ "log_odds_chosen": -3.9110000133514404,
67
+ "log_odds_ratio": -8.825319290161133,
68
+ "logps/chosen": -16.157787322998047,
69
+ "logps/rejected": -12.249563217163086,
70
+ "loss": 296.8894,
71
+ "nll_loss": 5.963077545166016,
72
+ "po_loss": 3.314715623855591,
73
+ "rewards/accuracies": 0.44062501192092896,
74
+ "rewards/chosen": -8.078893661499023,
75
+ "rewards/margins": -1.954110860824585,
76
+ "rewards/rejected": -6.124781608581543,
77
+ "step": 20
78
+ },
79
+ {
80
+ "epoch": 0.23703703703703705,
81
+ "grad_norm": 309.38031005859375,
82
+ "learning_rate": 3.90625e-05,
83
+ "log_odds_chosen": -0.052033402025699615,
84
+ "log_odds_ratio": -1.5327208042144775,
85
+ "logps/chosen": -3.302628755569458,
86
+ "logps/rejected": -3.230381488800049,
87
+ "loss": 83.2066,
88
+ "nll_loss": 2.531467914581299,
89
+ "po_loss": 0.06873763352632523,
90
+ "rewards/accuracies": 0.5562499761581421,
91
+ "rewards/chosen": -1.651314377784729,
92
+ "rewards/margins": -0.036123599857091904,
93
+ "rewards/rejected": -1.6151907444000244,
94
+ "step": 25
95
+ },
96
+ {
97
+ "epoch": 0.28444444444444444,
98
+ "grad_norm": 1114.235107421875,
99
+ "learning_rate": 4.6875e-05,
100
+ "log_odds_chosen": 0.07359764724969864,
101
+ "log_odds_ratio": -0.942546546459198,
102
+ "logps/chosen": -1.9597572088241577,
103
+ "logps/rejected": -2.0194180011749268,
104
+ "loss": 64.8739,
105
+ "nll_loss": 2.057140827178955,
106
+ "po_loss": -0.029830310493707657,
107
+ "rewards/accuracies": 0.543749988079071,
108
+ "rewards/chosen": -0.9798786044120789,
109
+ "rewards/margins": 0.029830310493707657,
110
+ "rewards/rejected": -1.0097090005874634,
111
+ "step": 30
112
+ },
113
+ {
114
+ "epoch": 0.33185185185185184,
115
+ "grad_norm": 388.6631774902344,
116
+ "learning_rate": 4.998613757348784e-05,
117
+ "log_odds_chosen": 0.3581487536430359,
118
+ "log_odds_ratio": -0.841637909412384,
119
+ "logps/chosen": -1.9078718423843384,
120
+ "logps/rejected": -2.2348716259002686,
121
+ "loss": 58.6657,
122
+ "nll_loss": 1.9908554553985596,
123
+ "po_loss": -0.15755358338356018,
124
+ "rewards/accuracies": 0.543749988079071,
125
+ "rewards/chosen": -0.9539359211921692,
126
+ "rewards/margins": 0.1634998619556427,
127
+ "rewards/rejected": -1.1174358129501343,
128
+ "step": 35
129
+ },
130
+ {
131
+ "epoch": 0.37925925925925924,
132
+ "grad_norm": 215.6595458984375,
133
+ "learning_rate": 4.990147841143462e-05,
134
+ "log_odds_chosen": 0.49525055289268494,
135
+ "log_odds_ratio": -0.6980496644973755,
136
+ "logps/chosen": -1.7104499340057373,
137
+ "logps/rejected": -2.14109206199646,
138
+ "loss": 55.0966,
139
+ "nll_loss": 1.9370899200439453,
140
+ "po_loss": -0.21532109379768372,
141
+ "rewards/accuracies": 0.65625,
142
+ "rewards/chosen": -0.8552249670028687,
143
+ "rewards/margins": 0.21532109379768372,
144
+ "rewards/rejected": -1.07054603099823,
145
+ "step": 40
146
+ },
147
+ {
148
+ "epoch": 0.4266666666666667,
149
+ "grad_norm": 490.9734191894531,
150
+ "learning_rate": 4.97401218720448e-05,
151
+ "log_odds_chosen": 1.0681084394454956,
152
+ "log_odds_ratio": -0.748261034488678,
153
+ "logps/chosen": -1.9336668252944946,
154
+ "logps/rejected": -2.933429718017578,
155
+ "loss": 45.8169,
156
+ "nll_loss": 1.9130370616912842,
157
+ "po_loss": -0.4812597632408142,
158
+ "rewards/accuracies": 0.671875,
159
+ "rewards/chosen": -0.9668334126472473,
160
+ "rewards/margins": 0.4998815655708313,
161
+ "rewards/rejected": -1.466714859008789,
162
+ "step": 45
163
+ },
164
+ {
165
+ "epoch": 0.4740740740740741,
166
+ "grad_norm": 687.996826171875,
167
+ "learning_rate": 4.9502564938797946e-05,
168
+ "log_odds_chosen": 1.5803465843200684,
169
+ "log_odds_ratio": -0.8952202796936035,
170
+ "logps/chosen": -2.2172279357910156,
171
+ "logps/rejected": -3.716336488723755,
172
+ "loss": 51.598,
173
+ "nll_loss": 2.2828214168548584,
174
+ "po_loss": -0.6703845262527466,
175
+ "rewards/accuracies": 0.6937500238418579,
176
+ "rewards/chosen": -1.1086139678955078,
177
+ "rewards/margins": 0.7495540976524353,
178
+ "rewards/rejected": -1.8581682443618774,
179
+ "step": 50
180
+ },
181
+ {
182
+ "epoch": 0.5214814814814814,
183
+ "grad_norm": 4070.76123046875,
184
+ "learning_rate": 4.918953929490768e-05,
185
+ "log_odds_chosen": 2.9766428470611572,
186
+ "log_odds_ratio": -0.919791042804718,
187
+ "logps/chosen": -2.4137585163116455,
188
+ "logps/rejected": -5.309506416320801,
189
+ "loss": 37.3237,
190
+ "nll_loss": 2.3748691082000732,
191
+ "po_loss": -1.2085039615631104,
192
+ "rewards/accuracies": 0.6656249761581421,
193
+ "rewards/chosen": -1.2068792581558228,
194
+ "rewards/margins": 1.4478741884231567,
195
+ "rewards/rejected": -2.6547532081604004,
196
+ "step": 55
197
+ },
198
+ {
199
+ "epoch": 0.5688888888888889,
200
+ "grad_norm": 569.17724609375,
201
+ "learning_rate": 4.88020090697132e-05,
202
+ "log_odds_chosen": 5.00119686126709,
203
+ "log_odds_ratio": -1.5456047058105469,
204
+ "logps/chosen": -6.171454906463623,
205
+ "logps/rejected": -11.093535423278809,
206
+ "loss": 103.9391,
207
+ "nll_loss": 4.901269912719727,
208
+ "po_loss": -1.653172492980957,
209
+ "rewards/accuracies": 0.721875011920929,
210
+ "rewards/chosen": -3.0857274532318115,
211
+ "rewards/margins": 2.461040496826172,
212
+ "rewards/rejected": -5.546767711639404,
213
+ "step": 60
214
+ },
215
+ {
216
+ "epoch": 0.6162962962962963,
217
+ "grad_norm": 614.133544921875,
218
+ "learning_rate": 4.834116786912897e-05,
219
+ "log_odds_chosen": 1.351292371749878,
220
+ "log_odds_ratio": -1.3231507539749146,
221
+ "logps/chosen": -2.553697109222412,
222
+ "logps/rejected": -3.8612685203552246,
223
+ "loss": 63.6228,
224
+ "nll_loss": 2.556739091873169,
225
+ "po_loss": -0.5685266256332397,
226
+ "rewards/accuracies": 0.59375,
227
+ "rewards/chosen": -1.276848554611206,
228
+ "rewards/margins": 0.6537858843803406,
229
+ "rewards/rejected": -1.9306342601776123,
230
+ "step": 65
231
+ },
232
+ {
233
+ "epoch": 0.6637037037037037,
234
+ "grad_norm": 386.0463562011719,
235
+ "learning_rate": 4.7808435099299045e-05,
236
+ "log_odds_chosen": 2.089937210083008,
237
+ "log_odds_ratio": -0.670052170753479,
238
+ "logps/chosen": -1.7493152618408203,
239
+ "logps/rejected": -3.7693347930908203,
240
+ "loss": 34.9169,
241
+ "nll_loss": 1.8565772771835327,
242
+ "po_loss": -0.7654256820678711,
243
+ "rewards/accuracies": 0.6937500238418579,
244
+ "rewards/chosen": -0.8746576309204102,
245
+ "rewards/margins": 1.010009765625,
246
+ "rewards/rejected": -1.8846673965454102,
247
+ "step": 70
248
+ },
249
+ {
250
+ "epoch": 0.7111111111111111,
251
+ "grad_norm": 3206.72705078125,
252
+ "learning_rate": 4.720545159477922e-05,
253
+ "log_odds_chosen": 2.741664409637451,
254
+ "log_odds_ratio": -1.1608259677886963,
255
+ "logps/chosen": -3.1741726398468018,
256
+ "logps/rejected": -5.838489532470703,
257
+ "loss": 55.9823,
258
+ "nll_loss": 2.62483811378479,
259
+ "po_loss": -0.8753900527954102,
260
+ "rewards/accuracies": 0.6968749761581421,
261
+ "rewards/chosen": -1.5870863199234009,
262
+ "rewards/margins": 1.3321585655212402,
263
+ "rewards/rejected": -2.9192447662353516,
264
+ "step": 75
265
+ },
266
+ {
267
+ "epoch": 0.7585185185185185,
268
+ "grad_norm": 412.6952209472656,
269
+ "learning_rate": 4.653407456471222e-05,
270
+ "log_odds_chosen": 1.9728152751922607,
271
+ "log_odds_ratio": -0.9150087237358093,
272
+ "logps/chosen": -2.0698251724243164,
273
+ "logps/rejected": -3.9561798572540283,
274
+ "loss": 41.4899,
275
+ "nll_loss": 2.0737736225128174,
276
+ "po_loss": -0.7772130370140076,
277
+ "rewards/accuracies": 0.653124988079071,
278
+ "rewards/chosen": -1.0349125862121582,
279
+ "rewards/margins": 0.9431774020195007,
280
+ "rewards/rejected": -1.9780899286270142,
281
+ "step": 80
282
+ },
283
+ {
284
+ "epoch": 0.8059259259259259,
285
+ "grad_norm": 1100.900634765625,
286
+ "learning_rate": 4.579637187256222e-05,
287
+ "log_odds_chosen": 2.417811870574951,
288
+ "log_odds_ratio": -0.9469805955886841,
289
+ "logps/chosen": -2.1366405487060547,
290
+ "logps/rejected": -4.464547157287598,
291
+ "loss": 36.3301,
292
+ "nll_loss": 2.163440704345703,
293
+ "po_loss": -1.0281257629394531,
294
+ "rewards/accuracies": 0.6812499761581421,
295
+ "rewards/chosen": -1.0683202743530273,
296
+ "rewards/margins": 1.163953185081482,
297
+ "rewards/rejected": -2.232273578643799,
298
+ "step": 85
299
+ },
300
+ {
301
+ "epoch": 0.8533333333333334,
302
+ "grad_norm": 785.2890625,
303
+ "learning_rate": 4.499461566702685e-05,
304
+ "log_odds_chosen": 3.7706730365753174,
305
+ "log_odds_ratio": -1.2985801696777344,
306
+ "logps/chosen": -3.598789691925049,
307
+ "logps/rejected": -7.259984016418457,
308
+ "loss": 42.6548,
309
+ "nll_loss": 2.9308180809020996,
310
+ "po_loss": -1.597855567932129,
311
+ "rewards/accuracies": 0.690625011920929,
312
+ "rewards/chosen": -1.7993948459625244,
313
+ "rewards/margins": 1.830596923828125,
314
+ "rewards/rejected": -3.6299920082092285,
315
+ "step": 90
316
+ },
317
+ {
318
+ "epoch": 0.9007407407407407,
319
+ "grad_norm": 3162.0009765625,
320
+ "learning_rate": 4.413127538374411e-05,
321
+ "log_odds_chosen": 2.615981340408325,
322
+ "log_odds_ratio": -1.0526683330535889,
323
+ "logps/chosen": -2.3276286125183105,
324
+ "logps/rejected": -4.851325988769531,
325
+ "loss": 31.8051,
326
+ "nll_loss": 2.035627841949463,
327
+ "po_loss": -1.0417195558547974,
328
+ "rewards/accuracies": 0.6499999761581421,
329
+ "rewards/chosen": -1.1638143062591553,
330
+ "rewards/margins": 1.2618486881256104,
331
+ "rewards/rejected": -2.4256629943847656,
332
+ "step": 95
333
+ },
334
+ {
335
+ "epoch": 0.9481481481481482,
336
+ "grad_norm": 571.5673217773438,
337
+ "learning_rate": 4.320901013934887e-05,
338
+ "log_odds_chosen": 4.09200382232666,
339
+ "log_odds_ratio": -1.3848493099212646,
340
+ "logps/chosen": -3.318943500518799,
341
+ "logps/rejected": -7.310236930847168,
342
+ "loss": 37.0821,
343
+ "nll_loss": 2.7883472442626953,
344
+ "po_loss": -1.629533052444458,
345
+ "rewards/accuracies": 0.6812499761581421,
346
+ "rewards/chosen": -1.6594717502593994,
347
+ "rewards/margins": 1.9956468343734741,
348
+ "rewards/rejected": -3.655118465423584,
349
+ "step": 100
350
+ },
351
+ {
352
+ "epoch": 0.9955555555555555,
353
+ "grad_norm": 251.86697387695312,
354
+ "learning_rate": 4.223066054130568e-05,
355
+ "log_odds_chosen": 3.148193359375,
356
+ "log_odds_ratio": -0.6689270734786987,
357
+ "logps/chosen": -1.6926807165145874,
358
+ "logps/rejected": -4.753729820251465,
359
+ "loss": 24.9079,
360
+ "nll_loss": 1.8303325176239014,
361
+ "po_loss": -1.05196213722229,
362
+ "rewards/accuracies": 0.6781250238418579,
363
+ "rewards/chosen": -0.8463403582572937,
364
+ "rewards/margins": 1.5305248498916626,
365
+ "rewards/rejected": -2.3768649101257324,
366
+ "step": 105
367
+ },
368
+ {
369
+ "epoch": 1.037925925925926,
370
+ "grad_norm": 235.2350616455078,
371
+ "learning_rate": 4.1199239938743797e-05,
372
+ "log_odds_chosen": 3.255824565887451,
373
+ "log_odds_ratio": -0.6982910633087158,
374
+ "logps/chosen": -1.715062141418457,
375
+ "logps/rejected": -4.823176383972168,
376
+ "loss": 23.0323,
377
+ "nll_loss": 1.716023325920105,
378
+ "po_loss": -0.9106981754302979,
379
+ "rewards/accuracies": 0.7307692170143127,
380
+ "rewards/chosen": -0.8575310707092285,
381
+ "rewards/margins": 1.5540571212768555,
382
+ "rewards/rejected": -2.411588191986084,
383
+ "step": 110
384
+ },
385
+ {
386
+ "epoch": 1.0853333333333333,
387
+ "grad_norm": 457.39080810546875,
388
+ "learning_rate": 4.0117925141242174e-05,
389
+ "log_odds_chosen": 5.342764854431152,
390
+ "log_odds_ratio": -0.9483366012573242,
391
+ "logps/chosen": -2.369607448577881,
392
+ "logps/rejected": -7.539316654205322,
393
+ "loss": 17.0777,
394
+ "nll_loss": 2.0790820121765137,
395
+ "po_loss": -1.5454037189483643,
396
+ "rewards/accuracies": 0.7593749761581421,
397
+ "rewards/chosen": -1.1848037242889404,
398
+ "rewards/margins": 2.5848546028137207,
399
+ "rewards/rejected": -3.769658327102661,
400
+ "step": 115
401
+ },
402
+ {
403
+ "epoch": 1.1327407407407408,
404
+ "grad_norm": 478.9156799316406,
405
+ "learning_rate": 3.899004663415084e-05,
406
+ "log_odds_chosen": 4.5327959060668945,
407
+ "log_odds_ratio": -0.7982605695724487,
408
+ "logps/chosen": -2.2072598934173584,
409
+ "logps/rejected": -6.586657524108887,
410
+ "loss": 12.6686,
411
+ "nll_loss": 1.9718843698501587,
412
+ "po_loss": -1.5759899616241455,
413
+ "rewards/accuracies": 0.7406250238418579,
414
+ "rewards/chosen": -1.1036299467086792,
415
+ "rewards/margins": 2.1896984577178955,
416
+ "rewards/rejected": -3.2933287620544434,
417
+ "step": 120
418
+ },
419
+ {
420
+ "epoch": 1.1801481481481482,
421
+ "grad_norm": 760.638916015625,
422
+ "learning_rate": 3.781907832058587e-05,
423
+ "log_odds_chosen": 4.349984645843506,
424
+ "log_odds_ratio": -1.0327547788619995,
425
+ "logps/chosen": -2.662911891937256,
426
+ "logps/rejected": -6.870619297027588,
427
+ "loss": 32.4605,
428
+ "nll_loss": 2.5304787158966064,
429
+ "po_loss": -1.5160895586013794,
430
+ "rewards/accuracies": 0.7406250238418579,
431
+ "rewards/chosen": -1.331455945968628,
432
+ "rewards/margins": 2.103853702545166,
433
+ "rewards/rejected": -3.435309648513794,
434
+ "step": 125
435
+ },
436
+ {
437
+ "epoch": 1.2275555555555555,
438
+ "grad_norm": 1151.731201171875,
439
+ "learning_rate": 3.660862682169282e-05,
440
+ "log_odds_chosen": 6.421099662780762,
441
+ "log_odds_ratio": -1.0163640975952148,
442
+ "logps/chosen": -2.875323534011841,
443
+ "logps/rejected": -9.088385581970215,
444
+ "loss": -4.7801,
445
+ "nll_loss": 2.3773765563964844,
446
+ "po_loss": -2.5267534255981445,
447
+ "rewards/accuracies": 0.762499988079071,
448
+ "rewards/chosen": -1.4376617670059204,
449
+ "rewards/margins": 3.1065304279327393,
450
+ "rewards/rejected": -4.544192790985107,
451
+ "step": 130
452
+ },
453
+ {
454
+ "epoch": 1.274962962962963,
455
+ "grad_norm": 535.7034912109375,
456
+ "learning_rate": 3.5362420368134356e-05,
457
+ "log_odds_chosen": 3.287085771560669,
458
+ "log_odds_ratio": -0.6112745404243469,
459
+ "logps/chosen": -1.6216014623641968,
460
+ "logps/rejected": -4.728818893432617,
461
+ "loss": 9.9669,
462
+ "nll_loss": 1.7199676036834717,
463
+ "po_loss": -1.4085030555725098,
464
+ "rewards/accuracies": 0.7437499761581421,
465
+ "rewards/chosen": -0.8108007311820984,
466
+ "rewards/margins": 1.553608775138855,
467
+ "rewards/rejected": -2.3644094467163086,
468
+ "step": 135
469
+ },
470
+ {
471
+ "epoch": 1.3223703703703704,
472
+ "grad_norm": 1074.8087158203125,
473
+ "learning_rate": 3.408429731701635e-05,
474
+ "log_odds_chosen": 3.7467429637908936,
475
+ "log_odds_ratio": -0.7782642841339111,
476
+ "logps/chosen": -2.253432512283325,
477
+ "logps/rejected": -5.834208011627197,
478
+ "loss": 18.4068,
479
+ "nll_loss": 2.049558401107788,
480
+ "po_loss": -1.4743449687957764,
481
+ "rewards/accuracies": 0.734375,
482
+ "rewards/chosen": -1.1267162561416626,
483
+ "rewards/margins": 1.790387749671936,
484
+ "rewards/rejected": -2.9171040058135986,
485
+ "step": 140
486
+ },
487
+ {
488
+ "epoch": 1.3697777777777778,
489
+ "grad_norm": 2433.367919921875,
490
+ "learning_rate": 3.2778194329621104e-05,
491
+ "log_odds_chosen": 6.9977288246154785,
492
+ "log_odds_ratio": -0.8477832078933716,
493
+ "logps/chosen": -2.8123180866241455,
494
+ "logps/rejected": -9.62822151184082,
495
+ "loss": -7.1341,
496
+ "nll_loss": 2.3402633666992188,
497
+ "po_loss": -2.5632026195526123,
498
+ "rewards/accuracies": 0.7718750238418579,
499
+ "rewards/chosen": -1.4061590433120728,
500
+ "rewards/margins": 3.407951831817627,
501
+ "rewards/rejected": -4.81411075592041,
502
+ "step": 145
503
+ },
504
+ {
505
+ "epoch": 1.417185185185185,
506
+ "grad_norm": 1035.2568359375,
507
+ "learning_rate": 3.144813424636031e-05,
508
+ "log_odds_chosen": 7.531359672546387,
509
+ "log_odds_ratio": -0.9453743696212769,
510
+ "logps/chosen": -2.4776880741119385,
511
+ "logps/rejected": -9.81294059753418,
512
+ "loss": -8.8755,
513
+ "nll_loss": 2.127579689025879,
514
+ "po_loss": -2.404938220977783,
515
+ "rewards/accuracies": 0.762499988079071,
516
+ "rewards/chosen": -1.2388440370559692,
517
+ "rewards/margins": 3.667625904083252,
518
+ "rewards/rejected": -4.90647029876709,
519
+ "step": 150
520
+ },
521
+ {
522
+ "epoch": 1.4645925925925927,
523
+ "grad_norm": 1741.77978515625,
524
+ "learning_rate": 3.0098213696293542e-05,
525
+ "log_odds_chosen": 4.925074577331543,
526
+ "log_odds_ratio": -1.4115755558013916,
527
+ "logps/chosen": -2.6048836708068848,
528
+ "logps/rejected": -7.376564025878906,
529
+ "loss": 23.5026,
530
+ "nll_loss": 2.325968027114868,
531
+ "po_loss": -1.5915124416351318,
532
+ "rewards/accuracies": 0.6968749761581421,
533
+ "rewards/chosen": -1.3024418354034424,
534
+ "rewards/margins": 2.3858399391174316,
535
+ "rewards/rejected": -3.688282012939453,
536
+ "step": 155
537
+ },
538
+ {
539
+ "epoch": 1.512,
540
+ "grad_norm": 1471.9173583984375,
541
+ "learning_rate": 2.8732590479375165e-05,
542
+ "log_odds_chosen": 6.550325870513916,
543
+ "log_odds_ratio": -0.9376071691513062,
544
+ "logps/chosen": -2.9662528038024902,
545
+ "logps/rejected": -9.300813674926758,
546
+ "loss": -3.3757,
547
+ "nll_loss": 2.572293758392334,
548
+ "po_loss": -2.677783250808716,
549
+ "rewards/accuracies": 0.762499988079071,
550
+ "rewards/chosen": -1.4831264019012451,
551
+ "rewards/margins": 3.1672801971435547,
552
+ "rewards/rejected": -4.650406837463379,
553
+ "step": 160
554
+ },
555
+ {
556
+ "epoch": 1.5594074074074074,
557
+ "grad_norm": 562.9378051757812,
558
+ "learning_rate": 2.7355470760292956e-05,
559
+ "log_odds_chosen": 5.505670547485352,
560
+ "log_odds_ratio": -0.8015021085739136,
561
+ "logps/chosen": -2.046816349029541,
562
+ "logps/rejected": -7.327232360839844,
563
+ "loss": -4.0531,
564
+ "nll_loss": 2.0913338661193848,
565
+ "po_loss": -2.217992067337036,
566
+ "rewards/accuracies": 0.768750011920929,
567
+ "rewards/chosen": -1.0234081745147705,
568
+ "rewards/margins": 2.6402080059051514,
569
+ "rewards/rejected": -3.663616180419922,
570
+ "step": 165
571
+ },
572
+ {
573
+ "epoch": 1.6068148148148147,
574
+ "grad_norm": 1370.6552734375,
575
+ "learning_rate": 2.597109611334169e-05,
576
+ "log_odds_chosen": 9.098596572875977,
577
+ "log_odds_ratio": -1.227461576461792,
578
+ "logps/chosen": -3.2510433197021484,
579
+ "logps/rejected": -12.151667594909668,
580
+ "loss": -13.6098,
581
+ "nll_loss": 2.593132495880127,
582
+ "po_loss": -3.018439769744873,
583
+ "rewards/accuracies": 0.7749999761581421,
584
+ "rewards/chosen": -1.6255216598510742,
585
+ "rewards/margins": 4.450311660766602,
586
+ "rewards/rejected": -6.075833797454834,
587
+ "step": 170
588
+ },
589
+ {
590
+ "epoch": 1.6542222222222223,
591
+ "grad_norm": 1004.2781982421875,
592
+ "learning_rate": 2.458373045823404e-05,
593
+ "log_odds_chosen": 6.2596330642700195,
594
+ "log_odds_ratio": -0.9497385025024414,
595
+ "logps/chosen": -2.4776344299316406,
596
+ "logps/rejected": -8.515233039855957,
597
+ "loss": -1.5757,
598
+ "nll_loss": 2.2205429077148438,
599
+ "po_loss": -2.2697837352752686,
600
+ "rewards/accuracies": 0.7718750238418579,
601
+ "rewards/chosen": -1.2388172149658203,
602
+ "rewards/margins": 3.018799066543579,
603
+ "rewards/rejected": -4.2576165199279785,
604
+ "step": 175
605
+ },
606
+ {
607
+ "epoch": 1.7016296296296296,
608
+ "grad_norm": 3024.766845703125,
609
+ "learning_rate": 2.3197646927086697e-05,
610
+ "log_odds_chosen": 7.638102054595947,
611
+ "log_odds_ratio": -1.0293067693710327,
612
+ "logps/chosen": -3.2467892169952393,
613
+ "logps/rejected": -10.673727989196777,
614
+ "loss": -11.2032,
615
+ "nll_loss": 2.4954748153686523,
616
+ "po_loss": -2.845574378967285,
617
+ "rewards/accuracies": 0.768750011920929,
618
+ "rewards/chosen": -1.6233946084976196,
619
+ "rewards/margins": 3.7134690284729004,
620
+ "rewards/rejected": -5.336863994598389,
621
+ "step": 180
622
+ },
623
+ {
624
+ "epoch": 1.749037037037037,
625
+ "grad_norm": 976.9481201171875,
626
+ "learning_rate": 2.1817114703032176e-05,
627
+ "log_odds_chosen": 7.674588203430176,
628
+ "log_odds_ratio": -1.1676676273345947,
629
+ "logps/chosen": -3.3499820232391357,
630
+ "logps/rejected": -10.849446296691895,
631
+ "loss": -16.751,
632
+ "nll_loss": 2.3611953258514404,
633
+ "po_loss": -2.8846640586853027,
634
+ "rewards/accuracies": 0.7593749761581421,
635
+ "rewards/chosen": -1.6749910116195679,
636
+ "rewards/margins": 3.749732255935669,
637
+ "rewards/rejected": -5.424723148345947,
638
+ "step": 185
639
+ },
640
+ {
641
+ "epoch": 1.7964444444444445,
642
+ "grad_norm": 1403.0980224609375,
643
+ "learning_rate": 2.0446385870993467e-05,
644
+ "log_odds_chosen": 8.678237915039062,
645
+ "log_odds_ratio": -0.6416040062904358,
646
+ "logps/chosen": -2.349724054336548,
647
+ "logps/rejected": -10.788393020629883,
648
+ "loss": -39.35,
649
+ "nll_loss": 1.945783257484436,
650
+ "po_loss": -3.175469398498535,
651
+ "rewards/accuracies": 0.815625011920929,
652
+ "rewards/chosen": -1.174862027168274,
653
+ "rewards/margins": 4.219334602355957,
654
+ "rewards/rejected": -5.394196510314941,
655
+ "step": 190
656
+ },
657
+ {
658
+ "epoch": 1.8438518518518519,
659
+ "grad_norm": 2194.176025390625,
660
+ "learning_rate": 1.9089682321121834e-05,
661
+ "log_odds_chosen": 11.073338508605957,
662
+ "log_odds_ratio": -1.7082958221435547,
663
+ "logps/chosen": -4.391510486602783,
664
+ "logps/rejected": -15.270744323730469,
665
+ "loss": -2.4223,
666
+ "nll_loss": 3.056588649749756,
667
+ "po_loss": -3.132284641265869,
668
+ "rewards/accuracies": 0.7718750238418579,
669
+ "rewards/chosen": -2.1957552433013916,
670
+ "rewards/margins": 5.439617156982422,
671
+ "rewards/rejected": -7.635372161865234,
672
+ "step": 195
673
+ },
674
+ {
675
+ "epoch": 1.8912592592592592,
676
+ "grad_norm": 1372.590576171875,
677
+ "learning_rate": 1.775118274523545e-05,
678
+ "log_odds_chosen": 7.143240928649902,
679
+ "log_odds_ratio": -1.3253802061080933,
680
+ "logps/chosen": -3.159841537475586,
681
+ "logps/rejected": -10.11519718170166,
682
+ "loss": -3.8252,
683
+ "nll_loss": 2.4054999351501465,
684
+ "po_loss": -2.525035858154297,
685
+ "rewards/accuracies": 0.715624988079071,
686
+ "rewards/chosen": -1.579920768737793,
687
+ "rewards/margins": 3.477677583694458,
688
+ "rewards/rejected": -5.05759859085083,
689
+ "step": 200
690
+ },
691
+ {
692
+ "epoch": 1.9386666666666668,
693
+ "grad_norm": 1031.638916015625,
694
+ "learning_rate": 1.643500976631037e-05,
695
+ "log_odds_chosen": 5.937264442443848,
696
+ "log_odds_ratio": -0.8166869878768921,
697
+ "logps/chosen": -2.044642210006714,
698
+ "logps/rejected": -7.761441230773926,
699
+ "loss": -10.1847,
700
+ "nll_loss": 1.9633842706680298,
701
+ "po_loss": -2.281655788421631,
702
+ "rewards/accuracies": 0.7906249761581421,
703
+ "rewards/chosen": -1.022321105003357,
704
+ "rewards/margins": 2.8583993911743164,
705
+ "rewards/rejected": -3.880720615386963,
706
+ "step": 205
707
+ },
708
+ {
709
+ "epoch": 1.986074074074074,
710
+ "grad_norm": 1609.1983642578125,
711
+ "learning_rate": 1.514521724066537e-05,
712
+ "log_odds_chosen": 7.844486236572266,
713
+ "log_odds_ratio": -1.5023125410079956,
714
+ "logps/chosen": -3.9588305950164795,
715
+ "logps/rejected": -11.614171981811523,
716
+ "loss": -3.4843,
717
+ "nll_loss": 2.696570873260498,
718
+ "po_loss": -2.8054568767547607,
719
+ "rewards/accuracies": 0.746874988079071,
720
+ "rewards/chosen": -1.9794152975082397,
721
+ "rewards/margins": 3.827669858932495,
722
+ "rewards/rejected": -5.807085990905762,
723
+ "step": 210
724
+ },
725
+ {
726
+ "epoch": 2.0284444444444443,
727
+ "grad_norm": 1158.2188720703125,
728
+ "learning_rate": 1.3885777771950348e-05,
729
+ "log_odds_chosen": 10.403746604919434,
730
+ "log_odds_ratio": -1.2075212001800537,
731
+ "logps/chosen": -3.1166441440582275,
732
+ "logps/rejected": -13.249801635742188,
733
+ "loss": -41.5962,
734
+ "nll_loss": 2.3673253059387207,
735
+ "po_loss": -3.821737766265869,
736
+ "rewards/accuracies": 0.7867133021354675,
737
+ "rewards/chosen": -1.5583220720291138,
738
+ "rewards/margins": 5.066579341888428,
739
+ "rewards/rejected": -6.624900817871094,
740
+ "step": 215
741
+ },
742
+ {
743
+ "epoch": 2.075851851851852,
744
+ "grad_norm": 1111.02734375,
745
+ "learning_rate": 1.2660570475395683e-05,
746
+ "log_odds_chosen": 8.10044002532959,
747
+ "log_odds_ratio": -1.2627264261245728,
748
+ "logps/chosen": -3.747706890106201,
749
+ "logps/rejected": -11.619832992553711,
750
+ "loss": -19.3237,
751
+ "nll_loss": 2.5686912536621094,
752
+ "po_loss": -3.172555923461914,
753
+ "rewards/accuracies": 0.765625,
754
+ "rewards/chosen": -1.8738534450531006,
755
+ "rewards/margins": 3.936063289642334,
756
+ "rewards/rejected": -5.8099164962768555,
757
+ "step": 220
758
+ },
759
+ {
760
+ "epoch": 2.1232592592592594,
761
+ "grad_norm": 2720.02490234375,
762
+ "learning_rate": 1.1473369030008974e-05,
763
+ "log_odds_chosen": 11.025764465332031,
764
+ "log_odds_ratio": -0.7701746821403503,
765
+ "logps/chosen": -3.573103666305542,
766
+ "logps/rejected": -14.34417724609375,
767
+ "loss": -58.0462,
768
+ "nll_loss": 2.4860126972198486,
769
+ "po_loss": -4.299956798553467,
770
+ "rewards/accuracies": 0.8374999761581421,
771
+ "rewards/chosen": -1.786551833152771,
772
+ "rewards/margins": 5.385537147521973,
773
+ "rewards/rejected": -7.172088623046875,
774
+ "step": 225
775
+ },
776
+ {
777
+ "epoch": 2.1706666666666665,
778
+ "grad_norm": 1029.630126953125,
779
+ "learning_rate": 1.0327830055518842e-05,
780
+ "log_odds_chosen": 10.32060432434082,
781
+ "log_odds_ratio": -0.8015488386154175,
782
+ "logps/chosen": -3.26597261428833,
783
+ "logps/rejected": -13.332249641418457,
784
+ "loss": -49.913,
785
+ "nll_loss": 2.330512762069702,
786
+ "po_loss": -3.890293836593628,
787
+ "rewards/accuracies": 0.8062499761581421,
788
+ "rewards/chosen": -1.632986307144165,
789
+ "rewards/margins": 5.033139228820801,
790
+ "rewards/rejected": -6.6661248207092285,
791
+ "step": 230
792
+ },
793
+ {
794
+ "epoch": 2.218074074074074,
795
+ "grad_norm": 1995.1678466796875,
796
+ "learning_rate": 9.227481849865235e-06,
797
+ "log_odds_chosen": 11.123466491699219,
798
+ "log_odds_ratio": -0.6873899102210999,
799
+ "logps/chosen": -2.032088041305542,
800
+ "logps/rejected": -12.853292465209961,
801
+ "loss": -69.5192,
802
+ "nll_loss": 1.9619925022125244,
803
+ "po_loss": -4.1344685554504395,
804
+ "rewards/accuracies": 0.8187500238418579,
805
+ "rewards/chosen": -1.016044020652771,
806
+ "rewards/margins": 5.410602569580078,
807
+ "rewards/rejected": -6.4266462326049805,
808
+ "step": 235
809
+ },
810
+ {
811
+ "epoch": 2.2654814814814817,
812
+ "grad_norm": 2002.389892578125,
813
+ "learning_rate": 8.175713521924978e-06,
814
+ "log_odds_chosen": 12.202537536621094,
815
+ "log_odds_ratio": -0.6559484004974365,
816
+ "logps/chosen": -2.459198236465454,
817
+ "logps/rejected": -14.35517692565918,
818
+ "loss": -66.0372,
819
+ "nll_loss": 2.1631743907928467,
820
+ "po_loss": -4.226839065551758,
821
+ "rewards/accuracies": 0.859375,
822
+ "rewards/chosen": -1.229599118232727,
823
+ "rewards/margins": 5.947988986968994,
824
+ "rewards/rejected": -7.17758846282959,
825
+ "step": 240
826
+ },
827
+ {
828
+ "epoch": 2.3128888888888888,
829
+ "grad_norm": 1811.5279541015625,
830
+ "learning_rate": 7.1757645529443665e-06,
831
+ "log_odds_chosen": 12.961212158203125,
832
+ "log_odds_ratio": -0.48759812116622925,
833
+ "logps/chosen": -2.229421615600586,
834
+ "logps/rejected": -14.900075912475586,
835
+ "loss": -85.5446,
836
+ "nll_loss": 1.9955413341522217,
837
+ "po_loss": -4.668810844421387,
838
+ "rewards/accuracies": 0.8531249761581421,
839
+ "rewards/chosen": -1.114710807800293,
840
+ "rewards/margins": 6.335326671600342,
841
+ "rewards/rejected": -7.450037956237793,
842
+ "step": 245
843
+ },
844
+ {
845
+ "epoch": 2.3602962962962963,
846
+ "grad_norm": 1374.3148193359375,
847
+ "learning_rate": 6.230714818829733e-06,
848
+ "log_odds_chosen": 13.162447929382324,
849
+ "log_odds_ratio": -0.579684853553772,
850
+ "logps/chosen": -2.2635672092437744,
851
+ "logps/rejected": -15.118783950805664,
852
+ "loss": -78.0242,
853
+ "nll_loss": 1.9523729085922241,
854
+ "po_loss": -4.39063024520874,
855
+ "rewards/accuracies": 0.824999988079071,
856
+ "rewards/chosen": -1.1317836046218872,
857
+ "rewards/margins": 6.427608489990234,
858
+ "rewards/rejected": -7.559391975402832,
859
+ "step": 250
860
+ },
861
+ {
862
+ "epoch": 2.407703703703704,
863
+ "grad_norm": 1782.409423828125,
864
+ "learning_rate": 5.343475104027743e-06,
865
+ "log_odds_chosen": 12.180627822875977,
866
+ "log_odds_ratio": -0.7269414663314819,
867
+ "logps/chosen": -2.7276577949523926,
868
+ "logps/rejected": -14.642044067382812,
869
+ "loss": -61.2515,
870
+ "nll_loss": 2.155933380126953,
871
+ "po_loss": -4.07004451751709,
872
+ "rewards/accuracies": 0.831250011920929,
873
+ "rewards/chosen": -1.3638288974761963,
874
+ "rewards/margins": 5.957192897796631,
875
+ "rewards/rejected": -7.321022033691406,
876
+ "step": 255
877
+ },
878
+ {
879
+ "epoch": 2.455111111111111,
880
+ "grad_norm": 1750.225341796875,
881
+ "learning_rate": 4.516778136213037e-06,
882
+ "log_odds_chosen": 12.096226692199707,
883
+ "log_odds_ratio": -0.6728618741035461,
884
+ "logps/chosen": -2.2444777488708496,
885
+ "logps/rejected": -14.059532165527344,
886
+ "loss": -72.1379,
887
+ "nll_loss": 1.928434133529663,
888
+ "po_loss": -4.182744026184082,
889
+ "rewards/accuracies": 0.84375,
890
+ "rewards/chosen": -1.1222388744354248,
891
+ "rewards/margins": 5.907527923583984,
892
+ "rewards/rejected": -7.029766082763672,
893
+ "step": 260
894
+ },
895
+ {
896
+ "epoch": 2.5025185185185186,
897
+ "grad_norm": 1919.8836669921875,
898
+ "learning_rate": 3.7531701693965554e-06,
899
+ "log_odds_chosen": 11.939556121826172,
900
+ "log_odds_ratio": -0.5605658888816833,
901
+ "logps/chosen": -2.554964780807495,
902
+ "logps/rejected": -14.228480339050293,
903
+ "loss": -72.1107,
904
+ "nll_loss": 2.096379518508911,
905
+ "po_loss": -4.3498382568359375,
906
+ "rewards/accuracies": 0.84375,
907
+ "rewards/chosen": -1.2774823904037476,
908
+ "rewards/margins": 5.836758136749268,
909
+ "rewards/rejected": -7.1142401695251465,
910
+ "step": 265
911
+ },
912
+ {
913
+ "epoch": 2.549925925925926,
914
+ "grad_norm": 3865.446044921875,
915
+ "learning_rate": 3.055003141378948e-06,
916
+ "log_odds_chosen": 13.130399703979492,
917
+ "log_odds_ratio": -0.7432569265365601,
918
+ "logps/chosen": -2.868731737136841,
919
+ "logps/rejected": -15.705291748046875,
920
+ "loss": -71.6455,
921
+ "nll_loss": 2.3375840187072754,
922
+ "po_loss": -4.576504230499268,
923
+ "rewards/accuracies": 0.8374999761581421,
924
+ "rewards/chosen": -1.4343658685684204,
925
+ "rewards/margins": 6.418280124664307,
926
+ "rewards/rejected": -7.8526458740234375,
927
+ "step": 270
928
+ },
929
+ {
930
+ "epoch": 2.5973333333333333,
931
+ "grad_norm": 1720.9822998046875,
932
+ "learning_rate": 2.424427429704365e-06,
933
+ "log_odds_chosen": 12.388298034667969,
934
+ "log_odds_ratio": -0.7917383313179016,
935
+ "logps/chosen": -2.539780378341675,
936
+ "logps/rejected": -14.631436347961426,
937
+ "loss": -74.1033,
938
+ "nll_loss": 2.096407651901245,
939
+ "po_loss": -4.412135124206543,
940
+ "rewards/accuracies": 0.862500011920929,
941
+ "rewards/chosen": -1.2698901891708374,
942
+ "rewards/margins": 6.045827865600586,
943
+ "rewards/rejected": -7.315718173980713,
944
+ "step": 275
945
+ },
946
+ {
947
+ "epoch": 2.644740740740741,
948
+ "grad_norm": 1882.84033203125,
949
+ "learning_rate": 1.8633852284264508e-06,
950
+ "log_odds_chosen": 13.360041618347168,
951
+ "log_odds_ratio": -0.7875075936317444,
952
+ "logps/chosen": -2.570291757583618,
953
+ "logps/rejected": -15.660143852233887,
954
+ "loss": -80.8755,
955
+ "nll_loss": 2.1218926906585693,
956
+ "po_loss": -4.649253845214844,
957
+ "rewards/accuracies": 0.840624988079071,
958
+ "rewards/chosen": -1.285145878791809,
959
+ "rewards/margins": 6.544926643371582,
960
+ "rewards/rejected": -7.830071926116943,
961
+ "step": 280
962
+ },
963
+ {
964
+ "epoch": 2.6921481481481484,
965
+ "grad_norm": 2601.312255859375,
966
+ "learning_rate": 1.3736045660864034e-06,
967
+ "log_odds_chosen": 12.91968059539795,
968
+ "log_odds_ratio": -0.935935378074646,
969
+ "logps/chosen": -3.055326461791992,
970
+ "logps/rejected": -15.716508865356445,
971
+ "loss": -68.9803,
972
+ "nll_loss": 2.237934112548828,
973
+ "po_loss": -4.393568515777588,
974
+ "rewards/accuracies": 0.846875011920929,
975
+ "rewards/chosen": -1.527663230895996,
976
+ "rewards/margins": 6.33059024810791,
977
+ "rewards/rejected": -7.858254432678223,
978
+ "step": 285
979
+ },
980
+ {
981
+ "epoch": 2.7395555555555555,
982
+ "grad_norm": 2271.935302734375,
983
+ "learning_rate": 9.565939833279192e-07,
984
+ "log_odds_chosen": 12.334630966186523,
985
+ "log_odds_ratio": -0.5848469734191895,
986
+ "logps/chosen": -3.0595030784606934,
987
+ "logps/rejected": -15.129420280456543,
988
+ "loss": -64.562,
989
+ "nll_loss": 2.4212985038757324,
990
+ "po_loss": -4.438861846923828,
991
+ "rewards/accuracies": 0.8374999761581421,
992
+ "rewards/chosen": -1.5297515392303467,
993
+ "rewards/margins": 6.034958362579346,
994
+ "rewards/rejected": -7.5647101402282715,
995
+ "step": 290
996
+ },
997
+ {
998
+ "epoch": 2.786962962962963,
999
+ "grad_norm": 1566.753173828125,
1000
+ "learning_rate": 6.136378865420872e-07,
1001
+ "log_odds_chosen": 11.438375473022461,
1002
+ "log_odds_ratio": -0.9271749258041382,
1003
+ "logps/chosen": -2.976278781890869,
1004
+ "logps/rejected": -14.156593322753906,
1005
+ "loss": -57.0123,
1006
+ "nll_loss": 2.327953577041626,
1007
+ "po_loss": -4.109589576721191,
1008
+ "rewards/accuracies": 0.8187500238418579,
1009
+ "rewards/chosen": -1.4881393909454346,
1010
+ "rewards/margins": 5.5901570320129395,
1011
+ "rewards/rejected": -7.078296661376953,
1012
+ "step": 295
1013
+ },
1014
+ {
1015
+ "epoch": 2.83437037037037,
1016
+ "grad_norm": 2821.74609375,
1017
+ "learning_rate": 3.45792591853214e-07,
1018
+ "log_odds_chosen": 11.852496147155762,
1019
+ "log_odds_ratio": -0.5676802396774292,
1020
+ "logps/chosen": -2.4296622276306152,
1021
+ "logps/rejected": -14.001643180847168,
1022
+ "loss": -73.6596,
1023
+ "nll_loss": 1.9556903839111328,
1024
+ "po_loss": -4.2575531005859375,
1025
+ "rewards/accuracies": 0.815625011920929,
1026
+ "rewards/chosen": -1.2148311138153076,
1027
+ "rewards/margins": 5.785989761352539,
1028
+ "rewards/rejected": -7.000821590423584,
1029
+ "step": 300
1030
+ },
1031
+ {
1032
+ "epoch": 2.8817777777777778,
1033
+ "grad_norm": 2135.990966796875,
1034
+ "learning_rate": 1.538830716302092e-07,
1035
+ "log_odds_chosen": 12.745733261108398,
1036
+ "log_odds_ratio": -0.7597203850746155,
1037
+ "logps/chosen": -2.1967830657958984,
1038
+ "logps/rejected": -14.625683784484863,
1039
+ "loss": -78.7585,
1040
+ "nll_loss": 1.9788535833358765,
1041
+ "po_loss": -4.440057277679443,
1042
+ "rewards/accuracies": 0.8500000238418579,
1043
+ "rewards/chosen": -1.0983915328979492,
1044
+ "rewards/margins": 6.214449882507324,
1045
+ "rewards/rejected": -7.312841892242432,
1046
+ "step": 305
1047
+ },
1048
+ {
1049
+ "epoch": 2.9291851851851853,
1050
+ "grad_norm": 1398.578369140625,
1051
+ "learning_rate": 3.8500413544415025e-08,
1052
+ "log_odds_chosen": 13.022871017456055,
1053
+ "log_odds_ratio": -0.4309860169887543,
1054
+ "logps/chosen": -2.221106767654419,
1055
+ "logps/rejected": -14.942235946655273,
1056
+ "loss": -86.5137,
1057
+ "nll_loss": 1.9662892818450928,
1058
+ "po_loss": -4.66984224319458,
1059
+ "rewards/accuracies": 0.871874988079071,
1060
+ "rewards/chosen": -1.1105533838272095,
1061
+ "rewards/margins": 6.360564708709717,
1062
+ "rewards/rejected": -7.471117973327637,
1063
+ "step": 310
1064
+ },
1065
+ {
1066
+ "epoch": 2.9765925925925925,
1067
+ "grad_norm": 1498.622314453125,
1068
+ "learning_rate": 0.0,
1069
+ "log_odds_chosen": 11.79759407043457,
1070
+ "log_odds_ratio": -0.305752694606781,
1071
+ "logps/chosen": -1.7462921142578125,
1072
+ "logps/rejected": -13.216029167175293,
1073
+ "loss": -82.7571,
1074
+ "nll_loss": 1.7008321285247803,
1075
+ "po_loss": -4.286990642547607,
1076
+ "rewards/accuracies": 0.859375,
1077
+ "rewards/chosen": -0.8731460571289062,
1078
+ "rewards/margins": 5.734868049621582,
1079
+ "rewards/rejected": -6.6080145835876465,
1080
+ "step": 315
1081
+ },
1082
+ {
1083
+ "epoch": 2.9765925925925925,
1084
+ "step": 315,
1085
+ "total_flos": 0.0,
1086
+ "train_loss": 16.270644454350546,
1087
+ "train_runtime": 9547.5166,
1088
+ "train_samples_per_second": 2.121,
1089
+ "train_steps_per_second": 0.033
1090
+ }
1091
+ ],
1092
+ "logging_steps": 5,
1093
+ "max_steps": 315,
1094
+ "num_input_tokens_seen": 0,
1095
+ "num_train_epochs": 3,
1096
+ "save_steps": 100000,
1097
+ "stateful_callbacks": {
1098
+ "TrainerControl": {
1099
+ "args": {
1100
+ "should_epoch_stop": false,
1101
+ "should_evaluate": false,
1102
+ "should_log": false,
1103
+ "should_save": true,
1104
+ "should_training_stop": true
1105
+ },
1106
+ "attributes": {}
1107
+ }
1108
+ },
1109
+ "total_flos": 0.0,
1110
+ "train_batch_size": 1,
1111
+ "trial_name": null,
1112
+ "trial_params": null
1113
+ }