martimfasantos commited on
Commit
7ae2afd
1 Parent(s): 23c88d5

Model save

Browse files
Files changed (5) hide show
  1. README.md +60 -0
  2. all_results.json +9 -0
  3. generation_config.json +11 -0
  4. train_results.json +9 -0
  5. trainer_state.json +1152 -0
README.md ADDED
@@ -0,0 +1,60 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: gemma
3
+ base_model: martimfasantos/gemma-2-2b-it-mt-sft-full_sardine
4
+ tags:
5
+ - trl
6
+ - dpo
7
+ - generated_from_trainer
8
+ model-index:
9
+ - name: gemma-2-2b-it-mt-dpo-full_sardine
10
+ results: []
11
+ ---
12
+
13
+ <!-- This model card has been generated automatically according to the information the Trainer had access to. You
14
+ should probably proofread and complete it, then remove this comment. -->
15
+
16
+ # gemma-2-2b-it-mt-dpo-full_sardine
17
+
18
+ This model is a fine-tuned version of [martimfasantos/gemma-2-2b-it-mt-sft-full_sardine](https://huggingface.co/martimfasantos/gemma-2-2b-it-mt-sft-full_sardine) on an unknown dataset.
19
+
20
+ ## Model description
21
+
22
+ More information needed
23
+
24
+ ## Intended uses & limitations
25
+
26
+ More information needed
27
+
28
+ ## Training and evaluation data
29
+
30
+ More information needed
31
+
32
+ ## Training procedure
33
+
34
+ ### Training hyperparameters
35
+
36
+ The following hyperparameters were used during training:
37
+ - learning_rate: 1e-07
38
+ - train_batch_size: 1
39
+ - eval_batch_size: 4
40
+ - seed: 42
41
+ - distributed_type: multi-GPU
42
+ - num_devices: 2
43
+ - gradient_accumulation_steps: 32
44
+ - total_train_batch_size: 64
45
+ - total_eval_batch_size: 8
46
+ - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
47
+ - lr_scheduler_type: cosine
48
+ - lr_scheduler_warmup_ratio: 0.1
49
+ - num_epochs: 3
50
+
51
+ ### Training results
52
+
53
+
54
+
55
+ ### Framework versions
56
+
57
+ - Transformers 4.43.3
58
+ - Pytorch 2.1.2+cu121
59
+ - Datasets 2.20.0
60
+ - Tokenizers 0.19.1
all_results.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 2.9897455374097985,
3
+ "total_flos": 0.0,
4
+ "train_loss": 0.4158425409457871,
5
+ "train_runtime": 20368.7971,
6
+ "train_samples": 15798,
7
+ "train_samples_per_second": 2.327,
8
+ "train_steps_per_second": 0.036
9
+ }
generation_config.json ADDED
@@ -0,0 +1,11 @@
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 2,
4
+ "cache_implementation": "hybrid",
5
+ "eos_token_id": [
6
+ 1,
7
+ 107
8
+ ],
9
+ "pad_token_id": 0,
10
+ "transformers_version": "4.43.3"
11
+ }
train_results.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 2.9897455374097985,
3
+ "total_flos": 0.0,
4
+ "train_loss": 0.4158425409457871,
5
+ "train_runtime": 20368.7971,
6
+ "train_samples": 15798,
7
+ "train_samples_per_second": 2.327,
8
+ "train_steps_per_second": 0.036
9
+ }
trainer_state.json ADDED
@@ -0,0 +1,1152 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 2.9897455374097985,
5
+ "eval_steps": 800,
6
+ "global_step": 738,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.004051145714647424,
13
+ "grad_norm": 23.354017564608814,
14
+ "learning_rate": 1.3513513513513514e-09,
15
+ "logits/chosen": -8.401385307312012,
16
+ "logits/rejected": -7.887872695922852,
17
+ "logps/chosen": -22.588991165161133,
18
+ "logps/rejected": -32.73207473754883,
19
+ "loss": 0.6931,
20
+ "rewards/accuracies": 0.0,
21
+ "rewards/chosen": 0.0,
22
+ "rewards/margins": 0.0,
23
+ "rewards/rejected": 0.0,
24
+ "step": 1
25
+ },
26
+ {
27
+ "epoch": 0.040511457146474236,
28
+ "grad_norm": 23.005760986532493,
29
+ "learning_rate": 1.3513513513513514e-08,
30
+ "logits/chosen": -8.023005485534668,
31
+ "logits/rejected": -7.215701580047607,
32
+ "logps/chosen": -26.691545486450195,
33
+ "logps/rejected": -35.018611907958984,
34
+ "loss": 0.6932,
35
+ "rewards/accuracies": 0.3958333432674408,
36
+ "rewards/chosen": -0.00048815118498168886,
37
+ "rewards/margins": -0.001463304040953517,
38
+ "rewards/rejected": 0.0009751527686603367,
39
+ "step": 10
40
+ },
41
+ {
42
+ "epoch": 0.08102291429294847,
43
+ "grad_norm": 25.76788201635075,
44
+ "learning_rate": 2.7027027027027028e-08,
45
+ "logits/chosen": -8.08788776397705,
46
+ "logits/rejected": -7.492005825042725,
47
+ "logps/chosen": -28.649845123291016,
48
+ "logps/rejected": -38.781715393066406,
49
+ "loss": 0.6934,
50
+ "rewards/accuracies": 0.4937500059604645,
51
+ "rewards/chosen": 0.0005952938226982951,
52
+ "rewards/margins": 0.00045213793055154383,
53
+ "rewards/rejected": 0.0001431556447641924,
54
+ "step": 20
55
+ },
56
+ {
57
+ "epoch": 0.12153437143942271,
58
+ "grad_norm": 24.285898108451498,
59
+ "learning_rate": 4.054054054054054e-08,
60
+ "logits/chosen": -8.020319938659668,
61
+ "logits/rejected": -7.39060115814209,
62
+ "logps/chosen": -26.554407119750977,
63
+ "logps/rejected": -39.03184509277344,
64
+ "loss": 0.6916,
65
+ "rewards/accuracies": 0.518750011920929,
66
+ "rewards/chosen": 0.0008592296508140862,
67
+ "rewards/margins": 0.003104406874626875,
68
+ "rewards/rejected": -0.002245177049189806,
69
+ "step": 30
70
+ },
71
+ {
72
+ "epoch": 0.16204582858589694,
73
+ "grad_norm": 26.673301885189623,
74
+ "learning_rate": 5.4054054054054056e-08,
75
+ "logits/chosen": -8.143043518066406,
76
+ "logits/rejected": -7.433783531188965,
77
+ "logps/chosen": -25.945232391357422,
78
+ "logps/rejected": -37.73704147338867,
79
+ "loss": 0.6882,
80
+ "rewards/accuracies": 0.574999988079071,
81
+ "rewards/chosen": 9.569372195983306e-05,
82
+ "rewards/margins": 0.007951222360134125,
83
+ "rewards/rejected": -0.007855528965592384,
84
+ "step": 40
85
+ },
86
+ {
87
+ "epoch": 0.2025572857323712,
88
+ "grad_norm": 26.497438005541635,
89
+ "learning_rate": 6.756756756756756e-08,
90
+ "logits/chosen": -8.0227632522583,
91
+ "logits/rejected": -7.415746212005615,
92
+ "logps/chosen": -28.301464080810547,
93
+ "logps/rejected": -38.43585205078125,
94
+ "loss": 0.6843,
95
+ "rewards/accuracies": 0.6468750238418579,
96
+ "rewards/chosen": 0.0006854601087979972,
97
+ "rewards/margins": 0.0169823057949543,
98
+ "rewards/rejected": -0.01629684492945671,
99
+ "step": 50
100
+ },
101
+ {
102
+ "epoch": 0.24306874287884542,
103
+ "grad_norm": 24.919397084889898,
104
+ "learning_rate": 8.108108108108108e-08,
105
+ "logits/chosen": -8.055706024169922,
106
+ "logits/rejected": -7.392535209655762,
107
+ "logps/chosen": -26.291152954101562,
108
+ "logps/rejected": -37.135032653808594,
109
+ "loss": 0.6751,
110
+ "rewards/accuracies": 0.734375,
111
+ "rewards/chosen": 0.0014377154875546694,
112
+ "rewards/margins": 0.03630940988659859,
113
+ "rewards/rejected": -0.03487168997526169,
114
+ "step": 60
115
+ },
116
+ {
117
+ "epoch": 0.28358020002531964,
118
+ "grad_norm": 20.09272984199583,
119
+ "learning_rate": 9.45945945945946e-08,
120
+ "logits/chosen": -8.232405662536621,
121
+ "logits/rejected": -7.535184383392334,
122
+ "logps/chosen": -27.17940902709961,
123
+ "logps/rejected": -36.42096710205078,
124
+ "loss": 0.6635,
125
+ "rewards/accuracies": 0.824999988079071,
126
+ "rewards/chosen": 0.0027192712295800447,
127
+ "rewards/margins": 0.06449674069881439,
128
+ "rewards/rejected": -0.06177746504545212,
129
+ "step": 70
130
+ },
131
+ {
132
+ "epoch": 0.3240916571717939,
133
+ "grad_norm": 23.04151710177534,
134
+ "learning_rate": 9.997985455197112e-08,
135
+ "logits/chosen": -8.351137161254883,
136
+ "logits/rejected": -7.78473424911499,
137
+ "logps/chosen": -26.988977432250977,
138
+ "logps/rejected": -36.72052001953125,
139
+ "loss": 0.6537,
140
+ "rewards/accuracies": 0.753125011920929,
141
+ "rewards/chosen": 0.00015634755254723132,
142
+ "rewards/margins": 0.08445531874895096,
143
+ "rewards/rejected": -0.08429896831512451,
144
+ "step": 80
145
+ },
146
+ {
147
+ "epoch": 0.36460311431826814,
148
+ "grad_norm": 24.86811296840432,
149
+ "learning_rate": 9.98568022639826e-08,
150
+ "logits/chosen": -8.316564559936523,
151
+ "logits/rejected": -7.610459327697754,
152
+ "logps/chosen": -28.134241104125977,
153
+ "logps/rejected": -40.095638275146484,
154
+ "loss": 0.6307,
155
+ "rewards/accuracies": 0.862500011920929,
156
+ "rewards/chosen": 0.005406993441283703,
157
+ "rewards/margins": 0.13842421770095825,
158
+ "rewards/rejected": -0.13301721215248108,
159
+ "step": 90
160
+ },
161
+ {
162
+ "epoch": 0.4051145714647424,
163
+ "grad_norm": 23.930696194698072,
164
+ "learning_rate": 9.962216467480192e-08,
165
+ "logits/chosen": -8.399057388305664,
166
+ "logits/rejected": -7.728206634521484,
167
+ "logps/chosen": -27.389541625976562,
168
+ "logps/rejected": -42.16829299926758,
169
+ "loss": 0.6168,
170
+ "rewards/accuracies": 0.796875,
171
+ "rewards/chosen": -0.0014130814233794808,
172
+ "rewards/margins": 0.17133907973766327,
173
+ "rewards/rejected": -0.17275215685367584,
174
+ "step": 100
175
+ },
176
+ {
177
+ "epoch": 0.4456260286112166,
178
+ "grad_norm": 25.12250587863022,
179
+ "learning_rate": 9.927646693054496e-08,
180
+ "logits/chosen": -8.546191215515137,
181
+ "logits/rejected": -7.818883419036865,
182
+ "logps/chosen": -26.858484268188477,
183
+ "logps/rejected": -42.138423919677734,
184
+ "loss": 0.5877,
185
+ "rewards/accuracies": 0.846875011920929,
186
+ "rewards/chosen": -0.002900606021285057,
187
+ "rewards/margins": 0.24653196334838867,
188
+ "rewards/rejected": -0.24943256378173828,
189
+ "step": 110
190
+ },
191
+ {
192
+ "epoch": 0.48613748575769083,
193
+ "grad_norm": 23.993302988236724,
194
+ "learning_rate": 9.882048274282503e-08,
195
+ "logits/chosen": -8.554715156555176,
196
+ "logits/rejected": -7.8766679763793945,
197
+ "logps/chosen": -26.874399185180664,
198
+ "logps/rejected": -40.70722961425781,
199
+ "loss": 0.5763,
200
+ "rewards/accuracies": 0.8031250238418579,
201
+ "rewards/chosen": -0.026281490921974182,
202
+ "rewards/margins": 0.26420071721076965,
203
+ "rewards/rejected": -0.29048219323158264,
204
+ "step": 120
205
+ },
206
+ {
207
+ "epoch": 0.5266489429041651,
208
+ "grad_norm": 19.760387337865797,
209
+ "learning_rate": 9.825523265709666e-08,
210
+ "logits/chosen": -8.872574806213379,
211
+ "logits/rejected": -8.05439281463623,
212
+ "logps/chosen": -29.766094207763672,
213
+ "logps/rejected": -39.4333381652832,
214
+ "loss": 0.5627,
215
+ "rewards/accuracies": 0.84375,
216
+ "rewards/chosen": -0.01587745174765587,
217
+ "rewards/margins": 0.3382584750652313,
218
+ "rewards/rejected": -0.3541359305381775,
219
+ "step": 130
220
+ },
221
+ {
222
+ "epoch": 0.5671604000506393,
223
+ "grad_norm": 20.405644698254697,
224
+ "learning_rate": 9.758198176855646e-08,
225
+ "logits/chosen": -8.83372974395752,
226
+ "logits/rejected": -8.060525894165039,
227
+ "logps/chosen": -27.360458374023438,
228
+ "logps/rejected": -41.453636169433594,
229
+ "loss": 0.5348,
230
+ "rewards/accuracies": 0.846875011920929,
231
+ "rewards/chosen": -0.028619933873414993,
232
+ "rewards/margins": 0.39799413084983826,
233
+ "rewards/rejected": -0.42661410570144653,
234
+ "step": 140
235
+ },
236
+ {
237
+ "epoch": 0.6076718571971136,
238
+ "grad_norm": 24.42451138080484,
239
+ "learning_rate": 9.680223689071363e-08,
240
+ "logits/chosen": -9.05488109588623,
241
+ "logits/rejected": -8.361872673034668,
242
+ "logps/chosen": -27.39214515686035,
243
+ "logps/rejected": -44.14860916137695,
244
+ "loss": 0.5253,
245
+ "rewards/accuracies": 0.856249988079071,
246
+ "rewards/chosen": -0.041482653468847275,
247
+ "rewards/margins": 0.43668460845947266,
248
+ "rewards/rejected": -0.47816720604896545,
249
+ "step": 150
250
+ },
251
+ {
252
+ "epoch": 0.6481833143435878,
253
+ "grad_norm": 22.08758150384141,
254
+ "learning_rate": 9.59177431829666e-08,
255
+ "logits/chosen": -9.038153648376465,
256
+ "logits/rejected": -8.328339576721191,
257
+ "logps/chosen": -27.219493865966797,
258
+ "logps/rejected": -43.879791259765625,
259
+ "loss": 0.5219,
260
+ "rewards/accuracies": 0.828125,
261
+ "rewards/chosen": -0.039076704531908035,
262
+ "rewards/margins": 0.47056007385253906,
263
+ "rewards/rejected": -0.5096367597579956,
264
+ "step": 160
265
+ },
266
+ {
267
+ "epoch": 0.6886947714900621,
268
+ "grad_norm": 22.525750174655215,
269
+ "learning_rate": 9.493048024473411e-08,
270
+ "logits/chosen": -9.239489555358887,
271
+ "logits/rejected": -8.590790748596191,
272
+ "logps/chosen": -27.6241455078125,
273
+ "logps/rejected": -43.37400817871094,
274
+ "loss": 0.5152,
275
+ "rewards/accuracies": 0.8187500238418579,
276
+ "rewards/chosen": -0.07299178838729858,
277
+ "rewards/margins": 0.4755212664604187,
278
+ "rewards/rejected": -0.5485130548477173,
279
+ "step": 170
280
+ },
281
+ {
282
+ "epoch": 0.7292062286365363,
283
+ "grad_norm": 21.336294616647535,
284
+ "learning_rate": 9.384265768488224e-08,
285
+ "logits/chosen": -9.427377700805664,
286
+ "logits/rejected": -8.607305526733398,
287
+ "logps/chosen": -27.55912208557129,
288
+ "logps/rejected": -42.556678771972656,
289
+ "loss": 0.4916,
290
+ "rewards/accuracies": 0.840624988079071,
291
+ "rewards/chosen": -0.07397119700908661,
292
+ "rewards/margins": 0.5928912162780762,
293
+ "rewards/rejected": -0.666862428188324,
294
+ "step": 180
295
+ },
296
+ {
297
+ "epoch": 0.7697176857830105,
298
+ "grad_norm": 20.933312614297105,
299
+ "learning_rate": 9.265671017636383e-08,
300
+ "logits/chosen": -9.61644458770752,
301
+ "logits/rejected": -8.950950622558594,
302
+ "logps/chosen": -28.78060531616211,
303
+ "logps/rejected": -45.048828125,
304
+ "loss": 0.4693,
305
+ "rewards/accuracies": 0.840624988079071,
306
+ "rewards/chosen": -0.06900081783533096,
307
+ "rewards/margins": 0.6354625821113586,
308
+ "rewards/rejected": -0.7044633030891418,
309
+ "step": 190
310
+ },
311
+ {
312
+ "epoch": 0.8102291429294848,
313
+ "grad_norm": 20.036505723069673,
314
+ "learning_rate": 9.137529200713809e-08,
315
+ "logits/chosen": -9.64765739440918,
316
+ "logits/rejected": -8.868057250976562,
317
+ "logps/chosen": -26.3610782623291,
318
+ "logps/rejected": -43.371124267578125,
319
+ "loss": 0.4555,
320
+ "rewards/accuracies": 0.8031250238418579,
321
+ "rewards/chosen": -0.10404930263757706,
322
+ "rewards/margins": 0.7085543870925903,
323
+ "rewards/rejected": -0.812603771686554,
324
+ "step": 200
325
+ },
326
+ {
327
+ "epoch": 0.850740600075959,
328
+ "grad_norm": 21.18417249103817,
329
+ "learning_rate": 9.000127113956672e-08,
330
+ "logits/chosen": -9.870735168457031,
331
+ "logits/rejected": -9.190484046936035,
332
+ "logps/chosen": -26.790430068969727,
333
+ "logps/rejected": -43.78266525268555,
334
+ "loss": 0.4639,
335
+ "rewards/accuracies": 0.8343750238418579,
336
+ "rewards/chosen": -0.10198328644037247,
337
+ "rewards/margins": 0.7101414799690247,
338
+ "rewards/rejected": -0.8121248483657837,
339
+ "step": 210
340
+ },
341
+ {
342
+ "epoch": 0.8912520572224332,
343
+ "grad_norm": 18.99507288099157,
344
+ "learning_rate": 8.853772279158165e-08,
345
+ "logits/chosen": -9.834332466125488,
346
+ "logits/rejected": -9.097946166992188,
347
+ "logps/chosen": -27.970256805419922,
348
+ "logps/rejected": -44.3878059387207,
349
+ "loss": 0.4372,
350
+ "rewards/accuracies": 0.84375,
351
+ "rewards/chosen": -0.1515013873577118,
352
+ "rewards/margins": 0.78831946849823,
353
+ "rewards/rejected": -0.9398208856582642,
354
+ "step": 220
355
+ },
356
+ {
357
+ "epoch": 0.9317635143689075,
358
+ "grad_norm": 22.643159035198256,
359
+ "learning_rate": 8.698792255399104e-08,
360
+ "logits/chosen": -10.250986099243164,
361
+ "logits/rejected": -9.542135238647461,
362
+ "logps/chosen": -28.57198715209961,
363
+ "logps/rejected": -47.67482376098633,
364
+ "loss": 0.4174,
365
+ "rewards/accuracies": 0.846875011920929,
366
+ "rewards/chosen": -0.16280151903629303,
367
+ "rewards/margins": 0.9155281186103821,
368
+ "rewards/rejected": -1.0783296823501587,
369
+ "step": 230
370
+ },
371
+ {
372
+ "epoch": 0.9722749715153817,
373
+ "grad_norm": 17.009463712911476,
374
+ "learning_rate": 8.535533905932736e-08,
375
+ "logits/chosen": -10.252339363098145,
376
+ "logits/rejected": -9.567429542541504,
377
+ "logps/chosen": -29.68218421936035,
378
+ "logps/rejected": -49.98845291137695,
379
+ "loss": 0.4311,
380
+ "rewards/accuracies": 0.840624988079071,
381
+ "rewards/chosen": -0.21189716458320618,
382
+ "rewards/margins": 0.9156039953231812,
383
+ "rewards/rejected": -1.1275012493133545,
384
+ "step": 240
385
+ },
386
+ {
387
+ "epoch": 1.012786428661856,
388
+ "grad_norm": 21.283801315155564,
389
+ "learning_rate": 8.364362621864594e-08,
390
+ "logits/chosen": -10.185906410217285,
391
+ "logits/rejected": -9.48774528503418,
392
+ "logps/chosen": -28.432941436767578,
393
+ "logps/rejected": -50.117576599121094,
394
+ "loss": 0.4013,
395
+ "rewards/accuracies": 0.859375,
396
+ "rewards/chosen": -0.21948130428791046,
397
+ "rewards/margins": 0.9736725687980652,
398
+ "rewards/rejected": -1.1931538581848145,
399
+ "step": 250
400
+ },
401
+ {
402
+ "epoch": 1.0532978858083302,
403
+ "grad_norm": 22.754490059016796,
404
+ "learning_rate": 8.185661504364844e-08,
405
+ "logits/chosen": -10.365915298461914,
406
+ "logits/rejected": -9.623150825500488,
407
+ "logps/chosen": -28.134822845458984,
408
+ "logps/rejected": -48.77445983886719,
409
+ "loss": 0.3799,
410
+ "rewards/accuracies": 0.8843749761581421,
411
+ "rewards/chosen": -0.22865962982177734,
412
+ "rewards/margins": 1.057023286819458,
413
+ "rewards/rejected": -1.2856829166412354,
414
+ "step": 260
415
+ },
416
+ {
417
+ "epoch": 1.0938093429548044,
418
+ "grad_norm": 29.101808968686573,
419
+ "learning_rate": 7.999830507243477e-08,
420
+ "logits/chosen": -10.540546417236328,
421
+ "logits/rejected": -9.749776840209961,
422
+ "logps/chosen": -32.03081512451172,
423
+ "logps/rejected": -50.996429443359375,
424
+ "loss": 0.4043,
425
+ "rewards/accuracies": 0.846875011920929,
426
+ "rewards/chosen": -0.30270159244537354,
427
+ "rewards/margins": 1.1776001453399658,
428
+ "rewards/rejected": -1.480301856994629,
429
+ "step": 270
430
+ },
431
+ {
432
+ "epoch": 1.1343208001012786,
433
+ "grad_norm": 22.52674198964978,
434
+ "learning_rate": 7.80728554180734e-08,
435
+ "logits/chosen": -10.482057571411133,
436
+ "logits/rejected": -9.802620887756348,
437
+ "logps/chosen": -29.116191864013672,
438
+ "logps/rejected": -51.501319885253906,
439
+ "loss": 0.3695,
440
+ "rewards/accuracies": 0.856249988079071,
441
+ "rewards/chosen": -0.2759430706501007,
442
+ "rewards/margins": 1.228001356124878,
443
+ "rewards/rejected": -1.5039442777633667,
444
+ "step": 280
445
+ },
446
+ {
447
+ "epoch": 1.174832257247753,
448
+ "grad_norm": 23.744363106936603,
449
+ "learning_rate": 7.608457546002422e-08,
450
+ "logits/chosen": -10.582446098327637,
451
+ "logits/rejected": -9.793972969055176,
452
+ "logps/chosen": -30.690120697021484,
453
+ "logps/rejected": -52.65431594848633,
454
+ "loss": 0.3608,
455
+ "rewards/accuracies": 0.846875011920929,
456
+ "rewards/chosen": -0.35969388484954834,
457
+ "rewards/margins": 1.3402436971664429,
458
+ "rewards/rejected": -1.6999378204345703,
459
+ "step": 290
460
+ },
461
+ {
462
+ "epoch": 1.2153437143942272,
463
+ "grad_norm": 16.96872058311971,
464
+ "learning_rate": 7.403791519924793e-08,
465
+ "logits/chosen": -10.636259078979492,
466
+ "logits/rejected": -9.861931800842285,
467
+ "logps/chosen": -30.6519718170166,
468
+ "logps/rejected": -55.83306884765625,
469
+ "loss": 0.3609,
470
+ "rewards/accuracies": 0.856249988079071,
471
+ "rewards/chosen": -0.34394341707229614,
472
+ "rewards/margins": 1.4830596446990967,
473
+ "rewards/rejected": -1.8270031213760376,
474
+ "step": 300
475
+ },
476
+ {
477
+ "epoch": 1.2558551715407014,
478
+ "grad_norm": 23.80487547845051,
479
+ "learning_rate": 7.193745529858826e-08,
480
+ "logits/chosen": -10.673724174499512,
481
+ "logits/rejected": -9.968966484069824,
482
+ "logps/chosen": -32.3521614074707,
483
+ "logps/rejected": -55.50031661987305,
484
+ "loss": 0.3796,
485
+ "rewards/accuracies": 0.878125011920929,
486
+ "rewards/chosen": -0.4293053150177002,
487
+ "rewards/margins": 1.2904064655303955,
488
+ "rewards/rejected": -1.7197116613388062,
489
+ "step": 310
490
+ },
491
+ {
492
+ "epoch": 1.2963666286871756,
493
+ "grad_norm": 26.02880426926903,
494
+ "learning_rate": 6.978789683071759e-08,
495
+ "logits/chosen": -10.829614639282227,
496
+ "logits/rejected": -10.158222198486328,
497
+ "logps/chosen": -30.30867576599121,
498
+ "logps/rejected": -54.475677490234375,
499
+ "loss": 0.3895,
500
+ "rewards/accuracies": 0.84375,
501
+ "rewards/chosen": -0.41875213384628296,
502
+ "rewards/margins": 1.3132567405700684,
503
+ "rewards/rejected": -1.732008695602417,
504
+ "step": 320
505
+ },
506
+ {
507
+ "epoch": 1.3368780858336498,
508
+ "grad_norm": 23.11525326850835,
509
+ "learning_rate": 6.759405075659165e-08,
510
+ "logits/chosen": -10.891769409179688,
511
+ "logits/rejected": -10.190674781799316,
512
+ "logps/chosen": -30.676244735717773,
513
+ "logps/rejected": -58.19831466674805,
514
+ "loss": 0.3621,
515
+ "rewards/accuracies": 0.8687499761581421,
516
+ "rewards/chosen": -0.4654254913330078,
517
+ "rewards/margins": 1.587354302406311,
518
+ "rewards/rejected": -2.0527796745300293,
519
+ "step": 330
520
+ },
521
+ {
522
+ "epoch": 1.377389542980124,
523
+ "grad_norm": 19.61508445315914,
524
+ "learning_rate": 6.536082715796124e-08,
525
+ "logits/chosen": -11.017595291137695,
526
+ "logits/rejected": -10.286384582519531,
527
+ "logps/chosen": -29.53006362915039,
528
+ "logps/rejected": -54.037132263183594,
529
+ "loss": 0.3455,
530
+ "rewards/accuracies": 0.8656250238418579,
531
+ "rewards/chosen": -0.3892853260040283,
532
+ "rewards/margins": 1.4755113124847412,
533
+ "rewards/rejected": -1.8647968769073486,
534
+ "step": 340
535
+ },
536
+ {
537
+ "epoch": 1.4179010001265984,
538
+ "grad_norm": 24.546782869534564,
539
+ "learning_rate": 6.309322424804034e-08,
540
+ "logits/chosen": -10.974300384521484,
541
+ "logits/rejected": -10.224813461303711,
542
+ "logps/chosen": -33.27606201171875,
543
+ "logps/rejected": -58.134117126464844,
544
+ "loss": 0.3673,
545
+ "rewards/accuracies": 0.8500000238418579,
546
+ "rewards/chosen": -0.5354558229446411,
547
+ "rewards/margins": 1.5486235618591309,
548
+ "rewards/rejected": -2.0840792655944824,
549
+ "step": 350
550
+ },
551
+ {
552
+ "epoch": 1.4584124572730726,
553
+ "grad_norm": 20.799419761461635,
554
+ "learning_rate": 6.079631718492568e-08,
555
+ "logits/chosen": -11.121304512023926,
556
+ "logits/rejected": -10.479646682739258,
557
+ "logps/chosen": -30.873401641845703,
558
+ "logps/rejected": -58.193626403808594,
559
+ "loss": 0.3602,
560
+ "rewards/accuracies": 0.875,
561
+ "rewards/chosen": -0.4879949986934662,
562
+ "rewards/margins": 1.5607774257659912,
563
+ "rewards/rejected": -2.0487725734710693,
564
+ "step": 360
565
+ },
566
+ {
567
+ "epoch": 1.4989239144195468,
568
+ "grad_norm": 22.15106589586126,
569
+ "learning_rate": 5.847524671280484e-08,
570
+ "logits/chosen": -11.006221771240234,
571
+ "logits/rejected": -10.338632583618164,
572
+ "logps/chosen": -31.7496395111084,
573
+ "logps/rejected": -56.75434112548828,
574
+ "loss": 0.355,
575
+ "rewards/accuracies": 0.846875011920929,
576
+ "rewards/chosen": -0.5273393392562866,
577
+ "rewards/margins": 1.4817734956741333,
578
+ "rewards/rejected": -2.00911283493042,
579
+ "step": 370
580
+ },
581
+ {
582
+ "epoch": 1.5394353715660212,
583
+ "grad_norm": 27.06131590311837,
584
+ "learning_rate": 5.6135207656374886e-08,
585
+ "logits/chosen": -11.114490509033203,
586
+ "logits/rejected": -10.422750473022461,
587
+ "logps/chosen": -31.69234848022461,
588
+ "logps/rejected": -55.545692443847656,
589
+ "loss": 0.3597,
590
+ "rewards/accuracies": 0.8531249761581421,
591
+ "rewards/chosen": -0.5423402190208435,
592
+ "rewards/margins": 1.4755096435546875,
593
+ "rewards/rejected": -2.017849922180176,
594
+ "step": 380
595
+ },
596
+ {
597
+ "epoch": 1.5799468287124951,
598
+ "grad_norm": 25.089579946286218,
599
+ "learning_rate": 5.3781437294222844e-08,
600
+ "logits/chosen": -11.122465133666992,
601
+ "logits/rejected": -10.4068603515625,
602
+ "logps/chosen": -32.763519287109375,
603
+ "logps/rejected": -58.76951217651367,
604
+ "loss": 0.3421,
605
+ "rewards/accuracies": 0.8500000238418579,
606
+ "rewards/chosen": -0.5654906034469604,
607
+ "rewards/margins": 1.569490671157837,
608
+ "rewards/rejected": -2.134981155395508,
609
+ "step": 390
610
+ },
611
+ {
612
+ "epoch": 1.6204582858589696,
613
+ "grad_norm": 25.42503001610947,
614
+ "learning_rate": 5.141920363718916e-08,
615
+ "logits/chosen": -11.278289794921875,
616
+ "logits/rejected": -10.557046890258789,
617
+ "logps/chosen": -33.54096221923828,
618
+ "logps/rejected": -57.26496505737305,
619
+ "loss": 0.3619,
620
+ "rewards/accuracies": 0.796875,
621
+ "rewards/chosen": -0.638907253742218,
622
+ "rewards/margins": 1.6008431911468506,
623
+ "rewards/rejected": -2.239750385284424,
624
+ "step": 400
625
+ },
626
+ {
627
+ "epoch": 1.6609697430054438,
628
+ "grad_norm": 19.761132187041504,
629
+ "learning_rate": 4.905379363794906e-08,
630
+ "logits/chosen": -11.1757230758667,
631
+ "logits/rejected": -10.361800193786621,
632
+ "logps/chosen": -31.939849853515625,
633
+ "logps/rejected": -58.90076446533203,
634
+ "loss": 0.3557,
635
+ "rewards/accuracies": 0.8500000238418579,
636
+ "rewards/chosen": -0.5620619058609009,
637
+ "rewards/margins": 1.6297252178192139,
638
+ "rewards/rejected": -2.1917872428894043,
639
+ "step": 410
640
+ },
641
+ {
642
+ "epoch": 1.701481200151918,
643
+ "grad_norm": 21.623030847191835,
644
+ "learning_rate": 4.669050135819966e-08,
645
+ "logits/chosen": -11.290115356445312,
646
+ "logits/rejected": -10.585660934448242,
647
+ "logps/chosen": -31.6041316986084,
648
+ "logps/rejected": -54.78083419799805,
649
+ "loss": 0.3602,
650
+ "rewards/accuracies": 0.8374999761581421,
651
+ "rewards/chosen": -0.6143931150436401,
652
+ "rewards/margins": 1.5332918167114258,
653
+ "rewards/rejected": -2.1476848125457764,
654
+ "step": 420
655
+ },
656
+ {
657
+ "epoch": 1.7419926572983924,
658
+ "grad_norm": 24.475959531037525,
659
+ "learning_rate": 4.4334616119936507e-08,
660
+ "logits/chosen": -11.401643753051758,
661
+ "logits/rejected": -10.660685539245605,
662
+ "logps/chosen": -31.317325592041016,
663
+ "logps/rejected": -57.25482940673828,
664
+ "loss": 0.3347,
665
+ "rewards/accuracies": 0.8687499761581421,
666
+ "rewards/chosen": -0.557376503944397,
667
+ "rewards/margins": 1.634526014328003,
668
+ "rewards/rejected": -2.1919026374816895,
669
+ "step": 430
670
+ },
671
+ {
672
+ "epoch": 1.7825041144448663,
673
+ "grad_norm": 19.486205802493238,
674
+ "learning_rate": 4.199141066733789e-08,
675
+ "logits/chosen": -11.200121879577637,
676
+ "logits/rejected": -10.525270462036133,
677
+ "logps/chosen": -29.716289520263672,
678
+ "logps/rejected": -55.837684631347656,
679
+ "loss": 0.3402,
680
+ "rewards/accuracies": 0.8500000238418579,
681
+ "rewards/chosen": -0.5918062925338745,
682
+ "rewards/margins": 1.5346314907073975,
683
+ "rewards/rejected": -2.1264379024505615,
684
+ "step": 440
685
+ },
686
+ {
687
+ "epoch": 1.8230155715913408,
688
+ "grad_norm": 33.35679620452287,
689
+ "learning_rate": 3.966612936575235e-08,
690
+ "logits/chosen": -11.320077896118164,
691
+ "logits/rejected": -10.705958366394043,
692
+ "logps/chosen": -31.9770450592041,
693
+ "logps/rejected": -57.85862350463867,
694
+ "loss": 0.3475,
695
+ "rewards/accuracies": 0.84375,
696
+ "rewards/chosen": -0.6092299818992615,
697
+ "rewards/margins": 1.5435864925384521,
698
+ "rewards/rejected": -2.1528162956237793,
699
+ "step": 450
700
+ },
701
+ {
702
+ "epoch": 1.863527028737815,
703
+ "grad_norm": 31.567941489065486,
704
+ "learning_rate": 3.736397646420134e-08,
705
+ "logits/chosen": -11.576532363891602,
706
+ "logits/rejected": -10.939656257629395,
707
+ "logps/chosen": -33.84613037109375,
708
+ "logps/rejected": -59.848854064941406,
709
+ "loss": 0.3416,
710
+ "rewards/accuracies": 0.871874988079071,
711
+ "rewards/chosen": -0.6238502860069275,
712
+ "rewards/margins": 1.6080996990203857,
713
+ "rewards/rejected": -2.231950283050537,
714
+ "step": 460
715
+ },
716
+ {
717
+ "epoch": 1.9040384858842891,
718
+ "grad_norm": 26.954070410698666,
719
+ "learning_rate": 3.509010444766674e-08,
720
+ "logits/chosen": -11.481412887573242,
721
+ "logits/rejected": -10.886090278625488,
722
+ "logps/chosen": -33.668968200683594,
723
+ "logps/rejected": -61.31293869018555,
724
+ "loss": 0.3405,
725
+ "rewards/accuracies": 0.8374999761581421,
726
+ "rewards/chosen": -0.6743445992469788,
727
+ "rewards/margins": 1.5814064741134644,
728
+ "rewards/rejected": -2.255751132965088,
729
+ "step": 470
730
+ },
731
+ {
732
+ "epoch": 1.9445499430307633,
733
+ "grad_norm": 26.90008640112988,
734
+ "learning_rate": 3.284960250523237e-08,
735
+ "logits/chosen": -11.358150482177734,
736
+ "logits/rejected": -10.692336082458496,
737
+ "logps/chosen": -32.4575309753418,
738
+ "logps/rejected": -58.795753479003906,
739
+ "loss": 0.3619,
740
+ "rewards/accuracies": 0.8812500238418579,
741
+ "rewards/chosen": -0.6383572816848755,
742
+ "rewards/margins": 1.5735241174697876,
743
+ "rewards/rejected": -2.211881160736084,
744
+ "step": 480
745
+ },
746
+ {
747
+ "epoch": 1.9850614001772375,
748
+ "grad_norm": 22.343884141895376,
749
+ "learning_rate": 3.064748513988914e-08,
750
+ "logits/chosen": -11.427927017211914,
751
+ "logits/rejected": -10.837053298950195,
752
+ "logps/chosen": -32.22171401977539,
753
+ "logps/rejected": -59.157142639160156,
754
+ "loss": 0.3589,
755
+ "rewards/accuracies": 0.887499988079071,
756
+ "rewards/chosen": -0.6551386117935181,
757
+ "rewards/margins": 1.6512985229492188,
758
+ "rewards/rejected": -2.3064370155334473,
759
+ "step": 490
760
+ },
761
+ {
762
+ "epoch": 2.025572857323712,
763
+ "grad_norm": 21.328409409312936,
764
+ "learning_rate": 2.8488680945496145e-08,
765
+ "logits/chosen": -11.407200813293457,
766
+ "logits/rejected": -10.8077974319458,
767
+ "logps/chosen": -32.2757453918457,
768
+ "logps/rejected": -58.08913040161133,
769
+ "loss": 0.3406,
770
+ "rewards/accuracies": 0.8500000238418579,
771
+ "rewards/chosen": -0.6142057776451111,
772
+ "rewards/margins": 1.6251834630966187,
773
+ "rewards/rejected": -2.239389419555664,
774
+ "step": 500
775
+ },
776
+ {
777
+ "epoch": 2.066084314470186,
778
+ "grad_norm": 21.92232136811071,
779
+ "learning_rate": 2.6378021576016464e-08,
780
+ "logits/chosen": -11.57470703125,
781
+ "logits/rejected": -10.841165542602539,
782
+ "logps/chosen": -33.64252471923828,
783
+ "logps/rejected": -60.10270309448242,
784
+ "loss": 0.3445,
785
+ "rewards/accuracies": 0.8500000238418579,
786
+ "rewards/chosen": -0.6759551763534546,
787
+ "rewards/margins": 1.7165590524673462,
788
+ "rewards/rejected": -2.3925139904022217,
789
+ "step": 510
790
+ },
791
+ {
792
+ "epoch": 2.1065957716166603,
793
+ "grad_norm": 23.376464913490395,
794
+ "learning_rate": 2.4320230931715697e-08,
795
+ "logits/chosen": -11.309526443481445,
796
+ "logits/rejected": -10.582537651062012,
797
+ "logps/chosen": -33.03705596923828,
798
+ "logps/rejected": -59.86212158203125,
799
+ "loss": 0.3228,
800
+ "rewards/accuracies": 0.871874988079071,
801
+ "rewards/chosen": -0.7096266746520996,
802
+ "rewards/margins": 1.673158884048462,
803
+ "rewards/rejected": -2.3827853202819824,
804
+ "step": 520
805
+ },
806
+ {
807
+ "epoch": 2.1471072287631348,
808
+ "grad_norm": 18.289595804333104,
809
+ "learning_rate": 2.2319914586525774e-08,
810
+ "logits/chosen": -11.554033279418945,
811
+ "logits/rejected": -10.901694297790527,
812
+ "logps/chosen": -34.2962646484375,
813
+ "logps/rejected": -61.11516189575195,
814
+ "loss": 0.3355,
815
+ "rewards/accuracies": 0.859375,
816
+ "rewards/chosen": -0.6974956393241882,
817
+ "rewards/margins": 1.66302490234375,
818
+ "rewards/rejected": -2.360520601272583,
819
+ "step": 530
820
+ },
821
+ {
822
+ "epoch": 2.1876186859096087,
823
+ "grad_norm": 23.20641172076809,
824
+ "learning_rate": 2.038154948023668e-08,
825
+ "logits/chosen": -11.543357849121094,
826
+ "logits/rejected": -10.89212703704834,
827
+ "logps/chosen": -35.47829055786133,
828
+ "logps/rejected": -60.59331130981445,
829
+ "loss": 0.3243,
830
+ "rewards/accuracies": 0.859375,
831
+ "rewards/chosen": -0.7092735171318054,
832
+ "rewards/margins": 1.7081466913223267,
833
+ "rewards/rejected": -2.4174201488494873,
834
+ "step": 540
835
+ },
836
+ {
837
+ "epoch": 2.228130143056083,
838
+ "grad_norm": 25.247728097262844,
839
+ "learning_rate": 1.8509473898586432e-08,
840
+ "logits/chosen": -11.410323143005371,
841
+ "logits/rejected": -10.694147109985352,
842
+ "logps/chosen": -32.15620803833008,
843
+ "logps/rejected": -62.000328063964844,
844
+ "loss": 0.2968,
845
+ "rewards/accuracies": 0.8999999761581421,
846
+ "rewards/chosen": -0.6627341508865356,
847
+ "rewards/margins": 1.8608602285385132,
848
+ "rewards/rejected": -2.523594617843628,
849
+ "step": 550
850
+ },
851
+ {
852
+ "epoch": 2.268641600202557,
853
+ "grad_norm": 24.11504934805817,
854
+ "learning_rate": 1.6707877763674887e-08,
855
+ "logits/chosen": -11.610152244567871,
856
+ "logits/rejected": -11.053690910339355,
857
+ "logps/chosen": -33.62152862548828,
858
+ "logps/rejected": -62.07065963745117,
859
+ "loss": 0.3314,
860
+ "rewards/accuracies": 0.846875011920929,
861
+ "rewards/chosen": -0.7055118083953857,
862
+ "rewards/margins": 1.7053735256195068,
863
+ "rewards/rejected": -2.4108853340148926,
864
+ "step": 560
865
+ },
866
+ {
867
+ "epoch": 2.3091530573490315,
868
+ "grad_norm": 24.563417003466505,
869
+ "learning_rate": 1.4980793256432472e-08,
870
+ "logits/chosen": -11.563300132751465,
871
+ "logits/rejected": -10.760797500610352,
872
+ "logps/chosen": -34.131736755371094,
873
+ "logps/rejected": -62.216285705566406,
874
+ "loss": 0.3008,
875
+ "rewards/accuracies": 0.893750011920929,
876
+ "rewards/chosen": -0.7754543423652649,
877
+ "rewards/margins": 1.8263394832611084,
878
+ "rewards/rejected": -2.6017937660217285,
879
+ "step": 570
880
+ },
881
+ {
882
+ "epoch": 2.349664514495506,
883
+ "grad_norm": 32.15938788887828,
884
+ "learning_rate": 1.3332085792131965e-08,
885
+ "logits/chosen": -11.64229679107666,
886
+ "logits/rejected": -10.900683403015137,
887
+ "logps/chosen": -35.07176971435547,
888
+ "logps/rejected": -61.276954650878906,
889
+ "loss": 0.3318,
890
+ "rewards/accuracies": 0.862500011920929,
891
+ "rewards/chosen": -0.7460302114486694,
892
+ "rewards/margins": 1.772905707359314,
893
+ "rewards/rejected": -2.5189361572265625,
894
+ "step": 580
895
+ },
896
+ {
897
+ "epoch": 2.39017597164198,
898
+ "grad_norm": 30.52452011331439,
899
+ "learning_rate": 1.1765445369141274e-08,
900
+ "logits/chosen": -11.567706108093262,
901
+ "logits/rejected": -10.895853042602539,
902
+ "logps/chosen": -33.69668960571289,
903
+ "logps/rejected": -62.25682830810547,
904
+ "loss": 0.3143,
905
+ "rewards/accuracies": 0.8656250238418579,
906
+ "rewards/chosen": -0.7219974398612976,
907
+ "rewards/margins": 1.7591053247451782,
908
+ "rewards/rejected": -2.481102466583252,
909
+ "step": 590
910
+ },
911
+ {
912
+ "epoch": 2.4306874287884543,
913
+ "grad_norm": 19.90507599157542,
914
+ "learning_rate": 1.0284378310279368e-08,
915
+ "logits/chosen": -11.508050918579102,
916
+ "logits/rejected": -10.74337387084961,
917
+ "logps/chosen": -32.797637939453125,
918
+ "logps/rejected": -63.94108200073242,
919
+ "loss": 0.303,
920
+ "rewards/accuracies": 0.8812500238418579,
921
+ "rewards/chosen": -0.6646064519882202,
922
+ "rewards/margins": 2.0007808208465576,
923
+ "rewards/rejected": -2.6653873920440674,
924
+ "step": 600
925
+ },
926
+ {
927
+ "epoch": 2.4711988859349283,
928
+ "grad_norm": 21.194773980773512,
929
+ "learning_rate": 8.892199415259499e-09,
930
+ "logits/chosen": -11.653422355651855,
931
+ "logits/rejected": -11.00390911102295,
932
+ "logps/chosen": -35.742061614990234,
933
+ "logps/rejected": -65.51417541503906,
934
+ "loss": 0.3164,
935
+ "rewards/accuracies": 0.8812500238418579,
936
+ "rewards/chosen": -0.7325908541679382,
937
+ "rewards/margins": 1.797716498374939,
938
+ "rewards/rejected": -2.5303072929382324,
939
+ "step": 610
940
+ },
941
+ {
942
+ "epoch": 2.5117103430814027,
943
+ "grad_norm": 32.194708817268754,
944
+ "learning_rate": 7.592024541783343e-09,
945
+ "logits/chosen": -11.667333602905273,
946
+ "logits/rejected": -10.975776672363281,
947
+ "logps/chosen": -34.44770812988281,
948
+ "logps/rejected": -59.715660095214844,
949
+ "loss": 0.3359,
950
+ "rewards/accuracies": 0.862500011920929,
951
+ "rewards/chosen": -0.762362003326416,
952
+ "rewards/margins": 1.7372121810913086,
953
+ "rewards/rejected": -2.4995741844177246,
954
+ "step": 620
955
+ },
956
+ {
957
+ "epoch": 2.552221800227877,
958
+ "grad_norm": 22.079251389305,
959
+ "learning_rate": 6.386763631890313e-09,
960
+ "logits/chosen": -11.638243675231934,
961
+ "logits/rejected": -10.932289123535156,
962
+ "logps/chosen": -32.0985107421875,
963
+ "logps/rejected": -59.693931579589844,
964
+ "loss": 0.3138,
965
+ "rewards/accuracies": 0.878125011920929,
966
+ "rewards/chosen": -0.6655905842781067,
967
+ "rewards/margins": 1.7755588293075562,
968
+ "rewards/rejected": -2.4411492347717285,
969
+ "step": 630
970
+ },
971
+ {
972
+ "epoch": 2.592733257374351,
973
+ "grad_norm": 27.178193998967842,
974
+ "learning_rate": 5.279114199170093e-09,
975
+ "logits/chosen": -11.623285293579102,
976
+ "logits/rejected": -10.849055290222168,
977
+ "logps/chosen": -34.02157211303711,
978
+ "logps/rejected": -63.82293701171875,
979
+ "loss": 0.2965,
980
+ "rewards/accuracies": 0.8812500238418579,
981
+ "rewards/chosen": -0.7236124873161316,
982
+ "rewards/margins": 1.9638967514038086,
983
+ "rewards/rejected": -2.687509059906006,
984
+ "step": 640
985
+ },
986
+ {
987
+ "epoch": 2.6332447145208255,
988
+ "grad_norm": 23.237750807734354,
989
+ "learning_rate": 4.271555291414636e-09,
990
+ "logits/chosen": -11.59123420715332,
991
+ "logits/rejected": -10.88310432434082,
992
+ "logps/chosen": -35.09485626220703,
993
+ "logps/rejected": -63.60078811645508,
994
+ "loss": 0.3114,
995
+ "rewards/accuracies": 0.8968750238418579,
996
+ "rewards/chosen": -0.6854656934738159,
997
+ "rewards/margins": 1.8889812231063843,
998
+ "rewards/rejected": -2.5744469165802,
999
+ "step": 650
1000
+ },
1001
+ {
1002
+ "epoch": 2.6737561716672995,
1003
+ "grad_norm": 23.37299143104811,
1004
+ "learning_rate": 3.3663419422218677e-09,
1005
+ "logits/chosen": -11.671598434448242,
1006
+ "logits/rejected": -10.90255355834961,
1007
+ "logps/chosen": -34.43938446044922,
1008
+ "logps/rejected": -63.91756057739258,
1009
+ "loss": 0.295,
1010
+ "rewards/accuracies": 0.887499988079071,
1011
+ "rewards/chosen": -0.690475344657898,
1012
+ "rewards/margins": 2.0302391052246094,
1013
+ "rewards/rejected": -2.7207143306732178,
1014
+ "step": 660
1015
+ },
1016
+ {
1017
+ "epoch": 2.714267628813774,
1018
+ "grad_norm": 21.948999572300334,
1019
+ "learning_rate": 2.565500123969183e-09,
1020
+ "logits/chosen": -11.691490173339844,
1021
+ "logits/rejected": -11.033750534057617,
1022
+ "logps/chosen": -34.22254180908203,
1023
+ "logps/rejected": -60.50248336791992,
1024
+ "loss": 0.3374,
1025
+ "rewards/accuracies": 0.840624988079071,
1026
+ "rewards/chosen": -0.8206373453140259,
1027
+ "rewards/margins": 1.7287733554840088,
1028
+ "rewards/rejected": -2.549410343170166,
1029
+ "step": 670
1030
+ },
1031
+ {
1032
+ "epoch": 2.754779085960248,
1033
+ "grad_norm": 29.12118466657181,
1034
+ "learning_rate": 1.8708222134525163e-09,
1035
+ "logits/chosen": -11.480539321899414,
1036
+ "logits/rejected": -10.761662483215332,
1037
+ "logps/chosen": -34.17140197753906,
1038
+ "logps/rejected": -62.354156494140625,
1039
+ "loss": 0.3059,
1040
+ "rewards/accuracies": 0.887499988079071,
1041
+ "rewards/chosen": -0.7504395246505737,
1042
+ "rewards/margins": 1.792453408241272,
1043
+ "rewards/rejected": -2.5428929328918457,
1044
+ "step": 680
1045
+ },
1046
+ {
1047
+ "epoch": 2.7952905431067223,
1048
+ "grad_norm": 28.127183360242164,
1049
+ "learning_rate": 1.283862980339334e-09,
1050
+ "logits/chosen": -11.640631675720215,
1051
+ "logits/rejected": -11.045839309692383,
1052
+ "logps/chosen": -36.884315490722656,
1053
+ "logps/rejected": -64.67533874511719,
1054
+ "loss": 0.3354,
1055
+ "rewards/accuracies": 0.846875011920929,
1056
+ "rewards/chosen": -0.8064830899238586,
1057
+ "rewards/margins": 1.7638835906982422,
1058
+ "rewards/rejected": -2.570366859436035,
1059
+ "step": 690
1060
+ },
1061
+ {
1062
+ "epoch": 2.8358020002531967,
1063
+ "grad_norm": 27.365840296062363,
1064
+ "learning_rate": 8.059361074139293e-10,
1065
+ "logits/chosen": -11.53732967376709,
1066
+ "logits/rejected": -10.982316970825195,
1067
+ "logps/chosen": -34.8204231262207,
1068
+ "logps/rejected": -60.4800910949707,
1069
+ "loss": 0.3765,
1070
+ "rewards/accuracies": 0.875,
1071
+ "rewards/chosen": -0.77055424451828,
1072
+ "rewards/margins": 1.6154954433441162,
1073
+ "rewards/rejected": -2.386049747467041,
1074
+ "step": 700
1075
+ },
1076
+ {
1077
+ "epoch": 2.8763134573996707,
1078
+ "grad_norm": 23.555963752594412,
1079
+ "learning_rate": 4.381112504031337e-10,
1080
+ "logits/chosen": -11.630104064941406,
1081
+ "logits/rejected": -10.956541061401367,
1082
+ "logps/chosen": -34.54252243041992,
1083
+ "logps/rejected": -60.827423095703125,
1084
+ "loss": 0.3183,
1085
+ "rewards/accuracies": 0.871874988079071,
1086
+ "rewards/chosen": -0.7591537833213806,
1087
+ "rewards/margins": 1.6892755031585693,
1088
+ "rewards/rejected": -2.4484293460845947,
1089
+ "step": 710
1090
+ },
1091
+ {
1092
+ "epoch": 2.916824914546145,
1093
+ "grad_norm": 21.502729719070256,
1094
+ "learning_rate": 1.8121164396283638e-10,
1095
+ "logits/chosen": -11.620569229125977,
1096
+ "logits/rejected": -10.932313919067383,
1097
+ "logps/chosen": -36.305519104003906,
1098
+ "logps/rejected": -66.49055480957031,
1099
+ "loss": 0.3241,
1100
+ "rewards/accuracies": 0.859375,
1101
+ "rewards/chosen": -0.7203236222267151,
1102
+ "rewards/margins": 1.99014413356781,
1103
+ "rewards/rejected": -2.71046781539917,
1104
+ "step": 720
1105
+ },
1106
+ {
1107
+ "epoch": 2.9573363716926195,
1108
+ "grad_norm": 22.58072518400253,
1109
+ "learning_rate": 3.581225918342645e-11,
1110
+ "logits/chosen": -11.6244535446167,
1111
+ "logits/rejected": -10.896235466003418,
1112
+ "logps/chosen": -31.406940460205078,
1113
+ "logps/rejected": -62.654396057128906,
1114
+ "loss": 0.3115,
1115
+ "rewards/accuracies": 0.887499988079071,
1116
+ "rewards/chosen": -0.7061868906021118,
1117
+ "rewards/margins": 1.8940362930297852,
1118
+ "rewards/rejected": -2.6002233028411865,
1119
+ "step": 730
1120
+ },
1121
+ {
1122
+ "epoch": 2.9897455374097985,
1123
+ "step": 738,
1124
+ "total_flos": 0.0,
1125
+ "train_loss": 0.4158425409457871,
1126
+ "train_runtime": 20368.7971,
1127
+ "train_samples_per_second": 2.327,
1128
+ "train_steps_per_second": 0.036
1129
+ }
1130
+ ],
1131
+ "logging_steps": 10,
1132
+ "max_steps": 738,
1133
+ "num_input_tokens_seen": 0,
1134
+ "num_train_epochs": 3,
1135
+ "save_steps": 100,
1136
+ "stateful_callbacks": {
1137
+ "TrainerControl": {
1138
+ "args": {
1139
+ "should_epoch_stop": false,
1140
+ "should_evaluate": false,
1141
+ "should_log": false,
1142
+ "should_save": true,
1143
+ "should_training_stop": true
1144
+ },
1145
+ "attributes": {}
1146
+ }
1147
+ },
1148
+ "total_flos": 0.0,
1149
+ "train_batch_size": 1,
1150
+ "trial_name": null,
1151
+ "trial_params": null
1152
+ }