Hugging Face's logo

EllieS
/
zephyr-dpo-timedial

PEFT
TensorBoard
Safetensors
mistral
alignment-handbook
trl
dpo
Generated from Trainer
Model card Files
xet
Metrics Community
EllieS commited on
Commit
e4d1012
·
verified ·
1 Parent(s): 2cd076b

Model save

Browse files
Files changed (6) hide show
  1. README.md +74 -0
  2. adapter_model.safetensors +1 -1
  3. all_results.json +8 -0
  4. runs/Apr11_12-13-45_612e66badb5c/events.out.tfevents.1712837726.612e66badb5c.4808.0 +2 -2
  5. train_results.json +8 -0
  6. trainer_state.json +497 -0
README.md ADDED
@@ -0,0 +1,74 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ library_name: peft
4
+ tags:
5
+ - trl
6
+ - dpo
7
+ - generated_from_trainer
8
+ base_model: alignment-handbook/zephyr-7b-sft-full
9
+ model-index:
10
+ - name: zephyr-dpo-timedial
11
+ results: []
12
+ ---
13
+
14
+ <!-- This model card has been generated automatically according to the information the Trainer had access to. You
15
+ should probably proofread and complete it, then remove this comment. -->
16
+
17
+ # zephyr-dpo-timedial
18
+
19
+ This model is a fine-tuned version of [alignment-handbook/zephyr-7b-sft-full](https://huggingface.co/alignment-handbook/zephyr-7b-sft-full) on the None dataset.
20
+ It achieves the following results on the evaluation set:
21
+ - Loss: 0.2236
22
+ - Rewards/chosen: 0.2987
23
+ - Rewards/rejected: -1.0958
24
+ - Rewards/accuracies: 1.0
25
+ - Rewards/margins: 1.3944
26
+ - Logps/rejected: -154.5925
27
+ - Logps/chosen: -0.6286
28
+ - Logits/rejected: -2.7419
29
+ - Logits/chosen: -2.7480
30
+
31
+ ## Model description
32
+
33
+ More information needed
34
+
35
+ ## Intended uses & limitations
36
+
37
+ More information needed
38
+
39
+ ## Training and evaluation data
40
+
41
+ More information needed
42
+
43
+ ## Training procedure
44
+
45
+ ### Training hyperparameters
46
+
47
+ The following hyperparameters were used during training:
48
+ - learning_rate: 5e-06
49
+ - train_batch_size: 2
50
+ - eval_batch_size: 2
51
+ - seed: 42
52
+ - distributed_type: multi-GPU
53
+ - gradient_accumulation_steps: 2
54
+ - total_train_batch_size: 4
55
+ - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
56
+ - lr_scheduler_type: cosine
57
+ - lr_scheduler_warmup_ratio: 0.1
58
+ - num_epochs: 1
59
+
60
+ ### Training results
61
+
62
+ | Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
63
+ |:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
64
+ | 0.3995 | 0.35 | 100 | 0.3705 | 0.2856 | -0.5202 | 1.0 | 0.8058 | -97.0354 | -1.9368 | -2.7815 | -2.7803 |
65
+ | 0.2236 | 0.69 | 200 | 0.2236 | 0.2987 | -1.0958 | 1.0 | 1.3944 | -154.5925 | -0.6286 | -2.7419 | -2.7480 |
66
+
67
+
68
+ ### Framework versions
69
+
70
+ - PEFT 0.7.1
71
+ - Transformers 4.39.0.dev0
72
+ - Pytorch 2.1.2+cu121
73
+ - Datasets 2.14.6
74
+ - Tokenizers 0.15.2
adapter_model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:913801ea55b598f7b58aaa007bb675321dd1e837e4dfa66a6569925581eb971d
3
  size 83946192
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3f036b11d8bb4ff3055489120cb1ebff9118f7b36514d8ba27aecd6b2e1c3f6d
3
  size 83946192
all_results.json ADDED
@@ -0,0 +1,8 @@
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 1.0,
3
+ "train_loss": 0.36633619635162884,
4
+ "train_runtime": 628.6241,
5
+ "train_samples": 1157,
6
+ "train_samples_per_second": 1.841,
7
+ "train_steps_per_second": 0.46
8
+ }
runs/Apr11_12-13-45_612e66badb5c/events.out.tfevents.1712837726.612e66badb5c.4808.0 CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:c958333c0e528a248dbd3c6fc69cdd9616580f3bae448122dcd2d1478c1bb160
3
- size 20525
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:80644563065d10ab3cd240ce2956e96847e989a3ae0f7e0ab142e5d5e2e733c1
3
+ size 26383
train_results.json ADDED
@@ -0,0 +1,8 @@
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 1.0,
3
+ "train_loss": 0.36633619635162884,
4
+ "train_runtime": 628.6241,
5
+ "train_samples": 1157,
6
+ "train_samples_per_second": 1.841,
7
+ "train_steps_per_second": 0.46
8
+ }
trainer_state.json ADDED
@@ -0,0 +1,497 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 0.998272884283247,
5
+ "eval_steps": 100,
6
+ "global_step": 289,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.0,
13
+ "grad_norm": 0.314453125,
14
+ "learning_rate": 1.7241379310344828e-07,
15
+ "logits/chosen": -2.648426055908203,
16
+ "logits/rejected": -2.6488823890686035,
17
+ "logps/chosen": -29.924301147460938,
18
+ "logps/rejected": -43.789119720458984,
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": 0.03,
28
+ "grad_norm": 0.294921875,
29
+ "learning_rate": 1.724137931034483e-06,
30
+ "logits/chosen": -2.791017532348633,
31
+ "logits/rejected": -2.792222023010254,
32
+ "logps/chosen": -30.778282165527344,
33
+ "logps/rejected": -44.59426498413086,
34
+ "loss": 0.6929,
35
+ "rewards/accuracies": 0.5,
36
+ "rewards/chosen": 0.0004109032452106476,
37
+ "rewards/margins": 0.0004249625199008733,
38
+ "rewards/rejected": -1.4059260138310492e-05,
39
+ "step": 10
40
+ },
41
+ {
42
+ "epoch": 0.07,
43
+ "grad_norm": 0.3203125,
44
+ "learning_rate": 3.448275862068966e-06,
45
+ "logits/chosen": -2.8013782501220703,
46
+ "logits/rejected": -2.80298113822937,
47
+ "logps/chosen": -30.407978057861328,
48
+ "logps/rejected": -46.99640655517578,
49
+ "loss": 0.6919,
50
+ "rewards/accuracies": 0.824999988079071,
51
+ "rewards/chosen": 0.0021479069255292416,
52
+ "rewards/margins": 0.0025836334098130465,
53
+ "rewards/rejected": -0.00043572625145316124,
54
+ "step": 20
55
+ },
56
+ {
57
+ "epoch": 0.1,
58
+ "grad_norm": 0.30078125,
59
+ "learning_rate": 4.999817502139027e-06,
60
+ "logits/chosen": -2.8514046669006348,
61
+ "logits/rejected": -2.851513385772705,
62
+ "logps/chosen": -29.340972900390625,
63
+ "logps/rejected": -46.38640594482422,
64
+ "loss": 0.6853,
65
+ "rewards/accuracies": 1.0,
66
+ "rewards/chosen": 0.016690988093614578,
67
+ "rewards/margins": 0.01585637591779232,
68
+ "rewards/rejected": 0.0008346118847839534,
69
+ "step": 30
70
+ },
71
+ {
72
+ "epoch": 0.14,
73
+ "grad_norm": 0.34765625,
74
+ "learning_rate": 4.977949980164773e-06,
75
+ "logits/chosen": -2.77081036567688,
76
+ "logits/rejected": -2.7683987617492676,
77
+ "logps/chosen": -24.99424934387207,
78
+ "logps/rejected": -43.76649475097656,
79
+ "loss": 0.6703,
80
+ "rewards/accuracies": 1.0,
81
+ "rewards/chosen": 0.05648866295814514,
82
+ "rewards/margins": 0.04632381349802017,
83
+ "rewards/rejected": 0.010164851322770119,
84
+ "step": 40
85
+ },
86
+ {
87
+ "epoch": 0.17,
88
+ "grad_norm": 0.53125,
89
+ "learning_rate": 4.919948367622307e-06,
90
+ "logits/chosen": -2.826796054840088,
91
+ "logits/rejected": -2.8314404487609863,
92
+ "logps/chosen": -20.64249610900879,
93
+ "logps/rejected": -46.383785247802734,
94
+ "loss": 0.6438,
95
+ "rewards/accuracies": 1.0,
96
+ "rewards/chosen": 0.09788469970226288,
97
+ "rewards/margins": 0.10135942697525024,
98
+ "rewards/rejected": -0.003474730532616377,
99
+ "step": 50
100
+ },
101
+ {
102
+ "epoch": 0.21,
103
+ "grad_norm": 0.57421875,
104
+ "learning_rate": 4.8266584586307555e-06,
105
+ "logits/chosen": -2.835439682006836,
106
+ "logits/rejected": -2.8375213146209717,
107
+ "logps/chosen": -19.745527267456055,
108
+ "logps/rejected": -48.08259582519531,
109
+ "loss": 0.6272,
110
+ "rewards/accuracies": 1.0,
111
+ "rewards/chosen": 0.10895518958568573,
112
+ "rewards/margins": 0.13693755865097046,
113
+ "rewards/rejected": -0.02798236906528473,
114
+ "step": 60
115
+ },
116
+ {
117
+ "epoch": 0.24,
118
+ "grad_norm": 0.80859375,
119
+ "learning_rate": 4.699440630133794e-06,
120
+ "logits/chosen": -2.8537039756774902,
121
+ "logits/rejected": -2.8549795150756836,
122
+ "logps/chosen": -16.767593383789062,
123
+ "logps/rejected": -53.961273193359375,
124
+ "loss": 0.5912,
125
+ "rewards/accuracies": 1.0,
126
+ "rewards/chosen": 0.13634786009788513,
127
+ "rewards/margins": 0.2162630558013916,
128
+ "rewards/rejected": -0.07991518080234528,
129
+ "step": 70
130
+ },
131
+ {
132
+ "epoch": 0.28,
133
+ "grad_norm": 0.7109375,
134
+ "learning_rate": 4.5401500045405126e-06,
135
+ "logits/chosen": -2.7856647968292236,
136
+ "logits/rejected": -2.7866556644439697,
137
+ "logps/chosen": -10.72642993927002,
138
+ "logps/rejected": -63.57000732421875,
139
+ "loss": 0.522,
140
+ "rewards/accuracies": 1.0,
141
+ "rewards/chosen": 0.20102627575397491,
142
+ "rewards/margins": 0.37968239188194275,
143
+ "rewards/rejected": -0.17865613102912903,
144
+ "step": 80
145
+ },
146
+ {
147
+ "epoch": 0.31,
148
+ "grad_norm": 0.73828125,
149
+ "learning_rate": 4.351109397863309e-06,
150
+ "logits/chosen": -2.784008026123047,
151
+ "logits/rejected": -2.7872986793518066,
152
+ "logps/chosen": -4.303569793701172,
153
+ "logps/rejected": -73.54255676269531,
154
+ "loss": 0.4602,
155
+ "rewards/accuracies": 1.0,
156
+ "rewards/chosen": 0.26261404156684875,
157
+ "rewards/margins": 0.5405046343803406,
158
+ "rewards/rejected": -0.277890682220459,
159
+ "step": 90
160
+ },
161
+ {
162
+ "epoch": 0.35,
163
+ "grad_norm": 0.671875,
164
+ "learning_rate": 4.135075447829912e-06,
165
+ "logits/chosen": -2.7918694019317627,
166
+ "logits/rejected": -2.797271490097046,
167
+ "logps/chosen": -2.3328211307525635,
168
+ "logps/rejected": -88.87330627441406,
169
+ "loss": 0.3995,
170
+ "rewards/accuracies": 1.0,
171
+ "rewards/chosen": 0.2833644449710846,
172
+ "rewards/margins": 0.7134484052658081,
173
+ "rewards/rejected": -0.4300839304924011,
174
+ "step": 100
175
+ },
176
+ {
177
+ "epoch": 0.35,
178
+ "eval_logits/chosen": -2.780285358428955,
179
+ "eval_logits/rejected": -2.7814674377441406,
180
+ "eval_logps/chosen": -1.9368014335632324,
181
+ "eval_logps/rejected": -97.03541564941406,
182
+ "eval_loss": 0.3704952299594879,
183
+ "eval_rewards/accuracies": 1.0,
184
+ "eval_rewards/chosen": 0.28558260202407837,
185
+ "eval_rewards/margins": 0.8057959675788879,
186
+ "eval_rewards/rejected": -0.5202134251594543,
187
+ "eval_runtime": 59.0118,
188
+ "eval_samples_per_second": 4.897,
189
+ "eval_steps_per_second": 2.457,
190
+ "step": 100
191
+ },
192
+ {
193
+ "epoch": 0.38,
194
+ "grad_norm": 0.87890625,
195
+ "learning_rate": 3.895198415897896e-06,
196
+ "logits/chosen": -2.757932186126709,
197
+ "logits/rejected": -2.75920033454895,
198
+ "logps/chosen": -1.1944591999053955,
199
+ "logps/rejected": -102.75828552246094,
200
+ "loss": 0.3536,
201
+ "rewards/accuracies": 1.0,
202
+ "rewards/chosen": 0.29440590739250183,
203
+ "rewards/margins": 0.8625706434249878,
204
+ "rewards/rejected": -0.5681648254394531,
205
+ "step": 110
206
+ },
207
+ {
208
+ "epoch": 0.41,
209
+ "grad_norm": 0.5546875,
210
+ "learning_rate": 3.634976249348867e-06,
211
+ "logits/chosen": -2.7818870544433594,
212
+ "logits/rejected": -2.7803432941436768,
213
+ "logps/chosen": -1.4344559907913208,
214
+ "logps/rejected": -116.496337890625,
215
+ "loss": 0.3157,
216
+ "rewards/accuracies": 1.0,
217
+ "rewards/chosen": 0.2936590015888214,
218
+ "rewards/margins": 0.9929488897323608,
219
+ "rewards/rejected": -0.6992899179458618,
220
+ "step": 120
221
+ },
222
+ {
223
+ "epoch": 0.45,
224
+ "grad_norm": 0.52734375,
225
+ "learning_rate": 3.3582035733403963e-06,
226
+ "logits/chosen": -2.784778594970703,
227
+ "logits/rejected": -2.78275465965271,
228
+ "logps/chosen": -1.0116920471191406,
229
+ "logps/rejected": -123.45560455322266,
230
+ "loss": 0.2951,
231
+ "rewards/accuracies": 1.0,
232
+ "rewards/chosen": 0.2951065003871918,
233
+ "rewards/margins": 1.0736355781555176,
234
+ "rewards/rejected": -0.7785290479660034,
235
+ "step": 130
236
+ },
237
+ {
238
+ "epoch": 0.48,
239
+ "grad_norm": 0.47265625,
240
+ "learning_rate": 3.068916356726475e-06,
241
+ "logits/chosen": -2.732774019241333,
242
+ "logits/rejected": -2.730987071990967,
243
+ "logps/chosen": -0.8254967927932739,
244
+ "logps/rejected": -134.45706176757812,
245
+ "loss": 0.2701,
246
+ "rewards/accuracies": 1.0,
247
+ "rewards/chosen": 0.29469770193099976,
248
+ "rewards/margins": 1.1738824844360352,
249
+ "rewards/rejected": -0.8791847229003906,
250
+ "step": 140
251
+ },
252
+ {
253
+ "epoch": 0.52,
254
+ "grad_norm": 0.451171875,
255
+ "learning_rate": 2.771333058543416e-06,
256
+ "logits/chosen": -2.7359633445739746,
257
+ "logits/rejected": -2.7314648628234863,
258
+ "logps/chosen": -0.7030662894248962,
259
+ "logps/rejected": -138.85061645507812,
260
+ "loss": 0.2582,
261
+ "rewards/accuracies": 1.0,
262
+ "rewards/chosen": 0.29896482825279236,
263
+ "rewards/margins": 1.2272465229034424,
264
+ "rewards/rejected": -0.9282817840576172,
265
+ "step": 150
266
+ },
267
+ {
268
+ "epoch": 0.55,
269
+ "grad_norm": 0.37109375,
270
+ "learning_rate": 2.4697931133779566e-06,
271
+ "logits/chosen": -2.760554790496826,
272
+ "logits/rejected": -2.754014492034912,
273
+ "logps/chosen": -4.322839736938477,
274
+ "logps/rejected": -146.63290405273438,
275
+ "loss": 0.2537,
276
+ "rewards/accuracies": 1.0,
277
+ "rewards/chosen": 0.26600345969200134,
278
+ "rewards/margins": 1.265166163444519,
279
+ "rewards/rejected": -0.9991627931594849,
280
+ "step": 160
281
+ },
282
+ {
283
+ "epoch": 0.59,
284
+ "grad_norm": 0.5,
285
+ "learning_rate": 2.168693652639432e-06,
286
+ "logits/chosen": -2.749978542327881,
287
+ "logits/rejected": -2.7451446056365967,
288
+ "logps/chosen": -0.6626571416854858,
289
+ "logps/rejected": -147.86549377441406,
290
+ "loss": 0.2367,
291
+ "rewards/accuracies": 1.0,
292
+ "rewards/chosen": 0.3002447485923767,
293
+ "rewards/margins": 1.324704647064209,
294
+ "rewards/rejected": -1.0244598388671875,
295
+ "step": 170
296
+ },
297
+ {
298
+ "epoch": 0.62,
299
+ "grad_norm": 0.50390625,
300
+ "learning_rate": 1.872425384482346e-06,
301
+ "logits/chosen": -2.7601447105407715,
302
+ "logits/rejected": -2.7531185150146484,
303
+ "logps/chosen": -0.5638567209243774,
304
+ "logps/rejected": -151.16998291015625,
305
+ "loss": 0.2307,
306
+ "rewards/accuracies": 1.0,
307
+ "rewards/chosen": 0.30341148376464844,
308
+ "rewards/margins": 1.3561573028564453,
309
+ "rewards/rejected": -1.0527456998825073,
310
+ "step": 180
311
+ },
312
+ {
313
+ "epoch": 0.66,
314
+ "grad_norm": 3.484375,
315
+ "learning_rate": 1.5853085673944695e-06,
316
+ "logits/chosen": -2.7549314498901367,
317
+ "logits/rejected": -2.748093366622925,
318
+ "logps/chosen": -1.0212732553482056,
319
+ "logps/rejected": -151.91207885742188,
320
+ "loss": 0.2298,
321
+ "rewards/accuracies": 1.0,
322
+ "rewards/chosen": 0.2937738597393036,
323
+ "rewards/margins": 1.36028254032135,
324
+ "rewards/rejected": -1.0665085315704346,
325
+ "step": 190
326
+ },
327
+ {
328
+ "epoch": 0.69,
329
+ "grad_norm": 0.37109375,
330
+ "learning_rate": 1.3115300110997097e-06,
331
+ "logits/chosen": -2.73872447013855,
332
+ "logits/rejected": -2.734360933303833,
333
+ "logps/chosen": -0.8876104354858398,
334
+ "logps/rejected": -154.93753051757812,
335
+ "loss": 0.2236,
336
+ "rewards/accuracies": 1.0,
337
+ "rewards/chosen": 0.2938212454319,
338
+ "rewards/margins": 1.3927398920059204,
339
+ "rewards/rejected": -1.0989186763763428,
340
+ "step": 200
341
+ },
342
+ {
343
+ "epoch": 0.69,
344
+ "eval_logits/chosen": -2.7480151653289795,
345
+ "eval_logits/rejected": -2.741860866546631,
346
+ "eval_logps/chosen": -0.6285810470581055,
347
+ "eval_logps/rejected": -154.59251403808594,
348
+ "eval_loss": 0.22360171377658844,
349
+ "eval_rewards/accuracies": 1.0,
350
+ "eval_rewards/chosen": 0.29866480827331543,
351
+ "eval_rewards/margins": 1.3944491147994995,
352
+ "eval_rewards/rejected": -1.095784306526184,
353
+ "eval_runtime": 58.9766,
354
+ "eval_samples_per_second": 4.9,
355
+ "eval_steps_per_second": 2.459,
356
+ "step": 200
357
+ },
358
+ {
359
+ "epoch": 0.73,
360
+ "grad_norm": 0.56640625,
361
+ "learning_rate": 1.0550820234444627e-06,
362
+ "logits/chosen": -2.7622146606445312,
363
+ "logits/rejected": -2.7575910091400146,
364
+ "logps/chosen": -0.2955681085586548,
365
+ "logps/rejected": -156.5863494873047,
366
+ "loss": 0.2204,
367
+ "rewards/accuracies": 1.0,
368
+ "rewards/chosen": 0.3056582808494568,
369
+ "rewards/margins": 1.4064788818359375,
370
+ "rewards/rejected": -1.1008204221725464,
371
+ "step": 210
372
+ },
373
+ {
374
+ "epoch": 0.76,
375
+ "grad_norm": 1.3125,
376
+ "learning_rate": 8.197041935593181e-07,
377
+ "logits/chosen": -2.7583320140838623,
378
+ "logits/rejected": -2.7535462379455566,
379
+ "logps/chosen": -0.26544955372810364,
380
+ "logps/rejected": -156.95286560058594,
381
+ "loss": 0.2193,
382
+ "rewards/accuracies": 1.0,
383
+ "rewards/chosen": 0.30188480019569397,
384
+ "rewards/margins": 1.4142131805419922,
385
+ "rewards/rejected": -1.112328290939331,
386
+ "step": 220
387
+ },
388
+ {
389
+ "epoch": 0.79,
390
+ "grad_norm": 0.373046875,
391
+ "learning_rate": 6.088288602287159e-07,
392
+ "logits/chosen": -2.723353385925293,
393
+ "logits/rejected": -2.71772837638855,
394
+ "logps/chosen": -0.17958000302314758,
395
+ "logps/rejected": -157.92813110351562,
396
+ "loss": 0.2166,
397
+ "rewards/accuracies": 1.0,
398
+ "rewards/chosen": 0.3015274405479431,
399
+ "rewards/margins": 1.4261300563812256,
400
+ "rewards/rejected": -1.1246027946472168,
401
+ "step": 230
402
+ },
403
+ {
404
+ "epoch": 0.83,
405
+ "grad_norm": 0.6328125,
406
+ "learning_rate": 4.255310606625124e-07,
407
+ "logits/chosen": -2.762199640274048,
408
+ "logits/rejected": -2.758517265319824,
409
+ "logps/chosen": -0.2033136636018753,
410
+ "logps/rejected": -158.05380249023438,
411
+ "loss": 0.2164,
412
+ "rewards/accuracies": 1.0,
413
+ "rewards/chosen": 0.3013545870780945,
414
+ "rewards/margins": 1.4260454177856445,
415
+ "rewards/rejected": -1.1246907711029053,
416
+ "step": 240
417
+ },
418
+ {
419
+ "epoch": 0.86,
420
+ "grad_norm": 0.478515625,
421
+ "learning_rate": 2.7248368952908055e-07,
422
+ "logits/chosen": -2.7318742275238037,
423
+ "logits/rejected": -2.724575996398926,
424
+ "logps/chosen": -0.2196466624736786,
425
+ "logps/rejected": -155.65301513671875,
426
+ "loss": 0.2205,
427
+ "rewards/accuracies": 1.0,
428
+ "rewards/chosen": 0.3047715723514557,
429
+ "rewards/margins": 1.406855583190918,
430
+ "rewards/rejected": -1.1020839214324951,
431
+ "step": 250
432
+ },
433
+ {
434
+ "epoch": 0.9,
435
+ "grad_norm": 0.4453125,
436
+ "learning_rate": 1.5191852213221198e-07,
437
+ "logits/chosen": -2.7768027782440186,
438
+ "logits/rejected": -2.7685720920562744,
439
+ "logps/chosen": -1.5333685874938965,
440
+ "logps/rejected": -156.28958129882812,
441
+ "loss": 0.2237,
442
+ "rewards/accuracies": 1.0,
443
+ "rewards/chosen": 0.29162219166755676,
444
+ "rewards/margins": 1.3938542604446411,
445
+ "rewards/rejected": -1.1022322177886963,
446
+ "step": 260
447
+ },
448
+ {
449
+ "epoch": 0.93,
450
+ "grad_norm": 0.345703125,
451
+ "learning_rate": 6.559367010166629e-08,
452
+ "logits/chosen": -2.699760675430298,
453
+ "logits/rejected": -2.696230411529541,
454
+ "logps/chosen": -0.23872259259223938,
455
+ "logps/rejected": -159.244140625,
456
+ "loss": 0.2139,
457
+ "rewards/accuracies": 1.0,
458
+ "rewards/chosen": 0.3005048334598541,
459
+ "rewards/margins": 1.4419862031936646,
460
+ "rewards/rejected": -1.1414815187454224,
461
+ "step": 270
462
+ },
463
+ {
464
+ "epoch": 0.97,
465
+ "grad_norm": 0.400390625,
466
+ "learning_rate": 1.4767944166687032e-08,
467
+ "logits/chosen": -2.699716091156006,
468
+ "logits/rejected": -2.6911072731018066,
469
+ "logps/chosen": -0.23203042149543762,
470
+ "logps/rejected": -159.73312377929688,
471
+ "loss": 0.2128,
472
+ "rewards/accuracies": 1.0,
473
+ "rewards/chosen": 0.3060615360736847,
474
+ "rewards/margins": 1.44607675075531,
475
+ "rewards/rejected": -1.1400153636932373,
476
+ "step": 280
477
+ },
478
+ {
479
+ "epoch": 1.0,
480
+ "step": 289,
481
+ "total_flos": 0.0,
482
+ "train_loss": 0.36633619635162884,
483
+ "train_runtime": 628.6241,
484
+ "train_samples_per_second": 1.841,
485
+ "train_steps_per_second": 0.46
486
+ }
487
+ ],
488
+ "logging_steps": 10,
489
+ "max_steps": 289,
490
+ "num_input_tokens_seen": 0,
491
+ "num_train_epochs": 1,
492
+ "save_steps": 100,
493
+ "total_flos": 0.0,
494
+ "train_batch_size": 2,
495
+ "trial_name": null,
496
+ "trial_params": null
497
+ }