Hugging Face's logo

joshuasundance
/
phi3-mini-4k-qlora-python-code-20k-mypo-4k-rfc-full

Text Generation
Transformers
Safetensors
English
phi3
python
dpo
mypo
conversational
custom_code
text-generation-inference
Inference Endpoints
4-bit precision
bitsandbytes
Model card Files Community
joshuasundance commited on
Commit
b91f3cc
·
verified ·
1 Parent(s): 6a0fc1e

Upload trainer_state.json

Browse files
Files changed (1) hide show
  1. trainer_state.json +273 -0
trainer_state.json ADDED
@@ -0,0 +1,273 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 1.0,
5
+ "eval_steps": 500,
6
+ "global_step": 800,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.0625,
13
+ "grad_norm": 4.144060134887695,
14
+ "learning_rate": 3.1249999999999997e-07,
15
+ "logits/chosen": 3.1186840534210205,
16
+ "logits/rejected": 3.2627792358398438,
17
+ "logps/chosen": -54.367942810058594,
18
+ "logps/rejected": -41.90058517456055,
19
+ "loss": 0.6909,
20
+ "rewards/accuracies": 0.5649999976158142,
21
+ "rewards/chosen": 0.004131872206926346,
22
+ "rewards/margins": 0.004790368489921093,
23
+ "rewards/rejected": -0.0006584958755411208,
24
+ "step": 50
25
+ },
26
+ {
27
+ "epoch": 0.125,
28
+ "grad_norm": 2.6885554790496826,
29
+ "learning_rate": 4.861111111111111e-07,
30
+ "logits/chosen": 3.1031572818756104,
31
+ "logits/rejected": 3.2696139812469482,
32
+ "logps/chosen": -54.12215805053711,
33
+ "logps/rejected": -40.67683029174805,
34
+ "loss": 0.6682,
35
+ "rewards/accuracies": 0.875,
36
+ "rewards/chosen": 0.03350482136011124,
37
+ "rewards/margins": 0.05115475505590439,
38
+ "rewards/rejected": -0.01764993742108345,
39
+ "step": 100
40
+ },
41
+ {
42
+ "epoch": 0.1875,
43
+ "grad_norm": 2.406733751296997,
44
+ "learning_rate": 4.513888888888889e-07,
45
+ "logits/chosen": 3.1083221435546875,
46
+ "logits/rejected": 3.2473440170288086,
47
+ "logps/chosen": -53.839054107666016,
48
+ "logps/rejected": -43.20619583129883,
49
+ "loss": 0.6225,
50
+ "rewards/accuracies": 0.9900000095367432,
51
+ "rewards/chosen": 0.08696311712265015,
52
+ "rewards/margins": 0.1482781618833542,
53
+ "rewards/rejected": -0.06131504476070404,
54
+ "step": 150
55
+ },
56
+ {
57
+ "epoch": 0.25,
58
+ "grad_norm": 2.196671724319458,
59
+ "learning_rate": 4.1666666666666667e-07,
60
+ "logits/chosen": 3.114607810974121,
61
+ "logits/rejected": 3.270479440689087,
62
+ "logps/chosen": -49.034297943115234,
63
+ "logps/rejected": -39.99443817138672,
64
+ "loss": 0.5765,
65
+ "rewards/accuracies": 0.9900000095367432,
66
+ "rewards/chosen": 0.15706641972064972,
67
+ "rewards/margins": 0.2536722421646118,
68
+ "rewards/rejected": -0.09660584479570389,
69
+ "step": 200
70
+ },
71
+ {
72
+ "epoch": 0.3125,
73
+ "grad_norm": 2.1391797065734863,
74
+ "learning_rate": 3.819444444444444e-07,
75
+ "logits/chosen": 3.0528652667999268,
76
+ "logits/rejected": 3.1882102489471436,
77
+ "logps/chosen": -50.195011138916016,
78
+ "logps/rejected": -42.24689865112305,
79
+ "loss": 0.5097,
80
+ "rewards/accuracies": 0.9950000047683716,
81
+ "rewards/chosen": 0.25386032462120056,
82
+ "rewards/margins": 0.4180786609649658,
83
+ "rewards/rejected": -0.16421833634376526,
84
+ "step": 250
85
+ },
86
+ {
87
+ "epoch": 0.375,
88
+ "grad_norm": 2.2767980098724365,
89
+ "learning_rate": 3.472222222222222e-07,
90
+ "logits/chosen": 2.9493257999420166,
91
+ "logits/rejected": 3.099841833114624,
92
+ "logps/chosen": -49.544803619384766,
93
+ "logps/rejected": -41.29837417602539,
94
+ "loss": 0.4578,
95
+ "rewards/accuracies": 1.0,
96
+ "rewards/chosen": 0.2961386442184448,
97
+ "rewards/margins": 0.5633676648139954,
98
+ "rewards/rejected": -0.2672290802001953,
99
+ "step": 300
100
+ },
101
+ {
102
+ "epoch": 0.4375,
103
+ "grad_norm": 1.8134651184082031,
104
+ "learning_rate": 3.1249999999999997e-07,
105
+ "logits/chosen": 2.9306814670562744,
106
+ "logits/rejected": 3.1076226234436035,
107
+ "logps/chosen": -49.26047897338867,
108
+ "logps/rejected": -43.91292953491211,
109
+ "loss": 0.419,
110
+ "rewards/accuracies": 1.0,
111
+ "rewards/chosen": 0.3598053753376007,
112
+ "rewards/margins": 0.677994966506958,
113
+ "rewards/rejected": -0.3181896507740021,
114
+ "step": 350
115
+ },
116
+ {
117
+ "epoch": 0.5,
118
+ "grad_norm": 2.273711919784546,
119
+ "learning_rate": 2.7777777777777776e-07,
120
+ "logits/chosen": 2.9157378673553467,
121
+ "logits/rejected": 3.065505266189575,
122
+ "logps/chosen": -47.57904815673828,
123
+ "logps/rejected": -43.553749084472656,
124
+ "loss": 0.3937,
125
+ "rewards/accuracies": 1.0,
126
+ "rewards/chosen": 0.33699238300323486,
127
+ "rewards/margins": 0.7744642496109009,
128
+ "rewards/rejected": -0.4374719560146332,
129
+ "step": 400
130
+ },
131
+ {
132
+ "epoch": 0.5625,
133
+ "grad_norm": 1.6072301864624023,
134
+ "learning_rate": 2.4305555555555555e-07,
135
+ "logits/chosen": 2.9038708209991455,
136
+ "logits/rejected": 3.0539231300354004,
137
+ "logps/chosen": -47.44837951660156,
138
+ "logps/rejected": -45.976158142089844,
139
+ "loss": 0.3513,
140
+ "rewards/accuracies": 1.0,
141
+ "rewards/chosen": 0.38920170068740845,
142
+ "rewards/margins": 0.9139933586120605,
143
+ "rewards/rejected": -0.5247916579246521,
144
+ "step": 450
145
+ },
146
+ {
147
+ "epoch": 0.625,
148
+ "grad_norm": 2.385540008544922,
149
+ "learning_rate": 2.0833333333333333e-07,
150
+ "logits/chosen": 2.953505754470825,
151
+ "logits/rejected": 3.1365463733673096,
152
+ "logps/chosen": -50.15205001831055,
153
+ "logps/rejected": -49.16044998168945,
154
+ "loss": 0.3337,
155
+ "rewards/accuracies": 1.0,
156
+ "rewards/chosen": 0.3933267295360565,
157
+ "rewards/margins": 0.9976506233215332,
158
+ "rewards/rejected": -0.6043238639831543,
159
+ "step": 500
160
+ },
161
+ {
162
+ "epoch": 0.6875,
163
+ "grad_norm": 2.3325698375701904,
164
+ "learning_rate": 1.736111111111111e-07,
165
+ "logits/chosen": 2.9089577198028564,
166
+ "logits/rejected": 3.068243980407715,
167
+ "logps/chosen": -48.45305633544922,
168
+ "logps/rejected": -48.769771575927734,
169
+ "loss": 0.2935,
170
+ "rewards/accuracies": 1.0,
171
+ "rewards/chosen": 0.41794300079345703,
172
+ "rewards/margins": 1.1487656831741333,
173
+ "rewards/rejected": -0.7308226227760315,
174
+ "step": 550
175
+ },
176
+ {
177
+ "epoch": 0.75,
178
+ "grad_norm": 2.3242642879486084,
179
+ "learning_rate": 1.3888888888888888e-07,
180
+ "logits/chosen": 2.916025400161743,
181
+ "logits/rejected": 3.075576066970825,
182
+ "logps/chosen": -48.39594268798828,
183
+ "logps/rejected": -47.85335922241211,
184
+ "loss": 0.2915,
185
+ "rewards/accuracies": 1.0,
186
+ "rewards/chosen": 0.4126347303390503,
187
+ "rewards/margins": 1.1786757707595825,
188
+ "rewards/rejected": -0.7660410404205322,
189
+ "step": 600
190
+ },
191
+ {
192
+ "epoch": 0.8125,
193
+ "grad_norm": 1.8319804668426514,
194
+ "learning_rate": 1.0416666666666667e-07,
195
+ "logits/chosen": 2.8800783157348633,
196
+ "logits/rejected": 3.0461764335632324,
197
+ "logps/chosen": -49.678184509277344,
198
+ "logps/rejected": -50.534385681152344,
199
+ "loss": 0.2617,
200
+ "rewards/accuracies": 1.0,
201
+ "rewards/chosen": 0.3970913290977478,
202
+ "rewards/margins": 1.3191609382629395,
203
+ "rewards/rejected": -0.9220696091651917,
204
+ "step": 650
205
+ },
206
+ {
207
+ "epoch": 0.875,
208
+ "grad_norm": 1.8796759843826294,
209
+ "learning_rate": 6.944444444444444e-08,
210
+ "logits/chosen": 2.8562612533569336,
211
+ "logits/rejected": 3.0434064865112305,
212
+ "logps/chosen": -50.938533782958984,
213
+ "logps/rejected": -51.9145622253418,
214
+ "loss": 0.2487,
215
+ "rewards/accuracies": 0.9950000047683716,
216
+ "rewards/chosen": 0.44071799516677856,
217
+ "rewards/margins": 1.3961044549942017,
218
+ "rewards/rejected": -0.9553865194320679,
219
+ "step": 700
220
+ },
221
+ {
222
+ "epoch": 0.9375,
223
+ "grad_norm": 1.6690614223480225,
224
+ "learning_rate": 3.472222222222222e-08,
225
+ "logits/chosen": 2.904910087585449,
226
+ "logits/rejected": 3.0448944568634033,
227
+ "logps/chosen": -50.198997497558594,
228
+ "logps/rejected": -52.53700637817383,
229
+ "loss": 0.2723,
230
+ "rewards/accuracies": 1.0,
231
+ "rewards/chosen": 0.3550363779067993,
232
+ "rewards/margins": 1.2905099391937256,
233
+ "rewards/rejected": -0.9354735612869263,
234
+ "step": 750
235
+ },
236
+ {
237
+ "epoch": 1.0,
238
+ "grad_norm": 1.8712382316589355,
239
+ "learning_rate": 0.0,
240
+ "logits/chosen": 2.856696128845215,
241
+ "logits/rejected": 3.0441014766693115,
242
+ "logps/chosen": -49.916290283203125,
243
+ "logps/rejected": -52.326080322265625,
244
+ "loss": 0.2438,
245
+ "rewards/accuracies": 1.0,
246
+ "rewards/chosen": 0.40912488102912903,
247
+ "rewards/margins": 1.3978087902069092,
248
+ "rewards/rejected": -0.9886838793754578,
249
+ "step": 800
250
+ }
251
+ ],
252
+ "logging_steps": 50,
253
+ "max_steps": 800,
254
+ "num_input_tokens_seen": 0,
255
+ "num_train_epochs": 1,
256
+ "save_steps": 500,
257
+ "stateful_callbacks": {
258
+ "TrainerControl": {
259
+ "args": {
260
+ "should_epoch_stop": false,
261
+ "should_evaluate": false,
262
+ "should_log": false,
263
+ "should_save": true,
264
+ "should_training_stop": true
265
+ },
266
+ "attributes": {}
267
+ }
268
+ },
269
+ "total_flos": 0.0,
270
+ "train_batch_size": 4,
271
+ "trial_name": null,
272
+ "trial_params": null
273
+ }