AmberYifan commited on
Commit
b4daa5f
1 Parent(s): 3a93399

Model save

Browse files
README.md CHANGED
@@ -2,10 +2,7 @@
2
  license: mit
3
  base_model: microsoft/Phi-3-small-8k-instruct
4
  tags:
5
- - alignment-handbook
6
  - generated_from_trainer
7
- datasets:
8
- - AmberYifan/spin-v
9
  model-index:
10
  - name: phi3-spin-zephyr-data
11
  results: []
@@ -16,15 +13,15 @@ should probably proofread and complete it, then remove this comment. -->
16
 
17
  # phi3-spin-zephyr-data
18
 
19
- This model is a fine-tuned version of [microsoft/Phi-3-small-8k-instruct](https://huggingface.co/microsoft/Phi-3-small-8k-instruct) on the AmberYifan/spin-v dataset.
20
  It achieves the following results on the evaluation set:
21
- - Loss: 0.1680
22
- - Rewards/real: -6.9251
23
- - Rewards/generated: -22.7311
24
  - Rewards/accuracies: 0.9443
25
- - Rewards/margins: 15.8060
26
- - Logps/generated: -518.7460
27
- - Logps/real: -321.6379
28
  - Logits/generated: -inf
29
  - Logits/real: -inf
30
 
@@ -62,7 +59,7 @@ The following hyperparameters were used during training:
62
 
63
  | Training Loss | Epoch | Step | Validation Loss | Rewards/real | Rewards/generated | Rewards/accuracies | Rewards/margins | Logps/generated | Logps/real | Logits/generated | Logits/real |
64
  |:-------------:|:-----:|:----:|:---------------:|:------------:|:-----------------:|:------------------:|:---------------:|:---------------:|:----------:|:----------------:|:-----------:|
65
- | 0.2807 | 0.64 | 500 | 0.1680 | -6.9251 | -22.7311 | 0.9443 | 15.8060 | -518.7460 | -321.6379 | -inf | -inf |
66
 
67
 
68
  ### Framework versions
 
2
  license: mit
3
  base_model: microsoft/Phi-3-small-8k-instruct
4
  tags:
 
5
  - generated_from_trainer
 
 
6
  model-index:
7
  - name: phi3-spin-zephyr-data
8
  results: []
 
13
 
14
  # phi3-spin-zephyr-data
15
 
16
+ This model is a fine-tuned version of [microsoft/Phi-3-small-8k-instruct](https://huggingface.co/microsoft/Phi-3-small-8k-instruct) on the None dataset.
17
  It achieves the following results on the evaluation set:
18
+ - Loss: 0.1748
19
+ - Rewards/real: -6.0641
20
+ - Rewards/generated: -21.7894
21
  - Rewards/accuracies: 0.9443
22
+ - Rewards/margins: 15.7252
23
+ - Logps/generated: -509.3286
24
+ - Logps/real: -313.0280
25
  - Logits/generated: -inf
26
  - Logits/real: -inf
27
 
 
59
 
60
  | Training Loss | Epoch | Step | Validation Loss | Rewards/real | Rewards/generated | Rewards/accuracies | Rewards/margins | Logps/generated | Logps/real | Logits/generated | Logits/real |
61
  |:-------------:|:-----:|:----:|:---------------:|:------------:|:-----------------:|:------------------:|:---------------:|:---------------:|:----------:|:----------------:|:-----------:|
62
+ | 0.2945 | 0.64 | 500 | 0.1748 | -6.0641 | -21.7894 | 0.9443 | 15.7252 | -509.3286 | -313.0280 | -inf | -inf |
63
 
64
 
65
  ### Framework versions
all_results.json CHANGED
@@ -1,8 +1,8 @@
1
  {
2
  "epoch": 1.0,
3
- "train_loss": 0.6284302195911938,
4
- "train_runtime": 5860.6637,
5
  "train_samples": 25000,
6
- "train_samples_per_second": 4.266,
7
- "train_steps_per_second": 0.133
8
  }
 
1
  {
2
  "epoch": 1.0,
3
+ "train_loss": 0.6284351689954493,
4
+ "train_runtime": 6560.8804,
5
  "train_samples": 25000,
6
+ "train_samples_per_second": 3.81,
7
+ "train_steps_per_second": 0.119
8
  }
model-00001-of-00004.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:e61a098b9291287a13df8ae4fcdb00643a5ec324bf65af655801e5cb47a21618
3
  size 4832943104
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1efa10e64d5859b0237937a454e8d0e674af6517f71086f2e7e98bc3147fe71c
3
  size 4832943104
model-00002-of-00004.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:1b112110dccae5c1237da32a3ef6bd78b31a00f092836ea5d6725a775764041d
3
  size 4799608224
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1ac4af61d8a50b24b2a0451b0d0d8d681f445298adb99946728417f8ea8a5f8c
3
  size 4799608224
model-00003-of-00004.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:6b8607487f2f6919b39039edda7f2d6f3a681042b5433ed53b46c2b0e1484738
3
  size 4799608240
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4cbc2426e9fb18df7c282f39a9fd4f5523abf48457b84c80ecdc7b84bf172b63
3
  size 4799608240
model-00004-of-00004.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:713883d1a11f48447b2142edf25e4ea6b77b0ed0851ce9d404c3ca3b854a8975
3
  size 352437304
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:856cab213b70e9a43d567a410c8964012fdc3434091381c99187734430660556
3
  size 352437304
runs/Jul24_14-12-34_gilbreth-j001.rcac.purdue.edu/events.out.tfevents.1721844898.gilbreth-j001.rcac.purdue.edu.5513.0 CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:a9c9931847440eb80ad54887342cd5c41ed40869da6dc121426354a790bd2778
3
- size 52207
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0ff6e5db5f8fbe5faa8099ace021ff1818e37a6f9aa6825947a27551f6fd0333
3
+ size 57609
train_results.json CHANGED
@@ -1,8 +1,8 @@
1
  {
2
  "epoch": 1.0,
3
- "train_loss": 0.6284302195911938,
4
- "train_runtime": 5860.6637,
5
  "train_samples": 25000,
6
- "train_samples_per_second": 4.266,
7
- "train_steps_per_second": 0.133
8
  }
 
1
  {
2
  "epoch": 1.0,
3
+ "train_loss": 0.6284351689954493,
4
+ "train_runtime": 6560.8804,
5
  "train_samples": 25000,
6
+ "train_samples_per_second": 3.81,
7
+ "train_steps_per_second": 0.119
8
  }
trainer_state.json CHANGED
@@ -27,13 +27,13 @@
27
  "learning_rate": 6.329113924050633e-08,
28
  "logits/generated": -Infinity,
29
  "logits/real": -Infinity,
30
- "logps/generated": -487.2403869628906,
31
- "logps/real": -384.84637451171875,
32
- "loss": 3.4512,
33
- "rewards/accuracies": 0.6666666865348816,
34
- "rewards/generated": -19.510425567626953,
35
- "rewards/margins": 6.074762344360352,
36
- "rewards/real": -13.435664176940918,
37
  "step": 10
38
  },
39
  {
@@ -41,13 +41,13 @@
41
  "learning_rate": 1.2658227848101266e-07,
42
  "logits/generated": -Infinity,
43
  "logits/real": -Infinity,
44
- "logps/generated": -473.37353515625,
45
- "logps/real": -413.13916015625,
46
- "loss": 3.9704,
47
  "rewards/accuracies": 0.637499988079071,
48
- "rewards/generated": -18.36065101623535,
49
- "rewards/margins": 3.8984241485595703,
50
- "rewards/real": -14.462226867675781,
51
  "step": 20
52
  },
53
  {
@@ -55,13 +55,13 @@
55
  "learning_rate": 1.89873417721519e-07,
56
  "logits/generated": -Infinity,
57
  "logits/real": -Infinity,
58
- "logps/generated": -510.23455810546875,
59
- "logps/real": -408.78350830078125,
60
- "loss": 3.6848,
61
  "rewards/accuracies": 0.6875,
62
- "rewards/generated": -21.037378311157227,
63
- "rewards/margins": 7.7357025146484375,
64
- "rewards/real": -13.301675796508789,
65
  "step": 30
66
  },
67
  {
@@ -69,13 +69,13 @@
69
  "learning_rate": 2.5316455696202533e-07,
70
  "logits/generated": -Infinity,
71
  "logits/real": -Infinity,
72
- "logps/generated": -451.669189453125,
73
- "logps/real": -368.1625061035156,
74
- "loss": 3.9079,
75
  "rewards/accuracies": 0.625,
76
- "rewards/generated": -16.798627853393555,
77
- "rewards/margins": 4.454809665679932,
78
- "rewards/real": -12.343819618225098,
79
  "step": 40
80
  },
81
  {
@@ -83,13 +83,13 @@
83
  "learning_rate": 3.1645569620253163e-07,
84
  "logits/generated": -Infinity,
85
  "logits/real": -Infinity,
86
- "logps/generated": -552.6080322265625,
87
- "logps/real": -397.91064453125,
88
- "loss": 2.9787,
89
  "rewards/accuracies": 0.737500011920929,
90
- "rewards/generated": -25.66506004333496,
91
- "rewards/margins": 10.62718391418457,
92
- "rewards/real": -15.037874221801758,
93
  "step": 50
94
  },
95
  {
@@ -97,13 +97,13 @@
97
  "learning_rate": 3.79746835443038e-07,
98
  "logits/generated": -Infinity,
99
  "logits/real": -Infinity,
100
- "logps/generated": -509.36578369140625,
101
- "logps/real": -390.8323059082031,
102
- "loss": 2.5737,
103
  "rewards/accuracies": 0.737500011920929,
104
- "rewards/generated": -22.55277442932129,
105
- "rewards/margins": 7.870760440826416,
106
- "rewards/real": -14.682014465332031,
107
  "step": 60
108
  },
109
  {
@@ -111,13 +111,13 @@
111
  "learning_rate": 4.4303797468354424e-07,
112
  "logits/generated": -Infinity,
113
  "logits/real": -Infinity,
114
- "logps/generated": -554.7327270507812,
115
- "logps/real": -382.78350830078125,
116
- "loss": 1.7978,
117
  "rewards/accuracies": 0.8374999761581421,
118
- "rewards/generated": -28.232168197631836,
119
- "rewards/margins": 14.969076156616211,
120
- "rewards/real": -13.263090133666992,
121
  "step": 70
122
  },
123
  {
@@ -125,13 +125,13 @@
125
  "learning_rate": 4.992887624466572e-07,
126
  "logits/generated": -Infinity,
127
  "logits/real": -Infinity,
128
- "logps/generated": -573.7469482421875,
129
- "logps/real": -402.5927429199219,
130
- "loss": 1.4178,
131
  "rewards/accuracies": 0.800000011920929,
132
- "rewards/generated": -28.089534759521484,
133
- "rewards/margins": 12.5045166015625,
134
- "rewards/real": -15.5850191116333,
135
  "step": 80
136
  },
137
  {
@@ -139,13 +139,13 @@
139
  "learning_rate": 4.92176386913229e-07,
140
  "logits/generated": -Infinity,
141
  "logits/real": -Infinity,
142
- "logps/generated": -576.7484741210938,
143
- "logps/real": -400.97186279296875,
144
- "loss": 1.2467,
145
- "rewards/accuracies": 0.7875000238418579,
146
- "rewards/generated": -30.259912490844727,
147
- "rewards/margins": 14.735580444335938,
148
- "rewards/real": -15.524327278137207,
149
  "step": 90
150
  },
151
  {
@@ -153,13 +153,13 @@
153
  "learning_rate": 4.850640113798008e-07,
154
  "logits/generated": -Infinity,
155
  "logits/real": -Infinity,
156
- "logps/generated": -654.0052490234375,
157
- "logps/real": -442.6397399902344,
158
- "loss": 0.9329,
159
  "rewards/accuracies": 0.8999999761581421,
160
- "rewards/generated": -35.535160064697266,
161
- "rewards/margins": 18.725465774536133,
162
- "rewards/real": -16.8096923828125,
163
  "step": 100
164
  },
165
  {
@@ -167,13 +167,13 @@
167
  "learning_rate": 4.779516358463727e-07,
168
  "logits/generated": -Infinity,
169
  "logits/real": -Infinity,
170
- "logps/generated": -638.0022583007812,
171
- "logps/real": -427.4214782714844,
172
- "loss": 1.1736,
173
  "rewards/accuracies": 0.875,
174
- "rewards/generated": -35.03214645385742,
175
- "rewards/margins": 18.788606643676758,
176
- "rewards/real": -16.24354362487793,
177
  "step": 110
178
  },
179
  {
@@ -181,13 +181,13 @@
181
  "learning_rate": 4.7083926031294454e-07,
182
  "logits/generated": -Infinity,
183
  "logits/real": -Infinity,
184
- "logps/generated": -647.80859375,
185
- "logps/real": -420.6051330566406,
186
- "loss": 0.7488,
187
  "rewards/accuracies": 0.9375,
188
- "rewards/generated": -35.87686538696289,
189
- "rewards/margins": 21.418781280517578,
190
- "rewards/real": -14.458081245422363,
191
  "step": 120
192
  },
193
  {
@@ -195,13 +195,13 @@
195
  "learning_rate": 4.6372688477951633e-07,
196
  "logits/generated": -Infinity,
197
  "logits/real": -Infinity,
198
- "logps/generated": -647.1038818359375,
199
- "logps/real": -412.8779296875,
200
- "loss": 0.6447,
201
  "rewards/accuracies": 0.875,
202
- "rewards/generated": -35.35771560668945,
203
- "rewards/margins": 20.0076904296875,
204
- "rewards/real": -15.350027084350586,
205
  "step": 130
206
  },
207
  {
@@ -209,13 +209,13 @@
209
  "learning_rate": 4.5661450924608817e-07,
210
  "logits/generated": -Infinity,
211
  "logits/real": -Infinity,
212
- "logps/generated": -699.005126953125,
213
- "logps/real": -401.8106384277344,
214
- "loss": 0.9455,
215
  "rewards/accuracies": 0.9375,
216
- "rewards/generated": -38.48512268066406,
217
- "rewards/margins": 23.671005249023438,
218
- "rewards/real": -14.814119338989258,
219
  "step": 140
220
  },
221
  {
@@ -223,13 +223,13 @@
223
  "learning_rate": 4.4950213371266e-07,
224
  "logits/generated": -Infinity,
225
  "logits/real": -Infinity,
226
- "logps/generated": -673.933349609375,
227
- "logps/real": -404.8531494140625,
228
- "loss": 0.6848,
229
- "rewards/accuracies": 0.9125000238418579,
230
- "rewards/generated": -38.79594039916992,
231
- "rewards/margins": 22.858118057250977,
232
- "rewards/real": -15.937822341918945,
233
  "step": 150
234
  },
235
  {
@@ -237,13 +237,13 @@
237
  "learning_rate": 4.4238975817923186e-07,
238
  "logits/generated": -Infinity,
239
  "logits/real": -Infinity,
240
- "logps/generated": -650.0132446289062,
241
- "logps/real": -380.55413818359375,
242
- "loss": 0.5912,
243
  "rewards/accuracies": 0.8999999761581421,
244
- "rewards/generated": -36.62251663208008,
245
- "rewards/margins": 22.439044952392578,
246
- "rewards/real": -14.183469772338867,
247
  "step": 160
248
  },
249
  {
@@ -251,13 +251,13 @@
251
  "learning_rate": 4.3527738264580364e-07,
252
  "logits/generated": -Infinity,
253
  "logits/real": -Infinity,
254
- "logps/generated": -676.5908203125,
255
- "logps/real": -413.38287353515625,
256
- "loss": 0.6238,
257
  "rewards/accuracies": 0.9125000238418579,
258
- "rewards/generated": -37.932899475097656,
259
- "rewards/margins": 22.803815841674805,
260
- "rewards/real": -15.129079818725586,
261
  "step": 170
262
  },
263
  {
@@ -265,13 +265,13 @@
265
  "learning_rate": 4.2816500711237554e-07,
266
  "logits/generated": -Infinity,
267
  "logits/real": -Infinity,
268
- "logps/generated": -660.9660034179688,
269
- "logps/real": -372.2694396972656,
270
- "loss": 0.3412,
271
  "rewards/accuracies": 0.9375,
272
- "rewards/generated": -37.31585693359375,
273
- "rewards/margins": 23.456205368041992,
274
- "rewards/real": -13.859651565551758,
275
  "step": 180
276
  },
277
  {
@@ -279,13 +279,13 @@
279
  "learning_rate": 4.2105263157894733e-07,
280
  "logits/generated": -Infinity,
281
  "logits/real": -Infinity,
282
- "logps/generated": -668.5888671875,
283
- "logps/real": -366.01507568359375,
284
- "loss": 0.5799,
285
  "rewards/accuracies": 0.9624999761581421,
286
- "rewards/generated": -37.69170379638672,
287
- "rewards/margins": 25.54647445678711,
288
- "rewards/real": -12.145231246948242,
289
  "step": 190
290
  },
291
  {
@@ -293,13 +293,13 @@
293
  "learning_rate": 4.1394025604551917e-07,
294
  "logits/generated": -Infinity,
295
  "logits/real": -Infinity,
296
- "logps/generated": -681.8294677734375,
297
- "logps/real": -392.57513427734375,
298
- "loss": 0.6742,
299
  "rewards/accuracies": 0.8999999761581421,
300
- "rewards/generated": -38.56768035888672,
301
- "rewards/margins": 24.48987579345703,
302
- "rewards/real": -14.077804565429688,
303
  "step": 200
304
  },
305
  {
@@ -307,13 +307,13 @@
307
  "learning_rate": 4.06827880512091e-07,
308
  "logits/generated": -Infinity,
309
  "logits/real": -Infinity,
310
- "logps/generated": -754.3670654296875,
311
- "logps/real": -383.51702880859375,
312
- "loss": 0.519,
313
  "rewards/accuracies": 0.925000011920929,
314
- "rewards/generated": -44.508426666259766,
315
- "rewards/margins": 30.150854110717773,
316
- "rewards/real": -14.357576370239258,
317
  "step": 210
318
  },
319
  {
@@ -321,13 +321,13 @@
321
  "learning_rate": 3.9971550497866285e-07,
322
  "logits/generated": -Infinity,
323
  "logits/real": -Infinity,
324
- "logps/generated": -669.1099243164062,
325
- "logps/real": -404.3431091308594,
326
- "loss": 0.3679,
327
  "rewards/accuracies": 0.9375,
328
- "rewards/generated": -39.06100845336914,
329
- "rewards/margins": 24.223697662353516,
330
- "rewards/real": -14.837308883666992,
331
  "step": 220
332
  },
333
  {
@@ -335,13 +335,13 @@
335
  "learning_rate": 3.926031294452347e-07,
336
  "logits/generated": -Infinity,
337
  "logits/real": -Infinity,
338
- "logps/generated": -688.25048828125,
339
- "logps/real": -396.00018310546875,
340
- "loss": 0.2677,
341
  "rewards/accuracies": 0.949999988079071,
342
- "rewards/generated": -39.11343765258789,
343
- "rewards/margins": 24.355987548828125,
344
- "rewards/real": -14.757448196411133,
345
  "step": 230
346
  },
347
  {
@@ -349,13 +349,13 @@
349
  "learning_rate": 3.8549075391180653e-07,
350
  "logits/generated": -Infinity,
351
  "logits/real": -Infinity,
352
- "logps/generated": -714.4971923828125,
353
- "logps/real": -415.25262451171875,
354
- "loss": 0.4572,
355
  "rewards/accuracies": 0.925000011920929,
356
- "rewards/generated": -40.9525032043457,
357
- "rewards/margins": 25.465791702270508,
358
- "rewards/real": -15.486714363098145,
359
  "step": 240
360
  },
361
  {
@@ -363,13 +363,13 @@
363
  "learning_rate": 3.783783783783784e-07,
364
  "logits/generated": -Infinity,
365
  "logits/real": -Infinity,
366
- "logps/generated": -729.0126342773438,
367
- "logps/real": -394.86724853515625,
368
- "loss": 0.345,
369
  "rewards/accuracies": 0.9624999761581421,
370
- "rewards/generated": -44.133033752441406,
371
- "rewards/margins": 29.94510841369629,
372
- "rewards/real": -14.187917709350586,
373
  "step": 250
374
  },
375
  {
@@ -377,13 +377,13 @@
377
  "learning_rate": 3.7126600284495016e-07,
378
  "logits/generated": -Infinity,
379
  "logits/real": -Infinity,
380
- "logps/generated": -722.02490234375,
381
- "logps/real": -403.8961181640625,
382
- "loss": 0.4853,
383
  "rewards/accuracies": 0.9375,
384
- "rewards/generated": -41.30921173095703,
385
- "rewards/margins": 26.302785873413086,
386
- "rewards/real": -15.006426811218262,
387
  "step": 260
388
  },
389
  {
@@ -391,13 +391,13 @@
391
  "learning_rate": 3.6415362731152206e-07,
392
  "logits/generated": -Infinity,
393
  "logits/real": -Infinity,
394
- "logps/generated": -695.8910522460938,
395
- "logps/real": -374.00433349609375,
396
- "loss": 0.3413,
397
  "rewards/accuracies": 0.949999988079071,
398
- "rewards/generated": -41.190391540527344,
399
- "rewards/margins": 27.258630752563477,
400
- "rewards/real": -13.931764602661133,
401
  "step": 270
402
  },
403
  {
@@ -405,13 +405,13 @@
405
  "learning_rate": 3.5704125177809385e-07,
406
  "logits/generated": -Infinity,
407
  "logits/real": -Infinity,
408
- "logps/generated": -734.4241943359375,
409
- "logps/real": -404.4153137207031,
410
- "loss": 0.4178,
411
  "rewards/accuracies": 0.9624999761581421,
412
- "rewards/generated": -43.61100387573242,
413
- "rewards/margins": 28.391124725341797,
414
- "rewards/real": -15.219881057739258,
415
  "step": 280
416
  },
417
  {
@@ -419,13 +419,13 @@
419
  "learning_rate": 3.4992887624466574e-07,
420
  "logits/generated": -Infinity,
421
  "logits/real": -Infinity,
422
- "logps/generated": -703.3042602539062,
423
- "logps/real": -400.789306640625,
424
- "loss": 0.358,
425
  "rewards/accuracies": 0.9624999761581421,
426
- "rewards/generated": -40.19432830810547,
427
- "rewards/margins": 25.275157928466797,
428
- "rewards/real": -14.919171333312988,
429
  "step": 290
430
  },
431
  {
@@ -433,13 +433,13 @@
433
  "learning_rate": 3.4281650071123753e-07,
434
  "logits/generated": -Infinity,
435
  "logits/real": -Infinity,
436
- "logps/generated": -717.597900390625,
437
- "logps/real": -387.6546325683594,
438
- "loss": 0.2285,
439
  "rewards/accuracies": 0.949999988079071,
440
- "rewards/generated": -42.98313522338867,
441
- "rewards/margins": 27.72798728942871,
442
- "rewards/real": -15.255144119262695,
443
  "step": 300
444
  },
445
  {
@@ -447,13 +447,13 @@
447
  "learning_rate": 3.3570412517780937e-07,
448
  "logits/generated": -Infinity,
449
  "logits/real": -Infinity,
450
- "logps/generated": -714.7513427734375,
451
- "logps/real": -424.5567932128906,
452
- "loss": 0.362,
453
- "rewards/accuracies": 0.8999999761581421,
454
- "rewards/generated": -42.02880859375,
455
- "rewards/margins": 26.301555633544922,
456
- "rewards/real": -15.727252006530762,
457
  "step": 310
458
  },
459
  {
@@ -461,13 +461,13 @@
461
  "learning_rate": 3.285917496443812e-07,
462
  "logits/generated": -Infinity,
463
  "logits/real": -Infinity,
464
- "logps/generated": -704.7682495117188,
465
- "logps/real": -389.437744140625,
466
- "loss": 0.3054,
467
  "rewards/accuracies": 0.925000011920929,
468
- "rewards/generated": -40.38294982910156,
469
- "rewards/margins": 26.326675415039062,
470
- "rewards/real": -14.05627727508545,
471
  "step": 320
472
  },
473
  {
@@ -475,13 +475,13 @@
475
  "learning_rate": 3.2147937411095305e-07,
476
  "logits/generated": -Infinity,
477
  "logits/real": -Infinity,
478
- "logps/generated": -788.1136474609375,
479
- "logps/real": -426.383544921875,
480
- "loss": 0.2255,
481
- "rewards/accuracies": 0.9750000238418579,
482
- "rewards/generated": -49.49579620361328,
483
- "rewards/margins": 32.605934143066406,
484
- "rewards/real": -16.889863967895508,
485
  "step": 330
486
  },
487
  {
@@ -489,13 +489,13 @@
489
  "learning_rate": 3.1436699857752484e-07,
490
  "logits/generated": -Infinity,
491
  "logits/real": -Infinity,
492
- "logps/generated": -720.3298950195312,
493
- "logps/real": -392.3839111328125,
494
- "loss": 0.3118,
495
  "rewards/accuracies": 0.9624999761581421,
496
- "rewards/generated": -42.088287353515625,
497
- "rewards/margins": 27.807758331298828,
498
- "rewards/real": -14.28053092956543,
499
  "step": 340
500
  },
501
  {
@@ -503,13 +503,13 @@
503
  "learning_rate": 3.0725462304409674e-07,
504
  "logits/generated": -Infinity,
505
  "logits/real": -Infinity,
506
- "logps/generated": -721.5670776367188,
507
- "logps/real": -399.3511047363281,
508
- "loss": 0.2108,
509
  "rewards/accuracies": 0.9375,
510
- "rewards/generated": -44.001930236816406,
511
- "rewards/margins": 28.8182430267334,
512
- "rewards/real": -15.183688163757324,
513
  "step": 350
514
  },
515
  {
@@ -517,13 +517,13 @@
517
  "learning_rate": 3.001422475106685e-07,
518
  "logits/generated": -Infinity,
519
  "logits/real": -Infinity,
520
- "logps/generated": -734.0841064453125,
521
- "logps/real": -473.2635192871094,
522
- "loss": 0.6012,
523
- "rewards/accuracies": 0.925000011920929,
524
- "rewards/generated": -44.092620849609375,
525
- "rewards/margins": 25.192630767822266,
526
- "rewards/real": -18.89999008178711,
527
  "step": 360
528
  },
529
  {
@@ -531,13 +531,13 @@
531
  "learning_rate": 2.9302987197724037e-07,
532
  "logits/generated": -Infinity,
533
  "logits/real": -Infinity,
534
- "logps/generated": -666.3681030273438,
535
- "logps/real": -369.57232666015625,
536
- "loss": 0.3811,
537
  "rewards/accuracies": 0.9624999761581421,
538
- "rewards/generated": -38.6079216003418,
539
- "rewards/margins": 23.917612075805664,
540
- "rewards/real": -14.690305709838867,
541
  "step": 370
542
  },
543
  {
@@ -545,13 +545,13 @@
545
  "learning_rate": 2.8591749644381226e-07,
546
  "logits/generated": -Infinity,
547
  "logits/real": -Infinity,
548
- "logps/generated": -709.2445068359375,
549
- "logps/real": -394.5869140625,
550
- "loss": 0.3208,
551
- "rewards/accuracies": 0.949999988079071,
552
- "rewards/generated": -42.30268096923828,
553
- "rewards/margins": 26.500690460205078,
554
- "rewards/real": -15.80199146270752,
555
  "step": 380
556
  },
557
  {
@@ -559,13 +559,13 @@
559
  "learning_rate": 2.7880512091038405e-07,
560
  "logits/generated": -Infinity,
561
  "logits/real": -Infinity,
562
- "logps/generated": -775.5260009765625,
563
- "logps/real": -426.09869384765625,
564
- "loss": 0.2884,
565
  "rewards/accuracies": 0.9624999761581421,
566
- "rewards/generated": -48.343849182128906,
567
- "rewards/margins": 31.10616683959961,
568
- "rewards/real": -17.237682342529297,
569
  "step": 390
570
  },
571
  {
@@ -573,13 +573,13 @@
573
  "learning_rate": 2.716927453769559e-07,
574
  "logits/generated": -Infinity,
575
  "logits/real": -Infinity,
576
- "logps/generated": -773.4555053710938,
577
- "logps/real": -450.7288513183594,
578
- "loss": 0.3376,
579
  "rewards/accuracies": 0.949999988079071,
580
- "rewards/generated": -47.49498748779297,
581
- "rewards/margins": 28.620986938476562,
582
- "rewards/real": -18.873998641967773,
583
  "step": 400
584
  },
585
  {
@@ -587,13 +587,13 @@
587
  "learning_rate": 2.6458036984352773e-07,
588
  "logits/generated": -Infinity,
589
  "logits/real": -Infinity,
590
- "logps/generated": -741.7351684570312,
591
- "logps/real": -449.17816162109375,
592
- "loss": 0.3754,
593
- "rewards/accuracies": 0.925000011920929,
594
- "rewards/generated": -46.99356460571289,
595
- "rewards/margins": 27.995086669921875,
596
- "rewards/real": -18.998476028442383,
597
  "step": 410
598
  },
599
  {
@@ -601,13 +601,13 @@
601
  "learning_rate": 2.574679943100996e-07,
602
  "logits/generated": -Infinity,
603
  "logits/real": -Infinity,
604
- "logps/generated": -738.3614501953125,
605
- "logps/real": -456.7314453125,
606
- "loss": 0.0844,
607
- "rewards/accuracies": 0.9375,
608
- "rewards/generated": -46.36874008178711,
609
- "rewards/margins": 27.84047508239746,
610
- "rewards/real": -18.528268814086914,
611
  "step": 420
612
  },
613
  {
@@ -615,13 +615,13 @@
615
  "learning_rate": 2.5035561877667136e-07,
616
  "logits/generated": -Infinity,
617
  "logits/real": -Infinity,
618
- "logps/generated": -768.1492309570312,
619
- "logps/real": -441.343017578125,
620
- "loss": 0.1898,
621
  "rewards/accuracies": 0.9750000238418579,
622
- "rewards/generated": -47.066619873046875,
623
- "rewards/margins": 29.435678482055664,
624
- "rewards/real": -17.63094139099121,
625
  "step": 430
626
  },
627
  {
@@ -629,13 +629,13 @@
629
  "learning_rate": 2.4324324324324326e-07,
630
  "logits/generated": -Infinity,
631
  "logits/real": -Infinity,
632
- "logps/generated": -787.4987182617188,
633
- "logps/real": -408.2841796875,
634
- "loss": 0.3922,
635
  "rewards/accuracies": 0.987500011920929,
636
- "rewards/generated": -50.27423095703125,
637
- "rewards/margins": 33.921939849853516,
638
- "rewards/real": -16.352293014526367,
639
  "step": 440
640
  },
641
  {
@@ -643,13 +643,13 @@
643
  "learning_rate": 2.3613086770981507e-07,
644
  "logits/generated": -Infinity,
645
  "logits/real": -Infinity,
646
- "logps/generated": -771.3111572265625,
647
- "logps/real": -408.37896728515625,
648
- "loss": 0.3608,
649
  "rewards/accuracies": 0.949999988079071,
650
- "rewards/generated": -48.568450927734375,
651
- "rewards/margins": 31.147785186767578,
652
- "rewards/real": -17.420665740966797,
653
  "step": 450
654
  },
655
  {
@@ -657,13 +657,13 @@
657
  "learning_rate": 2.290184921763869e-07,
658
  "logits/generated": -Infinity,
659
  "logits/real": -Infinity,
660
- "logps/generated": -827.6937255859375,
661
- "logps/real": -423.6630859375,
662
- "loss": 0.1253,
663
  "rewards/accuracies": 1.0,
664
- "rewards/generated": -51.94307327270508,
665
- "rewards/margins": 35.77669143676758,
666
- "rewards/real": -16.166385650634766,
667
  "step": 460
668
  },
669
  {
@@ -671,13 +671,13 @@
671
  "learning_rate": 2.2190611664295875e-07,
672
  "logits/generated": -Infinity,
673
  "logits/real": -Infinity,
674
- "logps/generated": -790.8399658203125,
675
- "logps/real": -440.3358459472656,
676
- "loss": 0.2949,
677
  "rewards/accuracies": 0.925000011920929,
678
- "rewards/generated": -48.574867248535156,
679
- "rewards/margins": 30.353759765625,
680
- "rewards/real": -18.22110939025879,
681
  "step": 470
682
  },
683
  {
@@ -685,13 +685,13 @@
685
  "learning_rate": 2.1479374110953057e-07,
686
  "logits/generated": -Infinity,
687
  "logits/real": -Infinity,
688
- "logps/generated": -781.35791015625,
689
- "logps/real": -390.54168701171875,
690
- "loss": 0.3877,
691
- "rewards/accuracies": 0.9624999761581421,
692
- "rewards/generated": -49.05020523071289,
693
- "rewards/margins": 32.30176544189453,
694
- "rewards/real": -16.748441696166992,
695
  "step": 480
696
  },
697
  {
@@ -699,13 +699,13 @@
699
  "learning_rate": 2.076813655761024e-07,
700
  "logits/generated": -Infinity,
701
  "logits/real": -Infinity,
702
- "logps/generated": -810.448486328125,
703
- "logps/real": -441.8717346191406,
704
- "loss": 0.2217,
705
  "rewards/accuracies": 0.987500011920929,
706
- "rewards/generated": -51.156005859375,
707
- "rewards/margins": 31.813852310180664,
708
- "rewards/real": -19.342153549194336,
709
  "step": 490
710
  },
711
  {
@@ -713,29 +713,29 @@
713
  "learning_rate": 2.0056899004267425e-07,
714
  "logits/generated": -Infinity,
715
  "logits/real": -Infinity,
716
- "logps/generated": -802.3043212890625,
717
- "logps/real": -487.0044860839844,
718
- "loss": 0.2807,
719
  "rewards/accuracies": 0.9624999761581421,
720
- "rewards/generated": -51.0858039855957,
721
- "rewards/margins": 30.591796875,
722
- "rewards/real": -20.494003295898438,
723
  "step": 500
724
  },
725
  {
726
  "epoch": 0.64,
727
  "eval_logits/generated": -Infinity,
728
  "eval_logits/real": -Infinity,
729
- "eval_logps/generated": -518.7460327148438,
730
- "eval_logps/real": -321.637939453125,
731
- "eval_loss": 0.16799330711364746,
732
  "eval_rewards/accuracies": 0.9442675113677979,
733
- "eval_rewards/generated": -22.73108673095703,
734
- "eval_rewards/margins": 15.805987358093262,
735
- "eval_rewards/real": -6.9250993728637695,
736
- "eval_runtime": 514.9684,
737
- "eval_samples_per_second": 9.709,
738
- "eval_steps_per_second": 0.305,
739
  "step": 500
740
  },
741
  {
@@ -743,13 +743,13 @@
743
  "learning_rate": 1.9345661450924607e-07,
744
  "logits/generated": -Infinity,
745
  "logits/real": -Infinity,
746
- "logps/generated": -764.2227783203125,
747
- "logps/real": -417.8741149902344,
748
- "loss": 0.2566,
749
- "rewards/accuracies": 0.9624999761581421,
750
- "rewards/generated": -46.21670150756836,
751
- "rewards/margins": 29.2624568939209,
752
- "rewards/real": -16.954248428344727,
753
  "step": 510
754
  },
755
  {
@@ -757,13 +757,13 @@
757
  "learning_rate": 1.863442389758179e-07,
758
  "logits/generated": -Infinity,
759
  "logits/real": -Infinity,
760
- "logps/generated": -785.6708984375,
761
- "logps/real": -436.8841857910156,
762
- "loss": 0.2175,
763
  "rewards/accuracies": 0.987500011920929,
764
- "rewards/generated": -50.03806686401367,
765
- "rewards/margins": 33.52396011352539,
766
- "rewards/real": -16.514102935791016,
767
  "step": 520
768
  },
769
  {
@@ -771,13 +771,13 @@
771
  "learning_rate": 1.7923186344238975e-07,
772
  "logits/generated": -Infinity,
773
  "logits/real": -Infinity,
774
- "logps/generated": -741.9736328125,
775
- "logps/real": -408.57763671875,
776
- "loss": 0.3024,
777
- "rewards/accuracies": 0.9750000238418579,
778
- "rewards/generated": -45.586387634277344,
779
- "rewards/margins": 29.933719635009766,
780
- "rewards/real": -15.652669906616211,
781
  "step": 530
782
  },
783
  {
@@ -785,13 +785,13 @@
785
  "learning_rate": 1.721194879089616e-07,
786
  "logits/generated": -Infinity,
787
  "logits/real": -Infinity,
788
- "logps/generated": -813.8468017578125,
789
- "logps/real": -430.96026611328125,
790
- "loss": 0.2112,
791
  "rewards/accuracies": 0.949999988079071,
792
- "rewards/generated": -52.11174392700195,
793
- "rewards/margins": 33.244110107421875,
794
- "rewards/real": -18.867637634277344,
795
  "step": 540
796
  },
797
  {
@@ -799,13 +799,13 @@
799
  "learning_rate": 1.650071123755334e-07,
800
  "logits/generated": -Infinity,
801
  "logits/real": -Infinity,
802
- "logps/generated": -813.7432250976562,
803
- "logps/real": -425.29754638671875,
804
- "loss": 0.3095,
805
  "rewards/accuracies": 0.949999988079071,
806
- "rewards/generated": -52.5365104675293,
807
- "rewards/margins": 34.71119689941406,
808
- "rewards/real": -17.825315475463867,
809
  "step": 550
810
  },
811
  {
@@ -813,13 +813,13 @@
813
  "learning_rate": 1.5789473684210525e-07,
814
  "logits/generated": -Infinity,
815
  "logits/real": -Infinity,
816
- "logps/generated": -790.2058715820312,
817
- "logps/real": -440.711669921875,
818
- "loss": 0.1864,
819
  "rewards/accuracies": 0.9624999761581421,
820
- "rewards/generated": -49.65260314941406,
821
- "rewards/margins": 31.609088897705078,
822
- "rewards/real": -18.04351234436035,
823
  "step": 560
824
  },
825
  {
@@ -827,13 +827,13 @@
827
  "learning_rate": 1.507823613086771e-07,
828
  "logits/generated": -Infinity,
829
  "logits/real": -Infinity,
830
- "logps/generated": -752.94677734375,
831
- "logps/real": -427.7891540527344,
832
- "loss": 0.3443,
833
- "rewards/accuracies": 0.925000011920929,
834
- "rewards/generated": -47.858802795410156,
835
- "rewards/margins": 30.949520111083984,
836
- "rewards/real": -16.90927505493164,
837
  "step": 570
838
  },
839
  {
@@ -841,13 +841,13 @@
841
  "learning_rate": 1.436699857752489e-07,
842
  "logits/generated": -Infinity,
843
  "logits/real": -Infinity,
844
- "logps/generated": -843.44140625,
845
- "logps/real": -425.5953063964844,
846
- "loss": 0.3088,
847
  "rewards/accuracies": 0.9750000238418579,
848
- "rewards/generated": -53.65172576904297,
849
- "rewards/margins": 36.3214111328125,
850
- "rewards/real": -17.330326080322266,
851
  "step": 580
852
  },
853
  {
@@ -855,13 +855,13 @@
855
  "learning_rate": 1.3655761024182077e-07,
856
  "logits/generated": -Infinity,
857
  "logits/real": -Infinity,
858
- "logps/generated": -816.5244750976562,
859
- "logps/real": -419.6240234375,
860
- "loss": 0.2445,
861
  "rewards/accuracies": 0.9624999761581421,
862
- "rewards/generated": -52.9045524597168,
863
- "rewards/margins": 36.242149353027344,
864
- "rewards/real": -16.662403106689453,
865
  "step": 590
866
  },
867
  {
@@ -869,13 +869,13 @@
869
  "learning_rate": 1.2944523470839261e-07,
870
  "logits/generated": -Infinity,
871
  "logits/real": -Infinity,
872
- "logps/generated": -824.7449340820312,
873
- "logps/real": -429.56414794921875,
874
- "loss": 0.3351,
875
- "rewards/accuracies": 0.949999988079071,
876
- "rewards/generated": -51.94647979736328,
877
- "rewards/margins": 34.617332458496094,
878
- "rewards/real": -17.329151153564453,
879
  "step": 600
880
  },
881
  {
@@ -883,13 +883,13 @@
883
  "learning_rate": 1.2233285917496443e-07,
884
  "logits/generated": -Infinity,
885
  "logits/real": -Infinity,
886
- "logps/generated": -775.5467529296875,
887
- "logps/real": -422.3451232910156,
888
- "loss": 0.07,
889
  "rewards/accuracies": 0.987500011920929,
890
- "rewards/generated": -48.54518508911133,
891
- "rewards/margins": 32.839900970458984,
892
- "rewards/real": -15.705281257629395,
893
  "step": 610
894
  },
895
  {
@@ -897,13 +897,13 @@
897
  "learning_rate": 1.1522048364153626e-07,
898
  "logits/generated": -Infinity,
899
  "logits/real": -Infinity,
900
- "logps/generated": -779.6900634765625,
901
- "logps/real": -441.49365234375,
902
- "loss": 0.2883,
903
  "rewards/accuracies": 0.9624999761581421,
904
- "rewards/generated": -49.64753341674805,
905
- "rewards/margins": 32.548099517822266,
906
- "rewards/real": -17.099435806274414,
907
  "step": 620
908
  },
909
  {
@@ -911,13 +911,13 @@
911
  "learning_rate": 1.0810810810810811e-07,
912
  "logits/generated": -Infinity,
913
  "logits/real": -Infinity,
914
- "logps/generated": -785.1602172851562,
915
- "logps/real": -403.9580993652344,
916
- "loss": 0.1209,
917
  "rewards/accuracies": 0.9624999761581421,
918
- "rewards/generated": -49.906761169433594,
919
- "rewards/margins": 33.77416229248047,
920
- "rewards/real": -16.13260269165039,
921
  "step": 630
922
  },
923
  {
@@ -925,13 +925,13 @@
925
  "learning_rate": 1.0099573257467994e-07,
926
  "logits/generated": -Infinity,
927
  "logits/real": -Infinity,
928
- "logps/generated": -830.7962036132812,
929
- "logps/real": -457.5830993652344,
930
- "loss": 0.2802,
931
  "rewards/accuracies": 0.949999988079071,
932
- "rewards/generated": -53.04398727416992,
933
- "rewards/margins": 34.22355270385742,
934
- "rewards/real": -18.820430755615234,
935
  "step": 640
936
  },
937
  {
@@ -939,13 +939,13 @@
939
  "learning_rate": 9.388335704125178e-08,
940
  "logits/generated": -Infinity,
941
  "logits/real": -Infinity,
942
- "logps/generated": -774.2494506835938,
943
- "logps/real": -407.559326171875,
944
- "loss": 0.1348,
945
  "rewards/accuracies": 1.0,
946
- "rewards/generated": -49.61153030395508,
947
- "rewards/margins": 32.225975036621094,
948
- "rewards/real": -17.385555267333984,
949
  "step": 650
950
  },
951
  {
@@ -953,13 +953,13 @@
953
  "learning_rate": 8.677098150782361e-08,
954
  "logits/generated": -Infinity,
955
  "logits/real": -Infinity,
956
- "logps/generated": -772.5739135742188,
957
- "logps/real": -409.3276672363281,
958
- "loss": 0.0997,
959
  "rewards/accuracies": 0.987500011920929,
960
- "rewards/generated": -48.17709732055664,
961
- "rewards/margins": 32.27601623535156,
962
- "rewards/real": -15.901082038879395,
963
  "step": 660
964
  },
965
  {
@@ -967,13 +967,13 @@
967
  "learning_rate": 7.965860597439544e-08,
968
  "logits/generated": -Infinity,
969
  "logits/real": -Infinity,
970
- "logps/generated": -764.9151000976562,
971
- "logps/real": -425.33709716796875,
972
- "loss": 0.2364,
973
  "rewards/accuracies": 0.925000011920929,
974
- "rewards/generated": -48.48712921142578,
975
- "rewards/margins": 31.54388427734375,
976
- "rewards/real": -16.9432430267334,
977
  "step": 670
978
  },
979
  {
@@ -981,13 +981,13 @@
981
  "learning_rate": 7.254623044096728e-08,
982
  "logits/generated": -Infinity,
983
  "logits/real": -Infinity,
984
- "logps/generated": -822.6394653320312,
985
- "logps/real": -434.67877197265625,
986
- "loss": 0.242,
987
  "rewards/accuracies": 0.987500011920929,
988
- "rewards/generated": -53.5615348815918,
989
- "rewards/margins": 35.22587585449219,
990
- "rewards/real": -18.33565902709961,
991
  "step": 680
992
  },
993
  {
@@ -995,13 +995,13 @@
995
  "learning_rate": 6.543385490753911e-08,
996
  "logits/generated": -Infinity,
997
  "logits/real": -Infinity,
998
- "logps/generated": -868.9513549804688,
999
- "logps/real": -439.777587890625,
1000
- "loss": 0.203,
1001
- "rewards/accuracies": 0.9624999761581421,
1002
- "rewards/generated": -56.48514938354492,
1003
- "rewards/margins": 37.33259201049805,
1004
- "rewards/real": -19.152555465698242,
1005
  "step": 690
1006
  },
1007
  {
@@ -1009,13 +1009,13 @@
1009
  "learning_rate": 5.832147937411095e-08,
1010
  "logits/generated": -Infinity,
1011
  "logits/real": -Infinity,
1012
- "logps/generated": -784.9397583007812,
1013
- "logps/real": -413.0848083496094,
1014
- "loss": 0.2787,
1015
- "rewards/accuracies": 0.925000011920929,
1016
- "rewards/generated": -47.916893005371094,
1017
- "rewards/margins": 29.88480567932129,
1018
- "rewards/real": -18.032087326049805,
1019
  "step": 700
1020
  },
1021
  {
@@ -1023,13 +1023,13 @@
1023
  "learning_rate": 5.120910384068278e-08,
1024
  "logits/generated": -Infinity,
1025
  "logits/real": -Infinity,
1026
- "logps/generated": -824.5567626953125,
1027
- "logps/real": -448.226318359375,
1028
- "loss": 0.2584,
1029
  "rewards/accuracies": 0.949999988079071,
1030
- "rewards/generated": -53.5278205871582,
1031
- "rewards/margins": 35.21510696411133,
1032
- "rewards/real": -18.312713623046875,
1033
  "step": 710
1034
  },
1035
  {
@@ -1037,13 +1037,13 @@
1037
  "learning_rate": 4.4096728307254624e-08,
1038
  "logits/generated": -Infinity,
1039
  "logits/real": -Infinity,
1040
- "logps/generated": -828.8753662109375,
1041
- "logps/real": -427.9056091308594,
1042
- "loss": 0.1088,
1043
  "rewards/accuracies": 0.987500011920929,
1044
- "rewards/generated": -53.2839469909668,
1045
- "rewards/margins": 35.713260650634766,
1046
- "rewards/real": -17.57068634033203,
1047
  "step": 720
1048
  },
1049
  {
@@ -1051,13 +1051,13 @@
1051
  "learning_rate": 3.698435277382646e-08,
1052
  "logits/generated": -Infinity,
1053
  "logits/real": -Infinity,
1054
- "logps/generated": -839.0155029296875,
1055
- "logps/real": -435.33428955078125,
1056
- "loss": 0.2107,
1057
- "rewards/accuracies": 1.0,
1058
- "rewards/generated": -55.168609619140625,
1059
- "rewards/margins": 37.164310455322266,
1060
- "rewards/real": -18.004301071166992,
1061
  "step": 730
1062
  },
1063
  {
@@ -1065,13 +1065,13 @@
1065
  "learning_rate": 2.9871977240398294e-08,
1066
  "logits/generated": -Infinity,
1067
  "logits/real": -Infinity,
1068
- "logps/generated": -805.4227905273438,
1069
- "logps/real": -428.77923583984375,
1070
- "loss": 0.2453,
1071
  "rewards/accuracies": 0.9624999761581421,
1072
- "rewards/generated": -52.77501678466797,
1073
- "rewards/margins": 34.709877014160156,
1074
- "rewards/real": -18.065141677856445,
1075
  "step": 740
1076
  },
1077
  {
@@ -1079,13 +1079,13 @@
1079
  "learning_rate": 2.275960170697013e-08,
1080
  "logits/generated": -Infinity,
1081
  "logits/real": -Infinity,
1082
- "logps/generated": -834.6575927734375,
1083
- "logps/real": -415.8843688964844,
1084
- "loss": 0.1762,
1085
  "rewards/accuracies": 0.9750000238418579,
1086
- "rewards/generated": -52.49644088745117,
1087
- "rewards/margins": 36.04848098754883,
1088
- "rewards/real": -16.447967529296875,
1089
  "step": 750
1090
  },
1091
  {
@@ -1093,13 +1093,13 @@
1093
  "learning_rate": 1.564722617354196e-08,
1094
  "logits/generated": -Infinity,
1095
  "logits/real": -Infinity,
1096
- "logps/generated": -850.2274169921875,
1097
- "logps/real": -395.8170471191406,
1098
- "loss": 0.1395,
1099
  "rewards/accuracies": 0.9750000238418579,
1100
- "rewards/generated": -55.13444900512695,
1101
- "rewards/margins": 38.56990432739258,
1102
- "rewards/real": -16.564544677734375,
1103
  "step": 760
1104
  },
1105
  {
@@ -1107,13 +1107,13 @@
1107
  "learning_rate": 8.534850640113798e-09,
1108
  "logits/generated": -Infinity,
1109
  "logits/real": -Infinity,
1110
- "logps/generated": -807.279296875,
1111
- "logps/real": -415.69586181640625,
1112
- "loss": 0.1756,
1113
- "rewards/accuracies": 0.9750000238418579,
1114
- "rewards/generated": -52.691368103027344,
1115
- "rewards/margins": 36.51746368408203,
1116
- "rewards/real": -16.173904418945312,
1117
  "step": 770
1118
  },
1119
  {
@@ -1121,23 +1121,23 @@
1121
  "learning_rate": 1.422475106685633e-09,
1122
  "logits/generated": -Infinity,
1123
  "logits/real": -Infinity,
1124
- "logps/generated": -819.2388916015625,
1125
- "logps/real": -390.07391357421875,
1126
- "loss": 0.2143,
1127
  "rewards/accuracies": 0.9750000238418579,
1128
- "rewards/generated": -52.7367057800293,
1129
- "rewards/margins": 36.12593460083008,
1130
- "rewards/real": -16.610774993896484,
1131
  "step": 780
1132
  },
1133
  {
1134
  "epoch": 1.0,
1135
  "step": 782,
1136
  "total_flos": 0.0,
1137
- "train_loss": 0.6284302195911938,
1138
- "train_runtime": 5860.6637,
1139
- "train_samples_per_second": 4.266,
1140
- "train_steps_per_second": 0.133
1141
  }
1142
  ],
1143
  "logging_steps": 10,
 
27
  "learning_rate": 6.329113924050633e-08,
28
  "logits/generated": -Infinity,
29
  "logits/real": -Infinity,
30
+ "logps/generated": -487.2853088378906,
31
+ "logps/real": -384.8033142089844,
32
+ "loss": 3.4516,
33
+ "rewards/accuracies": 0.6111111044883728,
34
+ "rewards/generated": -19.51491928100586,
35
+ "rewards/margins": 6.083561420440674,
36
+ "rewards/real": -13.431358337402344,
37
  "step": 10
38
  },
39
  {
 
41
  "learning_rate": 1.2658227848101266e-07,
42
  "logits/generated": -Infinity,
43
  "logits/real": -Infinity,
44
+ "logps/generated": -473.446044921875,
45
+ "logps/real": -413.171630859375,
46
+ "loss": 3.9705,
47
  "rewards/accuracies": 0.637499988079071,
48
+ "rewards/generated": -18.367900848388672,
49
+ "rewards/margins": 3.9024269580841064,
50
+ "rewards/real": -14.465472221374512,
51
  "step": 20
52
  },
53
  {
 
55
  "learning_rate": 1.89873417721519e-07,
56
  "logits/generated": -Infinity,
57
  "logits/real": -Infinity,
58
+ "logps/generated": -510.4585876464844,
59
+ "logps/real": -408.74212646484375,
60
+ "loss": 3.6955,
61
  "rewards/accuracies": 0.6875,
62
+ "rewards/generated": -21.059778213500977,
63
+ "rewards/margins": 7.762242317199707,
64
+ "rewards/real": -13.297533988952637,
65
  "step": 30
66
  },
67
  {
 
69
  "learning_rate": 2.5316455696202533e-07,
70
  "logits/generated": -Infinity,
71
  "logits/real": -Infinity,
72
+ "logps/generated": -451.734130859375,
73
+ "logps/real": -368.16998291015625,
74
+ "loss": 3.9056,
75
  "rewards/accuracies": 0.625,
76
+ "rewards/generated": -16.805126190185547,
77
+ "rewards/margins": 4.4605584144592285,
78
+ "rewards/real": -12.344568252563477,
79
  "step": 40
80
  },
81
  {
 
83
  "learning_rate": 3.1645569620253163e-07,
84
  "logits/generated": -Infinity,
85
  "logits/real": -Infinity,
86
+ "logps/generated": -552.9141845703125,
87
+ "logps/real": -397.95318603515625,
88
+ "loss": 2.9742,
89
  "rewards/accuracies": 0.737500011920929,
90
+ "rewards/generated": -25.695674896240234,
91
+ "rewards/margins": 10.653547286987305,
92
+ "rewards/real": -15.042126655578613,
93
  "step": 50
94
  },
95
  {
 
97
  "learning_rate": 3.79746835443038e-07,
98
  "logits/generated": -Infinity,
99
  "logits/real": -Infinity,
100
+ "logps/generated": -509.5240783691406,
101
+ "logps/real": -390.6533203125,
102
+ "loss": 2.5682,
103
  "rewards/accuracies": 0.737500011920929,
104
+ "rewards/generated": -22.568603515625,
105
+ "rewards/margins": 7.904494285583496,
106
+ "rewards/real": -14.664111137390137,
107
  "step": 60
108
  },
109
  {
 
111
  "learning_rate": 4.4303797468354424e-07,
112
  "logits/generated": -Infinity,
113
  "logits/real": -Infinity,
114
+ "logps/generated": -554.5638427734375,
115
+ "logps/real": -382.84832763671875,
116
+ "loss": 1.7793,
117
  "rewards/accuracies": 0.8374999761581421,
118
+ "rewards/generated": -28.215290069580078,
119
+ "rewards/margins": 14.945714950561523,
120
+ "rewards/real": -13.269571304321289,
121
  "step": 70
122
  },
123
  {
 
125
  "learning_rate": 4.992887624466572e-07,
126
  "logits/generated": -Infinity,
127
  "logits/real": -Infinity,
128
+ "logps/generated": -574.03759765625,
129
+ "logps/real": -402.69232177734375,
130
+ "loss": 1.4194,
131
  "rewards/accuracies": 0.800000011920929,
132
+ "rewards/generated": -28.118602752685547,
133
+ "rewards/margins": 12.523630142211914,
134
+ "rewards/real": -15.594972610473633,
135
  "step": 80
136
  },
137
  {
 
139
  "learning_rate": 4.92176386913229e-07,
140
  "logits/generated": -Infinity,
141
  "logits/real": -Infinity,
142
+ "logps/generated": -577.7903442382812,
143
+ "logps/real": -401.81561279296875,
144
+ "loss": 1.2635,
145
+ "rewards/accuracies": 0.7749999761581421,
146
+ "rewards/generated": -30.364093780517578,
147
+ "rewards/margins": 14.755389213562012,
148
+ "rewards/real": -15.608701705932617,
149
  "step": 90
150
  },
151
  {
 
153
  "learning_rate": 4.850640113798008e-07,
154
  "logits/generated": -Infinity,
155
  "logits/real": -Infinity,
156
+ "logps/generated": -654.4276733398438,
157
+ "logps/real": -442.893798828125,
158
+ "loss": 0.9213,
159
  "rewards/accuracies": 0.8999999761581421,
160
+ "rewards/generated": -35.57740020751953,
161
+ "rewards/margins": 18.7423038482666,
162
+ "rewards/real": -16.835100173950195,
163
  "step": 100
164
  },
165
  {
 
167
  "learning_rate": 4.779516358463727e-07,
168
  "logits/generated": -Infinity,
169
  "logits/real": -Infinity,
170
+ "logps/generated": -637.9244384765625,
171
+ "logps/real": -426.7471618652344,
172
+ "loss": 1.0798,
173
  "rewards/accuracies": 0.875,
174
+ "rewards/generated": -35.024356842041016,
175
+ "rewards/margins": 18.848251342773438,
176
+ "rewards/real": -16.17610740661621,
177
  "step": 110
178
  },
179
  {
 
181
  "learning_rate": 4.7083926031294454e-07,
182
  "logits/generated": -Infinity,
183
  "logits/real": -Infinity,
184
+ "logps/generated": -647.3193969726562,
185
+ "logps/real": -419.98095703125,
186
+ "loss": 0.7428,
187
  "rewards/accuracies": 0.9375,
188
+ "rewards/generated": -35.827938079833984,
189
+ "rewards/margins": 21.43227767944336,
190
+ "rewards/real": -14.395665168762207,
191
  "step": 120
192
  },
193
  {
 
195
  "learning_rate": 4.6372688477951633e-07,
196
  "logits/generated": -Infinity,
197
  "logits/real": -Infinity,
198
+ "logps/generated": -647.6121826171875,
199
+ "logps/real": -413.42401123046875,
200
+ "loss": 0.6566,
201
  "rewards/accuracies": 0.875,
202
+ "rewards/generated": -35.408538818359375,
203
+ "rewards/margins": 20.003904342651367,
204
+ "rewards/real": -15.404635429382324,
205
  "step": 130
206
  },
207
  {
 
209
  "learning_rate": 4.5661450924608817e-07,
210
  "logits/generated": -Infinity,
211
  "logits/real": -Infinity,
212
+ "logps/generated": -697.9991455078125,
213
+ "logps/real": -401.5552673339844,
214
+ "loss": 0.9421,
215
  "rewards/accuracies": 0.9375,
216
+ "rewards/generated": -38.38452911376953,
217
+ "rewards/margins": 23.595943450927734,
218
+ "rewards/real": -14.788581848144531,
219
  "step": 140
220
  },
221
  {
 
223
  "learning_rate": 4.4950213371266e-07,
224
  "logits/generated": -Infinity,
225
  "logits/real": -Infinity,
226
+ "logps/generated": -672.8363037109375,
227
+ "logps/real": -401.61590576171875,
228
+ "loss": 0.6587,
229
+ "rewards/accuracies": 0.8999999761581421,
230
+ "rewards/generated": -38.686241149902344,
231
+ "rewards/margins": 23.072139739990234,
232
+ "rewards/real": -15.614102363586426,
233
  "step": 150
234
  },
235
  {
 
237
  "learning_rate": 4.4238975817923186e-07,
238
  "logits/generated": -Infinity,
239
  "logits/real": -Infinity,
240
+ "logps/generated": -641.0380859375,
241
+ "logps/real": -377.9221496582031,
242
+ "loss": 0.6261,
243
  "rewards/accuracies": 0.8999999761581421,
244
+ "rewards/generated": -35.72499084472656,
245
+ "rewards/margins": 21.804719924926758,
246
+ "rewards/real": -13.920272827148438,
247
  "step": 160
248
  },
249
  {
 
251
  "learning_rate": 4.3527738264580364e-07,
252
  "logits/generated": -Infinity,
253
  "logits/real": -Infinity,
254
+ "logps/generated": -676.1116943359375,
255
+ "logps/real": -410.857421875,
256
+ "loss": 0.6189,
257
  "rewards/accuracies": 0.9125000238418579,
258
+ "rewards/generated": -37.884979248046875,
259
+ "rewards/margins": 23.008445739746094,
260
+ "rewards/real": -14.876535415649414,
261
  "step": 170
262
  },
263
  {
 
265
  "learning_rate": 4.2816500711237554e-07,
266
  "logits/generated": -Infinity,
267
  "logits/real": -Infinity,
268
+ "logps/generated": -654.6505737304688,
269
+ "logps/real": -370.24224853515625,
270
+ "loss": 0.3247,
271
  "rewards/accuracies": 0.9375,
272
+ "rewards/generated": -36.68430709838867,
273
+ "rewards/margins": 23.027376174926758,
274
+ "rewards/real": -13.656933784484863,
275
  "step": 180
276
  },
277
  {
 
279
  "learning_rate": 4.2105263157894733e-07,
280
  "logits/generated": -Infinity,
281
  "logits/real": -Infinity,
282
+ "logps/generated": -665.4993286132812,
283
+ "logps/real": -364.58880615234375,
284
+ "loss": 0.5709,
285
  "rewards/accuracies": 0.9624999761581421,
286
+ "rewards/generated": -37.38275146484375,
287
+ "rewards/margins": 25.380136489868164,
288
+ "rewards/real": -12.002609252929688,
289
  "step": 190
290
  },
291
  {
 
293
  "learning_rate": 4.1394025604551917e-07,
294
  "logits/generated": -Infinity,
295
  "logits/real": -Infinity,
296
+ "logps/generated": -684.380615234375,
297
+ "logps/real": -392.02288818359375,
298
+ "loss": 0.6823,
299
  "rewards/accuracies": 0.8999999761581421,
300
+ "rewards/generated": -38.82280349731445,
301
+ "rewards/margins": 24.800228118896484,
302
+ "rewards/real": -14.02257251739502,
303
  "step": 200
304
  },
305
  {
 
307
  "learning_rate": 4.06827880512091e-07,
308
  "logits/generated": -Infinity,
309
  "logits/real": -Infinity,
310
+ "logps/generated": -741.0132446289062,
311
+ "logps/real": -383.6390075683594,
312
+ "loss": 0.5369,
313
  "rewards/accuracies": 0.925000011920929,
314
+ "rewards/generated": -43.17304229736328,
315
+ "rewards/margins": 28.80326271057129,
316
+ "rewards/real": -14.369776725769043,
317
  "step": 210
318
  },
319
  {
 
321
  "learning_rate": 3.9971550497866285e-07,
322
  "logits/generated": -Infinity,
323
  "logits/real": -Infinity,
324
+ "logps/generated": -665.7184448242188,
325
+ "logps/real": -401.6400146484375,
326
+ "loss": 0.3696,
327
  "rewards/accuracies": 0.9375,
328
+ "rewards/generated": -38.721866607666016,
329
+ "rewards/margins": 24.154865264892578,
330
+ "rewards/real": -14.567001342773438,
331
  "step": 220
332
  },
333
  {
 
335
  "learning_rate": 3.926031294452347e-07,
336
  "logits/generated": -Infinity,
337
  "logits/real": -Infinity,
338
+ "logps/generated": -681.5037231445312,
339
+ "logps/real": -394.03509521484375,
340
+ "loss": 0.2532,
341
  "rewards/accuracies": 0.949999988079071,
342
+ "rewards/generated": -38.438758850097656,
343
+ "rewards/margins": 23.877817153930664,
344
+ "rewards/real": -14.560938835144043,
345
  "step": 230
346
  },
347
  {
 
349
  "learning_rate": 3.8549075391180653e-07,
350
  "logits/generated": -Infinity,
351
  "logits/real": -Infinity,
352
+ "logps/generated": -705.4946899414062,
353
+ "logps/real": -414.25146484375,
354
+ "loss": 0.4719,
355
  "rewards/accuracies": 0.925000011920929,
356
+ "rewards/generated": -40.05225372314453,
357
+ "rewards/margins": 24.665658950805664,
358
+ "rewards/real": -15.386594772338867,
359
  "step": 240
360
  },
361
  {
 
363
  "learning_rate": 3.783783783783784e-07,
364
  "logits/generated": -Infinity,
365
  "logits/real": -Infinity,
366
+ "logps/generated": -718.469970703125,
367
+ "logps/real": -393.9696350097656,
368
+ "loss": 0.326,
369
  "rewards/accuracies": 0.9624999761581421,
370
+ "rewards/generated": -43.078765869140625,
371
+ "rewards/margins": 28.980609893798828,
372
+ "rewards/real": -14.09815788269043,
373
  "step": 250
374
  },
375
  {
 
377
  "learning_rate": 3.7126600284495016e-07,
378
  "logits/generated": -Infinity,
379
  "logits/real": -Infinity,
380
+ "logps/generated": -721.9268798828125,
381
+ "logps/real": -403.2998046875,
382
+ "loss": 0.4804,
383
  "rewards/accuracies": 0.9375,
384
+ "rewards/generated": -41.29940414428711,
385
+ "rewards/margins": 26.352609634399414,
386
+ "rewards/real": -14.946797370910645,
387
  "step": 260
388
  },
389
  {
 
391
  "learning_rate": 3.6415362731152206e-07,
392
  "logits/generated": -Infinity,
393
  "logits/real": -Infinity,
394
+ "logps/generated": -690.9553833007812,
395
+ "logps/real": -374.5037536621094,
396
+ "loss": 0.3827,
397
  "rewards/accuracies": 0.949999988079071,
398
+ "rewards/generated": -40.69682693481445,
399
+ "rewards/margins": 26.715112686157227,
400
+ "rewards/real": -13.981710433959961,
401
  "step": 270
402
  },
403
  {
 
405
  "learning_rate": 3.5704125177809385e-07,
406
  "logits/generated": -Infinity,
407
  "logits/real": -Infinity,
408
+ "logps/generated": -725.0104370117188,
409
+ "logps/real": -406.488037109375,
410
+ "loss": 0.4101,
411
  "rewards/accuracies": 0.9624999761581421,
412
+ "rewards/generated": -42.66962432861328,
413
+ "rewards/margins": 27.242467880249023,
414
+ "rewards/real": -15.427154541015625,
415
  "step": 280
416
  },
417
  {
 
419
  "learning_rate": 3.4992887624466574e-07,
420
  "logits/generated": -Infinity,
421
  "logits/real": -Infinity,
422
+ "logps/generated": -697.3206787109375,
423
+ "logps/real": -400.65753173828125,
424
+ "loss": 0.3629,
425
  "rewards/accuracies": 0.9624999761581421,
426
+ "rewards/generated": -39.595970153808594,
427
+ "rewards/margins": 24.689983367919922,
428
+ "rewards/real": -14.90599250793457,
429
  "step": 290
430
  },
431
  {
 
433
  "learning_rate": 3.4281650071123753e-07,
434
  "logits/generated": -Infinity,
435
  "logits/real": -Infinity,
436
+ "logps/generated": -711.7612915039062,
437
+ "logps/real": -388.0057678222656,
438
+ "loss": 0.2122,
439
  "rewards/accuracies": 0.949999988079071,
440
+ "rewards/generated": -42.39947509765625,
441
+ "rewards/margins": 27.109222412109375,
442
+ "rewards/real": -15.290254592895508,
443
  "step": 300
444
  },
445
  {
 
447
  "learning_rate": 3.3570412517780937e-07,
448
  "logits/generated": -Infinity,
449
  "logits/real": -Infinity,
450
+ "logps/generated": -713.0751953125,
451
+ "logps/real": -423.7640686035156,
452
+ "loss": 0.3657,
453
+ "rewards/accuracies": 0.9125000238418579,
454
+ "rewards/generated": -41.861183166503906,
455
+ "rewards/margins": 26.213199615478516,
456
+ "rewards/real": -15.647982597351074,
457
  "step": 310
458
  },
459
  {
 
461
  "learning_rate": 3.285917496443812e-07,
462
  "logits/generated": -Infinity,
463
  "logits/real": -Infinity,
464
+ "logps/generated": -695.4622192382812,
465
+ "logps/real": -386.4324645996094,
466
+ "loss": 0.3005,
467
  "rewards/accuracies": 0.925000011920929,
468
+ "rewards/generated": -39.45234298706055,
469
+ "rewards/margins": 25.696590423583984,
470
+ "rewards/real": -13.755752563476562,
471
  "step": 320
472
  },
473
  {
 
475
  "learning_rate": 3.2147937411095305e-07,
476
  "logits/generated": -Infinity,
477
  "logits/real": -Infinity,
478
+ "logps/generated": -774.9369506835938,
479
+ "logps/real": -426.2724609375,
480
+ "loss": 0.2325,
481
+ "rewards/accuracies": 0.9624999761581421,
482
+ "rewards/generated": -48.17813491821289,
483
+ "rewards/margins": 31.29937171936035,
484
+ "rewards/real": -16.87876319885254,
485
  "step": 330
486
  },
487
  {
 
489
  "learning_rate": 3.1436699857752484e-07,
490
  "logits/generated": -Infinity,
491
  "logits/real": -Infinity,
492
+ "logps/generated": -714.9234008789062,
493
+ "logps/real": -395.0069885253906,
494
+ "loss": 0.2981,
495
  "rewards/accuracies": 0.9624999761581421,
496
+ "rewards/generated": -41.547645568847656,
497
+ "rewards/margins": 27.00480079650879,
498
+ "rewards/real": -14.54284381866455,
499
  "step": 340
500
  },
501
  {
 
503
  "learning_rate": 3.0725462304409674e-07,
504
  "logits/generated": -Infinity,
505
  "logits/real": -Infinity,
506
+ "logps/generated": -721.8656005859375,
507
+ "logps/real": -400.1606750488281,
508
+ "loss": 0.2019,
509
  "rewards/accuracies": 0.9375,
510
+ "rewards/generated": -44.031803131103516,
511
+ "rewards/margins": 28.76715087890625,
512
+ "rewards/real": -15.264646530151367,
513
  "step": 350
514
  },
515
  {
 
517
  "learning_rate": 3.001422475106685e-07,
518
  "logits/generated": -Infinity,
519
  "logits/real": -Infinity,
520
+ "logps/generated": -728.89599609375,
521
+ "logps/real": -470.1971130371094,
522
+ "loss": 0.5931,
523
+ "rewards/accuracies": 0.9375,
524
+ "rewards/generated": -43.573814392089844,
525
+ "rewards/margins": 24.9804630279541,
526
+ "rewards/real": -18.593351364135742,
527
  "step": 360
528
  },
529
  {
 
531
  "learning_rate": 2.9302987197724037e-07,
532
  "logits/generated": -Infinity,
533
  "logits/real": -Infinity,
534
+ "logps/generated": -667.4353637695312,
535
+ "logps/real": -369.68328857421875,
536
+ "loss": 0.4031,
537
  "rewards/accuracies": 0.9624999761581421,
538
+ "rewards/generated": -38.71464920043945,
539
+ "rewards/margins": 24.013246536254883,
540
+ "rewards/real": -14.701400756835938,
541
  "step": 370
542
  },
543
  {
 
545
  "learning_rate": 2.8591749644381226e-07,
546
  "logits/generated": -Infinity,
547
  "logits/real": -Infinity,
548
+ "logps/generated": -708.6209716796875,
549
+ "logps/real": -395.7604675292969,
550
+ "loss": 0.3691,
551
+ "rewards/accuracies": 0.9375,
552
+ "rewards/generated": -42.240318298339844,
553
+ "rewards/margins": 26.32097816467285,
554
+ "rewards/real": -15.919347763061523,
555
  "step": 380
556
  },
557
  {
 
559
  "learning_rate": 2.7880512091038405e-07,
560
  "logits/generated": -Infinity,
561
  "logits/real": -Infinity,
562
+ "logps/generated": -766.4454345703125,
563
+ "logps/real": -423.30755615234375,
564
+ "loss": 0.2758,
565
  "rewards/accuracies": 0.9624999761581421,
566
+ "rewards/generated": -47.435791015625,
567
+ "rewards/margins": 30.477214813232422,
568
+ "rewards/real": -16.958572387695312,
569
  "step": 390
570
  },
571
  {
 
573
  "learning_rate": 2.716927453769559e-07,
574
  "logits/generated": -Infinity,
575
  "logits/real": -Infinity,
576
+ "logps/generated": -768.6455078125,
577
+ "logps/real": -449.0997009277344,
578
+ "loss": 0.3455,
579
  "rewards/accuracies": 0.949999988079071,
580
+ "rewards/generated": -47.01398849487305,
581
+ "rewards/margins": 28.302906036376953,
582
+ "rewards/real": -18.711084365844727,
583
  "step": 400
584
  },
585
  {
 
587
  "learning_rate": 2.6458036984352773e-07,
588
  "logits/generated": -Infinity,
589
  "logits/real": -Infinity,
590
+ "logps/generated": -734.2765502929688,
591
+ "logps/real": -447.5357971191406,
592
+ "loss": 0.375,
593
+ "rewards/accuracies": 0.9375,
594
+ "rewards/generated": -46.2476921081543,
595
+ "rewards/margins": 27.413455963134766,
596
+ "rewards/real": -18.834239959716797,
597
  "step": 410
598
  },
599
  {
 
601
  "learning_rate": 2.574679943100996e-07,
602
  "logits/generated": -Infinity,
603
  "logits/real": -Infinity,
604
+ "logps/generated": -730.5106201171875,
605
+ "logps/real": -453.9371643066406,
606
+ "loss": 0.0816,
607
+ "rewards/accuracies": 0.925000011920929,
608
+ "rewards/generated": -45.583648681640625,
609
+ "rewards/margins": 27.334802627563477,
610
+ "rewards/real": -18.248844146728516,
611
  "step": 420
612
  },
613
  {
 
615
  "learning_rate": 2.5035561877667136e-07,
616
  "logits/generated": -Infinity,
617
  "logits/real": -Infinity,
618
+ "logps/generated": -766.5603637695312,
619
+ "logps/real": -438.89788818359375,
620
+ "loss": 0.1735,
621
  "rewards/accuracies": 0.9750000238418579,
622
+ "rewards/generated": -46.9077262878418,
623
+ "rewards/margins": 29.52130126953125,
624
+ "rewards/real": -17.386432647705078,
625
  "step": 430
626
  },
627
  {
 
629
  "learning_rate": 2.4324324324324326e-07,
630
  "logits/generated": -Infinity,
631
  "logits/real": -Infinity,
632
+ "logps/generated": -781.421630859375,
633
+ "logps/real": -405.1501770019531,
634
+ "loss": 0.3965,
635
  "rewards/accuracies": 0.987500011920929,
636
+ "rewards/generated": -49.6665153503418,
637
+ "rewards/margins": 33.62762451171875,
638
+ "rewards/real": -16.03889274597168,
639
  "step": 440
640
  },
641
  {
 
643
  "learning_rate": 2.3613086770981507e-07,
644
  "logits/generated": -Infinity,
645
  "logits/real": -Infinity,
646
+ "logps/generated": -760.2469482421875,
647
+ "logps/real": -404.4407653808594,
648
+ "loss": 0.4128,
649
  "rewards/accuracies": 0.949999988079071,
650
+ "rewards/generated": -47.4620361328125,
651
+ "rewards/margins": 30.4351863861084,
652
+ "rewards/real": -17.026851654052734,
653
  "step": 450
654
  },
655
  {
 
657
  "learning_rate": 2.290184921763869e-07,
658
  "logits/generated": -Infinity,
659
  "logits/real": -Infinity,
660
+ "logps/generated": -820.1843872070312,
661
+ "logps/real": -420.02587890625,
662
+ "loss": 0.1306,
663
  "rewards/accuracies": 1.0,
664
+ "rewards/generated": -51.192138671875,
665
+ "rewards/margins": 35.38946533203125,
666
+ "rewards/real": -15.8026704788208,
667
  "step": 460
668
  },
669
  {
 
671
  "learning_rate": 2.2190611664295875e-07,
672
  "logits/generated": -Infinity,
673
  "logits/real": -Infinity,
674
+ "logps/generated": -759.2930908203125,
675
+ "logps/real": -438.888916015625,
676
+ "loss": 0.2487,
677
  "rewards/accuracies": 0.925000011920929,
678
+ "rewards/generated": -45.420188903808594,
679
+ "rewards/margins": 27.34377670288086,
680
+ "rewards/real": -18.0764102935791,
681
  "step": 470
682
  },
683
  {
 
685
  "learning_rate": 2.1479374110953057e-07,
686
  "logits/generated": -Infinity,
687
  "logits/real": -Infinity,
688
+ "logps/generated": -777.7584838867188,
689
+ "logps/real": -388.6899108886719,
690
+ "loss": 0.3922,
691
+ "rewards/accuracies": 0.949999988079071,
692
+ "rewards/generated": -48.69025421142578,
693
+ "rewards/margins": 32.126991271972656,
694
+ "rewards/real": -16.563264846801758,
695
  "step": 480
696
  },
697
  {
 
699
  "learning_rate": 2.076813655761024e-07,
700
  "logits/generated": -Infinity,
701
  "logits/real": -Infinity,
702
+ "logps/generated": -804.2822875976562,
703
+ "logps/real": -438.558837890625,
704
+ "loss": 0.2558,
705
  "rewards/accuracies": 0.987500011920929,
706
+ "rewards/generated": -50.539390563964844,
707
+ "rewards/margins": 31.528533935546875,
708
+ "rewards/real": -19.0108585357666,
709
  "step": 490
710
  },
711
  {
 
713
  "learning_rate": 2.0056899004267425e-07,
714
  "logits/generated": -Infinity,
715
  "logits/real": -Infinity,
716
+ "logps/generated": -799.6202392578125,
717
+ "logps/real": -483.55096435546875,
718
+ "loss": 0.2945,
719
  "rewards/accuracies": 0.9624999761581421,
720
+ "rewards/generated": -50.81739807128906,
721
+ "rewards/margins": 30.668746948242188,
722
+ "rewards/real": -20.14865493774414,
723
  "step": 500
724
  },
725
  {
726
  "epoch": 0.64,
727
  "eval_logits/generated": -Infinity,
728
  "eval_logits/real": -Infinity,
729
+ "eval_logps/generated": -509.32861328125,
730
+ "eval_logps/real": -313.0279541015625,
731
+ "eval_loss": 0.1748354285955429,
732
  "eval_rewards/accuracies": 0.9442675113677979,
733
+ "eval_rewards/generated": -21.789350509643555,
734
+ "eval_rewards/margins": 15.725247383117676,
735
+ "eval_rewards/real": -6.0641021728515625,
736
+ "eval_runtime": 590.026,
737
+ "eval_samples_per_second": 8.474,
738
+ "eval_steps_per_second": 0.266,
739
  "step": 500
740
  },
741
  {
 
743
  "learning_rate": 1.9345661450924607e-07,
744
  "logits/generated": -Infinity,
745
  "logits/real": -Infinity,
746
+ "logps/generated": -757.3658447265625,
747
+ "logps/real": -412.8812561035156,
748
+ "loss": 0.2523,
749
+ "rewards/accuracies": 0.9750000238418579,
750
+ "rewards/generated": -45.53099822998047,
751
+ "rewards/margins": 29.076038360595703,
752
+ "rewards/real": -16.454959869384766,
753
  "step": 510
754
  },
755
  {
 
757
  "learning_rate": 1.863442389758179e-07,
758
  "logits/generated": -Infinity,
759
  "logits/real": -Infinity,
760
+ "logps/generated": -780.4212036132812,
761
+ "logps/real": -437.5557556152344,
762
+ "loss": 0.2147,
763
  "rewards/accuracies": 0.987500011920929,
764
+ "rewards/generated": -49.51309585571289,
765
+ "rewards/margins": 32.93183517456055,
766
+ "rewards/real": -16.581256866455078,
767
  "step": 520
768
  },
769
  {
 
771
  "learning_rate": 1.7923186344238975e-07,
772
  "logits/generated": -Infinity,
773
  "logits/real": -Infinity,
774
+ "logps/generated": -735.9564208984375,
775
+ "logps/real": -407.6651306152344,
776
+ "loss": 0.2975,
777
+ "rewards/accuracies": 0.9624999761581421,
778
+ "rewards/generated": -44.98466491699219,
779
+ "rewards/margins": 29.423254013061523,
780
+ "rewards/real": -15.561413764953613,
781
  "step": 530
782
  },
783
  {
 
785
  "learning_rate": 1.721194879089616e-07,
786
  "logits/generated": -Infinity,
787
  "logits/real": -Infinity,
788
+ "logps/generated": -807.4033203125,
789
+ "logps/real": -429.4000549316406,
790
+ "loss": 0.2254,
791
  "rewards/accuracies": 0.949999988079071,
792
+ "rewards/generated": -51.46739959716797,
793
+ "rewards/margins": 32.75578689575195,
794
+ "rewards/real": -18.711612701416016,
795
  "step": 540
796
  },
797
  {
 
799
  "learning_rate": 1.650071123755334e-07,
800
  "logits/generated": -Infinity,
801
  "logits/real": -Infinity,
802
+ "logps/generated": -811.77099609375,
803
+ "logps/real": -425.84649658203125,
804
+ "loss": 0.3105,
805
  "rewards/accuracies": 0.949999988079071,
806
+ "rewards/generated": -52.33929443359375,
807
+ "rewards/margins": 34.459083557128906,
808
+ "rewards/real": -17.88020896911621,
809
  "step": 550
810
  },
811
  {
 
813
  "learning_rate": 1.5789473684210525e-07,
814
  "logits/generated": -Infinity,
815
  "logits/real": -Infinity,
816
+ "logps/generated": -788.242919921875,
817
+ "logps/real": -440.695556640625,
818
+ "loss": 0.1702,
819
  "rewards/accuracies": 0.9624999761581421,
820
+ "rewards/generated": -49.456298828125,
821
+ "rewards/margins": 31.41439437866211,
822
+ "rewards/real": -18.041906356811523,
823
  "step": 560
824
  },
825
  {
 
827
  "learning_rate": 1.507823613086771e-07,
828
  "logits/generated": -Infinity,
829
  "logits/real": -Infinity,
830
+ "logps/generated": -752.26611328125,
831
+ "logps/real": -427.10113525390625,
832
+ "loss": 0.3281,
833
+ "rewards/accuracies": 0.9375,
834
+ "rewards/generated": -47.79071807861328,
835
+ "rewards/margins": 30.950246810913086,
836
+ "rewards/real": -16.84047508239746,
837
  "step": 570
838
  },
839
  {
 
841
  "learning_rate": 1.436699857752489e-07,
842
  "logits/generated": -Infinity,
843
  "logits/real": -Infinity,
844
+ "logps/generated": -826.0186767578125,
845
+ "logps/real": -424.66680908203125,
846
+ "loss": 0.3042,
847
  "rewards/accuracies": 0.9750000238418579,
848
+ "rewards/generated": -51.909454345703125,
849
+ "rewards/margins": 34.6719856262207,
850
+ "rewards/real": -17.237468719482422,
851
  "step": 580
852
  },
853
  {
 
855
  "learning_rate": 1.3655761024182077e-07,
856
  "logits/generated": -Infinity,
857
  "logits/real": -Infinity,
858
+ "logps/generated": -816.61962890625,
859
+ "logps/real": -420.100830078125,
860
+ "loss": 0.2635,
861
  "rewards/accuracies": 0.9624999761581421,
862
+ "rewards/generated": -52.91407012939453,
863
+ "rewards/margins": 36.203983306884766,
864
+ "rewards/real": -16.710086822509766,
865
  "step": 590
866
  },
867
  {
 
869
  "learning_rate": 1.2944523470839261e-07,
870
  "logits/generated": -Infinity,
871
  "logits/real": -Infinity,
872
+ "logps/generated": -820.4493408203125,
873
+ "logps/real": -428.4378967285156,
874
+ "loss": 0.347,
875
+ "rewards/accuracies": 0.9375,
876
+ "rewards/generated": -51.51692581176758,
877
+ "rewards/margins": 34.30039596557617,
878
+ "rewards/real": -17.216527938842773,
879
  "step": 600
880
  },
881
  {
 
883
  "learning_rate": 1.2233285917496443e-07,
884
  "logits/generated": -Infinity,
885
  "logits/real": -Infinity,
886
+ "logps/generated": -768.8092041015625,
887
+ "logps/real": -420.62200927734375,
888
+ "loss": 0.0588,
889
  "rewards/accuracies": 0.987500011920929,
890
+ "rewards/generated": -47.871437072753906,
891
+ "rewards/margins": 32.33845901489258,
892
+ "rewards/real": -15.532976150512695,
893
  "step": 610
894
  },
895
  {
 
897
  "learning_rate": 1.1522048364153626e-07,
898
  "logits/generated": -Infinity,
899
  "logits/real": -Infinity,
900
+ "logps/generated": -786.0013427734375,
901
+ "logps/real": -442.2757873535156,
902
+ "loss": 0.2483,
903
  "rewards/accuracies": 0.9624999761581421,
904
+ "rewards/generated": -50.278663635253906,
905
+ "rewards/margins": 33.101016998291016,
906
+ "rewards/real": -17.17764663696289,
907
  "step": 620
908
  },
909
  {
 
911
  "learning_rate": 1.0810810810810811e-07,
912
  "logits/generated": -Infinity,
913
  "logits/real": -Infinity,
914
+ "logps/generated": -787.63818359375,
915
+ "logps/real": -404.19122314453125,
916
+ "loss": 0.1336,
917
  "rewards/accuracies": 0.9624999761581421,
918
+ "rewards/generated": -50.15456008911133,
919
+ "rewards/margins": 33.99864959716797,
920
+ "rewards/real": -16.15591049194336,
921
  "step": 630
922
  },
923
  {
 
925
  "learning_rate": 1.0099573257467994e-07,
926
  "logits/generated": -Infinity,
927
  "logits/real": -Infinity,
928
+ "logps/generated": -830.99462890625,
929
+ "logps/real": -457.3229064941406,
930
+ "loss": 0.2925,
931
  "rewards/accuracies": 0.949999988079071,
932
+ "rewards/generated": -53.06385040283203,
933
+ "rewards/margins": 34.269439697265625,
934
+ "rewards/real": -18.794404983520508,
935
  "step": 640
936
  },
937
  {
 
939
  "learning_rate": 9.388335704125178e-08,
940
  "logits/generated": -Infinity,
941
  "logits/real": -Infinity,
942
+ "logps/generated": -777.55517578125,
943
+ "logps/real": -405.3326110839844,
944
+ "loss": 0.137,
945
  "rewards/accuracies": 1.0,
946
+ "rewards/generated": -49.94211196899414,
947
+ "rewards/margins": 32.77922821044922,
948
+ "rewards/real": -17.16288948059082,
949
  "step": 650
950
  },
951
  {
 
953
  "learning_rate": 8.677098150782361e-08,
954
  "logits/generated": -Infinity,
955
  "logits/real": -Infinity,
956
+ "logps/generated": -775.7921142578125,
957
+ "logps/real": -408.35302734375,
958
+ "loss": 0.0992,
959
  "rewards/accuracies": 0.987500011920929,
960
+ "rewards/generated": -48.498905181884766,
961
+ "rewards/margins": 32.695289611816406,
962
+ "rewards/real": -15.803617477416992,
963
  "step": 660
964
  },
965
  {
 
967
  "learning_rate": 7.965860597439544e-08,
968
  "logits/generated": -Infinity,
969
  "logits/real": -Infinity,
970
+ "logps/generated": -764.4813232421875,
971
+ "logps/real": -427.140625,
972
+ "loss": 0.2396,
973
  "rewards/accuracies": 0.925000011920929,
974
+ "rewards/generated": -48.443748474121094,
975
+ "rewards/margins": 31.320148468017578,
976
+ "rewards/real": -17.123600006103516,
977
  "step": 670
978
  },
979
  {
 
981
  "learning_rate": 7.254623044096728e-08,
982
  "logits/generated": -Infinity,
983
  "logits/real": -Infinity,
984
+ "logps/generated": -803.4662475585938,
985
+ "logps/real": -435.402099609375,
986
+ "loss": 0.2362,
987
  "rewards/accuracies": 0.987500011920929,
988
+ "rewards/generated": -51.644203186035156,
989
+ "rewards/margins": 33.23621368408203,
990
+ "rewards/real": -18.40799331665039,
991
  "step": 680
992
  },
993
  {
 
995
  "learning_rate": 6.543385490753911e-08,
996
  "logits/generated": -Infinity,
997
  "logits/real": -Infinity,
998
+ "logps/generated": -863.4392700195312,
999
+ "logps/real": -439.1971130371094,
1000
+ "loss": 0.1949,
1001
+ "rewards/accuracies": 0.9750000238418579,
1002
+ "rewards/generated": -55.93394088745117,
1003
+ "rewards/margins": 36.83943176269531,
1004
+ "rewards/real": -19.094507217407227,
1005
  "step": 690
1006
  },
1007
  {
 
1009
  "learning_rate": 5.832147937411095e-08,
1010
  "logits/generated": -Infinity,
1011
  "logits/real": -Infinity,
1012
+ "logps/generated": -779.9309692382812,
1013
+ "logps/real": -411.91455078125,
1014
+ "loss": 0.2666,
1015
+ "rewards/accuracies": 0.9375,
1016
+ "rewards/generated": -47.416011810302734,
1017
+ "rewards/margins": 29.500951766967773,
1018
+ "rewards/real": -17.915063858032227,
1019
  "step": 700
1020
  },
1021
  {
 
1023
  "learning_rate": 5.120910384068278e-08,
1024
  "logits/generated": -Infinity,
1025
  "logits/real": -Infinity,
1026
+ "logps/generated": -826.6921997070312,
1027
+ "logps/real": -447.87762451171875,
1028
+ "loss": 0.258,
1029
  "rewards/accuracies": 0.949999988079071,
1030
+ "rewards/generated": -53.74137496948242,
1031
+ "rewards/margins": 35.463531494140625,
1032
+ "rewards/real": -18.277841567993164,
1033
  "step": 710
1034
  },
1035
  {
 
1037
  "learning_rate": 4.4096728307254624e-08,
1038
  "logits/generated": -Infinity,
1039
  "logits/real": -Infinity,
1040
+ "logps/generated": -834.2398681640625,
1041
+ "logps/real": -427.90966796875,
1042
+ "loss": 0.1272,
1043
  "rewards/accuracies": 0.987500011920929,
1044
+ "rewards/generated": -53.82038497924805,
1045
+ "rewards/margins": 36.249298095703125,
1046
+ "rewards/real": -17.571086883544922,
1047
  "step": 720
1048
  },
1049
  {
 
1051
  "learning_rate": 3.698435277382646e-08,
1052
  "logits/generated": -Infinity,
1053
  "logits/real": -Infinity,
1054
+ "logps/generated": -838.7642822265625,
1055
+ "logps/real": -436.14642333984375,
1056
+ "loss": 0.2078,
1057
+ "rewards/accuracies": 0.9750000238418579,
1058
+ "rewards/generated": -55.14348602294922,
1059
+ "rewards/margins": 37.05797576904297,
1060
+ "rewards/real": -18.08551597595215,
1061
  "step": 730
1062
  },
1063
  {
 
1065
  "learning_rate": 2.9871977240398294e-08,
1066
  "logits/generated": -Infinity,
1067
  "logits/real": -Infinity,
1068
+ "logps/generated": -798.5349731445312,
1069
+ "logps/real": -427.6378479003906,
1070
+ "loss": 0.2621,
1071
  "rewards/accuracies": 0.9624999761581421,
1072
+ "rewards/generated": -52.086219787597656,
1073
+ "rewards/margins": 34.135223388671875,
1074
+ "rewards/real": -17.95099449157715,
1075
  "step": 740
1076
  },
1077
  {
 
1079
  "learning_rate": 2.275960170697013e-08,
1080
  "logits/generated": -Infinity,
1081
  "logits/real": -Infinity,
1082
+ "logps/generated": -835.6724853515625,
1083
+ "logps/real": -414.00152587890625,
1084
+ "loss": 0.1564,
1085
  "rewards/accuracies": 0.9750000238418579,
1086
+ "rewards/generated": -52.597930908203125,
1087
+ "rewards/margins": 36.33824920654297,
1088
+ "rewards/real": -16.259681701660156,
1089
  "step": 750
1090
  },
1091
  {
 
1093
  "learning_rate": 1.564722617354196e-08,
1094
  "logits/generated": -Infinity,
1095
  "logits/real": -Infinity,
1096
+ "logps/generated": -850.904296875,
1097
+ "logps/real": -392.68310546875,
1098
+ "loss": 0.1515,
1099
  "rewards/accuracies": 0.9750000238418579,
1100
+ "rewards/generated": -55.202125549316406,
1101
+ "rewards/margins": 38.95097351074219,
1102
+ "rewards/real": -16.25115394592285,
1103
  "step": 760
1104
  },
1105
  {
 
1107
  "learning_rate": 8.534850640113798e-09,
1108
  "logits/generated": -Infinity,
1109
  "logits/real": -Infinity,
1110
+ "logps/generated": -814.2432250976562,
1111
+ "logps/real": -415.373779296875,
1112
+ "loss": 0.1867,
1113
+ "rewards/accuracies": 0.9624999761581421,
1114
+ "rewards/generated": -53.38776397705078,
1115
+ "rewards/margins": 37.246055603027344,
1116
+ "rewards/real": -16.141704559326172,
1117
  "step": 770
1118
  },
1119
  {
 
1121
  "learning_rate": 1.422475106685633e-09,
1122
  "logits/generated": -Infinity,
1123
  "logits/real": -Infinity,
1124
+ "logps/generated": -811.1026611328125,
1125
+ "logps/real": -388.99835205078125,
1126
+ "loss": 0.2403,
1127
  "rewards/accuracies": 0.9750000238418579,
1128
+ "rewards/generated": -51.92308807373047,
1129
+ "rewards/margins": 35.41986846923828,
1130
+ "rewards/real": -16.503215789794922,
1131
  "step": 780
1132
  },
1133
  {
1134
  "epoch": 1.0,
1135
  "step": 782,
1136
  "total_flos": 0.0,
1137
+ "train_loss": 0.6284351689954493,
1138
+ "train_runtime": 6560.8804,
1139
+ "train_samples_per_second": 3.81,
1140
+ "train_steps_per_second": 0.119
1141
  }
1142
  ],
1143
  "logging_steps": 10,