Hugging Face's logo

trevorbchen
/
diffusion_memorization

Model card Files
xet
 Community
trevorbchen commited on
Commit
f607a93
·
verified ·
1 Parent(s): a6d1ad4

Add files using upload-large-folder tool

Browse files
This view is limited to 50 files because it contains too many changes.   See raw diff
Files changed (50) hide show
  1. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed44/d60/runner.log +502 -0
  2. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed44/d70/config.json +26 -0
  3. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed44/d70/metrics.jsonl +0 -0
  4. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed44/d70/runner.log +502 -0
  5. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed44/d80/config.json +26 -0
  6. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed44/d80/metrics.jsonl +0 -0
  7. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed44/d80/runner.log +502 -0
  8. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed44/d90/config.json +26 -0
  9. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed44/d90/metrics.jsonl +0 -0
  10. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed44/d90/runner.log +502 -0
  11. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d10/config.json +26 -0
  12. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d10/metrics.jsonl +0 -0
  13. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d10/runner.log +502 -0
  14. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d100/config.json +26 -0
  15. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d100/metrics.jsonl +0 -0
  16. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d100/runner.log +502 -0
  17. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d120/config.json +26 -0
  18. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d120/metrics.jsonl +0 -0
  19. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d120/runner.log +502 -0
  20. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d140/config.json +26 -0
  21. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d140/metrics.jsonl +0 -0
  22. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d140/runner.log +502 -0
  23. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d160/config.json +26 -0
  24. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d160/metrics.jsonl +0 -0
  25. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d160/runner.log +502 -0
  26. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d180/config.json +26 -0
  27. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d180/metrics.jsonl +0 -0
  28. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d180/runner.log +502 -0
  29. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d20/config.json +26 -0
  30. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d20/metrics.jsonl +0 -0
  31. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d20/runner.log +502 -0
  32. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d200/config.json +26 -0
  33. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d200/metrics.jsonl +0 -0
  34. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d200/runner.log +502 -0
  35. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d30/config.json +26 -0
  36. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d30/metrics.jsonl +0 -0
  37. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d30/runner.log +502 -0
  38. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d40/config.json +26 -0
  39. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d40/metrics.jsonl +0 -0
  40. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d40/runner.log +502 -0
  41. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d50/config.json +26 -0
  42. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d50/metrics.jsonl +0 -0
  43. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d50/runner.log +502 -0
  44. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d60/config.json +26 -0
  45. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d60/metrics.jsonl +0 -0
  46. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d60/runner.log +502 -0
  47. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d70/config.json +26 -0
  48. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d70/metrics.jsonl +0 -0
  49. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d70/runner.log +502 -0
  50. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d80/config.json +26 -0
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed44/d60/runner.log ADDED
@@ -0,0 +1,502 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ FID real-side stats cached over 1000 diverse train images
2
+ step 1 | 71s | train_loss=0.9909 | fid=299.2898
3
+ step 10000 | 156s | train_loss=0.8307
4
+ step 20000 | 250s | train_loss=0.7545
5
+ step 30000 | 344s | train_loss=0.6456
6
+ step 40000 | 438s | train_loss=0.4892
7
+ step 50000 | 531s | train_loss=0.3340
8
+ step 60000 | 625s | train_loss=0.2378
9
+ step 70000 | 718s | train_loss=0.1880
10
+ step 80000 | 812s | train_loss=0.1610
11
+ step 90000 | 904s | train_loss=0.1441
12
+ step 100000 | 1072s | train_loss=0.1338 | fid=57.9142
13
+ step 110000 | 1153s | train_loss=0.1238
14
+ step 120000 | 1246s | train_loss=0.1167
15
+ step 130000 | 1340s | train_loss=0.1104
16
+ step 140000 | 1433s | train_loss=0.1052
17
+ step 150000 | 1527s | train_loss=0.1005
18
+ step 160000 | 1621s | train_loss=0.0963
19
+ step 170000 | 1715s | train_loss=0.0929
20
+ step 180000 | 1808s | train_loss=0.0885
21
+ step 190000 | 1901s | train_loss=0.0867
22
+ step 200000 | 2067s | train_loss=0.0840 | fid=54.6814
23
+ step 210000 | 2152s | train_loss=0.0802
24
+ step 220000 | 2246s | train_loss=0.0786
25
+ step 230000 | 2340s | train_loss=0.0771
26
+ step 240000 | 2433s | train_loss=0.0743
27
+ step 250000 | 2527s | train_loss=0.0716
28
+ step 260000 | 2621s | train_loss=0.0711
29
+ step 270000 | 2715s | train_loss=0.0686
30
+ step 280000 | 2807s | train_loss=0.0670
31
+ step 290000 | 2900s | train_loss=0.0660
32
+ step 300000 | 3067s | train_loss=0.0637 | fid=53.2268
33
+ step 310000 | 3154s | train_loss=0.0634
34
+ step 320000 | 3248s | train_loss=0.0619
35
+ step 330000 | 3342s | train_loss=0.0607
36
+ step 340000 | 3436s | train_loss=0.0595
37
+ step 350000 | 3530s | train_loss=0.0578
38
+ step 360000 | 3624s | train_loss=0.0577
39
+ step 370000 | 3717s | train_loss=0.0569
40
+ step 380000 | 3810s | train_loss=0.0557
41
+ step 390000 | 3902s | train_loss=0.0539
42
+ step 400000 | 4070s | train_loss=0.0539 | fid=51.6088
43
+ step 410000 | 4158s | train_loss=0.0521
44
+ step 420000 | 4252s | train_loss=0.0520
45
+ step 430000 | 4345s | train_loss=0.0512
46
+ step 440000 | 4439s | train_loss=0.0505
47
+ step 450000 | 4533s | train_loss=0.0495
48
+ step 460000 | 4626s | train_loss=0.0493
49
+ step 470000 | 4720s | train_loss=0.0485
50
+ step 480000 | 4811s | train_loss=0.0483
51
+ step 490000 | 4905s | train_loss=0.0474
52
+ step 500000 | 5072s | train_loss=0.0480 | fid=50.2064
53
+ step 510000 | 5160s | train_loss=0.0465
54
+ step 520000 | 5254s | train_loss=0.0463
55
+ step 530000 | 5348s | train_loss=0.0452
56
+ step 540000 | 5442s | train_loss=0.0454
57
+ step 550000 | 5536s | train_loss=0.0442
58
+ step 560000 | 5629s | train_loss=0.0439
59
+ step 570000 | 5722s | train_loss=0.0433
60
+ step 580000 | 5814s | train_loss=0.0432
61
+ step 590000 | 5908s | train_loss=0.0428
62
+ step 600000 | 6075s | train_loss=0.0420 | fid=49.0724
63
+ step 610000 | 6163s | train_loss=0.0417
64
+ step 620000 | 6257s | train_loss=0.0420
65
+ step 630000 | 6351s | train_loss=0.0411
66
+ step 640000 | 6444s | train_loss=0.0407
67
+ step 650000 | 6538s | train_loss=0.0406
68
+ step 660000 | 6631s | train_loss=0.0400
69
+ step 670000 | 6722s | train_loss=0.0395
70
+ step 680000 | 6816s | train_loss=0.0398
71
+ step 690000 | 6910s | train_loss=0.0387
72
+ step 700000 | 7077s | train_loss=0.0384 | fid=47.4482
73
+ step 710000 | 7165s | train_loss=0.0390
74
+ step 720000 | 7259s | train_loss=0.0391
75
+ step 730000 | 7353s | train_loss=0.0376
76
+ step 740000 | 7447s | train_loss=0.0378
77
+ step 750000 | 7540s | train_loss=0.0370
78
+ step 760000 | 7634s | train_loss=0.0370
79
+ step 770000 | 7726s | train_loss=0.0367
80
+ step 780000 | 7820s | train_loss=0.0375
81
+ step 790000 | 7914s | train_loss=0.0362
82
+ step 800000 | 8083s | train_loss=0.0363 | fid=47.2832
83
+ step 810000 | 8170s | train_loss=0.0350
84
+ step 820000 | 8264s | train_loss=0.0356
85
+ step 830000 | 8358s | train_loss=0.0346
86
+ step 840000 | 8451s | train_loss=0.0347
87
+ step 850000 | 8545s | train_loss=0.0346
88
+ step 860000 | 8637s | train_loss=0.0348
89
+ step 870000 | 8731s | train_loss=0.0339
90
+ step 880000 | 8825s | train_loss=0.0339
91
+ step 890000 | 8918s | train_loss=0.0350
92
+ step 900000 | 9086s | train_loss=0.0339 | fid=46.3076
93
+ step 910000 | 9174s | train_loss=0.0334
94
+ step 920000 | 9267s | train_loss=0.0331
95
+ step 930000 | 9361s | train_loss=0.0333
96
+ step 940000 | 9454s | train_loss=0.0334
97
+ step 950000 | 9548s | train_loss=0.0335
98
+ step 960000 | 9640s | train_loss=0.0327
99
+ step 970000 | 9734s | train_loss=0.0323
100
+ step 980000 | 9828s | train_loss=0.0320
101
+ step 990000 | 9922s | train_loss=0.0315
102
+ step 1000000 | 10091s | train_loss=0.0319 | fid=46.4769
103
+ step 1010000 | 10179s | train_loss=0.0314
104
+ step 1020000 | 10272s | train_loss=0.0314
105
+ step 1030000 | 10365s | train_loss=0.0313
106
+ step 1040000 | 10459s | train_loss=0.0322
107
+ step 1050000 | 10551s | train_loss=0.0302
108
+ step 1060000 | 10644s | train_loss=0.0301
109
+ step 1070000 | 10739s | train_loss=0.0308
110
+ step 1080000 | 10832s | train_loss=0.0303
111
+ step 1090000 | 10926s | train_loss=0.0303
112
+ step 1100000 | 11095s | train_loss=0.0295 | fid=46.0691
113
+ step 1110000 | 11187s | train_loss=0.0299
114
+ step 1120000 | 11276s | train_loss=0.0293
115
+ step 1130000 | 11369s | train_loss=0.0294
116
+ step 1140000 | 11461s | train_loss=0.0296
117
+ step 1150000 | 11554s | train_loss=0.0295
118
+ step 1160000 | 11648s | train_loss=0.0284
119
+ step 1170000 | 11742s | train_loss=0.0288
120
+ step 1180000 | 11836s | train_loss=0.0286
121
+ step 1190000 | 11929s | train_loss=0.0281
122
+ step 1200000 | 12096s | train_loss=0.0284 | fid=45.6723
123
+ step 1210000 | 12189s | train_loss=0.0284
124
+ step 1220000 | 12277s | train_loss=0.0293
125
+ step 1230000 | 12371s | train_loss=0.0282
126
+ step 1240000 | 12463s | train_loss=0.0271
127
+ step 1250000 | 12556s | train_loss=0.0267
128
+ step 1260000 | 12650s | train_loss=0.0273
129
+ step 1270000 | 12744s | train_loss=0.0268
130
+ step 1280000 | 12838s | train_loss=0.0278
131
+ step 1290000 | 12932s | train_loss=0.0263
132
+ step 1300000 | 13100s | train_loss=0.0267 | fid=45.4684
133
+ step 1310000 | 13193s | train_loss=0.0264
134
+ step 1320000 | 13282s | train_loss=0.0261
135
+ step 1330000 | 13374s | train_loss=0.0258
136
+ step 1340000 | 13467s | train_loss=0.0264
137
+ step 1350000 | 13561s | train_loss=0.0257
138
+ step 1360000 | 13654s | train_loss=0.0259
139
+ step 1370000 | 13748s | train_loss=0.0259
140
+ step 1380000 | 13841s | train_loss=0.0259
141
+ step 1390000 | 13935s | train_loss=0.0252
142
+ step 1400000 | 14102s | train_loss=0.0251 | fid=45.4718
143
+ step 1410000 | 14195s | train_loss=0.0258
144
+ step 1420000 | 14285s | train_loss=0.0255
145
+ step 1430000 | 14377s | train_loss=0.0244
146
+ step 1440000 | 14470s | train_loss=0.0252
147
+ step 1450000 | 14564s | train_loss=0.0250
148
+ step 1460000 | 14658s | train_loss=0.0244
149
+ step 1470000 | 14752s | train_loss=0.0243
150
+ step 1480000 | 14846s | train_loss=0.0245
151
+ step 1490000 | 14939s | train_loss=0.0251
152
+ step 1500000 | 15106s | train_loss=0.0242 | fid=45.2966
153
+ step 1510000 | 15199s | train_loss=0.0234
154
+ step 1520000 | 15287s | train_loss=0.0240
155
+ step 1530000 | 15380s | train_loss=0.0239
156
+ step 1540000 | 15474s | train_loss=0.0241
157
+ step 1550000 | 15567s | train_loss=0.0240
158
+ step 1560000 | 15661s | train_loss=0.0230
159
+ step 1570000 | 15755s | train_loss=0.0234
160
+ step 1580000 | 15849s | train_loss=0.0241
161
+ step 1590000 | 15943s | train_loss=0.0236
162
+ step 1600000 | 16111s | train_loss=0.0228 | fid=44.2732
163
+ step 1610000 | 16202s | train_loss=0.0232
164
+ step 1620000 | 16292s | train_loss=0.0230
165
+ step 1630000 | 16386s | train_loss=0.0229
166
+ step 1640000 | 16479s | train_loss=0.0221
167
+ step 1650000 | 16573s | train_loss=0.0226
168
+ step 1660000 | 16667s | train_loss=0.0227
169
+ step 1670000 | 16760s | train_loss=0.0223
170
+ step 1680000 | 16854s | train_loss=0.0224
171
+ step 1690000 | 16948s | train_loss=0.0220
172
+ step 1700000 | 17115s | train_loss=0.0219 | fid=45.2342
173
+ step 1710000 | 17206s | train_loss=0.0220
174
+ step 1720000 | 17297s | train_loss=0.0214
175
+ step 1730000 | 17391s | train_loss=0.0219
176
+ step 1740000 | 17484s | train_loss=0.0213
177
+ step 1750000 | 17577s | train_loss=0.0221
178
+ step 1760000 | 17671s | train_loss=0.0219
179
+ step 1770000 | 17765s | train_loss=0.0215
180
+ step 1780000 | 17859s | train_loss=0.0224
181
+ step 1790000 | 17953s | train_loss=0.0214
182
+ step 1800000 | 18118s | train_loss=0.0211 | fid=44.7134
183
+ step 1810000 | 18211s | train_loss=0.0207
184
+ step 1820000 | 18302s | train_loss=0.0207
185
+ step 1830000 | 18395s | train_loss=0.0216
186
+ step 1840000 | 18488s | train_loss=0.0206
187
+ step 1850000 | 18582s | train_loss=0.0210
188
+ step 1860000 | 18676s | train_loss=0.0205
189
+ step 1870000 | 18770s | train_loss=0.0210
190
+ step 1880000 | 18863s | train_loss=0.0199
191
+ step 1890000 | 18957s | train_loss=0.0213
192
+ step 1900000 | 19124s | train_loss=0.0197 | fid=44.1733
193
+ step 1910000 | 19216s | train_loss=0.0203
194
+ step 1920000 | 19307s | train_loss=0.0201
195
+ step 1930000 | 19401s | train_loss=0.0207
196
+ step 1940000 | 19495s | train_loss=0.0194
197
+ step 1950000 | 19589s | train_loss=0.0197
198
+ step 1960000 | 19682s | train_loss=0.0202
199
+ step 1970000 | 19776s | train_loss=0.0192
200
+ step 1980000 | 19870s | train_loss=0.0200
201
+ step 1990000 | 19964s | train_loss=0.0199
202
+ step 2000000 | 20130s | train_loss=0.0197 | fid=44.6527
203
+ step 2010000 | 20223s | train_loss=0.0193
204
+ step 2020000 | 20313s | train_loss=0.0191
205
+ step 2030000 | 20406s | train_loss=0.0202
206
+ step 2040000 | 20500s | train_loss=0.0191
207
+ step 2050000 | 20593s | train_loss=0.0193
208
+ step 2060000 | 20687s | train_loss=0.0188
209
+ step 2070000 | 20781s | train_loss=0.0189
210
+ step 2080000 | 20874s | train_loss=0.0188
211
+ step 2090000 | 20966s | train_loss=0.0190
212
+ step 2100000 | 21134s | train_loss=0.0184 | fid=44.9874
213
+ step 2110000 | 21227s | train_loss=0.0182
214
+ step 2120000 | 21317s | train_loss=0.0197
215
+ step 2130000 | 21411s | train_loss=0.0190
216
+ step 2140000 | 21505s | train_loss=0.0192
217
+ step 2150000 | 21598s | train_loss=0.0184
218
+ step 2160000 | 21692s | train_loss=0.0181
219
+ step 2170000 | 21786s | train_loss=0.0193
220
+ step 2180000 | 21879s | train_loss=0.0185
221
+ step 2190000 | 21971s | train_loss=0.0188
222
+ step 2200000 | 22140s | train_loss=0.0183 | fid=44.0390
223
+ step 2210000 | 22233s | train_loss=0.0187
224
+ step 2220000 | 22324s | train_loss=0.0181
225
+ step 2230000 | 22417s | train_loss=0.0180
226
+ step 2240000 | 22511s | train_loss=0.0177
227
+ step 2250000 | 22604s | train_loss=0.0175
228
+ step 2260000 | 22697s | train_loss=0.0175
229
+ step 2270000 | 22791s | train_loss=0.0179
230
+ step 2280000 | 22882s | train_loss=0.0171
231
+ step 2290000 | 22976s | train_loss=0.0178
232
+ step 2300000 | 23145s | train_loss=0.0172 | fid=44.5845
233
+ step 2310000 | 23237s | train_loss=0.0171
234
+ step 2320000 | 23331s | train_loss=0.0170
235
+ step 2330000 | 23422s | train_loss=0.0177
236
+ step 2340000 | 23516s | train_loss=0.0168
237
+ step 2350000 | 23609s | train_loss=0.0168
238
+ step 2360000 | 23702s | train_loss=0.0165
239
+ step 2370000 | 23796s | train_loss=0.0169
240
+ step 2380000 | 23888s | train_loss=0.0172
241
+ step 2390000 | 23982s | train_loss=0.0167
242
+ step 2400000 | 24150s | train_loss=0.0166 | fid=44.3244
243
+ step 2410000 | 24242s | train_loss=0.0168
244
+ step 2420000 | 24337s | train_loss=0.0161
245
+ step 2430000 | 24427s | train_loss=0.0167
246
+ step 2440000 | 24521s | train_loss=0.0165
247
+ step 2450000 | 24614s | train_loss=0.0167
248
+ step 2460000 | 24708s | train_loss=0.0165
249
+ step 2470000 | 24799s | train_loss=0.0164
250
+ step 2480000 | 24893s | train_loss=0.0162
251
+ step 2490000 | 24987s | train_loss=0.0165
252
+ step 2500000 | 25156s | train_loss=0.0165 | fid=44.1289
253
+ step 2510000 | 25248s | train_loss=0.0161
254
+ step 2520000 | 25342s | train_loss=0.0159
255
+ step 2530000 | 25432s | train_loss=0.0154
256
+ step 2540000 | 25525s | train_loss=0.0158
257
+ step 2550000 | 25619s | train_loss=0.0162
258
+ step 2560000 | 25711s | train_loss=0.0160
259
+ step 2570000 | 25804s | train_loss=0.0160
260
+ step 2580000 | 25897s | train_loss=0.0152
261
+ step 2590000 | 25991s | train_loss=0.0158
262
+ step 2600000 | 26159s | train_loss=0.0160 | fid=44.7198
263
+ step 2610000 | 26252s | train_loss=0.0162
264
+ step 2620000 | 26346s | train_loss=0.0157
265
+ step 2630000 | 26437s | train_loss=0.0161
266
+ step 2640000 | 26530s | train_loss=0.0159
267
+ step 2650000 | 26624s | train_loss=0.0152
268
+ step 2660000 | 26716s | train_loss=0.0152
269
+ step 2670000 | 26809s | train_loss=0.0150
270
+ step 2680000 | 26903s | train_loss=0.0161
271
+ step 2690000 | 26997s | train_loss=0.0149
272
+ step 2700000 | 27165s | train_loss=0.0152 | fid=43.6736
273
+ step 2710000 | 27257s | train_loss=0.0148
274
+ step 2720000 | 27352s | train_loss=0.0153
275
+ step 2730000 | 27442s | train_loss=0.0150
276
+ step 2740000 | 27536s | train_loss=0.0144
277
+ step 2750000 | 27628s | train_loss=0.0147
278
+ step 2760000 | 27721s | train_loss=0.0147
279
+ step 2770000 | 27815s | train_loss=0.0145
280
+ step 2780000 | 27909s | train_loss=0.0150
281
+ step 2790000 | 28003s | train_loss=0.0141
282
+ step 2800000 | 28171s | train_loss=0.0144 | fid=44.6194
283
+ step 2810000 | 28264s | train_loss=0.0143
284
+ step 2820000 | 28358s | train_loss=0.0144
285
+ step 2830000 | 28448s | train_loss=0.0147
286
+ step 2840000 | 28542s | train_loss=0.0142
287
+ step 2850000 | 28634s | train_loss=0.0147
288
+ step 2860000 | 28728s | train_loss=0.0141
289
+ step 2870000 | 28822s | train_loss=0.0141
290
+ step 2880000 | 28915s | train_loss=0.0139
291
+ step 2890000 | 29009s | train_loss=0.0142
292
+ step 2900000 | 29177s | train_loss=0.0144 | fid=44.4455
293
+ step 2910000 | 29269s | train_loss=0.0136
294
+ step 2920000 | 29363s | train_loss=0.0137
295
+ step 2930000 | 29453s | train_loss=0.0142
296
+ step 2940000 | 29545s | train_loss=0.0138
297
+ step 2950000 | 29639s | train_loss=0.0136
298
+ step 2960000 | 29733s | train_loss=0.0139
299
+ step 2970000 | 29826s | train_loss=0.0148
300
+ step 2980000 | 29920s | train_loss=0.0136
301
+ step 2990000 | 30013s | train_loss=0.0144
302
+ step 3000000 | 30182s | train_loss=0.0136 | fid=44.5454
303
+ step 3010000 | 30274s | train_loss=0.0137
304
+ step 3020000 | 30368s | train_loss=0.0133
305
+ step 3030000 | 30457s | train_loss=0.0137
306
+ step 3040000 | 30551s | train_loss=0.0132
307
+ step 3050000 | 30645s | train_loss=0.0137
308
+ step 3060000 | 30739s | train_loss=0.0136
309
+ step 3070000 | 30832s | train_loss=0.0134
310
+ step 3080000 | 30926s | train_loss=0.0135
311
+ step 3090000 | 31019s | train_loss=0.0134
312
+ step 3100000 | 31187s | train_loss=0.0133 | fid=43.9052
313
+ step 3110000 | 31279s | train_loss=0.0136
314
+ step 3120000 | 31373s | train_loss=0.0136
315
+ step 3130000 | 31459s | train_loss=0.0136
316
+ step 3140000 | 31553s | train_loss=0.0133
317
+ step 3150000 | 31647s | train_loss=0.0132
318
+ step 3160000 | 31740s | train_loss=0.0129
319
+ step 3170000 | 31834s | train_loss=0.0137
320
+ step 3180000 | 31928s | train_loss=0.0127
321
+ step 3190000 | 32021s | train_loss=0.0134
322
+ step 3200000 | 32187s | train_loss=0.0125 | fid=44.1817
323
+ step 3210000 | 32280s | train_loss=0.0126
324
+ step 3220000 | 32372s | train_loss=0.0126
325
+ step 3230000 | 32463s | train_loss=0.0132
326
+ step 3240000 | 32557s | train_loss=0.0129
327
+ step 3250000 | 32651s | train_loss=0.0124
328
+ step 3260000 | 32745s | train_loss=0.0134
329
+ step 3270000 | 32838s | train_loss=0.0130
330
+ step 3280000 | 32931s | train_loss=0.0129
331
+ step 3290000 | 33025s | train_loss=0.0125
332
+ step 3300000 | 33193s | train_loss=0.0123 | fid=44.7283
333
+ step 3310000 | 33286s | train_loss=0.0122
334
+ step 3320000 | 33378s | train_loss=0.0128
335
+ step 3330000 | 33468s | train_loss=0.0125
336
+ step 3340000 | 33561s | train_loss=0.0123
337
+ step 3350000 | 33655s | train_loss=0.0128
338
+ step 3360000 | 33749s | train_loss=0.0121
339
+ step 3370000 | 33842s | train_loss=0.0124
340
+ step 3380000 | 33936s | train_loss=0.0120
341
+ step 3390000 | 34029s | train_loss=0.0122
342
+ step 3400000 | 34197s | train_loss=0.0119 | fid=44.3725
343
+ step 3410000 | 34288s | train_loss=0.0123
344
+ step 3420000 | 34381s | train_loss=0.0123
345
+ step 3430000 | 34472s | train_loss=0.0119
346
+ step 3440000 | 34566s | train_loss=0.0122
347
+ step 3450000 | 34660s | train_loss=0.0118
348
+ step 3460000 | 34753s | train_loss=0.0121
349
+ step 3470000 | 34847s | train_loss=0.0119
350
+ step 3480000 | 34941s | train_loss=0.0119
351
+ step 3490000 | 35034s | train_loss=0.0122
352
+ step 3500000 | 35201s | train_loss=0.0124 | fid=44.5385
353
+ step 3510000 | 35291s | train_loss=0.0117
354
+ step 3520000 | 35384s | train_loss=0.0119
355
+ step 3530000 | 35475s | train_loss=0.0115
356
+ step 3540000 | 35569s | train_loss=0.0121
357
+ step 3550000 | 35663s | train_loss=0.0118
358
+ step 3560000 | 35757s | train_loss=0.0119
359
+ step 3570000 | 35850s | train_loss=0.0118
360
+ step 3580000 | 35944s | train_loss=0.0122
361
+ step 3590000 | 36037s | train_loss=0.0112
362
+ step 3600000 | 36202s | train_loss=0.0115 | fid=44.4538
363
+ step 3610000 | 36294s | train_loss=0.0122
364
+ step 3620000 | 36388s | train_loss=0.0119
365
+ step 3630000 | 36479s | train_loss=0.0117
366
+ step 3640000 | 36572s | train_loss=0.0114
367
+ step 3650000 | 36666s | train_loss=0.0115
368
+ step 3660000 | 36760s | train_loss=0.0112
369
+ step 3670000 | 36854s | train_loss=0.0108
370
+ step 3680000 | 36947s | train_loss=0.0115
371
+ step 3690000 | 37041s | train_loss=0.0108
372
+ step 3700000 | 37205s | train_loss=0.0111 | fid=44.2348
373
+ step 3710000 | 37297s | train_loss=0.0112
374
+ step 3720000 | 37391s | train_loss=0.0111
375
+ step 3730000 | 37482s | train_loss=0.0114
376
+ step 3740000 | 37575s | train_loss=0.0114
377
+ step 3750000 | 37669s | train_loss=0.0118
378
+ step 3760000 | 37762s | train_loss=0.0110
379
+ step 3770000 | 37856s | train_loss=0.0114
380
+ step 3780000 | 37949s | train_loss=0.0108
381
+ step 3790000 | 38043s | train_loss=0.0110
382
+ step 3800000 | 38208s | train_loss=0.0106 | fid=43.9729
383
+ step 3810000 | 38300s | train_loss=0.0107
384
+ step 3820000 | 38393s | train_loss=0.0108
385
+ step 3830000 | 38486s | train_loss=0.0111
386
+ step 3840000 | 38578s | train_loss=0.0106
387
+ step 3850000 | 38671s | train_loss=0.0109
388
+ step 3860000 | 38765s | train_loss=0.0106
389
+ step 3870000 | 38858s | train_loss=0.0106
390
+ step 3880000 | 38952s | train_loss=0.0109
391
+ step 3890000 | 39043s | train_loss=0.0112
392
+ step 3900000 | 39210s | train_loss=0.0109 | fid=44.4597
393
+ step 3910000 | 39302s | train_loss=0.0109
394
+ step 3920000 | 39395s | train_loss=0.0105
395
+ step 3930000 | 39488s | train_loss=0.0105
396
+ step 3940000 | 39579s | train_loss=0.0104
397
+ step 3950000 | 39673s | train_loss=0.0106
398
+ step 3960000 | 39766s | train_loss=0.0105
399
+ step 3970000 | 39860s | train_loss=0.0108
400
+ step 3980000 | 39953s | train_loss=0.0104
401
+ step 3990000 | 40045s | train_loss=0.0104
402
+ step 4000000 | 40212s | train_loss=0.0110 | fid=44.3524
403
+ step 4010000 | 40303s | train_loss=0.0102
404
+ step 4020000 | 40397s | train_loss=0.0111
405
+ step 4030000 | 40490s | train_loss=0.0099
406
+ step 4040000 | 40581s | train_loss=0.0104
407
+ step 4050000 | 40674s | train_loss=0.0107
408
+ step 4060000 | 40768s | train_loss=0.0104
409
+ step 4070000 | 40861s | train_loss=0.0100
410
+ step 4080000 | 40952s | train_loss=0.0103
411
+ step 4090000 | 41046s | train_loss=0.0101
412
+ step 4100000 | 41214s | train_loss=0.0105 | fid=43.9690
413
+ step 4110000 | 41306s | train_loss=0.0097
414
+ step 4120000 | 41400s | train_loss=0.0100
415
+ step 4130000 | 41493s | train_loss=0.0100
416
+ step 4140000 | 41583s | train_loss=0.0104
417
+ step 4150000 | 41677s | train_loss=0.0103
418
+ step 4160000 | 41771s | train_loss=0.0100
419
+ step 4170000 | 41864s | train_loss=0.0101
420
+ step 4180000 | 41956s | train_loss=0.0105
421
+ step 4190000 | 42049s | train_loss=0.0101
422
+ step 4200000 | 42216s | train_loss=0.0097 | fid=44.5055
423
+ step 4210000 | 42308s | train_loss=0.0101
424
+ step 4220000 | 42402s | train_loss=0.0095
425
+ step 4230000 | 42496s | train_loss=0.0101
426
+ step 4240000 | 42586s | train_loss=0.0100
427
+ step 4250000 | 42679s | train_loss=0.0100
428
+ step 4260000 | 42773s | train_loss=0.0097
429
+ step 4270000 | 42865s | train_loss=0.0098
430
+ step 4280000 | 42958s | train_loss=0.0098
431
+ step 4290000 | 43052s | train_loss=0.0098
432
+ step 4300000 | 43221s | train_loss=0.0098 | fid=44.1862
433
+ step 4310000 | 43312s | train_loss=0.0097
434
+ step 4320000 | 43406s | train_loss=0.0095
435
+ step 4330000 | 43500s | train_loss=0.0099
436
+ step 4340000 | 43590s | train_loss=0.0095
437
+ step 4350000 | 43684s | train_loss=0.0095
438
+ step 4360000 | 43777s | train_loss=0.0099
439
+ step 4370000 | 43869s | train_loss=0.0095
440
+ step 4380000 | 43962s | train_loss=0.0092
441
+ step 4390000 | 44056s | train_loss=0.0097
442
+ step 4400000 | 44224s | train_loss=0.0095 | fid=43.7924
443
+ step 4410000 | 44316s | train_loss=0.0100
444
+ step 4420000 | 44409s | train_loss=0.0101
445
+ step 4430000 | 44503s | train_loss=0.0099
446
+ step 4440000 | 44593s | train_loss=0.0097
447
+ step 4450000 | 44687s | train_loss=0.0094
448
+ step 4460000 | 44778s | train_loss=0.0099
449
+ step 4470000 | 44872s | train_loss=0.0097
450
+ step 4480000 | 44965s | train_loss=0.0099
451
+ step 4490000 | 45059s | train_loss=0.0096
452
+ step 4500000 | 45226s | train_loss=0.0092 | fid=44.8565
453
+ step 4510000 | 45319s | train_loss=0.0094
454
+ step 4520000 | 45412s | train_loss=0.0091
455
+ step 4530000 | 45506s | train_loss=0.0093
456
+ step 4540000 | 45597s | train_loss=0.0093
457
+ step 4550000 | 45690s | train_loss=0.0094
458
+ step 4560000 | 45782s | train_loss=0.0095
459
+ step 4570000 | 45876s | train_loss=0.0092
460
+ step 4580000 | 45970s | train_loss=0.0094
461
+ step 4590000 | 46063s | train_loss=0.0095
462
+ step 4600000 | 46231s | train_loss=0.0091 | fid=44.1523
463
+ step 4610000 | 46323s | train_loss=0.0088
464
+ step 4620000 | 46416s | train_loss=0.0095
465
+ step 4630000 | 46510s | train_loss=0.0094
466
+ step 4640000 | 46600s | train_loss=0.0099
467
+ step 4650000 | 46692s | train_loss=0.0096
468
+ step 4660000 | 46786s | train_loss=0.0092
469
+ step 4670000 | 46881s | train_loss=0.0091
470
+ step 4680000 | 46974s | train_loss=0.0091
471
+ step 4690000 | 47068s | train_loss=0.0093
472
+ step 4700000 | 47236s | train_loss=0.0090 | fid=44.1023
473
+ step 4710000 | 47329s | train_loss=0.0091
474
+ step 4720000 | 47423s | train_loss=0.0096
475
+ step 4730000 | 47516s | train_loss=0.0092
476
+ step 4740000 | 47602s | train_loss=0.0094
477
+ step 4750000 | 47678s | train_loss=0.0088
478
+ step 4760000 | 47754s | train_loss=0.0092
479
+ step 4770000 | 47830s | train_loss=0.0096
480
+ step 4780000 | 47905s | train_loss=0.0096
481
+ step 4790000 | 47981s | train_loss=0.0090
482
+ step 4800000 | 48119s | train_loss=0.0088 | fid=44.2178
483
+ step 4810000 | 48193s | train_loss=0.0090
484
+ step 4820000 | 48269s | train_loss=0.0096
485
+ step 4830000 | 48345s | train_loss=0.0091
486
+ step 4840000 | 48418s | train_loss=0.0092
487
+ step 4850000 | 48492s | train_loss=0.0094
488
+ step 4860000 | 48568s | train_loss=0.0087
489
+ step 4870000 | 48644s | train_loss=0.0087
490
+ step 4880000 | 48720s | train_loss=0.0086
491
+ step 4890000 | 48796s | train_loss=0.0091
492
+ step 4900000 | 48932s | train_loss=0.0085 | fid=43.7390
493
+ step 4910000 | 49006s | train_loss=0.0088
494
+ step 4920000 | 49083s | train_loss=0.0086
495
+ step 4930000 | 49159s | train_loss=0.0085
496
+ step 4940000 | 49233s | train_loss=0.0088
497
+ step 4950000 | 49305s | train_loss=0.0088
498
+ step 4960000 | 49381s | train_loss=0.0087
499
+ step 4970000 | 49457s | train_loss=0.0085
500
+ step 4980000 | 49533s | train_loss=0.0082
501
+ step 4990000 | 49608s | train_loss=0.0086
502
+ step 5000000 | 49745s | train_loss=0.0091 | fid=44.5720
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed44/d70/config.json ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "vae_checkpoint": "vae_checkpoints/celeba_resnet_modernloss_d70/vae.pt",
3
+ "out": "results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed44/d70",
4
+ "subset_path": "diagnostics/celeba_diverse_1k_mean_plus_sd/subset_indices.json",
5
+ "train_tar": "/overflow/data/cerberus/celeba_hq_256_partitioned_ws/train/celeba_hq_256-train-0000.tar",
6
+ "n_train": 1000,
7
+ "total_steps": 5000000,
8
+ "eval_interval": 10000,
9
+ "mem_interval": 100000,
10
+ "fid_interval": 100000,
11
+ "fid_n_real": 1000,
12
+ "fid_n_gen": 10000,
13
+ "n_gen_samples": 10000,
14
+ "n_sde_steps": 500,
15
+ "batch_size": 512,
16
+ "hidden": 1024,
17
+ "depth": 5,
18
+ "optimizer": "sgd",
19
+ "lr": 0.001,
20
+ "momentum": 0.8,
21
+ "t_min": 0.01,
22
+ "t_max": 3.0,
23
+ "t_eval": 0.1,
24
+ "seed": 44,
25
+ "device": null
26
+ }
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed44/d70/metrics.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed44/d70/runner.log ADDED
@@ -0,0 +1,502 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ FID real-side stats cached over 1000 diverse train images
2
+ step 1 | 74s | train_loss=0.9936 | fid=282.2327
3
+ step 10000 | 168s | train_loss=0.8238
4
+ step 20000 | 262s | train_loss=0.7326
5
+ step 30000 | 355s | train_loss=0.6079
6
+ step 40000 | 449s | train_loss=0.4422
7
+ step 50000 | 542s | train_loss=0.3002
8
+ step 60000 | 636s | train_loss=0.2197
9
+ step 70000 | 730s | train_loss=0.1762
10
+ step 80000 | 823s | train_loss=0.1539
11
+ step 90000 | 916s | train_loss=0.1389
12
+ step 100000 | 1082s | train_loss=0.1291 | fid=56.3282
13
+ step 110000 | 1175s | train_loss=0.1215
14
+ step 120000 | 1268s | train_loss=0.1143
15
+ step 130000 | 1362s | train_loss=0.1082
16
+ step 140000 | 1456s | train_loss=0.1045
17
+ step 150000 | 1549s | train_loss=0.0984
18
+ step 160000 | 1643s | train_loss=0.0965
19
+ step 170000 | 1737s | train_loss=0.0918
20
+ step 180000 | 1830s | train_loss=0.0883
21
+ step 190000 | 1922s | train_loss=0.0853
22
+ step 200000 | 2090s | train_loss=0.0832 | fid=56.3814
23
+ step 210000 | 2181s | train_loss=0.0804
24
+ step 220000 | 2275s | train_loss=0.0777
25
+ step 230000 | 2369s | train_loss=0.0754
26
+ step 240000 | 2462s | train_loss=0.0738
27
+ step 250000 | 2556s | train_loss=0.0719
28
+ step 260000 | 2650s | train_loss=0.0711
29
+ step 270000 | 2744s | train_loss=0.0683
30
+ step 280000 | 2835s | train_loss=0.0672
31
+ step 290000 | 2927s | train_loss=0.0648
32
+ step 300000 | 3095s | train_loss=0.0634 | fid=53.7341
33
+ step 310000 | 3184s | train_loss=0.0629
34
+ step 320000 | 3278s | train_loss=0.0629
35
+ step 330000 | 3372s | train_loss=0.0606
36
+ step 340000 | 3465s | train_loss=0.0592
37
+ step 350000 | 3559s | train_loss=0.0600
38
+ step 360000 | 3653s | train_loss=0.0569
39
+ step 370000 | 3746s | train_loss=0.0575
40
+ step 380000 | 3839s | train_loss=0.0565
41
+ step 390000 | 3931s | train_loss=0.0551
42
+ step 400000 | 4099s | train_loss=0.0541 | fid=50.6200
43
+ step 410000 | 4189s | train_loss=0.0542
44
+ step 420000 | 4283s | train_loss=0.0534
45
+ step 430000 | 4376s | train_loss=0.0524
46
+ step 440000 | 4470s | train_loss=0.0509
47
+ step 450000 | 4564s | train_loss=0.0503
48
+ step 460000 | 4657s | train_loss=0.0496
49
+ step 470000 | 4750s | train_loss=0.0500
50
+ step 480000 | 4842s | train_loss=0.0496
51
+ step 490000 | 4936s | train_loss=0.0486
52
+ step 500000 | 5103s | train_loss=0.0481 | fid=49.4685
53
+ step 510000 | 5192s | train_loss=0.0470
54
+ step 520000 | 5286s | train_loss=0.0467
55
+ step 530000 | 5381s | train_loss=0.0463
56
+ step 540000 | 5474s | train_loss=0.0462
57
+ step 550000 | 5568s | train_loss=0.0452
58
+ step 560000 | 5661s | train_loss=0.0447
59
+ step 570000 | 5754s | train_loss=0.0444
60
+ step 580000 | 5847s | train_loss=0.0445
61
+ step 590000 | 5941s | train_loss=0.0447
62
+ step 600000 | 6108s | train_loss=0.0431 | fid=48.1780
63
+ step 610000 | 6197s | train_loss=0.0434
64
+ step 620000 | 6291s | train_loss=0.0437
65
+ step 630000 | 6385s | train_loss=0.0428
66
+ step 640000 | 6479s | train_loss=0.0419
67
+ step 650000 | 6572s | train_loss=0.0421
68
+ step 660000 | 6665s | train_loss=0.0414
69
+ step 670000 | 6757s | train_loss=0.0402
70
+ step 680000 | 6850s | train_loss=0.0409
71
+ step 690000 | 6943s | train_loss=0.0401
72
+ step 700000 | 7110s | train_loss=0.0398 | fid=47.8908
73
+ step 710000 | 7199s | train_loss=0.0395
74
+ step 720000 | 7293s | train_loss=0.0397
75
+ step 730000 | 7387s | train_loss=0.0388
76
+ step 740000 | 7481s | train_loss=0.0390
77
+ step 750000 | 7574s | train_loss=0.0388
78
+ step 760000 | 7666s | train_loss=0.0386
79
+ step 770000 | 7760s | train_loss=0.0379
80
+ step 780000 | 7854s | train_loss=0.0374
81
+ step 790000 | 7947s | train_loss=0.0371
82
+ step 800000 | 8114s | train_loss=0.0383 | fid=45.7041
83
+ step 810000 | 8203s | train_loss=0.0368
84
+ step 820000 | 8297s | train_loss=0.0366
85
+ step 830000 | 8391s | train_loss=0.0366
86
+ step 840000 | 8485s | train_loss=0.0358
87
+ step 850000 | 8578s | train_loss=0.0368
88
+ step 860000 | 8670s | train_loss=0.0361
89
+ step 870000 | 8763s | train_loss=0.0351
90
+ step 880000 | 8857s | train_loss=0.0355
91
+ step 890000 | 8951s | train_loss=0.0353
92
+ step 900000 | 9118s | train_loss=0.0358 | fid=45.9028
93
+ step 910000 | 9207s | train_loss=0.0345
94
+ step 920000 | 9301s | train_loss=0.0340
95
+ step 930000 | 9395s | train_loss=0.0343
96
+ step 940000 | 9489s | train_loss=0.0340
97
+ step 950000 | 9580s | train_loss=0.0341
98
+ step 960000 | 9674s | train_loss=0.0338
99
+ step 970000 | 9768s | train_loss=0.0344
100
+ step 980000 | 9862s | train_loss=0.0327
101
+ step 990000 | 9956s | train_loss=0.0326
102
+ step 1000000 | 10123s | train_loss=0.0327 | fid=44.1092
103
+ step 1010000 | 10211s | train_loss=0.0322
104
+ step 1020000 | 10305s | train_loss=0.0319
105
+ step 1030000 | 10399s | train_loss=0.0329
106
+ step 1040000 | 10492s | train_loss=0.0325
107
+ step 1050000 | 10584s | train_loss=0.0322
108
+ step 1060000 | 10678s | train_loss=0.0319
109
+ step 1070000 | 10772s | train_loss=0.0318
110
+ step 1080000 | 10866s | train_loss=0.0314
111
+ step 1090000 | 10960s | train_loss=0.0319
112
+ step 1100000 | 11126s | train_loss=0.0314 | fid=44.5369
113
+ step 1110000 | 11216s | train_loss=0.0313
114
+ step 1120000 | 11308s | train_loss=0.0311
115
+ step 1130000 | 11401s | train_loss=0.0305
116
+ step 1140000 | 11493s | train_loss=0.0313
117
+ step 1150000 | 11587s | train_loss=0.0300
118
+ step 1160000 | 11680s | train_loss=0.0293
119
+ step 1170000 | 11774s | train_loss=0.0301
120
+ step 1180000 | 11868s | train_loss=0.0292
121
+ step 1190000 | 11962s | train_loss=0.0289
122
+ step 1200000 | 12128s | train_loss=0.0295 | fid=43.0816
123
+ step 1210000 | 12218s | train_loss=0.0293
124
+ step 1220000 | 12310s | train_loss=0.0289
125
+ step 1230000 | 12403s | train_loss=0.0288
126
+ step 1240000 | 12496s | train_loss=0.0292
127
+ step 1250000 | 12590s | train_loss=0.0277
128
+ step 1260000 | 12683s | train_loss=0.0306
129
+ step 1270000 | 12778s | train_loss=0.0277
130
+ step 1280000 | 12871s | train_loss=0.0284
131
+ step 1290000 | 12965s | train_loss=0.0280
132
+ step 1300000 | 13131s | train_loss=0.0271 | fid=43.4576
133
+ step 1310000 | 13221s | train_loss=0.0278
134
+ step 1320000 | 13315s | train_loss=0.0271
135
+ step 1330000 | 13407s | train_loss=0.0274
136
+ step 1340000 | 13500s | train_loss=0.0272
137
+ step 1350000 | 13594s | train_loss=0.0277
138
+ step 1360000 | 13687s | train_loss=0.0266
139
+ step 1370000 | 13781s | train_loss=0.0271
140
+ step 1380000 | 13875s | train_loss=0.0274
141
+ step 1390000 | 13968s | train_loss=0.0270
142
+ step 1400000 | 14134s | train_loss=0.0268 | fid=43.9846
143
+ step 1410000 | 14225s | train_loss=0.0277
144
+ step 1420000 | 14316s | train_loss=0.0265
145
+ step 1430000 | 14409s | train_loss=0.0257
146
+ step 1440000 | 14503s | train_loss=0.0255
147
+ step 1450000 | 14597s | train_loss=0.0255
148
+ step 1460000 | 14690s | train_loss=0.0258
149
+ step 1470000 | 14784s | train_loss=0.0259
150
+ step 1480000 | 14878s | train_loss=0.0259
151
+ step 1490000 | 14972s | train_loss=0.0262
152
+ step 1500000 | 15138s | train_loss=0.0254 | fid=43.0346
153
+ step 1510000 | 15232s | train_loss=0.0256
154
+ step 1520000 | 15320s | train_loss=0.0251
155
+ step 1530000 | 15414s | train_loss=0.0246
156
+ step 1540000 | 15507s | train_loss=0.0251
157
+ step 1550000 | 15601s | train_loss=0.0247
158
+ step 1560000 | 15695s | train_loss=0.0248
159
+ step 1570000 | 15789s | train_loss=0.0248
160
+ step 1580000 | 15883s | train_loss=0.0252
161
+ step 1590000 | 15977s | train_loss=0.0248
162
+ step 1600000 | 16143s | train_loss=0.0240 | fid=42.4704
163
+ step 1610000 | 16236s | train_loss=0.0236
164
+ step 1620000 | 16326s | train_loss=0.0249
165
+ step 1630000 | 16419s | train_loss=0.0236
166
+ step 1640000 | 16513s | train_loss=0.0233
167
+ step 1650000 | 16607s | train_loss=0.0237
168
+ step 1660000 | 16701s | train_loss=0.0245
169
+ step 1670000 | 16794s | train_loss=0.0234
170
+ step 1680000 | 16888s | train_loss=0.0240
171
+ step 1690000 | 16982s | train_loss=0.0236
172
+ step 1700000 | 17149s | train_loss=0.0225 | fid=42.2384
173
+ step 1710000 | 17243s | train_loss=0.0229
174
+ step 1720000 | 17333s | train_loss=0.0230
175
+ step 1730000 | 17426s | train_loss=0.0226
176
+ step 1740000 | 17520s | train_loss=0.0222
177
+ step 1750000 | 17614s | train_loss=0.0230
178
+ step 1760000 | 17708s | train_loss=0.0226
179
+ step 1770000 | 17802s | train_loss=0.0233
180
+ step 1780000 | 17896s | train_loss=0.0224
181
+ step 1790000 | 17989s | train_loss=0.0226
182
+ step 1800000 | 18153s | train_loss=0.0219 | fid=41.7921
183
+ step 1810000 | 18248s | train_loss=0.0217
184
+ step 1820000 | 18338s | train_loss=0.0217
185
+ step 1830000 | 18432s | train_loss=0.0229
186
+ step 1840000 | 18525s | train_loss=0.0221
187
+ step 1850000 | 18619s | train_loss=0.0224
188
+ step 1860000 | 18713s | train_loss=0.0214
189
+ step 1870000 | 18807s | train_loss=0.0219
190
+ step 1880000 | 18900s | train_loss=0.0215
191
+ step 1890000 | 18994s | train_loss=0.0225
192
+ step 1900000 | 19160s | train_loss=0.0215 | fid=41.8357
193
+ step 1910000 | 19254s | train_loss=0.0210
194
+ step 1920000 | 19344s | train_loss=0.0214
195
+ step 1930000 | 19438s | train_loss=0.0213
196
+ step 1940000 | 19532s | train_loss=0.0203
197
+ step 1950000 | 19626s | train_loss=0.0205
198
+ step 1960000 | 19720s | train_loss=0.0204
199
+ step 1970000 | 19814s | train_loss=0.0204
200
+ step 1980000 | 19907s | train_loss=0.0213
201
+ step 1990000 | 20001s | train_loss=0.0200
202
+ step 2000000 | 20167s | train_loss=0.0211 | fid=42.3155
203
+ step 2010000 | 20261s | train_loss=0.0202
204
+ step 2020000 | 20351s | train_loss=0.0200
205
+ step 2030000 | 20444s | train_loss=0.0205
206
+ step 2040000 | 20538s | train_loss=0.0197
207
+ step 2050000 | 20632s | train_loss=0.0200
208
+ step 2060000 | 20726s | train_loss=0.0200
209
+ step 2070000 | 20819s | train_loss=0.0200
210
+ step 2080000 | 20913s | train_loss=0.0206
211
+ step 2090000 | 21005s | train_loss=0.0198
212
+ step 2100000 | 21172s | train_loss=0.0195 | fid=42.1325
213
+ step 2110000 | 21266s | train_loss=0.0191
214
+ step 2120000 | 21356s | train_loss=0.0198
215
+ step 2130000 | 21450s | train_loss=0.0194
216
+ step 2140000 | 21544s | train_loss=0.0196
217
+ step 2150000 | 21637s | train_loss=0.0188
218
+ step 2160000 | 21730s | train_loss=0.0199
219
+ step 2170000 | 21824s | train_loss=0.0195
220
+ step 2180000 | 21916s | train_loss=0.0191
221
+ step 2190000 | 22010s | train_loss=0.0192
222
+ step 2200000 | 22177s | train_loss=0.0191 | fid=41.8721
223
+ step 2210000 | 22270s | train_loss=0.0197
224
+ step 2220000 | 22361s | train_loss=0.0184
225
+ step 2230000 | 22455s | train_loss=0.0183
226
+ step 2240000 | 22548s | train_loss=0.0185
227
+ step 2250000 | 22642s | train_loss=0.0189
228
+ step 2260000 | 22735s | train_loss=0.0184
229
+ step 2270000 | 22828s | train_loss=0.0189
230
+ step 2280000 | 22920s | train_loss=0.0179
231
+ step 2290000 | 23014s | train_loss=0.0181
232
+ step 2300000 | 23181s | train_loss=0.0185 | fid=41.0660
233
+ step 2310000 | 23275s | train_loss=0.0177
234
+ step 2320000 | 23366s | train_loss=0.0180
235
+ step 2330000 | 23459s | train_loss=0.0185
236
+ step 2340000 | 23553s | train_loss=0.0180
237
+ step 2350000 | 23646s | train_loss=0.0174
238
+ step 2360000 | 23739s | train_loss=0.0171
239
+ step 2370000 | 23831s | train_loss=0.0175
240
+ step 2380000 | 23925s | train_loss=0.0180
241
+ step 2390000 | 24019s | train_loss=0.0172
242
+ step 2400000 | 24186s | train_loss=0.0174 | fid=42.1488
243
+ step 2410000 | 24279s | train_loss=0.0183
244
+ step 2420000 | 24371s | train_loss=0.0168
245
+ step 2430000 | 24464s | train_loss=0.0173
246
+ step 2440000 | 24558s | train_loss=0.0171
247
+ step 2450000 | 24652s | train_loss=0.0174
248
+ step 2460000 | 24745s | train_loss=0.0173
249
+ step 2470000 | 24837s | train_loss=0.0167
250
+ step 2480000 | 24931s | train_loss=0.0172
251
+ step 2490000 | 25024s | train_loss=0.0172
252
+ step 2500000 | 25191s | train_loss=0.0171 | fid=41.8417
253
+ step 2510000 | 25285s | train_loss=0.0168
254
+ step 2520000 | 25376s | train_loss=0.0167
255
+ step 2530000 | 25470s | train_loss=0.0167
256
+ step 2540000 | 25563s | train_loss=0.0165
257
+ step 2550000 | 25657s | train_loss=0.0164
258
+ step 2560000 | 25748s | train_loss=0.0165
259
+ step 2570000 | 25841s | train_loss=0.0159
260
+ step 2580000 | 25935s | train_loss=0.0161
261
+ step 2590000 | 26029s | train_loss=0.0167
262
+ step 2600000 | 26195s | train_loss=0.0164 | fid=41.8686
263
+ step 2610000 | 26289s | train_loss=0.0169
264
+ step 2620000 | 26380s | train_loss=0.0168
265
+ step 2630000 | 26474s | train_loss=0.0167
266
+ step 2640000 | 26567s | train_loss=0.0160
267
+ step 2650000 | 26659s | train_loss=0.0158
268
+ step 2660000 | 26753s | train_loss=0.0155
269
+ step 2670000 | 26846s | train_loss=0.0157
270
+ step 2680000 | 26940s | train_loss=0.0159
271
+ step 2690000 | 27034s | train_loss=0.0156
272
+ step 2700000 | 27201s | train_loss=0.0153 | fid=41.3771
273
+ step 2710000 | 27294s | train_loss=0.0156
274
+ step 2720000 | 27386s | train_loss=0.0160
275
+ step 2730000 | 27479s | train_loss=0.0162
276
+ step 2740000 | 27573s | train_loss=0.0155
277
+ step 2750000 | 27665s | train_loss=0.0148
278
+ step 2760000 | 27759s | train_loss=0.0155
279
+ step 2770000 | 27853s | train_loss=0.0151
280
+ step 2780000 | 27946s | train_loss=0.0156
281
+ step 2790000 | 28040s | train_loss=0.0145
282
+ step 2800000 | 28207s | train_loss=0.0147 | fid=41.5180
283
+ step 2810000 | 28301s | train_loss=0.0148
284
+ step 2820000 | 28392s | train_loss=0.0154
285
+ step 2830000 | 28486s | train_loss=0.0155
286
+ step 2840000 | 28578s | train_loss=0.0149
287
+ step 2850000 | 28672s | train_loss=0.0152
288
+ step 2860000 | 28766s | train_loss=0.0146
289
+ step 2870000 | 28859s | train_loss=0.0148
290
+ step 2880000 | 28953s | train_loss=0.0147
291
+ step 2890000 | 29047s | train_loss=0.0143
292
+ step 2900000 | 29214s | train_loss=0.0151 | fid=41.6079
293
+ step 2910000 | 29308s | train_loss=0.0140
294
+ step 2920000 | 29398s | train_loss=0.0143
295
+ step 2930000 | 29492s | train_loss=0.0150
296
+ step 2940000 | 29584s | train_loss=0.0141
297
+ step 2950000 | 29677s | train_loss=0.0148
298
+ step 2960000 | 29771s | train_loss=0.0142
299
+ step 2970000 | 29865s | train_loss=0.0145
300
+ step 2980000 | 29958s | train_loss=0.0139
301
+ step 2990000 | 30052s | train_loss=0.0147
302
+ step 3000000 | 30219s | train_loss=0.0141 | fid=41.6935
303
+ step 3010000 | 30312s | train_loss=0.0141
304
+ step 3020000 | 30404s | train_loss=0.0139
305
+ step 3030000 | 30495s | train_loss=0.0138
306
+ step 3040000 | 30589s | train_loss=0.0141
307
+ step 3050000 | 30683s | train_loss=0.0139
308
+ step 3060000 | 30776s | train_loss=0.0141
309
+ step 3070000 | 30870s | train_loss=0.0140
310
+ step 3080000 | 30963s | train_loss=0.0138
311
+ step 3090000 | 31057s | train_loss=0.0140
312
+ step 3100000 | 31223s | train_loss=0.0143 | fid=41.0898
313
+ step 3110000 | 31317s | train_loss=0.0144
314
+ step 3120000 | 31410s | train_loss=0.0145
315
+ step 3130000 | 31497s | train_loss=0.0135
316
+ step 3140000 | 31591s | train_loss=0.0138
317
+ step 3150000 | 31685s | train_loss=0.0141
318
+ step 3160000 | 31778s | train_loss=0.0132
319
+ step 3170000 | 31872s | train_loss=0.0138
320
+ step 3180000 | 31965s | train_loss=0.0133
321
+ step 3190000 | 32059s | train_loss=0.0137
322
+ step 3200000 | 32226s | train_loss=0.0129 | fid=41.4764
323
+ step 3210000 | 32319s | train_loss=0.0135
324
+ step 3220000 | 32412s | train_loss=0.0136
325
+ step 3230000 | 32503s | train_loss=0.0139
326
+ step 3240000 | 32597s | train_loss=0.0130
327
+ step 3250000 | 32690s | train_loss=0.0136
328
+ step 3260000 | 32784s | train_loss=0.0142
329
+ step 3270000 | 32878s | train_loss=0.0136
330
+ step 3280000 | 32971s | train_loss=0.0129
331
+ step 3290000 | 33064s | train_loss=0.0129
332
+ step 3300000 | 33230s | train_loss=0.0131 | fid=41.4101
333
+ step 3310000 | 33321s | train_loss=0.0128
334
+ step 3320000 | 33416s | train_loss=0.0137
335
+ step 3330000 | 33505s | train_loss=0.0136
336
+ step 3340000 | 33598s | train_loss=0.0128
337
+ step 3350000 | 33693s | train_loss=0.0137
338
+ step 3360000 | 33786s | train_loss=0.0128
339
+ step 3370000 | 33880s | train_loss=0.0130
340
+ step 3380000 | 33974s | train_loss=0.0131
341
+ step 3390000 | 34067s | train_loss=0.0128
342
+ step 3400000 | 34234s | train_loss=0.0130 | fid=41.8922
343
+ step 3410000 | 34325s | train_loss=0.0123
344
+ step 3420000 | 34419s | train_loss=0.0130
345
+ step 3430000 | 34510s | train_loss=0.0126
346
+ step 3440000 | 34603s | train_loss=0.0133
347
+ step 3450000 | 34697s | train_loss=0.0126
348
+ step 3460000 | 34791s | train_loss=0.0128
349
+ step 3470000 | 34884s | train_loss=0.0124
350
+ step 3480000 | 34978s | train_loss=0.0126
351
+ step 3490000 | 35070s | train_loss=0.0126
352
+ step 3500000 | 35234s | train_loss=0.0126 | fid=41.7623
353
+ step 3510000 | 35328s | train_loss=0.0124
354
+ step 3520000 | 35422s | train_loss=0.0133
355
+ step 3530000 | 35512s | train_loss=0.0121
356
+ step 3540000 | 35606s | train_loss=0.0133
357
+ step 3550000 | 35699s | train_loss=0.0126
358
+ step 3560000 | 35793s | train_loss=0.0129
359
+ step 3570000 | 35887s | train_loss=0.0127
360
+ step 3580000 | 35980s | train_loss=0.0124
361
+ step 3590000 | 36073s | train_loss=0.0116
362
+ step 3600000 | 36235s | train_loss=0.0121 | fid=41.7174
363
+ step 3610000 | 36328s | train_loss=0.0125
364
+ step 3620000 | 36423s | train_loss=0.0121
365
+ step 3630000 | 36514s | train_loss=0.0125
366
+ step 3640000 | 36607s | train_loss=0.0118
367
+ step 3650000 | 36701s | train_loss=0.0117
368
+ step 3660000 | 36794s | train_loss=0.0120
369
+ step 3670000 | 36888s | train_loss=0.0116
370
+ step 3680000 | 36981s | train_loss=0.0123
371
+ step 3690000 | 37074s | train_loss=0.0118
372
+ step 3700000 | 37237s | train_loss=0.0115 | fid=41.3975
373
+ step 3710000 | 37331s | train_loss=0.0128
374
+ step 3720000 | 37425s | train_loss=0.0120
375
+ step 3730000 | 37515s | train_loss=0.0116
376
+ step 3740000 | 37609s | train_loss=0.0114
377
+ step 3750000 | 37702s | train_loss=0.0120
378
+ step 3760000 | 37796s | train_loss=0.0120
379
+ step 3770000 | 37889s | train_loss=0.0119
380
+ step 3780000 | 37983s | train_loss=0.0113
381
+ step 3790000 | 38075s | train_loss=0.0114
382
+ step 3800000 | 38240s | train_loss=0.0116 | fid=42.3031
383
+ step 3810000 | 38334s | train_loss=0.0110
384
+ step 3820000 | 38428s | train_loss=0.0117
385
+ step 3830000 | 38518s | train_loss=0.0116
386
+ step 3840000 | 38612s | train_loss=0.0111
387
+ step 3850000 | 38706s | train_loss=0.0113
388
+ step 3860000 | 38800s | train_loss=0.0116
389
+ step 3870000 | 38893s | train_loss=0.0111
390
+ step 3880000 | 38987s | train_loss=0.0115
391
+ step 3890000 | 39078s | train_loss=0.0117
392
+ step 3900000 | 39244s | train_loss=0.0113 | fid=41.0003
393
+ step 3910000 | 39338s | train_loss=0.0115
394
+ step 3920000 | 39431s | train_loss=0.0108
395
+ step 3930000 | 39521s | train_loss=0.0115
396
+ step 3940000 | 39615s | train_loss=0.0113
397
+ step 3950000 | 39708s | train_loss=0.0110
398
+ step 3960000 | 39802s | train_loss=0.0113
399
+ step 3970000 | 39896s | train_loss=0.0110
400
+ step 3980000 | 39987s | train_loss=0.0111
401
+ step 3990000 | 40081s | train_loss=0.0116
402
+ step 4000000 | 40246s | train_loss=0.0114 | fid=42.4209
403
+ step 4010000 | 40340s | train_loss=0.0111
404
+ step 4020000 | 40434s | train_loss=0.0113
405
+ step 4030000 | 40524s | train_loss=0.0104
406
+ step 4040000 | 40617s | train_loss=0.0109
407
+ step 4050000 | 40711s | train_loss=0.0115
408
+ step 4060000 | 40805s | train_loss=0.0110
409
+ step 4070000 | 40898s | train_loss=0.0107
410
+ step 4080000 | 40989s | train_loss=0.0109
411
+ step 4090000 | 41083s | train_loss=0.0109
412
+ step 4100000 | 41249s | train_loss=0.0110 | fid=41.7577
413
+ step 4110000 | 41343s | train_loss=0.0107
414
+ step 4120000 | 41437s | train_loss=0.0111
415
+ step 4130000 | 41528s | train_loss=0.0108
416
+ step 4140000 | 41619s | train_loss=0.0112
417
+ step 4150000 | 41713s | train_loss=0.0113
418
+ step 4160000 | 41807s | train_loss=0.0109
419
+ step 4170000 | 41899s | train_loss=0.0114
420
+ step 4180000 | 41992s | train_loss=0.0111
421
+ step 4190000 | 42086s | train_loss=0.0106
422
+ step 4200000 | 42252s | train_loss=0.0109 | fid=41.6909
423
+ step 4210000 | 42346s | train_loss=0.0108
424
+ step 4220000 | 42439s | train_loss=0.0103
425
+ step 4230000 | 42532s | train_loss=0.0110
426
+ step 4240000 | 42623s | train_loss=0.0104
427
+ step 4250000 | 42717s | train_loss=0.0104
428
+ step 4260000 | 42810s | train_loss=0.0103
429
+ step 4270000 | 42902s | train_loss=0.0107
430
+ step 4280000 | 42995s | train_loss=0.0106
431
+ step 4290000 | 43089s | train_loss=0.0104
432
+ step 4300000 | 43256s | train_loss=0.0105 | fid=42.2357
433
+ step 4310000 | 43349s | train_loss=0.0102
434
+ step 4320000 | 43443s | train_loss=0.0103
435
+ step 4330000 | 43536s | train_loss=0.0104
436
+ step 4340000 | 43627s | train_loss=0.0101
437
+ step 4350000 | 43721s | train_loss=0.0103
438
+ step 4360000 | 43812s | train_loss=0.0107
439
+ step 4370000 | 43906s | train_loss=0.0100
440
+ step 4380000 | 44000s | train_loss=0.0102
441
+ step 4390000 | 44093s | train_loss=0.0107
442
+ step 4400000 | 44259s | train_loss=0.0106 | fid=41.4018
443
+ step 4410000 | 44353s | train_loss=0.0109
444
+ step 4420000 | 44446s | train_loss=0.0108
445
+ step 4430000 | 44539s | train_loss=0.0106
446
+ step 4440000 | 44630s | train_loss=0.0104
447
+ step 4450000 | 44724s | train_loss=0.0102
448
+ step 4460000 | 44815s | train_loss=0.0109
449
+ step 4470000 | 44909s | train_loss=0.0101
450
+ step 4480000 | 45003s | train_loss=0.0104
451
+ step 4490000 | 45096s | train_loss=0.0103
452
+ step 4500000 | 45263s | train_loss=0.0102 | fid=42.6356
453
+ step 4510000 | 45357s | train_loss=0.0101
454
+ step 4520000 | 45450s | train_loss=0.0100
455
+ step 4530000 | 45543s | train_loss=0.0101
456
+ step 4540000 | 45635s | train_loss=0.0105
457
+ step 4550000 | 45727s | train_loss=0.0100
458
+ step 4560000 | 45820s | train_loss=0.0102
459
+ step 4570000 | 45914s | train_loss=0.0096
460
+ step 4580000 | 46009s | train_loss=0.0101
461
+ step 4590000 | 46102s | train_loss=0.0099
462
+ step 4600000 | 46269s | train_loss=0.0095 | fid=41.8796
463
+ step 4610000 | 46362s | train_loss=0.0094
464
+ step 4620000 | 46455s | train_loss=0.0103
465
+ step 4630000 | 46547s | train_loss=0.0101
466
+ step 4640000 | 46638s | train_loss=0.0104
467
+ step 4650000 | 46732s | train_loss=0.0104
468
+ step 4660000 | 46826s | train_loss=0.0105
469
+ step 4670000 | 46920s | train_loss=0.0098
470
+ step 4680000 | 47014s | train_loss=0.0098
471
+ step 4690000 | 47108s | train_loss=0.0098
472
+ step 4700000 | 47274s | train_loss=0.0095 | fid=41.8283
473
+ step 4710000 | 47367s | train_loss=0.0098
474
+ step 4720000 | 47461s | train_loss=0.0100
475
+ step 4730000 | 47554s | train_loss=0.0104
476
+ step 4740000 | 47633s | train_loss=0.0099
477
+ step 4750000 | 47709s | train_loss=0.0097
478
+ step 4760000 | 47785s | train_loss=0.0100
479
+ step 4770000 | 47861s | train_loss=0.0107
480
+ step 4780000 | 47937s | train_loss=0.0100
481
+ step 4790000 | 48013s | train_loss=0.0098
482
+ step 4800000 | 48148s | train_loss=0.0097 | fid=42.2165
483
+ step 4810000 | 48224s | train_loss=0.0096
484
+ step 4820000 | 48300s | train_loss=0.0097
485
+ step 4830000 | 48375s | train_loss=0.0102
486
+ step 4840000 | 48447s | train_loss=0.0097
487
+ step 4850000 | 48523s | train_loss=0.0101
488
+ step 4860000 | 48599s | train_loss=0.0095
489
+ step 4870000 | 48675s | train_loss=0.0098
490
+ step 4880000 | 48751s | train_loss=0.0094
491
+ step 4890000 | 48827s | train_loss=0.0101
492
+ step 4900000 | 48963s | train_loss=0.0093 | fid=42.1123
493
+ step 4910000 | 49039s | train_loss=0.0100
494
+ step 4920000 | 49115s | train_loss=0.0093
495
+ step 4930000 | 49190s | train_loss=0.0092
496
+ step 4940000 | 49262s | train_loss=0.0094
497
+ step 4950000 | 49338s | train_loss=0.0094
498
+ step 4960000 | 49413s | train_loss=0.0093
499
+ step 4970000 | 49489s | train_loss=0.0089
500
+ step 4980000 | 49565s | train_loss=0.0093
501
+ step 4990000 | 49641s | train_loss=0.0096
502
+ step 5000000 | 49770s | train_loss=0.0095 | fid=42.3553
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed44/d80/config.json ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "vae_checkpoint": "vae_checkpoints/celeba_resnet_modernloss_d80/vae.pt",
3
+ "out": "results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed44/d80",
4
+ "subset_path": "diagnostics/celeba_diverse_1k_mean_plus_sd/subset_indices.json",
5
+ "train_tar": "/overflow/data/cerberus/celeba_hq_256_partitioned_ws/train/celeba_hq_256-train-0000.tar",
6
+ "n_train": 1000,
7
+ "total_steps": 5000000,
8
+ "eval_interval": 10000,
9
+ "mem_interval": 100000,
10
+ "fid_interval": 100000,
11
+ "fid_n_real": 1000,
12
+ "fid_n_gen": 10000,
13
+ "n_gen_samples": 10000,
14
+ "n_sde_steps": 500,
15
+ "batch_size": 512,
16
+ "hidden": 1024,
17
+ "depth": 5,
18
+ "optimizer": "sgd",
19
+ "lr": 0.001,
20
+ "momentum": 0.8,
21
+ "t_min": 0.01,
22
+ "t_max": 3.0,
23
+ "t_eval": 0.1,
24
+ "seed": 44,
25
+ "device": null
26
+ }
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed44/d80/metrics.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed44/d80/runner.log ADDED
@@ -0,0 +1,502 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ FID real-side stats cached over 1000 diverse train images
2
+ step 1 | 73s | train_loss=0.9947 | fid=276.8765
3
+ step 10000 | 167s | train_loss=0.8193
4
+ step 20000 | 260s | train_loss=0.7124
5
+ step 30000 | 355s | train_loss=0.5812
6
+ step 40000 | 449s | train_loss=0.4235
7
+ step 50000 | 543s | train_loss=0.2915
8
+ step 60000 | 637s | train_loss=0.2145
9
+ step 70000 | 731s | train_loss=0.1746
10
+ step 80000 | 824s | train_loss=0.1513
11
+ step 90000 | 917s | train_loss=0.1375
12
+ step 100000 | 1083s | train_loss=0.1268 | fid=56.7187
13
+ step 110000 | 1177s | train_loss=0.1196
14
+ step 120000 | 1272s | train_loss=0.1151
15
+ step 130000 | 1366s | train_loss=0.1074
16
+ step 140000 | 1460s | train_loss=0.1022
17
+ step 150000 | 1555s | train_loss=0.0972
18
+ step 160000 | 1649s | train_loss=0.0958
19
+ step 170000 | 1744s | train_loss=0.0928
20
+ step 180000 | 1837s | train_loss=0.0881
21
+ step 190000 | 1930s | train_loss=0.0858
22
+ step 200000 | 2094s | train_loss=0.0825 | fid=55.2490
23
+ step 210000 | 2188s | train_loss=0.0809
24
+ step 220000 | 2282s | train_loss=0.0782
25
+ step 230000 | 2376s | train_loss=0.0758
26
+ step 240000 | 2470s | train_loss=0.0737
27
+ step 250000 | 2565s | train_loss=0.0718
28
+ step 260000 | 2659s | train_loss=0.0720
29
+ step 270000 | 2753s | train_loss=0.0696
30
+ step 280000 | 2846s | train_loss=0.0677
31
+ step 290000 | 2938s | train_loss=0.0674
32
+ step 300000 | 3102s | train_loss=0.0644 | fid=51.8577
33
+ step 310000 | 3196s | train_loss=0.0653
34
+ step 320000 | 3290s | train_loss=0.0620
35
+ step 330000 | 3385s | train_loss=0.0615
36
+ step 340000 | 3479s | train_loss=0.0602
37
+ step 350000 | 3574s | train_loss=0.0596
38
+ step 360000 | 3668s | train_loss=0.0587
39
+ step 370000 | 3761s | train_loss=0.0579
40
+ step 380000 | 3854s | train_loss=0.0576
41
+ step 390000 | 3948s | train_loss=0.0557
42
+ step 400000 | 4111s | train_loss=0.0559 | fid=51.3264
43
+ step 410000 | 4205s | train_loss=0.0542
44
+ step 420000 | 4300s | train_loss=0.0539
45
+ step 430000 | 4394s | train_loss=0.0531
46
+ step 440000 | 4488s | train_loss=0.0526
47
+ step 450000 | 4583s | train_loss=0.0520
48
+ step 460000 | 4676s | train_loss=0.0507
49
+ step 470000 | 4770s | train_loss=0.0508
50
+ step 480000 | 4863s | train_loss=0.0507
51
+ step 490000 | 4957s | train_loss=0.0493
52
+ step 500000 | 5121s | train_loss=0.0498 | fid=49.9983
53
+ step 510000 | 5215s | train_loss=0.0478
54
+ step 520000 | 5310s | train_loss=0.0478
55
+ step 530000 | 5405s | train_loss=0.0474
56
+ step 540000 | 5499s | train_loss=0.0479
57
+ step 550000 | 5594s | train_loss=0.0468
58
+ step 560000 | 5687s | train_loss=0.0463
59
+ step 570000 | 5779s | train_loss=0.0463
60
+ step 580000 | 5873s | train_loss=0.0458
61
+ step 590000 | 5968s | train_loss=0.0458
62
+ step 600000 | 6131s | train_loss=0.0452 | fid=48.5033
63
+ step 610000 | 6225s | train_loss=0.0445
64
+ step 620000 | 6319s | train_loss=0.0451
65
+ step 630000 | 6414s | train_loss=0.0439
66
+ step 640000 | 6507s | train_loss=0.0437
67
+ step 650000 | 6601s | train_loss=0.0428
68
+ step 660000 | 6695s | train_loss=0.0431
69
+ step 670000 | 6787s | train_loss=0.0429
70
+ step 680000 | 6882s | train_loss=0.0431
71
+ step 690000 | 6975s | train_loss=0.0413
72
+ step 700000 | 7140s | train_loss=0.0412 | fid=47.5669
73
+ step 710000 | 7234s | train_loss=0.0417
74
+ step 720000 | 7328s | train_loss=0.0412
75
+ step 730000 | 7423s | train_loss=0.0413
76
+ step 740000 | 7517s | train_loss=0.0403
77
+ step 750000 | 7611s | train_loss=0.0418
78
+ step 760000 | 7704s | train_loss=0.0415
79
+ step 770000 | 7798s | train_loss=0.0403
80
+ step 780000 | 7893s | train_loss=0.0391
81
+ step 790000 | 7987s | train_loss=0.0386
82
+ step 800000 | 8152s | train_loss=0.0403 | fid=46.5982
83
+ step 810000 | 8247s | train_loss=0.0386
84
+ step 820000 | 8341s | train_loss=0.0379
85
+ step 830000 | 8436s | train_loss=0.0382
86
+ step 840000 | 8530s | train_loss=0.0388
87
+ step 850000 | 8622s | train_loss=0.0378
88
+ step 860000 | 8716s | train_loss=0.0378
89
+ step 870000 | 8810s | train_loss=0.0368
90
+ step 880000 | 8905s | train_loss=0.0374
91
+ step 890000 | 8999s | train_loss=0.0380
92
+ step 900000 | 9166s | train_loss=0.0371 | fid=46.7047
93
+ step 910000 | 9259s | train_loss=0.0364
94
+ step 920000 | 9354s | train_loss=0.0361
95
+ step 930000 | 9448s | train_loss=0.0362
96
+ step 940000 | 9542s | train_loss=0.0373
97
+ step 950000 | 9634s | train_loss=0.0353
98
+ step 960000 | 9729s | train_loss=0.0366
99
+ step 970000 | 9823s | train_loss=0.0356
100
+ step 980000 | 9918s | train_loss=0.0346
101
+ step 990000 | 10013s | train_loss=0.0356
102
+ step 1000000 | 10179s | train_loss=0.0351 | fid=45.3806
103
+ step 1010000 | 10273s | train_loss=0.0344
104
+ step 1020000 | 10367s | train_loss=0.0346
105
+ step 1030000 | 10461s | train_loss=0.0346
106
+ step 1040000 | 10553s | train_loss=0.0349
107
+ step 1050000 | 10647s | train_loss=0.0332
108
+ step 1060000 | 10742s | train_loss=0.0339
109
+ step 1070000 | 10836s | train_loss=0.0331
110
+ step 1080000 | 10931s | train_loss=0.0343
111
+ step 1090000 | 11025s | train_loss=0.0330
112
+ step 1100000 | 11192s | train_loss=0.0327 | fid=44.4908
113
+ step 1110000 | 11285s | train_loss=0.0332
114
+ step 1120000 | 11379s | train_loss=0.0332
115
+ step 1130000 | 11471s | train_loss=0.0328
116
+ step 1140000 | 11565s | train_loss=0.0324
117
+ step 1150000 | 11659s | train_loss=0.0318
118
+ step 1160000 | 11754s | train_loss=0.0322
119
+ step 1170000 | 11848s | train_loss=0.0319
120
+ step 1180000 | 11942s | train_loss=0.0319
121
+ step 1190000 | 12036s | train_loss=0.0312
122
+ step 1200000 | 12202s | train_loss=0.0309 | fid=44.0628
123
+ step 1210000 | 12296s | train_loss=0.0317
124
+ step 1220000 | 12390s | train_loss=0.0310
125
+ step 1230000 | 12482s | train_loss=0.0311
126
+ step 1240000 | 12576s | train_loss=0.0308
127
+ step 1250000 | 12670s | train_loss=0.0298
128
+ step 1260000 | 12764s | train_loss=0.0307
129
+ step 1270000 | 12859s | train_loss=0.0295
130
+ step 1280000 | 12953s | train_loss=0.0312
131
+ step 1290000 | 13048s | train_loss=0.0302
132
+ step 1300000 | 13214s | train_loss=0.0299 | fid=42.2414
133
+ step 1310000 | 13308s | train_loss=0.0298
134
+ step 1320000 | 13401s | train_loss=0.0292
135
+ step 1330000 | 13494s | train_loss=0.0295
136
+ step 1340000 | 13589s | train_loss=0.0294
137
+ step 1350000 | 13683s | train_loss=0.0291
138
+ step 1360000 | 13777s | train_loss=0.0288
139
+ step 1370000 | 13871s | train_loss=0.0286
140
+ step 1380000 | 13965s | train_loss=0.0294
141
+ step 1390000 | 14060s | train_loss=0.0284
142
+ step 1400000 | 14225s | train_loss=0.0278 | fid=42.2193
143
+ step 1410000 | 14317s | train_loss=0.0284
144
+ step 1420000 | 14411s | train_loss=0.0276
145
+ step 1430000 | 14505s | train_loss=0.0274
146
+ step 1440000 | 14599s | train_loss=0.0283
147
+ step 1450000 | 14694s | train_loss=0.0270
148
+ step 1460000 | 14788s | train_loss=0.0276
149
+ step 1470000 | 14883s | train_loss=0.0278
150
+ step 1480000 | 14978s | train_loss=0.0274
151
+ step 1490000 | 15072s | train_loss=0.0281
152
+ step 1500000 | 15238s | train_loss=0.0267 | fid=42.8647
153
+ step 1510000 | 15330s | train_loss=0.0264
154
+ step 1520000 | 15424s | train_loss=0.0270
155
+ step 1530000 | 15519s | train_loss=0.0266
156
+ step 1540000 | 15613s | train_loss=0.0264
157
+ step 1550000 | 15708s | train_loss=0.0272
158
+ step 1560000 | 15802s | train_loss=0.0264
159
+ step 1570000 | 15897s | train_loss=0.0270
160
+ step 1580000 | 15991s | train_loss=0.0261
161
+ step 1590000 | 16085s | train_loss=0.0257
162
+ step 1600000 | 16247s | train_loss=0.0266 | fid=42.6558
163
+ step 1610000 | 16342s | train_loss=0.0257
164
+ step 1620000 | 16435s | train_loss=0.0253
165
+ step 1630000 | 16530s | train_loss=0.0254
166
+ step 1640000 | 16624s | train_loss=0.0247
167
+ step 1650000 | 16719s | train_loss=0.0250
168
+ step 1660000 | 16814s | train_loss=0.0248
169
+ step 1670000 | 16908s | train_loss=0.0249
170
+ step 1680000 | 17002s | train_loss=0.0249
171
+ step 1690000 | 17097s | train_loss=0.0244
172
+ step 1700000 | 17260s | train_loss=0.0238 | fid=42.6949
173
+ step 1710000 | 17354s | train_loss=0.0251
174
+ step 1720000 | 17448s | train_loss=0.0248
175
+ step 1730000 | 17542s | train_loss=0.0243
176
+ step 1740000 | 17637s | train_loss=0.0242
177
+ step 1750000 | 17731s | train_loss=0.0240
178
+ step 1760000 | 17826s | train_loss=0.0244
179
+ step 1770000 | 17921s | train_loss=0.0235
180
+ step 1780000 | 18014s | train_loss=0.0241
181
+ step 1790000 | 18107s | train_loss=0.0234
182
+ step 1800000 | 18273s | train_loss=0.0237 | fid=41.4449
183
+ step 1810000 | 18367s | train_loss=0.0231
184
+ step 1820000 | 18461s | train_loss=0.0231
185
+ step 1830000 | 18555s | train_loss=0.0240
186
+ step 1840000 | 18649s | train_loss=0.0230
187
+ step 1850000 | 18744s | train_loss=0.0228
188
+ step 1860000 | 18838s | train_loss=0.0227
189
+ step 1870000 | 18932s | train_loss=0.0229
190
+ step 1880000 | 19027s | train_loss=0.0225
191
+ step 1890000 | 19119s | train_loss=0.0242
192
+ step 1900000 | 19285s | train_loss=0.0229 | fid=41.0483
193
+ step 1910000 | 19379s | train_loss=0.0219
194
+ step 1920000 | 19474s | train_loss=0.0237
195
+ step 1930000 | 19568s | train_loss=0.0229
196
+ step 1940000 | 19663s | train_loss=0.0219
197
+ step 1950000 | 19757s | train_loss=0.0219
198
+ step 1960000 | 19852s | train_loss=0.0219
199
+ step 1970000 | 19946s | train_loss=0.0227
200
+ step 1980000 | 20038s | train_loss=0.0223
201
+ step 1990000 | 20132s | train_loss=0.0218
202
+ step 2000000 | 20299s | train_loss=0.0219 | fid=41.7057
203
+ step 2010000 | 20393s | train_loss=0.0211
204
+ step 2020000 | 20487s | train_loss=0.0215
205
+ step 2030000 | 20582s | train_loss=0.0217
206
+ step 2040000 | 20676s | train_loss=0.0215
207
+ step 2050000 | 20771s | train_loss=0.0223
208
+ step 2060000 | 20865s | train_loss=0.0215
209
+ step 2070000 | 20957s | train_loss=0.0211
210
+ step 2080000 | 21052s | train_loss=0.0221
211
+ step 2090000 | 21145s | train_loss=0.0206
212
+ step 2100000 | 21311s | train_loss=0.0205 | fid=41.4031
213
+ step 2110000 | 21406s | train_loss=0.0205
214
+ step 2120000 | 21500s | train_loss=0.0201
215
+ step 2130000 | 21595s | train_loss=0.0210
216
+ step 2140000 | 21689s | train_loss=0.0208
217
+ step 2150000 | 21783s | train_loss=0.0210
218
+ step 2160000 | 21877s | train_loss=0.0205
219
+ step 2170000 | 21970s | train_loss=0.0204
220
+ step 2180000 | 22064s | train_loss=0.0208
221
+ step 2190000 | 22158s | train_loss=0.0201
222
+ step 2200000 | 22325s | train_loss=0.0199 | fid=40.7561
223
+ step 2210000 | 22419s | train_loss=0.0216
224
+ step 2220000 | 22514s | train_loss=0.0202
225
+ step 2230000 | 22608s | train_loss=0.0201
226
+ step 2240000 | 22701s | train_loss=0.0196
227
+ step 2250000 | 22796s | train_loss=0.0200
228
+ step 2260000 | 22887s | train_loss=0.0196
229
+ step 2270000 | 22982s | train_loss=0.0204
230
+ step 2280000 | 23076s | train_loss=0.0196
231
+ step 2290000 | 23170s | train_loss=0.0196
232
+ step 2300000 | 23336s | train_loss=0.0194 | fid=40.8227
233
+ step 2310000 | 23430s | train_loss=0.0195
234
+ step 2320000 | 23525s | train_loss=0.0193
235
+ step 2330000 | 23619s | train_loss=0.0192
236
+ step 2340000 | 23712s | train_loss=0.0200
237
+ step 2350000 | 23805s | train_loss=0.0196
238
+ step 2360000 | 23898s | train_loss=0.0190
239
+ step 2370000 | 23993s | train_loss=0.0193
240
+ step 2380000 | 24087s | train_loss=0.0194
241
+ step 2390000 | 24180s | train_loss=0.0190
242
+ step 2400000 | 24345s | train_loss=0.0193 | fid=40.2737
243
+ step 2410000 | 24440s | train_loss=0.0201
244
+ step 2420000 | 24534s | train_loss=0.0186
245
+ step 2430000 | 24628s | train_loss=0.0187
246
+ step 2440000 | 24722s | train_loss=0.0184
247
+ step 2450000 | 24814s | train_loss=0.0185
248
+ step 2460000 | 24908s | train_loss=0.0185
249
+ step 2470000 | 25002s | train_loss=0.0187
250
+ step 2480000 | 25097s | train_loss=0.0180
251
+ step 2490000 | 25189s | train_loss=0.0188
252
+ step 2500000 | 25357s | train_loss=0.0183 | fid=40.6182
253
+ step 2510000 | 25451s | train_loss=0.0179
254
+ step 2520000 | 25545s | train_loss=0.0178
255
+ step 2530000 | 25639s | train_loss=0.0175
256
+ step 2540000 | 25731s | train_loss=0.0184
257
+ step 2550000 | 25825s | train_loss=0.0182
258
+ step 2560000 | 25919s | train_loss=0.0179
259
+ step 2570000 | 26013s | train_loss=0.0178
260
+ step 2580000 | 26108s | train_loss=0.0178
261
+ step 2590000 | 26200s | train_loss=0.0186
262
+ step 2600000 | 26367s | train_loss=0.0179 | fid=39.7483
263
+ step 2610000 | 26462s | train_loss=0.0178
264
+ step 2620000 | 26556s | train_loss=0.0169
265
+ step 2630000 | 26650s | train_loss=0.0183
266
+ step 2640000 | 26742s | train_loss=0.0180
267
+ step 2650000 | 26836s | train_loss=0.0176
268
+ step 2660000 | 26931s | train_loss=0.0173
269
+ step 2670000 | 27025s | train_loss=0.0172
270
+ step 2680000 | 27119s | train_loss=0.0177
271
+ step 2690000 | 27211s | train_loss=0.0170
272
+ step 2700000 | 27378s | train_loss=0.0166 | fid=39.7586
273
+ step 2710000 | 27472s | train_loss=0.0173
274
+ step 2720000 | 27566s | train_loss=0.0175
275
+ step 2730000 | 27658s | train_loss=0.0176
276
+ step 2740000 | 27752s | train_loss=0.0166
277
+ step 2750000 | 27847s | train_loss=0.0165
278
+ step 2760000 | 27941s | train_loss=0.0169
279
+ step 2770000 | 28036s | train_loss=0.0170
280
+ step 2780000 | 28130s | train_loss=0.0175
281
+ step 2790000 | 28222s | train_loss=0.0164
282
+ step 2800000 | 28388s | train_loss=0.0162 | fid=39.9894
283
+ step 2810000 | 28482s | train_loss=0.0167
284
+ step 2820000 | 28574s | train_loss=0.0168
285
+ step 2830000 | 28669s | train_loss=0.0169
286
+ step 2840000 | 28763s | train_loss=0.0167
287
+ step 2850000 | 28857s | train_loss=0.0171
288
+ step 2860000 | 28951s | train_loss=0.0164
289
+ step 2870000 | 29045s | train_loss=0.0161
290
+ step 2880000 | 29140s | train_loss=0.0159
291
+ step 2890000 | 29231s | train_loss=0.0163
292
+ step 2900000 | 29398s | train_loss=0.0172 | fid=40.4625
293
+ step 2910000 | 29491s | train_loss=0.0159
294
+ step 2920000 | 29584s | train_loss=0.0158
295
+ step 2930000 | 29678s | train_loss=0.0158
296
+ step 2940000 | 29773s | train_loss=0.0162
297
+ step 2950000 | 29867s | train_loss=0.0158
298
+ step 2960000 | 29961s | train_loss=0.0165
299
+ step 2970000 | 30055s | train_loss=0.0162
300
+ step 2980000 | 30150s | train_loss=0.0154
301
+ step 2990000 | 30241s | train_loss=0.0168
302
+ step 3000000 | 30407s | train_loss=0.0159 | fid=40.1383
303
+ step 3010000 | 30500s | train_loss=0.0157
304
+ step 3020000 | 30594s | train_loss=0.0159
305
+ step 3030000 | 30689s | train_loss=0.0155
306
+ step 3040000 | 30783s | train_loss=0.0160
307
+ step 3050000 | 30877s | train_loss=0.0158
308
+ step 3060000 | 30971s | train_loss=0.0163
309
+ step 3070000 | 31066s | train_loss=0.0155
310
+ step 3080000 | 31160s | train_loss=0.0155
311
+ step 3090000 | 31251s | train_loss=0.0152
312
+ step 3100000 | 31420s | train_loss=0.0151 | fid=39.5050
313
+ step 3110000 | 31514s | train_loss=0.0164
314
+ step 3120000 | 31608s | train_loss=0.0166
315
+ step 3130000 | 31703s | train_loss=0.0155
316
+ step 3140000 | 31797s | train_loss=0.0155
317
+ step 3150000 | 31891s | train_loss=0.0155
318
+ step 3160000 | 31984s | train_loss=0.0146
319
+ step 3170000 | 32079s | train_loss=0.0152
320
+ step 3180000 | 32174s | train_loss=0.0150
321
+ step 3190000 | 32265s | train_loss=0.0150
322
+ step 3200000 | 32429s | train_loss=0.0152 | fid=39.2929
323
+ step 3210000 | 32524s | train_loss=0.0152
324
+ step 3220000 | 32618s | train_loss=0.0157
325
+ step 3230000 | 32713s | train_loss=0.0154
326
+ step 3240000 | 32807s | train_loss=0.0148
327
+ step 3250000 | 32901s | train_loss=0.0152
328
+ step 3260000 | 32994s | train_loss=0.0156
329
+ step 3270000 | 33089s | train_loss=0.0158
330
+ step 3280000 | 33184s | train_loss=0.0148
331
+ step 3290000 | 33275s | train_loss=0.0148
332
+ step 3300000 | 33443s | train_loss=0.0149 | fid=39.2265
333
+ step 3310000 | 33537s | train_loss=0.0148
334
+ step 3320000 | 33631s | train_loss=0.0157
335
+ step 3330000 | 33725s | train_loss=0.0148
336
+ step 3340000 | 33819s | train_loss=0.0147
337
+ step 3350000 | 33913s | train_loss=0.0162
338
+ step 3360000 | 34007s | train_loss=0.0140
339
+ step 3370000 | 34101s | train_loss=0.0147
340
+ step 3380000 | 34195s | train_loss=0.0147
341
+ step 3390000 | 34285s | train_loss=0.0142
342
+ step 3400000 | 34453s | train_loss=0.0147 | fid=39.2717
343
+ step 3410000 | 34548s | train_loss=0.0152
344
+ step 3420000 | 34642s | train_loss=0.0143
345
+ step 3430000 | 34737s | train_loss=0.0144
346
+ step 3440000 | 34831s | train_loss=0.0144
347
+ step 3450000 | 34925s | train_loss=0.0138
348
+ step 3460000 | 35018s | train_loss=0.0149
349
+ step 3470000 | 35112s | train_loss=0.0138
350
+ step 3480000 | 35205s | train_loss=0.0143
351
+ step 3490000 | 35295s | train_loss=0.0142
352
+ step 3500000 | 35464s | train_loss=0.0146 | fid=39.4914
353
+ step 3510000 | 35558s | train_loss=0.0143
354
+ step 3520000 | 35653s | train_loss=0.0141
355
+ step 3530000 | 35747s | train_loss=0.0139
356
+ step 3540000 | 35841s | train_loss=0.0142
357
+ step 3550000 | 35935s | train_loss=0.0143
358
+ step 3560000 | 36029s | train_loss=0.0144
359
+ step 3570000 | 36123s | train_loss=0.0138
360
+ step 3580000 | 36212s | train_loss=0.0140
361
+ step 3590000 | 36305s | train_loss=0.0136
362
+ step 3600000 | 36474s | train_loss=0.0145 | fid=39.7520
363
+ step 3610000 | 36567s | train_loss=0.0141
364
+ step 3620000 | 36662s | train_loss=0.0144
365
+ step 3630000 | 36756s | train_loss=0.0140
366
+ step 3640000 | 36850s | train_loss=0.0137
367
+ step 3650000 | 36944s | train_loss=0.0146
368
+ step 3660000 | 37038s | train_loss=0.0135
369
+ step 3670000 | 37128s | train_loss=0.0133
370
+ step 3680000 | 37221s | train_loss=0.0143
371
+ step 3690000 | 37314s | train_loss=0.0132
372
+ step 3700000 | 37483s | train_loss=0.0132 | fid=39.6327
373
+ step 3710000 | 37577s | train_loss=0.0137
374
+ step 3720000 | 37671s | train_loss=0.0136
375
+ step 3730000 | 37766s | train_loss=0.0135
376
+ step 3740000 | 37859s | train_loss=0.0132
377
+ step 3750000 | 37953s | train_loss=0.0135
378
+ step 3760000 | 38048s | train_loss=0.0140
379
+ step 3770000 | 38139s | train_loss=0.0138
380
+ step 3780000 | 38232s | train_loss=0.0126
381
+ step 3790000 | 38324s | train_loss=0.0131
382
+ step 3800000 | 38495s | train_loss=0.0131 | fid=39.1071
383
+ step 3810000 | 38590s | train_loss=0.0129
384
+ step 3820000 | 38685s | train_loss=0.0133
385
+ step 3830000 | 38779s | train_loss=0.0131
386
+ step 3840000 | 38873s | train_loss=0.0125
387
+ step 3850000 | 38967s | train_loss=0.0128
388
+ step 3860000 | 39058s | train_loss=0.0129
389
+ step 3870000 | 39152s | train_loss=0.0129
390
+ step 3880000 | 39243s | train_loss=0.0136
391
+ step 3890000 | 39337s | train_loss=0.0132
392
+ step 3900000 | 39509s | train_loss=0.0127 | fid=38.8551
393
+ step 3910000 | 39603s | train_loss=0.0128
394
+ step 3920000 | 39697s | train_loss=0.0129
395
+ step 3930000 | 39791s | train_loss=0.0131
396
+ step 3940000 | 39885s | train_loss=0.0132
397
+ step 3950000 | 39977s | train_loss=0.0134
398
+ step 3960000 | 40071s | train_loss=0.0128
399
+ step 3970000 | 40165s | train_loss=0.0125
400
+ step 3980000 | 40255s | train_loss=0.0127
401
+ step 3990000 | 40350s | train_loss=0.0126
402
+ step 4000000 | 40520s | train_loss=0.0125 | fid=38.9166
403
+ step 4010000 | 40615s | train_loss=0.0128
404
+ step 4020000 | 40709s | train_loss=0.0136
405
+ step 4030000 | 40803s | train_loss=0.0120
406
+ step 4040000 | 40897s | train_loss=0.0125
407
+ step 4050000 | 40989s | train_loss=0.0133
408
+ step 4060000 | 41084s | train_loss=0.0122
409
+ step 4070000 | 41178s | train_loss=0.0124
410
+ step 4080000 | 41269s | train_loss=0.0133
411
+ step 4090000 | 41363s | train_loss=0.0122
412
+ step 4100000 | 41534s | train_loss=0.0127 | fid=39.4288
413
+ step 4110000 | 41628s | train_loss=0.0122
414
+ step 4120000 | 41722s | train_loss=0.0124
415
+ step 4130000 | 41817s | train_loss=0.0122
416
+ step 4140000 | 41908s | train_loss=0.0131
417
+ step 4150000 | 42002s | train_loss=0.0132
418
+ step 4160000 | 42097s | train_loss=0.0117
419
+ step 4170000 | 42192s | train_loss=0.0130
420
+ step 4180000 | 42283s | train_loss=0.0122
421
+ step 4190000 | 42377s | train_loss=0.0121
422
+ step 4200000 | 42545s | train_loss=0.0122 | fid=39.0692
423
+ step 4210000 | 42640s | train_loss=0.0120
424
+ step 4220000 | 42734s | train_loss=0.0116
425
+ step 4230000 | 42828s | train_loss=0.0124
426
+ step 4240000 | 42920s | train_loss=0.0118
427
+ step 4250000 | 43014s | train_loss=0.0125
428
+ step 4260000 | 43108s | train_loss=0.0122
429
+ step 4270000 | 43203s | train_loss=0.0122
430
+ step 4280000 | 43294s | train_loss=0.0118
431
+ step 4290000 | 43387s | train_loss=0.0125
432
+ step 4300000 | 43558s | train_loss=0.0116 | fid=39.6727
433
+ step 4310000 | 43652s | train_loss=0.0118
434
+ step 4320000 | 43746s | train_loss=0.0119
435
+ step 4330000 | 43838s | train_loss=0.0117
436
+ step 4340000 | 43932s | train_loss=0.0116
437
+ step 4350000 | 44026s | train_loss=0.0112
438
+ step 4360000 | 44121s | train_loss=0.0118
439
+ step 4370000 | 44215s | train_loss=0.0117
440
+ step 4380000 | 44306s | train_loss=0.0114
441
+ step 4390000 | 44400s | train_loss=0.0127
442
+ step 4400000 | 44568s | train_loss=0.0118 | fid=38.6289
443
+ step 4410000 | 44663s | train_loss=0.0125
444
+ step 4420000 | 44754s | train_loss=0.0120
445
+ step 4430000 | 44848s | train_loss=0.0119
446
+ step 4440000 | 44943s | train_loss=0.0121
447
+ step 4450000 | 45037s | train_loss=0.0117
448
+ step 4460000 | 45131s | train_loss=0.0116
449
+ step 4470000 | 45225s | train_loss=0.0114
450
+ step 4480000 | 45318s | train_loss=0.0117
451
+ step 4490000 | 45412s | train_loss=0.0119
452
+ step 4500000 | 45579s | train_loss=0.0117 | fid=39.4435
453
+ step 4510000 | 45673s | train_loss=0.0115
454
+ step 4520000 | 45765s | train_loss=0.0114
455
+ step 4530000 | 45860s | train_loss=0.0111
456
+ step 4540000 | 45954s | train_loss=0.0113
457
+ step 4550000 | 46048s | train_loss=0.0109
458
+ step 4560000 | 46143s | train_loss=0.0115
459
+ step 4570000 | 46235s | train_loss=0.0112
460
+ step 4580000 | 46327s | train_loss=0.0115
461
+ step 4590000 | 46422s | train_loss=0.0115
462
+ step 4600000 | 46589s | train_loss=0.0110 | fid=39.2207
463
+ step 4610000 | 46681s | train_loss=0.0111
464
+ step 4620000 | 46776s | train_loss=0.0118
465
+ step 4630000 | 46871s | train_loss=0.0117
466
+ step 4640000 | 46966s | train_loss=0.0125
467
+ step 4650000 | 47060s | train_loss=0.0127
468
+ step 4660000 | 47155s | train_loss=0.0116
469
+ step 4670000 | 47249s | train_loss=0.0118
470
+ step 4680000 | 47341s | train_loss=0.0110
471
+ step 4690000 | 47436s | train_loss=0.0116
472
+ step 4700000 | 47602s | train_loss=0.0107 | fid=39.0497
473
+ step 4710000 | 47678s | train_loss=0.0110
474
+ step 4720000 | 47755s | train_loss=0.0115
475
+ step 4730000 | 47831s | train_loss=0.0116
476
+ step 4740000 | 47907s | train_loss=0.0114
477
+ step 4750000 | 47983s | train_loss=0.0110
478
+ step 4760000 | 48059s | train_loss=0.0115
479
+ step 4770000 | 48134s | train_loss=0.0117
480
+ step 4780000 | 48208s | train_loss=0.0118
481
+ step 4790000 | 48284s | train_loss=0.0108
482
+ step 4800000 | 48420s | train_loss=0.0108 | fid=40.3370
483
+ step 4810000 | 48496s | train_loss=0.0111
484
+ step 4820000 | 48572s | train_loss=0.0116
485
+ step 4830000 | 48648s | train_loss=0.0118
486
+ step 4840000 | 48724s | train_loss=0.0110
487
+ step 4850000 | 48801s | train_loss=0.0118
488
+ step 4860000 | 48877s | train_loss=0.0106
489
+ step 4870000 | 48952s | train_loss=0.0112
490
+ step 4880000 | 49026s | train_loss=0.0108
491
+ step 4890000 | 49102s | train_loss=0.0110
492
+ step 4900000 | 49239s | train_loss=0.0105 | fid=38.9635
493
+ step 4910000 | 49315s | train_loss=0.0112
494
+ step 4920000 | 49391s | train_loss=0.0106
495
+ step 4930000 | 49467s | train_loss=0.0104
496
+ step 4940000 | 49544s | train_loss=0.0104
497
+ step 4950000 | 49620s | train_loss=0.0104
498
+ step 4960000 | 49696s | train_loss=0.0105
499
+ step 4970000 | 49764s | train_loss=0.0106
500
+ step 4980000 | 49801s | train_loss=0.0103
501
+ step 4990000 | 49839s | train_loss=0.0102
502
+ step 5000000 | 49898s | train_loss=0.0107 | fid=40.1450
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed44/d90/config.json ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "vae_checkpoint": "vae_checkpoints/celeba_resnet_modernloss_d90/vae.pt",
3
+ "out": "results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed44/d90",
4
+ "subset_path": "diagnostics/celeba_diverse_1k_mean_plus_sd/subset_indices.json",
5
+ "train_tar": "/overflow/data/cerberus/celeba_hq_256_partitioned_ws/train/celeba_hq_256-train-0000.tar",
6
+ "n_train": 1000,
7
+ "total_steps": 5000000,
8
+ "eval_interval": 10000,
9
+ "mem_interval": 100000,
10
+ "fid_interval": 100000,
11
+ "fid_n_real": 1000,
12
+ "fid_n_gen": 10000,
13
+ "n_gen_samples": 10000,
14
+ "n_sde_steps": 500,
15
+ "batch_size": 512,
16
+ "hidden": 1024,
17
+ "depth": 5,
18
+ "optimizer": "sgd",
19
+ "lr": 0.001,
20
+ "momentum": 0.8,
21
+ "t_min": 0.01,
22
+ "t_max": 3.0,
23
+ "t_eval": 0.1,
24
+ "seed": 44,
25
+ "device": null
26
+ }
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed44/d90/metrics.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed44/d90/runner.log ADDED
@@ -0,0 +1,502 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ FID real-side stats cached over 1000 diverse train images
2
+ step 1 | 73s | train_loss=0.9954 | fid=308.9319
3
+ step 10000 | 166s | train_loss=0.8115
4
+ step 20000 | 260s | train_loss=0.6939
5
+ step 30000 | 354s | train_loss=0.5605
6
+ step 40000 | 448s | train_loss=0.4136
7
+ step 50000 | 542s | train_loss=0.2936
8
+ step 60000 | 636s | train_loss=0.2179
9
+ step 70000 | 730s | train_loss=0.1781
10
+ step 80000 | 824s | train_loss=0.1529
11
+ step 90000 | 917s | train_loss=0.1384
12
+ step 100000 | 1082s | train_loss=0.1277 | fid=60.9427
13
+ step 110000 | 1176s | train_loss=0.1218
14
+ step 120000 | 1270s | train_loss=0.1144
15
+ step 130000 | 1364s | train_loss=0.1080
16
+ step 140000 | 1458s | train_loss=0.1046
17
+ step 150000 | 1553s | train_loss=0.1010
18
+ step 160000 | 1647s | train_loss=0.0982
19
+ step 170000 | 1741s | train_loss=0.0954
20
+ step 180000 | 1835s | train_loss=0.0887
21
+ step 190000 | 1928s | train_loss=0.0863
22
+ step 200000 | 2092s | train_loss=0.0857 | fid=59.7462
23
+ step 210000 | 2185s | train_loss=0.0817
24
+ step 220000 | 2280s | train_loss=0.0788
25
+ step 230000 | 2374s | train_loss=0.0773
26
+ step 240000 | 2468s | train_loss=0.0746
27
+ step 250000 | 2563s | train_loss=0.0730
28
+ step 260000 | 2657s | train_loss=0.0717
29
+ step 270000 | 2751s | train_loss=0.0691
30
+ step 280000 | 2844s | train_loss=0.0696
31
+ step 290000 | 2936s | train_loss=0.0681
32
+ step 300000 | 3101s | train_loss=0.0667 | fid=56.3101
33
+ step 310000 | 3195s | train_loss=0.0643
34
+ step 320000 | 3289s | train_loss=0.0640
35
+ step 330000 | 3384s | train_loss=0.0625
36
+ step 340000 | 3478s | train_loss=0.0616
37
+ step 350000 | 3572s | train_loss=0.0614
38
+ step 360000 | 3667s | train_loss=0.0599
39
+ step 370000 | 3760s | train_loss=0.0598
40
+ step 380000 | 3853s | train_loss=0.0581
41
+ step 390000 | 3946s | train_loss=0.0575
42
+ step 400000 | 4110s | train_loss=0.0566 | fid=54.5504
43
+ step 410000 | 4204s | train_loss=0.0572
44
+ step 420000 | 4298s | train_loss=0.0553
45
+ step 430000 | 4393s | train_loss=0.0560
46
+ step 440000 | 4487s | train_loss=0.0547
47
+ step 450000 | 4581s | train_loss=0.0530
48
+ step 460000 | 4674s | train_loss=0.0517
49
+ step 470000 | 4768s | train_loss=0.0533
50
+ step 480000 | 4861s | train_loss=0.0529
51
+ step 490000 | 4955s | train_loss=0.0521
52
+ step 500000 | 5119s | train_loss=0.0506 | fid=52.7284
53
+ step 510000 | 5213s | train_loss=0.0513
54
+ step 520000 | 5307s | train_loss=0.0501
55
+ step 530000 | 5402s | train_loss=0.0493
56
+ step 540000 | 5496s | train_loss=0.0483
57
+ step 550000 | 5591s | train_loss=0.0487
58
+ step 560000 | 5684s | train_loss=0.0478
59
+ step 570000 | 5777s | train_loss=0.0493
60
+ step 580000 | 5871s | train_loss=0.0467
61
+ step 590000 | 5965s | train_loss=0.0462
62
+ step 600000 | 6130s | train_loss=0.0481 | fid=53.2857
63
+ step 610000 | 6223s | train_loss=0.0471
64
+ step 620000 | 6318s | train_loss=0.0468
65
+ step 630000 | 6412s | train_loss=0.0449
66
+ step 640000 | 6506s | train_loss=0.0442
67
+ step 650000 | 6600s | train_loss=0.0457
68
+ step 660000 | 6693s | train_loss=0.0453
69
+ step 670000 | 6786s | train_loss=0.0439
70
+ step 680000 | 6880s | train_loss=0.0427
71
+ step 690000 | 6974s | train_loss=0.0441
72
+ step 700000 | 7140s | train_loss=0.0426 | fid=51.2337
73
+ step 710000 | 7233s | train_loss=0.0432
74
+ step 720000 | 7328s | train_loss=0.0418
75
+ step 730000 | 7422s | train_loss=0.0431
76
+ step 740000 | 7516s | train_loss=0.0428
77
+ step 750000 | 7611s | train_loss=0.0406
78
+ step 760000 | 7703s | train_loss=0.0402
79
+ step 770000 | 7798s | train_loss=0.0410
80
+ step 780000 | 7892s | train_loss=0.0401
81
+ step 790000 | 7987s | train_loss=0.0396
82
+ step 800000 | 8153s | train_loss=0.0413 | fid=50.2696
83
+ step 810000 | 8246s | train_loss=0.0401
84
+ step 820000 | 8341s | train_loss=0.0395
85
+ step 830000 | 8435s | train_loss=0.0387
86
+ step 840000 | 8529s | train_loss=0.0394
87
+ step 850000 | 8622s | train_loss=0.0392
88
+ step 860000 | 8716s | train_loss=0.0390
89
+ step 870000 | 8810s | train_loss=0.0393
90
+ step 880000 | 8905s | train_loss=0.0383
91
+ step 890000 | 8999s | train_loss=0.0394
92
+ step 900000 | 9165s | train_loss=0.0374 | fid=49.2707
93
+ step 910000 | 9258s | train_loss=0.0376
94
+ step 920000 | 9352s | train_loss=0.0381
95
+ step 930000 | 9447s | train_loss=0.0374
96
+ step 940000 | 9541s | train_loss=0.0363
97
+ step 950000 | 9633s | train_loss=0.0367
98
+ step 960000 | 9728s | train_loss=0.0365
99
+ step 970000 | 9822s | train_loss=0.0379
100
+ step 980000 | 9917s | train_loss=0.0352
101
+ step 990000 | 10011s | train_loss=0.0358
102
+ step 1000000 | 10178s | train_loss=0.0356 | fid=48.5643
103
+ step 1010000 | 10271s | train_loss=0.0347
104
+ step 1020000 | 10366s | train_loss=0.0354
105
+ step 1030000 | 10460s | train_loss=0.0354
106
+ step 1040000 | 10553s | train_loss=0.0357
107
+ step 1050000 | 10647s | train_loss=0.0363
108
+ step 1060000 | 10741s | train_loss=0.0345
109
+ step 1070000 | 10836s | train_loss=0.0341
110
+ step 1080000 | 10930s | train_loss=0.0345
111
+ step 1090000 | 11024s | train_loss=0.0348
112
+ step 1100000 | 11191s | train_loss=0.0342 | fid=46.8351
113
+ step 1110000 | 11284s | train_loss=0.0338
114
+ step 1120000 | 11378s | train_loss=0.0336
115
+ step 1130000 | 11470s | train_loss=0.0346
116
+ step 1140000 | 11564s | train_loss=0.0334
117
+ step 1150000 | 11658s | train_loss=0.0322
118
+ step 1160000 | 11753s | train_loss=0.0332
119
+ step 1170000 | 11847s | train_loss=0.0337
120
+ step 1180000 | 11941s | train_loss=0.0335
121
+ step 1190000 | 12035s | train_loss=0.0326
122
+ step 1200000 | 12201s | train_loss=0.0325 | fid=46.0755
123
+ step 1210000 | 12294s | train_loss=0.0312
124
+ step 1220000 | 12388s | train_loss=0.0321
125
+ step 1230000 | 12480s | train_loss=0.0320
126
+ step 1240000 | 12574s | train_loss=0.0325
127
+ step 1250000 | 12668s | train_loss=0.0316
128
+ step 1260000 | 12763s | train_loss=0.0315
129
+ step 1270000 | 12857s | train_loss=0.0326
130
+ step 1280000 | 12952s | train_loss=0.0311
131
+ step 1290000 | 13047s | train_loss=0.0318
132
+ step 1300000 | 13213s | train_loss=0.0309 | fid=46.4013
133
+ step 1310000 | 13306s | train_loss=0.0301
134
+ step 1320000 | 13398s | train_loss=0.0304
135
+ step 1330000 | 13492s | train_loss=0.0299
136
+ step 1340000 | 13586s | train_loss=0.0303
137
+ step 1350000 | 13680s | train_loss=0.0307
138
+ step 1360000 | 13774s | train_loss=0.0294
139
+ step 1370000 | 13868s | train_loss=0.0303
140
+ step 1380000 | 13962s | train_loss=0.0296
141
+ step 1390000 | 14057s | train_loss=0.0293
142
+ step 1400000 | 14223s | train_loss=0.0297 | fid=45.4498
143
+ step 1410000 | 14315s | train_loss=0.0288
144
+ step 1420000 | 14409s | train_loss=0.0297
145
+ step 1430000 | 14502s | train_loss=0.0294
146
+ step 1440000 | 14597s | train_loss=0.0291
147
+ step 1450000 | 14692s | train_loss=0.0286
148
+ step 1460000 | 14786s | train_loss=0.0282
149
+ step 1470000 | 14880s | train_loss=0.0302
150
+ step 1480000 | 14975s | train_loss=0.0285
151
+ step 1490000 | 15069s | train_loss=0.0294
152
+ step 1500000 | 15235s | train_loss=0.0286 | fid=45.0048
153
+ step 1510000 | 15327s | train_loss=0.0292
154
+ step 1520000 | 15421s | train_loss=0.0280
155
+ step 1530000 | 15515s | train_loss=0.0286
156
+ step 1540000 | 15609s | train_loss=0.0270
157
+ step 1550000 | 15704s | train_loss=0.0271
158
+ step 1560000 | 15798s | train_loss=0.0269
159
+ step 1570000 | 15893s | train_loss=0.0272
160
+ step 1580000 | 15987s | train_loss=0.0271
161
+ step 1590000 | 16082s | train_loss=0.0276
162
+ step 1600000 | 16246s | train_loss=0.0265 | fid=43.5951
163
+ step 1610000 | 16340s | train_loss=0.0267
164
+ step 1620000 | 16433s | train_loss=0.0270
165
+ step 1630000 | 16527s | train_loss=0.0262
166
+ step 1640000 | 16622s | train_loss=0.0267
167
+ step 1650000 | 16716s | train_loss=0.0269
168
+ step 1660000 | 16811s | train_loss=0.0259
169
+ step 1670000 | 16905s | train_loss=0.0273
170
+ step 1680000 | 16999s | train_loss=0.0263
171
+ step 1690000 | 17094s | train_loss=0.0262
172
+ step 1700000 | 17257s | train_loss=0.0254 | fid=42.6083
173
+ step 1710000 | 17351s | train_loss=0.0264
174
+ step 1720000 | 17444s | train_loss=0.0257
175
+ step 1730000 | 17539s | train_loss=0.0248
176
+ step 1740000 | 17633s | train_loss=0.0250
177
+ step 1750000 | 17728s | train_loss=0.0264
178
+ step 1760000 | 17822s | train_loss=0.0255
179
+ step 1770000 | 17917s | train_loss=0.0256
180
+ step 1780000 | 18011s | train_loss=0.0255
181
+ step 1790000 | 18105s | train_loss=0.0246
182
+ step 1800000 | 18269s | train_loss=0.0245 | fid=42.3962
183
+ step 1810000 | 18362s | train_loss=0.0243
184
+ step 1820000 | 18456s | train_loss=0.0245
185
+ step 1830000 | 18549s | train_loss=0.0247
186
+ step 1840000 | 18644s | train_loss=0.0245
187
+ step 1850000 | 18739s | train_loss=0.0246
188
+ step 1860000 | 18833s | train_loss=0.0246
189
+ step 1870000 | 18927s | train_loss=0.0250
190
+ step 1880000 | 19022s | train_loss=0.0240
191
+ step 1890000 | 19116s | train_loss=0.0245
192
+ step 1900000 | 19282s | train_loss=0.0232 | fid=42.1686
193
+ step 1910000 | 19375s | train_loss=0.0237
194
+ step 1920000 | 19470s | train_loss=0.0237
195
+ step 1930000 | 19564s | train_loss=0.0233
196
+ step 1940000 | 19659s | train_loss=0.0230
197
+ step 1950000 | 19753s | train_loss=0.0239
198
+ step 1960000 | 19848s | train_loss=0.0228
199
+ step 1970000 | 19942s | train_loss=0.0235
200
+ step 1980000 | 20035s | train_loss=0.0233
201
+ step 1990000 | 20129s | train_loss=0.0231
202
+ step 2000000 | 20297s | train_loss=0.0232 | fid=41.4898
203
+ step 2010000 | 20389s | train_loss=0.0242
204
+ step 2020000 | 20484s | train_loss=0.0237
205
+ step 2030000 | 20578s | train_loss=0.0230
206
+ step 2040000 | 20672s | train_loss=0.0229
207
+ step 2050000 | 20767s | train_loss=0.0223
208
+ step 2060000 | 20861s | train_loss=0.0224
209
+ step 2070000 | 20955s | train_loss=0.0232
210
+ step 2080000 | 21048s | train_loss=0.0226
211
+ step 2090000 | 21142s | train_loss=0.0220
212
+ step 2100000 | 21308s | train_loss=0.0232 | fid=41.5736
213
+ step 2110000 | 21402s | train_loss=0.0217
214
+ step 2120000 | 21496s | train_loss=0.0218
215
+ step 2130000 | 21590s | train_loss=0.0221
216
+ step 2140000 | 21685s | train_loss=0.0225
217
+ step 2150000 | 21779s | train_loss=0.0219
218
+ step 2160000 | 21873s | train_loss=0.0220
219
+ step 2170000 | 21966s | train_loss=0.0224
220
+ step 2180000 | 22060s | train_loss=0.0226
221
+ step 2190000 | 22154s | train_loss=0.0211
222
+ step 2200000 | 22321s | train_loss=0.0219 | fid=41.2080
223
+ step 2210000 | 22414s | train_loss=0.0219
224
+ step 2220000 | 22509s | train_loss=0.0216
225
+ step 2230000 | 22603s | train_loss=0.0218
226
+ step 2240000 | 22697s | train_loss=0.0223
227
+ step 2250000 | 22791s | train_loss=0.0206
228
+ step 2260000 | 22883s | train_loss=0.0217
229
+ step 2270000 | 22977s | train_loss=0.0210
230
+ step 2280000 | 23071s | train_loss=0.0209
231
+ step 2290000 | 23165s | train_loss=0.0210
232
+ step 2300000 | 23332s | train_loss=0.0209 | fid=41.1284
233
+ step 2310000 | 23424s | train_loss=0.0218
234
+ step 2320000 | 23519s | train_loss=0.0208
235
+ step 2330000 | 23613s | train_loss=0.0204
236
+ step 2340000 | 23706s | train_loss=0.0200
237
+ step 2350000 | 23800s | train_loss=0.0195
238
+ step 2360000 | 23893s | train_loss=0.0209
239
+ step 2370000 | 23987s | train_loss=0.0201
240
+ step 2380000 | 24081s | train_loss=0.0206
241
+ step 2390000 | 24175s | train_loss=0.0200
242
+ step 2400000 | 24341s | train_loss=0.0201 | fid=41.0360
243
+ step 2410000 | 24435s | train_loss=0.0198
244
+ step 2420000 | 24530s | train_loss=0.0208
245
+ step 2430000 | 24624s | train_loss=0.0214
246
+ step 2440000 | 24717s | train_loss=0.0195
247
+ step 2450000 | 24810s | train_loss=0.0195
248
+ step 2460000 | 24904s | train_loss=0.0209
249
+ step 2470000 | 24998s | train_loss=0.0197
250
+ step 2480000 | 25093s | train_loss=0.0195
251
+ step 2490000 | 25186s | train_loss=0.0198
252
+ step 2500000 | 25354s | train_loss=0.0194 | fid=40.5056
253
+ step 2510000 | 25447s | train_loss=0.0197
254
+ step 2520000 | 25541s | train_loss=0.0200
255
+ step 2530000 | 25635s | train_loss=0.0203
256
+ step 2540000 | 25726s | train_loss=0.0190
257
+ step 2550000 | 25820s | train_loss=0.0191
258
+ step 2560000 | 25915s | train_loss=0.0196
259
+ step 2570000 | 26009s | train_loss=0.0185
260
+ step 2580000 | 26103s | train_loss=0.0187
261
+ step 2590000 | 26195s | train_loss=0.0194
262
+ step 2600000 | 26363s | train_loss=0.0188 | fid=40.2189
263
+ step 2610000 | 26456s | train_loss=0.0195
264
+ step 2620000 | 26550s | train_loss=0.0193
265
+ step 2630000 | 26644s | train_loss=0.0185
266
+ step 2640000 | 26737s | train_loss=0.0181
267
+ step 2650000 | 26831s | train_loss=0.0192
268
+ step 2660000 | 26925s | train_loss=0.0184
269
+ step 2670000 | 27019s | train_loss=0.0184
270
+ step 2680000 | 27113s | train_loss=0.0187
271
+ step 2690000 | 27205s | train_loss=0.0185
272
+ step 2700000 | 27373s | train_loss=0.0181 | fid=40.0348
273
+ step 2710000 | 27466s | train_loss=0.0183
274
+ step 2720000 | 27560s | train_loss=0.0187
275
+ step 2730000 | 27652s | train_loss=0.0178
276
+ step 2740000 | 27746s | train_loss=0.0183
277
+ step 2750000 | 27840s | train_loss=0.0180
278
+ step 2760000 | 27934s | train_loss=0.0187
279
+ step 2770000 | 28028s | train_loss=0.0195
280
+ step 2780000 | 28123s | train_loss=0.0184
281
+ step 2790000 | 28214s | train_loss=0.0182
282
+ step 2800000 | 28382s | train_loss=0.0186 | fid=40.1194
283
+ step 2810000 | 28475s | train_loss=0.0184
284
+ step 2820000 | 28568s | train_loss=0.0181
285
+ step 2830000 | 28662s | train_loss=0.0181
286
+ step 2840000 | 28757s | train_loss=0.0185
287
+ step 2850000 | 28850s | train_loss=0.0177
288
+ step 2860000 | 28945s | train_loss=0.0182
289
+ step 2870000 | 29039s | train_loss=0.0170
290
+ step 2880000 | 29134s | train_loss=0.0178
291
+ step 2890000 | 29226s | train_loss=0.0179
292
+ step 2900000 | 29393s | train_loss=0.0176 | fid=40.1471
293
+ step 2910000 | 29486s | train_loss=0.0165
294
+ step 2920000 | 29579s | train_loss=0.0185
295
+ step 2930000 | 29673s | train_loss=0.0171
296
+ step 2940000 | 29767s | train_loss=0.0170
297
+ step 2950000 | 29862s | train_loss=0.0170
298
+ step 2960000 | 29955s | train_loss=0.0175
299
+ step 2970000 | 30050s | train_loss=0.0169
300
+ step 2980000 | 30144s | train_loss=0.0169
301
+ step 2990000 | 30235s | train_loss=0.0174
302
+ step 3000000 | 30403s | train_loss=0.0175 | fid=38.9835
303
+ step 3010000 | 30495s | train_loss=0.0165
304
+ step 3020000 | 30589s | train_loss=0.0169
305
+ step 3030000 | 30684s | train_loss=0.0164
306
+ step 3040000 | 30778s | train_loss=0.0161
307
+ step 3050000 | 30872s | train_loss=0.0159
308
+ step 3060000 | 30966s | train_loss=0.0168
309
+ step 3070000 | 31061s | train_loss=0.0166
310
+ step 3080000 | 31155s | train_loss=0.0166
311
+ step 3090000 | 31246s | train_loss=0.0164
312
+ step 3100000 | 31414s | train_loss=0.0170 | fid=39.9573
313
+ step 3110000 | 31507s | train_loss=0.0161
314
+ step 3120000 | 31601s | train_loss=0.0168
315
+ step 3130000 | 31695s | train_loss=0.0174
316
+ step 3140000 | 31789s | train_loss=0.0167
317
+ step 3150000 | 31883s | train_loss=0.0160
318
+ step 3160000 | 31977s | train_loss=0.0168
319
+ step 3170000 | 32071s | train_loss=0.0163
320
+ step 3180000 | 32166s | train_loss=0.0161
321
+ step 3190000 | 32258s | train_loss=0.0159
322
+ step 3200000 | 32424s | train_loss=0.0167 | fid=39.4870
323
+ step 3210000 | 32517s | train_loss=0.0174
324
+ step 3220000 | 32612s | train_loss=0.0169
325
+ step 3230000 | 32706s | train_loss=0.0158
326
+ step 3240000 | 32800s | train_loss=0.0161
327
+ step 3250000 | 32894s | train_loss=0.0152
328
+ step 3260000 | 32988s | train_loss=0.0161
329
+ step 3270000 | 33082s | train_loss=0.0166
330
+ step 3280000 | 33177s | train_loss=0.0160
331
+ step 3290000 | 33268s | train_loss=0.0168
332
+ step 3300000 | 33435s | train_loss=0.0160 | fid=40.0068
333
+ step 3310000 | 33527s | train_loss=0.0158
334
+ step 3320000 | 33621s | train_loss=0.0156
335
+ step 3330000 | 33715s | train_loss=0.0155
336
+ step 3340000 | 33809s | train_loss=0.0168
337
+ step 3350000 | 33904s | train_loss=0.0163
338
+ step 3360000 | 33998s | train_loss=0.0154
339
+ step 3370000 | 34092s | train_loss=0.0166
340
+ step 3380000 | 34187s | train_loss=0.0158
341
+ step 3390000 | 34276s | train_loss=0.0159
342
+ step 3400000 | 34445s | train_loss=0.0157 | fid=38.9844
343
+ step 3410000 | 34537s | train_loss=0.0157
344
+ step 3420000 | 34631s | train_loss=0.0170
345
+ step 3430000 | 34725s | train_loss=0.0151
346
+ step 3440000 | 34819s | train_loss=0.0164
347
+ step 3450000 | 34914s | train_loss=0.0150
348
+ step 3460000 | 35007s | train_loss=0.0155
349
+ step 3470000 | 35101s | train_loss=0.0149
350
+ step 3480000 | 35195s | train_loss=0.0160
351
+ step 3490000 | 35284s | train_loss=0.0156
352
+ step 3500000 | 35453s | train_loss=0.0153 | fid=39.2143
353
+ step 3510000 | 35544s | train_loss=0.0152
354
+ step 3520000 | 35638s | train_loss=0.0157
355
+ step 3530000 | 35733s | train_loss=0.0149
356
+ step 3540000 | 35827s | train_loss=0.0152
357
+ step 3550000 | 35921s | train_loss=0.0146
358
+ step 3560000 | 36015s | train_loss=0.0149
359
+ step 3570000 | 36108s | train_loss=0.0143
360
+ step 3580000 | 36199s | train_loss=0.0160
361
+ step 3590000 | 36291s | train_loss=0.0148
362
+ step 3600000 | 36459s | train_loss=0.0158 | fid=38.7056
363
+ step 3610000 | 36551s | train_loss=0.0164
364
+ step 3620000 | 36645s | train_loss=0.0147
365
+ step 3630000 | 36739s | train_loss=0.0156
366
+ step 3640000 | 36833s | train_loss=0.0142
367
+ step 3650000 | 36927s | train_loss=0.0146
368
+ step 3660000 | 37021s | train_loss=0.0154
369
+ step 3670000 | 37114s | train_loss=0.0140
370
+ step 3680000 | 37205s | train_loss=0.0140
371
+ step 3690000 | 37298s | train_loss=0.0156
372
+ step 3700000 | 37466s | train_loss=0.0146 | fid=39.4540
373
+ step 3710000 | 37558s | train_loss=0.0144
374
+ step 3720000 | 37652s | train_loss=0.0144
375
+ step 3730000 | 37746s | train_loss=0.0146
376
+ step 3740000 | 37840s | train_loss=0.0141
377
+ step 3750000 | 37934s | train_loss=0.0146
378
+ step 3760000 | 38028s | train_loss=0.0141
379
+ step 3770000 | 38119s | train_loss=0.0146
380
+ step 3780000 | 38213s | train_loss=0.0140
381
+ step 3790000 | 38305s | train_loss=0.0149
382
+ step 3800000 | 38474s | train_loss=0.0143 | fid=38.9222
383
+ step 3810000 | 38566s | train_loss=0.0148
384
+ step 3820000 | 38660s | train_loss=0.0158
385
+ step 3830000 | 38755s | train_loss=0.0144
386
+ step 3840000 | 38848s | train_loss=0.0149
387
+ step 3850000 | 38943s | train_loss=0.0140
388
+ step 3860000 | 39034s | train_loss=0.0146
389
+ step 3870000 | 39128s | train_loss=0.0146
390
+ step 3880000 | 39221s | train_loss=0.0156
391
+ step 3890000 | 39313s | train_loss=0.0140
392
+ step 3900000 | 39481s | train_loss=0.0146 | fid=38.8990
393
+ step 3910000 | 39573s | train_loss=0.0142
394
+ step 3920000 | 39667s | train_loss=0.0141
395
+ step 3930000 | 39760s | train_loss=0.0135
396
+ step 3940000 | 39855s | train_loss=0.0143
397
+ step 3950000 | 39949s | train_loss=0.0143
398
+ step 3960000 | 40041s | train_loss=0.0136
399
+ step 3970000 | 40135s | train_loss=0.0142
400
+ step 3980000 | 40227s | train_loss=0.0142
401
+ step 3990000 | 40320s | train_loss=0.0139
402
+ step 4000000 | 40488s | train_loss=0.0138 | fid=39.4720
403
+ step 4010000 | 40580s | train_loss=0.0136
404
+ step 4020000 | 40674s | train_loss=0.0145
405
+ step 4030000 | 40768s | train_loss=0.0135
406
+ step 4040000 | 40862s | train_loss=0.0135
407
+ step 4050000 | 40954s | train_loss=0.0130
408
+ step 4060000 | 41048s | train_loss=0.0137
409
+ step 4070000 | 41142s | train_loss=0.0134
410
+ step 4080000 | 41236s | train_loss=0.0135
411
+ step 4090000 | 41327s | train_loss=0.0138
412
+ step 4100000 | 41496s | train_loss=0.0136 | fid=38.9457
413
+ step 4110000 | 41587s | train_loss=0.0135
414
+ step 4120000 | 41682s | train_loss=0.0127
415
+ step 4130000 | 41776s | train_loss=0.0133
416
+ step 4140000 | 41871s | train_loss=0.0131
417
+ step 4150000 | 41962s | train_loss=0.0130
418
+ step 4160000 | 42056s | train_loss=0.0133
419
+ step 4170000 | 42151s | train_loss=0.0125
420
+ step 4180000 | 42245s | train_loss=0.0140
421
+ step 4190000 | 42337s | train_loss=0.0135
422
+ step 4200000 | 42505s | train_loss=0.0133 | fid=38.8201
423
+ step 4210000 | 42597s | train_loss=0.0127
424
+ step 4220000 | 42692s | train_loss=0.0136
425
+ step 4230000 | 42786s | train_loss=0.0131
426
+ step 4240000 | 42877s | train_loss=0.0134
427
+ step 4250000 | 42971s | train_loss=0.0135
428
+ step 4260000 | 43066s | train_loss=0.0141
429
+ step 4270000 | 43160s | train_loss=0.0134
430
+ step 4280000 | 43253s | train_loss=0.0128
431
+ step 4290000 | 43345s | train_loss=0.0131
432
+ step 4300000 | 43513s | train_loss=0.0136 | fid=39.0948
433
+ step 4310000 | 43605s | train_loss=0.0126
434
+ step 4320000 | 43699s | train_loss=0.0133
435
+ step 4330000 | 43791s | train_loss=0.0126
436
+ step 4340000 | 43885s | train_loss=0.0132
437
+ step 4350000 | 43980s | train_loss=0.0128
438
+ step 4360000 | 44074s | train_loss=0.0128
439
+ step 4370000 | 44168s | train_loss=0.0129
440
+ step 4380000 | 44260s | train_loss=0.0137
441
+ step 4390000 | 44354s | train_loss=0.0128
442
+ step 4400000 | 44521s | train_loss=0.0130 | fid=39.0468
443
+ step 4410000 | 44613s | train_loss=0.0133
444
+ step 4420000 | 44707s | train_loss=0.0130
445
+ step 4430000 | 44799s | train_loss=0.0129
446
+ step 4440000 | 44893s | train_loss=0.0128
447
+ step 4450000 | 44988s | train_loss=0.0125
448
+ step 4460000 | 45082s | train_loss=0.0139
449
+ step 4470000 | 45176s | train_loss=0.0144
450
+ step 4480000 | 45268s | train_loss=0.0120
451
+ step 4490000 | 45363s | train_loss=0.0125
452
+ step 4500000 | 45531s | train_loss=0.0127 | fid=39.4384
453
+ step 4510000 | 45624s | train_loss=0.0122
454
+ step 4520000 | 45716s | train_loss=0.0128
455
+ step 4530000 | 45810s | train_loss=0.0143
456
+ step 4540000 | 45904s | train_loss=0.0129
457
+ step 4550000 | 45999s | train_loss=0.0120
458
+ step 4560000 | 46093s | train_loss=0.0126
459
+ step 4570000 | 46187s | train_loss=0.0122
460
+ step 4580000 | 46278s | train_loss=0.0125
461
+ step 4590000 | 46372s | train_loss=0.0127
462
+ step 4600000 | 46539s | train_loss=0.0124 | fid=39.2159
463
+ step 4610000 | 46632s | train_loss=0.0129
464
+ step 4620000 | 46725s | train_loss=0.0130
465
+ step 4630000 | 46820s | train_loss=0.0128
466
+ step 4640000 | 46914s | train_loss=0.0120
467
+ step 4650000 | 47009s | train_loss=0.0119
468
+ step 4660000 | 47104s | train_loss=0.0131
469
+ step 4670000 | 47198s | train_loss=0.0126
470
+ step 4680000 | 47290s | train_loss=0.0125
471
+ step 4690000 | 47384s | train_loss=0.0144
472
+ step 4700000 | 47553s | train_loss=0.0129 | fid=39.5688
473
+ step 4710000 | 47633s | train_loss=0.0131
474
+ step 4720000 | 47709s | train_loss=0.0123
475
+ step 4730000 | 47786s | train_loss=0.0123
476
+ step 4740000 | 47862s | train_loss=0.0119
477
+ step 4750000 | 47938s | train_loss=0.0130
478
+ step 4760000 | 48014s | train_loss=0.0119
479
+ step 4770000 | 48091s | train_loss=0.0115
480
+ step 4780000 | 48164s | train_loss=0.0117
481
+ step 4790000 | 48240s | train_loss=0.0124
482
+ step 4800000 | 48376s | train_loss=0.0131 | fid=38.6088
483
+ step 4810000 | 48449s | train_loss=0.0122
484
+ step 4820000 | 48526s | train_loss=0.0122
485
+ step 4830000 | 48602s | train_loss=0.0122
486
+ step 4840000 | 48679s | train_loss=0.0116
487
+ step 4850000 | 48755s | train_loss=0.0121
488
+ step 4860000 | 48832s | train_loss=0.0123
489
+ step 4870000 | 48908s | train_loss=0.0117
490
+ step 4880000 | 48980s | train_loss=0.0120
491
+ step 4890000 | 49057s | train_loss=0.0127
492
+ step 4900000 | 49194s | train_loss=0.0123 | fid=39.4494
493
+ step 4910000 | 49267s | train_loss=0.0116
494
+ step 4920000 | 49343s | train_loss=0.0120
495
+ step 4930000 | 49419s | train_loss=0.0119
496
+ step 4940000 | 49495s | train_loss=0.0119
497
+ step 4950000 | 49571s | train_loss=0.0118
498
+ step 4960000 | 49648s | train_loss=0.0112
499
+ step 4970000 | 49724s | train_loss=0.0114
500
+ step 4980000 | 49778s | train_loss=0.0113
501
+ step 4990000 | 49816s | train_loss=0.0118
502
+ step 5000000 | 49888s | train_loss=0.0125 | fid=39.3011
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d10/config.json ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "vae_checkpoint": "vae_checkpoints/celeba_resnet_modernloss_d10/vae.pt",
3
+ "out": "results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d10",
4
+ "subset_path": "diagnostics/celeba_diverse_1k_mean_plus_sd/subset_indices.json",
5
+ "train_tar": "/overflow/data/cerberus/celeba_hq_256_partitioned_ws/train/celeba_hq_256-train-0000.tar",
6
+ "n_train": 1000,
7
+ "total_steps": 5000000,
8
+ "eval_interval": 10000,
9
+ "mem_interval": 100000,
10
+ "fid_interval": 100000,
11
+ "fid_n_real": 1000,
12
+ "fid_n_gen": 10000,
13
+ "n_gen_samples": 10000,
14
+ "n_sde_steps": 500,
15
+ "batch_size": 512,
16
+ "hidden": 1024,
17
+ "depth": 5,
18
+ "optimizer": "sgd",
19
+ "lr": 0.001,
20
+ "momentum": 0.8,
21
+ "t_min": 0.01,
22
+ "t_max": 3.0,
23
+ "t_eval": 0.1,
24
+ "seed": 45,
25
+ "device": null
26
+ }
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d10/metrics.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d10/runner.log ADDED
@@ -0,0 +1,502 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ FID real-side stats cached over 1000 diverse train images
2
+ step 1 | 72s | train_loss=1.0279 | fid=193.6372
3
+ step 10000 | 163s | train_loss=0.9494
4
+ step 20000 | 255s | train_loss=0.9471
5
+ step 30000 | 348s | train_loss=0.9457
6
+ step 40000 | 440s | train_loss=0.9449
7
+ step 50000 | 532s | train_loss=0.9441
8
+ step 60000 | 625s | train_loss=0.9433
9
+ step 70000 | 717s | train_loss=0.9427
10
+ step 80000 | 809s | train_loss=0.9421
11
+ step 90000 | 901s | train_loss=0.9414
12
+ step 100000 | 1064s | train_loss=0.9409 | fid=87.4519
13
+ step 110000 | 1152s | train_loss=0.9403
14
+ step 120000 | 1244s | train_loss=0.9397
15
+ step 130000 | 1337s | train_loss=0.9391
16
+ step 140000 | 1429s | train_loss=0.9384
17
+ step 150000 | 1522s | train_loss=0.9378
18
+ step 160000 | 1615s | train_loss=0.9370
19
+ step 170000 | 1707s | train_loss=0.9360
20
+ step 180000 | 1800s | train_loss=0.9355
21
+ step 190000 | 1893s | train_loss=0.9344
22
+ step 200000 | 2058s | train_loss=0.9336 | fid=87.6089
23
+ step 210000 | 2146s | train_loss=0.9325
24
+ step 220000 | 2239s | train_loss=0.9312
25
+ step 230000 | 2332s | train_loss=0.9300
26
+ step 240000 | 2425s | train_loss=0.9283
27
+ step 250000 | 2518s | train_loss=0.9266
28
+ step 260000 | 2611s | train_loss=0.9247
29
+ step 270000 | 2704s | train_loss=0.9223
30
+ step 280000 | 2797s | train_loss=0.9192
31
+ step 290000 | 2891s | train_loss=0.9155
32
+ step 300000 | 3056s | train_loss=0.9110 | fid=86.3083
33
+ step 310000 | 3146s | train_loss=0.9043
34
+ step 320000 | 3237s | train_loss=0.8952
35
+ step 330000 | 3331s | train_loss=0.8836
36
+ step 340000 | 3424s | train_loss=0.8656
37
+ step 350000 | 3517s | train_loss=0.8458
38
+ step 360000 | 3610s | train_loss=0.8218
39
+ step 370000 | 3703s | train_loss=0.7929
40
+ step 380000 | 3796s | train_loss=0.7524
41
+ step 390000 | 3887s | train_loss=0.7176
42
+ step 400000 | 4054s | train_loss=0.6671 | fid=87.9228
43
+ step 410000 | 4144s | train_loss=0.6275
44
+ step 420000 | 4236s | train_loss=0.5853
45
+ step 430000 | 4329s | train_loss=0.5472
46
+ step 440000 | 4422s | train_loss=0.5108
47
+ step 450000 | 4515s | train_loss=0.4741
48
+ step 460000 | 4608s | train_loss=0.4502
49
+ step 470000 | 4701s | train_loss=0.4150
50
+ step 480000 | 4791s | train_loss=0.3974
51
+ step 490000 | 4884s | train_loss=0.3743
52
+ step 500000 | 5051s | train_loss=0.3618 | fid=85.9419
53
+ step 510000 | 5141s | train_loss=0.3464
54
+ step 520000 | 5234s | train_loss=0.3273
55
+ step 530000 | 5326s | train_loss=0.3228
56
+ step 540000 | 5419s | train_loss=0.3042
57
+ step 550000 | 5512s | train_loss=0.3001
58
+ step 560000 | 5605s | train_loss=0.2849
59
+ step 570000 | 5698s | train_loss=0.2810
60
+ step 580000 | 5788s | train_loss=0.2762
61
+ step 590000 | 5881s | train_loss=0.2715
62
+ step 600000 | 6048s | train_loss=0.2545 | fid=87.9286
63
+ step 610000 | 6138s | train_loss=0.2470
64
+ step 620000 | 6231s | train_loss=0.2458
65
+ step 630000 | 6323s | train_loss=0.2429
66
+ step 640000 | 6416s | train_loss=0.2370
67
+ step 650000 | 6509s | train_loss=0.2286
68
+ step 660000 | 6602s | train_loss=0.2254
69
+ step 670000 | 6692s | train_loss=0.2245
70
+ step 680000 | 6785s | train_loss=0.2170
71
+ step 690000 | 6878s | train_loss=0.2137
72
+ step 700000 | 7044s | train_loss=0.2096 | fid=86.7016
73
+ step 710000 | 7135s | train_loss=0.2052
74
+ step 720000 | 7227s | train_loss=0.2009
75
+ step 730000 | 7321s | train_loss=0.2007
76
+ step 740000 | 7412s | train_loss=0.1939
77
+ step 750000 | 7505s | train_loss=0.1943
78
+ step 760000 | 7598s | train_loss=0.1935
79
+ step 770000 | 7688s | train_loss=0.1922
80
+ step 780000 | 7781s | train_loss=0.1874
81
+ step 790000 | 7874s | train_loss=0.1856
82
+ step 800000 | 8041s | train_loss=0.1849 | fid=87.1326
83
+ step 810000 | 8132s | train_loss=0.1789
84
+ step 820000 | 8225s | train_loss=0.1800
85
+ step 830000 | 8318s | train_loss=0.1759
86
+ step 840000 | 8410s | train_loss=0.1743
87
+ step 850000 | 8503s | train_loss=0.1745
88
+ step 860000 | 8592s | train_loss=0.1727
89
+ step 870000 | 8685s | train_loss=0.1696
90
+ step 880000 | 8778s | train_loss=0.1680
91
+ step 890000 | 8871s | train_loss=0.1651
92
+ step 900000 | 9038s | train_loss=0.1646 | fid=88.1491
93
+ step 910000 | 9129s | train_loss=0.1635
94
+ step 920000 | 9220s | train_loss=0.1600
95
+ step 930000 | 9314s | train_loss=0.1624
96
+ step 940000 | 9407s | train_loss=0.1600
97
+ step 950000 | 9498s | train_loss=0.1584
98
+ step 960000 | 9588s | train_loss=0.1574
99
+ step 970000 | 9681s | train_loss=0.1554
100
+ step 980000 | 9775s | train_loss=0.1521
101
+ step 990000 | 9868s | train_loss=0.1539
102
+ step 1000000 | 10034s | train_loss=0.1541 | fid=87.8470
103
+ step 1010000 | 10127s | train_loss=0.1537
104
+ step 1020000 | 10218s | train_loss=0.1497
105
+ step 1030000 | 10311s | train_loss=0.1470
106
+ step 1040000 | 10403s | train_loss=0.1481
107
+ step 1050000 | 10491s | train_loss=0.1453
108
+ step 1060000 | 10584s | train_loss=0.1451
109
+ step 1070000 | 10677s | train_loss=0.1454
110
+ step 1080000 | 10770s | train_loss=0.1420
111
+ step 1090000 | 10863s | train_loss=0.1451
112
+ step 1100000 | 11029s | train_loss=0.1454 | fid=87.9224
113
+ step 1110000 | 11121s | train_loss=0.1376
114
+ step 1120000 | 11213s | train_loss=0.1411
115
+ step 1130000 | 11305s | train_loss=0.1388
116
+ step 1140000 | 11398s | train_loss=0.1392
117
+ step 1150000 | 11489s | train_loss=0.1371
118
+ step 1160000 | 11580s | train_loss=0.1369
119
+ step 1170000 | 11673s | train_loss=0.1385
120
+ step 1180000 | 11766s | train_loss=0.1358
121
+ step 1190000 | 11859s | train_loss=0.1327
122
+ step 1200000 | 12026s | train_loss=0.1318 | fid=87.6148
123
+ step 1210000 | 12118s | train_loss=0.1312
124
+ step 1220000 | 12209s | train_loss=0.1325
125
+ step 1230000 | 12302s | train_loss=0.1300
126
+ step 1240000 | 12392s | train_loss=0.1265
127
+ step 1250000 | 12485s | train_loss=0.1291
128
+ step 1260000 | 12576s | train_loss=0.1288
129
+ step 1270000 | 12669s | train_loss=0.1298
130
+ step 1280000 | 12762s | train_loss=0.1283
131
+ step 1290000 | 12855s | train_loss=0.1256
132
+ step 1300000 | 13022s | train_loss=0.1239 | fid=87.8883
133
+ step 1310000 | 13114s | train_loss=0.1254
134
+ step 1320000 | 13206s | train_loss=0.1241
135
+ step 1330000 | 13298s | train_loss=0.1232
136
+ step 1340000 | 13388s | train_loss=0.1229
137
+ step 1350000 | 13481s | train_loss=0.1218
138
+ step 1360000 | 13575s | train_loss=0.1207
139
+ step 1370000 | 13666s | train_loss=0.1189
140
+ step 1380000 | 13759s | train_loss=0.1214
141
+ step 1390000 | 13851s | train_loss=0.1198
142
+ step 1400000 | 14018s | train_loss=0.1199 | fid=88.2587
143
+ step 1410000 | 14110s | train_loss=0.1201
144
+ step 1420000 | 14201s | train_loss=0.1184
145
+ step 1430000 | 14290s | train_loss=0.1182
146
+ step 1440000 | 14382s | train_loss=0.1163
147
+ step 1450000 | 14475s | train_loss=0.1148
148
+ step 1460000 | 14568s | train_loss=0.1170
149
+ step 1470000 | 14660s | train_loss=0.1148
150
+ step 1480000 | 14753s | train_loss=0.1160
151
+ step 1490000 | 14846s | train_loss=0.1146
152
+ step 1500000 | 15012s | train_loss=0.1132 | fid=86.5842
153
+ step 1510000 | 15104s | train_loss=0.1121
154
+ step 1520000 | 15196s | train_loss=0.1151
155
+ step 1530000 | 15283s | train_loss=0.1142
156
+ step 1540000 | 15376s | train_loss=0.1130
157
+ step 1550000 | 15469s | train_loss=0.1124
158
+ step 1560000 | 15562s | train_loss=0.1115
159
+ step 1570000 | 15654s | train_loss=0.1095
160
+ step 1580000 | 15747s | train_loss=0.1101
161
+ step 1590000 | 15840s | train_loss=0.1114
162
+ step 1600000 | 16007s | train_loss=0.1100 | fid=86.8865
163
+ step 1610000 | 16098s | train_loss=0.1082
164
+ step 1620000 | 16182s | train_loss=0.1074
165
+ step 1630000 | 16275s | train_loss=0.1079
166
+ step 1640000 | 16367s | train_loss=0.1066
167
+ step 1650000 | 16460s | train_loss=0.1088
168
+ step 1660000 | 16554s | train_loss=0.1081
169
+ step 1670000 | 16647s | train_loss=0.1084
170
+ step 1680000 | 16739s | train_loss=0.1051
171
+ step 1690000 | 16832s | train_loss=0.1059
172
+ step 1700000 | 16998s | train_loss=0.1055 | fid=87.4939
173
+ step 1710000 | 17089s | train_loss=0.1052
174
+ step 1720000 | 17172s | train_loss=0.1048
175
+ step 1730000 | 17265s | train_loss=0.1022
176
+ step 1740000 | 17358s | train_loss=0.1023
177
+ step 1750000 | 17450s | train_loss=0.1024
178
+ step 1760000 | 17543s | train_loss=0.1012
179
+ step 1770000 | 17636s | train_loss=0.1001
180
+ step 1780000 | 17728s | train_loss=0.1010
181
+ step 1790000 | 17821s | train_loss=0.1008
182
+ step 1800000 | 17988s | train_loss=0.1010 | fid=87.0795
183
+ step 1810000 | 18080s | train_loss=0.1011
184
+ step 1820000 | 18164s | train_loss=0.1001
185
+ step 1830000 | 18255s | train_loss=0.0992
186
+ step 1840000 | 18348s | train_loss=0.1002
187
+ step 1850000 | 18440s | train_loss=0.0992
188
+ step 1860000 | 18534s | train_loss=0.0988
189
+ step 1870000 | 18627s | train_loss=0.0979
190
+ step 1880000 | 18720s | train_loss=0.0969
191
+ step 1890000 | 18811s | train_loss=0.0972
192
+ step 1900000 | 18978s | train_loss=0.0975 | fid=87.7265
193
+ step 1910000 | 19069s | train_loss=0.0960
194
+ step 1920000 | 19152s | train_loss=0.0955
195
+ step 1930000 | 19243s | train_loss=0.0949
196
+ step 1940000 | 19336s | train_loss=0.0968
197
+ step 1950000 | 19429s | train_loss=0.0969
198
+ step 1960000 | 19522s | train_loss=0.0957
199
+ step 1970000 | 19615s | train_loss=0.0953
200
+ step 1980000 | 19708s | train_loss=0.0956
201
+ step 1990000 | 19800s | train_loss=0.0948
202
+ step 2000000 | 19965s | train_loss=0.0933 | fid=88.1092
203
+ step 2010000 | 20054s | train_loss=0.0933
204
+ step 2020000 | 20140s | train_loss=0.0932
205
+ step 2030000 | 20231s | train_loss=0.0936
206
+ step 2040000 | 20324s | train_loss=0.0929
207
+ step 2050000 | 20417s | train_loss=0.0925
208
+ step 2060000 | 20510s | train_loss=0.0909
209
+ step 2070000 | 20603s | train_loss=0.0913
210
+ step 2080000 | 20696s | train_loss=0.0919
211
+ step 2090000 | 20789s | train_loss=0.0929
212
+ step 2100000 | 20953s | train_loss=0.0931 | fid=84.6373
213
+ step 2110000 | 21040s | train_loss=0.0905
214
+ step 2120000 | 21131s | train_loss=0.0898
215
+ step 2130000 | 21223s | train_loss=0.0900
216
+ step 2140000 | 21316s | train_loss=0.0891
217
+ step 2150000 | 21412s | train_loss=0.0885
218
+ step 2160000 | 21507s | train_loss=0.0896
219
+ step 2170000 | 21601s | train_loss=0.0901
220
+ step 2180000 | 21696s | train_loss=0.0880
221
+ step 2190000 | 21790s | train_loss=0.0895
222
+ step 2200000 | 21957s | train_loss=0.0885 | fid=86.6952
223
+ step 2210000 | 22037s | train_loss=0.0878
224
+ step 2220000 | 22128s | train_loss=0.0881
225
+ step 2230000 | 22222s | train_loss=0.0881
226
+ step 2240000 | 22317s | train_loss=0.0862
227
+ step 2250000 | 22413s | train_loss=0.0869
228
+ step 2260000 | 22508s | train_loss=0.0875
229
+ step 2270000 | 22602s | train_loss=0.0875
230
+ step 2280000 | 22696s | train_loss=0.0862
231
+ step 2290000 | 22791s | train_loss=0.0874
232
+ step 2300000 | 22959s | train_loss=0.0860 | fid=86.4463
233
+ step 2310000 | 23041s | train_loss=0.0843
234
+ step 2320000 | 23131s | train_loss=0.0843
235
+ step 2330000 | 23224s | train_loss=0.0851
236
+ step 2340000 | 23320s | train_loss=0.0836
237
+ step 2350000 | 23415s | train_loss=0.0841
238
+ step 2360000 | 23510s | train_loss=0.0832
239
+ step 2370000 | 23603s | train_loss=0.0843
240
+ step 2380000 | 23696s | train_loss=0.0851
241
+ step 2390000 | 23789s | train_loss=0.0847
242
+ step 2400000 | 23954s | train_loss=0.0846 | fid=87.9533
243
+ step 2410000 | 24040s | train_loss=0.0845
244
+ step 2420000 | 24131s | train_loss=0.0836
245
+ step 2430000 | 24223s | train_loss=0.0826
246
+ step 2440000 | 24316s | train_loss=0.0824
247
+ step 2450000 | 24409s | train_loss=0.0832
248
+ step 2460000 | 24502s | train_loss=0.0830
249
+ step 2470000 | 24594s | train_loss=0.0833
250
+ step 2480000 | 24687s | train_loss=0.0813
251
+ step 2490000 | 24779s | train_loss=0.0831
252
+ step 2500000 | 24946s | train_loss=0.0817 | fid=88.6282
253
+ step 2510000 | 25025s | train_loss=0.0806
254
+ step 2520000 | 25116s | train_loss=0.0802
255
+ step 2530000 | 25208s | train_loss=0.0808
256
+ step 2540000 | 25301s | train_loss=0.0808
257
+ step 2550000 | 25394s | train_loss=0.0798
258
+ step 2560000 | 25487s | train_loss=0.0803
259
+ step 2570000 | 25580s | train_loss=0.0801
260
+ step 2580000 | 25673s | train_loss=0.0799
261
+ step 2590000 | 25766s | train_loss=0.0794
262
+ step 2600000 | 25932s | train_loss=0.0805 | fid=86.1920
263
+ step 2610000 | 26013s | train_loss=0.0803
264
+ step 2620000 | 26104s | train_loss=0.0792
265
+ step 2630000 | 26195s | train_loss=0.0790
266
+ step 2640000 | 26288s | train_loss=0.0802
267
+ step 2650000 | 26381s | train_loss=0.0797
268
+ step 2660000 | 26474s | train_loss=0.0788
269
+ step 2670000 | 26567s | train_loss=0.0765
270
+ step 2680000 | 26659s | train_loss=0.0771
271
+ step 2690000 | 26752s | train_loss=0.0761
272
+ step 2700000 | 26917s | train_loss=0.0772 | fid=86.2549
273
+ step 2710000 | 27003s | train_loss=0.0767
274
+ step 2720000 | 27093s | train_loss=0.0777
275
+ step 2730000 | 27184s | train_loss=0.0786
276
+ step 2740000 | 27277s | train_loss=0.0767
277
+ step 2750000 | 27370s | train_loss=0.0767
278
+ step 2760000 | 27463s | train_loss=0.0770
279
+ step 2770000 | 27556s | train_loss=0.0776
280
+ step 2780000 | 27649s | train_loss=0.0762
281
+ step 2790000 | 27741s | train_loss=0.0757
282
+ step 2800000 | 27905s | train_loss=0.0755 | fid=87.1931
283
+ step 2810000 | 27993s | train_loss=0.0766
284
+ step 2820000 | 28082s | train_loss=0.0752
285
+ step 2830000 | 28174s | train_loss=0.0758
286
+ step 2840000 | 28267s | train_loss=0.0746
287
+ step 2850000 | 28359s | train_loss=0.0749
288
+ step 2860000 | 28452s | train_loss=0.0763
289
+ step 2870000 | 28545s | train_loss=0.0747
290
+ step 2880000 | 28638s | train_loss=0.0738
291
+ step 2890000 | 28730s | train_loss=0.0743
292
+ step 2900000 | 28892s | train_loss=0.0739 | fid=87.3832
293
+ step 2910000 | 28983s | train_loss=0.0735
294
+ step 2920000 | 29073s | train_loss=0.0734
295
+ step 2930000 | 29164s | train_loss=0.0740
296
+ step 2940000 | 29257s | train_loss=0.0741
297
+ step 2950000 | 29350s | train_loss=0.0745
298
+ step 2960000 | 29443s | train_loss=0.0738
299
+ step 2970000 | 29536s | train_loss=0.0741
300
+ step 2980000 | 29629s | train_loss=0.0711
301
+ step 2990000 | 29723s | train_loss=0.0733
302
+ step 3000000 | 29887s | train_loss=0.0731 | fid=88.3028
303
+ step 3010000 | 29982s | train_loss=0.0737
304
+ step 3020000 | 30076s | train_loss=0.0731
305
+ step 3030000 | 30167s | train_loss=0.0724
306
+ step 3040000 | 30263s | train_loss=0.0733
307
+ step 3050000 | 30357s | train_loss=0.0723
308
+ step 3060000 | 30452s | train_loss=0.0721
309
+ step 3070000 | 30547s | train_loss=0.0717
310
+ step 3080000 | 30642s | train_loss=0.0727
311
+ step 3090000 | 30737s | train_loss=0.0714
312
+ step 3100000 | 30893s | train_loss=0.0723 | fid=87.4747
313
+ step 3110000 | 30989s | train_loss=0.0713
314
+ step 3120000 | 31083s | train_loss=0.0707
315
+ step 3130000 | 31175s | train_loss=0.0713
316
+ step 3140000 | 31269s | train_loss=0.0714
317
+ step 3150000 | 31365s | train_loss=0.0731
318
+ step 3160000 | 31462s | train_loss=0.0707
319
+ step 3170000 | 31557s | train_loss=0.0715
320
+ step 3180000 | 31652s | train_loss=0.0712
321
+ step 3190000 | 31748s | train_loss=0.0709
322
+ step 3200000 | 31908s | train_loss=0.0710 | fid=87.5707
323
+ step 3210000 | 32003s | train_loss=0.0705
324
+ step 3220000 | 32097s | train_loss=0.0705
325
+ step 3230000 | 32190s | train_loss=0.0714
326
+ step 3240000 | 32282s | train_loss=0.0698
327
+ step 3250000 | 32377s | train_loss=0.0687
328
+ step 3260000 | 32473s | train_loss=0.0697
329
+ step 3270000 | 32568s | train_loss=0.0699
330
+ step 3280000 | 32664s | train_loss=0.0697
331
+ step 3290000 | 32760s | train_loss=0.0689
332
+ step 3300000 | 32913s | train_loss=0.0686 | fid=87.7791
333
+ step 3310000 | 33010s | train_loss=0.0679
334
+ step 3320000 | 33093s | train_loss=0.0701
335
+ step 3330000 | 33186s | train_loss=0.0693
336
+ step 3340000 | 33280s | train_loss=0.0700
337
+ step 3350000 | 33376s | train_loss=0.0668
338
+ step 3360000 | 33470s | train_loss=0.0682
339
+ step 3370000 | 33565s | train_loss=0.0673
340
+ step 3380000 | 33661s | train_loss=0.0675
341
+ step 3390000 | 33756s | train_loss=0.0677
342
+ step 3400000 | 33908s | train_loss=0.0675 | fid=87.6004
343
+ step 3410000 | 34003s | train_loss=0.0676
344
+ step 3420000 | 34096s | train_loss=0.0665
345
+ step 3430000 | 34182s | train_loss=0.0676
346
+ step 3440000 | 34275s | train_loss=0.0673
347
+ step 3450000 | 34372s | train_loss=0.0674
348
+ step 3460000 | 34466s | train_loss=0.0662
349
+ step 3470000 | 34561s | train_loss=0.0666
350
+ step 3480000 | 34657s | train_loss=0.0673
351
+ step 3490000 | 34753s | train_loss=0.0656
352
+ step 3500000 | 34903s | train_loss=0.0666 | fid=87.3525
353
+ step 3510000 | 34997s | train_loss=0.0676
354
+ step 3520000 | 35091s | train_loss=0.0654
355
+ step 3530000 | 35175s | train_loss=0.0660
356
+ step 3540000 | 35270s | train_loss=0.0670
357
+ step 3550000 | 35364s | train_loss=0.0661
358
+ step 3560000 | 35457s | train_loss=0.0668
359
+ step 3570000 | 35553s | train_loss=0.0669
360
+ step 3580000 | 35648s | train_loss=0.0680
361
+ step 3590000 | 35744s | train_loss=0.0662
362
+ step 3600000 | 35895s | train_loss=0.0651 | fid=87.5705
363
+ step 3610000 | 35990s | train_loss=0.0661
364
+ step 3620000 | 36084s | train_loss=0.0662
365
+ step 3630000 | 36169s | train_loss=0.0655
366
+ step 3640000 | 36256s | train_loss=0.0657
367
+ step 3650000 | 36350s | train_loss=0.0659
368
+ step 3660000 | 36444s | train_loss=0.0664
369
+ step 3670000 | 36540s | train_loss=0.0652
370
+ step 3680000 | 36637s | train_loss=0.0654
371
+ step 3690000 | 36732s | train_loss=0.0648
372
+ step 3700000 | 36888s | train_loss=0.0651 | fid=87.5314
373
+ step 3710000 | 36984s | train_loss=0.0647
374
+ step 3720000 | 37077s | train_loss=0.0650
375
+ step 3730000 | 37172s | train_loss=0.0651
376
+ step 3740000 | 37255s | train_loss=0.0648
377
+ step 3750000 | 37351s | train_loss=0.0641
378
+ step 3760000 | 37444s | train_loss=0.0644
379
+ step 3770000 | 37540s | train_loss=0.0644
380
+ step 3780000 | 37636s | train_loss=0.0636
381
+ step 3790000 | 37731s | train_loss=0.0630
382
+ step 3800000 | 37881s | train_loss=0.0647 | fid=87.3148
383
+ step 3810000 | 37976s | train_loss=0.0632
384
+ step 3820000 | 38069s | train_loss=0.0641
385
+ step 3830000 | 38163s | train_loss=0.0634
386
+ step 3840000 | 38253s | train_loss=0.0625
387
+ step 3850000 | 38345s | train_loss=0.0639
388
+ step 3860000 | 38439s | train_loss=0.0632
389
+ step 3870000 | 38534s | train_loss=0.0621
390
+ step 3880000 | 38630s | train_loss=0.0649
391
+ step 3890000 | 38726s | train_loss=0.0638
392
+ step 3900000 | 38883s | train_loss=0.0626 | fid=86.9503
393
+ step 3910000 | 38977s | train_loss=0.0630
394
+ step 3920000 | 39071s | train_loss=0.0614
395
+ step 3930000 | 39164s | train_loss=0.0623
396
+ step 3940000 | 39259s | train_loss=0.0635
397
+ step 3950000 | 39352s | train_loss=0.0638
398
+ step 3960000 | 39446s | train_loss=0.0625
399
+ step 3970000 | 39542s | train_loss=0.0623
400
+ step 3980000 | 39638s | train_loss=0.0621
401
+ step 3990000 | 39733s | train_loss=0.0634
402
+ step 4000000 | 39883s | train_loss=0.0636 | fid=87.0629
403
+ step 4010000 | 39978s | train_loss=0.0627
404
+ step 4020000 | 40074s | train_loss=0.0627
405
+ step 4030000 | 40167s | train_loss=0.0616
406
+ step 4040000 | 40260s | train_loss=0.0612
407
+ step 4050000 | 40355s | train_loss=0.0620
408
+ step 4060000 | 40451s | train_loss=0.0617
409
+ step 4070000 | 40546s | train_loss=0.0620
410
+ step 4080000 | 40642s | train_loss=0.0627
411
+ step 4090000 | 40737s | train_loss=0.0603
412
+ step 4100000 | 40887s | train_loss=0.0606 | fid=87.7867
413
+ step 4110000 | 40982s | train_loss=0.0625
414
+ step 4120000 | 41077s | train_loss=0.0610
415
+ step 4130000 | 41170s | train_loss=0.0620
416
+ step 4140000 | 41262s | train_loss=0.0609
417
+ step 4150000 | 41359s | train_loss=0.0617
418
+ step 4160000 | 41456s | train_loss=0.0611
419
+ step 4170000 | 41551s | train_loss=0.0617
420
+ step 4180000 | 41648s | train_loss=0.0626
421
+ step 4190000 | 41746s | train_loss=0.0614
422
+ step 4200000 | 41899s | train_loss=0.0602 | fid=88.2921
423
+ step 4210000 | 41997s | train_loss=0.0618
424
+ step 4220000 | 42094s | train_loss=0.0599
425
+ step 4230000 | 42186s | train_loss=0.0611
426
+ step 4240000 | 42279s | train_loss=0.0619
427
+ step 4250000 | 42376s | train_loss=0.0605
428
+ step 4260000 | 42473s | train_loss=0.0602
429
+ step 4270000 | 42568s | train_loss=0.0599
430
+ step 4280000 | 42665s | train_loss=0.0612
431
+ step 4290000 | 42762s | train_loss=0.0611
432
+ step 4300000 | 42917s | train_loss=0.0589 | fid=86.9896
433
+ step 4310000 | 43014s | train_loss=0.0617
434
+ step 4320000 | 43111s | train_loss=0.0600
435
+ step 4330000 | 43203s | train_loss=0.0593
436
+ step 4340000 | 43296s | train_loss=0.0593
437
+ step 4350000 | 43394s | train_loss=0.0585
438
+ step 4360000 | 43491s | train_loss=0.0606
439
+ step 4370000 | 43588s | train_loss=0.0586
440
+ step 4380000 | 43683s | train_loss=0.0599
441
+ step 4390000 | 43778s | train_loss=0.0587
442
+ step 4400000 | 43934s | train_loss=0.0595 | fid=87.7202
443
+ step 4410000 | 44031s | train_loss=0.0601
444
+ step 4420000 | 44127s | train_loss=0.0592
445
+ step 4430000 | 44221s | train_loss=0.0592
446
+ step 4440000 | 44315s | train_loss=0.0595
447
+ step 4450000 | 44413s | train_loss=0.0590
448
+ step 4460000 | 44509s | train_loss=0.0588
449
+ step 4470000 | 44606s | train_loss=0.0597
450
+ step 4480000 | 44700s | train_loss=0.0584
451
+ step 4490000 | 44791s | train_loss=0.0587
452
+ step 4500000 | 44947s | train_loss=0.0583 | fid=87.7308
453
+ step 4510000 | 45044s | train_loss=0.0574
454
+ step 4520000 | 45138s | train_loss=0.0590
455
+ step 4530000 | 45234s | train_loss=0.0581
456
+ step 4540000 | 45328s | train_loss=0.0588
457
+ step 4550000 | 45424s | train_loss=0.0578
458
+ step 4560000 | 45521s | train_loss=0.0588
459
+ step 4570000 | 45619s | train_loss=0.0580
460
+ step 4580000 | 45713s | train_loss=0.0588
461
+ step 4590000 | 45805s | train_loss=0.0575
462
+ step 4600000 | 45967s | train_loss=0.0587 | fid=85.6555
463
+ step 4610000 | 46063s | train_loss=0.0591
464
+ step 4620000 | 46159s | train_loss=0.0582
465
+ step 4630000 | 46254s | train_loss=0.0584
466
+ step 4640000 | 46348s | train_loss=0.0582
467
+ step 4650000 | 46444s | train_loss=0.0587
468
+ step 4660000 | 46541s | train_loss=0.0578
469
+ step 4670000 | 46637s | train_loss=0.0572
470
+ step 4680000 | 46733s | train_loss=0.0571
471
+ step 4690000 | 46818s | train_loss=0.0575
472
+ step 4700000 | 46983s | train_loss=0.0576 | fid=86.9889
473
+ step 4710000 | 47078s | train_loss=0.0581
474
+ step 4720000 | 47175s | train_loss=0.0579
475
+ step 4730000 | 47271s | train_loss=0.0577
476
+ step 4740000 | 47367s | train_loss=0.0574
477
+ step 4750000 | 47462s | train_loss=0.0570
478
+ step 4760000 | 47558s | train_loss=0.0582
479
+ step 4770000 | 47654s | train_loss=0.0570
480
+ step 4780000 | 47752s | train_loss=0.0576
481
+ step 4790000 | 47838s | train_loss=0.0561
482
+ step 4800000 | 48009s | train_loss=0.0570 | fid=88.2981
483
+ step 4810000 | 48107s | train_loss=0.0565
484
+ step 4820000 | 48202s | train_loss=0.0570
485
+ step 4830000 | 48299s | train_loss=0.0558
486
+ step 4840000 | 48394s | train_loss=0.0562
487
+ step 4850000 | 48490s | train_loss=0.0571
488
+ step 4860000 | 48587s | train_loss=0.0573
489
+ step 4870000 | 48684s | train_loss=0.0567
490
+ step 4880000 | 48774s | train_loss=0.0566
491
+ step 4890000 | 48851s | train_loss=0.0567
492
+ step 4900000 | 48992s | train_loss=0.0568 | fid=87.3401
493
+ step 4910000 | 49072s | train_loss=0.0565
494
+ step 4920000 | 49148s | train_loss=0.0567
495
+ step 4930000 | 49227s | train_loss=0.0567
496
+ step 4940000 | 49303s | train_loss=0.0576
497
+ step 4950000 | 49382s | train_loss=0.0566
498
+ step 4960000 | 49462s | train_loss=0.0565
499
+ step 4970000 | 49542s | train_loss=0.0548
500
+ step 4980000 | 49623s | train_loss=0.0557
501
+ step 4990000 | 49699s | train_loss=0.0565
502
+ step 5000000 | 49843s | train_loss=0.0553 | fid=86.7828
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d100/config.json ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "vae_checkpoint": "vae_checkpoints/celeba_resnet_modernloss_d100/vae.pt",
3
+ "out": "results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d100",
4
+ "subset_path": "diagnostics/celeba_diverse_1k_mean_plus_sd/subset_indices.json",
5
+ "train_tar": "/overflow/data/cerberus/celeba_hq_256_partitioned_ws/train/celeba_hq_256-train-0000.tar",
6
+ "n_train": 1000,
7
+ "total_steps": 5000000,
8
+ "eval_interval": 10000,
9
+ "mem_interval": 100000,
10
+ "fid_interval": 100000,
11
+ "fid_n_real": 1000,
12
+ "fid_n_gen": 10000,
13
+ "n_gen_samples": 10000,
14
+ "n_sde_steps": 500,
15
+ "batch_size": 512,
16
+ "hidden": 1024,
17
+ "depth": 5,
18
+ "optimizer": "sgd",
19
+ "lr": 0.001,
20
+ "momentum": 0.8,
21
+ "t_min": 0.01,
22
+ "t_max": 3.0,
23
+ "t_eval": 0.1,
24
+ "seed": 45,
25
+ "device": null
26
+ }
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d100/metrics.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d100/runner.log ADDED
@@ -0,0 +1,502 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ FID real-side stats cached over 1000 diverse train images
2
+ step 1 | 71s | train_loss=1.0036 | fid=291.3698
3
+ step 10000 | 166s | train_loss=0.8106
4
+ step 20000 | 262s | train_loss=0.6865
5
+ step 30000 | 358s | train_loss=0.5505
6
+ step 40000 | 454s | train_loss=0.4129
7
+ step 50000 | 551s | train_loss=0.2994
8
+ step 60000 | 647s | train_loss=0.2273
9
+ step 70000 | 743s | train_loss=0.1856
10
+ step 80000 | 839s | train_loss=0.1585
11
+ step 90000 | 936s | train_loss=0.1448
12
+ step 100000 | 1096s | train_loss=0.1290 | fid=60.3562
13
+ step 110000 | 1192s | train_loss=0.1215
14
+ step 120000 | 1288s | train_loss=0.1153
15
+ step 130000 | 1384s | train_loss=0.1092
16
+ step 140000 | 1481s | train_loss=0.1040
17
+ step 150000 | 1579s | train_loss=0.1017
18
+ step 160000 | 1676s | train_loss=0.0968
19
+ step 170000 | 1773s | train_loss=0.0924
20
+ step 180000 | 1870s | train_loss=0.0903
21
+ step 190000 | 1963s | train_loss=0.0883
22
+ step 200000 | 2130s | train_loss=0.0856 | fid=59.1669
23
+ step 210000 | 2227s | train_loss=0.0845
24
+ step 220000 | 2324s | train_loss=0.0813
25
+ step 230000 | 2421s | train_loss=0.0781
26
+ step 240000 | 2518s | train_loss=0.0780
27
+ step 250000 | 2615s | train_loss=0.0765
28
+ step 260000 | 2713s | train_loss=0.0749
29
+ step 270000 | 2810s | train_loss=0.0719
30
+ step 280000 | 2903s | train_loss=0.0718
31
+ step 290000 | 3000s | train_loss=0.0712
32
+ step 300000 | 3169s | train_loss=0.0679 | fid=57.1085
33
+ step 310000 | 3267s | train_loss=0.0668
34
+ step 320000 | 3364s | train_loss=0.0659
35
+ step 330000 | 3461s | train_loss=0.0649
36
+ step 340000 | 3559s | train_loss=0.0647
37
+ step 350000 | 3656s | train_loss=0.0631
38
+ step 360000 | 3754s | train_loss=0.0623
39
+ step 370000 | 3846s | train_loss=0.0601
40
+ step 380000 | 3944s | train_loss=0.0613
41
+ step 390000 | 4041s | train_loss=0.0588
42
+ step 400000 | 4210s | train_loss=0.0597 | fid=53.3376
43
+ step 410000 | 4307s | train_loss=0.0572
44
+ step 420000 | 4404s | train_loss=0.0566
45
+ step 430000 | 4502s | train_loss=0.0566
46
+ step 440000 | 4599s | train_loss=0.0574
47
+ step 450000 | 4696s | train_loss=0.0550
48
+ step 460000 | 4789s | train_loss=0.0541
49
+ step 470000 | 4884s | train_loss=0.0542
50
+ step 480000 | 4982s | train_loss=0.0533
51
+ step 490000 | 5078s | train_loss=0.0525
52
+ step 500000 | 5246s | train_loss=0.0515 | fid=53.3169
53
+ step 510000 | 5343s | train_loss=0.0512
54
+ step 520000 | 5440s | train_loss=0.0513
55
+ step 530000 | 5537s | train_loss=0.0512
56
+ step 540000 | 5635s | train_loss=0.0504
57
+ step 550000 | 5729s | train_loss=0.0508
58
+ step 560000 | 5824s | train_loss=0.0475
59
+ step 570000 | 5921s | train_loss=0.0503
60
+ step 580000 | 6019s | train_loss=0.0488
61
+ step 590000 | 6114s | train_loss=0.0481
62
+ step 600000 | 6283s | train_loss=0.0478 | fid=51.1817
63
+ step 610000 | 6380s | train_loss=0.0481
64
+ step 620000 | 6477s | train_loss=0.0495
65
+ step 630000 | 6575s | train_loss=0.0471
66
+ step 640000 | 6669s | train_loss=0.0455
67
+ step 650000 | 6762s | train_loss=0.0456
68
+ step 660000 | 6859s | train_loss=0.0444
69
+ step 670000 | 6957s | train_loss=0.0463
70
+ step 680000 | 7053s | train_loss=0.0448
71
+ step 690000 | 7150s | train_loss=0.0442
72
+ step 700000 | 7318s | train_loss=0.0442 | fid=50.9893
73
+ step 710000 | 7416s | train_loss=0.0444
74
+ step 720000 | 7513s | train_loss=0.0432
75
+ step 730000 | 7607s | train_loss=0.0431
76
+ step 740000 | 7701s | train_loss=0.0434
77
+ step 750000 | 7798s | train_loss=0.0432
78
+ step 760000 | 7895s | train_loss=0.0439
79
+ step 770000 | 7993s | train_loss=0.0428
80
+ step 780000 | 8089s | train_loss=0.0430
81
+ step 790000 | 8186s | train_loss=0.0428
82
+ step 800000 | 8354s | train_loss=0.0427 | fid=50.1734
83
+ step 810000 | 8451s | train_loss=0.0412
84
+ step 820000 | 8545s | train_loss=0.0420
85
+ step 830000 | 8639s | train_loss=0.0414
86
+ step 840000 | 8736s | train_loss=0.0397
87
+ step 850000 | 8833s | train_loss=0.0404
88
+ step 860000 | 8931s | train_loss=0.0403
89
+ step 870000 | 9026s | train_loss=0.0405
90
+ step 880000 | 9123s | train_loss=0.0400
91
+ step 890000 | 9218s | train_loss=0.0382
92
+ step 900000 | 9389s | train_loss=0.0410 | fid=50.8127
93
+ step 910000 | 9484s | train_loss=0.0405
94
+ step 920000 | 9578s | train_loss=0.0397
95
+ step 930000 | 9675s | train_loss=0.0385
96
+ step 940000 | 9772s | train_loss=0.0386
97
+ step 950000 | 9870s | train_loss=0.0399
98
+ step 960000 | 9967s | train_loss=0.0384
99
+ step 970000 | 10062s | train_loss=0.0371
100
+ step 980000 | 10160s | train_loss=0.0379
101
+ step 990000 | 10254s | train_loss=0.0369
102
+ step 1000000 | 10423s | train_loss=0.0373 | fid=50.6578
103
+ step 1010000 | 10516s | train_loss=0.0369
104
+ step 1020000 | 10614s | train_loss=0.0362
105
+ step 1030000 | 10711s | train_loss=0.0367
106
+ step 1040000 | 10807s | train_loss=0.0372
107
+ step 1050000 | 10904s | train_loss=0.0378
108
+ step 1060000 | 11001s | train_loss=0.0379
109
+ step 1070000 | 11097s | train_loss=0.0365
110
+ step 1080000 | 11195s | train_loss=0.0362
111
+ step 1090000 | 11289s | train_loss=0.0369
112
+ step 1100000 | 11454s | train_loss=0.0356 | fid=51.7471
113
+ step 1110000 | 11551s | train_loss=0.0358
114
+ step 1120000 | 11648s | train_loss=0.0351
115
+ step 1130000 | 11746s | train_loss=0.0359
116
+ step 1140000 | 11843s | train_loss=0.0346
117
+ step 1150000 | 11940s | train_loss=0.0344
118
+ step 1160000 | 12036s | train_loss=0.0341
119
+ step 1170000 | 12133s | train_loss=0.0340
120
+ step 1180000 | 12231s | train_loss=0.0349
121
+ step 1190000 | 12322s | train_loss=0.0338
122
+ step 1200000 | 12490s | train_loss=0.0343 | fid=53.1193
123
+ step 1210000 | 12587s | train_loss=0.0336
124
+ step 1220000 | 12684s | train_loss=0.0337
125
+ step 1230000 | 12782s | train_loss=0.0341
126
+ step 1240000 | 12879s | train_loss=0.0330
127
+ step 1250000 | 12977s | train_loss=0.0337
128
+ step 1260000 | 13073s | train_loss=0.0333
129
+ step 1270000 | 13170s | train_loss=0.0331
130
+ step 1280000 | 13263s | train_loss=0.0328
131
+ step 1290000 | 13355s | train_loss=0.0332
132
+ step 1300000 | 13526s | train_loss=0.0337 | fid=53.9262
133
+ step 1310000 | 13624s | train_loss=0.0327
134
+ step 1320000 | 13720s | train_loss=0.0323
135
+ step 1330000 | 13817s | train_loss=0.0313
136
+ step 1340000 | 13913s | train_loss=0.0320
137
+ step 1350000 | 14009s | train_loss=0.0316
138
+ step 1360000 | 14106s | train_loss=0.0323
139
+ step 1370000 | 14199s | train_loss=0.0328
140
+ step 1380000 | 14289s | train_loss=0.0311
141
+ step 1390000 | 14386s | train_loss=0.0318
142
+ step 1400000 | 14556s | train_loss=0.0323 | fid=53.9213
143
+ step 1410000 | 14653s | train_loss=0.0312
144
+ step 1420000 | 14750s | train_loss=0.0317
145
+ step 1430000 | 14847s | train_loss=0.0306
146
+ step 1440000 | 14944s | train_loss=0.0303
147
+ step 1450000 | 15039s | train_loss=0.0303
148
+ step 1460000 | 15133s | train_loss=0.0308
149
+ step 1470000 | 15223s | train_loss=0.0297
150
+ step 1480000 | 15313s | train_loss=0.0315
151
+ step 1490000 | 15411s | train_loss=0.0313
152
+ step 1500000 | 15582s | train_loss=0.0309 | fid=52.2456
153
+ step 1510000 | 15678s | train_loss=0.0307
154
+ step 1520000 | 15775s | train_loss=0.0308
155
+ step 1530000 | 15873s | train_loss=0.0299
156
+ step 1540000 | 15968s | train_loss=0.0298
157
+ step 1550000 | 16057s | train_loss=0.0297
158
+ step 1560000 | 16150s | train_loss=0.0298
159
+ step 1570000 | 16243s | train_loss=0.0284
160
+ step 1580000 | 16331s | train_loss=0.0303
161
+ step 1590000 | 16427s | train_loss=0.0284
162
+ step 1600000 | 16597s | train_loss=0.0291 | fid=49.7995
163
+ step 1610000 | 16694s | train_loss=0.0291
164
+ step 1620000 | 16792s | train_loss=0.0297
165
+ step 1630000 | 16889s | train_loss=0.0294
166
+ step 1640000 | 16980s | train_loss=0.0280
167
+ step 1650000 | 17073s | train_loss=0.0282
168
+ step 1660000 | 17161s | train_loss=0.0278
169
+ step 1670000 | 17258s | train_loss=0.0276
170
+ step 1680000 | 17350s | train_loss=0.0280
171
+ step 1690000 | 17444s | train_loss=0.0274
172
+ step 1700000 | 17615s | train_loss=0.0278 | fid=48.8564
173
+ step 1710000 | 17711s | train_loss=0.0283
174
+ step 1720000 | 17808s | train_loss=0.0274
175
+ step 1730000 | 17906s | train_loss=0.0272
176
+ step 1740000 | 17994s | train_loss=0.0280
177
+ step 1750000 | 18081s | train_loss=0.0274
178
+ step 1760000 | 18177s | train_loss=0.0273
179
+ step 1770000 | 18275s | train_loss=0.0267
180
+ step 1780000 | 18370s | train_loss=0.0286
181
+ step 1790000 | 18462s | train_loss=0.0263
182
+ step 1800000 | 18634s | train_loss=0.0270 | fid=49.6749
183
+ step 1810000 | 18731s | train_loss=0.0277
184
+ step 1820000 | 18829s | train_loss=0.0263
185
+ step 1830000 | 18915s | train_loss=0.0275
186
+ step 1840000 | 19010s | train_loss=0.0259
187
+ step 1850000 | 19100s | train_loss=0.0264
188
+ step 1860000 | 19197s | train_loss=0.0264
189
+ step 1870000 | 19294s | train_loss=0.0264
190
+ step 1880000 | 19389s | train_loss=0.0255
191
+ step 1890000 | 19478s | train_loss=0.0262
192
+ step 1900000 | 19649s | train_loss=0.0263 | fid=47.5337
193
+ step 1910000 | 19746s | train_loss=0.0269
194
+ step 1920000 | 19837s | train_loss=0.0251
195
+ step 1930000 | 19930s | train_loss=0.0254
196
+ step 1940000 | 20017s | train_loss=0.0252
197
+ step 1950000 | 20114s | train_loss=0.0252
198
+ step 1960000 | 20211s | train_loss=0.0249
199
+ step 1970000 | 20308s | train_loss=0.0254
200
+ step 1980000 | 20406s | train_loss=0.0263
201
+ step 1990000 | 20490s | train_loss=0.0253
202
+ step 2000000 | 20662s | train_loss=0.0254 | fid=47.4961
203
+ step 2010000 | 20759s | train_loss=0.0256
204
+ step 2020000 | 20848s | train_loss=0.0243
205
+ step 2030000 | 20940s | train_loss=0.0242
206
+ step 2040000 | 21034s | train_loss=0.0249
207
+ step 2050000 | 21131s | train_loss=0.0255
208
+ step 2060000 | 21228s | train_loss=0.0246
209
+ step 2070000 | 21326s | train_loss=0.0252
210
+ step 2080000 | 21424s | train_loss=0.0256
211
+ step 2090000 | 21507s | train_loss=0.0242
212
+ step 2100000 | 21675s | train_loss=0.0250 | fid=48.8376
213
+ step 2110000 | 21759s | train_loss=0.0247
214
+ step 2120000 | 21853s | train_loss=0.0240
215
+ step 2130000 | 21933s | train_loss=0.0236
216
+ step 2140000 | 22028s | train_loss=0.0229
217
+ step 2150000 | 22126s | train_loss=0.0235
218
+ step 2160000 | 22223s | train_loss=0.0232
219
+ step 2170000 | 22321s | train_loss=0.0231
220
+ step 2180000 | 22419s | train_loss=0.0235
221
+ step 2190000 | 22512s | train_loss=0.0237
222
+ step 2200000 | 22675s | train_loss=0.0240 | fid=47.1066
223
+ step 2210000 | 22759s | train_loss=0.0239
224
+ step 2220000 | 22856s | train_loss=0.0235
225
+ step 2230000 | 22939s | train_loss=0.0231
226
+ step 2240000 | 23037s | train_loss=0.0235
227
+ step 2250000 | 23134s | train_loss=0.0237
228
+ step 2260000 | 23232s | train_loss=0.0241
229
+ step 2270000 | 23330s | train_loss=0.0242
230
+ step 2280000 | 23427s | train_loss=0.0232
231
+ step 2290000 | 23525s | train_loss=0.0227
232
+ step 2300000 | 23680s | train_loss=0.0221 | fid=50.1308
233
+ step 2310000 | 23775s | train_loss=0.0222
234
+ step 2320000 | 23866s | train_loss=0.0218
235
+ step 2330000 | 23960s | train_loss=0.0224
236
+ step 2340000 | 24056s | train_loss=0.0229
237
+ step 2350000 | 24153s | train_loss=0.0227
238
+ step 2360000 | 24249s | train_loss=0.0231
239
+ step 2370000 | 24346s | train_loss=0.0229
240
+ step 2380000 | 24443s | train_loss=0.0224
241
+ step 2390000 | 24540s | train_loss=0.0213
242
+ step 2400000 | 24692s | train_loss=0.0220 | fid=46.6076
243
+ step 2410000 | 24786s | train_loss=0.0217
244
+ step 2420000 | 24876s | train_loss=0.0218
245
+ step 2430000 | 24972s | train_loss=0.0221
246
+ step 2440000 | 25069s | train_loss=0.0218
247
+ step 2450000 | 25165s | train_loss=0.0211
248
+ step 2460000 | 25262s | train_loss=0.0225
249
+ step 2470000 | 25359s | train_loss=0.0208
250
+ step 2480000 | 25455s | train_loss=0.0224
251
+ step 2490000 | 25548s | train_loss=0.0223
252
+ step 2500000 | 25705s | train_loss=0.0206 | fid=48.6467
253
+ step 2510000 | 25791s | train_loss=0.0221
254
+ step 2520000 | 25888s | train_loss=0.0211
255
+ step 2530000 | 25984s | train_loss=0.0201
256
+ step 2540000 | 26081s | train_loss=0.0219
257
+ step 2550000 | 26178s | train_loss=0.0202
258
+ step 2560000 | 26275s | train_loss=0.0211
259
+ step 2570000 | 26371s | train_loss=0.0216
260
+ step 2580000 | 26468s | train_loss=0.0212
261
+ step 2590000 | 26554s | train_loss=0.0210
262
+ step 2600000 | 26715s | train_loss=0.0211 | fid=47.4626
263
+ step 2610000 | 26804s | train_loss=0.0206
264
+ step 2620000 | 26900s | train_loss=0.0215
265
+ step 2630000 | 26996s | train_loss=0.0206
266
+ step 2640000 | 27093s | train_loss=0.0204
267
+ step 2650000 | 27190s | train_loss=0.0209
268
+ step 2660000 | 27287s | train_loss=0.0203
269
+ step 2670000 | 27384s | train_loss=0.0211
270
+ step 2680000 | 27472s | train_loss=0.0221
271
+ step 2690000 | 27569s | train_loss=0.0203
272
+ step 2700000 | 27718s | train_loss=0.0203 | fid=46.7321
273
+ step 2710000 | 27815s | train_loss=0.0196
274
+ step 2720000 | 27912s | train_loss=0.0212
275
+ step 2730000 | 28008s | train_loss=0.0207
276
+ step 2740000 | 28105s | train_loss=0.0206
277
+ step 2750000 | 28202s | train_loss=0.0204
278
+ step 2760000 | 28298s | train_loss=0.0206
279
+ step 2770000 | 28392s | train_loss=0.0206
280
+ step 2780000 | 28482s | train_loss=0.0198
281
+ step 2790000 | 28579s | train_loss=0.0205
282
+ step 2800000 | 28727s | train_loss=0.0198 | fid=47.5052
283
+ step 2810000 | 28824s | train_loss=0.0190
284
+ step 2820000 | 28921s | train_loss=0.0197
285
+ step 2830000 | 29017s | train_loss=0.0198
286
+ step 2840000 | 29115s | train_loss=0.0194
287
+ step 2850000 | 29212s | train_loss=0.0198
288
+ step 2860000 | 29309s | train_loss=0.0200
289
+ step 2870000 | 29399s | train_loss=0.0207
290
+ step 2880000 | 29496s | train_loss=0.0197
291
+ step 2890000 | 29591s | train_loss=0.0190
292
+ step 2900000 | 29739s | train_loss=0.0193 | fid=46.1672
293
+ step 2910000 | 29835s | train_loss=0.0197
294
+ step 2920000 | 29932s | train_loss=0.0191
295
+ step 2930000 | 30030s | train_loss=0.0188
296
+ step 2940000 | 30127s | train_loss=0.0190
297
+ step 2950000 | 30225s | train_loss=0.0199
298
+ step 2960000 | 30313s | train_loss=0.0198
299
+ step 2970000 | 30407s | train_loss=0.0198
300
+ step 2980000 | 30505s | train_loss=0.0192
301
+ step 2990000 | 30591s | train_loss=0.0187
302
+ step 3000000 | 30757s | train_loss=0.0198 | fid=45.8308
303
+ step 3010000 | 30841s | train_loss=0.0186
304
+ step 3020000 | 30938s | train_loss=0.0180
305
+ step 3030000 | 31036s | train_loss=0.0195
306
+ step 3040000 | 31133s | train_loss=0.0191
307
+ step 3050000 | 31230s | train_loss=0.0183
308
+ step 3060000 | 31318s | train_loss=0.0191
309
+ step 3070000 | 31416s | train_loss=0.0190
310
+ step 3080000 | 31511s | train_loss=0.0189
311
+ step 3090000 | 31601s | train_loss=0.0191
312
+ step 3100000 | 31773s | train_loss=0.0188 | fid=45.4173
313
+ step 3110000 | 31859s | train_loss=0.0196
314
+ step 3120000 | 31957s | train_loss=0.0190
315
+ step 3130000 | 32054s | train_loss=0.0187
316
+ step 3140000 | 32152s | train_loss=0.0192
317
+ step 3150000 | 32238s | train_loss=0.0176
318
+ step 3160000 | 32336s | train_loss=0.0180
319
+ step 3170000 | 32434s | train_loss=0.0191
320
+ step 3180000 | 32516s | train_loss=0.0181
321
+ step 3190000 | 32610s | train_loss=0.0175
322
+ step 3200000 | 32780s | train_loss=0.0189 | fid=46.2402
323
+ step 3210000 | 32865s | train_loss=0.0182
324
+ step 3220000 | 32962s | train_loss=0.0193
325
+ step 3230000 | 33060s | train_loss=0.0185
326
+ step 3240000 | 33154s | train_loss=0.0179
327
+ step 3250000 | 33245s | train_loss=0.0183
328
+ step 3260000 | 33343s | train_loss=0.0202
329
+ step 3270000 | 33439s | train_loss=0.0176
330
+ step 3280000 | 33526s | train_loss=0.0182
331
+ step 3290000 | 33624s | train_loss=0.0176
332
+ step 3300000 | 33795s | train_loss=0.0178 | fid=44.9881
333
+ step 3310000 | 33886s | train_loss=0.0183
334
+ step 3320000 | 33970s | train_loss=0.0171
335
+ step 3330000 | 34068s | train_loss=0.0187
336
+ step 3340000 | 34155s | train_loss=0.0175
337
+ step 3350000 | 34252s | train_loss=0.0172
338
+ step 3360000 | 34350s | train_loss=0.0173
339
+ step 3370000 | 34438s | train_loss=0.0192
340
+ step 3380000 | 34527s | train_loss=0.0182
341
+ step 3390000 | 34625s | train_loss=0.0173
342
+ step 3400000 | 34796s | train_loss=0.0179 | fid=44.7150
343
+ step 3410000 | 34894s | train_loss=0.0178
344
+ step 3420000 | 34977s | train_loss=0.0182
345
+ step 3430000 | 35064s | train_loss=0.0179
346
+ step 3440000 | 35155s | train_loss=0.0179
347
+ step 3450000 | 35253s | train_loss=0.0171
348
+ step 3460000 | 35351s | train_loss=0.0175
349
+ step 3470000 | 35434s | train_loss=0.0168
350
+ step 3480000 | 35528s | train_loss=0.0167
351
+ step 3490000 | 35626s | train_loss=0.0172
352
+ step 3500000 | 35797s | train_loss=0.0173 | fid=44.0635
353
+ step 3510000 | 35894s | train_loss=0.0166
354
+ step 3520000 | 35986s | train_loss=0.0171
355
+ step 3530000 | 36063s | train_loss=0.0177
356
+ step 3540000 | 36160s | train_loss=0.0185
357
+ step 3550000 | 36258s | train_loss=0.0170
358
+ step 3560000 | 36354s | train_loss=0.0178
359
+ step 3570000 | 36435s | train_loss=0.0171
360
+ step 3580000 | 36533s | train_loss=0.0176
361
+ step 3590000 | 36630s | train_loss=0.0168
362
+ step 3600000 | 36801s | train_loss=0.0166 | fid=44.6981
363
+ step 3610000 | 36898s | train_loss=0.0170
364
+ step 3620000 | 36991s | train_loss=0.0172
365
+ step 3630000 | 37066s | train_loss=0.0168
366
+ step 3640000 | 37163s | train_loss=0.0168
367
+ step 3650000 | 37261s | train_loss=0.0178
368
+ step 3660000 | 37354s | train_loss=0.0168
369
+ step 3670000 | 37443s | train_loss=0.0172
370
+ step 3680000 | 37540s | train_loss=0.0165
371
+ step 3690000 | 37638s | train_loss=0.0171
372
+ step 3700000 | 37809s | train_loss=0.0165 | fid=42.0926
373
+ step 3710000 | 37906s | train_loss=0.0173
374
+ step 3720000 | 37993s | train_loss=0.0169
375
+ step 3730000 | 38068s | train_loss=0.0161
376
+ step 3740000 | 38165s | train_loss=0.0174
377
+ step 3750000 | 38262s | train_loss=0.0168
378
+ step 3760000 | 38347s | train_loss=0.0165
379
+ step 3770000 | 38444s | train_loss=0.0175
380
+ step 3780000 | 38541s | train_loss=0.0160
381
+ step 3790000 | 38639s | train_loss=0.0163
382
+ step 3800000 | 38809s | train_loss=0.0161 | fid=44.2697
383
+ step 3810000 | 38907s | train_loss=0.0164
384
+ step 3820000 | 38989s | train_loss=0.0163
385
+ step 3830000 | 39080s | train_loss=0.0169
386
+ step 3840000 | 39166s | train_loss=0.0170
387
+ step 3850000 | 39260s | train_loss=0.0157
388
+ step 3860000 | 39350s | train_loss=0.0160
389
+ step 3870000 | 39448s | train_loss=0.0167
390
+ step 3880000 | 39545s | train_loss=0.0153
391
+ step 3890000 | 39643s | train_loss=0.0166
392
+ step 3900000 | 39814s | train_loss=0.0163 | fid=42.3317
393
+ step 3910000 | 39908s | train_loss=0.0160
394
+ step 3920000 | 39990s | train_loss=0.0154
395
+ step 3930000 | 40088s | train_loss=0.0161
396
+ step 3940000 | 40166s | train_loss=0.0155
397
+ step 3950000 | 40251s | train_loss=0.0162
398
+ step 3960000 | 40348s | train_loss=0.0154
399
+ step 3970000 | 40445s | train_loss=0.0174
400
+ step 3980000 | 40543s | train_loss=0.0157
401
+ step 3990000 | 40640s | train_loss=0.0162
402
+ step 4000000 | 40812s | train_loss=0.0153 | fid=43.2825
403
+ step 4010000 | 40903s | train_loss=0.0153
404
+ step 4020000 | 40993s | train_loss=0.0161
405
+ step 4030000 | 41091s | train_loss=0.0166
406
+ step 4040000 | 41177s | train_loss=0.0149
407
+ step 4050000 | 41257s | train_loss=0.0161
408
+ step 4060000 | 41354s | train_loss=0.0161
409
+ step 4070000 | 41451s | train_loss=0.0150
410
+ step 4080000 | 41548s | train_loss=0.0150
411
+ step 4090000 | 41646s | train_loss=0.0148
412
+ step 4100000 | 41816s | train_loss=0.0153 | fid=41.7485
413
+ step 4110000 | 41900s | train_loss=0.0156
414
+ step 4120000 | 41994s | train_loss=0.0149
415
+ step 4130000 | 42090s | train_loss=0.0156
416
+ step 4140000 | 42177s | train_loss=0.0163
417
+ step 4150000 | 42253s | train_loss=0.0148
418
+ step 4160000 | 42350s | train_loss=0.0166
419
+ step 4170000 | 42447s | train_loss=0.0162
420
+ step 4180000 | 42544s | train_loss=0.0150
421
+ step 4190000 | 42642s | train_loss=0.0160
422
+ step 4200000 | 42813s | train_loss=0.0149 | fid=41.1087
423
+ step 4210000 | 42901s | train_loss=0.0153
424
+ step 4220000 | 42999s | train_loss=0.0154
425
+ step 4230000 | 43095s | train_loss=0.0145
426
+ step 4240000 | 43177s | train_loss=0.0155
427
+ step 4250000 | 43261s | train_loss=0.0151
428
+ step 4260000 | 43356s | train_loss=0.0152
429
+ step 4270000 | 43453s | train_loss=0.0152
430
+ step 4280000 | 43550s | train_loss=0.0155
431
+ step 4290000 | 43648s | train_loss=0.0158
432
+ step 4300000 | 43808s | train_loss=0.0154 | fid=40.5707
433
+ step 4310000 | 43905s | train_loss=0.0160
434
+ step 4320000 | 44003s | train_loss=0.0147
435
+ step 4330000 | 44096s | train_loss=0.0151
436
+ step 4340000 | 44182s | train_loss=0.0150
437
+ step 4350000 | 44275s | train_loss=0.0154
438
+ step 4360000 | 44363s | train_loss=0.0162
439
+ step 4370000 | 44460s | train_loss=0.0153
440
+ step 4380000 | 44558s | train_loss=0.0149
441
+ step 4390000 | 44655s | train_loss=0.0151
442
+ step 4400000 | 44813s | train_loss=0.0149 | fid=40.4938
443
+ step 4410000 | 44910s | train_loss=0.0148
444
+ step 4420000 | 45006s | train_loss=0.0147
445
+ step 4430000 | 45094s | train_loss=0.0146
446
+ step 4440000 | 45185s | train_loss=0.0151
447
+ step 4450000 | 45283s | train_loss=0.0146
448
+ step 4460000 | 45364s | train_loss=0.0144
449
+ step 4470000 | 45461s | train_loss=0.0157
450
+ step 4480000 | 45558s | train_loss=0.0139
451
+ step 4490000 | 45655s | train_loss=0.0140
452
+ step 4500000 | 45814s | train_loss=0.0142 | fid=39.7126
453
+ step 4510000 | 45911s | train_loss=0.0143
454
+ step 4520000 | 46007s | train_loss=0.0141
455
+ step 4530000 | 46092s | train_loss=0.0160
456
+ step 4540000 | 46189s | train_loss=0.0142
457
+ step 4550000 | 46287s | train_loss=0.0143
458
+ step 4560000 | 46376s | train_loss=0.0142
459
+ step 4570000 | 46468s | train_loss=0.0143
460
+ step 4580000 | 46565s | train_loss=0.0137
461
+ step 4590000 | 46658s | train_loss=0.0157
462
+ step 4600000 | 46827s | train_loss=0.0139 | fid=40.5857
463
+ step 4610000 | 46924s | train_loss=0.0139
464
+ step 4620000 | 47017s | train_loss=0.0150
465
+ step 4630000 | 47103s | train_loss=0.0158
466
+ step 4640000 | 47201s | train_loss=0.0149
467
+ step 4650000 | 47298s | train_loss=0.0150
468
+ step 4660000 | 47393s | train_loss=0.0147
469
+ step 4670000 | 47477s | train_loss=0.0137
470
+ step 4680000 | 47573s | train_loss=0.0142
471
+ step 4690000 | 47652s | train_loss=0.0142
472
+ step 4700000 | 47792s | train_loss=0.0144 | fid=40.3667
473
+ step 4710000 | 47865s | train_loss=0.0140
474
+ step 4720000 | 47934s | train_loss=0.0141
475
+ step 4730000 | 47994s | train_loss=0.0135
476
+ step 4740000 | 48054s | train_loss=0.0138
477
+ step 4750000 | 48114s | train_loss=0.0142
478
+ step 4760000 | 48175s | train_loss=0.0142
479
+ step 4770000 | 48229s | train_loss=0.0141
480
+ step 4780000 | 48289s | train_loss=0.0138
481
+ step 4790000 | 48350s | train_loss=0.0131
482
+ step 4800000 | 48459s | train_loss=0.0139 | fid=40.3406
483
+ step 4810000 | 48499s | train_loss=0.0146
484
+ step 4820000 | 48539s | train_loss=0.0148
485
+ step 4830000 | 48580s | train_loss=0.0137
486
+ step 4840000 | 48620s | train_loss=0.0139
487
+ step 4850000 | 48660s | train_loss=0.0132
488
+ step 4860000 | 48700s | train_loss=0.0140
489
+ step 4870000 | 48733s | train_loss=0.0138
490
+ step 4880000 | 48774s | train_loss=0.0136
491
+ step 4890000 | 48814s | train_loss=0.0144
492
+ step 4900000 | 48890s | train_loss=0.0138 | fid=41.0696
493
+ step 4910000 | 48930s | train_loss=0.0137
494
+ step 4920000 | 48970s | train_loss=0.0149
495
+ step 4930000 | 49011s | train_loss=0.0138
496
+ step 4940000 | 49051s | train_loss=0.0134
497
+ step 4950000 | 49091s | train_loss=0.0146
498
+ step 4960000 | 49129s | train_loss=0.0141
499
+ step 4970000 | 49165s | train_loss=0.0131
500
+ step 4980000 | 49206s | train_loss=0.0140
501
+ step 4990000 | 49246s | train_loss=0.0131
502
+ step 5000000 | 49320s | train_loss=0.0128 | fid=40.0213
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d120/config.json ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "vae_checkpoint": "vae_checkpoints/celeba_resnet_modernloss_d120/vae.pt",
3
+ "out": "results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d120",
4
+ "subset_path": "diagnostics/celeba_diverse_1k_mean_plus_sd/subset_indices.json",
5
+ "train_tar": "/overflow/data/cerberus/celeba_hq_256_partitioned_ws/train/celeba_hq_256-train-0000.tar",
6
+ "n_train": 1000,
7
+ "total_steps": 5000000,
8
+ "eval_interval": 10000,
9
+ "mem_interval": 100000,
10
+ "fid_interval": 100000,
11
+ "fid_n_real": 1000,
12
+ "fid_n_gen": 10000,
13
+ "n_gen_samples": 10000,
14
+ "n_sde_steps": 500,
15
+ "batch_size": 512,
16
+ "hidden": 1024,
17
+ "depth": 5,
18
+ "optimizer": "sgd",
19
+ "lr": 0.001,
20
+ "momentum": 0.8,
21
+ "t_min": 0.01,
22
+ "t_max": 3.0,
23
+ "t_eval": 0.1,
24
+ "seed": 45,
25
+ "device": null
26
+ }
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d120/metrics.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d120/runner.log ADDED
@@ -0,0 +1,502 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ FID real-side stats cached over 1000 diverse train images
2
+ step 1 | 74s | train_loss=1.0001 | fid=265.9095
3
+ step 10000 | 170s | train_loss=0.7912
4
+ step 20000 | 268s | train_loss=0.6559
5
+ step 30000 | 367s | train_loss=0.5214
6
+ step 40000 | 465s | train_loss=0.3989
7
+ step 50000 | 563s | train_loss=0.3015
8
+ step 60000 | 662s | train_loss=0.2352
9
+ step 70000 | 760s | train_loss=0.1914
10
+ step 80000 | 859s | train_loss=0.1640
11
+ step 90000 | 957s | train_loss=0.1453
12
+ step 100000 | 1126s | train_loss=0.1334 | fid=63.9819
13
+ step 110000 | 1216s | train_loss=0.1242
14
+ step 120000 | 1315s | train_loss=0.1169
15
+ step 130000 | 1413s | train_loss=0.1112
16
+ step 140000 | 1512s | train_loss=0.1062
17
+ step 150000 | 1610s | train_loss=0.1050
18
+ step 160000 | 1707s | train_loss=0.0993
19
+ step 170000 | 1807s | train_loss=0.0943
20
+ step 180000 | 1907s | train_loss=0.0939
21
+ step 190000 | 2007s | train_loss=0.0897
22
+ step 200000 | 2179s | train_loss=0.0872 | fid=64.7048
23
+ step 210000 | 2274s | train_loss=0.0844
24
+ step 220000 | 2373s | train_loss=0.0829
25
+ step 230000 | 2474s | train_loss=0.0831
26
+ step 240000 | 2575s | train_loss=0.0796
27
+ step 250000 | 2674s | train_loss=0.0786
28
+ step 260000 | 2772s | train_loss=0.0767
29
+ step 270000 | 2871s | train_loss=0.0754
30
+ step 280000 | 2971s | train_loss=0.0742
31
+ step 290000 | 3067s | train_loss=0.0715
32
+ step 300000 | 3242s | train_loss=0.0711 | fid=61.4377
33
+ step 310000 | 3336s | train_loss=0.0709
34
+ step 320000 | 3436s | train_loss=0.0674
35
+ step 330000 | 3536s | train_loss=0.0657
36
+ step 340000 | 3635s | train_loss=0.0665
37
+ step 350000 | 3735s | train_loss=0.0662
38
+ step 360000 | 3833s | train_loss=0.0630
39
+ step 370000 | 3931s | train_loss=0.0624
40
+ step 380000 | 4028s | train_loss=0.0629
41
+ step 390000 | 4128s | train_loss=0.0619
42
+ step 400000 | 4302s | train_loss=0.0611 | fid=60.0681
43
+ step 410000 | 4398s | train_loss=0.0594
44
+ step 420000 | 4497s | train_loss=0.0596
45
+ step 430000 | 4597s | train_loss=0.0595
46
+ step 440000 | 4697s | train_loss=0.0576
47
+ step 450000 | 4797s | train_loss=0.0564
48
+ step 460000 | 4895s | train_loss=0.0560
49
+ step 470000 | 4991s | train_loss=0.0565
50
+ step 480000 | 5090s | train_loss=0.0543
51
+ step 490000 | 5190s | train_loss=0.0538
52
+ step 500000 | 5365s | train_loss=0.0554 | fid=60.2788
53
+ step 510000 | 5463s | train_loss=0.0534
54
+ step 520000 | 5560s | train_loss=0.0528
55
+ step 530000 | 5661s | train_loss=0.0520
56
+ step 540000 | 5760s | train_loss=0.0522
57
+ step 550000 | 5859s | train_loss=0.0528
58
+ step 560000 | 5956s | train_loss=0.0513
59
+ step 570000 | 6053s | train_loss=0.0526
60
+ step 580000 | 6152s | train_loss=0.0511
61
+ step 590000 | 6252s | train_loss=0.0498
62
+ step 600000 | 6428s | train_loss=0.0485 | fid=60.8280
63
+ step 610000 | 6527s | train_loss=0.0491
64
+ step 620000 | 6625s | train_loss=0.0523
65
+ step 630000 | 6725s | train_loss=0.0486
66
+ step 640000 | 6824s | train_loss=0.0480
67
+ step 650000 | 6925s | train_loss=0.0489
68
+ step 660000 | 7020s | train_loss=0.0488
69
+ step 670000 | 7119s | train_loss=0.0500
70
+ step 680000 | 7218s | train_loss=0.0490
71
+ step 690000 | 7317s | train_loss=0.0485
72
+ step 700000 | 7490s | train_loss=0.0458 | fid=58.8146
73
+ step 710000 | 7590s | train_loss=0.0470
74
+ step 720000 | 7688s | train_loss=0.0473
75
+ step 730000 | 7788s | train_loss=0.0461
76
+ step 740000 | 7888s | train_loss=0.0459
77
+ step 750000 | 7984s | train_loss=0.0449
78
+ step 760000 | 8084s | train_loss=0.0447
79
+ step 770000 | 8185s | train_loss=0.0456
80
+ step 780000 | 8284s | train_loss=0.0457
81
+ step 790000 | 8383s | train_loss=0.0457
82
+ step 800000 | 8555s | train_loss=0.0440 | fid=61.4269
83
+ step 810000 | 8654s | train_loss=0.0436
84
+ step 820000 | 8754s | train_loss=0.0432
85
+ step 830000 | 8853s | train_loss=0.0443
86
+ step 840000 | 8947s | train_loss=0.0441
87
+ step 850000 | 9036s | train_loss=0.0437
88
+ step 860000 | 9133s | train_loss=0.0420
89
+ step 870000 | 9233s | train_loss=0.0430
90
+ step 880000 | 9333s | train_loss=0.0420
91
+ step 890000 | 9430s | train_loss=0.0421
92
+ step 900000 | 9602s | train_loss=0.0411 | fid=61.4851
93
+ step 910000 | 9702s | train_loss=0.0435
94
+ step 920000 | 9803s | train_loss=0.0421
95
+ step 930000 | 9902s | train_loss=0.0403
96
+ step 940000 | 9999s | train_loss=0.0426
97
+ step 950000 | 10099s | train_loss=0.0406
98
+ step 960000 | 10199s | train_loss=0.0409
99
+ step 970000 | 10298s | train_loss=0.0399
100
+ step 980000 | 10398s | train_loss=0.0406
101
+ step 990000 | 10497s | train_loss=0.0403
102
+ step 1000000 | 10668s | train_loss=0.0404 | fid=61.8049
103
+ step 1010000 | 10768s | train_loss=0.0421
104
+ step 1020000 | 10868s | train_loss=0.0383
105
+ step 1030000 | 10964s | train_loss=0.0397
106
+ step 1040000 | 11063s | train_loss=0.0396
107
+ step 1050000 | 11163s | train_loss=0.0385
108
+ step 1060000 | 11263s | train_loss=0.0403
109
+ step 1070000 | 11361s | train_loss=0.0393
110
+ step 1080000 | 11461s | train_loss=0.0390
111
+ step 1090000 | 11561s | train_loss=0.0387
112
+ step 1100000 | 11733s | train_loss=0.0385 | fid=62.7754
113
+ step 1110000 | 11833s | train_loss=0.0385
114
+ step 1120000 | 11933s | train_loss=0.0390
115
+ step 1130000 | 12021s | train_loss=0.0388
116
+ step 1140000 | 12121s | train_loss=0.0375
117
+ step 1150000 | 12221s | train_loss=0.0380
118
+ step 1160000 | 12320s | train_loss=0.0378
119
+ step 1170000 | 12418s | train_loss=0.0369
120
+ step 1180000 | 12518s | train_loss=0.0378
121
+ step 1190000 | 12617s | train_loss=0.0363
122
+ step 1200000 | 12789s | train_loss=0.0375 | fid=66.6862
123
+ step 1210000 | 12888s | train_loss=0.0367
124
+ step 1220000 | 12986s | train_loss=0.0369
125
+ step 1230000 | 13087s | train_loss=0.0360
126
+ step 1240000 | 13184s | train_loss=0.0360
127
+ step 1250000 | 13285s | train_loss=0.0372
128
+ step 1260000 | 13383s | train_loss=0.0364
129
+ step 1270000 | 13483s | train_loss=0.0351
130
+ step 1280000 | 13582s | train_loss=0.0361
131
+ step 1290000 | 13682s | train_loss=0.0365
132
+ step 1300000 | 13853s | train_loss=0.0362 | fid=67.0263
133
+ step 1310000 | 13952s | train_loss=0.0373
134
+ step 1320000 | 14040s | train_loss=0.0353
135
+ step 1330000 | 14140s | train_loss=0.0346
136
+ step 1340000 | 14239s | train_loss=0.0343
137
+ step 1350000 | 14339s | train_loss=0.0351
138
+ step 1360000 | 14436s | train_loss=0.0350
139
+ step 1370000 | 14536s | train_loss=0.0339
140
+ step 1380000 | 14636s | train_loss=0.0339
141
+ step 1390000 | 14735s | train_loss=0.0342
142
+ step 1400000 | 14906s | train_loss=0.0334 | fid=67.0722
143
+ step 1410000 | 15001s | train_loss=0.0344
144
+ step 1420000 | 15096s | train_loss=0.0338
145
+ step 1430000 | 15196s | train_loss=0.0331
146
+ step 1440000 | 15294s | train_loss=0.0339
147
+ step 1450000 | 15393s | train_loss=0.0333
148
+ step 1460000 | 15491s | train_loss=0.0336
149
+ step 1470000 | 15592s | train_loss=0.0324
150
+ step 1480000 | 15692s | train_loss=0.0332
151
+ step 1490000 | 15791s | train_loss=0.0341
152
+ step 1500000 | 15962s | train_loss=0.0328 | fid=66.5760
153
+ step 1510000 | 16053s | train_loss=0.0330
154
+ step 1520000 | 16153s | train_loss=0.0335
155
+ step 1530000 | 16252s | train_loss=0.0318
156
+ step 1540000 | 16350s | train_loss=0.0329
157
+ step 1550000 | 16448s | train_loss=0.0331
158
+ step 1560000 | 16545s | train_loss=0.0317
159
+ step 1570000 | 16645s | train_loss=0.0310
160
+ step 1580000 | 16745s | train_loss=0.0330
161
+ step 1590000 | 16844s | train_loss=0.0329
162
+ step 1600000 | 17016s | train_loss=0.0320 | fid=67.8003
163
+ step 1610000 | 17103s | train_loss=0.0325
164
+ step 1620000 | 17203s | train_loss=0.0316
165
+ step 1630000 | 17302s | train_loss=0.0318
166
+ step 1640000 | 17402s | train_loss=0.0313
167
+ step 1650000 | 17499s | train_loss=0.0303
168
+ step 1660000 | 17598s | train_loss=0.0302
169
+ step 1670000 | 17698s | train_loss=0.0305
170
+ step 1680000 | 17798s | train_loss=0.0300
171
+ step 1690000 | 17896s | train_loss=0.0305
172
+ step 1700000 | 18066s | train_loss=0.0305 | fid=65.0535
173
+ step 1710000 | 18155s | train_loss=0.0307
174
+ step 1720000 | 18255s | train_loss=0.0299
175
+ step 1730000 | 18353s | train_loss=0.0309
176
+ step 1740000 | 18453s | train_loss=0.0306
177
+ step 1750000 | 18552s | train_loss=0.0301
178
+ step 1760000 | 18651s | train_loss=0.0299
179
+ step 1770000 | 18751s | train_loss=0.0299
180
+ step 1780000 | 18850s | train_loss=0.0299
181
+ step 1790000 | 18950s | train_loss=0.0291
182
+ step 1800000 | 19117s | train_loss=0.0284 | fid=64.2844
183
+ step 1810000 | 19210s | train_loss=0.0301
184
+ step 1820000 | 19309s | train_loss=0.0290
185
+ step 1830000 | 19410s | train_loss=0.0287
186
+ step 1840000 | 19509s | train_loss=0.0278
187
+ step 1850000 | 19607s | train_loss=0.0286
188
+ step 1860000 | 19707s | train_loss=0.0289
189
+ step 1870000 | 19806s | train_loss=0.0292
190
+ step 1880000 | 19905s | train_loss=0.0275
191
+ step 1890000 | 20005s | train_loss=0.0284
192
+ step 1900000 | 20168s | train_loss=0.0291 | fid=63.2072
193
+ step 1910000 | 20268s | train_loss=0.0273
194
+ step 1920000 | 20368s | train_loss=0.0271
195
+ step 1930000 | 20468s | train_loss=0.0289
196
+ step 1940000 | 20567s | train_loss=0.0275
197
+ step 1950000 | 20667s | train_loss=0.0269
198
+ step 1960000 | 20765s | train_loss=0.0267
199
+ step 1970000 | 20863s | train_loss=0.0298
200
+ step 1980000 | 20963s | train_loss=0.0272
201
+ step 1990000 | 21062s | train_loss=0.0288
202
+ step 2000000 | 21220s | train_loss=0.0271 | fid=62.6215
203
+ step 2010000 | 21319s | train_loss=0.0273
204
+ step 2020000 | 21419s | train_loss=0.0276
205
+ step 2030000 | 21519s | train_loss=0.0278
206
+ step 2040000 | 21617s | train_loss=0.0262
207
+ step 2050000 | 21716s | train_loss=0.0269
208
+ step 2060000 | 21814s | train_loss=0.0287
209
+ step 2070000 | 21912s | train_loss=0.0269
210
+ step 2080000 | 22012s | train_loss=0.0279
211
+ step 2090000 | 22109s | train_loss=0.0255
212
+ step 2100000 | 22269s | train_loss=0.0261 | fid=60.3396
213
+ step 2110000 | 22368s | train_loss=0.0270
214
+ step 2120000 | 22468s | train_loss=0.0273
215
+ step 2130000 | 22569s | train_loss=0.0264
216
+ step 2140000 | 22669s | train_loss=0.0251
217
+ step 2150000 | 22769s | train_loss=0.0259
218
+ step 2160000 | 22866s | train_loss=0.0253
219
+ step 2170000 | 22966s | train_loss=0.0256
220
+ step 2180000 | 23064s | train_loss=0.0252
221
+ step 2190000 | 23163s | train_loss=0.0262
222
+ step 2200000 | 23323s | train_loss=0.0253 | fid=57.6667
223
+ step 2210000 | 23423s | train_loss=0.0249
224
+ step 2220000 | 23523s | train_loss=0.0260
225
+ step 2230000 | 23624s | train_loss=0.0247
226
+ step 2240000 | 23723s | train_loss=0.0261
227
+ step 2250000 | 23822s | train_loss=0.0264
228
+ step 2260000 | 23919s | train_loss=0.0249
229
+ step 2270000 | 24018s | train_loss=0.0243
230
+ step 2280000 | 24117s | train_loss=0.0252
231
+ step 2290000 | 24215s | train_loss=0.0254
232
+ step 2300000 | 24380s | train_loss=0.0238 | fid=57.7434
233
+ step 2310000 | 24480s | train_loss=0.0254
234
+ step 2320000 | 24580s | train_loss=0.0245
235
+ step 2330000 | 24680s | train_loss=0.0240
236
+ step 2340000 | 24779s | train_loss=0.0239
237
+ step 2350000 | 24879s | train_loss=0.0241
238
+ step 2360000 | 24979s | train_loss=0.0244
239
+ step 2370000 | 25076s | train_loss=0.0246
240
+ step 2380000 | 25175s | train_loss=0.0238
241
+ step 2390000 | 25262s | train_loss=0.0240
242
+ step 2400000 | 25432s | train_loss=0.0248 | fid=57.9408
243
+ step 2410000 | 25532s | train_loss=0.0240
244
+ step 2420000 | 25632s | train_loss=0.0237
245
+ step 2430000 | 25732s | train_loss=0.0245
246
+ step 2440000 | 25832s | train_loss=0.0232
247
+ step 2450000 | 25931s | train_loss=0.0234
248
+ step 2460000 | 26031s | train_loss=0.0256
249
+ step 2470000 | 26129s | train_loss=0.0238
250
+ step 2480000 | 26227s | train_loss=0.0253
251
+ step 2490000 | 26313s | train_loss=0.0253
252
+ step 2500000 | 26486s | train_loss=0.0235 | fid=56.5293
253
+ step 2510000 | 26586s | train_loss=0.0234
254
+ step 2520000 | 26686s | train_loss=0.0232
255
+ step 2530000 | 26785s | train_loss=0.0225
256
+ step 2540000 | 26884s | train_loss=0.0237
257
+ step 2550000 | 26984s | train_loss=0.0226
258
+ step 2560000 | 27083s | train_loss=0.0232
259
+ step 2570000 | 27181s | train_loss=0.0227
260
+ step 2580000 | 27278s | train_loss=0.0237
261
+ step 2590000 | 27365s | train_loss=0.0237
262
+ step 2600000 | 27537s | train_loss=0.0228 | fid=56.3741
263
+ step 2610000 | 27637s | train_loss=0.0220
264
+ step 2620000 | 27736s | train_loss=0.0226
265
+ step 2630000 | 27835s | train_loss=0.0225
266
+ step 2640000 | 27935s | train_loss=0.0210
267
+ step 2650000 | 28034s | train_loss=0.0224
268
+ step 2660000 | 28134s | train_loss=0.0220
269
+ step 2670000 | 28235s | train_loss=0.0221
270
+ step 2680000 | 28325s | train_loss=0.0239
271
+ step 2690000 | 28423s | train_loss=0.0219
272
+ step 2700000 | 28594s | train_loss=0.0222 | fid=55.5351
273
+ step 2710000 | 28694s | train_loss=0.0220
274
+ step 2720000 | 28793s | train_loss=0.0231
275
+ step 2730000 | 28893s | train_loss=0.0219
276
+ step 2740000 | 28993s | train_loss=0.0221
277
+ step 2750000 | 29092s | train_loss=0.0223
278
+ step 2760000 | 29190s | train_loss=0.0224
279
+ step 2770000 | 29291s | train_loss=0.0217
280
+ step 2780000 | 29385s | train_loss=0.0236
281
+ step 2790000 | 29484s | train_loss=0.0239
282
+ step 2800000 | 29658s | train_loss=0.0218 | fid=55.6380
283
+ step 2810000 | 29757s | train_loss=0.0221
284
+ step 2820000 | 29857s | train_loss=0.0220
285
+ step 2830000 | 29958s | train_loss=0.0213
286
+ step 2840000 | 30056s | train_loss=0.0209
287
+ step 2850000 | 30156s | train_loss=0.0212
288
+ step 2860000 | 30257s | train_loss=0.0214
289
+ step 2870000 | 30348s | train_loss=0.0224
290
+ step 2880000 | 30440s | train_loss=0.0210
291
+ step 2890000 | 30538s | train_loss=0.0213
292
+ step 2900000 | 30712s | train_loss=0.0213 | fid=54.4276
293
+ step 2910000 | 30811s | train_loss=0.0215
294
+ step 2920000 | 30911s | train_loss=0.0215
295
+ step 2930000 | 31010s | train_loss=0.0219
296
+ step 2940000 | 31110s | train_loss=0.0225
297
+ step 2950000 | 31210s | train_loss=0.0213
298
+ step 2960000 | 31311s | train_loss=0.0209
299
+ step 2970000 | 31401s | train_loss=0.0220
300
+ step 2980000 | 31500s | train_loss=0.0199
301
+ step 2990000 | 31597s | train_loss=0.0207
302
+ step 3000000 | 31770s | train_loss=0.0214 | fid=56.8594
303
+ step 3010000 | 31870s | train_loss=0.0209
304
+ step 3020000 | 31969s | train_loss=0.0198
305
+ step 3030000 | 32070s | train_loss=0.0202
306
+ step 3040000 | 32169s | train_loss=0.0215
307
+ step 3050000 | 32269s | train_loss=0.0208
308
+ step 3060000 | 32369s | train_loss=0.0215
309
+ step 3070000 | 32463s | train_loss=0.0205
310
+ step 3080000 | 32563s | train_loss=0.0204
311
+ step 3090000 | 32658s | train_loss=0.0206
312
+ step 3100000 | 32831s | train_loss=0.0206 | fid=53.6776
313
+ step 3110000 | 32931s | train_loss=0.0209
314
+ step 3120000 | 33032s | train_loss=0.0206
315
+ step 3130000 | 33132s | train_loss=0.0199
316
+ step 3140000 | 33232s | train_loss=0.0200
317
+ step 3150000 | 33332s | train_loss=0.0201
318
+ step 3160000 | 33429s | train_loss=0.0199
319
+ step 3170000 | 33528s | train_loss=0.0213
320
+ step 3180000 | 33629s | train_loss=0.0201
321
+ step 3190000 | 33724s | train_loss=0.0193
322
+ step 3200000 | 33899s | train_loss=0.0209 | fid=51.7212
323
+ step 3210000 | 34000s | train_loss=0.0196
324
+ step 3220000 | 34100s | train_loss=0.0200
325
+ step 3230000 | 34200s | train_loss=0.0207
326
+ step 3240000 | 34299s | train_loss=0.0206
327
+ step 3250000 | 34390s | train_loss=0.0202
328
+ step 3260000 | 34489s | train_loss=0.0219
329
+ step 3270000 | 34589s | train_loss=0.0196
330
+ step 3280000 | 34689s | train_loss=0.0188
331
+ step 3290000 | 34784s | train_loss=0.0184
332
+ step 3300000 | 34958s | train_loss=0.0194 | fid=53.2961
333
+ step 3310000 | 35059s | train_loss=0.0192
334
+ step 3320000 | 35158s | train_loss=0.0194
335
+ step 3330000 | 35257s | train_loss=0.0192
336
+ step 3340000 | 35357s | train_loss=0.0189
337
+ step 3350000 | 35448s | train_loss=0.0188
338
+ step 3360000 | 35548s | train_loss=0.0190
339
+ step 3370000 | 35648s | train_loss=0.0207
340
+ step 3380000 | 35747s | train_loss=0.0189
341
+ step 3390000 | 35844s | train_loss=0.0184
342
+ step 3400000 | 36016s | train_loss=0.0189 | fid=49.9300
343
+ step 3410000 | 36116s | train_loss=0.0192
344
+ step 3420000 | 36217s | train_loss=0.0195
345
+ step 3430000 | 36317s | train_loss=0.0185
346
+ step 3440000 | 36408s | train_loss=0.0196
347
+ step 3450000 | 36506s | train_loss=0.0191
348
+ step 3460000 | 36606s | train_loss=0.0192
349
+ step 3470000 | 36705s | train_loss=0.0198
350
+ step 3480000 | 36806s | train_loss=0.0187
351
+ step 3490000 | 36901s | train_loss=0.0183
352
+ step 3500000 | 37075s | train_loss=0.0190 | fid=48.5364
353
+ step 3510000 | 37175s | train_loss=0.0187
354
+ step 3520000 | 37275s | train_loss=0.0180
355
+ step 3530000 | 37374s | train_loss=0.0191
356
+ step 3540000 | 37463s | train_loss=0.0190
357
+ step 3550000 | 37562s | train_loss=0.0183
358
+ step 3560000 | 37661s | train_loss=0.0188
359
+ step 3570000 | 37761s | train_loss=0.0194
360
+ step 3580000 | 37860s | train_loss=0.0187
361
+ step 3590000 | 37958s | train_loss=0.0183
362
+ step 3600000 | 38132s | train_loss=0.0182 | fid=47.7330
363
+ step 3610000 | 38231s | train_loss=0.0195
364
+ step 3620000 | 38332s | train_loss=0.0189
365
+ step 3630000 | 38428s | train_loss=0.0181
366
+ step 3640000 | 38520s | train_loss=0.0185
367
+ step 3650000 | 38619s | train_loss=0.0195
368
+ step 3660000 | 38718s | train_loss=0.0181
369
+ step 3670000 | 38819s | train_loss=0.0194
370
+ step 3680000 | 38918s | train_loss=0.0184
371
+ step 3690000 | 39017s | train_loss=0.0187
372
+ step 3700000 | 39191s | train_loss=0.0181 | fid=47.0487
373
+ step 3710000 | 39292s | train_loss=0.0187
374
+ step 3720000 | 39391s | train_loss=0.0183
375
+ step 3730000 | 39481s | train_loss=0.0186
376
+ step 3740000 | 39578s | train_loss=0.0184
377
+ step 3750000 | 39678s | train_loss=0.0177
378
+ step 3760000 | 39778s | train_loss=0.0177
379
+ step 3770000 | 39879s | train_loss=0.0196
380
+ step 3780000 | 39978s | train_loss=0.0184
381
+ step 3790000 | 40077s | train_loss=0.0179
382
+ step 3800000 | 40250s | train_loss=0.0176 | fid=44.7128
383
+ step 3810000 | 40351s | train_loss=0.0173
384
+ step 3820000 | 40450s | train_loss=0.0178
385
+ step 3830000 | 40537s | train_loss=0.0180
386
+ step 3840000 | 40637s | train_loss=0.0181
387
+ step 3850000 | 40737s | train_loss=0.0183
388
+ step 3860000 | 40836s | train_loss=0.0173
389
+ step 3870000 | 40935s | train_loss=0.0191
390
+ step 3880000 | 41030s | train_loss=0.0180
391
+ step 3890000 | 41130s | train_loss=0.0204
392
+ step 3900000 | 41303s | train_loss=0.0175 | fid=43.6378
393
+ step 3910000 | 41403s | train_loss=0.0170
394
+ step 3920000 | 41501s | train_loss=0.0173
395
+ step 3930000 | 41591s | train_loss=0.0167
396
+ step 3940000 | 41691s | train_loss=0.0172
397
+ step 3950000 | 41790s | train_loss=0.0172
398
+ step 3960000 | 41889s | train_loss=0.0171
399
+ step 3970000 | 41989s | train_loss=0.0193
400
+ step 3980000 | 42086s | train_loss=0.0172
401
+ step 3990000 | 42186s | train_loss=0.0173
402
+ step 4000000 | 42362s | train_loss=0.0173 | fid=44.6570
403
+ step 4010000 | 42462s | train_loss=0.0180
404
+ step 4020000 | 42553s | train_loss=0.0183
405
+ step 4030000 | 42653s | train_loss=0.0180
406
+ step 4040000 | 42751s | train_loss=0.0166
407
+ step 4050000 | 42851s | train_loss=0.0178
408
+ step 4060000 | 42952s | train_loss=0.0175
409
+ step 4070000 | 43051s | train_loss=0.0170
410
+ step 4080000 | 43149s | train_loss=0.0166
411
+ step 4090000 | 43250s | train_loss=0.0160
412
+ step 4100000 | 43426s | train_loss=0.0164 | fid=44.0303
413
+ step 4110000 | 43524s | train_loss=0.0167
414
+ step 4120000 | 43614s | train_loss=0.0162
415
+ step 4130000 | 43714s | train_loss=0.0172
416
+ step 4140000 | 43813s | train_loss=0.0172
417
+ step 4150000 | 43913s | train_loss=0.0171
418
+ step 4160000 | 44014s | train_loss=0.0168
419
+ step 4170000 | 44115s | train_loss=0.0184
420
+ step 4180000 | 44212s | train_loss=0.0170
421
+ step 4190000 | 44313s | train_loss=0.0172
422
+ step 4200000 | 44488s | train_loss=0.0163 | fid=43.7797
423
+ step 4210000 | 44576s | train_loss=0.0167
424
+ step 4220000 | 44676s | train_loss=0.0176
425
+ step 4230000 | 44774s | train_loss=0.0162
426
+ step 4240000 | 44874s | train_loss=0.0179
427
+ step 4250000 | 44974s | train_loss=0.0167
428
+ step 4260000 | 45074s | train_loss=0.0160
429
+ step 4270000 | 45175s | train_loss=0.0163
430
+ step 4280000 | 45271s | train_loss=0.0176
431
+ step 4290000 | 45371s | train_loss=0.0173
432
+ step 4300000 | 45546s | train_loss=0.0173 | fid=42.4960
433
+ step 4310000 | 45636s | train_loss=0.0167
434
+ step 4320000 | 45736s | train_loss=0.0164
435
+ step 4330000 | 45835s | train_loss=0.0172
436
+ step 4340000 | 45936s | train_loss=0.0164
437
+ step 4350000 | 46037s | train_loss=0.0176
438
+ step 4360000 | 46138s | train_loss=0.0183
439
+ step 4370000 | 46238s | train_loss=0.0175
440
+ step 4380000 | 46334s | train_loss=0.0164
441
+ step 4390000 | 46434s | train_loss=0.0172
442
+ step 4400000 | 46599s | train_loss=0.0165 | fid=42.9901
443
+ step 4410000 | 46697s | train_loss=0.0166
444
+ step 4420000 | 46794s | train_loss=0.0161
445
+ step 4430000 | 46895s | train_loss=0.0166
446
+ step 4440000 | 46995s | train_loss=0.0164
447
+ step 4450000 | 47095s | train_loss=0.0165
448
+ step 4460000 | 47195s | train_loss=0.0158
449
+ step 4470000 | 47293s | train_loss=0.0167
450
+ step 4480000 | 47391s | train_loss=0.0154
451
+ step 4490000 | 47492s | train_loss=0.0158
452
+ step 4500000 | 47654s | train_loss=0.0155 | fid=42.7685
453
+ step 4510000 | 47754s | train_loss=0.0155
454
+ step 4520000 | 47853s | train_loss=0.0161
455
+ step 4530000 | 47954s | train_loss=0.0177
456
+ step 4540000 | 48054s | train_loss=0.0164
457
+ step 4550000 | 48152s | train_loss=0.0164
458
+ step 4560000 | 48252s | train_loss=0.0159
459
+ step 4570000 | 48351s | train_loss=0.0161
460
+ step 4580000 | 48449s | train_loss=0.0152
461
+ step 4590000 | 48550s | train_loss=0.0182
462
+ step 4600000 | 48721s | train_loss=0.0160 | fid=42.4447
463
+ step 4610000 | 48821s | train_loss=0.0154
464
+ step 4620000 | 48920s | train_loss=0.0166
465
+ step 4630000 | 49018s | train_loss=0.0171
466
+ step 4640000 | 49119s | train_loss=0.0160
467
+ step 4650000 | 49220s | train_loss=0.0165
468
+ step 4660000 | 49321s | train_loss=0.0165
469
+ step 4670000 | 49419s | train_loss=0.0161
470
+ step 4680000 | 49518s | train_loss=0.0159
471
+ step 4690000 | 49614s | train_loss=0.0159
472
+ step 4700000 | 49789s | train_loss=0.0165 | fid=41.6823
473
+ step 4710000 | 49887s | train_loss=0.0163
474
+ step 4720000 | 49985s | train_loss=0.0152
475
+ step 4730000 | 50083s | train_loss=0.0153
476
+ step 4740000 | 50181s | train_loss=0.0163
477
+ step 4750000 | 50280s | train_loss=0.0163
478
+ step 4760000 | 50378s | train_loss=0.0162
479
+ step 4770000 | 50474s | train_loss=0.0150
480
+ step 4780000 | 50571s | train_loss=0.0156
481
+ step 4790000 | 50652s | train_loss=0.0146
482
+ step 4800000 | 50793s | train_loss=0.0156 | fid=40.5713
483
+ step 4810000 | 50873s | train_loss=0.0150
484
+ step 4820000 | 50953s | train_loss=0.0156
485
+ step 4830000 | 51031s | train_loss=0.0156
486
+ step 4840000 | 51111s | train_loss=0.0149
487
+ step 4850000 | 51190s | train_loss=0.0143
488
+ step 4860000 | 51270s | train_loss=0.0156
489
+ step 4870000 | 51347s | train_loss=0.0151
490
+ step 4880000 | 51424s | train_loss=0.0144
491
+ step 4890000 | 51504s | train_loss=0.0154
492
+ step 4900000 | 51647s | train_loss=0.0161 | fid=41.3939
493
+ step 4910000 | 51727s | train_loss=0.0161
494
+ step 4920000 | 51807s | train_loss=0.0162
495
+ step 4930000 | 51883s | train_loss=0.0162
496
+ step 4940000 | 51962s | train_loss=0.0159
497
+ step 4950000 | 52042s | train_loss=0.0166
498
+ step 4960000 | 52122s | train_loss=0.0152
499
+ step 4970000 | 52202s | train_loss=0.0157
500
+ step 4980000 | 52247s | train_loss=0.0151
501
+ step 4990000 | 52287s | train_loss=0.0145
502
+ step 5000000 | 52362s | train_loss=0.0143 | fid=41.9502
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d140/config.json ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "vae_checkpoint": "vae_checkpoints/celeba_resnet_modernloss_d140/vae.pt",
3
+ "out": "results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d140",
4
+ "subset_path": "diagnostics/celeba_diverse_1k_mean_plus_sd/subset_indices.json",
5
+ "train_tar": "/overflow/data/cerberus/celeba_hq_256_partitioned_ws/train/celeba_hq_256-train-0000.tar",
6
+ "n_train": 1000,
7
+ "total_steps": 5000000,
8
+ "eval_interval": 10000,
9
+ "mem_interval": 100000,
10
+ "fid_interval": 100000,
11
+ "fid_n_real": 1000,
12
+ "fid_n_gen": 10000,
13
+ "n_gen_samples": 10000,
14
+ "n_sde_steps": 500,
15
+ "batch_size": 512,
16
+ "hidden": 1024,
17
+ "depth": 5,
18
+ "optimizer": "sgd",
19
+ "lr": 0.001,
20
+ "momentum": 0.8,
21
+ "t_min": 0.01,
22
+ "t_max": 3.0,
23
+ "t_eval": 0.1,
24
+ "seed": 45,
25
+ "device": null
26
+ }
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d140/metrics.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d140/runner.log ADDED
@@ -0,0 +1,502 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ FID real-side stats cached over 1000 diverse train images
2
+ step 1 | 74s | train_loss=0.9993 | fid=310.2282
3
+ step 10000 | 160s | train_loss=0.7574
4
+ step 20000 | 251s | train_loss=0.6301
5
+ step 30000 | 343s | train_loss=0.5174
6
+ step 40000 | 435s | train_loss=0.4146
7
+ step 50000 | 526s | train_loss=0.3325
8
+ step 60000 | 618s | train_loss=0.2684
9
+ step 70000 | 710s | train_loss=0.2194
10
+ step 80000 | 802s | train_loss=0.1860
11
+ step 90000 | 893s | train_loss=0.1685
12
+ step 100000 | 1057s | train_loss=0.1496 | fid=63.5919
13
+ step 110000 | 1143s | train_loss=0.1354
14
+ step 120000 | 1233s | train_loss=0.1252
15
+ step 130000 | 1324s | train_loss=0.1177
16
+ step 140000 | 1416s | train_loss=0.1115
17
+ step 150000 | 1508s | train_loss=0.1109
18
+ step 160000 | 1599s | train_loss=0.1031
19
+ step 170000 | 1690s | train_loss=0.0976
20
+ step 180000 | 1782s | train_loss=0.0945
21
+ step 190000 | 1874s | train_loss=0.0895
22
+ step 200000 | 2037s | train_loss=0.0911 | fid=69.7963
23
+ step 210000 | 2129s | train_loss=0.0855
24
+ step 220000 | 2217s | train_loss=0.0838
25
+ step 230000 | 2307s | train_loss=0.0825
26
+ step 240000 | 2399s | train_loss=0.0789
27
+ step 250000 | 2491s | train_loss=0.0783
28
+ step 260000 | 2583s | train_loss=0.0809
29
+ step 270000 | 2675s | train_loss=0.0731
30
+ step 280000 | 2767s | train_loss=0.0743
31
+ step 290000 | 2858s | train_loss=0.0730
32
+ step 300000 | 3023s | train_loss=0.0691 | fid=68.4308
33
+ step 310000 | 3115s | train_loss=0.0687
34
+ step 320000 | 3207s | train_loss=0.0702
35
+ step 330000 | 3294s | train_loss=0.0659
36
+ step 340000 | 3384s | train_loss=0.0677
37
+ step 350000 | 3477s | train_loss=0.0655
38
+ step 360000 | 3569s | train_loss=0.0647
39
+ step 370000 | 3661s | train_loss=0.0611
40
+ step 380000 | 3752s | train_loss=0.0612
41
+ step 390000 | 3844s | train_loss=0.0607
42
+ step 400000 | 4010s | train_loss=0.0614 | fid=65.0841
43
+ step 410000 | 4102s | train_loss=0.0619
44
+ step 420000 | 4194s | train_loss=0.0578
45
+ step 430000 | 4286s | train_loss=0.0580
46
+ step 440000 | 4373s | train_loss=0.0568
47
+ step 450000 | 4464s | train_loss=0.0561
48
+ step 460000 | 4556s | train_loss=0.0561
49
+ step 470000 | 4648s | train_loss=0.0572
50
+ step 480000 | 4740s | train_loss=0.0531
51
+ step 490000 | 4831s | train_loss=0.0536
52
+ step 500000 | 4996s | train_loss=0.0543 | fid=64.3177
53
+ step 510000 | 5088s | train_loss=0.0525
54
+ step 520000 | 5180s | train_loss=0.0523
55
+ step 530000 | 5271s | train_loss=0.0530
56
+ step 540000 | 5364s | train_loss=0.0505
57
+ step 550000 | 5450s | train_loss=0.0536
58
+ step 560000 | 5541s | train_loss=0.0501
59
+ step 570000 | 5633s | train_loss=0.0525
60
+ step 580000 | 5725s | train_loss=0.0503
61
+ step 590000 | 5816s | train_loss=0.0518
62
+ step 600000 | 5982s | train_loss=0.0500 | fid=62.7529
63
+ step 610000 | 6074s | train_loss=0.0488
64
+ step 620000 | 6166s | train_loss=0.0505
65
+ step 630000 | 6258s | train_loss=0.0476
66
+ step 640000 | 6350s | train_loss=0.0458
67
+ step 650000 | 6433s | train_loss=0.0480
68
+ step 660000 | 6525s | train_loss=0.0474
69
+ step 670000 | 6615s | train_loss=0.0467
70
+ step 680000 | 6707s | train_loss=0.0460
71
+ step 690000 | 6799s | train_loss=0.0452
72
+ step 700000 | 6965s | train_loss=0.0461 | fid=61.4376
73
+ step 710000 | 7057s | train_loss=0.0472
74
+ step 720000 | 7149s | train_loss=0.0459
75
+ step 730000 | 7240s | train_loss=0.0443
76
+ step 740000 | 7332s | train_loss=0.0451
77
+ step 750000 | 7424s | train_loss=0.0430
78
+ step 760000 | 7509s | train_loss=0.0427
79
+ step 770000 | 7601s | train_loss=0.0440
80
+ step 780000 | 7692s | train_loss=0.0428
81
+ step 790000 | 7784s | train_loss=0.0430
82
+ step 800000 | 7950s | train_loss=0.0434 | fid=59.8803
83
+ step 810000 | 8043s | train_loss=0.0423
84
+ step 820000 | 8134s | train_loss=0.0415
85
+ step 830000 | 8227s | train_loss=0.0435
86
+ step 840000 | 8318s | train_loss=0.0433
87
+ step 850000 | 8410s | train_loss=0.0412
88
+ step 860000 | 8503s | train_loss=0.0406
89
+ step 870000 | 8589s | train_loss=0.0414
90
+ step 880000 | 8681s | train_loss=0.0412
91
+ step 890000 | 8772s | train_loss=0.0400
92
+ step 900000 | 8996s | train_loss=0.0431 | fid=63.1167
93
+ step 910000 | 9088s | train_loss=0.0406
94
+ step 920000 | 9179s | train_loss=0.0407
95
+ step 930000 | 9271s | train_loss=0.0379
96
+ step 940000 | 9363s | train_loss=0.0412
97
+ step 950000 | 9454s | train_loss=0.0381
98
+ step 960000 | 9546s | train_loss=0.0385
99
+ step 970000 | 9634s | train_loss=0.0402
100
+ step 980000 | 9726s | train_loss=0.0392
101
+ step 990000 | 9818s | train_loss=0.0389
102
+ step 1000000 | 9982s | train_loss=0.0375 | fid=60.1986
103
+ step 1010000 | 10074s | train_loss=0.0395
104
+ step 1020000 | 10166s | train_loss=0.0382
105
+ step 1030000 | 10258s | train_loss=0.0387
106
+ step 1040000 | 10349s | train_loss=0.0396
107
+ step 1050000 | 10441s | train_loss=0.0373
108
+ step 1060000 | 10533s | train_loss=0.0389
109
+ step 1070000 | 10626s | train_loss=0.0370
110
+ step 1080000 | 10712s | train_loss=0.0373
111
+ step 1090000 | 10804s | train_loss=0.0364
112
+ step 1100000 | 10970s | train_loss=0.0375 | fid=63.4053
113
+ step 1110000 | 11062s | train_loss=0.0362
114
+ step 1120000 | 11154s | train_loss=0.0370
115
+ step 1130000 | 11246s | train_loss=0.0375
116
+ step 1140000 | 11338s | train_loss=0.0371
117
+ step 1150000 | 11430s | train_loss=0.0365
118
+ step 1160000 | 11521s | train_loss=0.0361
119
+ step 1170000 | 11613s | train_loss=0.0357
120
+ step 1180000 | 11702s | train_loss=0.0352
121
+ step 1190000 | 11793s | train_loss=0.0358
122
+ step 1200000 | 11994s | train_loss=0.0343 | fid=63.2619
123
+ step 1210000 | 12084s | train_loss=0.0352
124
+ step 1220000 | 12176s | train_loss=0.0340
125
+ step 1230000 | 12268s | train_loss=0.0346
126
+ step 1240000 | 12360s | train_loss=0.0341
127
+ step 1250000 | 12451s | train_loss=0.0369
128
+ step 1260000 | 12543s | train_loss=0.0365
129
+ step 1270000 | 12635s | train_loss=0.0340
130
+ step 1280000 | 12728s | train_loss=0.0324
131
+ step 1290000 | 12814s | train_loss=0.0352
132
+ step 1300000 | 12978s | train_loss=0.0330 | fid=62.7009
133
+ step 1310000 | 13071s | train_loss=0.0331
134
+ step 1320000 | 13161s | train_loss=0.0325
135
+ step 1330000 | 13253s | train_loss=0.0317
136
+ step 1340000 | 13345s | train_loss=0.0318
137
+ step 1350000 | 13437s | train_loss=0.0329
138
+ step 1360000 | 13529s | train_loss=0.0315
139
+ step 1370000 | 13620s | train_loss=0.0329
140
+ step 1380000 | 13712s | train_loss=0.0324
141
+ step 1390000 | 13803s | train_loss=0.0325
142
+ step 1400000 | 14005s | train_loss=0.0319 | fid=62.8718
143
+ step 1410000 | 14097s | train_loss=0.0317
144
+ step 1420000 | 14187s | train_loss=0.0321
145
+ step 1430000 | 14279s | train_loss=0.0306
146
+ step 1440000 | 14371s | train_loss=0.0315
147
+ step 1450000 | 14463s | train_loss=0.0329
148
+ step 1460000 | 14555s | train_loss=0.0307
149
+ step 1470000 | 14647s | train_loss=0.0304
150
+ step 1480000 | 14739s | train_loss=0.0311
151
+ step 1490000 | 14831s | train_loss=0.0310
152
+ step 1500000 | 15022s | train_loss=0.0291 | fid=62.5233
153
+ step 1510000 | 15114s | train_loss=0.0303
154
+ step 1520000 | 15206s | train_loss=0.0308
155
+ step 1530000 | 15296s | train_loss=0.0300
156
+ step 1540000 | 15388s | train_loss=0.0294
157
+ step 1550000 | 15479s | train_loss=0.0294
158
+ step 1560000 | 15572s | train_loss=0.0307
159
+ step 1570000 | 15664s | train_loss=0.0295
160
+ step 1580000 | 15755s | train_loss=0.0309
161
+ step 1590000 | 15847s | train_loss=0.0297
162
+ step 1600000 | 16032s | train_loss=0.0300 | fid=61.0695
163
+ step 1610000 | 16124s | train_loss=0.0297
164
+ step 1620000 | 16216s | train_loss=0.0297
165
+ step 1630000 | 16308s | train_loss=0.0295
166
+ step 1640000 | 16398s | train_loss=0.0289
167
+ step 1650000 | 16489s | train_loss=0.0282
168
+ step 1660000 | 16581s | train_loss=0.0280
169
+ step 1670000 | 16673s | train_loss=0.0274
170
+ step 1680000 | 16765s | train_loss=0.0269
171
+ step 1690000 | 16856s | train_loss=0.0286
172
+ step 1700000 | 17072s | train_loss=0.0284 | fid=62.9326
173
+ step 1710000 | 17164s | train_loss=0.0303
174
+ step 1720000 | 17256s | train_loss=0.0280
175
+ step 1730000 | 17348s | train_loss=0.0285
176
+ step 1740000 | 17438s | train_loss=0.0277
177
+ step 1750000 | 17530s | train_loss=0.0283
178
+ step 1760000 | 17622s | train_loss=0.0275
179
+ step 1770000 | 17714s | train_loss=0.0275
180
+ step 1780000 | 17806s | train_loss=0.0294
181
+ step 1790000 | 17897s | train_loss=0.0265
182
+ step 1800000 | 18112s | train_loss=0.0280 | fid=61.3208
183
+ step 1810000 | 18204s | train_loss=0.0290
184
+ step 1820000 | 18296s | train_loss=0.0274
185
+ step 1830000 | 18388s | train_loss=0.0268
186
+ step 1840000 | 18480s | train_loss=0.0268
187
+ step 1850000 | 18570s | train_loss=0.0271
188
+ step 1860000 | 18662s | train_loss=0.0269
189
+ step 1870000 | 18754s | train_loss=0.0277
190
+ step 1880000 | 18846s | train_loss=0.0258
191
+ step 1890000 | 18938s | train_loss=0.0273
192
+ step 1900000 | 19145s | train_loss=0.0265 | fid=64.8369
193
+ step 1910000 | 19237s | train_loss=0.0260
194
+ step 1920000 | 19329s | train_loss=0.0262
195
+ step 1930000 | 19421s | train_loss=0.0276
196
+ step 1940000 | 19513s | train_loss=0.0259
197
+ step 1950000 | 19603s | train_loss=0.0264
198
+ step 1960000 | 19695s | train_loss=0.0249
199
+ step 1970000 | 19787s | train_loss=0.0252
200
+ step 1980000 | 19879s | train_loss=0.0270
201
+ step 1990000 | 19970s | train_loss=0.0254
202
+ step 2000000 | 20163s | train_loss=0.0255 | fid=64.2448
203
+ step 2010000 | 20254s | train_loss=0.0259
204
+ step 2020000 | 20346s | train_loss=0.0271
205
+ step 2030000 | 20438s | train_loss=0.0265
206
+ step 2040000 | 20530s | train_loss=0.0248
207
+ step 2050000 | 20622s | train_loss=0.0259
208
+ step 2060000 | 20713s | train_loss=0.0251
209
+ step 2070000 | 20804s | train_loss=0.0252
210
+ step 2080000 | 20896s | train_loss=0.0245
211
+ step 2090000 | 20988s | train_loss=0.0240
212
+ step 2100000 | 21196s | train_loss=0.0260 | fid=61.2561
213
+ step 2110000 | 21286s | train_loss=0.0242
214
+ step 2120000 | 21378s | train_loss=0.0251
215
+ step 2130000 | 21470s | train_loss=0.0248
216
+ step 2140000 | 21562s | train_loss=0.0236
217
+ step 2150000 | 21653s | train_loss=0.0237
218
+ step 2160000 | 21745s | train_loss=0.0231
219
+ step 2170000 | 21835s | train_loss=0.0246
220
+ step 2180000 | 21927s | train_loss=0.0231
221
+ step 2190000 | 22019s | train_loss=0.0228
222
+ step 2200000 | 22219s | train_loss=0.0255 | fid=61.9191
223
+ step 2210000 | 22309s | train_loss=0.0232
224
+ step 2220000 | 22401s | train_loss=0.0235
225
+ step 2230000 | 22493s | train_loss=0.0238
226
+ step 2240000 | 22585s | train_loss=0.0242
227
+ step 2250000 | 22677s | train_loss=0.0256
228
+ step 2260000 | 22769s | train_loss=0.0244
229
+ step 2270000 | 22859s | train_loss=0.0245
230
+ step 2280000 | 22951s | train_loss=0.0231
231
+ step 2290000 | 23043s | train_loss=0.0228
232
+ step 2300000 | 23230s | train_loss=0.0230 | fid=62.5856
233
+ step 2310000 | 23321s | train_loss=0.0231
234
+ step 2320000 | 23413s | train_loss=0.0241
235
+ step 2330000 | 23505s | train_loss=0.0232
236
+ step 2340000 | 23597s | train_loss=0.0220
237
+ step 2350000 | 23689s | train_loss=0.0223
238
+ step 2360000 | 23780s | train_loss=0.0224
239
+ step 2370000 | 23873s | train_loss=0.0233
240
+ step 2380000 | 23962s | train_loss=0.0229
241
+ step 2390000 | 24054s | train_loss=0.0228
242
+ step 2400000 | 24240s | train_loss=0.0230 | fid=60.9137
243
+ step 2410000 | 24329s | train_loss=0.0232
244
+ step 2420000 | 24420s | train_loss=0.0225
245
+ step 2430000 | 24512s | train_loss=0.0226
246
+ step 2440000 | 24604s | train_loss=0.0220
247
+ step 2450000 | 24696s | train_loss=0.0215
248
+ step 2460000 | 24788s | train_loss=0.0235
249
+ step 2470000 | 24880s | train_loss=0.0224
250
+ step 2480000 | 24972s | train_loss=0.0225
251
+ step 2490000 | 25062s | train_loss=0.0246
252
+ step 2500000 | 25268s | train_loss=0.0211 | fid=61.5708
253
+ step 2510000 | 25356s | train_loss=0.0223
254
+ step 2520000 | 25447s | train_loss=0.0226
255
+ step 2530000 | 25539s | train_loss=0.0208
256
+ step 2540000 | 25631s | train_loss=0.0226
257
+ step 2550000 | 25723s | train_loss=0.0216
258
+ step 2560000 | 25815s | train_loss=0.0219
259
+ step 2570000 | 25907s | train_loss=0.0220
260
+ step 2580000 | 25999s | train_loss=0.0208
261
+ step 2590000 | 26089s | train_loss=0.0215
262
+ step 2600000 | 26300s | train_loss=0.0212 | fid=58.0981
263
+ step 2610000 | 26389s | train_loss=0.0209
264
+ step 2620000 | 26480s | train_loss=0.0217
265
+ step 2630000 | 26571s | train_loss=0.0209
266
+ step 2640000 | 26663s | train_loss=0.0206
267
+ step 2650000 | 26755s | train_loss=0.0212
268
+ step 2660000 | 26847s | train_loss=0.0206
269
+ step 2670000 | 26939s | train_loss=0.0209
270
+ step 2680000 | 27031s | train_loss=0.0229
271
+ step 2690000 | 27123s | train_loss=0.0209
272
+ step 2700000 | 27330s | train_loss=0.0210 | fid=58.7416
273
+ step 2710000 | 27419s | train_loss=0.0205
274
+ step 2720000 | 27512s | train_loss=0.0225
275
+ step 2730000 | 27602s | train_loss=0.0206
276
+ step 2740000 | 27694s | train_loss=0.0208
277
+ step 2750000 | 27786s | train_loss=0.0224
278
+ step 2760000 | 27878s | train_loss=0.0207
279
+ step 2770000 | 27970s | train_loss=0.0214
280
+ step 2780000 | 28061s | train_loss=0.0213
281
+ step 2790000 | 28153s | train_loss=0.0231
282
+ step 2800000 | 28340s | train_loss=0.0204 | fid=56.8607
283
+ step 2810000 | 28431s | train_loss=0.0202
284
+ step 2820000 | 28522s | train_loss=0.0206
285
+ step 2830000 | 28613s | train_loss=0.0201
286
+ step 2840000 | 28705s | train_loss=0.0203
287
+ step 2850000 | 28797s | train_loss=0.0197
288
+ step 2860000 | 28889s | train_loss=0.0205
289
+ step 2870000 | 28981s | train_loss=0.0206
290
+ step 2880000 | 29072s | train_loss=0.0213
291
+ step 2890000 | 29164s | train_loss=0.0209
292
+ step 2900000 | 29331s | train_loss=0.0198 | fid=53.5597
293
+ step 2910000 | 29423s | train_loss=0.0194
294
+ step 2920000 | 29512s | train_loss=0.0207
295
+ step 2930000 | 29604s | train_loss=0.0202
296
+ step 2940000 | 29694s | train_loss=0.0208
297
+ step 2950000 | 29786s | train_loss=0.0199
298
+ step 2960000 | 29878s | train_loss=0.0210
299
+ step 2970000 | 29970s | train_loss=0.0198
300
+ step 2980000 | 30061s | train_loss=0.0197
301
+ step 2990000 | 30153s | train_loss=0.0194
302
+ step 3000000 | 30374s | train_loss=0.0202 | fid=54.6823
303
+ step 3010000 | 30464s | train_loss=0.0193
304
+ step 3020000 | 30554s | train_loss=0.0196
305
+ step 3030000 | 30646s | train_loss=0.0193
306
+ step 3040000 | 30736s | train_loss=0.0200
307
+ step 3050000 | 30828s | train_loss=0.0199
308
+ step 3060000 | 30920s | train_loss=0.0203
309
+ step 3070000 | 31012s | train_loss=0.0193
310
+ step 3080000 | 31103s | train_loss=0.0203
311
+ step 3090000 | 31195s | train_loss=0.0200
312
+ step 3100000 | 31392s | train_loss=0.0189 | fid=50.3611
313
+ step 3110000 | 31485s | train_loss=0.0207
314
+ step 3120000 | 31574s | train_loss=0.0207
315
+ step 3130000 | 31664s | train_loss=0.0188
316
+ step 3140000 | 31756s | train_loss=0.0192
317
+ step 3150000 | 31846s | train_loss=0.0188
318
+ step 3160000 | 31937s | train_loss=0.0188
319
+ step 3170000 | 32029s | train_loss=0.0194
320
+ step 3180000 | 32121s | train_loss=0.0203
321
+ step 3190000 | 32213s | train_loss=0.0189
322
+ step 3200000 | 32392s | train_loss=0.0201 | fid=51.4138
323
+ step 3210000 | 32484s | train_loss=0.0189
324
+ step 3220000 | 32577s | train_loss=0.0194
325
+ step 3230000 | 32664s | train_loss=0.0199
326
+ step 3240000 | 32755s | train_loss=0.0186
327
+ step 3250000 | 32845s | train_loss=0.0203
328
+ step 3260000 | 32937s | train_loss=0.0216
329
+ step 3270000 | 33029s | train_loss=0.0185
330
+ step 3280000 | 33121s | train_loss=0.0180
331
+ step 3290000 | 33213s | train_loss=0.0175
332
+ step 3300000 | 33376s | train_loss=0.0183 | fid=50.1554
333
+ step 3310000 | 33468s | train_loss=0.0180
334
+ step 3320000 | 33560s | train_loss=0.0177
335
+ step 3330000 | 33653s | train_loss=0.0192
336
+ step 3340000 | 33740s | train_loss=0.0178
337
+ step 3350000 | 33832s | train_loss=0.0182
338
+ step 3360000 | 33922s | train_loss=0.0189
339
+ step 3370000 | 34014s | train_loss=0.0198
340
+ step 3380000 | 34106s | train_loss=0.0181
341
+ step 3390000 | 34198s | train_loss=0.0177
342
+ step 3400000 | 34380s | train_loss=0.0176 | fid=48.5737
343
+ step 3410000 | 34472s | train_loss=0.0192
344
+ step 3420000 | 34564s | train_loss=0.0184
345
+ step 3430000 | 34656s | train_loss=0.0185
346
+ step 3440000 | 34745s | train_loss=0.0186
347
+ step 3450000 | 34835s | train_loss=0.0185
348
+ step 3460000 | 34927s | train_loss=0.0181
349
+ step 3470000 | 35017s | train_loss=0.0182
350
+ step 3480000 | 35109s | train_loss=0.0181
351
+ step 3490000 | 35201s | train_loss=0.0175
352
+ step 3500000 | 35398s | train_loss=0.0182 | fid=49.4720
353
+ step 3510000 | 35490s | train_loss=0.0172
354
+ step 3520000 | 35582s | train_loss=0.0171
355
+ step 3530000 | 35673s | train_loss=0.0175
356
+ step 3540000 | 35766s | train_loss=0.0177
357
+ step 3550000 | 35854s | train_loss=0.0174
358
+ step 3560000 | 35946s | train_loss=0.0179
359
+ step 3570000 | 36036s | train_loss=0.0181
360
+ step 3580000 | 36128s | train_loss=0.0179
361
+ step 3590000 | 36220s | train_loss=0.0179
362
+ step 3600000 | 36416s | train_loss=0.0185 | fid=46.4105
363
+ step 3610000 | 36508s | train_loss=0.0176
364
+ step 3620000 | 36600s | train_loss=0.0181
365
+ step 3630000 | 36692s | train_loss=0.0182
366
+ step 3640000 | 36784s | train_loss=0.0178
367
+ step 3650000 | 36873s | train_loss=0.0188
368
+ step 3660000 | 36964s | train_loss=0.0176
369
+ step 3670000 | 37056s | train_loss=0.0184
370
+ step 3680000 | 37146s | train_loss=0.0176
371
+ step 3690000 | 37238s | train_loss=0.0182
372
+ step 3700000 | 37440s | train_loss=0.0184 | fid=43.7232
373
+ step 3710000 | 37532s | train_loss=0.0173
374
+ step 3720000 | 37623s | train_loss=0.0173
375
+ step 3730000 | 37715s | train_loss=0.0175
376
+ step 3740000 | 37807s | train_loss=0.0184
377
+ step 3750000 | 37896s | train_loss=0.0167
378
+ step 3760000 | 37989s | train_loss=0.0168
379
+ step 3770000 | 38079s | train_loss=0.0178
380
+ step 3780000 | 38169s | train_loss=0.0169
381
+ step 3790000 | 38261s | train_loss=0.0170
382
+ step 3800000 | 38463s | train_loss=0.0172 | fid=43.5575
383
+ step 3810000 | 38555s | train_loss=0.0173
384
+ step 3820000 | 38646s | train_loss=0.0169
385
+ step 3830000 | 38738s | train_loss=0.0171
386
+ step 3840000 | 38830s | train_loss=0.0172
387
+ step 3850000 | 38921s | train_loss=0.0174
388
+ step 3860000 | 39012s | train_loss=0.0163
389
+ step 3870000 | 39104s | train_loss=0.0191
390
+ step 3880000 | 39193s | train_loss=0.0165
391
+ step 3890000 | 39285s | train_loss=0.0182
392
+ step 3900000 | 39490s | train_loss=0.0171 | fid=42.6690
393
+ step 3910000 | 39581s | train_loss=0.0165
394
+ step 3920000 | 39673s | train_loss=0.0173
395
+ step 3930000 | 39765s | train_loss=0.0167
396
+ step 3940000 | 39857s | train_loss=0.0165
397
+ step 3950000 | 39949s | train_loss=0.0171
398
+ step 3960000 | 40039s | train_loss=0.0174
399
+ step 3970000 | 40131s | train_loss=0.0179
400
+ step 3980000 | 40221s | train_loss=0.0172
401
+ step 3990000 | 40312s | train_loss=0.0164
402
+ step 4000000 | 40521s | train_loss=0.0168 | fid=41.8213
403
+ step 4010000 | 40612s | train_loss=0.0168
404
+ step 4020000 | 40705s | train_loss=0.0178
405
+ step 4030000 | 40796s | train_loss=0.0178
406
+ step 4040000 | 40888s | train_loss=0.0165
407
+ step 4050000 | 40980s | train_loss=0.0169
408
+ step 4060000 | 41068s | train_loss=0.0168
409
+ step 4070000 | 41160s | train_loss=0.0163
410
+ step 4080000 | 41253s | train_loss=0.0166
411
+ step 4090000 | 41341s | train_loss=0.0156
412
+ step 4100000 | 41543s | train_loss=0.0153 | fid=40.3233
413
+ step 4110000 | 41635s | train_loss=0.0168
414
+ step 4120000 | 41727s | train_loss=0.0160
415
+ step 4130000 | 41819s | train_loss=0.0178
416
+ step 4140000 | 41910s | train_loss=0.0185
417
+ step 4150000 | 42003s | train_loss=0.0158
418
+ step 4160000 | 42092s | train_loss=0.0161
419
+ step 4170000 | 42184s | train_loss=0.0175
420
+ step 4180000 | 42276s | train_loss=0.0166
421
+ step 4190000 | 42364s | train_loss=0.0162
422
+ step 4200000 | 42551s | train_loss=0.0160 | fid=41.2841
423
+ step 4210000 | 42643s | train_loss=0.0157
424
+ step 4220000 | 42735s | train_loss=0.0165
425
+ step 4230000 | 42827s | train_loss=0.0161
426
+ step 4240000 | 42919s | train_loss=0.0162
427
+ step 4250000 | 43011s | train_loss=0.0166
428
+ step 4260000 | 43102s | train_loss=0.0162
429
+ step 4270000 | 43192s | train_loss=0.0161
430
+ step 4280000 | 43285s | train_loss=0.0176
431
+ step 4290000 | 43377s | train_loss=0.0172
432
+ step 4300000 | 43561s | train_loss=0.0167 | fid=39.8893
433
+ step 4310000 | 43653s | train_loss=0.0159
434
+ step 4320000 | 43745s | train_loss=0.0164
435
+ step 4330000 | 43837s | train_loss=0.0166
436
+ step 4340000 | 43929s | train_loss=0.0159
437
+ step 4350000 | 44021s | train_loss=0.0165
438
+ step 4360000 | 44113s | train_loss=0.0174
439
+ step 4370000 | 44202s | train_loss=0.0166
440
+ step 4380000 | 44295s | train_loss=0.0162
441
+ step 4390000 | 44387s | train_loss=0.0168
442
+ step 4400000 | 44576s | train_loss=0.0164 | fid=37.9295
443
+ step 4410000 | 44668s | train_loss=0.0155
444
+ step 4420000 | 44759s | train_loss=0.0158
445
+ step 4430000 | 44851s | train_loss=0.0156
446
+ step 4440000 | 44943s | train_loss=0.0172
447
+ step 4450000 | 45035s | train_loss=0.0161
448
+ step 4460000 | 45127s | train_loss=0.0153
449
+ step 4470000 | 45218s | train_loss=0.0162
450
+ step 4480000 | 45308s | train_loss=0.0152
451
+ step 4490000 | 45400s | train_loss=0.0160
452
+ step 4500000 | 45584s | train_loss=0.0145 | fid=39.8140
453
+ step 4510000 | 45674s | train_loss=0.0148
454
+ step 4520000 | 45766s | train_loss=0.0155
455
+ step 4530000 | 45858s | train_loss=0.0168
456
+ step 4540000 | 45950s | train_loss=0.0158
457
+ step 4550000 | 46042s | train_loss=0.0155
458
+ step 4560000 | 46134s | train_loss=0.0150
459
+ step 4570000 | 46227s | train_loss=0.0154
460
+ step 4580000 | 46315s | train_loss=0.0145
461
+ step 4590000 | 46407s | train_loss=0.0167
462
+ step 4600000 | 46610s | train_loss=0.0153 | fid=39.8361
463
+ step 4610000 | 46702s | train_loss=0.0151
464
+ step 4620000 | 46793s | train_loss=0.0171
465
+ step 4630000 | 46885s | train_loss=0.0162
466
+ step 4640000 | 46977s | train_loss=0.0160
467
+ step 4650000 | 47068s | train_loss=0.0161
468
+ step 4660000 | 47160s | train_loss=0.0159
469
+ step 4670000 | 47252s | train_loss=0.0150
470
+ step 4680000 | 47341s | train_loss=0.0155
471
+ step 4690000 | 47433s | train_loss=0.0159
472
+ step 4700000 | 47636s | train_loss=0.0154 | fid=38.0869
473
+ step 4710000 | 47726s | train_loss=0.0159
474
+ step 4720000 | 47817s | train_loss=0.0155
475
+ step 4730000 | 47909s | train_loss=0.0146
476
+ step 4740000 | 48001s | train_loss=0.0165
477
+ step 4750000 | 48092s | train_loss=0.0158
478
+ step 4760000 | 48184s | train_loss=0.0163
479
+ step 4770000 | 48276s | train_loss=0.0149
480
+ step 4780000 | 48366s | train_loss=0.0147
481
+ step 4790000 | 48457s | train_loss=0.0141
482
+ step 4800000 | 48622s | train_loss=0.0149 | fid=38.0455
483
+ step 4810000 | 48714s | train_loss=0.0152
484
+ step 4820000 | 48804s | train_loss=0.0162
485
+ step 4830000 | 48894s | train_loss=0.0144
486
+ step 4840000 | 48986s | train_loss=0.0145
487
+ step 4850000 | 49078s | train_loss=0.0138
488
+ step 4860000 | 49170s | train_loss=0.0159
489
+ step 4870000 | 49262s | train_loss=0.0148
490
+ step 4880000 | 49354s | train_loss=0.0142
491
+ step 4890000 | 49443s | train_loss=0.0152
492
+ step 4900000 | 49608s | train_loss=0.0168 | fid=36.4833
493
+ step 4910000 | 49700s | train_loss=0.0156
494
+ step 4920000 | 49790s | train_loss=0.0160
495
+ step 4930000 | 49882s | train_loss=0.0150
496
+ step 4940000 | 49973s | train_loss=0.0151
497
+ step 4950000 | 50065s | train_loss=0.0155
498
+ step 4960000 | 50156s | train_loss=0.0151
499
+ step 4970000 | 50248s | train_loss=0.0156
500
+ step 4980000 | 50340s | train_loss=0.0154
501
+ step 4990000 | 50432s | train_loss=0.0141
502
+ step 5000000 | 50588s | train_loss=0.0148 | fid=37.5819
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d160/config.json ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "vae_checkpoint": "vae_checkpoints/celeba_resnet_modernloss_d160/vae.pt",
3
+ "out": "results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d160",
4
+ "subset_path": "diagnostics/celeba_diverse_1k_mean_plus_sd/subset_indices.json",
5
+ "train_tar": "/overflow/data/cerberus/celeba_hq_256_partitioned_ws/train/celeba_hq_256-train-0000.tar",
6
+ "n_train": 1000,
7
+ "total_steps": 5000000,
8
+ "eval_interval": 10000,
9
+ "mem_interval": 100000,
10
+ "fid_interval": 100000,
11
+ "fid_n_real": 1000,
12
+ "fid_n_gen": 10000,
13
+ "n_gen_samples": 10000,
14
+ "n_sde_steps": 500,
15
+ "batch_size": 512,
16
+ "hidden": 1024,
17
+ "depth": 5,
18
+ "optimizer": "sgd",
19
+ "lr": 0.001,
20
+ "momentum": 0.8,
21
+ "t_min": 0.01,
22
+ "t_max": 3.0,
23
+ "t_eval": 0.1,
24
+ "seed": 45,
25
+ "device": null
26
+ }
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d160/metrics.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d160/runner.log ADDED
@@ -0,0 +1,502 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ FID real-side stats cached over 1000 diverse train images
2
+ step 1 | 75s | train_loss=1.0002 | fid=289.7242
3
+ step 10000 | 173s | train_loss=0.7164
4
+ step 20000 | 273s | train_loss=0.6056
5
+ step 30000 | 373s | train_loss=0.5184
6
+ step 40000 | 473s | train_loss=0.4327
7
+ step 50000 | 573s | train_loss=0.3418
8
+ step 60000 | 673s | train_loss=0.2813
9
+ step 70000 | 773s | train_loss=0.2211
10
+ step 80000 | 873s | train_loss=0.1963
11
+ step 90000 | 973s | train_loss=0.1716
12
+ step 100000 | 1143s | train_loss=0.1631 | fid=69.9215
13
+ step 110000 | 1241s | train_loss=0.1382
14
+ step 120000 | 1341s | train_loss=0.1350
15
+ step 130000 | 1440s | train_loss=0.1287
16
+ step 140000 | 1540s | train_loss=0.1315
17
+ step 150000 | 1639s | train_loss=0.1162
18
+ step 160000 | 1736s | train_loss=0.1096
19
+ step 170000 | 1835s | train_loss=0.1061
20
+ step 180000 | 1934s | train_loss=0.1051
21
+ step 190000 | 2032s | train_loss=0.0983
22
+ step 200000 | 2201s | train_loss=0.0949 | fid=87.1248
23
+ step 210000 | 2299s | train_loss=0.0988
24
+ step 220000 | 2398s | train_loss=0.0936
25
+ step 230000 | 2497s | train_loss=0.0942
26
+ step 240000 | 2596s | train_loss=0.0953
27
+ step 250000 | 2695s | train_loss=0.0911
28
+ step 260000 | 2792s | train_loss=0.0858
29
+ step 270000 | 2891s | train_loss=0.0831
30
+ step 280000 | 2990s | train_loss=0.0842
31
+ step 290000 | 3086s | train_loss=0.0821
32
+ step 300000 | 3257s | train_loss=0.0812 | fid=98.6816
33
+ step 310000 | 3354s | train_loss=0.0805
34
+ step 320000 | 3453s | train_loss=0.0775
35
+ step 330000 | 3553s | train_loss=0.0758
36
+ step 340000 | 3651s | train_loss=0.0771
37
+ step 350000 | 3750s | train_loss=0.0786
38
+ step 360000 | 3848s | train_loss=0.0735
39
+ step 370000 | 3945s | train_loss=0.0721
40
+ step 380000 | 4041s | train_loss=0.0735
41
+ step 390000 | 4141s | train_loss=0.0700
42
+ step 400000 | 4313s | train_loss=0.0739 | fid=104.1463
43
+ step 410000 | 4411s | train_loss=0.0702
44
+ step 420000 | 4511s | train_loss=0.0646
45
+ step 430000 | 4610s | train_loss=0.0693
46
+ step 440000 | 4709s | train_loss=0.0644
47
+ step 450000 | 4808s | train_loss=0.0654
48
+ step 460000 | 4906s | train_loss=0.0637
49
+ step 470000 | 5000s | train_loss=0.0673
50
+ step 480000 | 5099s | train_loss=0.0609
51
+ step 490000 | 5198s | train_loss=0.0621
52
+ step 500000 | 5371s | train_loss=0.0636 | fid=107.9804
53
+ step 510000 | 5469s | train_loss=0.0622
54
+ step 520000 | 5566s | train_loss=0.0619
55
+ step 530000 | 5666s | train_loss=0.0600
56
+ step 540000 | 5764s | train_loss=0.0587
57
+ step 550000 | 5862s | train_loss=0.0634
58
+ step 560000 | 5960s | train_loss=0.0557
59
+ step 570000 | 6055s | train_loss=0.0608
60
+ step 580000 | 6154s | train_loss=0.0573
61
+ step 590000 | 6252s | train_loss=0.0594
62
+ step 600000 | 6428s | train_loss=0.0598 | fid=109.1809
63
+ step 610000 | 6526s | train_loss=0.0563
64
+ step 620000 | 6623s | train_loss=0.0572
65
+ step 630000 | 6722s | train_loss=0.0541
66
+ step 640000 | 6820s | train_loss=0.0517
67
+ step 650000 | 6920s | train_loss=0.0555
68
+ step 660000 | 7014s | train_loss=0.0553
69
+ step 670000 | 7113s | train_loss=0.0556
70
+ step 680000 | 7211s | train_loss=0.0556
71
+ step 690000 | 7310s | train_loss=0.0510
72
+ step 700000 | 7485s | train_loss=0.0523 | fid=99.8189
73
+ step 710000 | 7581s | train_loss=0.0517
74
+ step 720000 | 7679s | train_loss=0.0549
75
+ step 730000 | 7777s | train_loss=0.0517
76
+ step 740000 | 7876s | train_loss=0.0517
77
+ step 750000 | 7972s | train_loss=0.0491
78
+ step 760000 | 8070s | train_loss=0.0473
79
+ step 770000 | 8169s | train_loss=0.0485
80
+ step 780000 | 8268s | train_loss=0.0481
81
+ step 790000 | 8366s | train_loss=0.0495
82
+ step 800000 | 8541s | train_loss=0.0498 | fid=99.4145
83
+ step 810000 | 8637s | train_loss=0.0466
84
+ step 820000 | 8736s | train_loss=0.0490
85
+ step 830000 | 8833s | train_loss=0.0485
86
+ step 840000 | 8931s | train_loss=0.0469
87
+ step 850000 | 9014s | train_loss=0.0514
88
+ step 860000 | 9111s | train_loss=0.0469
89
+ step 870000 | 9209s | train_loss=0.0481
90
+ step 880000 | 9308s | train_loss=0.0464
91
+ step 890000 | 9407s | train_loss=0.0446
92
+ step 900000 | 9581s | train_loss=0.0472 | fid=93.7293
93
+ step 910000 | 9677s | train_loss=0.0464
94
+ step 920000 | 9776s | train_loss=0.0469
95
+ step 930000 | 9874s | train_loss=0.0437
96
+ step 940000 | 9973s | train_loss=0.0455
97
+ step 950000 | 10069s | train_loss=0.0443
98
+ step 960000 | 10168s | train_loss=0.0444
99
+ step 970000 | 10267s | train_loss=0.0452
100
+ step 980000 | 10366s | train_loss=0.0436
101
+ step 990000 | 10464s | train_loss=0.0462
102
+ step 1000000 | 10639s | train_loss=0.0436 | fid=93.1966
103
+ step 1010000 | 10735s | train_loss=0.0504
104
+ step 1020000 | 10834s | train_loss=0.0411
105
+ step 1030000 | 10932s | train_loss=0.0433
106
+ step 1040000 | 11028s | train_loss=0.0436
107
+ step 1050000 | 11127s | train_loss=0.0437
108
+ step 1060000 | 11226s | train_loss=0.0476
109
+ step 1070000 | 11324s | train_loss=0.0408
110
+ step 1080000 | 11423s | train_loss=0.0428
111
+ step 1090000 | 11521s | train_loss=0.0403
112
+ step 1100000 | 11696s | train_loss=0.0436 | fid=87.4146
113
+ step 1110000 | 11793s | train_loss=0.0418
114
+ step 1120000 | 11891s | train_loss=0.0413
115
+ step 1130000 | 11982s | train_loss=0.0411
116
+ step 1140000 | 12077s | train_loss=0.0400
117
+ step 1150000 | 12176s | train_loss=0.0400
118
+ step 1160000 | 12275s | train_loss=0.0387
119
+ step 1170000 | 12373s | train_loss=0.0394
120
+ step 1180000 | 12472s | train_loss=0.0408
121
+ step 1190000 | 12570s | train_loss=0.0398
122
+ step 1200000 | 12744s | train_loss=0.0405 | fid=80.5894
123
+ step 1210000 | 12840s | train_loss=0.0399
124
+ step 1220000 | 12939s | train_loss=0.0385
125
+ step 1230000 | 13036s | train_loss=0.0408
126
+ step 1240000 | 13133s | train_loss=0.0381
127
+ step 1250000 | 13232s | train_loss=0.0417
128
+ step 1260000 | 13330s | train_loss=0.0411
129
+ step 1270000 | 13429s | train_loss=0.0404
130
+ step 1280000 | 13528s | train_loss=0.0362
131
+ step 1290000 | 13627s | train_loss=0.0394
132
+ step 1300000 | 13802s | train_loss=0.0406 | fid=77.7974
133
+ step 1310000 | 13898s | train_loss=0.0379
134
+ step 1320000 | 13988s | train_loss=0.0354
135
+ step 1330000 | 14083s | train_loss=0.0360
136
+ step 1340000 | 14182s | train_loss=0.0369
137
+ step 1350000 | 14280s | train_loss=0.0359
138
+ step 1360000 | 14378s | train_loss=0.0377
139
+ step 1370000 | 14476s | train_loss=0.0371
140
+ step 1380000 | 14575s | train_loss=0.0365
141
+ step 1390000 | 14674s | train_loss=0.0364
142
+ step 1400000 | 14849s | train_loss=0.0360 | fid=78.8131
143
+ step 1410000 | 14945s | train_loss=0.0386
144
+ step 1420000 | 15034s | train_loss=0.0387
145
+ step 1430000 | 15133s | train_loss=0.0344
146
+ step 1440000 | 15232s | train_loss=0.0353
147
+ step 1450000 | 15329s | train_loss=0.0354
148
+ step 1460000 | 15428s | train_loss=0.0341
149
+ step 1470000 | 15526s | train_loss=0.0339
150
+ step 1480000 | 15625s | train_loss=0.0385
151
+ step 1490000 | 15724s | train_loss=0.0362
152
+ step 1500000 | 15897s | train_loss=0.0360 | fid=76.4109
153
+ step 1510000 | 15992s | train_loss=0.0354
154
+ step 1520000 | 16082s | train_loss=0.0361
155
+ step 1530000 | 16181s | train_loss=0.0336
156
+ step 1540000 | 16280s | train_loss=0.0347
157
+ step 1550000 | 16376s | train_loss=0.0364
158
+ step 1560000 | 16475s | train_loss=0.0355
159
+ step 1570000 | 16572s | train_loss=0.0336
160
+ step 1580000 | 16671s | train_loss=0.0360
161
+ step 1590000 | 16770s | train_loss=0.0341
162
+ step 1600000 | 16945s | train_loss=0.0350 | fid=75.3955
163
+ step 1610000 | 17033s | train_loss=0.0350
164
+ step 1620000 | 17123s | train_loss=0.0367
165
+ step 1630000 | 17222s | train_loss=0.0381
166
+ step 1640000 | 17320s | train_loss=0.0332
167
+ step 1650000 | 17419s | train_loss=0.0330
168
+ step 1660000 | 17516s | train_loss=0.0308
169
+ step 1670000 | 17614s | train_loss=0.0312
170
+ step 1680000 | 17714s | train_loss=0.0309
171
+ step 1690000 | 17812s | train_loss=0.0322
172
+ step 1700000 | 17987s | train_loss=0.0320 | fid=70.7230
173
+ step 1710000 | 18074s | train_loss=0.0328
174
+ step 1720000 | 18165s | train_loss=0.0325
175
+ step 1730000 | 18263s | train_loss=0.0321
176
+ step 1740000 | 18361s | train_loss=0.0343
177
+ step 1750000 | 18460s | train_loss=0.0325
178
+ step 1760000 | 18558s | train_loss=0.0313
179
+ step 1770000 | 18656s | train_loss=0.0323
180
+ step 1780000 | 18754s | train_loss=0.0330
181
+ step 1790000 | 18853s | train_loss=0.0302
182
+ step 1800000 | 19028s | train_loss=0.0315 | fid=76.0825
183
+ step 1810000 | 19116s | train_loss=0.0351
184
+ step 1820000 | 19208s | train_loss=0.0321
185
+ step 1830000 | 19306s | train_loss=0.0322
186
+ step 1840000 | 19406s | train_loss=0.0297
187
+ step 1850000 | 19504s | train_loss=0.0315
188
+ step 1860000 | 19601s | train_loss=0.0299
189
+ step 1870000 | 19700s | train_loss=0.0311
190
+ step 1880000 | 19798s | train_loss=0.0295
191
+ step 1890000 | 19897s | train_loss=0.0330
192
+ step 1900000 | 20070s | train_loss=0.0314 | fid=68.7358
193
+ step 1910000 | 20159s | train_loss=0.0299
194
+ step 1920000 | 20254s | train_loss=0.0294
195
+ step 1930000 | 20353s | train_loss=0.0308
196
+ step 1940000 | 20452s | train_loss=0.0296
197
+ step 1950000 | 20550s | train_loss=0.0314
198
+ step 1960000 | 20649s | train_loss=0.0282
199
+ step 1970000 | 20745s | train_loss=0.0333
200
+ step 1980000 | 20844s | train_loss=0.0299
201
+ step 1990000 | 20943s | train_loss=0.0334
202
+ step 2000000 | 21116s | train_loss=0.0291 | fid=72.6757
203
+ step 2010000 | 21199s | train_loss=0.0297
204
+ step 2020000 | 21296s | train_loss=0.0295
205
+ step 2030000 | 21395s | train_loss=0.0290
206
+ step 2040000 | 21494s | train_loss=0.0290
207
+ step 2050000 | 21592s | train_loss=0.0292
208
+ step 2060000 | 21690s | train_loss=0.0307
209
+ step 2070000 | 21787s | train_loss=0.0290
210
+ step 2080000 | 21884s | train_loss=0.0297
211
+ step 2090000 | 21983s | train_loss=0.0321
212
+ step 2100000 | 22155s | train_loss=0.0297 | fid=68.2670
213
+ step 2110000 | 22240s | train_loss=0.0293
214
+ step 2120000 | 22337s | train_loss=0.0303
215
+ step 2130000 | 22436s | train_loss=0.0298
216
+ step 2140000 | 22535s | train_loss=0.0273
217
+ step 2150000 | 22634s | train_loss=0.0273
218
+ step 2160000 | 22733s | train_loss=0.0265
219
+ step 2170000 | 22830s | train_loss=0.0271
220
+ step 2180000 | 22929s | train_loss=0.0270
221
+ step 2190000 | 23027s | train_loss=0.0303
222
+ step 2200000 | 23200s | train_loss=0.0357 | fid=59.5094
223
+ step 2210000 | 23288s | train_loss=0.0265
224
+ step 2220000 | 23386s | train_loss=0.0271
225
+ step 2230000 | 23485s | train_loss=0.0270
226
+ step 2240000 | 23583s | train_loss=0.0315
227
+ step 2250000 | 23682s | train_loss=0.0299
228
+ step 2260000 | 23780s | train_loss=0.0280
229
+ step 2270000 | 23880s | train_loss=0.0267
230
+ step 2280000 | 23976s | train_loss=0.0268
231
+ step 2290000 | 24073s | train_loss=0.0262
232
+ step 2300000 | 24241s | train_loss=0.0264 | fid=59.9976
233
+ step 2310000 | 24339s | train_loss=0.0268
234
+ step 2320000 | 24436s | train_loss=0.0271
235
+ step 2330000 | 24535s | train_loss=0.0257
236
+ step 2340000 | 24634s | train_loss=0.0250
237
+ step 2350000 | 24733s | train_loss=0.0269
238
+ step 2360000 | 24832s | train_loss=0.0249
239
+ step 2370000 | 24931s | train_loss=0.0264
240
+ step 2380000 | 25028s | train_loss=0.0279
241
+ step 2390000 | 25125s | train_loss=0.0260
242
+ step 2400000 | 25287s | train_loss=0.0270 | fid=63.5444
243
+ step 2410000 | 25384s | train_loss=0.0267
244
+ step 2420000 | 25482s | train_loss=0.0250
245
+ step 2430000 | 25581s | train_loss=0.0245
246
+ step 2440000 | 25680s | train_loss=0.0250
247
+ step 2450000 | 25779s | train_loss=0.0235
248
+ step 2460000 | 25877s | train_loss=0.0278
249
+ step 2470000 | 25976s | train_loss=0.0265
250
+ step 2480000 | 26075s | train_loss=0.0265
251
+ step 2490000 | 26172s | train_loss=0.0267
252
+ step 2500000 | 26333s | train_loss=0.0240 | fid=58.8742
253
+ step 2510000 | 26430s | train_loss=0.0277
254
+ step 2520000 | 26527s | train_loss=0.0252
255
+ step 2530000 | 26626s | train_loss=0.0255
256
+ step 2540000 | 26725s | train_loss=0.0252
257
+ step 2550000 | 26824s | train_loss=0.0234
258
+ step 2560000 | 26923s | train_loss=0.0259
259
+ step 2570000 | 27021s | train_loss=0.0253
260
+ step 2580000 | 27120s | train_loss=0.0264
261
+ step 2590000 | 27217s | train_loss=0.0246
262
+ step 2600000 | 27377s | train_loss=0.0249 | fid=60.7260
263
+ step 2610000 | 27475s | train_loss=0.0237
264
+ step 2620000 | 27573s | train_loss=0.0271
265
+ step 2630000 | 27672s | train_loss=0.0254
266
+ step 2640000 | 27770s | train_loss=0.0236
267
+ step 2650000 | 27868s | train_loss=0.0255
268
+ step 2660000 | 27967s | train_loss=0.0235
269
+ step 2670000 | 28066s | train_loss=0.0240
270
+ step 2680000 | 28165s | train_loss=0.0291
271
+ step 2690000 | 28262s | train_loss=0.0245
272
+ step 2700000 | 28427s | train_loss=0.0228 | fid=52.6742
273
+ step 2710000 | 28524s | train_loss=0.0233
274
+ step 2720000 | 28622s | train_loss=0.0250
275
+ step 2730000 | 28721s | train_loss=0.0243
276
+ step 2740000 | 28820s | train_loss=0.0252
277
+ step 2750000 | 28919s | train_loss=0.0234
278
+ step 2760000 | 29018s | train_loss=0.0245
279
+ step 2770000 | 29116s | train_loss=0.0247
280
+ step 2780000 | 29214s | train_loss=0.0242
281
+ step 2790000 | 29313s | train_loss=0.0291
282
+ step 2800000 | 29482s | train_loss=0.0240 | fid=56.2641
283
+ step 2810000 | 29579s | train_loss=0.0233
284
+ step 2820000 | 29676s | train_loss=0.0228
285
+ step 2830000 | 29774s | train_loss=0.0228
286
+ step 2840000 | 29873s | train_loss=0.0243
287
+ step 2850000 | 29972s | train_loss=0.0227
288
+ step 2860000 | 30069s | train_loss=0.0233
289
+ step 2870000 | 30168s | train_loss=0.0245
290
+ step 2880000 | 30267s | train_loss=0.0232
291
+ step 2890000 | 30355s | train_loss=0.0229
292
+ step 2900000 | 30522s | train_loss=0.0240 | fid=56.5517
293
+ step 2910000 | 30619s | train_loss=0.0224
294
+ step 2920000 | 30716s | train_loss=0.0220
295
+ step 2930000 | 30815s | train_loss=0.0218
296
+ step 2940000 | 30913s | train_loss=0.0282
297
+ step 2950000 | 31011s | train_loss=0.0234
298
+ step 2960000 | 31110s | train_loss=0.0231
299
+ step 2970000 | 31209s | train_loss=0.0241
300
+ step 2980000 | 31308s | train_loss=0.0236
301
+ step 2990000 | 31397s | train_loss=0.0226
302
+ step 3000000 | 31569s | train_loss=0.0233 | fid=57.8875
303
+ step 3010000 | 31666s | train_loss=0.0222
304
+ step 3020000 | 31764s | train_loss=0.0215
305
+ step 3030000 | 31861s | train_loss=0.0217
306
+ step 3040000 | 31959s | train_loss=0.0237
307
+ step 3050000 | 32057s | train_loss=0.0246
308
+ step 3060000 | 32156s | train_loss=0.0230
309
+ step 3070000 | 32255s | train_loss=0.0233
310
+ step 3080000 | 32353s | train_loss=0.0238
311
+ step 3090000 | 32446s | train_loss=0.0233
312
+ step 3100000 | 32617s | train_loss=0.0218 | fid=55.5270
313
+ step 3110000 | 32713s | train_loss=0.0244
314
+ step 3120000 | 32812s | train_loss=0.0221
315
+ step 3130000 | 32908s | train_loss=0.0213
316
+ step 3140000 | 33007s | train_loss=0.0215
317
+ step 3150000 | 33106s | train_loss=0.0209
318
+ step 3160000 | 33204s | train_loss=0.0218
319
+ step 3170000 | 33303s | train_loss=0.0248
320
+ step 3180000 | 33399s | train_loss=0.0243
321
+ step 3190000 | 33497s | train_loss=0.0209
322
+ step 3200000 | 33672s | train_loss=0.0238 | fid=53.8577
323
+ step 3210000 | 33767s | train_loss=0.0217
324
+ step 3220000 | 33866s | train_loss=0.0223
325
+ step 3230000 | 33963s | train_loss=0.0222
326
+ step 3240000 | 34062s | train_loss=0.0234
327
+ step 3250000 | 34161s | train_loss=0.0254
328
+ step 3260000 | 34260s | train_loss=0.0265
329
+ step 3270000 | 34357s | train_loss=0.0230
330
+ step 3280000 | 34450s | train_loss=0.0207
331
+ step 3290000 | 34549s | train_loss=0.0203
332
+ step 3300000 | 34723s | train_loss=0.0213 | fid=51.4384
333
+ step 3310000 | 34818s | train_loss=0.0218
334
+ step 3320000 | 34917s | train_loss=0.0210
335
+ step 3330000 | 35014s | train_loss=0.0215
336
+ step 3340000 | 35113s | train_loss=0.0212
337
+ step 3350000 | 35211s | train_loss=0.0206
338
+ step 3360000 | 35309s | train_loss=0.0215
339
+ step 3370000 | 35398s | train_loss=0.0249
340
+ step 3380000 | 35497s | train_loss=0.0208
341
+ step 3390000 | 35596s | train_loss=0.0222
342
+ step 3400000 | 35769s | train_loss=0.0206 | fid=53.1034
343
+ step 3410000 | 35865s | train_loss=0.0233
344
+ step 3420000 | 35964s | train_loss=0.0235
345
+ step 3430000 | 36061s | train_loss=0.0215
346
+ step 3440000 | 36159s | train_loss=0.0214
347
+ step 3450000 | 36258s | train_loss=0.0222
348
+ step 3460000 | 36357s | train_loss=0.0220
349
+ step 3470000 | 36445s | train_loss=0.0206
350
+ step 3480000 | 36544s | train_loss=0.0204
351
+ step 3490000 | 36642s | train_loss=0.0209
352
+ step 3500000 | 36817s | train_loss=0.0215 | fid=50.7960
353
+ step 3510000 | 36915s | train_loss=0.0201
354
+ step 3520000 | 37011s | train_loss=0.0198
355
+ step 3530000 | 37109s | train_loss=0.0216
356
+ step 3540000 | 37208s | train_loss=0.0231
357
+ step 3550000 | 37306s | train_loss=0.0206
358
+ step 3560000 | 37402s | train_loss=0.0202
359
+ step 3570000 | 37492s | train_loss=0.0208
360
+ step 3580000 | 37590s | train_loss=0.0201
361
+ step 3590000 | 37688s | train_loss=0.0200
362
+ step 3600000 | 37863s | train_loss=0.0196 | fid=48.3375
363
+ step 3610000 | 37960s | train_loss=0.0221
364
+ step 3620000 | 38058s | train_loss=0.0197
365
+ step 3630000 | 38155s | train_loss=0.0207
366
+ step 3640000 | 38253s | train_loss=0.0206
367
+ step 3650000 | 38352s | train_loss=0.0211
368
+ step 3660000 | 38443s | train_loss=0.0203
369
+ step 3670000 | 38537s | train_loss=0.0247
370
+ step 3680000 | 38636s | train_loss=0.0209
371
+ step 3690000 | 38734s | train_loss=0.0201
372
+ step 3700000 | 38908s | train_loss=0.0205 | fid=50.6653
373
+ step 3710000 | 39006s | train_loss=0.0206
374
+ step 3720000 | 39104s | train_loss=0.0205
375
+ step 3730000 | 39200s | train_loss=0.0196
376
+ step 3740000 | 39299s | train_loss=0.0197
377
+ step 3750000 | 39397s | train_loss=0.0208
378
+ step 3760000 | 39485s | train_loss=0.0199
379
+ step 3770000 | 39582s | train_loss=0.0202
380
+ step 3780000 | 39681s | train_loss=0.0202
381
+ step 3790000 | 39780s | train_loss=0.0207
382
+ step 3800000 | 39953s | train_loss=0.0204 | fid=51.7770
383
+ step 3810000 | 40050s | train_loss=0.0203
384
+ step 3820000 | 40149s | train_loss=0.0191
385
+ step 3830000 | 40245s | train_loss=0.0210
386
+ step 3840000 | 40343s | train_loss=0.0196
387
+ step 3850000 | 40441s | train_loss=0.0205
388
+ step 3860000 | 40527s | train_loss=0.0187
389
+ step 3870000 | 40626s | train_loss=0.0219
390
+ step 3880000 | 40725s | train_loss=0.0198
391
+ step 3890000 | 40823s | train_loss=0.0199
392
+ step 3900000 | 40997s | train_loss=0.0213 | fid=46.5770
393
+ step 3910000 | 41093s | train_loss=0.0181
394
+ step 3920000 | 41191s | train_loss=0.0199
395
+ step 3930000 | 41289s | train_loss=0.0184
396
+ step 3940000 | 41385s | train_loss=0.0184
397
+ step 3950000 | 41484s | train_loss=0.0196
398
+ step 3960000 | 41571s | train_loss=0.0189
399
+ step 3970000 | 41670s | train_loss=0.0211
400
+ step 3980000 | 41768s | train_loss=0.0196
401
+ step 3990000 | 41866s | train_loss=0.0191
402
+ step 4000000 | 42041s | train_loss=0.0196 | fid=47.8376
403
+ step 4010000 | 42138s | train_loss=0.0194
404
+ step 4020000 | 42237s | train_loss=0.0232
405
+ step 4030000 | 42336s | train_loss=0.0200
406
+ step 4040000 | 42431s | train_loss=0.0188
407
+ step 4050000 | 42527s | train_loss=0.0196
408
+ step 4060000 | 42620s | train_loss=0.0194
409
+ step 4070000 | 42719s | train_loss=0.0206
410
+ step 4080000 | 42817s | train_loss=0.0182
411
+ step 4090000 | 42916s | train_loss=0.0185
412
+ step 4100000 | 43091s | train_loss=0.0186 | fid=48.0144
413
+ step 4110000 | 43189s | train_loss=0.0226
414
+ step 4120000 | 43288s | train_loss=0.0180
415
+ step 4130000 | 43387s | train_loss=0.0210
416
+ step 4140000 | 43483s | train_loss=0.0200
417
+ step 4150000 | 43572s | train_loss=0.0182
418
+ step 4160000 | 43671s | train_loss=0.0182
419
+ step 4170000 | 43769s | train_loss=0.0209
420
+ step 4180000 | 43868s | train_loss=0.0207
421
+ step 4190000 | 43967s | train_loss=0.0198
422
+ step 4200000 | 44142s | train_loss=0.0182 | fid=47.6259
423
+ step 4210000 | 44240s | train_loss=0.0199
424
+ step 4220000 | 44339s | train_loss=0.0201
425
+ step 4230000 | 44437s | train_loss=0.0185
426
+ step 4240000 | 44534s | train_loss=0.0195
427
+ step 4250000 | 44622s | train_loss=0.0207
428
+ step 4260000 | 44721s | train_loss=0.0205
429
+ step 4270000 | 44819s | train_loss=0.0192
430
+ step 4280000 | 44918s | train_loss=0.0192
431
+ step 4290000 | 45016s | train_loss=0.0187
432
+ step 4300000 | 45191s | train_loss=0.0188 | fid=50.9833
433
+ step 4310000 | 45288s | train_loss=0.0186
434
+ step 4320000 | 45387s | train_loss=0.0188
435
+ step 4330000 | 45486s | train_loss=0.0189
436
+ step 4340000 | 45577s | train_loss=0.0187
437
+ step 4350000 | 45672s | train_loss=0.0191
438
+ step 4360000 | 45771s | train_loss=0.0201
439
+ step 4370000 | 45870s | train_loss=0.0184
440
+ step 4380000 | 45969s | train_loss=0.0200
441
+ step 4390000 | 46068s | train_loss=0.0195
442
+ step 4400000 | 46242s | train_loss=0.0193 | fid=45.9827
443
+ step 4410000 | 46339s | train_loss=0.0218
444
+ step 4420000 | 46438s | train_loss=0.0181
445
+ step 4430000 | 46537s | train_loss=0.0186
446
+ step 4440000 | 46621s | train_loss=0.0191
447
+ step 4450000 | 46718s | train_loss=0.0191
448
+ step 4460000 | 46816s | train_loss=0.0178
449
+ step 4470000 | 46915s | train_loss=0.0187
450
+ step 4480000 | 47014s | train_loss=0.0169
451
+ step 4490000 | 47112s | train_loss=0.0186
452
+ step 4500000 | 47286s | train_loss=0.0178 | fid=46.0720
453
+ step 4510000 | 47383s | train_loss=0.0184
454
+ step 4520000 | 47482s | train_loss=0.0200
455
+ step 4530000 | 47580s | train_loss=0.0223
456
+ step 4540000 | 47666s | train_loss=0.0179
457
+ step 4550000 | 47764s | train_loss=0.0189
458
+ step 4560000 | 47861s | train_loss=0.0192
459
+ step 4570000 | 47960s | train_loss=0.0181
460
+ step 4580000 | 48059s | train_loss=0.0176
461
+ step 4590000 | 48157s | train_loss=0.0197
462
+ step 4600000 | 48332s | train_loss=0.0169 | fid=43.7691
463
+ step 4610000 | 48429s | train_loss=0.0176
464
+ step 4620000 | 48528s | train_loss=0.0189
465
+ step 4630000 | 48623s | train_loss=0.0203
466
+ step 4640000 | 48720s | train_loss=0.0195
467
+ step 4650000 | 48819s | train_loss=0.0199
468
+ step 4660000 | 48916s | train_loss=0.0194
469
+ step 4670000 | 49014s | train_loss=0.0176
470
+ step 4680000 | 49113s | train_loss=0.0182
471
+ step 4690000 | 49212s | train_loss=0.0190
472
+ step 4700000 | 49387s | train_loss=0.0188 | fid=46.6573
473
+ step 4710000 | 49484s | train_loss=0.0203
474
+ step 4720000 | 49583s | train_loss=0.0178
475
+ step 4730000 | 49679s | train_loss=0.0177
476
+ step 4740000 | 49778s | train_loss=0.0191
477
+ step 4750000 | 49875s | train_loss=0.0194
478
+ step 4760000 | 49973s | train_loss=0.0184
479
+ step 4770000 | 50073s | train_loss=0.0173
480
+ step 4780000 | 50173s | train_loss=0.0171
481
+ step 4790000 | 50272s | train_loss=0.0174
482
+ step 4800000 | 50451s | train_loss=0.0202 | fid=50.1268
483
+ step 4810000 | 50549s | train_loss=0.0172
484
+ step 4820000 | 50636s | train_loss=0.0192
485
+ step 4830000 | 50716s | train_loss=0.0167
486
+ step 4840000 | 50795s | train_loss=0.0173
487
+ step 4850000 | 50876s | train_loss=0.0167
488
+ step 4860000 | 50957s | train_loss=0.0195
489
+ step 4870000 | 51037s | train_loss=0.0172
490
+ step 4880000 | 51118s | train_loss=0.0175
491
+ step 4890000 | 51199s | train_loss=0.0185
492
+ step 4900000 | 51346s | train_loss=0.0205 | fid=49.5645
493
+ step 4910000 | 51424s | train_loss=0.0190
494
+ step 4920000 | 51505s | train_loss=0.0182
495
+ step 4930000 | 51586s | train_loss=0.0198
496
+ step 4940000 | 51665s | train_loss=0.0167
497
+ step 4950000 | 51746s | train_loss=0.0169
498
+ step 4960000 | 51826s | train_loss=0.0177
499
+ step 4970000 | 51905s | train_loss=0.0187
500
+ step 4980000 | 51986s | train_loss=0.0172
501
+ step 4990000 | 52067s | train_loss=0.0179
502
+ step 5000000 | 52213s | train_loss=0.0171 | fid=47.3853
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d180/config.json ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "vae_checkpoint": "vae_checkpoints/celeba_resnet_modernloss_d180/vae.pt",
3
+ "out": "results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d180",
4
+ "subset_path": "diagnostics/celeba_diverse_1k_mean_plus_sd/subset_indices.json",
5
+ "train_tar": "/overflow/data/cerberus/celeba_hq_256_partitioned_ws/train/celeba_hq_256-train-0000.tar",
6
+ "n_train": 1000,
7
+ "total_steps": 5000000,
8
+ "eval_interval": 10000,
9
+ "mem_interval": 100000,
10
+ "fid_interval": 100000,
11
+ "fid_n_real": 1000,
12
+ "fid_n_gen": 10000,
13
+ "n_gen_samples": 10000,
14
+ "n_sde_steps": 500,
15
+ "batch_size": 512,
16
+ "hidden": 1024,
17
+ "depth": 5,
18
+ "optimizer": "sgd",
19
+ "lr": 0.001,
20
+ "momentum": 0.8,
21
+ "t_min": 0.01,
22
+ "t_max": 3.0,
23
+ "t_eval": 0.1,
24
+ "seed": 45,
25
+ "device": null
26
+ }
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d180/metrics.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d180/runner.log ADDED
@@ -0,0 +1,502 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ FID real-side stats cached over 1000 diverse train images
2
+ step 1 | 77s | train_loss=1.0006 | fid=345.6789
3
+ step 10000 | 176s | train_loss=0.6679
4
+ step 20000 | 275s | train_loss=0.5572
5
+ step 30000 | 374s | train_loss=0.4889
6
+ step 40000 | 473s | train_loss=0.4154
7
+ step 50000 | 573s | train_loss=0.3410
8
+ step 60000 | 672s | train_loss=0.2825
9
+ step 70000 | 772s | train_loss=0.2377
10
+ step 80000 | 871s | train_loss=0.2026
11
+ step 90000 | 970s | train_loss=0.1829
12
+ step 100000 | 1142s | train_loss=0.1621 | fid=65.9873
13
+ step 110000 | 1237s | train_loss=0.1508
14
+ step 120000 | 1337s | train_loss=0.1404
15
+ step 130000 | 1436s | train_loss=0.1315
16
+ step 140000 | 1535s | train_loss=0.1309
17
+ step 150000 | 1634s | train_loss=0.1233
18
+ step 160000 | 1731s | train_loss=0.1203
19
+ step 170000 | 1830s | train_loss=0.1141
20
+ step 180000 | 1929s | train_loss=0.1144
21
+ step 190000 | 2027s | train_loss=0.1093
22
+ step 200000 | 2199s | train_loss=0.1068 | fid=75.1920
23
+ step 210000 | 2296s | train_loss=0.1006
24
+ step 220000 | 2395s | train_loss=0.0996
25
+ step 230000 | 2494s | train_loss=0.0976
26
+ step 240000 | 2593s | train_loss=0.0992
27
+ step 250000 | 2691s | train_loss=0.0979
28
+ step 260000 | 2789s | train_loss=0.0948
29
+ step 270000 | 2887s | train_loss=0.0953
30
+ step 280000 | 2986s | train_loss=0.0890
31
+ step 290000 | 3081s | train_loss=0.0888
32
+ step 300000 | 3256s | train_loss=0.0880 | fid=77.0286
33
+ step 310000 | 3353s | train_loss=0.0863
34
+ step 320000 | 3452s | train_loss=0.0883
35
+ step 330000 | 3551s | train_loss=0.0800
36
+ step 340000 | 3649s | train_loss=0.0821
37
+ step 350000 | 3748s | train_loss=0.0794
38
+ step 360000 | 3846s | train_loss=0.0796
39
+ step 370000 | 3944s | train_loss=0.0795
40
+ step 380000 | 4039s | train_loss=0.0780
41
+ step 390000 | 4138s | train_loss=0.0760
42
+ step 400000 | 4313s | train_loss=0.0793 | fid=84.1725
43
+ step 410000 | 4411s | train_loss=0.0730
44
+ step 420000 | 4510s | train_loss=0.0721
45
+ step 430000 | 4609s | train_loss=0.0717
46
+ step 440000 | 4708s | train_loss=0.0724
47
+ step 450000 | 4807s | train_loss=0.0715
48
+ step 460000 | 4904s | train_loss=0.0698
49
+ step 470000 | 4999s | train_loss=0.0685
50
+ step 480000 | 5098s | train_loss=0.0695
51
+ step 490000 | 5197s | train_loss=0.0683
52
+ step 500000 | 5372s | train_loss=0.0682 | fid=87.8656
53
+ step 510000 | 5471s | train_loss=0.0705
54
+ step 520000 | 5568s | train_loss=0.0682
55
+ step 530000 | 5667s | train_loss=0.0642
56
+ step 540000 | 5766s | train_loss=0.0626
57
+ step 550000 | 5864s | train_loss=0.0658
58
+ step 560000 | 5962s | train_loss=0.0672
59
+ step 570000 | 6057s | train_loss=0.0654
60
+ step 580000 | 6156s | train_loss=0.0643
61
+ step 590000 | 6255s | train_loss=0.0630
62
+ step 600000 | 6430s | train_loss=0.0622 | fid=89.4142
63
+ step 610000 | 6529s | train_loss=0.0615
64
+ step 620000 | 6626s | train_loss=0.0654
65
+ step 630000 | 6725s | train_loss=0.0584
66
+ step 640000 | 6823s | train_loss=0.0622
67
+ step 650000 | 6922s | train_loss=0.0589
68
+ step 660000 | 7017s | train_loss=0.0585
69
+ step 670000 | 7116s | train_loss=0.0589
70
+ step 680000 | 7214s | train_loss=0.0598
71
+ step 690000 | 7312s | train_loss=0.0559
72
+ step 700000 | 7488s | train_loss=0.0547 | fid=88.0518
73
+ step 710000 | 7586s | train_loss=0.0562
74
+ step 720000 | 7683s | train_loss=0.0556
75
+ step 730000 | 7782s | train_loss=0.0539
76
+ step 740000 | 7881s | train_loss=0.0556
77
+ step 750000 | 7976s | train_loss=0.0539
78
+ step 760000 | 8075s | train_loss=0.0528
79
+ step 770000 | 8174s | train_loss=0.0539
80
+ step 780000 | 8272s | train_loss=0.0548
81
+ step 790000 | 8370s | train_loss=0.0517
82
+ step 800000 | 8546s | train_loss=0.0519 | fid=87.5763
83
+ step 810000 | 8642s | train_loss=0.0508
84
+ step 820000 | 8741s | train_loss=0.0526
85
+ step 830000 | 8838s | train_loss=0.0543
86
+ step 840000 | 8934s | train_loss=0.0526
87
+ step 850000 | 9019s | train_loss=0.0537
88
+ step 860000 | 9115s | train_loss=0.0505
89
+ step 870000 | 9214s | train_loss=0.0548
90
+ step 880000 | 9313s | train_loss=0.0496
91
+ step 890000 | 9411s | train_loss=0.0497
92
+ step 900000 | 9585s | train_loss=0.0521 | fid=88.6370
93
+ step 910000 | 9682s | train_loss=0.0509
94
+ step 920000 | 9781s | train_loss=0.0513
95
+ step 930000 | 9879s | train_loss=0.0501
96
+ step 940000 | 9977s | train_loss=0.0474
97
+ step 950000 | 10074s | train_loss=0.0471
98
+ step 960000 | 10173s | train_loss=0.0471
99
+ step 970000 | 10271s | train_loss=0.0473
100
+ step 980000 | 10370s | train_loss=0.0505
101
+ step 990000 | 10468s | train_loss=0.0464
102
+ step 1000000 | 10642s | train_loss=0.0478 | fid=77.9495
103
+ step 1010000 | 10739s | train_loss=0.0477
104
+ step 1020000 | 10838s | train_loss=0.0473
105
+ step 1030000 | 10935s | train_loss=0.0471
106
+ step 1040000 | 11031s | train_loss=0.0462
107
+ step 1050000 | 11130s | train_loss=0.0483
108
+ step 1060000 | 11229s | train_loss=0.0449
109
+ step 1070000 | 11328s | train_loss=0.0507
110
+ step 1080000 | 11426s | train_loss=0.0442
111
+ step 1090000 | 11525s | train_loss=0.0471
112
+ step 1100000 | 11701s | train_loss=0.0440 | fid=77.0646
113
+ step 1110000 | 11798s | train_loss=0.0481
114
+ step 1120000 | 11897s | train_loss=0.0438
115
+ step 1130000 | 11987s | train_loss=0.0434
116
+ step 1140000 | 12082s | train_loss=0.0443
117
+ step 1150000 | 12181s | train_loss=0.0471
118
+ step 1160000 | 12280s | train_loss=0.0447
119
+ step 1170000 | 12378s | train_loss=0.0457
120
+ step 1180000 | 12476s | train_loss=0.0463
121
+ step 1190000 | 12575s | train_loss=0.0422
122
+ step 1200000 | 12750s | train_loss=0.0437 | fid=75.0992
123
+ step 1210000 | 12847s | train_loss=0.0430
124
+ step 1220000 | 12945s | train_loss=0.0425
125
+ step 1230000 | 13042s | train_loss=0.0431
126
+ step 1240000 | 13139s | train_loss=0.0479
127
+ step 1250000 | 13239s | train_loss=0.0406
128
+ step 1260000 | 13337s | train_loss=0.0439
129
+ step 1270000 | 13436s | train_loss=0.0399
130
+ step 1280000 | 13534s | train_loss=0.0426
131
+ step 1290000 | 13633s | train_loss=0.0444
132
+ step 1300000 | 13808s | train_loss=0.0391 | fid=66.5521
133
+ step 1310000 | 13905s | train_loss=0.0416
134
+ step 1320000 | 13994s | train_loss=0.0414
135
+ step 1330000 | 14091s | train_loss=0.0408
136
+ step 1340000 | 14188s | train_loss=0.0445
137
+ step 1350000 | 14287s | train_loss=0.0421
138
+ step 1360000 | 14384s | train_loss=0.0398
139
+ step 1370000 | 14483s | train_loss=0.0378
140
+ step 1380000 | 14582s | train_loss=0.0395
141
+ step 1390000 | 14680s | train_loss=0.0402
142
+ step 1400000 | 14855s | train_loss=0.0404 | fid=63.2339
143
+ step 1410000 | 14952s | train_loss=0.0382
144
+ step 1420000 | 15041s | train_loss=0.0388
145
+ step 1430000 | 15140s | train_loss=0.0425
146
+ step 1440000 | 15239s | train_loss=0.0397
147
+ step 1450000 | 15336s | train_loss=0.0403
148
+ step 1460000 | 15434s | train_loss=0.0390
149
+ step 1470000 | 15532s | train_loss=0.0391
150
+ step 1480000 | 15631s | train_loss=0.0368
151
+ step 1490000 | 15730s | train_loss=0.0408
152
+ step 1500000 | 15903s | train_loss=0.0371 | fid=64.6179
153
+ step 1510000 | 16000s | train_loss=0.0372
154
+ step 1520000 | 16090s | train_loss=0.0373
155
+ step 1530000 | 16189s | train_loss=0.0380
156
+ step 1540000 | 16287s | train_loss=0.0352
157
+ step 1550000 | 16383s | train_loss=0.0401
158
+ step 1560000 | 16482s | train_loss=0.0386
159
+ step 1570000 | 16579s | train_loss=0.0355
160
+ step 1580000 | 16678s | train_loss=0.0396
161
+ step 1590000 | 16776s | train_loss=0.0348
162
+ step 1600000 | 16951s | train_loss=0.0373 | fid=61.6693
163
+ step 1610000 | 17040s | train_loss=0.0374
164
+ step 1620000 | 17131s | train_loss=0.0377
165
+ step 1630000 | 17229s | train_loss=0.0361
166
+ step 1640000 | 17328s | train_loss=0.0345
167
+ step 1650000 | 17426s | train_loss=0.0371
168
+ step 1660000 | 17523s | train_loss=0.0382
169
+ step 1670000 | 17621s | train_loss=0.0371
170
+ step 1680000 | 17720s | train_loss=0.0357
171
+ step 1690000 | 17818s | train_loss=0.0345
172
+ step 1700000 | 17993s | train_loss=0.0363 | fid=60.2383
173
+ step 1710000 | 18081s | train_loss=0.0352
174
+ step 1720000 | 18172s | train_loss=0.0378
175
+ step 1730000 | 18270s | train_loss=0.0354
176
+ step 1740000 | 18368s | train_loss=0.0385
177
+ step 1750000 | 18467s | train_loss=0.0346
178
+ step 1760000 | 18564s | train_loss=0.0345
179
+ step 1770000 | 18662s | train_loss=0.0348
180
+ step 1780000 | 18761s | train_loss=0.0363
181
+ step 1790000 | 18859s | train_loss=0.0331
182
+ step 1800000 | 19034s | train_loss=0.0356 | fid=57.2581
183
+ step 1810000 | 19123s | train_loss=0.0345
184
+ step 1820000 | 19216s | train_loss=0.0326
185
+ step 1830000 | 19315s | train_loss=0.0352
186
+ step 1840000 | 19413s | train_loss=0.0331
187
+ step 1850000 | 19511s | train_loss=0.0323
188
+ step 1860000 | 19609s | train_loss=0.0320
189
+ step 1870000 | 19707s | train_loss=0.0336
190
+ step 1880000 | 19805s | train_loss=0.0348
191
+ step 1890000 | 19904s | train_loss=0.0338
192
+ step 1900000 | 20078s | train_loss=0.0341 | fid=54.9191
193
+ step 1910000 | 20166s | train_loss=0.0339
194
+ step 1920000 | 20263s | train_loss=0.0350
195
+ step 1930000 | 20362s | train_loss=0.0333
196
+ step 1940000 | 20461s | train_loss=0.0328
197
+ step 1950000 | 20559s | train_loss=0.0375
198
+ step 1960000 | 20658s | train_loss=0.0314
199
+ step 1970000 | 20754s | train_loss=0.0353
200
+ step 1980000 | 20852s | train_loss=0.0326
201
+ step 1990000 | 20951s | train_loss=0.0332
202
+ step 2000000 | 21125s | train_loss=0.0340 | fid=54.1497
203
+ step 2010000 | 21210s | train_loss=0.0307
204
+ step 2020000 | 21307s | train_loss=0.0317
205
+ step 2030000 | 21406s | train_loss=0.0309
206
+ step 2040000 | 21505s | train_loss=0.0332
207
+ step 2050000 | 21602s | train_loss=0.0319
208
+ step 2060000 | 21701s | train_loss=0.0323
209
+ step 2070000 | 21797s | train_loss=0.0358
210
+ step 2080000 | 21895s | train_loss=0.0302
211
+ step 2090000 | 21994s | train_loss=0.0332
212
+ step 2100000 | 22167s | train_loss=0.0325 | fid=54.9512
213
+ step 2110000 | 22254s | train_loss=0.0322
214
+ step 2120000 | 22350s | train_loss=0.0304
215
+ step 2130000 | 22449s | train_loss=0.0306
216
+ step 2140000 | 22548s | train_loss=0.0331
217
+ step 2150000 | 22647s | train_loss=0.0322
218
+ step 2160000 | 22746s | train_loss=0.0310
219
+ step 2170000 | 22842s | train_loss=0.0331
220
+ step 2180000 | 22941s | train_loss=0.0297
221
+ step 2190000 | 23039s | train_loss=0.0316
222
+ step 2200000 | 23209s | train_loss=0.0303 | fid=54.6915
223
+ step 2210000 | 23301s | train_loss=0.0303
224
+ step 2220000 | 23398s | train_loss=0.0333
225
+ step 2230000 | 23497s | train_loss=0.0283
226
+ step 2240000 | 23596s | train_loss=0.0288
227
+ step 2250000 | 23694s | train_loss=0.0297
228
+ step 2260000 | 23792s | train_loss=0.0295
229
+ step 2270000 | 23891s | train_loss=0.0306
230
+ step 2280000 | 23987s | train_loss=0.0290
231
+ step 2290000 | 24085s | train_loss=0.0342
232
+ step 2300000 | 24250s | train_loss=0.0283 | fid=52.2723
233
+ step 2310000 | 24349s | train_loss=0.0303
234
+ step 2320000 | 24447s | train_loss=0.0302
235
+ step 2330000 | 24545s | train_loss=0.0289
236
+ step 2340000 | 24644s | train_loss=0.0292
237
+ step 2350000 | 24743s | train_loss=0.0302
238
+ step 2360000 | 24842s | train_loss=0.0308
239
+ step 2370000 | 24941s | train_loss=0.0287
240
+ step 2380000 | 25037s | train_loss=0.0320
241
+ step 2390000 | 25135s | train_loss=0.0280
242
+ step 2400000 | 25296s | train_loss=0.0286 | fid=52.9045
243
+ step 2410000 | 25395s | train_loss=0.0272
244
+ step 2420000 | 25492s | train_loss=0.0289
245
+ step 2430000 | 25591s | train_loss=0.0285
246
+ step 2440000 | 25690s | train_loss=0.0265
247
+ step 2450000 | 25788s | train_loss=0.0288
248
+ step 2460000 | 25887s | train_loss=0.0307
249
+ step 2470000 | 25986s | train_loss=0.0279
250
+ step 2480000 | 26083s | train_loss=0.0300
251
+ step 2490000 | 26181s | train_loss=0.0285
252
+ step 2500000 | 26342s | train_loss=0.0309 | fid=52.4323
253
+ step 2510000 | 26441s | train_loss=0.0274
254
+ step 2520000 | 26538s | train_loss=0.0303
255
+ step 2530000 | 26637s | train_loss=0.0280
256
+ step 2540000 | 26735s | train_loss=0.0287
257
+ step 2550000 | 26834s | train_loss=0.0287
258
+ step 2560000 | 26933s | train_loss=0.0299
259
+ step 2570000 | 27032s | train_loss=0.0272
260
+ step 2580000 | 27130s | train_loss=0.0270
261
+ step 2590000 | 27227s | train_loss=0.0291
262
+ step 2600000 | 27387s | train_loss=0.0277 | fid=54.4033
263
+ step 2610000 | 27486s | train_loss=0.0269
264
+ step 2620000 | 27583s | train_loss=0.0285
265
+ step 2630000 | 27682s | train_loss=0.0263
266
+ step 2640000 | 27780s | train_loss=0.0274
267
+ step 2650000 | 27879s | train_loss=0.0289
268
+ step 2660000 | 27977s | train_loss=0.0298
269
+ step 2670000 | 28076s | train_loss=0.0270
270
+ step 2680000 | 28175s | train_loss=0.0279
271
+ step 2690000 | 28271s | train_loss=0.0285
272
+ step 2700000 | 28437s | train_loss=0.0263 | fid=56.2437
273
+ step 2710000 | 28536s | train_loss=0.0300
274
+ step 2720000 | 28633s | train_loss=0.0276
275
+ step 2730000 | 28732s | train_loss=0.0276
276
+ step 2740000 | 28831s | train_loss=0.0265
277
+ step 2750000 | 28930s | train_loss=0.0262
278
+ step 2760000 | 29029s | train_loss=0.0278
279
+ step 2770000 | 29127s | train_loss=0.0246
280
+ step 2780000 | 29225s | train_loss=0.0256
281
+ step 2790000 | 29323s | train_loss=0.0310
282
+ step 2800000 | 29493s | train_loss=0.0257 | fid=51.7445
283
+ step 2810000 | 29591s | train_loss=0.0272
284
+ step 2820000 | 29689s | train_loss=0.0289
285
+ step 2830000 | 29787s | train_loss=0.0268
286
+ step 2840000 | 29886s | train_loss=0.0247
287
+ step 2850000 | 29984s | train_loss=0.0257
288
+ step 2860000 | 30082s | train_loss=0.0279
289
+ step 2870000 | 30181s | train_loss=0.0291
290
+ step 2880000 | 30280s | train_loss=0.0321
291
+ step 2890000 | 30365s | train_loss=0.0247
292
+ step 2900000 | 30536s | train_loss=0.0244 | fid=49.6990
293
+ step 2910000 | 30635s | train_loss=0.0312
294
+ step 2920000 | 30732s | train_loss=0.0252
295
+ step 2930000 | 30831s | train_loss=0.0266
296
+ step 2940000 | 30929s | train_loss=0.0247
297
+ step 2950000 | 31027s | train_loss=0.0257
298
+ step 2960000 | 31126s | train_loss=0.0251
299
+ step 2970000 | 31225s | train_loss=0.0258
300
+ step 2980000 | 31324s | train_loss=0.0258
301
+ step 2990000 | 31412s | train_loss=0.0281
302
+ step 3000000 | 31586s | train_loss=0.0261 | fid=49.6994
303
+ step 3010000 | 31685s | train_loss=0.0271
304
+ step 3020000 | 31781s | train_loss=0.0255
305
+ step 3030000 | 31879s | train_loss=0.0239
306
+ step 3040000 | 31977s | train_loss=0.0255
307
+ step 3050000 | 32076s | train_loss=0.0258
308
+ step 3060000 | 32175s | train_loss=0.0280
309
+ step 3070000 | 32273s | train_loss=0.0243
310
+ step 3080000 | 32371s | train_loss=0.0258
311
+ step 3090000 | 32464s | train_loss=0.0228
312
+ step 3100000 | 32638s | train_loss=0.0247 | fid=50.2273
313
+ step 3110000 | 32735s | train_loss=0.0257
314
+ step 3120000 | 32832s | train_loss=0.0294
315
+ step 3130000 | 32930s | train_loss=0.0256
316
+ step 3140000 | 33029s | train_loss=0.0237
317
+ step 3150000 | 33128s | train_loss=0.0246
318
+ step 3160000 | 33226s | train_loss=0.0259
319
+ step 3170000 | 33325s | train_loss=0.0240
320
+ step 3180000 | 33421s | train_loss=0.0252
321
+ step 3190000 | 33519s | train_loss=0.0251
322
+ step 3200000 | 33692s | train_loss=0.0245 | fid=50.5826
323
+ step 3210000 | 33791s | train_loss=0.0233
324
+ step 3220000 | 33890s | train_loss=0.0261
325
+ step 3230000 | 33987s | train_loss=0.0252
326
+ step 3240000 | 34086s | train_loss=0.0259
327
+ step 3250000 | 34184s | train_loss=0.0243
328
+ step 3260000 | 34282s | train_loss=0.0231
329
+ step 3270000 | 34375s | train_loss=0.0251
330
+ step 3280000 | 34472s | train_loss=0.0262
331
+ step 3290000 | 34571s | train_loss=0.0231
332
+ step 3300000 | 34745s | train_loss=0.0251 | fid=52.0869
333
+ step 3310000 | 34842s | train_loss=0.0229
334
+ step 3320000 | 34941s | train_loss=0.0268
335
+ step 3330000 | 35038s | train_loss=0.0242
336
+ step 3340000 | 35136s | train_loss=0.0259
337
+ step 3350000 | 35235s | train_loss=0.0224
338
+ step 3360000 | 35333s | train_loss=0.0235
339
+ step 3370000 | 35422s | train_loss=0.0247
340
+ step 3380000 | 35521s | train_loss=0.0234
341
+ step 3390000 | 35619s | train_loss=0.0237
342
+ step 3400000 | 35794s | train_loss=0.0266 | fid=45.3250
343
+ step 3410000 | 35891s | train_loss=0.0245
344
+ step 3420000 | 35990s | train_loss=0.0232
345
+ step 3430000 | 36086s | train_loss=0.0257
346
+ step 3440000 | 36185s | train_loss=0.0219
347
+ step 3450000 | 36284s | train_loss=0.0246
348
+ step 3460000 | 36381s | train_loss=0.0246
349
+ step 3470000 | 36471s | train_loss=0.0241
350
+ step 3480000 | 36570s | train_loss=0.0230
351
+ step 3490000 | 36668s | train_loss=0.0246
352
+ step 3500000 | 36843s | train_loss=0.0239 | fid=52.9903
353
+ step 3510000 | 36940s | train_loss=0.0227
354
+ step 3520000 | 37039s | train_loss=0.0228
355
+ step 3530000 | 37136s | train_loss=0.0229
356
+ step 3540000 | 37235s | train_loss=0.0255
357
+ step 3550000 | 37333s | train_loss=0.0237
358
+ step 3560000 | 37424s | train_loss=0.0226
359
+ step 3570000 | 37519s | train_loss=0.0251
360
+ step 3580000 | 37617s | train_loss=0.0230
361
+ step 3590000 | 37716s | train_loss=0.0244
362
+ step 3600000 | 37889s | train_loss=0.0233 | fid=49.9103
363
+ step 3610000 | 37988s | train_loss=0.0233
364
+ step 3620000 | 38085s | train_loss=0.0220
365
+ step 3630000 | 38182s | train_loss=0.0227
366
+ step 3640000 | 38281s | train_loss=0.0238
367
+ step 3650000 | 38379s | train_loss=0.0234
368
+ step 3660000 | 38466s | train_loss=0.0231
369
+ step 3670000 | 38565s | train_loss=0.0234
370
+ step 3680000 | 38663s | train_loss=0.0236
371
+ step 3690000 | 38761s | train_loss=0.0225
372
+ step 3700000 | 38935s | train_loss=0.0237 | fid=47.6845
373
+ step 3710000 | 39034s | train_loss=0.0234
374
+ step 3720000 | 39132s | train_loss=0.0228
375
+ step 3730000 | 39229s | train_loss=0.0218
376
+ step 3740000 | 39327s | train_loss=0.0209
377
+ step 3750000 | 39425s | train_loss=0.0228
378
+ step 3760000 | 39512s | train_loss=0.0214
379
+ step 3770000 | 39611s | train_loss=0.0222
380
+ step 3780000 | 39710s | train_loss=0.0243
381
+ step 3790000 | 39808s | train_loss=0.0232
382
+ step 3800000 | 39983s | train_loss=0.0227 | fid=46.7846
383
+ step 3810000 | 40082s | train_loss=0.0209
384
+ step 3820000 | 40181s | train_loss=0.0219
385
+ step 3830000 | 40276s | train_loss=0.0232
386
+ step 3840000 | 40375s | train_loss=0.0218
387
+ step 3850000 | 40472s | train_loss=0.0236
388
+ step 3860000 | 40559s | train_loss=0.0216
389
+ step 3870000 | 40657s | train_loss=0.0224
390
+ step 3880000 | 40756s | train_loss=0.0227
391
+ step 3890000 | 40854s | train_loss=0.0222
392
+ step 3900000 | 41027s | train_loss=0.0236 | fid=43.8041
393
+ step 3910000 | 41126s | train_loss=0.0219
394
+ step 3920000 | 41224s | train_loss=0.0210
395
+ step 3930000 | 41319s | train_loss=0.0218
396
+ step 3940000 | 41418s | train_loss=0.0202
397
+ step 3950000 | 41512s | train_loss=0.0212
398
+ step 3960000 | 41603s | train_loss=0.0208
399
+ step 3970000 | 41702s | train_loss=0.0212
400
+ step 3980000 | 41800s | train_loss=0.0234
401
+ step 3990000 | 41898s | train_loss=0.0198
402
+ step 4000000 | 42073s | train_loss=0.0223 | fid=51.3707
403
+ step 4010000 | 42171s | train_loss=0.0217
404
+ step 4020000 | 42270s | train_loss=0.0224
405
+ step 4030000 | 42364s | train_loss=0.0209
406
+ step 4040000 | 42463s | train_loss=0.0212
407
+ step 4050000 | 42554s | train_loss=0.0224
408
+ step 4060000 | 42652s | train_loss=0.0209
409
+ step 4070000 | 42750s | train_loss=0.0221
410
+ step 4080000 | 42849s | train_loss=0.0225
411
+ step 4090000 | 42948s | train_loss=0.0224
412
+ step 4100000 | 43122s | train_loss=0.0198 | fid=46.7357
413
+ step 4110000 | 43220s | train_loss=0.0221
414
+ step 4120000 | 43319s | train_loss=0.0231
415
+ step 4130000 | 43419s | train_loss=0.0206
416
+ step 4140000 | 43514s | train_loss=0.0221
417
+ step 4150000 | 43603s | train_loss=0.0215
418
+ step 4160000 | 43702s | train_loss=0.0215
419
+ step 4170000 | 43800s | train_loss=0.0217
420
+ step 4180000 | 43899s | train_loss=0.0211
421
+ step 4190000 | 43998s | train_loss=0.0227
422
+ step 4200000 | 44173s | train_loss=0.0198 | fid=47.7708
423
+ step 4210000 | 44272s | train_loss=0.0203
424
+ step 4220000 | 44371s | train_loss=0.0200
425
+ step 4230000 | 44470s | train_loss=0.0195
426
+ step 4240000 | 44559s | train_loss=0.0198
427
+ step 4250000 | 44654s | train_loss=0.0221
428
+ step 4260000 | 44752s | train_loss=0.0209
429
+ step 4270000 | 44850s | train_loss=0.0208
430
+ step 4280000 | 44949s | train_loss=0.0206
431
+ step 4290000 | 45048s | train_loss=0.0210
432
+ step 4300000 | 45223s | train_loss=0.0213 | fid=50.6745
433
+ step 4310000 | 45322s | train_loss=0.0201
434
+ step 4320000 | 45420s | train_loss=0.0203
435
+ step 4330000 | 45519s | train_loss=0.0212
436
+ step 4340000 | 45609s | train_loss=0.0195
437
+ step 4350000 | 45705s | train_loss=0.0239
438
+ step 4360000 | 45804s | train_loss=0.0199
439
+ step 4370000 | 45903s | train_loss=0.0200
440
+ step 4380000 | 46001s | train_loss=0.0206
441
+ step 4390000 | 46100s | train_loss=0.0205
442
+ step 4400000 | 46275s | train_loss=0.0204 | fid=47.4342
443
+ step 4410000 | 46374s | train_loss=0.0198
444
+ step 4420000 | 46472s | train_loss=0.0197
445
+ step 4430000 | 46569s | train_loss=0.0218
446
+ step 4440000 | 46655s | train_loss=0.0242
447
+ step 4450000 | 46752s | train_loss=0.0208
448
+ step 4460000 | 46850s | train_loss=0.0213
449
+ step 4470000 | 46949s | train_loss=0.0210
450
+ step 4480000 | 47048s | train_loss=0.0208
451
+ step 4490000 | 47146s | train_loss=0.0194
452
+ step 4500000 | 47321s | train_loss=0.0193 | fid=48.1663
453
+ step 4510000 | 47420s | train_loss=0.0192
454
+ step 4520000 | 47519s | train_loss=0.0193
455
+ step 4530000 | 47611s | train_loss=0.0192
456
+ step 4540000 | 47703s | train_loss=0.0227
457
+ step 4550000 | 47800s | train_loss=0.0197
458
+ step 4560000 | 47899s | train_loss=0.0205
459
+ step 4570000 | 47998s | train_loss=0.0201
460
+ step 4580000 | 48096s | train_loss=0.0211
461
+ step 4590000 | 48194s | train_loss=0.0208
462
+ step 4600000 | 48369s | train_loss=0.0193 | fid=45.0917
463
+ step 4610000 | 48467s | train_loss=0.0204
464
+ step 4620000 | 48566s | train_loss=0.0185
465
+ step 4630000 | 48662s | train_loss=0.0189
466
+ step 4640000 | 48758s | train_loss=0.0191
467
+ step 4650000 | 48857s | train_loss=0.0182
468
+ step 4660000 | 48954s | train_loss=0.0201
469
+ step 4670000 | 49053s | train_loss=0.0213
470
+ step 4680000 | 49151s | train_loss=0.0203
471
+ step 4690000 | 49250s | train_loss=0.0200
472
+ step 4700000 | 49425s | train_loss=0.0201 | fid=45.4369
473
+ step 4710000 | 49524s | train_loss=0.0195
474
+ step 4720000 | 49620s | train_loss=0.0205
475
+ step 4730000 | 49718s | train_loss=0.0197
476
+ step 4740000 | 49815s | train_loss=0.0192
477
+ step 4750000 | 49915s | train_loss=0.0182
478
+ step 4760000 | 50013s | train_loss=0.0221
479
+ step 4770000 | 50112s | train_loss=0.0187
480
+ step 4780000 | 50211s | train_loss=0.0182
481
+ step 4790000 | 50310s | train_loss=0.0201
482
+ step 4800000 | 50484s | train_loss=0.0195 | fid=45.5588
483
+ step 4810000 | 50583s | train_loss=0.0218
484
+ step 4820000 | 50663s | train_loss=0.0230
485
+ step 4830000 | 50743s | train_loss=0.0197
486
+ step 4840000 | 50822s | train_loss=0.0197
487
+ step 4850000 | 50903s | train_loss=0.0189
488
+ step 4860000 | 50982s | train_loss=0.0198
489
+ step 4870000 | 51062s | train_loss=0.0209
490
+ step 4880000 | 51143s | train_loss=0.0198
491
+ step 4890000 | 51223s | train_loss=0.0205
492
+ step 4900000 | 51365s | train_loss=0.0193 | fid=46.0616
493
+ step 4910000 | 51445s | train_loss=0.0188
494
+ step 4920000 | 51525s | train_loss=0.0205
495
+ step 4930000 | 51606s | train_loss=0.0198
496
+ step 4940000 | 51685s | train_loss=0.0212
497
+ step 4950000 | 51765s | train_loss=0.0190
498
+ step 4960000 | 51842s | train_loss=0.0201
499
+ step 4970000 | 51922s | train_loss=0.0197
500
+ step 4980000 | 52003s | train_loss=0.0198
501
+ step 4990000 | 52084s | train_loss=0.0197
502
+ step 5000000 | 52224s | train_loss=0.0198 | fid=48.9713
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d20/config.json ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "vae_checkpoint": "vae_checkpoints/celeba_resnet_modernloss_d20/vae.pt",
3
+ "out": "results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d20",
4
+ "subset_path": "diagnostics/celeba_diverse_1k_mean_plus_sd/subset_indices.json",
5
+ "train_tar": "/overflow/data/cerberus/celeba_hq_256_partitioned_ws/train/celeba_hq_256-train-0000.tar",
6
+ "n_train": 1000,
7
+ "total_steps": 5000000,
8
+ "eval_interval": 10000,
9
+ "mem_interval": 100000,
10
+ "fid_interval": 100000,
11
+ "fid_n_real": 1000,
12
+ "fid_n_gen": 10000,
13
+ "n_gen_samples": 10000,
14
+ "n_sde_steps": 500,
15
+ "batch_size": 512,
16
+ "hidden": 1024,
17
+ "depth": 5,
18
+ "optimizer": "sgd",
19
+ "lr": 0.001,
20
+ "momentum": 0.8,
21
+ "t_min": 0.01,
22
+ "t_max": 3.0,
23
+ "t_eval": 0.1,
24
+ "seed": 45,
25
+ "device": null
26
+ }
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d20/metrics.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d20/runner.log ADDED
@@ -0,0 +1,502 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ FID real-side stats cached over 1000 diverse train images
2
+ step 1 | 72s | train_loss=1.0034 | fid=225.4772
3
+ step 10000 | 163s | train_loss=0.8829
4
+ step 20000 | 256s | train_loss=0.8783
5
+ step 30000 | 349s | train_loss=0.8744
6
+ step 40000 | 442s | train_loss=0.8688
7
+ step 50000 | 535s | train_loss=0.8628
8
+ step 60000 | 628s | train_loss=0.8554
9
+ step 70000 | 720s | train_loss=0.8444
10
+ step 80000 | 813s | train_loss=0.8241
11
+ step 90000 | 906s | train_loss=0.7909
12
+ step 100000 | 1068s | train_loss=0.7462 | fid=71.4024
13
+ step 110000 | 1157s | train_loss=0.6856
14
+ step 120000 | 1250s | train_loss=0.6048
15
+ step 130000 | 1343s | train_loss=0.5183
16
+ step 140000 | 1435s | train_loss=0.4334
17
+ step 150000 | 1529s | train_loss=0.3631
18
+ step 160000 | 1622s | train_loss=0.3097
19
+ step 170000 | 1715s | train_loss=0.2704
20
+ step 180000 | 1809s | train_loss=0.2443
21
+ step 190000 | 1902s | train_loss=0.2212
22
+ step 200000 | 2066s | train_loss=0.2073 | fid=71.2299
23
+ step 210000 | 2157s | train_loss=0.1953
24
+ step 220000 | 2251s | train_loss=0.1827
25
+ step 230000 | 2344s | train_loss=0.1742
26
+ step 240000 | 2438s | train_loss=0.1652
27
+ step 250000 | 2531s | train_loss=0.1578
28
+ step 260000 | 2625s | train_loss=0.1521
29
+ step 270000 | 2719s | train_loss=0.1455
30
+ step 280000 | 2812s | train_loss=0.1390
31
+ step 290000 | 2904s | train_loss=0.1357
32
+ step 300000 | 3070s | train_loss=0.1305 | fid=71.6060
33
+ step 310000 | 3163s | train_loss=0.1267
34
+ step 320000 | 3255s | train_loss=0.1222
35
+ step 330000 | 3349s | train_loss=0.1195
36
+ step 340000 | 3442s | train_loss=0.1157
37
+ step 350000 | 3536s | train_loss=0.1117
38
+ step 360000 | 3630s | train_loss=0.1098
39
+ step 370000 | 3723s | train_loss=0.1069
40
+ step 380000 | 3816s | train_loss=0.1040
41
+ step 390000 | 3908s | train_loss=0.1011
42
+ step 400000 | 4074s | train_loss=0.0995 | fid=69.0447
43
+ step 410000 | 4166s | train_loss=0.0968
44
+ step 420000 | 4259s | train_loss=0.0949
45
+ step 430000 | 4352s | train_loss=0.0926
46
+ step 440000 | 4446s | train_loss=0.0902
47
+ step 450000 | 4540s | train_loss=0.0886
48
+ step 460000 | 4633s | train_loss=0.0870
49
+ step 470000 | 4727s | train_loss=0.0864
50
+ step 480000 | 4817s | train_loss=0.0842
51
+ step 490000 | 4910s | train_loss=0.0827
52
+ step 500000 | 5077s | train_loss=0.0811 | fid=68.6743
53
+ step 510000 | 5168s | train_loss=0.0791
54
+ step 520000 | 5262s | train_loss=0.0785
55
+ step 530000 | 5354s | train_loss=0.0773
56
+ step 540000 | 5447s | train_loss=0.0768
57
+ step 550000 | 5541s | train_loss=0.0752
58
+ step 560000 | 5634s | train_loss=0.0743
59
+ step 570000 | 5727s | train_loss=0.0727
60
+ step 580000 | 5818s | train_loss=0.0708
61
+ step 590000 | 5912s | train_loss=0.0711
62
+ step 600000 | 6077s | train_loss=0.0697 | fid=69.2410
63
+ step 610000 | 6169s | train_loss=0.0683
64
+ step 620000 | 6263s | train_loss=0.0677
65
+ step 630000 | 6355s | train_loss=0.0668
66
+ step 640000 | 6449s | train_loss=0.0665
67
+ step 650000 | 6542s | train_loss=0.0649
68
+ step 660000 | 6636s | train_loss=0.0648
69
+ step 670000 | 6726s | train_loss=0.0636
70
+ step 680000 | 6819s | train_loss=0.0627
71
+ step 690000 | 6913s | train_loss=0.0624
72
+ step 700000 | 7078s | train_loss=0.0614 | fid=69.3402
73
+ step 710000 | 7170s | train_loss=0.0613
74
+ step 720000 | 7263s | train_loss=0.0601
75
+ step 730000 | 7356s | train_loss=0.0601
76
+ step 740000 | 7449s | train_loss=0.0587
77
+ step 750000 | 7543s | train_loss=0.0587
78
+ step 760000 | 7633s | train_loss=0.0586
79
+ step 770000 | 7727s | train_loss=0.0569
80
+ step 780000 | 7820s | train_loss=0.0564
81
+ step 790000 | 7914s | train_loss=0.0562
82
+ step 800000 | 8081s | train_loss=0.0554 | fid=69.2346
83
+ step 810000 | 8173s | train_loss=0.0548
84
+ step 820000 | 8267s | train_loss=0.0552
85
+ step 830000 | 8361s | train_loss=0.0540
86
+ step 840000 | 8453s | train_loss=0.0536
87
+ step 850000 | 8547s | train_loss=0.0531
88
+ step 860000 | 8637s | train_loss=0.0521
89
+ step 870000 | 8730s | train_loss=0.0527
90
+ step 880000 | 8824s | train_loss=0.0517
91
+ step 890000 | 8917s | train_loss=0.0511
92
+ step 900000 | 9083s | train_loss=0.0507 | fid=68.3387
93
+ step 910000 | 9175s | train_loss=0.0503
94
+ step 920000 | 9269s | train_loss=0.0497
95
+ step 930000 | 9363s | train_loss=0.0494
96
+ step 940000 | 9455s | train_loss=0.0490
97
+ step 950000 | 9546s | train_loss=0.0492
98
+ step 960000 | 9639s | train_loss=0.0495
99
+ step 970000 | 9733s | train_loss=0.0479
100
+ step 980000 | 9826s | train_loss=0.0470
101
+ step 990000 | 9920s | train_loss=0.0473
102
+ step 1000000 | 10086s | train_loss=0.0467 | fid=68.6119
103
+ step 1010000 | 10178s | train_loss=0.0463
104
+ step 1020000 | 10272s | train_loss=0.0458
105
+ step 1030000 | 10365s | train_loss=0.0454
106
+ step 1040000 | 10459s | train_loss=0.0454
107
+ step 1050000 | 10547s | train_loss=0.0459
108
+ step 1060000 | 10640s | train_loss=0.0443
109
+ step 1070000 | 10734s | train_loss=0.0444
110
+ step 1080000 | 10827s | train_loss=0.0444
111
+ step 1090000 | 10920s | train_loss=0.0436
112
+ step 1100000 | 11087s | train_loss=0.0433 | fid=68.1778
113
+ step 1110000 | 11179s | train_loss=0.0432
114
+ step 1120000 | 11272s | train_loss=0.0430
115
+ step 1130000 | 11365s | train_loss=0.0431
116
+ step 1140000 | 11456s | train_loss=0.0423
117
+ step 1150000 | 11548s | train_loss=0.0421
118
+ step 1160000 | 11642s | train_loss=0.0421
119
+ step 1170000 | 11736s | train_loss=0.0425
120
+ step 1180000 | 11829s | train_loss=0.0417
121
+ step 1190000 | 11923s | train_loss=0.0410
122
+ step 1200000 | 12089s | train_loss=0.0411 | fid=67.8462
123
+ step 1210000 | 12181s | train_loss=0.0406
124
+ step 1220000 | 12275s | train_loss=0.0404
125
+ step 1230000 | 12365s | train_loss=0.0403
126
+ step 1240000 | 12458s | train_loss=0.0397
127
+ step 1250000 | 12551s | train_loss=0.0403
128
+ step 1260000 | 12644s | train_loss=0.0396
129
+ step 1270000 | 12737s | train_loss=0.0390
130
+ step 1280000 | 12831s | train_loss=0.0386
131
+ step 1290000 | 12924s | train_loss=0.0387
132
+ step 1300000 | 13090s | train_loss=0.0388 | fid=68.3523
133
+ step 1310000 | 13182s | train_loss=0.0384
134
+ step 1320000 | 13275s | train_loss=0.0383
135
+ step 1330000 | 13366s | train_loss=0.0384
136
+ step 1340000 | 13459s | train_loss=0.0378
137
+ step 1350000 | 13553s | train_loss=0.0376
138
+ step 1360000 | 13645s | train_loss=0.0375
139
+ step 1370000 | 13739s | train_loss=0.0374
140
+ step 1380000 | 13832s | train_loss=0.0368
141
+ step 1390000 | 13925s | train_loss=0.0369
142
+ step 1400000 | 14090s | train_loss=0.0367 | fid=67.9202
143
+ step 1410000 | 14182s | train_loss=0.0363
144
+ step 1420000 | 14272s | train_loss=0.0361
145
+ step 1430000 | 14365s | train_loss=0.0363
146
+ step 1440000 | 14458s | train_loss=0.0355
147
+ step 1450000 | 14551s | train_loss=0.0356
148
+ step 1460000 | 14644s | train_loss=0.0356
149
+ step 1470000 | 14737s | train_loss=0.0351
150
+ step 1480000 | 14831s | train_loss=0.0349
151
+ step 1490000 | 14924s | train_loss=0.0349
152
+ step 1500000 | 15089s | train_loss=0.0346 | fid=66.7456
153
+ step 1510000 | 15182s | train_loss=0.0344
154
+ step 1520000 | 15270s | train_loss=0.0340
155
+ step 1530000 | 15363s | train_loss=0.0343
156
+ step 1540000 | 15456s | train_loss=0.0344
157
+ step 1550000 | 15550s | train_loss=0.0338
158
+ step 1560000 | 15644s | train_loss=0.0338
159
+ step 1570000 | 15736s | train_loss=0.0336
160
+ step 1580000 | 15829s | train_loss=0.0335
161
+ step 1590000 | 15923s | train_loss=0.0332
162
+ step 1600000 | 16089s | train_loss=0.0329 | fid=68.4116
163
+ step 1610000 | 16174s | train_loss=0.0331
164
+ step 1620000 | 16266s | train_loss=0.0330
165
+ step 1630000 | 16360s | train_loss=0.0324
166
+ step 1640000 | 16453s | train_loss=0.0329
167
+ step 1650000 | 16547s | train_loss=0.0322
168
+ step 1660000 | 16641s | train_loss=0.0324
169
+ step 1670000 | 16733s | train_loss=0.0324
170
+ step 1680000 | 16827s | train_loss=0.0323
171
+ step 1690000 | 16921s | train_loss=0.0318
172
+ step 1700000 | 17087s | train_loss=0.0322 | fid=67.7311
173
+ step 1710000 | 17170s | train_loss=0.0313
174
+ step 1720000 | 17263s | train_loss=0.0321
175
+ step 1730000 | 17357s | train_loss=0.0313
176
+ step 1740000 | 17450s | train_loss=0.0309
177
+ step 1750000 | 17543s | train_loss=0.0312
178
+ step 1760000 | 17637s | train_loss=0.0307
179
+ step 1770000 | 17729s | train_loss=0.0304
180
+ step 1780000 | 17823s | train_loss=0.0309
181
+ step 1790000 | 17916s | train_loss=0.0302
182
+ step 1800000 | 18082s | train_loss=0.0303 | fid=68.1400
183
+ step 1810000 | 18167s | train_loss=0.0302
184
+ step 1820000 | 18259s | train_loss=0.0302
185
+ step 1830000 | 18352s | train_loss=0.0299
186
+ step 1840000 | 18446s | train_loss=0.0295
187
+ step 1850000 | 18539s | train_loss=0.0301
188
+ step 1860000 | 18633s | train_loss=0.0301
189
+ step 1870000 | 18727s | train_loss=0.0295
190
+ step 1880000 | 18819s | train_loss=0.0298
191
+ step 1890000 | 18912s | train_loss=0.0294
192
+ step 1900000 | 19077s | train_loss=0.0296 | fid=68.1098
193
+ step 1910000 | 19165s | train_loss=0.0291
194
+ step 1920000 | 19256s | train_loss=0.0288
195
+ step 1930000 | 19350s | train_loss=0.0293
196
+ step 1940000 | 19443s | train_loss=0.0289
197
+ step 1950000 | 19537s | train_loss=0.0284
198
+ step 1960000 | 19631s | train_loss=0.0284
199
+ step 1970000 | 19724s | train_loss=0.0284
200
+ step 1980000 | 19816s | train_loss=0.0284
201
+ step 1990000 | 19909s | train_loss=0.0281
202
+ step 2000000 | 20069s | train_loss=0.0283 | fid=67.1521
203
+ step 2010000 | 20163s | train_loss=0.0279
204
+ step 2020000 | 20254s | train_loss=0.0280
205
+ step 2030000 | 20348s | train_loss=0.0279
206
+ step 2040000 | 20441s | train_loss=0.0277
207
+ step 2050000 | 20535s | train_loss=0.0277
208
+ step 2060000 | 20628s | train_loss=0.0275
209
+ step 2070000 | 20722s | train_loss=0.0273
210
+ step 2080000 | 20815s | train_loss=0.0277
211
+ step 2090000 | 20909s | train_loss=0.0272
212
+ step 2100000 | 21067s | train_loss=0.0271 | fid=66.8267
213
+ step 2110000 | 21161s | train_loss=0.0270
214
+ step 2120000 | 21253s | train_loss=0.0273
215
+ step 2130000 | 21347s | train_loss=0.0271
216
+ step 2140000 | 21442s | train_loss=0.0271
217
+ step 2150000 | 21536s | train_loss=0.0271
218
+ step 2160000 | 21631s | train_loss=0.0262
219
+ step 2170000 | 21725s | train_loss=0.0266
220
+ step 2180000 | 21820s | train_loss=0.0266
221
+ step 2190000 | 21913s | train_loss=0.0265
222
+ step 2200000 | 22063s | train_loss=0.0264 | fid=67.1224
223
+ step 2210000 | 22158s | train_loss=0.0260
224
+ step 2220000 | 22252s | train_loss=0.0259
225
+ step 2230000 | 22346s | train_loss=0.0262
226
+ step 2240000 | 22441s | train_loss=0.0261
227
+ step 2250000 | 22535s | train_loss=0.0258
228
+ step 2260000 | 22630s | train_loss=0.0259
229
+ step 2270000 | 22723s | train_loss=0.0255
230
+ step 2280000 | 22818s | train_loss=0.0255
231
+ step 2290000 | 22912s | train_loss=0.0256
232
+ step 2300000 | 23064s | train_loss=0.0251 | fid=67.9543
233
+ step 2310000 | 23159s | train_loss=0.0253
234
+ step 2320000 | 23252s | train_loss=0.0252
235
+ step 2330000 | 23347s | train_loss=0.0251
236
+ step 2340000 | 23441s | train_loss=0.0252
237
+ step 2350000 | 23535s | train_loss=0.0256
238
+ step 2360000 | 23629s | train_loss=0.0253
239
+ step 2370000 | 23722s | train_loss=0.0246
240
+ step 2380000 | 23815s | train_loss=0.0248
241
+ step 2390000 | 23909s | train_loss=0.0245
242
+ step 2400000 | 24066s | train_loss=0.0245 | fid=67.3231
243
+ step 2410000 | 24160s | train_loss=0.0246
244
+ step 2420000 | 24252s | train_loss=0.0249
245
+ step 2430000 | 24345s | train_loss=0.0241
246
+ step 2440000 | 24439s | train_loss=0.0240
247
+ step 2450000 | 24532s | train_loss=0.0243
248
+ step 2460000 | 24625s | train_loss=0.0243
249
+ step 2470000 | 24719s | train_loss=0.0240
250
+ step 2480000 | 24812s | train_loss=0.0241
251
+ step 2490000 | 24905s | train_loss=0.0237
252
+ step 2500000 | 25058s | train_loss=0.0242 | fid=68.3144
253
+ step 2510000 | 25152s | train_loss=0.0235
254
+ step 2520000 | 25244s | train_loss=0.0233
255
+ step 2530000 | 25337s | train_loss=0.0239
256
+ step 2540000 | 25430s | train_loss=0.0233
257
+ step 2550000 | 25524s | train_loss=0.0234
258
+ step 2560000 | 25617s | train_loss=0.0239
259
+ step 2570000 | 25711s | train_loss=0.0234
260
+ step 2580000 | 25804s | train_loss=0.0231
261
+ step 2590000 | 25898s | train_loss=0.0234
262
+ step 2600000 | 26051s | train_loss=0.0231 | fid=68.3706
263
+ step 2610000 | 26145s | train_loss=0.0228
264
+ step 2620000 | 26237s | train_loss=0.0236
265
+ step 2630000 | 26330s | train_loss=0.0233
266
+ step 2640000 | 26424s | train_loss=0.0225
267
+ step 2650000 | 26517s | train_loss=0.0224
268
+ step 2660000 | 26611s | train_loss=0.0227
269
+ step 2670000 | 26704s | train_loss=0.0229
270
+ step 2680000 | 26796s | train_loss=0.0226
271
+ step 2690000 | 26890s | train_loss=0.0225
272
+ step 2700000 | 27045s | train_loss=0.0224 | fid=67.6936
273
+ step 2710000 | 27139s | train_loss=0.0227
274
+ step 2720000 | 27231s | train_loss=0.0225
275
+ step 2730000 | 27325s | train_loss=0.0223
276
+ step 2740000 | 27419s | train_loss=0.0221
277
+ step 2750000 | 27512s | train_loss=0.0219
278
+ step 2760000 | 27605s | train_loss=0.0222
279
+ step 2770000 | 27698s | train_loss=0.0220
280
+ step 2780000 | 27791s | train_loss=0.0215
281
+ step 2790000 | 27885s | train_loss=0.0218
282
+ step 2800000 | 28039s | train_loss=0.0221 | fid=67.6056
283
+ step 2810000 | 28131s | train_loss=0.0219
284
+ step 2820000 | 28223s | train_loss=0.0221
285
+ step 2830000 | 28317s | train_loss=0.0217
286
+ step 2840000 | 28410s | train_loss=0.0215
287
+ step 2850000 | 28503s | train_loss=0.0215
288
+ step 2860000 | 28596s | train_loss=0.0212
289
+ step 2870000 | 28689s | train_loss=0.0214
290
+ step 2880000 | 28783s | train_loss=0.0213
291
+ step 2890000 | 28875s | train_loss=0.0215
292
+ step 2900000 | 29032s | train_loss=0.0214 | fid=67.3933
293
+ step 2910000 | 29124s | train_loss=0.0213
294
+ step 2920000 | 29216s | train_loss=0.0214
295
+ step 2930000 | 29310s | train_loss=0.0210
296
+ step 2940000 | 29404s | train_loss=0.0209
297
+ step 2950000 | 29497s | train_loss=0.0210
298
+ step 2960000 | 29591s | train_loss=0.0208
299
+ step 2970000 | 29684s | train_loss=0.0212
300
+ step 2980000 | 29779s | train_loss=0.0207
301
+ step 2990000 | 29869s | train_loss=0.0210
302
+ step 3000000 | 30036s | train_loss=0.0210 | fid=66.8639
303
+ step 3010000 | 30130s | train_loss=0.0205
304
+ step 3020000 | 30223s | train_loss=0.0208
305
+ step 3030000 | 30318s | train_loss=0.0207
306
+ step 3040000 | 30413s | train_loss=0.0208
307
+ step 3050000 | 30508s | train_loss=0.0204
308
+ step 3060000 | 30602s | train_loss=0.0206
309
+ step 3070000 | 30697s | train_loss=0.0207
310
+ step 3080000 | 30792s | train_loss=0.0203
311
+ step 3090000 | 30874s | train_loss=0.0203
312
+ step 3100000 | 31042s | train_loss=0.0202 | fid=67.6991
313
+ step 3110000 | 31138s | train_loss=0.0202
314
+ step 3120000 | 31230s | train_loss=0.0201
315
+ step 3130000 | 31324s | train_loss=0.0198
316
+ step 3140000 | 31419s | train_loss=0.0202
317
+ step 3150000 | 31514s | train_loss=0.0202
318
+ step 3160000 | 31608s | train_loss=0.0199
319
+ step 3170000 | 31703s | train_loss=0.0195
320
+ step 3180000 | 31797s | train_loss=0.0196
321
+ step 3190000 | 31885s | train_loss=0.0197
322
+ step 3200000 | 32052s | train_loss=0.0199 | fid=67.4498
323
+ step 3210000 | 32148s | train_loss=0.0196
324
+ step 3220000 | 32239s | train_loss=0.0194
325
+ step 3230000 | 32334s | train_loss=0.0195
326
+ step 3240000 | 32428s | train_loss=0.0197
327
+ step 3250000 | 32523s | train_loss=0.0193
328
+ step 3260000 | 32618s | train_loss=0.0195
329
+ step 3270000 | 32713s | train_loss=0.0197
330
+ step 3280000 | 32800s | train_loss=0.0192
331
+ step 3290000 | 32887s | train_loss=0.0195
332
+ step 3300000 | 33092s | train_loss=0.0195 | fid=68.0454
333
+ step 3310000 | 33184s | train_loss=0.0192
334
+ step 3320000 | 33279s | train_loss=0.0193
335
+ step 3330000 | 33374s | train_loss=0.0193
336
+ step 3340000 | 33468s | train_loss=0.0189
337
+ step 3350000 | 33562s | train_loss=0.0189
338
+ step 3360000 | 33657s | train_loss=0.0189
339
+ step 3370000 | 33752s | train_loss=0.0189
340
+ step 3380000 | 33831s | train_loss=0.0185
341
+ step 3390000 | 33923s | train_loss=0.0193
342
+ step 3400000 | 34118s | train_loss=0.0187 | fid=67.1745
343
+ step 3410000 | 34211s | train_loss=0.0189
344
+ step 3420000 | 34305s | train_loss=0.0188
345
+ step 3430000 | 34400s | train_loss=0.0184
346
+ step 3440000 | 34494s | train_loss=0.0189
347
+ step 3450000 | 34589s | train_loss=0.0187
348
+ step 3460000 | 34684s | train_loss=0.0183
349
+ step 3470000 | 34776s | train_loss=0.0185
350
+ step 3480000 | 34856s | train_loss=0.0181
351
+ step 3490000 | 34949s | train_loss=0.0185
352
+ step 3500000 | 35147s | train_loss=0.0182 | fid=67.2195
353
+ step 3510000 | 35240s | train_loss=0.0183
354
+ step 3520000 | 35333s | train_loss=0.0182
355
+ step 3530000 | 35428s | train_loss=0.0185
356
+ step 3540000 | 35522s | train_loss=0.0185
357
+ step 3550000 | 35617s | train_loss=0.0181
358
+ step 3560000 | 35712s | train_loss=0.0183
359
+ step 3570000 | 35799s | train_loss=0.0184
360
+ step 3580000 | 35885s | train_loss=0.0182
361
+ step 3590000 | 35978s | train_loss=0.0180
362
+ step 3600000 | 36202s | train_loss=0.0180 | fid=67.6915
363
+ step 3610000 | 36297s | train_loss=0.0180
364
+ step 3620000 | 36390s | train_loss=0.0181
365
+ step 3630000 | 36485s | train_loss=0.0177
366
+ step 3640000 | 36581s | train_loss=0.0177
367
+ step 3650000 | 36676s | train_loss=0.0179
368
+ step 3660000 | 36769s | train_loss=0.0179
369
+ step 3670000 | 36853s | train_loss=0.0180
370
+ step 3680000 | 36947s | train_loss=0.0177
371
+ step 3690000 | 37040s | train_loss=0.0179
372
+ step 3700000 | 37242s | train_loss=0.0176 | fid=67.6529
373
+ step 3710000 | 37338s | train_loss=0.0176
374
+ step 3720000 | 37431s | train_loss=0.0174
375
+ step 3730000 | 37526s | train_loss=0.0174
376
+ step 3740000 | 37621s | train_loss=0.0175
377
+ step 3750000 | 37716s | train_loss=0.0175
378
+ step 3760000 | 37797s | train_loss=0.0174
379
+ step 3770000 | 37885s | train_loss=0.0171
380
+ step 3780000 | 37979s | train_loss=0.0172
381
+ step 3790000 | 38073s | train_loss=0.0173
382
+ step 3800000 | 38258s | train_loss=0.0173 | fid=67.3541
383
+ step 3810000 | 38352s | train_loss=0.0169
384
+ step 3820000 | 38446s | train_loss=0.0170
385
+ step 3830000 | 38541s | train_loss=0.0168
386
+ step 3840000 | 38636s | train_loss=0.0171
387
+ step 3850000 | 38731s | train_loss=0.0173
388
+ step 3860000 | 38815s | train_loss=0.0169
389
+ step 3870000 | 38908s | train_loss=0.0169
390
+ step 3880000 | 39002s | train_loss=0.0168
391
+ step 3890000 | 39096s | train_loss=0.0169
392
+ step 3900000 | 39263s | train_loss=0.0169 | fid=67.2425
393
+ step 3910000 | 39358s | train_loss=0.0165
394
+ step 3920000 | 39451s | train_loss=0.0168
395
+ step 3930000 | 39546s | train_loss=0.0169
396
+ step 3940000 | 39641s | train_loss=0.0168
397
+ step 3950000 | 39736s | train_loss=0.0167
398
+ step 3960000 | 39813s | train_loss=0.0166
399
+ step 3970000 | 39905s | train_loss=0.0165
400
+ step 3980000 | 39999s | train_loss=0.0163
401
+ step 3990000 | 40094s | train_loss=0.0166
402
+ step 4000000 | 40260s | train_loss=0.0164 | fid=67.5797
403
+ step 4010000 | 40355s | train_loss=0.0164
404
+ step 4020000 | 40451s | train_loss=0.0160
405
+ step 4030000 | 40545s | train_loss=0.0164
406
+ step 4040000 | 40640s | train_loss=0.0164
407
+ step 4050000 | 40735s | train_loss=0.0162
408
+ step 4060000 | 40817s | train_loss=0.0163
409
+ step 4070000 | 40904s | train_loss=0.0160
410
+ step 4080000 | 40999s | train_loss=0.0161
411
+ step 4090000 | 41094s | train_loss=0.0160
412
+ step 4100000 | 41260s | train_loss=0.0161 | fid=67.3912
413
+ step 4110000 | 41356s | train_loss=0.0159
414
+ step 4120000 | 41451s | train_loss=0.0161
415
+ step 4130000 | 41545s | train_loss=0.0159
416
+ step 4140000 | 41641s | train_loss=0.0160
417
+ step 4150000 | 41737s | train_loss=0.0165
418
+ step 4160000 | 41821s | train_loss=0.0161
419
+ step 4170000 | 41911s | train_loss=0.0159
420
+ step 4180000 | 42007s | train_loss=0.0160
421
+ step 4190000 | 42103s | train_loss=0.0161
422
+ step 4200000 | 42268s | train_loss=0.0161 | fid=67.2090
423
+ step 4210000 | 42364s | train_loss=0.0157
424
+ step 4220000 | 42460s | train_loss=0.0156
425
+ step 4230000 | 42554s | train_loss=0.0160
426
+ step 4240000 | 42649s | train_loss=0.0161
427
+ step 4250000 | 42745s | train_loss=0.0159
428
+ step 4260000 | 42827s | train_loss=0.0156
429
+ step 4270000 | 42919s | train_loss=0.0155
430
+ step 4280000 | 43015s | train_loss=0.0155
431
+ step 4290000 | 43110s | train_loss=0.0157
432
+ step 4300000 | 43276s | train_loss=0.0154 | fid=67.6258
433
+ step 4310000 | 43372s | train_loss=0.0156
434
+ step 4320000 | 43468s | train_loss=0.0154
435
+ step 4330000 | 43565s | train_loss=0.0154
436
+ step 4340000 | 43658s | train_loss=0.0153
437
+ step 4350000 | 43754s | train_loss=0.0153
438
+ step 4360000 | 43834s | train_loss=0.0156
439
+ step 4370000 | 43929s | train_loss=0.0151
440
+ step 4380000 | 44024s | train_loss=0.0152
441
+ step 4390000 | 44119s | train_loss=0.0153
442
+ step 4400000 | 44285s | train_loss=0.0155 | fid=66.9013
443
+ step 4410000 | 44381s | train_loss=0.0153
444
+ step 4420000 | 44476s | train_loss=0.0149
445
+ step 4430000 | 44572s | train_loss=0.0152
446
+ step 4440000 | 44666s | train_loss=0.0151
447
+ step 4450000 | 44761s | train_loss=0.0152
448
+ step 4460000 | 44840s | train_loss=0.0152
449
+ step 4470000 | 44933s | train_loss=0.0153
450
+ step 4480000 | 45027s | train_loss=0.0154
451
+ step 4490000 | 45121s | train_loss=0.0150
452
+ step 4500000 | 45290s | train_loss=0.0151 | fid=67.3875
453
+ step 4510000 | 45385s | train_loss=0.0149
454
+ step 4520000 | 45481s | train_loss=0.0153
455
+ step 4530000 | 45577s | train_loss=0.0150
456
+ step 4540000 | 45673s | train_loss=0.0148
457
+ step 4550000 | 45766s | train_loss=0.0149
458
+ step 4560000 | 45848s | train_loss=0.0147
459
+ step 4570000 | 45945s | train_loss=0.0147
460
+ step 4580000 | 46038s | train_loss=0.0148
461
+ step 4590000 | 46134s | train_loss=0.0147
462
+ step 4600000 | 46302s | train_loss=0.0149 | fid=67.9078
463
+ step 4610000 | 46397s | train_loss=0.0148
464
+ step 4620000 | 46493s | train_loss=0.0144
465
+ step 4630000 | 46589s | train_loss=0.0146
466
+ step 4640000 | 46684s | train_loss=0.0146
467
+ step 4650000 | 46777s | train_loss=0.0146
468
+ step 4660000 | 46858s | train_loss=0.0145
469
+ step 4670000 | 46954s | train_loss=0.0145
470
+ step 4680000 | 47047s | train_loss=0.0147
471
+ step 4690000 | 47143s | train_loss=0.0144
472
+ step 4700000 | 47310s | train_loss=0.0143 | fid=67.8041
473
+ step 4710000 | 47406s | train_loss=0.0145
474
+ step 4720000 | 47501s | train_loss=0.0144
475
+ step 4730000 | 47596s | train_loss=0.0145
476
+ step 4740000 | 47692s | train_loss=0.0146
477
+ step 4750000 | 47787s | train_loss=0.0147
478
+ step 4760000 | 47874s | train_loss=0.0144
479
+ step 4770000 | 47970s | train_loss=0.0147
480
+ step 4780000 | 48064s | train_loss=0.0144
481
+ step 4790000 | 48160s | train_loss=0.0142
482
+ step 4800000 | 48327s | train_loss=0.0145 | fid=67.1548
483
+ step 4810000 | 48423s | train_loss=0.0140
484
+ step 4820000 | 48517s | train_loss=0.0142
485
+ step 4830000 | 48613s | train_loss=0.0141
486
+ step 4840000 | 48708s | train_loss=0.0142
487
+ step 4850000 | 48794s | train_loss=0.0144
488
+ step 4860000 | 48870s | train_loss=0.0142
489
+ step 4870000 | 48948s | train_loss=0.0144
490
+ step 4880000 | 49024s | train_loss=0.0141
491
+ step 4890000 | 49103s | train_loss=0.0139
492
+ step 4900000 | 49241s | train_loss=0.0140 | fid=66.9205
493
+ step 4910000 | 49318s | train_loss=0.0140
494
+ step 4920000 | 49397s | train_loss=0.0138
495
+ step 4930000 | 49475s | train_loss=0.0140
496
+ step 4940000 | 49553s | train_loss=0.0140
497
+ step 4950000 | 49631s | train_loss=0.0137
498
+ step 4960000 | 49707s | train_loss=0.0138
499
+ step 4970000 | 49786s | train_loss=0.0138
500
+ step 4980000 | 49857s | train_loss=0.0142
501
+ step 4990000 | 49915s | train_loss=0.0139
502
+ step 5000000 | 49999s | train_loss=0.0138 | fid=67.4795
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d200/config.json ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "vae_checkpoint": "vae_checkpoints/celeba_resnet_modernloss_d200/vae.pt",
3
+ "out": "results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d200",
4
+ "subset_path": "diagnostics/celeba_diverse_1k_mean_plus_sd/subset_indices.json",
5
+ "train_tar": "/overflow/data/cerberus/celeba_hq_256_partitioned_ws/train/celeba_hq_256-train-0000.tar",
6
+ "n_train": 1000,
7
+ "total_steps": 5000000,
8
+ "eval_interval": 10000,
9
+ "mem_interval": 100000,
10
+ "fid_interval": 100000,
11
+ "fid_n_real": 1000,
12
+ "fid_n_gen": 10000,
13
+ "n_gen_samples": 10000,
14
+ "n_sde_steps": 500,
15
+ "batch_size": 512,
16
+ "hidden": 1024,
17
+ "depth": 5,
18
+ "optimizer": "sgd",
19
+ "lr": 0.001,
20
+ "momentum": 0.8,
21
+ "t_min": 0.01,
22
+ "t_max": 3.0,
23
+ "t_eval": 0.1,
24
+ "seed": 45,
25
+ "device": null
26
+ }
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d200/metrics.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d200/runner.log ADDED
@@ -0,0 +1,502 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ FID real-side stats cached over 1000 diverse train images
2
+ step 1 | 77s | train_loss=1.0028 | fid=309.1138
3
+ step 10000 | 177s | train_loss=0.6356
4
+ step 20000 | 278s | train_loss=0.5214
5
+ step 30000 | 378s | train_loss=0.4640
6
+ step 40000 | 478s | train_loss=0.4210
7
+ step 50000 | 578s | train_loss=0.3664
8
+ step 60000 | 679s | train_loss=0.3122
9
+ step 70000 | 779s | train_loss=0.2724
10
+ step 80000 | 880s | train_loss=0.2302
11
+ step 90000 | 980s | train_loss=0.2040
12
+ step 100000 | 1151s | train_loss=0.1854 | fid=67.6428
13
+ step 110000 | 1251s | train_loss=0.1658
14
+ step 120000 | 1351s | train_loss=0.1546
15
+ step 130000 | 1452s | train_loss=0.1443
16
+ step 140000 | 1553s | train_loss=0.1432
17
+ step 150000 | 1653s | train_loss=0.1308
18
+ step 160000 | 1753s | train_loss=0.1258
19
+ step 170000 | 1856s | train_loss=0.1232
20
+ step 180000 | 1959s | train_loss=0.1169
21
+ step 190000 | 2058s | train_loss=0.1142
22
+ step 200000 | 2232s | train_loss=0.1110 | fid=77.0928
23
+ step 210000 | 2335s | train_loss=0.1047
24
+ step 220000 | 2438s | train_loss=0.1078
25
+ step 230000 | 2541s | train_loss=0.0999
26
+ step 240000 | 2643s | train_loss=0.0981
27
+ step 250000 | 2745s | train_loss=0.0985
28
+ step 260000 | 2845s | train_loss=0.0974
29
+ step 270000 | 2947s | train_loss=0.0969
30
+ step 280000 | 3046s | train_loss=0.0891
31
+ step 290000 | 3148s | train_loss=0.0892
32
+ step 300000 | 3323s | train_loss=0.0903 | fid=89.0995
33
+ step 310000 | 3426s | train_loss=0.0875
34
+ step 320000 | 3529s | train_loss=0.0867
35
+ step 330000 | 3631s | train_loss=0.0802
36
+ step 340000 | 3733s | train_loss=0.0805
37
+ step 350000 | 3834s | train_loss=0.0767
38
+ step 360000 | 3935s | train_loss=0.0792
39
+ step 370000 | 4034s | train_loss=0.0802
40
+ step 380000 | 4136s | train_loss=0.0771
41
+ step 390000 | 4238s | train_loss=0.0750
42
+ step 400000 | 4412s | train_loss=0.0771 | fid=96.2196
43
+ step 410000 | 4515s | train_loss=0.0733
44
+ step 420000 | 4617s | train_loss=0.0720
45
+ step 430000 | 4720s | train_loss=0.0693
46
+ step 440000 | 4822s | train_loss=0.0709
47
+ step 450000 | 4923s | train_loss=0.0698
48
+ step 460000 | 5021s | train_loss=0.0682
49
+ step 470000 | 5123s | train_loss=0.0665
50
+ step 480000 | 5225s | train_loss=0.0692
51
+ step 490000 | 5328s | train_loss=0.0668
52
+ step 500000 | 5502s | train_loss=0.0634 | fid=95.0143
53
+ step 510000 | 5604s | train_loss=0.0701
54
+ step 520000 | 5707s | train_loss=0.0637
55
+ step 530000 | 5808s | train_loss=0.0622
56
+ step 540000 | 5911s | train_loss=0.0613
57
+ step 550000 | 6008s | train_loss=0.0613
58
+ step 560000 | 6110s | train_loss=0.0683
59
+ step 570000 | 6212s | train_loss=0.0622
60
+ step 580000 | 6314s | train_loss=0.0632
61
+ step 590000 | 6418s | train_loss=0.0575
62
+ step 600000 | 6587s | train_loss=0.0585 | fid=97.6045
63
+ step 610000 | 6690s | train_loss=0.0604
64
+ step 620000 | 6791s | train_loss=0.0594
65
+ step 630000 | 6894s | train_loss=0.0556
66
+ step 640000 | 6992s | train_loss=0.0574
67
+ step 650000 | 7095s | train_loss=0.0554
68
+ step 660000 | 7196s | train_loss=0.0577
69
+ step 670000 | 7298s | train_loss=0.0570
70
+ step 680000 | 7399s | train_loss=0.0574
71
+ step 690000 | 7495s | train_loss=0.0549
72
+ step 700000 | 7674s | train_loss=0.0545 | fid=99.6299
73
+ step 710000 | 7775s | train_loss=0.0559
74
+ step 720000 | 7877s | train_loss=0.0528
75
+ step 730000 | 7977s | train_loss=0.0564
76
+ step 740000 | 8079s | train_loss=0.0551
77
+ step 750000 | 8182s | train_loss=0.0514
78
+ step 760000 | 8283s | train_loss=0.0498
79
+ step 770000 | 8385s | train_loss=0.0531
80
+ step 780000 | 8488s | train_loss=0.0562
81
+ step 790000 | 8585s | train_loss=0.0491
82
+ step 800000 | 8764s | train_loss=0.0511 | fid=99.1503
83
+ step 810000 | 8867s | train_loss=0.0506
84
+ step 820000 | 8961s | train_loss=0.0518
85
+ step 830000 | 9055s | train_loss=0.0545
86
+ step 840000 | 9155s | train_loss=0.0541
87
+ step 850000 | 9257s | train_loss=0.0523
88
+ step 860000 | 9359s | train_loss=0.0494
89
+ step 870000 | 9459s | train_loss=0.0497
90
+ step 880000 | 9562s | train_loss=0.0470
91
+ step 890000 | 9659s | train_loss=0.0469
92
+ step 900000 | 9839s | train_loss=0.0484 | fid=96.2394
93
+ step 910000 | 9941s | train_loss=0.0509
94
+ step 920000 | 10041s | train_loss=0.0488
95
+ step 930000 | 10144s | train_loss=0.0481
96
+ step 940000 | 10246s | train_loss=0.0490
97
+ step 950000 | 10348s | train_loss=0.0442
98
+ step 960000 | 10449s | train_loss=0.0463
99
+ step 970000 | 10550s | train_loss=0.0473
100
+ step 980000 | 10650s | train_loss=0.0472
101
+ step 990000 | 10751s | train_loss=0.0457
102
+ step 1000000 | 10930s | train_loss=0.0467 | fid=92.2198
103
+ step 1010000 | 11030s | train_loss=0.0461
104
+ step 1020000 | 11132s | train_loss=0.0459
105
+ step 1030000 | 11234s | train_loss=0.0454
106
+ step 1040000 | 11336s | train_loss=0.0447
107
+ step 1050000 | 11438s | train_loss=0.0467
108
+ step 1060000 | 11540s | train_loss=0.0444
109
+ step 1070000 | 11642s | train_loss=0.0514
110
+ step 1080000 | 11740s | train_loss=0.0426
111
+ step 1090000 | 11842s | train_loss=0.0462
112
+ step 1100000 | 12012s | train_loss=0.0447 | fid=87.1018
113
+ step 1110000 | 12115s | train_loss=0.0448
114
+ step 1120000 | 12217s | train_loss=0.0431
115
+ step 1130000 | 12319s | train_loss=0.0411
116
+ step 1140000 | 12420s | train_loss=0.0445
117
+ step 1150000 | 12522s | train_loss=0.0462
118
+ step 1160000 | 12624s | train_loss=0.0451
119
+ step 1170000 | 12727s | train_loss=0.0459
120
+ step 1180000 | 12824s | train_loss=0.0471
121
+ step 1190000 | 12927s | train_loss=0.0430
122
+ step 1200000 | 13104s | train_loss=0.0432 | fid=85.1261
123
+ step 1210000 | 13207s | train_loss=0.0447
124
+ step 1220000 | 13309s | train_loss=0.0433
125
+ step 1230000 | 13411s | train_loss=0.0424
126
+ step 1240000 | 13513s | train_loss=0.0453
127
+ step 1250000 | 13615s | train_loss=0.0403
128
+ step 1260000 | 13717s | train_loss=0.0436
129
+ step 1270000 | 13816s | train_loss=0.0404
130
+ step 1280000 | 13917s | train_loss=0.0413
131
+ step 1290000 | 14008s | train_loss=0.0432
132
+ step 1300000 | 14188s | train_loss=0.0395 | fid=78.5018
133
+ step 1310000 | 14290s | train_loss=0.0413
134
+ step 1320000 | 14391s | train_loss=0.0419
135
+ step 1330000 | 14493s | train_loss=0.0407
136
+ step 1340000 | 14596s | train_loss=0.0433
137
+ step 1350000 | 14698s | train_loss=0.0402
138
+ step 1360000 | 14800s | train_loss=0.0380
139
+ step 1370000 | 14900s | train_loss=0.0383
140
+ step 1380000 | 14996s | train_loss=0.0398
141
+ step 1390000 | 15093s | train_loss=0.0423
142
+ step 1400000 | 15272s | train_loss=0.0380 | fid=78.8938
143
+ step 1410000 | 15375s | train_loss=0.0387
144
+ step 1420000 | 15475s | train_loss=0.0394
145
+ step 1430000 | 15578s | train_loss=0.0404
146
+ step 1440000 | 15680s | train_loss=0.0390
147
+ step 1450000 | 15782s | train_loss=0.0386
148
+ step 1460000 | 15884s | train_loss=0.0382
149
+ step 1470000 | 15983s | train_loss=0.0394
150
+ step 1480000 | 16076s | train_loss=0.0374
151
+ step 1490000 | 16178s | train_loss=0.0441
152
+ step 1500000 | 16356s | train_loss=0.0378 | fid=75.2170
153
+ step 1510000 | 16458s | train_loss=0.0374
154
+ step 1520000 | 16558s | train_loss=0.0374
155
+ step 1530000 | 16661s | train_loss=0.0389
156
+ step 1540000 | 16763s | train_loss=0.0359
157
+ step 1550000 | 16865s | train_loss=0.0398
158
+ step 1560000 | 16964s | train_loss=0.0382
159
+ step 1570000 | 17053s | train_loss=0.0355
160
+ step 1580000 | 17151s | train_loss=0.0382
161
+ step 1590000 | 17253s | train_loss=0.0348
162
+ step 1600000 | 17432s | train_loss=0.0349 | fid=70.6419
163
+ step 1610000 | 17534s | train_loss=0.0382
164
+ step 1620000 | 17636s | train_loss=0.0385
165
+ step 1630000 | 17739s | train_loss=0.0356
166
+ step 1640000 | 17839s | train_loss=0.0362
167
+ step 1650000 | 17942s | train_loss=0.0360
168
+ step 1660000 | 18042s | train_loss=0.0372
169
+ step 1670000 | 18126s | train_loss=0.0360
170
+ step 1680000 | 18229s | train_loss=0.0341
171
+ step 1690000 | 18329s | train_loss=0.0352
172
+ step 1700000 | 18509s | train_loss=0.0349 | fid=72.9898
173
+ step 1710000 | 18611s | train_loss=0.0343
174
+ step 1720000 | 18713s | train_loss=0.0371
175
+ step 1730000 | 18815s | train_loss=0.0352
176
+ step 1740000 | 18917s | train_loss=0.0380
177
+ step 1750000 | 19021s | train_loss=0.0354
178
+ step 1760000 | 19113s | train_loss=0.0348
179
+ step 1770000 | 19206s | train_loss=0.0343
180
+ step 1780000 | 19307s | train_loss=0.0361
181
+ step 1790000 | 19410s | train_loss=0.0326
182
+ step 1800000 | 19589s | train_loss=0.0343 | fid=66.6996
183
+ step 1810000 | 19691s | train_loss=0.0341
184
+ step 1820000 | 19793s | train_loss=0.0326
185
+ step 1830000 | 19895s | train_loss=0.0358
186
+ step 1840000 | 19997s | train_loss=0.0325
187
+ step 1850000 | 20097s | train_loss=0.0320
188
+ step 1860000 | 20187s | train_loss=0.0327
189
+ step 1870000 | 20289s | train_loss=0.0333
190
+ step 1880000 | 20392s | train_loss=0.0344
191
+ step 1890000 | 20494s | train_loss=0.0342
192
+ step 1900000 | 20673s | train_loss=0.0351 | fid=69.0640
193
+ step 1910000 | 20775s | train_loss=0.0335
194
+ step 1920000 | 20876s | train_loss=0.0340
195
+ step 1930000 | 20978s | train_loss=0.0350
196
+ step 1940000 | 21081s | train_loss=0.0321
197
+ step 1950000 | 21172s | train_loss=0.0364
198
+ step 1960000 | 21266s | train_loss=0.0317
199
+ step 1970000 | 21369s | train_loss=0.0344
200
+ step 1980000 | 21472s | train_loss=0.0319
201
+ step 1990000 | 21574s | train_loss=0.0333
202
+ step 2000000 | 21753s | train_loss=0.0329 | fid=63.7743
203
+ step 2010000 | 21854s | train_loss=0.0319
204
+ step 2020000 | 21955s | train_loss=0.0321
205
+ step 2030000 | 22058s | train_loss=0.0311
206
+ step 2040000 | 22157s | train_loss=0.0317
207
+ step 2050000 | 22245s | train_loss=0.0321
208
+ step 2060000 | 22346s | train_loss=0.0316
209
+ step 2070000 | 22448s | train_loss=0.0385
210
+ step 2080000 | 22551s | train_loss=0.0302
211
+ step 2090000 | 22653s | train_loss=0.0333
212
+ step 2100000 | 22833s | train_loss=0.0327 | fid=62.7510
213
+ step 2110000 | 22935s | train_loss=0.0345
214
+ step 2120000 | 23037s | train_loss=0.0314
215
+ step 2130000 | 23137s | train_loss=0.0320
216
+ step 2140000 | 23226s | train_loss=0.0315
217
+ step 2150000 | 23326s | train_loss=0.0326
218
+ step 2160000 | 23429s | train_loss=0.0310
219
+ step 2170000 | 23532s | train_loss=0.0355
220
+ step 2180000 | 23634s | train_loss=0.0305
221
+ step 2190000 | 23736s | train_loss=0.0317
222
+ step 2200000 | 23914s | train_loss=0.0308 | fid=65.3953
223
+ step 2210000 | 24015s | train_loss=0.0309
224
+ step 2220000 | 24117s | train_loss=0.0321
225
+ step 2230000 | 24217s | train_loss=0.0297
226
+ step 2240000 | 24310s | train_loss=0.0278
227
+ step 2250000 | 24411s | train_loss=0.0315
228
+ step 2260000 | 24513s | train_loss=0.0296
229
+ step 2270000 | 24616s | train_loss=0.0306
230
+ step 2280000 | 24718s | train_loss=0.0291
231
+ step 2290000 | 24821s | train_loss=0.0361
232
+ step 2300000 | 25000s | train_loss=0.0302 | fid=60.9445
233
+ step 2310000 | 25102s | train_loss=0.0311
234
+ step 2320000 | 25202s | train_loss=0.0305
235
+ step 2330000 | 25290s | train_loss=0.0292
236
+ step 2340000 | 25391s | train_loss=0.0303
237
+ step 2350000 | 25492s | train_loss=0.0310
238
+ step 2360000 | 25595s | train_loss=0.0301
239
+ step 2370000 | 25698s | train_loss=0.0303
240
+ step 2380000 | 25800s | train_loss=0.0315
241
+ step 2390000 | 25902s | train_loss=0.0277
242
+ step 2400000 | 26081s | train_loss=0.0296 | fid=61.3778
243
+ step 2410000 | 26183s | train_loss=0.0297
244
+ step 2420000 | 26277s | train_loss=0.0294
245
+ step 2430000 | 26371s | train_loss=0.0292
246
+ step 2440000 | 26474s | train_loss=0.0274
247
+ step 2450000 | 26575s | train_loss=0.0283
248
+ step 2460000 | 26678s | train_loss=0.0321
249
+ step 2470000 | 26780s | train_loss=0.0301
250
+ step 2480000 | 26882s | train_loss=0.0289
251
+ step 2490000 | 26984s | train_loss=0.0282
252
+ step 2500000 | 27164s | train_loss=0.0299 | fid=57.3555
253
+ step 2510000 | 27265s | train_loss=0.0279
254
+ step 2520000 | 27353s | train_loss=0.0326
255
+ step 2530000 | 27453s | train_loss=0.0283
256
+ step 2540000 | 27554s | train_loss=0.0297
257
+ step 2550000 | 27657s | train_loss=0.0294
258
+ step 2560000 | 27758s | train_loss=0.0296
259
+ step 2570000 | 27860s | train_loss=0.0271
260
+ step 2580000 | 27963s | train_loss=0.0270
261
+ step 2590000 | 28065s | train_loss=0.0301
262
+ step 2600000 | 28244s | train_loss=0.0287 | fid=55.8671
263
+ step 2610000 | 28337s | train_loss=0.0275
264
+ step 2620000 | 28437s | train_loss=0.0285
265
+ step 2630000 | 28540s | train_loss=0.0275
266
+ step 2640000 | 28641s | train_loss=0.0272
267
+ step 2650000 | 28744s | train_loss=0.0283
268
+ step 2660000 | 28846s | train_loss=0.0273
269
+ step 2670000 | 28949s | train_loss=0.0266
270
+ step 2680000 | 29051s | train_loss=0.0276
271
+ step 2690000 | 29153s | train_loss=0.0277
272
+ step 2700000 | 29332s | train_loss=0.0267 | fid=55.4955
273
+ step 2710000 | 29434s | train_loss=0.0295
274
+ step 2720000 | 29534s | train_loss=0.0273
275
+ step 2730000 | 29637s | train_loss=0.0284
276
+ step 2740000 | 29737s | train_loss=0.0267
277
+ step 2750000 | 29839s | train_loss=0.0259
278
+ step 2760000 | 29942s | train_loss=0.0288
279
+ step 2770000 | 30044s | train_loss=0.0252
280
+ step 2780000 | 30146s | train_loss=0.0255
281
+ step 2790000 | 30248s | train_loss=0.0322
282
+ step 2800000 | 30413s | train_loss=0.0263 | fid=55.5426
283
+ step 2810000 | 30517s | train_loss=0.0286
284
+ step 2820000 | 30616s | train_loss=0.0297
285
+ step 2830000 | 30717s | train_loss=0.0277
286
+ step 2840000 | 30819s | train_loss=0.0249
287
+ step 2850000 | 30921s | train_loss=0.0254
288
+ step 2860000 | 31023s | train_loss=0.0268
289
+ step 2870000 | 31125s | train_loss=0.0281
290
+ step 2880000 | 31229s | train_loss=0.0319
291
+ step 2890000 | 31332s | train_loss=0.0252
292
+ step 2900000 | 31503s | train_loss=0.0242 | fid=53.2558
293
+ step 2910000 | 31603s | train_loss=0.0336
294
+ step 2920000 | 31704s | train_loss=0.0266
295
+ step 2930000 | 31804s | train_loss=0.0280
296
+ step 2940000 | 31906s | train_loss=0.0264
297
+ step 2950000 | 32008s | train_loss=0.0268
298
+ step 2960000 | 32111s | train_loss=0.0259
299
+ step 2970000 | 32213s | train_loss=0.0269
300
+ step 2980000 | 32316s | train_loss=0.0264
301
+ step 2990000 | 32412s | train_loss=0.0300
302
+ step 3000000 | 32592s | train_loss=0.0292 | fid=52.0522
303
+ step 3010000 | 32690s | train_loss=0.0264
304
+ step 3020000 | 32792s | train_loss=0.0255
305
+ step 3030000 | 32893s | train_loss=0.0250
306
+ step 3040000 | 32995s | train_loss=0.0255
307
+ step 3050000 | 33098s | train_loss=0.0263
308
+ step 3060000 | 33200s | train_loss=0.0272
309
+ step 3070000 | 33302s | train_loss=0.0250
310
+ step 3080000 | 33402s | train_loss=0.0252
311
+ step 3090000 | 33504s | train_loss=0.0241
312
+ step 3100000 | 33681s | train_loss=0.0255 | fid=50.0098
313
+ step 3110000 | 33782s | train_loss=0.0252
314
+ step 3120000 | 33885s | train_loss=0.0272
315
+ step 3130000 | 33986s | train_loss=0.0278
316
+ step 3140000 | 34088s | train_loss=0.0242
317
+ step 3150000 | 34191s | train_loss=0.0243
318
+ step 3160000 | 34293s | train_loss=0.0252
319
+ step 3170000 | 34388s | train_loss=0.0241
320
+ step 3180000 | 34491s | train_loss=0.0254
321
+ step 3190000 | 34593s | train_loss=0.0251
322
+ step 3200000 | 34769s | train_loss=0.0247 | fid=52.3011
323
+ step 3210000 | 34871s | train_loss=0.0235
324
+ step 3220000 | 34972s | train_loss=0.0254
325
+ step 3230000 | 35074s | train_loss=0.0258
326
+ step 3240000 | 35176s | train_loss=0.0254
327
+ step 3250000 | 35278s | train_loss=0.0245
328
+ step 3260000 | 35377s | train_loss=0.0240
329
+ step 3270000 | 35474s | train_loss=0.0257
330
+ step 3280000 | 35576s | train_loss=0.0260
331
+ step 3290000 | 35678s | train_loss=0.0241
332
+ step 3300000 | 35854s | train_loss=0.0258 | fid=53.3655
333
+ step 3310000 | 35957s | train_loss=0.0238
334
+ step 3320000 | 36057s | train_loss=0.0280
335
+ step 3330000 | 36160s | train_loss=0.0252
336
+ step 3340000 | 36263s | train_loss=0.0259
337
+ step 3350000 | 36364s | train_loss=0.0222
338
+ step 3360000 | 36456s | train_loss=0.0248
339
+ step 3370000 | 36559s | train_loss=0.0262
340
+ step 3380000 | 36661s | train_loss=0.0243
341
+ step 3390000 | 36763s | train_loss=0.0242
342
+ step 3400000 | 36940s | train_loss=0.0260 | fid=48.4075
343
+ step 3410000 | 37043s | train_loss=0.0254
344
+ step 3420000 | 37144s | train_loss=0.0242
345
+ step 3430000 | 37247s | train_loss=0.0265
346
+ step 3440000 | 37349s | train_loss=0.0224
347
+ step 3450000 | 37440s | train_loss=0.0263
348
+ step 3460000 | 37543s | train_loss=0.0229
349
+ step 3470000 | 37644s | train_loss=0.0239
350
+ step 3480000 | 37746s | train_loss=0.0241
351
+ step 3490000 | 37850s | train_loss=0.0257
352
+ step 3500000 | 38027s | train_loss=0.0243 | fid=48.7212
353
+ step 3510000 | 38128s | train_loss=0.0243
354
+ step 3520000 | 38229s | train_loss=0.0231
355
+ step 3530000 | 38332s | train_loss=0.0242
356
+ step 3540000 | 38430s | train_loss=0.0244
357
+ step 3550000 | 38525s | train_loss=0.0243
358
+ step 3560000 | 38626s | train_loss=0.0238
359
+ step 3570000 | 38728s | train_loss=0.0246
360
+ step 3580000 | 38830s | train_loss=0.0239
361
+ step 3590000 | 38932s | train_loss=0.0270
362
+ step 3600000 | 39110s | train_loss=0.0233 | fid=47.7787
363
+ step 3610000 | 39210s | train_loss=0.0240
364
+ step 3620000 | 39313s | train_loss=0.0227
365
+ step 3630000 | 39415s | train_loss=0.0231
366
+ step 3640000 | 39505s | train_loss=0.0247
367
+ step 3650000 | 39607s | train_loss=0.0232
368
+ step 3660000 | 39710s | train_loss=0.0236
369
+ step 3670000 | 39813s | train_loss=0.0236
370
+ step 3680000 | 39915s | train_loss=0.0242
371
+ step 3690000 | 40015s | train_loss=0.0219
372
+ step 3700000 | 40195s | train_loss=0.0246 | fid=45.5566
373
+ step 3710000 | 40296s | train_loss=0.0246
374
+ step 3720000 | 40398s | train_loss=0.0253
375
+ step 3730000 | 40494s | train_loss=0.0222
376
+ step 3740000 | 40590s | train_loss=0.0229
377
+ step 3750000 | 40692s | train_loss=0.0237
378
+ step 3760000 | 40794s | train_loss=0.0211
379
+ step 3770000 | 40896s | train_loss=0.0225
380
+ step 3780000 | 40997s | train_loss=0.0237
381
+ step 3790000 | 41096s | train_loss=0.0245
382
+ step 3800000 | 41275s | train_loss=0.0239 | fid=44.9775
383
+ step 3810000 | 41376s | train_loss=0.0212
384
+ step 3820000 | 41478s | train_loss=0.0231
385
+ step 3830000 | 41569s | train_loss=0.0231
386
+ step 3840000 | 41671s | train_loss=0.0217
387
+ step 3850000 | 41774s | train_loss=0.0238
388
+ step 3860000 | 41875s | train_loss=0.0226
389
+ step 3870000 | 41978s | train_loss=0.0220
390
+ step 3880000 | 42078s | train_loss=0.0225
391
+ step 3890000 | 42180s | train_loss=0.0222
392
+ step 3900000 | 42361s | train_loss=0.0228 | fid=42.8499
393
+ step 3910000 | 42464s | train_loss=0.0238
394
+ step 3920000 | 42558s | train_loss=0.0220
395
+ step 3930000 | 42661s | train_loss=0.0229
396
+ step 3940000 | 42762s | train_loss=0.0217
397
+ step 3950000 | 42865s | train_loss=0.0230
398
+ step 3960000 | 42968s | train_loss=0.0212
399
+ step 3970000 | 43071s | train_loss=0.0223
400
+ step 3980000 | 43171s | train_loss=0.0238
401
+ step 3990000 | 43274s | train_loss=0.0201
402
+ step 4000000 | 43453s | train_loss=0.0220 | fid=45.6634
403
+ step 4010000 | 43549s | train_loss=0.0228
404
+ step 4020000 | 43648s | train_loss=0.0224
405
+ step 4030000 | 43750s | train_loss=0.0208
406
+ step 4040000 | 43853s | train_loss=0.0218
407
+ step 4050000 | 43955s | train_loss=0.0213
408
+ step 4060000 | 44058s | train_loss=0.0223
409
+ step 4070000 | 44160s | train_loss=0.0216
410
+ step 4080000 | 44261s | train_loss=0.0242
411
+ step 4090000 | 44364s | train_loss=0.0225
412
+ step 4100000 | 44540s | train_loss=0.0201 | fid=46.4232
413
+ step 4110000 | 44633s | train_loss=0.0217
414
+ step 4120000 | 44735s | train_loss=0.0240
415
+ step 4130000 | 44837s | train_loss=0.0208
416
+ step 4140000 | 44939s | train_loss=0.0222
417
+ step 4150000 | 45042s | train_loss=0.0206
418
+ step 4160000 | 45145s | train_loss=0.0219
419
+ step 4170000 | 45245s | train_loss=0.0232
420
+ step 4180000 | 45347s | train_loss=0.0209
421
+ step 4190000 | 45449s | train_loss=0.0228
422
+ step 4200000 | 45619s | train_loss=0.0204 | fid=48.0027
423
+ step 4210000 | 45722s | train_loss=0.0214
424
+ step 4220000 | 45824s | train_loss=0.0207
425
+ step 4230000 | 45927s | train_loss=0.0201
426
+ step 4240000 | 46030s | train_loss=0.0204
427
+ step 4250000 | 46133s | train_loss=0.0228
428
+ step 4260000 | 46236s | train_loss=0.0216
429
+ step 4270000 | 46335s | train_loss=0.0217
430
+ step 4280000 | 46437s | train_loss=0.0220
431
+ step 4290000 | 46539s | train_loss=0.0210
432
+ step 4300000 | 46705s | train_loss=0.0210 | fid=47.1222
433
+ step 4310000 | 46806s | train_loss=0.0205
434
+ step 4320000 | 46909s | train_loss=0.0203
435
+ step 4330000 | 47012s | train_loss=0.0220
436
+ step 4340000 | 47113s | train_loss=0.0208
437
+ step 4350000 | 47216s | train_loss=0.0263
438
+ step 4360000 | 47317s | train_loss=0.0202
439
+ step 4370000 | 47418s | train_loss=0.0203
440
+ step 4380000 | 47521s | train_loss=0.0210
441
+ step 4390000 | 47617s | train_loss=0.0208
442
+ step 4400000 | 47790s | train_loss=0.0211 | fid=42.9458
443
+ step 4410000 | 47893s | train_loss=0.0222
444
+ step 4420000 | 47996s | train_loss=0.0205
445
+ step 4430000 | 48097s | train_loss=0.0232
446
+ step 4440000 | 48200s | train_loss=0.0248
447
+ step 4450000 | 48303s | train_loss=0.0207
448
+ step 4460000 | 48402s | train_loss=0.0223
449
+ step 4470000 | 48504s | train_loss=0.0220
450
+ step 4480000 | 48607s | train_loss=0.0227
451
+ step 4490000 | 48706s | train_loss=0.0202
452
+ step 4500000 | 48883s | train_loss=0.0194 | fid=42.0542
453
+ step 4510000 | 48984s | train_loss=0.0209
454
+ step 4520000 | 49087s | train_loss=0.0202
455
+ step 4530000 | 49190s | train_loss=0.0200
456
+ step 4540000 | 49292s | train_loss=0.0233
457
+ step 4550000 | 49394s | train_loss=0.0201
458
+ step 4560000 | 49495s | train_loss=0.0206
459
+ step 4570000 | 49598s | train_loss=0.0203
460
+ step 4580000 | 49697s | train_loss=0.0211
461
+ step 4590000 | 49796s | train_loss=0.0212
462
+ step 4600000 | 49974s | train_loss=0.0192 | fid=45.0506
463
+ step 4610000 | 50074s | train_loss=0.0223
464
+ step 4620000 | 50174s | train_loss=0.0196
465
+ step 4630000 | 50275s | train_loss=0.0197
466
+ step 4640000 | 50374s | train_loss=0.0196
467
+ step 4650000 | 50472s | train_loss=0.0192
468
+ step 4660000 | 50571s | train_loss=0.0207
469
+ step 4670000 | 50654s | train_loss=0.0207
470
+ step 4680000 | 50735s | train_loss=0.0202
471
+ step 4690000 | 50815s | train_loss=0.0207
472
+ step 4700000 | 50959s | train_loss=0.0209 | fid=42.9424
473
+ step 4710000 | 51040s | train_loss=0.0201
474
+ step 4720000 | 51121s | train_loss=0.0206
475
+ step 4730000 | 51202s | train_loss=0.0200
476
+ step 4740000 | 51283s | train_loss=0.0198
477
+ step 4750000 | 51361s | train_loss=0.0190
478
+ step 4760000 | 51441s | train_loss=0.0237
479
+ step 4770000 | 51522s | train_loss=0.0194
480
+ step 4780000 | 51603s | train_loss=0.0180
481
+ step 4790000 | 51682s | train_loss=0.0194
482
+ step 4800000 | 51832s | train_loss=0.0200 | fid=44.1091
483
+ step 4810000 | 51914s | train_loss=0.0214
484
+ step 4820000 | 51995s | train_loss=0.0231
485
+ step 4830000 | 52076s | train_loss=0.0202
486
+ step 4840000 | 52158s | train_loss=0.0200
487
+ step 4850000 | 52226s | train_loss=0.0191
488
+ step 4860000 | 52267s | train_loss=0.0200
489
+ step 4870000 | 52307s | train_loss=0.0216
490
+ step 4880000 | 52348s | train_loss=0.0204
491
+ step 4890000 | 52375s | train_loss=0.0217
492
+ step 4900000 | 52420s | train_loss=0.0195 | fid=46.4559
493
+ step 4910000 | 52443s | train_loss=0.0198
494
+ step 4920000 | 52467s | train_loss=0.0200
495
+ step 4930000 | 52490s | train_loss=0.0204
496
+ step 4940000 | 52514s | train_loss=0.0211
497
+ step 4950000 | 52537s | train_loss=0.0192
498
+ step 4960000 | 52561s | train_loss=0.0210
499
+ step 4970000 | 52584s | train_loss=0.0206
500
+ step 4980000 | 52608s | train_loss=0.0208
501
+ step 4990000 | 52631s | train_loss=0.0207
502
+ step 5000000 | 52676s | train_loss=0.0194 | fid=42.5611
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d30/config.json ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "vae_checkpoint": "vae_checkpoints/celeba_resnet_modernloss_d30/vae.pt",
3
+ "out": "results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d30",
4
+ "subset_path": "diagnostics/celeba_diverse_1k_mean_plus_sd/subset_indices.json",
5
+ "train_tar": "/overflow/data/cerberus/celeba_hq_256_partitioned_ws/train/celeba_hq_256-train-0000.tar",
6
+ "n_train": 1000,
7
+ "total_steps": 5000000,
8
+ "eval_interval": 10000,
9
+ "mem_interval": 100000,
10
+ "fid_interval": 100000,
11
+ "fid_n_real": 1000,
12
+ "fid_n_gen": 10000,
13
+ "n_gen_samples": 10000,
14
+ "n_sde_steps": 500,
15
+ "batch_size": 512,
16
+ "hidden": 1024,
17
+ "depth": 5,
18
+ "optimizer": "sgd",
19
+ "lr": 0.001,
20
+ "momentum": 0.8,
21
+ "t_min": 0.01,
22
+ "t_max": 3.0,
23
+ "t_eval": 0.1,
24
+ "seed": 45,
25
+ "device": null
26
+ }
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d30/metrics.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d30/runner.log ADDED
@@ -0,0 +1,502 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ FID real-side stats cached over 1000 diverse train images
2
+ step 1 | 74s | train_loss=0.9941 | fid=248.5675
3
+ step 10000 | 170s | train_loss=0.8646
4
+ step 20000 | 267s | train_loss=0.8536
5
+ step 30000 | 364s | train_loss=0.8382
6
+ step 40000 | 461s | train_loss=0.8082
7
+ step 50000 | 558s | train_loss=0.7582
8
+ step 60000 | 655s | train_loss=0.6820
9
+ step 70000 | 750s | train_loss=0.5728
10
+ step 80000 | 847s | train_loss=0.4513
11
+ step 90000 | 944s | train_loss=0.3512
12
+ step 100000 | 1105s | train_loss=0.2815 | fid=63.7853
13
+ step 110000 | 1201s | train_loss=0.2342
14
+ step 120000 | 1298s | train_loss=0.2061
15
+ step 130000 | 1394s | train_loss=0.1841
16
+ step 140000 | 1491s | train_loss=0.1711
17
+ step 150000 | 1589s | train_loss=0.1560
18
+ step 160000 | 1686s | train_loss=0.1460
19
+ step 170000 | 1783s | train_loss=0.1386
20
+ step 180000 | 1880s | train_loss=0.1305
21
+ step 190000 | 1975s | train_loss=0.1234
22
+ step 200000 | 2142s | train_loss=0.1173 | fid=60.1689
23
+ step 210000 | 2239s | train_loss=0.1119
24
+ step 220000 | 2336s | train_loss=0.1072
25
+ step 230000 | 2434s | train_loss=0.1032
26
+ step 240000 | 2531s | train_loss=0.0994
27
+ step 250000 | 2629s | train_loss=0.0948
28
+ step 260000 | 2726s | train_loss=0.0912
29
+ step 270000 | 2824s | train_loss=0.0886
30
+ step 280000 | 2919s | train_loss=0.0856
31
+ step 290000 | 3017s | train_loss=0.0828
32
+ step 300000 | 3183s | train_loss=0.0806 | fid=60.6603
33
+ step 310000 | 3281s | train_loss=0.0789
34
+ step 320000 | 3378s | train_loss=0.0760
35
+ step 330000 | 3476s | train_loss=0.0749
36
+ step 340000 | 3573s | train_loss=0.0727
37
+ step 350000 | 3671s | train_loss=0.0705
38
+ step 360000 | 3768s | train_loss=0.0703
39
+ step 370000 | 3863s | train_loss=0.0669
40
+ step 380000 | 3961s | train_loss=0.0661
41
+ step 390000 | 4057s | train_loss=0.0650
42
+ step 400000 | 4225s | train_loss=0.0638 | fid=59.7074
43
+ step 410000 | 4322s | train_loss=0.0624
44
+ step 420000 | 4420s | train_loss=0.0607
45
+ step 430000 | 4518s | train_loss=0.0601
46
+ step 440000 | 4615s | train_loss=0.0583
47
+ step 450000 | 4712s | train_loss=0.0580
48
+ step 460000 | 4806s | train_loss=0.0572
49
+ step 470000 | 4904s | train_loss=0.0555
50
+ step 480000 | 5001s | train_loss=0.0550
51
+ step 490000 | 5096s | train_loss=0.0544
52
+ step 500000 | 5264s | train_loss=0.0541 | fid=58.6863
53
+ step 510000 | 5361s | train_loss=0.0524
54
+ step 520000 | 5459s | train_loss=0.0525
55
+ step 530000 | 5556s | train_loss=0.0519
56
+ step 540000 | 5653s | train_loss=0.0506
57
+ step 550000 | 5748s | train_loss=0.0491
58
+ step 560000 | 5845s | train_loss=0.0491
59
+ step 570000 | 5943s | train_loss=0.0481
60
+ step 580000 | 6041s | train_loss=0.0479
61
+ step 590000 | 6134s | train_loss=0.0470
62
+ step 600000 | 6305s | train_loss=0.0473 | fid=58.6661
63
+ step 610000 | 6402s | train_loss=0.0455
64
+ step 620000 | 6499s | train_loss=0.0455
65
+ step 630000 | 6597s | train_loss=0.0446
66
+ step 640000 | 6691s | train_loss=0.0439
67
+ step 650000 | 6789s | train_loss=0.0438
68
+ step 660000 | 6886s | train_loss=0.0428
69
+ step 670000 | 6983s | train_loss=0.0432
70
+ step 680000 | 7079s | train_loss=0.0425
71
+ step 690000 | 7174s | train_loss=0.0421
72
+ step 700000 | 7345s | train_loss=0.0415 | fid=59.0550
73
+ step 710000 | 7442s | train_loss=0.0412
74
+ step 720000 | 7540s | train_loss=0.0405
75
+ step 730000 | 7634s | train_loss=0.0398
76
+ step 740000 | 7731s | train_loss=0.0405
77
+ step 750000 | 7828s | train_loss=0.0400
78
+ step 760000 | 7926s | train_loss=0.0391
79
+ step 770000 | 8024s | train_loss=0.0389
80
+ step 780000 | 8119s | train_loss=0.0383
81
+ step 790000 | 8214s | train_loss=0.0381
82
+ step 800000 | 8384s | train_loss=0.0377 | fid=57.3125
83
+ step 810000 | 8482s | train_loss=0.0374
84
+ step 820000 | 8576s | train_loss=0.0369
85
+ step 830000 | 8673s | train_loss=0.0363
86
+ step 840000 | 8770s | train_loss=0.0372
87
+ step 850000 | 8868s | train_loss=0.0367
88
+ step 860000 | 8965s | train_loss=0.0363
89
+ step 870000 | 9062s | train_loss=0.0353
90
+ step 880000 | 9156s | train_loss=0.0353
91
+ step 890000 | 9253s | train_loss=0.0347
92
+ step 900000 | 9424s | train_loss=0.0350 | fid=57.1111
93
+ step 910000 | 9519s | train_loss=0.0339
94
+ step 920000 | 9615s | train_loss=0.0345
95
+ step 930000 | 9713s | train_loss=0.0342
96
+ step 940000 | 9810s | train_loss=0.0335
97
+ step 950000 | 9907s | train_loss=0.0337
98
+ step 960000 | 10005s | train_loss=0.0335
99
+ step 970000 | 10100s | train_loss=0.0327
100
+ step 980000 | 10195s | train_loss=0.0327
101
+ step 990000 | 10293s | train_loss=0.0328
102
+ step 1000000 | 10463s | train_loss=0.0318 | fid=57.7842
103
+ step 1010000 | 10558s | train_loss=0.0319
104
+ step 1020000 | 10656s | train_loss=0.0318
105
+ step 1030000 | 10753s | train_loss=0.0317
106
+ step 1040000 | 10850s | train_loss=0.0310
107
+ step 1050000 | 10948s | train_loss=0.0314
108
+ step 1060000 | 11044s | train_loss=0.0305
109
+ step 1070000 | 11140s | train_loss=0.0307
110
+ step 1080000 | 11236s | train_loss=0.0304
111
+ step 1090000 | 11333s | train_loss=0.0306
112
+ step 1100000 | 11501s | train_loss=0.0299 | fid=56.8743
113
+ step 1110000 | 11598s | train_loss=0.0303
114
+ step 1120000 | 11696s | train_loss=0.0298
115
+ step 1130000 | 11793s | train_loss=0.0296
116
+ step 1140000 | 11890s | train_loss=0.0294
117
+ step 1150000 | 11988s | train_loss=0.0292
118
+ step 1160000 | 12084s | train_loss=0.0290
119
+ step 1170000 | 12178s | train_loss=0.0290
120
+ step 1180000 | 12276s | train_loss=0.0285
121
+ step 1190000 | 12370s | train_loss=0.0284
122
+ step 1200000 | 12540s | train_loss=0.0286 | fid=56.7023
123
+ step 1210000 | 12637s | train_loss=0.0284
124
+ step 1220000 | 12734s | train_loss=0.0280
125
+ step 1230000 | 12832s | train_loss=0.0277
126
+ step 1240000 | 12929s | train_loss=0.0277
127
+ step 1250000 | 13026s | train_loss=0.0275
128
+ step 1260000 | 13123s | train_loss=0.0275
129
+ step 1270000 | 13218s | train_loss=0.0274
130
+ step 1280000 | 13312s | train_loss=0.0269
131
+ step 1290000 | 13409s | train_loss=0.0268
132
+ step 1300000 | 13580s | train_loss=0.0266 | fid=56.9452
133
+ step 1310000 | 13677s | train_loss=0.0271
134
+ step 1320000 | 13774s | train_loss=0.0266
135
+ step 1330000 | 13871s | train_loss=0.0263
136
+ step 1340000 | 13969s | train_loss=0.0260
137
+ step 1350000 | 14065s | train_loss=0.0259
138
+ step 1360000 | 14160s | train_loss=0.0261
139
+ step 1370000 | 14254s | train_loss=0.0257
140
+ step 1380000 | 14349s | train_loss=0.0263
141
+ step 1390000 | 14446s | train_loss=0.0256
142
+ step 1400000 | 14617s | train_loss=0.0256 | fid=56.7034
143
+ step 1410000 | 14714s | train_loss=0.0251
144
+ step 1420000 | 14811s | train_loss=0.0251
145
+ step 1430000 | 14909s | train_loss=0.0246
146
+ step 1440000 | 15006s | train_loss=0.0252
147
+ step 1450000 | 15100s | train_loss=0.0245
148
+ step 1460000 | 15196s | train_loss=0.0244
149
+ step 1470000 | 15288s | train_loss=0.0242
150
+ step 1480000 | 15385s | train_loss=0.0244
151
+ step 1490000 | 15482s | train_loss=0.0245
152
+ step 1500000 | 15653s | train_loss=0.0245 | fid=56.7846
153
+ step 1510000 | 15751s | train_loss=0.0241
154
+ step 1520000 | 15848s | train_loss=0.0240
155
+ step 1530000 | 15945s | train_loss=0.0236
156
+ step 1540000 | 16041s | train_loss=0.0236
157
+ step 1550000 | 16137s | train_loss=0.0236
158
+ step 1560000 | 16224s | train_loss=0.0239
159
+ step 1570000 | 16321s | train_loss=0.0231
160
+ step 1580000 | 16419s | train_loss=0.0234
161
+ step 1590000 | 16516s | train_loss=0.0229
162
+ step 1600000 | 16686s | train_loss=0.0230 | fid=57.1424
163
+ step 1610000 | 16784s | train_loss=0.0232
164
+ step 1620000 | 16881s | train_loss=0.0228
165
+ step 1630000 | 16979s | train_loss=0.0232
166
+ step 1640000 | 17076s | train_loss=0.0228
167
+ step 1650000 | 17162s | train_loss=0.0225
168
+ step 1660000 | 17258s | train_loss=0.0224
169
+ step 1670000 | 17355s | train_loss=0.0227
170
+ step 1680000 | 17452s | train_loss=0.0220
171
+ step 1690000 | 17549s | train_loss=0.0225
172
+ step 1700000 | 17720s | train_loss=0.0222 | fid=56.4287
173
+ step 1710000 | 17817s | train_loss=0.0217
174
+ step 1720000 | 17915s | train_loss=0.0221
175
+ step 1730000 | 18011s | train_loss=0.0219
176
+ step 1740000 | 18103s | train_loss=0.0215
177
+ step 1750000 | 18193s | train_loss=0.0215
178
+ step 1760000 | 18290s | train_loss=0.0216
179
+ step 1770000 | 18387s | train_loss=0.0216
180
+ step 1780000 | 18485s | train_loss=0.0219
181
+ step 1790000 | 18583s | train_loss=0.0213
182
+ step 1800000 | 18753s | train_loss=0.0209 | fid=56.9210
183
+ step 1810000 | 18851s | train_loss=0.0209
184
+ step 1820000 | 18948s | train_loss=0.0210
185
+ step 1830000 | 19044s | train_loss=0.0209
186
+ step 1840000 | 19129s | train_loss=0.0205
187
+ step 1850000 | 19225s | train_loss=0.0208
188
+ step 1860000 | 19323s | train_loss=0.0204
189
+ step 1870000 | 19420s | train_loss=0.0207
190
+ step 1880000 | 19517s | train_loss=0.0208
191
+ step 1890000 | 19614s | train_loss=0.0204
192
+ step 1900000 | 19784s | train_loss=0.0200 | fid=56.2112
193
+ step 1910000 | 19881s | train_loss=0.0205
194
+ step 1920000 | 19977s | train_loss=0.0202
195
+ step 1930000 | 20066s | train_loss=0.0204
196
+ step 1940000 | 20161s | train_loss=0.0200
197
+ step 1950000 | 20257s | train_loss=0.0200
198
+ step 1960000 | 20354s | train_loss=0.0204
199
+ step 1970000 | 20451s | train_loss=0.0202
200
+ step 1980000 | 20549s | train_loss=0.0197
201
+ step 1990000 | 20646s | train_loss=0.0193
202
+ step 2000000 | 20815s | train_loss=0.0197 | fid=56.9324
203
+ step 2010000 | 20913s | train_loss=0.0195
204
+ step 2020000 | 21009s | train_loss=0.0195
205
+ step 2030000 | 21098s | train_loss=0.0195
206
+ step 2040000 | 21194s | train_loss=0.0190
207
+ step 2050000 | 21291s | train_loss=0.0195
208
+ step 2060000 | 21390s | train_loss=0.0190
209
+ step 2070000 | 21489s | train_loss=0.0195
210
+ step 2080000 | 21587s | train_loss=0.0188
211
+ step 2090000 | 21686s | train_loss=0.0193
212
+ step 2100000 | 21858s | train_loss=0.0190 | fid=56.3060
213
+ step 2110000 | 21955s | train_loss=0.0187
214
+ step 2120000 | 22038s | train_loss=0.0189
215
+ step 2130000 | 22133s | train_loss=0.0184
216
+ step 2140000 | 22230s | train_loss=0.0186
217
+ step 2150000 | 22329s | train_loss=0.0190
218
+ step 2160000 | 22428s | train_loss=0.0184
219
+ step 2170000 | 22527s | train_loss=0.0184
220
+ step 2180000 | 22625s | train_loss=0.0182
221
+ step 2190000 | 22723s | train_loss=0.0181
222
+ step 2200000 | 22895s | train_loss=0.0178 | fid=56.5650
223
+ step 2210000 | 22991s | train_loss=0.0182
224
+ step 2220000 | 23074s | train_loss=0.0181
225
+ step 2230000 | 23171s | train_loss=0.0187
226
+ step 2240000 | 23270s | train_loss=0.0182
227
+ step 2250000 | 23369s | train_loss=0.0179
228
+ step 2260000 | 23467s | train_loss=0.0179
229
+ step 2270000 | 23566s | train_loss=0.0176
230
+ step 2280000 | 23663s | train_loss=0.0179
231
+ step 2290000 | 23760s | train_loss=0.0175
232
+ step 2300000 | 23931s | train_loss=0.0177 | fid=57.2124
233
+ step 2310000 | 24019s | train_loss=0.0176
234
+ step 2320000 | 24115s | train_loss=0.0176
235
+ step 2330000 | 24211s | train_loss=0.0174
236
+ step 2340000 | 24308s | train_loss=0.0180
237
+ step 2350000 | 24405s | train_loss=0.0177
238
+ step 2360000 | 24503s | train_loss=0.0174
239
+ step 2370000 | 24599s | train_loss=0.0170
240
+ step 2380000 | 24696s | train_loss=0.0176
241
+ step 2390000 | 24793s | train_loss=0.0171
242
+ step 2400000 | 24962s | train_loss=0.0171 | fid=56.0559
243
+ step 2410000 | 25051s | train_loss=0.0174
244
+ step 2420000 | 25147s | train_loss=0.0171
245
+ step 2430000 | 25242s | train_loss=0.0175
246
+ step 2440000 | 25339s | train_loss=0.0168
247
+ step 2450000 | 25437s | train_loss=0.0167
248
+ step 2460000 | 25534s | train_loss=0.0171
249
+ step 2470000 | 25631s | train_loss=0.0167
250
+ step 2480000 | 25728s | train_loss=0.0166
251
+ step 2490000 | 25825s | train_loss=0.0162
252
+ step 2500000 | 25984s | train_loss=0.0168 | fid=57.1036
253
+ step 2510000 | 26080s | train_loss=0.0164
254
+ step 2520000 | 26176s | train_loss=0.0164
255
+ step 2530000 | 26273s | train_loss=0.0162
256
+ step 2540000 | 26371s | train_loss=0.0168
257
+ step 2550000 | 26468s | train_loss=0.0163
258
+ step 2560000 | 26565s | train_loss=0.0165
259
+ step 2570000 | 26661s | train_loss=0.0161
260
+ step 2580000 | 26758s | train_loss=0.0161
261
+ step 2590000 | 26855s | train_loss=0.0161
262
+ step 2600000 | 27015s | train_loss=0.0161 | fid=56.3202
263
+ step 2610000 | 27111s | train_loss=0.0160
264
+ step 2620000 | 27207s | train_loss=0.0160
265
+ step 2630000 | 27304s | train_loss=0.0160
266
+ step 2640000 | 27401s | train_loss=0.0158
267
+ step 2650000 | 27498s | train_loss=0.0159
268
+ step 2660000 | 27596s | train_loss=0.0160
269
+ step 2670000 | 27692s | train_loss=0.0157
270
+ step 2680000 | 27789s | train_loss=0.0158
271
+ step 2690000 | 27886s | train_loss=0.0157
272
+ step 2700000 | 28043s | train_loss=0.0159 | fid=57.3615
273
+ step 2710000 | 28141s | train_loss=0.0157
274
+ step 2720000 | 28236s | train_loss=0.0153
275
+ step 2730000 | 28334s | train_loss=0.0153
276
+ step 2740000 | 28430s | train_loss=0.0154
277
+ step 2750000 | 28528s | train_loss=0.0156
278
+ step 2760000 | 28625s | train_loss=0.0156
279
+ step 2770000 | 28721s | train_loss=0.0155
280
+ step 2780000 | 28818s | train_loss=0.0152
281
+ step 2790000 | 28904s | train_loss=0.0151
282
+ step 2800000 | 29073s | train_loss=0.0155 | fid=56.9385
283
+ step 2810000 | 29169s | train_loss=0.0152
284
+ step 2820000 | 29266s | train_loss=0.0153
285
+ step 2830000 | 29364s | train_loss=0.0156
286
+ step 2840000 | 29461s | train_loss=0.0157
287
+ step 2850000 | 29559s | train_loss=0.0151
288
+ step 2860000 | 29656s | train_loss=0.0148
289
+ step 2870000 | 29754s | train_loss=0.0149
290
+ step 2880000 | 29850s | train_loss=0.0148
291
+ step 2890000 | 29945s | train_loss=0.0151
292
+ step 2900000 | 30116s | train_loss=0.0149 | fid=56.5078
293
+ step 2910000 | 30213s | train_loss=0.0148
294
+ step 2920000 | 30312s | train_loss=0.0146
295
+ step 2930000 | 30411s | train_loss=0.0153
296
+ step 2940000 | 30510s | train_loss=0.0147
297
+ step 2950000 | 30607s | train_loss=0.0148
298
+ step 2960000 | 30707s | train_loss=0.0144
299
+ step 2970000 | 30804s | train_loss=0.0145
300
+ step 2980000 | 30890s | train_loss=0.0146
301
+ step 2990000 | 30989s | train_loss=0.0150
302
+ step 3000000 | 31160s | train_loss=0.0147 | fid=56.1263
303
+ step 3010000 | 31256s | train_loss=0.0145
304
+ step 3020000 | 31355s | train_loss=0.0143
305
+ step 3030000 | 31454s | train_loss=0.0145
306
+ step 3040000 | 31553s | train_loss=0.0146
307
+ step 3050000 | 31652s | train_loss=0.0144
308
+ step 3060000 | 31751s | train_loss=0.0142
309
+ step 3070000 | 31841s | train_loss=0.0142
310
+ step 3080000 | 31938s | train_loss=0.0140
311
+ step 3090000 | 32037s | train_loss=0.0143
312
+ step 3100000 | 32206s | train_loss=0.0142 | fid=56.8218
313
+ step 3110000 | 32304s | train_loss=0.0140
314
+ step 3120000 | 32403s | train_loss=0.0143
315
+ step 3130000 | 32502s | train_loss=0.0141
316
+ step 3140000 | 32601s | train_loss=0.0142
317
+ step 3150000 | 32700s | train_loss=0.0137
318
+ step 3160000 | 32793s | train_loss=0.0142
319
+ step 3170000 | 32883s | train_loss=0.0142
320
+ step 3180000 | 32980s | train_loss=0.0143
321
+ step 3190000 | 33073s | train_loss=0.0139
322
+ step 3200000 | 33237s | train_loss=0.0138 | fid=56.4313
323
+ step 3210000 | 33336s | train_loss=0.0137
324
+ step 3220000 | 33435s | train_loss=0.0138
325
+ step 3230000 | 33532s | train_loss=0.0145
326
+ step 3240000 | 33632s | train_loss=0.0142
327
+ step 3250000 | 33731s | train_loss=0.0135
328
+ step 3260000 | 33816s | train_loss=0.0136
329
+ step 3270000 | 33910s | train_loss=0.0137
330
+ step 3280000 | 34008s | train_loss=0.0136
331
+ step 3290000 | 34102s | train_loss=0.0134
332
+ step 3300000 | 34268s | train_loss=0.0135 | fid=56.3616
333
+ step 3310000 | 34367s | train_loss=0.0136
334
+ step 3320000 | 34466s | train_loss=0.0134
335
+ step 3330000 | 34565s | train_loss=0.0132
336
+ step 3340000 | 34664s | train_loss=0.0136
337
+ step 3350000 | 34762s | train_loss=0.0133
338
+ step 3360000 | 34844s | train_loss=0.0133
339
+ step 3370000 | 34941s | train_loss=0.0130
340
+ step 3380000 | 35039s | train_loss=0.0134
341
+ step 3390000 | 35132s | train_loss=0.0132
342
+ step 3400000 | 35298s | train_loss=0.0136 | fid=56.8462
343
+ step 3410000 | 35397s | train_loss=0.0133
344
+ step 3420000 | 35495s | train_loss=0.0132
345
+ step 3430000 | 35594s | train_loss=0.0133
346
+ step 3440000 | 35693s | train_loss=0.0132
347
+ step 3450000 | 35787s | train_loss=0.0130
348
+ step 3460000 | 35874s | train_loss=0.0134
349
+ step 3470000 | 35971s | train_loss=0.0129
350
+ step 3480000 | 36069s | train_loss=0.0130
351
+ step 3490000 | 36160s | train_loss=0.0133
352
+ step 3500000 | 36321s | train_loss=0.0131 | fid=57.0411
353
+ step 3510000 | 36420s | train_loss=0.0127
354
+ step 3520000 | 36519s | train_loss=0.0133
355
+ step 3530000 | 36618s | train_loss=0.0130
356
+ step 3540000 | 36717s | train_loss=0.0127
357
+ step 3550000 | 36805s | train_loss=0.0129
358
+ step 3560000 | 36901s | train_loss=0.0124
359
+ step 3570000 | 37000s | train_loss=0.0126
360
+ step 3580000 | 37098s | train_loss=0.0126
361
+ step 3590000 | 37195s | train_loss=0.0126
362
+ step 3600000 | 37354s | train_loss=0.0126 | fid=55.8085
363
+ step 3610000 | 37453s | train_loss=0.0128
364
+ step 3620000 | 37552s | train_loss=0.0128
365
+ step 3630000 | 37651s | train_loss=0.0127
366
+ step 3640000 | 37749s | train_loss=0.0123
367
+ step 3650000 | 37830s | train_loss=0.0125
368
+ step 3660000 | 37927s | train_loss=0.0122
369
+ step 3670000 | 38025s | train_loss=0.0125
370
+ step 3680000 | 38124s | train_loss=0.0127
371
+ step 3690000 | 38221s | train_loss=0.0121
372
+ step 3700000 | 38386s | train_loss=0.0125 | fid=56.8114
373
+ step 3710000 | 38485s | train_loss=0.0125
374
+ step 3720000 | 38585s | train_loss=0.0123
375
+ step 3730000 | 38683s | train_loss=0.0127
376
+ step 3740000 | 38777s | train_loss=0.0123
377
+ step 3750000 | 38870s | train_loss=0.0122
378
+ step 3760000 | 38967s | train_loss=0.0125
379
+ step 3770000 | 39065s | train_loss=0.0124
380
+ step 3780000 | 39162s | train_loss=0.0121
381
+ step 3790000 | 39260s | train_loss=0.0122
382
+ step 3800000 | 39430s | train_loss=0.0119 | fid=56.4150
383
+ step 3810000 | 39530s | train_loss=0.0121
384
+ step 3820000 | 39629s | train_loss=0.0120
385
+ step 3830000 | 39727s | train_loss=0.0120
386
+ step 3840000 | 39809s | train_loss=0.0120
387
+ step 3850000 | 39904s | train_loss=0.0121
388
+ step 3860000 | 40003s | train_loss=0.0120
389
+ step 3870000 | 40102s | train_loss=0.0120
390
+ step 3880000 | 40199s | train_loss=0.0123
391
+ step 3890000 | 40296s | train_loss=0.0117
392
+ step 3900000 | 40469s | train_loss=0.0118 | fid=56.8252
393
+ step 3910000 | 40568s | train_loss=0.0120
394
+ step 3920000 | 40668s | train_loss=0.0123
395
+ step 3930000 | 40766s | train_loss=0.0120
396
+ step 3940000 | 40847s | train_loss=0.0119
397
+ step 3950000 | 40944s | train_loss=0.0117
398
+ step 3960000 | 41042s | train_loss=0.0117
399
+ step 3970000 | 41142s | train_loss=0.0119
400
+ step 3980000 | 41238s | train_loss=0.0117
401
+ step 3990000 | 41335s | train_loss=0.0117
402
+ step 4000000 | 41508s | train_loss=0.0117 | fid=56.4946
403
+ step 4010000 | 41608s | train_loss=0.0114
404
+ step 4020000 | 41708s | train_loss=0.0119
405
+ step 4030000 | 41801s | train_loss=0.0114
406
+ step 4040000 | 41892s | train_loss=0.0114
407
+ step 4050000 | 41990s | train_loss=0.0115
408
+ step 4060000 | 42090s | train_loss=0.0114
409
+ step 4070000 | 42185s | train_loss=0.0117
410
+ step 4080000 | 42282s | train_loss=0.0113
411
+ step 4090000 | 42382s | train_loss=0.0114
412
+ step 4100000 | 42555s | train_loss=0.0116 | fid=56.2068
413
+ step 4110000 | 42654s | train_loss=0.0113
414
+ step 4120000 | 42754s | train_loss=0.0114
415
+ step 4130000 | 42839s | train_loss=0.0119
416
+ step 4140000 | 42936s | train_loss=0.0112
417
+ step 4150000 | 43036s | train_loss=0.0115
418
+ step 4160000 | 43135s | train_loss=0.0114
419
+ step 4170000 | 43230s | train_loss=0.0116
420
+ step 4180000 | 43327s | train_loss=0.0115
421
+ step 4190000 | 43427s | train_loss=0.0113
422
+ step 4200000 | 43600s | train_loss=0.0114 | fid=56.6341
423
+ step 4210000 | 43700s | train_loss=0.0115
424
+ step 4220000 | 43794s | train_loss=0.0115
425
+ step 4230000 | 43884s | train_loss=0.0110
426
+ step 4240000 | 43981s | train_loss=0.0110
427
+ step 4250000 | 44081s | train_loss=0.0112
428
+ step 4260000 | 44178s | train_loss=0.0112
429
+ step 4270000 | 44277s | train_loss=0.0113
430
+ step 4280000 | 44375s | train_loss=0.0112
431
+ step 4290000 | 44475s | train_loss=0.0115
432
+ step 4300000 | 44648s | train_loss=0.0112 | fid=56.6374
433
+ step 4310000 | 44747s | train_loss=0.0110
434
+ step 4320000 | 44832s | train_loss=0.0111
435
+ step 4330000 | 44928s | train_loss=0.0111
436
+ step 4340000 | 45026s | train_loss=0.0112
437
+ step 4350000 | 45124s | train_loss=0.0107
438
+ step 4360000 | 45222s | train_loss=0.0109
439
+ step 4370000 | 45319s | train_loss=0.0110
440
+ step 4380000 | 45418s | train_loss=0.0107
441
+ step 4390000 | 45518s | train_loss=0.0109
442
+ step 4400000 | 45692s | train_loss=0.0110 | fid=57.6194
443
+ step 4410000 | 45790s | train_loss=0.0108
444
+ step 4420000 | 45877s | train_loss=0.0106
445
+ step 4430000 | 45975s | train_loss=0.0107
446
+ step 4440000 | 46074s | train_loss=0.0109
447
+ step 4450000 | 46173s | train_loss=0.0111
448
+ step 4460000 | 46272s | train_loss=0.0110
449
+ step 4470000 | 46369s | train_loss=0.0102
450
+ step 4480000 | 46468s | train_loss=0.0105
451
+ step 4490000 | 46567s | train_loss=0.0106
452
+ step 4500000 | 46740s | train_loss=0.0106 | fid=57.7695
453
+ step 4510000 | 46828s | train_loss=0.0109
454
+ step 4520000 | 46925s | train_loss=0.0102
455
+ step 4530000 | 47022s | train_loss=0.0105
456
+ step 4540000 | 47121s | train_loss=0.0107
457
+ step 4550000 | 47219s | train_loss=0.0105
458
+ step 4560000 | 47318s | train_loss=0.0105
459
+ step 4570000 | 47417s | train_loss=0.0104
460
+ step 4580000 | 47516s | train_loss=0.0110
461
+ step 4590000 | 47615s | train_loss=0.0105
462
+ step 4600000 | 47788s | train_loss=0.0104 | fid=56.4702
463
+ step 4610000 | 47882s | train_loss=0.0105
464
+ step 4620000 | 47981s | train_loss=0.0106
465
+ step 4630000 | 48079s | train_loss=0.0103
466
+ step 4640000 | 48179s | train_loss=0.0103
467
+ step 4650000 | 48276s | train_loss=0.0108
468
+ step 4660000 | 48374s | train_loss=0.0103
469
+ step 4670000 | 48473s | train_loss=0.0104
470
+ step 4680000 | 48573s | train_loss=0.0103
471
+ step 4690000 | 48672s | train_loss=0.0103
472
+ step 4700000 | 48828s | train_loss=0.0103 | fid=56.7216
473
+ step 4710000 | 48909s | train_loss=0.0103
474
+ step 4720000 | 48989s | train_loss=0.0101
475
+ step 4730000 | 49069s | train_loss=0.0101
476
+ step 4740000 | 49148s | train_loss=0.0107
477
+ step 4750000 | 49229s | train_loss=0.0102
478
+ step 4760000 | 49307s | train_loss=0.0104
479
+ step 4770000 | 49388s | train_loss=0.0101
480
+ step 4780000 | 49469s | train_loss=0.0102
481
+ step 4790000 | 49550s | train_loss=0.0096
482
+ step 4800000 | 49693s | train_loss=0.0100 | fid=56.5474
483
+ step 4810000 | 49775s | train_loss=0.0098
484
+ step 4820000 | 49850s | train_loss=0.0098
485
+ step 4830000 | 49911s | train_loss=0.0100
486
+ step 4840000 | 49959s | train_loss=0.0098
487
+ step 4850000 | 49993s | train_loss=0.0103
488
+ step 4860000 | 50015s | train_loss=0.0100
489
+ step 4870000 | 50038s | train_loss=0.0101
490
+ step 4880000 | 50060s | train_loss=0.0103
491
+ step 4890000 | 50082s | train_loss=0.0099
492
+ step 4900000 | 50123s | train_loss=0.0103 | fid=56.7505
493
+ step 4910000 | 50148s | train_loss=0.0100
494
+ step 4920000 | 50172s | train_loss=0.0100
495
+ step 4930000 | 50196s | train_loss=0.0099
496
+ step 4940000 | 50220s | train_loss=0.0100
497
+ step 4950000 | 50244s | train_loss=0.0102
498
+ step 4960000 | 50267s | train_loss=0.0099
499
+ step 4970000 | 50290s | train_loss=0.0100
500
+ step 4980000 | 50313s | train_loss=0.0099
501
+ step 4990000 | 50336s | train_loss=0.0100
502
+ step 5000000 | 50377s | train_loss=0.0099 | fid=55.9065
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d40/config.json ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "vae_checkpoint": "vae_checkpoints/celeba_resnet_modernloss_d40/vae.pt",
3
+ "out": "results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d40",
4
+ "subset_path": "diagnostics/celeba_diverse_1k_mean_plus_sd/subset_indices.json",
5
+ "train_tar": "/overflow/data/cerberus/celeba_hq_256_partitioned_ws/train/celeba_hq_256-train-0000.tar",
6
+ "n_train": 1000,
7
+ "total_steps": 5000000,
8
+ "eval_interval": 10000,
9
+ "mem_interval": 100000,
10
+ "fid_interval": 100000,
11
+ "fid_n_real": 1000,
12
+ "fid_n_gen": 10000,
13
+ "n_gen_samples": 10000,
14
+ "n_sde_steps": 500,
15
+ "batch_size": 512,
16
+ "hidden": 1024,
17
+ "depth": 5,
18
+ "optimizer": "sgd",
19
+ "lr": 0.001,
20
+ "momentum": 0.8,
21
+ "t_min": 0.01,
22
+ "t_max": 3.0,
23
+ "t_eval": 0.1,
24
+ "seed": 45,
25
+ "device": null
26
+ }
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d40/metrics.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d40/runner.log ADDED
@@ -0,0 +1,502 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ FID real-side stats cached over 1000 diverse train images
2
+ step 1 | 74s | train_loss=0.9971 | fid=275.8813
3
+ step 10000 | 167s | train_loss=0.8480
4
+ step 20000 | 261s | train_loss=0.8191
5
+ step 30000 | 355s | train_loss=0.7638
6
+ step 40000 | 448s | train_loss=0.6788
7
+ step 50000 | 542s | train_loss=0.5490
8
+ step 60000 | 636s | train_loss=0.4041
9
+ step 70000 | 729s | train_loss=0.2958
10
+ step 80000 | 822s | train_loss=0.2308
11
+ step 90000 | 916s | train_loss=0.1923
12
+ step 100000 | 1076s | train_loss=0.1703 | fid=58.3601
13
+ step 110000 | 1168s | train_loss=0.1551
14
+ step 120000 | 1262s | train_loss=0.1418
15
+ step 130000 | 1355s | train_loss=0.1305
16
+ step 140000 | 1448s | train_loss=0.1240
17
+ step 150000 | 1542s | train_loss=0.1157
18
+ step 160000 | 1636s | train_loss=0.1101
19
+ step 170000 | 1730s | train_loss=0.1060
20
+ step 180000 | 1824s | train_loss=0.1013
21
+ step 190000 | 1918s | train_loss=0.0964
22
+ step 200000 | 2079s | train_loss=0.0939 | fid=56.6309
23
+ step 210000 | 2172s | train_loss=0.0899
24
+ step 220000 | 2266s | train_loss=0.0868
25
+ step 230000 | 2360s | train_loss=0.0839
26
+ step 240000 | 2454s | train_loss=0.0800
27
+ step 250000 | 2548s | train_loss=0.0780
28
+ step 260000 | 2642s | train_loss=0.0765
29
+ step 270000 | 2736s | train_loss=0.0735
30
+ step 280000 | 2830s | train_loss=0.0721
31
+ step 290000 | 2922s | train_loss=0.0711
32
+ step 300000 | 3087s | train_loss=0.0679 | fid=56.6406
33
+ step 310000 | 3181s | train_loss=0.0669
34
+ step 320000 | 3274s | train_loss=0.0655
35
+ step 330000 | 3368s | train_loss=0.0648
36
+ step 340000 | 3462s | train_loss=0.0624
37
+ step 350000 | 3557s | train_loss=0.0603
38
+ step 360000 | 3651s | train_loss=0.0594
39
+ step 370000 | 3745s | train_loss=0.0586
40
+ step 380000 | 3838s | train_loss=0.0579
41
+ step 390000 | 3930s | train_loss=0.0570
42
+ step 400000 | 4096s | train_loss=0.0563 | fid=56.6248
43
+ step 410000 | 4190s | train_loss=0.0542
44
+ step 420000 | 4283s | train_loss=0.0542
45
+ step 430000 | 4377s | train_loss=0.0529
46
+ step 440000 | 4471s | train_loss=0.0517
47
+ step 450000 | 4565s | train_loss=0.0515
48
+ step 460000 | 4660s | train_loss=0.0505
49
+ step 470000 | 4753s | train_loss=0.0495
50
+ step 480000 | 4844s | train_loss=0.0489
51
+ step 490000 | 4938s | train_loss=0.0487
52
+ step 500000 | 5104s | train_loss=0.0482 | fid=55.8854
53
+ step 510000 | 5198s | train_loss=0.0474
54
+ step 520000 | 5291s | train_loss=0.0472
55
+ step 530000 | 5385s | train_loss=0.0459
56
+ step 540000 | 5478s | train_loss=0.0459
57
+ step 550000 | 5573s | train_loss=0.0451
58
+ step 560000 | 5667s | train_loss=0.0442
59
+ step 570000 | 5757s | train_loss=0.0435
60
+ step 580000 | 5851s | train_loss=0.0438
61
+ step 590000 | 5946s | train_loss=0.0431
62
+ step 600000 | 6111s | train_loss=0.0426 | fid=54.4745
63
+ step 610000 | 6205s | train_loss=0.0420
64
+ step 620000 | 6300s | train_loss=0.0414
65
+ step 630000 | 6393s | train_loss=0.0413
66
+ step 640000 | 6486s | train_loss=0.0408
67
+ step 650000 | 6581s | train_loss=0.0400
68
+ step 660000 | 6674s | train_loss=0.0401
69
+ step 670000 | 6765s | train_loss=0.0395
70
+ step 680000 | 6859s | train_loss=0.0392
71
+ step 690000 | 6953s | train_loss=0.0396
72
+ step 700000 | 7119s | train_loss=0.0392 | fid=54.3737
73
+ step 710000 | 7212s | train_loss=0.0388
74
+ step 720000 | 7307s | train_loss=0.0380
75
+ step 730000 | 7399s | train_loss=0.0374
76
+ step 740000 | 7493s | train_loss=0.0375
77
+ step 750000 | 7587s | train_loss=0.0372
78
+ step 760000 | 7678s | train_loss=0.0369
79
+ step 770000 | 7772s | train_loss=0.0361
80
+ step 780000 | 7866s | train_loss=0.0361
81
+ step 790000 | 7960s | train_loss=0.0357
82
+ step 800000 | 8125s | train_loss=0.0356 | fid=53.9977
83
+ step 810000 | 8219s | train_loss=0.0351
84
+ step 820000 | 8313s | train_loss=0.0351
85
+ step 830000 | 8406s | train_loss=0.0349
86
+ step 840000 | 8501s | train_loss=0.0345
87
+ step 850000 | 8591s | train_loss=0.0344
88
+ step 860000 | 8685s | train_loss=0.0341
89
+ step 870000 | 8779s | train_loss=0.0338
90
+ step 880000 | 8873s | train_loss=0.0336
91
+ step 890000 | 8967s | train_loss=0.0341
92
+ step 900000 | 9132s | train_loss=0.0337 | fid=53.2102
93
+ step 910000 | 9226s | train_loss=0.0326
94
+ step 920000 | 9320s | train_loss=0.0332
95
+ step 930000 | 9414s | train_loss=0.0327
96
+ step 940000 | 9506s | train_loss=0.0323
97
+ step 950000 | 9597s | train_loss=0.0323
98
+ step 960000 | 9691s | train_loss=0.0317
99
+ step 970000 | 9785s | train_loss=0.0315
100
+ step 980000 | 9879s | train_loss=0.0314
101
+ step 990000 | 9973s | train_loss=0.0310
102
+ step 1000000 | 10138s | train_loss=0.0311 | fid=53.2736
103
+ step 1010000 | 10232s | train_loss=0.0305
104
+ step 1020000 | 10327s | train_loss=0.0308
105
+ step 1030000 | 10421s | train_loss=0.0308
106
+ step 1040000 | 10509s | train_loss=0.0303
107
+ step 1050000 | 10603s | train_loss=0.0307
108
+ step 1060000 | 10698s | train_loss=0.0297
109
+ step 1070000 | 10791s | train_loss=0.0302
110
+ step 1080000 | 10885s | train_loss=0.0294
111
+ step 1090000 | 10980s | train_loss=0.0292
112
+ step 1100000 | 11145s | train_loss=0.0289 | fid=52.9604
113
+ step 1110000 | 11239s | train_loss=0.0289
114
+ step 1120000 | 11333s | train_loss=0.0288
115
+ step 1130000 | 11424s | train_loss=0.0285
116
+ step 1140000 | 11516s | train_loss=0.0283
117
+ step 1150000 | 11611s | train_loss=0.0280
118
+ step 1160000 | 11705s | train_loss=0.0279
119
+ step 1170000 | 11799s | train_loss=0.0281
120
+ step 1180000 | 11893s | train_loss=0.0279
121
+ step 1190000 | 11987s | train_loss=0.0275
122
+ step 1200000 | 12152s | train_loss=0.0278 | fid=52.4570
123
+ step 1210000 | 12246s | train_loss=0.0279
124
+ step 1220000 | 12340s | train_loss=0.0276
125
+ step 1230000 | 12431s | train_loss=0.0271
126
+ step 1240000 | 12526s | train_loss=0.0269
127
+ step 1250000 | 12617s | train_loss=0.0274
128
+ step 1260000 | 12711s | train_loss=0.0267
129
+ step 1270000 | 12805s | train_loss=0.0268
130
+ step 1280000 | 12900s | train_loss=0.0263
131
+ step 1290000 | 12994s | train_loss=0.0264
132
+ step 1300000 | 13160s | train_loss=0.0255 | fid=52.7532
133
+ step 1310000 | 13254s | train_loss=0.0264
134
+ step 1320000 | 13344s | train_loss=0.0257
135
+ step 1330000 | 13438s | train_loss=0.0256
136
+ step 1340000 | 13532s | train_loss=0.0261
137
+ step 1350000 | 13625s | train_loss=0.0263
138
+ step 1360000 | 13719s | train_loss=0.0257
139
+ step 1370000 | 13812s | train_loss=0.0252
140
+ step 1380000 | 13906s | train_loss=0.0253
141
+ step 1390000 | 14001s | train_loss=0.0253
142
+ step 1400000 | 14165s | train_loss=0.0250 | fid=52.0091
143
+ step 1410000 | 14257s | train_loss=0.0245
144
+ step 1420000 | 14350s | train_loss=0.0251
145
+ step 1430000 | 14443s | train_loss=0.0242
146
+ step 1440000 | 14537s | train_loss=0.0242
147
+ step 1450000 | 14629s | train_loss=0.0246
148
+ step 1460000 | 14723s | train_loss=0.0240
149
+ step 1470000 | 14818s | train_loss=0.0240
150
+ step 1480000 | 14912s | train_loss=0.0245
151
+ step 1490000 | 15006s | train_loss=0.0242
152
+ step 1500000 | 15171s | train_loss=0.0236 | fid=51.4374
153
+ step 1510000 | 15260s | train_loss=0.0236
154
+ step 1520000 | 15353s | train_loss=0.0232
155
+ step 1530000 | 15447s | train_loss=0.0234
156
+ step 1540000 | 15541s | train_loss=0.0238
157
+ step 1550000 | 15636s | train_loss=0.0231
158
+ step 1560000 | 15728s | train_loss=0.0230
159
+ step 1570000 | 15822s | train_loss=0.0233
160
+ step 1580000 | 15916s | train_loss=0.0227
161
+ step 1590000 | 16009s | train_loss=0.0226
162
+ step 1600000 | 16171s | train_loss=0.0227 | fid=51.9132
163
+ step 1610000 | 16263s | train_loss=0.0229
164
+ step 1620000 | 16357s | train_loss=0.0225
165
+ step 1630000 | 16452s | train_loss=0.0220
166
+ step 1640000 | 16546s | train_loss=0.0221
167
+ step 1650000 | 16640s | train_loss=0.0221
168
+ step 1660000 | 16733s | train_loss=0.0228
169
+ step 1670000 | 16827s | train_loss=0.0220
170
+ step 1680000 | 16921s | train_loss=0.0218
171
+ step 1690000 | 17014s | train_loss=0.0220
172
+ step 1700000 | 17171s | train_loss=0.0215 | fid=52.4610
173
+ step 1710000 | 17265s | train_loss=0.0217
174
+ step 1720000 | 17359s | train_loss=0.0214
175
+ step 1730000 | 17453s | train_loss=0.0224
176
+ step 1740000 | 17547s | train_loss=0.0214
177
+ step 1750000 | 17641s | train_loss=0.0217
178
+ step 1760000 | 17734s | train_loss=0.0211
179
+ step 1770000 | 17828s | train_loss=0.0211
180
+ step 1780000 | 17922s | train_loss=0.0212
181
+ step 1790000 | 18015s | train_loss=0.0215
182
+ step 1800000 | 18172s | train_loss=0.0209 | fid=52.1867
183
+ step 1810000 | 18266s | train_loss=0.0209
184
+ step 1820000 | 18360s | train_loss=0.0207
185
+ step 1830000 | 18454s | train_loss=0.0210
186
+ step 1840000 | 18548s | train_loss=0.0206
187
+ step 1850000 | 18643s | train_loss=0.0206
188
+ step 1860000 | 18737s | train_loss=0.0206
189
+ step 1870000 | 18829s | train_loss=0.0201
190
+ step 1880000 | 18923s | train_loss=0.0208
191
+ step 1890000 | 19016s | train_loss=0.0202
192
+ step 1900000 | 19171s | train_loss=0.0196 | fid=52.3842
193
+ step 1910000 | 19265s | train_loss=0.0204
194
+ step 1920000 | 19359s | train_loss=0.0198
195
+ step 1930000 | 19453s | train_loss=0.0204
196
+ step 1940000 | 19547s | train_loss=0.0196
197
+ step 1950000 | 19641s | train_loss=0.0198
198
+ step 1960000 | 19735s | train_loss=0.0198
199
+ step 1970000 | 19827s | train_loss=0.0197
200
+ step 1980000 | 19921s | train_loss=0.0197
201
+ step 1990000 | 20013s | train_loss=0.0200
202
+ step 2000000 | 20170s | train_loss=0.0196 | fid=51.9667
203
+ step 2010000 | 20264s | train_loss=0.0191
204
+ step 2020000 | 20358s | train_loss=0.0198
205
+ step 2030000 | 20452s | train_loss=0.0194
206
+ step 2040000 | 20546s | train_loss=0.0191
207
+ step 2050000 | 20640s | train_loss=0.0188
208
+ step 2060000 | 20735s | train_loss=0.0195
209
+ step 2070000 | 20827s | train_loss=0.0191
210
+ step 2080000 | 20922s | train_loss=0.0192
211
+ step 2090000 | 21014s | train_loss=0.0189
212
+ step 2100000 | 21172s | train_loss=0.0188 | fid=52.3143
213
+ step 2110000 | 21267s | train_loss=0.0183
214
+ step 2120000 | 21361s | train_loss=0.0184
215
+ step 2130000 | 21457s | train_loss=0.0190
216
+ step 2140000 | 21551s | train_loss=0.0187
217
+ step 2150000 | 21646s | train_loss=0.0184
218
+ step 2160000 | 21740s | train_loss=0.0179
219
+ step 2170000 | 21835s | train_loss=0.0189
220
+ step 2180000 | 21928s | train_loss=0.0188
221
+ step 2190000 | 22012s | train_loss=0.0182
222
+ step 2200000 | 22171s | train_loss=0.0179 | fid=51.8352
223
+ step 2210000 | 22267s | train_loss=0.0184
224
+ step 2220000 | 22362s | train_loss=0.0185
225
+ step 2230000 | 22457s | train_loss=0.0180
226
+ step 2240000 | 22551s | train_loss=0.0181
227
+ step 2250000 | 22646s | train_loss=0.0181
228
+ step 2260000 | 22739s | train_loss=0.0176
229
+ step 2270000 | 22834s | train_loss=0.0181
230
+ step 2280000 | 22928s | train_loss=0.0181
231
+ step 2290000 | 23015s | train_loss=0.0176
232
+ step 2300000 | 23173s | train_loss=0.0175 | fid=51.8267
233
+ step 2310000 | 23268s | train_loss=0.0171
234
+ step 2320000 | 23362s | train_loss=0.0173
235
+ step 2330000 | 23457s | train_loss=0.0172
236
+ step 2340000 | 23551s | train_loss=0.0173
237
+ step 2350000 | 23645s | train_loss=0.0174
238
+ step 2360000 | 23739s | train_loss=0.0173
239
+ step 2370000 | 23833s | train_loss=0.0176
240
+ step 2380000 | 23927s | train_loss=0.0170
241
+ step 2390000 | 24012s | train_loss=0.0176
242
+ step 2400000 | 24177s | train_loss=0.0167 | fid=51.9540
243
+ step 2410000 | 24271s | train_loss=0.0170
244
+ step 2420000 | 24365s | train_loss=0.0170
245
+ step 2430000 | 24459s | train_loss=0.0169
246
+ step 2440000 | 24553s | train_loss=0.0167
247
+ step 2450000 | 24646s | train_loss=0.0168
248
+ step 2460000 | 24739s | train_loss=0.0171
249
+ step 2470000 | 24833s | train_loss=0.0168
250
+ step 2480000 | 24928s | train_loss=0.0168
251
+ step 2490000 | 25006s | train_loss=0.0165
252
+ step 2500000 | 25171s | train_loss=0.0163 | fid=51.7713
253
+ step 2510000 | 25265s | train_loss=0.0170
254
+ step 2520000 | 25359s | train_loss=0.0160
255
+ step 2530000 | 25453s | train_loss=0.0162
256
+ step 2540000 | 25547s | train_loss=0.0157
257
+ step 2550000 | 25641s | train_loss=0.0162
258
+ step 2560000 | 25735s | train_loss=0.0156
259
+ step 2570000 | 25828s | train_loss=0.0163
260
+ step 2580000 | 25923s | train_loss=0.0166
261
+ step 2590000 | 26003s | train_loss=0.0159
262
+ step 2600000 | 26169s | train_loss=0.0158 | fid=52.1889
263
+ step 2610000 | 26263s | train_loss=0.0159
264
+ step 2620000 | 26357s | train_loss=0.0159
265
+ step 2630000 | 26451s | train_loss=0.0157
266
+ step 2640000 | 26545s | train_loss=0.0156
267
+ step 2650000 | 26638s | train_loss=0.0162
268
+ step 2660000 | 26732s | train_loss=0.0155
269
+ step 2670000 | 26825s | train_loss=0.0160
270
+ step 2680000 | 26916s | train_loss=0.0155
271
+ step 2690000 | 27003s | train_loss=0.0156
272
+ step 2700000 | 27166s | train_loss=0.0158 | fid=51.7881
273
+ step 2710000 | 27260s | train_loss=0.0153
274
+ step 2720000 | 27354s | train_loss=0.0155
275
+ step 2730000 | 27449s | train_loss=0.0150
276
+ step 2740000 | 27542s | train_loss=0.0150
277
+ step 2750000 | 27636s | train_loss=0.0155
278
+ step 2760000 | 27729s | train_loss=0.0152
279
+ step 2770000 | 27823s | train_loss=0.0158
280
+ step 2780000 | 27909s | train_loss=0.0152
281
+ step 2790000 | 27998s | train_loss=0.0153
282
+ step 2800000 | 28162s | train_loss=0.0155 | fid=51.7967
283
+ step 2810000 | 28257s | train_loss=0.0152
284
+ step 2820000 | 28351s | train_loss=0.0147
285
+ step 2830000 | 28444s | train_loss=0.0157
286
+ step 2840000 | 28538s | train_loss=0.0153
287
+ step 2850000 | 28632s | train_loss=0.0151
288
+ step 2860000 | 28725s | train_loss=0.0148
289
+ step 2870000 | 28819s | train_loss=0.0144
290
+ step 2880000 | 28901s | train_loss=0.0146
291
+ step 2890000 | 28995s | train_loss=0.0143
292
+ step 2900000 | 29161s | train_loss=0.0145 | fid=52.3694
293
+ step 2910000 | 29255s | train_loss=0.0146
294
+ step 2920000 | 29349s | train_loss=0.0146
295
+ step 2930000 | 29443s | train_loss=0.0151
296
+ step 2940000 | 29536s | train_loss=0.0144
297
+ step 2950000 | 29630s | train_loss=0.0144
298
+ step 2960000 | 29724s | train_loss=0.0140
299
+ step 2970000 | 29819s | train_loss=0.0141
300
+ step 2980000 | 29907s | train_loss=0.0140
301
+ step 2990000 | 30002s | train_loss=0.0145
302
+ step 3000000 | 30168s | train_loss=0.0149 | fid=52.0317
303
+ step 3010000 | 30264s | train_loss=0.0151
304
+ step 3020000 | 30358s | train_loss=0.0141
305
+ step 3030000 | 30453s | train_loss=0.0142
306
+ step 3040000 | 30548s | train_loss=0.0143
307
+ step 3050000 | 30642s | train_loss=0.0141
308
+ step 3060000 | 30737s | train_loss=0.0142
309
+ step 3070000 | 30825s | train_loss=0.0139
310
+ step 3080000 | 30912s | train_loss=0.0135
311
+ step 3090000 | 31008s | train_loss=0.0139
312
+ step 3100000 | 31172s | train_loss=0.0138 | fid=52.5469
313
+ step 3110000 | 31267s | train_loss=0.0140
314
+ step 3120000 | 31361s | train_loss=0.0135
315
+ step 3130000 | 31456s | train_loss=0.0140
316
+ step 3140000 | 31551s | train_loss=0.0135
317
+ step 3150000 | 31646s | train_loss=0.0135
318
+ step 3160000 | 31741s | train_loss=0.0135
319
+ step 3170000 | 31828s | train_loss=0.0133
320
+ step 3180000 | 31921s | train_loss=0.0140
321
+ step 3190000 | 32016s | train_loss=0.0132
322
+ step 3200000 | 32184s | train_loss=0.0133 | fid=51.7096
323
+ step 3210000 | 32276s | train_loss=0.0134
324
+ step 3220000 | 32371s | train_loss=0.0135
325
+ step 3230000 | 32466s | train_loss=0.0140
326
+ step 3240000 | 32560s | train_loss=0.0141
327
+ step 3250000 | 32656s | train_loss=0.0132
328
+ step 3260000 | 32751s | train_loss=0.0137
329
+ step 3270000 | 32830s | train_loss=0.0132
330
+ step 3280000 | 32924s | train_loss=0.0131
331
+ step 3290000 | 33020s | train_loss=0.0133
332
+ step 3300000 | 33177s | train_loss=0.0128 | fid=53.2340
333
+ step 3310000 | 33271s | train_loss=0.0131
334
+ step 3320000 | 33366s | train_loss=0.0131
335
+ step 3330000 | 33461s | train_loss=0.0133
336
+ step 3340000 | 33555s | train_loss=0.0129
337
+ step 3350000 | 33650s | train_loss=0.0132
338
+ step 3360000 | 33745s | train_loss=0.0127
339
+ step 3370000 | 33825s | train_loss=0.0136
340
+ step 3380000 | 33917s | train_loss=0.0128
341
+ step 3390000 | 34011s | train_loss=0.0128
342
+ step 3400000 | 34171s | train_loss=0.0130 | fid=51.7356
343
+ step 3410000 | 34265s | train_loss=0.0134
344
+ step 3420000 | 34359s | train_loss=0.0127
345
+ step 3430000 | 34454s | train_loss=0.0128
346
+ step 3440000 | 34549s | train_loss=0.0130
347
+ step 3450000 | 34644s | train_loss=0.0126
348
+ step 3460000 | 34739s | train_loss=0.0132
349
+ step 3470000 | 34820s | train_loss=0.0125
350
+ step 3480000 | 34910s | train_loss=0.0131
351
+ step 3490000 | 35004s | train_loss=0.0124
352
+ step 3500000 | 35162s | train_loss=0.0134 | fid=52.5019
353
+ step 3510000 | 35257s | train_loss=0.0126
354
+ step 3520000 | 35350s | train_loss=0.0128
355
+ step 3530000 | 35444s | train_loss=0.0122
356
+ step 3540000 | 35539s | train_loss=0.0124
357
+ step 3550000 | 35634s | train_loss=0.0125
358
+ step 3560000 | 35729s | train_loss=0.0120
359
+ step 3570000 | 35813s | train_loss=0.0120
360
+ step 3580000 | 35901s | train_loss=0.0119
361
+ step 3590000 | 35996s | train_loss=0.0124
362
+ step 3600000 | 36161s | train_loss=0.0126 | fid=51.8156
363
+ step 3610000 | 36246s | train_loss=0.0120
364
+ step 3620000 | 36339s | train_loss=0.0117
365
+ step 3630000 | 36434s | train_loss=0.0126
366
+ step 3640000 | 36529s | train_loss=0.0117
367
+ step 3650000 | 36624s | train_loss=0.0120
368
+ step 3660000 | 36719s | train_loss=0.0118
369
+ step 3670000 | 36803s | train_loss=0.0122
370
+ step 3680000 | 36895s | train_loss=0.0121
371
+ step 3690000 | 36990s | train_loss=0.0120
372
+ step 3700000 | 37158s | train_loss=0.0130 | fid=51.9879
373
+ step 3710000 | 37242s | train_loss=0.0117
374
+ step 3720000 | 37338s | train_loss=0.0121
375
+ step 3730000 | 37430s | train_loss=0.0123
376
+ step 3740000 | 37526s | train_loss=0.0118
377
+ step 3750000 | 37621s | train_loss=0.0120
378
+ step 3760000 | 37716s | train_loss=0.0119
379
+ step 3770000 | 37797s | train_loss=0.0118
380
+ step 3780000 | 37886s | train_loss=0.0117
381
+ step 3790000 | 37980s | train_loss=0.0119
382
+ step 3800000 | 38148s | train_loss=0.0119 | fid=53.0154
383
+ step 3810000 | 38241s | train_loss=0.0114
384
+ step 3820000 | 38332s | train_loss=0.0116
385
+ step 3830000 | 38425s | train_loss=0.0115
386
+ step 3840000 | 38520s | train_loss=0.0116
387
+ step 3850000 | 38615s | train_loss=0.0118
388
+ step 3860000 | 38710s | train_loss=0.0114
389
+ step 3870000 | 38794s | train_loss=0.0115
390
+ step 3880000 | 38887s | train_loss=0.0112
391
+ step 3890000 | 38981s | train_loss=0.0111
392
+ step 3900000 | 39149s | train_loss=0.0115 | fid=51.3210
393
+ step 3910000 | 39245s | train_loss=0.0115
394
+ step 3920000 | 39337s | train_loss=0.0119
395
+ step 3930000 | 39431s | train_loss=0.0111
396
+ step 3940000 | 39526s | train_loss=0.0112
397
+ step 3950000 | 39621s | train_loss=0.0111
398
+ step 3960000 | 39716s | train_loss=0.0115
399
+ step 3970000 | 39797s | train_loss=0.0111
400
+ step 3980000 | 39886s | train_loss=0.0108
401
+ step 3990000 | 39980s | train_loss=0.0114
402
+ step 4000000 | 40149s | train_loss=0.0112 | fid=51.7592
403
+ step 4010000 | 40244s | train_loss=0.0111
404
+ step 4020000 | 40337s | train_loss=0.0112
405
+ step 4030000 | 40433s | train_loss=0.0108
406
+ step 4040000 | 40526s | train_loss=0.0108
407
+ step 4050000 | 40621s | train_loss=0.0109
408
+ step 4060000 | 40716s | train_loss=0.0109
409
+ step 4070000 | 40801s | train_loss=0.0108
410
+ step 4080000 | 40886s | train_loss=0.0108
411
+ step 4090000 | 40980s | train_loss=0.0107
412
+ step 4100000 | 41149s | train_loss=0.0106 | fid=51.3785
413
+ step 4110000 | 41244s | train_loss=0.0109
414
+ step 4120000 | 41337s | train_loss=0.0105
415
+ step 4130000 | 41432s | train_loss=0.0109
416
+ step 4140000 | 41526s | train_loss=0.0108
417
+ step 4150000 | 41621s | train_loss=0.0106
418
+ step 4160000 | 41716s | train_loss=0.0108
419
+ step 4170000 | 41805s | train_loss=0.0105
420
+ step 4180000 | 41893s | train_loss=0.0107
421
+ step 4190000 | 41986s | train_loss=0.0104
422
+ step 4200000 | 42156s | train_loss=0.0107 | fid=51.7140
423
+ step 4210000 | 42249s | train_loss=0.0110
424
+ step 4220000 | 42343s | train_loss=0.0108
425
+ step 4230000 | 42438s | train_loss=0.0105
426
+ step 4240000 | 42534s | train_loss=0.0106
427
+ step 4250000 | 42627s | train_loss=0.0106
428
+ step 4260000 | 42723s | train_loss=0.0107
429
+ step 4270000 | 42809s | train_loss=0.0107
430
+ step 4280000 | 42898s | train_loss=0.0104
431
+ step 4290000 | 42992s | train_loss=0.0102
432
+ step 4300000 | 43161s | train_loss=0.0101 | fid=52.1782
433
+ step 4310000 | 43255s | train_loss=0.0104
434
+ step 4320000 | 43349s | train_loss=0.0108
435
+ step 4330000 | 43444s | train_loss=0.0105
436
+ step 4340000 | 43540s | train_loss=0.0109
437
+ step 4350000 | 43634s | train_loss=0.0101
438
+ step 4360000 | 43729s | train_loss=0.0106
439
+ step 4370000 | 43814s | train_loss=0.0107
440
+ step 4380000 | 43904s | train_loss=0.0102
441
+ step 4390000 | 43998s | train_loss=0.0104
442
+ step 4400000 | 44167s | train_loss=0.0102 | fid=51.4253
443
+ step 4410000 | 44263s | train_loss=0.0102
444
+ step 4420000 | 44356s | train_loss=0.0100
445
+ step 4430000 | 44452s | train_loss=0.0104
446
+ step 4440000 | 44547s | train_loss=0.0100
447
+ step 4450000 | 44642s | train_loss=0.0105
448
+ step 4460000 | 44735s | train_loss=0.0103
449
+ step 4470000 | 44819s | train_loss=0.0102
450
+ step 4480000 | 44908s | train_loss=0.0099
451
+ step 4490000 | 45002s | train_loss=0.0098
452
+ step 4500000 | 45171s | train_loss=0.0099 | fid=52.7478
453
+ step 4510000 | 45267s | train_loss=0.0101
454
+ step 4520000 | 45359s | train_loss=0.0101
455
+ step 4530000 | 45455s | train_loss=0.0107
456
+ step 4540000 | 45550s | train_loss=0.0098
457
+ step 4550000 | 45646s | train_loss=0.0105
458
+ step 4560000 | 45739s | train_loss=0.0096
459
+ step 4570000 | 45826s | train_loss=0.0100
460
+ step 4580000 | 45917s | train_loss=0.0102
461
+ step 4590000 | 46011s | train_loss=0.0104
462
+ step 4600000 | 46180s | train_loss=0.0097 | fid=50.8286
463
+ step 4610000 | 46276s | train_loss=0.0099
464
+ step 4620000 | 46369s | train_loss=0.0102
465
+ step 4630000 | 46464s | train_loss=0.0095
466
+ step 4640000 | 46559s | train_loss=0.0096
467
+ step 4650000 | 46654s | train_loss=0.0095
468
+ step 4660000 | 46747s | train_loss=0.0096
469
+ step 4670000 | 46831s | train_loss=0.0095
470
+ step 4680000 | 46924s | train_loss=0.0102
471
+ step 4690000 | 47018s | train_loss=0.0094
472
+ step 4700000 | 47188s | train_loss=0.0097 | fid=51.9800
473
+ step 4710000 | 47284s | train_loss=0.0098
474
+ step 4720000 | 47378s | train_loss=0.0097
475
+ step 4730000 | 47472s | train_loss=0.0093
476
+ step 4740000 | 47567s | train_loss=0.0109
477
+ step 4750000 | 47662s | train_loss=0.0100
478
+ step 4760000 | 47758s | train_loss=0.0097
479
+ step 4770000 | 47845s | train_loss=0.0094
480
+ step 4780000 | 47939s | train_loss=0.0095
481
+ step 4790000 | 48033s | train_loss=0.0095
482
+ step 4800000 | 48203s | train_loss=0.0095 | fid=51.4121
483
+ step 4810000 | 48299s | train_loss=0.0095
484
+ step 4820000 | 48393s | train_loss=0.0093
485
+ step 4830000 | 48488s | train_loss=0.0093
486
+ step 4840000 | 48583s | train_loss=0.0093
487
+ step 4850000 | 48678s | train_loss=0.0096
488
+ step 4860000 | 48769s | train_loss=0.0091
489
+ step 4870000 | 48845s | train_loss=0.0093
490
+ step 4880000 | 48923s | train_loss=0.0097
491
+ step 4890000 | 48999s | train_loss=0.0094
492
+ step 4900000 | 49139s | train_loss=0.0099 | fid=52.4284
493
+ step 4910000 | 49216s | train_loss=0.0097
494
+ step 4920000 | 49291s | train_loss=0.0091
495
+ step 4930000 | 49369s | train_loss=0.0090
496
+ step 4940000 | 49447s | train_loss=0.0090
497
+ step 4950000 | 49525s | train_loss=0.0096
498
+ step 4960000 | 49603s | train_loss=0.0092
499
+ step 4970000 | 49681s | train_loss=0.0095
500
+ step 4980000 | 49756s | train_loss=0.0092
501
+ step 4990000 | 49835s | train_loss=0.0091
502
+ step 5000000 | 49941s | train_loss=0.0096 | fid=51.4480
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d50/config.json ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "vae_checkpoint": "vae_checkpoints/celeba_resnet_modernloss_d50/vae.pt",
3
+ "out": "results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d50",
4
+ "subset_path": "diagnostics/celeba_diverse_1k_mean_plus_sd/subset_indices.json",
5
+ "train_tar": "/overflow/data/cerberus/celeba_hq_256_partitioned_ws/train/celeba_hq_256-train-0000.tar",
6
+ "n_train": 1000,
7
+ "total_steps": 5000000,
8
+ "eval_interval": 10000,
9
+ "mem_interval": 100000,
10
+ "fid_interval": 100000,
11
+ "fid_n_real": 1000,
12
+ "fid_n_gen": 10000,
13
+ "n_gen_samples": 10000,
14
+ "n_sde_steps": 500,
15
+ "batch_size": 512,
16
+ "hidden": 1024,
17
+ "depth": 5,
18
+ "optimizer": "sgd",
19
+ "lr": 0.001,
20
+ "momentum": 0.8,
21
+ "t_min": 0.01,
22
+ "t_max": 3.0,
23
+ "t_eval": 0.1,
24
+ "seed": 45,
25
+ "device": null
26
+ }
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d50/metrics.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d50/runner.log ADDED
@@ -0,0 +1,502 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ FID real-side stats cached over 1000 diverse train images
2
+ step 1 | 70s | train_loss=1.0033 | fid=242.1853
3
+ step 10000 | 152s | train_loss=0.8568
4
+ step 20000 | 240s | train_loss=0.7990
5
+ step 30000 | 327s | train_loss=0.7114
6
+ step 40000 | 415s | train_loss=0.5788
7
+ step 50000 | 502s | train_loss=0.4155
8
+ step 60000 | 590s | train_loss=0.2905
9
+ step 70000 | 678s | train_loss=0.2224
10
+ step 80000 | 765s | train_loss=0.1832
11
+ step 90000 | 852s | train_loss=0.1617
12
+ step 100000 | 1011s | train_loss=0.1460 | fid=57.2314
13
+ step 110000 | 1095s | train_loss=0.1354
14
+ step 120000 | 1180s | train_loss=0.1266
15
+ step 130000 | 1268s | train_loss=0.1203
16
+ step 140000 | 1355s | train_loss=0.1120
17
+ step 150000 | 1443s | train_loss=0.1071
18
+ step 160000 | 1531s | train_loss=0.1022
19
+ step 170000 | 1619s | train_loss=0.0960
20
+ step 180000 | 1707s | train_loss=0.0926
21
+ step 190000 | 1795s | train_loss=0.0907
22
+ step 200000 | 1953s | train_loss=0.0867 | fid=54.7920
23
+ step 210000 | 2043s | train_loss=0.0830
24
+ step 220000 | 2126s | train_loss=0.0801
25
+ step 230000 | 2213s | train_loss=0.0780
26
+ step 240000 | 2301s | train_loss=0.0767
27
+ step 250000 | 2389s | train_loss=0.0729
28
+ step 260000 | 2477s | train_loss=0.0717
29
+ step 270000 | 2565s | train_loss=0.0698
30
+ step 280000 | 2653s | train_loss=0.0688
31
+ step 290000 | 2741s | train_loss=0.0664
32
+ step 300000 | 2897s | train_loss=0.0649 | fid=54.4773
33
+ step 310000 | 2985s | train_loss=0.0638
34
+ step 320000 | 3071s | train_loss=0.0620
35
+ step 330000 | 3158s | train_loss=0.0610
36
+ step 340000 | 3244s | train_loss=0.0603
37
+ step 350000 | 3332s | train_loss=0.0577
38
+ step 360000 | 3421s | train_loss=0.0572
39
+ step 370000 | 3509s | train_loss=0.0566
40
+ step 380000 | 3597s | train_loss=0.0552
41
+ step 390000 | 3685s | train_loss=0.0555
42
+ step 400000 | 3842s | train_loss=0.0539 | fid=52.0939
43
+ step 410000 | 3931s | train_loss=0.0526
44
+ step 420000 | 4019s | train_loss=0.0519
45
+ step 430000 | 4103s | train_loss=0.0518
46
+ step 440000 | 4191s | train_loss=0.0505
47
+ step 450000 | 4278s | train_loss=0.0503
48
+ step 460000 | 4366s | train_loss=0.0497
49
+ step 470000 | 4454s | train_loss=0.0491
50
+ step 480000 | 4542s | train_loss=0.0490
51
+ step 490000 | 4631s | train_loss=0.0477
52
+ step 500000 | 4789s | train_loss=0.0477 | fid=52.4585
53
+ step 510000 | 4877s | train_loss=0.0471
54
+ step 520000 | 4965s | train_loss=0.0473
55
+ step 530000 | 5052s | train_loss=0.0454
56
+ step 540000 | 5137s | train_loss=0.0452
57
+ step 550000 | 5226s | train_loss=0.0450
58
+ step 560000 | 5312s | train_loss=0.0447
59
+ step 570000 | 5400s | train_loss=0.0439
60
+ step 580000 | 5488s | train_loss=0.0434
61
+ step 590000 | 5576s | train_loss=0.0437
62
+ step 600000 | 5736s | train_loss=0.0426 | fid=50.6226
63
+ step 610000 | 5824s | train_loss=0.0428
64
+ step 620000 | 5912s | train_loss=0.0412
65
+ step 630000 | 6000s | train_loss=0.0428
66
+ step 640000 | 6086s | train_loss=0.0405
67
+ step 650000 | 6172s | train_loss=0.0407
68
+ step 660000 | 6261s | train_loss=0.0405
69
+ step 670000 | 6348s | train_loss=0.0400
70
+ step 680000 | 6436s | train_loss=0.0400
71
+ step 690000 | 6524s | train_loss=0.0394
72
+ step 700000 | 6682s | train_loss=0.0389 | fid=49.0087
73
+ step 710000 | 6770s | train_loss=0.0390
74
+ step 720000 | 6858s | train_loss=0.0377
75
+ step 730000 | 6946s | train_loss=0.0387
76
+ step 740000 | 7035s | train_loss=0.0380
77
+ step 750000 | 7119s | train_loss=0.0375
78
+ step 760000 | 7207s | train_loss=0.0372
79
+ step 770000 | 7295s | train_loss=0.0367
80
+ step 780000 | 7381s | train_loss=0.0370
81
+ step 790000 | 7470s | train_loss=0.0367
82
+ step 800000 | 7630s | train_loss=0.0362 | fid=49.7423
83
+ step 810000 | 7718s | train_loss=0.0358
84
+ step 820000 | 7806s | train_loss=0.0356
85
+ step 830000 | 7894s | train_loss=0.0353
86
+ step 840000 | 7982s | train_loss=0.0349
87
+ step 850000 | 8070s | train_loss=0.0348
88
+ step 860000 | 8155s | train_loss=0.0348
89
+ step 870000 | 8243s | train_loss=0.0348
90
+ step 880000 | 8331s | train_loss=0.0338
91
+ step 890000 | 8418s | train_loss=0.0345
92
+ step 900000 | 8578s | train_loss=0.0336 | fid=49.3932
93
+ step 910000 | 8666s | train_loss=0.0336
94
+ step 920000 | 8754s | train_loss=0.0333
95
+ step 930000 | 8842s | train_loss=0.0333
96
+ step 940000 | 8930s | train_loss=0.0325
97
+ step 950000 | 9019s | train_loss=0.0323
98
+ step 960000 | 9104s | train_loss=0.0327
99
+ step 970000 | 9190s | train_loss=0.0324
100
+ step 980000 | 9278s | train_loss=0.0330
101
+ step 990000 | 9367s | train_loss=0.0312
102
+ step 1000000 | 9523s | train_loss=0.0314 | fid=48.8980
103
+ step 1010000 | 9611s | train_loss=0.0311
104
+ step 1020000 | 9699s | train_loss=0.0311
105
+ step 1030000 | 9787s | train_loss=0.0311
106
+ step 1040000 | 9875s | train_loss=0.0305
107
+ step 1050000 | 9963s | train_loss=0.0306
108
+ step 1060000 | 10050s | train_loss=0.0300
109
+ step 1070000 | 10136s | train_loss=0.0306
110
+ step 1080000 | 10223s | train_loss=0.0299
111
+ step 1090000 | 10312s | train_loss=0.0301
112
+ step 1100000 | 10471s | train_loss=0.0300 | fid=48.6430
113
+ step 1110000 | 10559s | train_loss=0.0293
114
+ step 1120000 | 10647s | train_loss=0.0298
115
+ step 1130000 | 10735s | train_loss=0.0292
116
+ step 1140000 | 10823s | train_loss=0.0300
117
+ step 1150000 | 10911s | train_loss=0.0291
118
+ step 1160000 | 11000s | train_loss=0.0284
119
+ step 1170000 | 11085s | train_loss=0.0279
120
+ step 1180000 | 11172s | train_loss=0.0286
121
+ step 1190000 | 11260s | train_loss=0.0278
122
+ step 1200000 | 11418s | train_loss=0.0282 | fid=47.9624
123
+ step 1210000 | 11505s | train_loss=0.0286
124
+ step 1220000 | 11593s | train_loss=0.0284
125
+ step 1230000 | 11681s | train_loss=0.0277
126
+ step 1240000 | 11769s | train_loss=0.0274
127
+ step 1250000 | 11857s | train_loss=0.0275
128
+ step 1260000 | 11945s | train_loss=0.0269
129
+ step 1270000 | 12032s | train_loss=0.0267
130
+ step 1280000 | 12119s | train_loss=0.0265
131
+ step 1290000 | 12206s | train_loss=0.0272
132
+ step 1300000 | 12365s | train_loss=0.0268 | fid=48.2839
133
+ step 1310000 | 12454s | train_loss=0.0268
134
+ step 1320000 | 12540s | train_loss=0.0261
135
+ step 1330000 | 12628s | train_loss=0.0259
136
+ step 1340000 | 12716s | train_loss=0.0267
137
+ step 1350000 | 12804s | train_loss=0.0259
138
+ step 1360000 | 12893s | train_loss=0.0254
139
+ step 1370000 | 12981s | train_loss=0.0257
140
+ step 1380000 | 13068s | train_loss=0.0252
141
+ step 1390000 | 13155s | train_loss=0.0251
142
+ step 1400000 | 13313s | train_loss=0.0252 | fid=47.6149
143
+ step 1410000 | 13401s | train_loss=0.0252
144
+ step 1420000 | 13490s | train_loss=0.0250
145
+ step 1430000 | 13577s | train_loss=0.0247
146
+ step 1440000 | 13664s | train_loss=0.0250
147
+ step 1450000 | 13752s | train_loss=0.0251
148
+ step 1460000 | 13840s | train_loss=0.0246
149
+ step 1470000 | 13928s | train_loss=0.0250
150
+ step 1480000 | 14015s | train_loss=0.0250
151
+ step 1490000 | 14102s | train_loss=0.0248
152
+ step 1500000 | 14260s | train_loss=0.0246 | fid=47.7566
153
+ step 1510000 | 14348s | train_loss=0.0243
154
+ step 1520000 | 14436s | train_loss=0.0239
155
+ step 1530000 | 14524s | train_loss=0.0237
156
+ step 1540000 | 14612s | train_loss=0.0241
157
+ step 1550000 | 14698s | train_loss=0.0238
158
+ step 1560000 | 14787s | train_loss=0.0232
159
+ step 1570000 | 14875s | train_loss=0.0238
160
+ step 1580000 | 14963s | train_loss=0.0248
161
+ step 1590000 | 15050s | train_loss=0.0233
162
+ step 1600000 | 15215s | train_loss=0.0225 | fid=48.4557
163
+ step 1610000 | 15303s | train_loss=0.0232
164
+ step 1620000 | 15391s | train_loss=0.0230
165
+ step 1630000 | 15479s | train_loss=0.0227
166
+ step 1640000 | 15567s | train_loss=0.0229
167
+ step 1650000 | 15654s | train_loss=0.0231
168
+ step 1660000 | 15742s | train_loss=0.0222
169
+ step 1670000 | 15830s | train_loss=0.0218
170
+ step 1680000 | 15918s | train_loss=0.0224
171
+ step 1690000 | 16005s | train_loss=0.0217
172
+ step 1700000 | 16179s | train_loss=0.0217 | fid=47.3467
173
+ step 1710000 | 16267s | train_loss=0.0224
174
+ step 1720000 | 16355s | train_loss=0.0222
175
+ step 1730000 | 16443s | train_loss=0.0216
176
+ step 1740000 | 16531s | train_loss=0.0218
177
+ step 1750000 | 16619s | train_loss=0.0214
178
+ step 1760000 | 16706s | train_loss=0.0221
179
+ step 1770000 | 16794s | train_loss=0.0216
180
+ step 1780000 | 16882s | train_loss=0.0212
181
+ step 1790000 | 16971s | train_loss=0.0209
182
+ step 1800000 | 17153s | train_loss=0.0213 | fid=47.3519
183
+ step 1810000 | 17240s | train_loss=0.0212
184
+ step 1820000 | 17328s | train_loss=0.0207
185
+ step 1830000 | 17416s | train_loss=0.0210
186
+ step 1840000 | 17504s | train_loss=0.0209
187
+ step 1850000 | 17592s | train_loss=0.0208
188
+ step 1860000 | 17680s | train_loss=0.0204
189
+ step 1870000 | 17767s | train_loss=0.0209
190
+ step 1880000 | 17855s | train_loss=0.0210
191
+ step 1890000 | 17943s | train_loss=0.0206
192
+ step 1900000 | 18125s | train_loss=0.0205 | fid=47.2173
193
+ step 1910000 | 18212s | train_loss=0.0212
194
+ step 1920000 | 18300s | train_loss=0.0199
195
+ step 1930000 | 18388s | train_loss=0.0199
196
+ step 1940000 | 18476s | train_loss=0.0196
197
+ step 1950000 | 18564s | train_loss=0.0203
198
+ step 1960000 | 18652s | train_loss=0.0209
199
+ step 1970000 | 18741s | train_loss=0.0207
200
+ step 1980000 | 18827s | train_loss=0.0199
201
+ step 1990000 | 18915s | train_loss=0.0198
202
+ step 2000000 | 19099s | train_loss=0.0195 | fid=47.6085
203
+ step 2010000 | 19186s | train_loss=0.0193
204
+ step 2020000 | 19274s | train_loss=0.0203
205
+ step 2030000 | 19362s | train_loss=0.0190
206
+ step 2040000 | 19450s | train_loss=0.0192
207
+ step 2050000 | 19538s | train_loss=0.0188
208
+ step 2060000 | 19626s | train_loss=0.0191
209
+ step 2070000 | 19714s | train_loss=0.0195
210
+ step 2080000 | 19801s | train_loss=0.0196
211
+ step 2090000 | 19889s | train_loss=0.0191
212
+ step 2100000 | 20073s | train_loss=0.0187 | fid=48.0229
213
+ step 2110000 | 20162s | train_loss=0.0182
214
+ step 2120000 | 20248s | train_loss=0.0190
215
+ step 2130000 | 20336s | train_loss=0.0193
216
+ step 2140000 | 20424s | train_loss=0.0187
217
+ step 2150000 | 20512s | train_loss=0.0184
218
+ step 2160000 | 20600s | train_loss=0.0184
219
+ step 2170000 | 20688s | train_loss=0.0185
220
+ step 2180000 | 20776s | train_loss=0.0191
221
+ step 2190000 | 20863s | train_loss=0.0181
222
+ step 2200000 | 21041s | train_loss=0.0183 | fid=47.4267
223
+ step 2210000 | 21128s | train_loss=0.0182
224
+ step 2220000 | 21215s | train_loss=0.0184
225
+ step 2230000 | 21303s | train_loss=0.0179
226
+ step 2240000 | 21391s | train_loss=0.0180
227
+ step 2250000 | 21480s | train_loss=0.0184
228
+ step 2260000 | 21568s | train_loss=0.0174
229
+ step 2270000 | 21657s | train_loss=0.0177
230
+ step 2280000 | 21745s | train_loss=0.0178
231
+ step 2290000 | 21834s | train_loss=0.0175
232
+ step 2300000 | 22046s | train_loss=0.0171 | fid=46.7614
233
+ step 2310000 | 22133s | train_loss=0.0171
234
+ step 2320000 | 22220s | train_loss=0.0169
235
+ step 2330000 | 22309s | train_loss=0.0179
236
+ step 2340000 | 22397s | train_loss=0.0168
237
+ step 2350000 | 22486s | train_loss=0.0173
238
+ step 2360000 | 22574s | train_loss=0.0171
239
+ step 2370000 | 22663s | train_loss=0.0178
240
+ step 2380000 | 22751s | train_loss=0.0175
241
+ step 2390000 | 22839s | train_loss=0.0170
242
+ step 2400000 | 23052s | train_loss=0.0164 | fid=46.9243
243
+ step 2410000 | 23139s | train_loss=0.0169
244
+ step 2420000 | 23226s | train_loss=0.0165
245
+ step 2430000 | 23314s | train_loss=0.0163
246
+ step 2440000 | 23403s | train_loss=0.0165
247
+ step 2450000 | 23491s | train_loss=0.0163
248
+ step 2460000 | 23580s | train_loss=0.0169
249
+ step 2470000 | 23667s | train_loss=0.0161
250
+ step 2480000 | 23755s | train_loss=0.0164
251
+ step 2490000 | 23844s | train_loss=0.0166
252
+ step 2500000 | 24018s | train_loss=0.0168 | fid=46.9119
253
+ step 2510000 | 24104s | train_loss=0.0169
254
+ step 2520000 | 24191s | train_loss=0.0155
255
+ step 2530000 | 24279s | train_loss=0.0161
256
+ step 2540000 | 24367s | train_loss=0.0159
257
+ step 2550000 | 24455s | train_loss=0.0160
258
+ step 2560000 | 24543s | train_loss=0.0151
259
+ step 2570000 | 24630s | train_loss=0.0159
260
+ step 2580000 | 24718s | train_loss=0.0160
261
+ step 2590000 | 24806s | train_loss=0.0155
262
+ step 2600000 | 24998s | train_loss=0.0154 | fid=47.2171
263
+ step 2610000 | 25085s | train_loss=0.0157
264
+ step 2620000 | 25172s | train_loss=0.0157
265
+ step 2630000 | 25260s | train_loss=0.0152
266
+ step 2640000 | 25347s | train_loss=0.0150
267
+ step 2650000 | 25436s | train_loss=0.0166
268
+ step 2660000 | 25524s | train_loss=0.0149
269
+ step 2670000 | 25611s | train_loss=0.0154
270
+ step 2680000 | 25700s | train_loss=0.0150
271
+ step 2690000 | 25787s | train_loss=0.0152
272
+ step 2700000 | 25978s | train_loss=0.0151 | fid=47.0906
273
+ step 2710000 | 26064s | train_loss=0.0150
274
+ step 2720000 | 26152s | train_loss=0.0152
275
+ step 2730000 | 26238s | train_loss=0.0145
276
+ step 2740000 | 26326s | train_loss=0.0156
277
+ step 2750000 | 26414s | train_loss=0.0142
278
+ step 2760000 | 26502s | train_loss=0.0146
279
+ step 2770000 | 26590s | train_loss=0.0147
280
+ step 2780000 | 26678s | train_loss=0.0142
281
+ step 2790000 | 26766s | train_loss=0.0145
282
+ step 2800000 | 26947s | train_loss=0.0146 | fid=47.4098
283
+ step 2810000 | 27034s | train_loss=0.0145
284
+ step 2820000 | 27120s | train_loss=0.0144
285
+ step 2830000 | 27207s | train_loss=0.0155
286
+ step 2840000 | 27295s | train_loss=0.0143
287
+ step 2850000 | 27383s | train_loss=0.0149
288
+ step 2860000 | 27471s | train_loss=0.0142
289
+ step 2870000 | 27559s | train_loss=0.0150
290
+ step 2880000 | 27647s | train_loss=0.0141
291
+ step 2890000 | 27734s | train_loss=0.0142
292
+ step 2900000 | 27929s | train_loss=0.0145 | fid=47.1559
293
+ step 2910000 | 28018s | train_loss=0.0139
294
+ step 2920000 | 28102s | train_loss=0.0138
295
+ step 2930000 | 28188s | train_loss=0.0147
296
+ step 2940000 | 28276s | train_loss=0.0136
297
+ step 2950000 | 28364s | train_loss=0.0144
298
+ step 2960000 | 28452s | train_loss=0.0138
299
+ step 2970000 | 28540s | train_loss=0.0137
300
+ step 2980000 | 28628s | train_loss=0.0136
301
+ step 2990000 | 28716s | train_loss=0.0144
302
+ step 3000000 | 28901s | train_loss=0.0137 | fid=47.5040
303
+ step 3010000 | 28989s | train_loss=0.0141
304
+ step 3020000 | 29074s | train_loss=0.0141
305
+ step 3030000 | 29161s | train_loss=0.0133
306
+ step 3040000 | 29249s | train_loss=0.0138
307
+ step 3050000 | 29337s | train_loss=0.0137
308
+ step 3060000 | 29425s | train_loss=0.0134
309
+ step 3070000 | 29513s | train_loss=0.0135
310
+ step 3080000 | 29601s | train_loss=0.0134
311
+ step 3090000 | 29689s | train_loss=0.0133
312
+ step 3100000 | 29860s | train_loss=0.0129 | fid=47.8724
313
+ step 3110000 | 29947s | train_loss=0.0139
314
+ step 3120000 | 30034s | train_loss=0.0133
315
+ step 3130000 | 30121s | train_loss=0.0133
316
+ step 3140000 | 30208s | train_loss=0.0135
317
+ step 3150000 | 30297s | train_loss=0.0127
318
+ step 3160000 | 30385s | train_loss=0.0131
319
+ step 3170000 | 30474s | train_loss=0.0125
320
+ step 3180000 | 30563s | train_loss=0.0134
321
+ step 3190000 | 30651s | train_loss=0.0128
322
+ step 3200000 | 30851s | train_loss=0.0129 | fid=47.3366
323
+ step 3210000 | 30938s | train_loss=0.0133
324
+ step 3220000 | 31027s | train_loss=0.0124
325
+ step 3230000 | 31114s | train_loss=0.0128
326
+ step 3240000 | 31200s | train_loss=0.0132
327
+ step 3250000 | 31288s | train_loss=0.0124
328
+ step 3260000 | 31377s | train_loss=0.0125
329
+ step 3270000 | 31465s | train_loss=0.0125
330
+ step 3280000 | 31554s | train_loss=0.0129
331
+ step 3290000 | 31642s | train_loss=0.0124
332
+ step 3300000 | 31825s | train_loss=0.0128 | fid=46.8399
333
+ step 3310000 | 31912s | train_loss=0.0125
334
+ step 3320000 | 32001s | train_loss=0.0127
335
+ step 3330000 | 32088s | train_loss=0.0126
336
+ step 3340000 | 32177s | train_loss=0.0122
337
+ step 3350000 | 32262s | train_loss=0.0125
338
+ step 3360000 | 32351s | train_loss=0.0130
339
+ step 3370000 | 32439s | train_loss=0.0125
340
+ step 3380000 | 32528s | train_loss=0.0123
341
+ step 3390000 | 32616s | train_loss=0.0124
342
+ step 3400000 | 32829s | train_loss=0.0126 | fid=47.4712
343
+ step 3410000 | 32916s | train_loss=0.0121
344
+ step 3420000 | 33005s | train_loss=0.0123
345
+ step 3430000 | 33084s | train_loss=0.0119
346
+ step 3440000 | 33171s | train_loss=0.0125
347
+ step 3450000 | 33257s | train_loss=0.0122
348
+ step 3460000 | 33346s | train_loss=0.0123
349
+ step 3470000 | 33434s | train_loss=0.0118
350
+ step 3480000 | 33522s | train_loss=0.0120
351
+ step 3490000 | 33611s | train_loss=0.0123
352
+ step 3500000 | 33831s | train_loss=0.0123 | fid=47.7040
353
+ step 3510000 | 33918s | train_loss=0.0117
354
+ step 3520000 | 34006s | train_loss=0.0118
355
+ step 3530000 | 34093s | train_loss=0.0121
356
+ step 3540000 | 34173s | train_loss=0.0118
357
+ step 3550000 | 34262s | train_loss=0.0118
358
+ step 3560000 | 34349s | train_loss=0.0114
359
+ step 3570000 | 34438s | train_loss=0.0116
360
+ step 3580000 | 34526s | train_loss=0.0114
361
+ step 3590000 | 34615s | train_loss=0.0117
362
+ step 3600000 | 34836s | train_loss=0.0116 | fid=46.3962
363
+ step 3610000 | 34923s | train_loss=0.0114
364
+ step 3620000 | 35011s | train_loss=0.0111
365
+ step 3630000 | 35100s | train_loss=0.0118
366
+ step 3640000 | 35178s | train_loss=0.0113
367
+ step 3650000 | 35267s | train_loss=0.0116
368
+ step 3660000 | 35354s | train_loss=0.0111
369
+ step 3670000 | 35442s | train_loss=0.0111
370
+ step 3680000 | 35531s | train_loss=0.0111
371
+ step 3690000 | 35619s | train_loss=0.0114
372
+ step 3700000 | 35833s | train_loss=0.0122 | fid=46.8601
373
+ step 3710000 | 35921s | train_loss=0.0111
374
+ step 3720000 | 36009s | train_loss=0.0114
375
+ step 3730000 | 36098s | train_loss=0.0114
376
+ step 3740000 | 36175s | train_loss=0.0114
377
+ step 3750000 | 36257s | train_loss=0.0111
378
+ step 3760000 | 36344s | train_loss=0.0112
379
+ step 3770000 | 36432s | train_loss=0.0111
380
+ step 3780000 | 36521s | train_loss=0.0110
381
+ step 3790000 | 36610s | train_loss=0.0113
382
+ step 3800000 | 36811s | train_loss=0.0113 | fid=46.6843
383
+ step 3810000 | 36898s | train_loss=0.0108
384
+ step 3820000 | 36987s | train_loss=0.0109
385
+ step 3830000 | 37075s | train_loss=0.0111
386
+ step 3840000 | 37163s | train_loss=0.0111
387
+ step 3850000 | 37241s | train_loss=0.0108
388
+ step 3860000 | 37330s | train_loss=0.0111
389
+ step 3870000 | 37416s | train_loss=0.0111
390
+ step 3880000 | 37505s | train_loss=0.0105
391
+ step 3890000 | 37594s | train_loss=0.0105
392
+ step 3900000 | 37818s | train_loss=0.0108 | fid=46.6478
393
+ step 3910000 | 37906s | train_loss=0.0107
394
+ step 3920000 | 37994s | train_loss=0.0107
395
+ step 3930000 | 38082s | train_loss=0.0106
396
+ step 3940000 | 38169s | train_loss=0.0107
397
+ step 3950000 | 38253s | train_loss=0.0103
398
+ step 3960000 | 38340s | train_loss=0.0103
399
+ step 3970000 | 38427s | train_loss=0.0105
400
+ step 3980000 | 38516s | train_loss=0.0102
401
+ step 3990000 | 38605s | train_loss=0.0108
402
+ step 4000000 | 38797s | train_loss=0.0102 | fid=47.3203
403
+ step 4010000 | 38884s | train_loss=0.0101
404
+ step 4020000 | 38972s | train_loss=0.0102
405
+ step 4030000 | 39061s | train_loss=0.0101
406
+ step 4040000 | 39148s | train_loss=0.0100
407
+ step 4050000 | 39237s | train_loss=0.0102
408
+ step 4060000 | 39323s | train_loss=0.0108
409
+ step 4070000 | 39412s | train_loss=0.0102
410
+ step 4080000 | 39499s | train_loss=0.0105
411
+ step 4090000 | 39588s | train_loss=0.0101
412
+ step 4100000 | 39818s | train_loss=0.0101 | fid=47.4568
413
+ step 4110000 | 39905s | train_loss=0.0102
414
+ step 4120000 | 39994s | train_loss=0.0101
415
+ step 4130000 | 40083s | train_loss=0.0102
416
+ step 4140000 | 40170s | train_loss=0.0100
417
+ step 4150000 | 40257s | train_loss=0.0103
418
+ step 4160000 | 40346s | train_loss=0.0099
419
+ step 4170000 | 40435s | train_loss=0.0099
420
+ step 4180000 | 40522s | train_loss=0.0100
421
+ step 4190000 | 40610s | train_loss=0.0102
422
+ step 4200000 | 40837s | train_loss=0.0102 | fid=46.3896
423
+ step 4210000 | 40924s | train_loss=0.0097
424
+ step 4220000 | 41012s | train_loss=0.0102
425
+ step 4230000 | 41101s | train_loss=0.0097
426
+ step 4240000 | 41188s | train_loss=0.0101
427
+ step 4250000 | 41275s | train_loss=0.0099
428
+ step 4260000 | 41364s | train_loss=0.0104
429
+ step 4270000 | 41453s | train_loss=0.0102
430
+ step 4280000 | 41540s | train_loss=0.0096
431
+ step 4290000 | 41629s | train_loss=0.0098
432
+ step 4300000 | 41833s | train_loss=0.0095 | fid=46.8647
433
+ step 4310000 | 41920s | train_loss=0.0100
434
+ step 4320000 | 42009s | train_loss=0.0097
435
+ step 4330000 | 42098s | train_loss=0.0098
436
+ step 4340000 | 42185s | train_loss=0.0099
437
+ step 4350000 | 42272s | train_loss=0.0095
438
+ step 4360000 | 42361s | train_loss=0.0103
439
+ step 4370000 | 42450s | train_loss=0.0101
440
+ step 4380000 | 42539s | train_loss=0.0096
441
+ step 4390000 | 42626s | train_loss=0.0097
442
+ step 4400000 | 42831s | train_loss=0.0097 | fid=47.2397
443
+ step 4410000 | 42918s | train_loss=0.0099
444
+ step 4420000 | 43007s | train_loss=0.0096
445
+ step 4430000 | 43096s | train_loss=0.0097
446
+ step 4440000 | 43183s | train_loss=0.0093
447
+ step 4450000 | 43271s | train_loss=0.0096
448
+ step 4460000 | 43359s | train_loss=0.0096
449
+ step 4470000 | 43448s | train_loss=0.0096
450
+ step 4480000 | 43537s | train_loss=0.0093
451
+ step 4490000 | 43624s | train_loss=0.0092
452
+ step 4500000 | 43833s | train_loss=0.0097 | fid=47.2448
453
+ step 4510000 | 43922s | train_loss=0.0097
454
+ step 4520000 | 44009s | train_loss=0.0094
455
+ step 4530000 | 44097s | train_loss=0.0095
456
+ step 4540000 | 44185s | train_loss=0.0094
457
+ step 4550000 | 44274s | train_loss=0.0095
458
+ step 4560000 | 44361s | train_loss=0.0091
459
+ step 4570000 | 44450s | train_loss=0.0096
460
+ step 4580000 | 44539s | train_loss=0.0099
461
+ step 4590000 | 44628s | train_loss=0.0091
462
+ step 4600000 | 44848s | train_loss=0.0094 | fid=47.4001
463
+ step 4610000 | 44938s | train_loss=0.0093
464
+ step 4620000 | 45025s | train_loss=0.0092
465
+ step 4630000 | 45113s | train_loss=0.0094
466
+ step 4640000 | 45201s | train_loss=0.0091
467
+ step 4650000 | 45288s | train_loss=0.0094
468
+ step 4660000 | 45376s | train_loss=0.0090
469
+ step 4670000 | 45465s | train_loss=0.0091
470
+ step 4680000 | 45554s | train_loss=0.0095
471
+ step 4690000 | 45644s | train_loss=0.0090
472
+ step 4700000 | 45843s | train_loss=0.0091 | fid=46.9980
473
+ step 4710000 | 45933s | train_loss=0.0091
474
+ step 4720000 | 46020s | train_loss=0.0091
475
+ step 4730000 | 46108s | train_loss=0.0089
476
+ step 4740000 | 46196s | train_loss=0.0096
477
+ step 4750000 | 46285s | train_loss=0.0089
478
+ step 4760000 | 46372s | train_loss=0.0091
479
+ step 4770000 | 46461s | train_loss=0.0089
480
+ step 4780000 | 46550s | train_loss=0.0090
481
+ step 4790000 | 46638s | train_loss=0.0089
482
+ step 4800000 | 46842s | train_loss=0.0088 | fid=46.4566
483
+ step 4810000 | 46931s | train_loss=0.0089
484
+ step 4820000 | 47018s | train_loss=0.0089
485
+ step 4830000 | 47107s | train_loss=0.0087
486
+ step 4840000 | 47194s | train_loss=0.0088
487
+ step 4850000 | 47284s | train_loss=0.0089
488
+ step 4860000 | 47371s | train_loss=0.0087
489
+ step 4870000 | 47460s | train_loss=0.0086
490
+ step 4880000 | 47548s | train_loss=0.0091
491
+ step 4890000 | 47637s | train_loss=0.0085
492
+ step 4900000 | 47809s | train_loss=0.0088 | fid=46.8814
493
+ step 4910000 | 47898s | train_loss=0.0088
494
+ step 4920000 | 47987s | train_loss=0.0084
495
+ step 4930000 | 48074s | train_loss=0.0084
496
+ step 4940000 | 48163s | train_loss=0.0086
497
+ step 4950000 | 48251s | train_loss=0.0085
498
+ step 4960000 | 48338s | train_loss=0.0087
499
+ step 4970000 | 48427s | train_loss=0.0086
500
+ step 4980000 | 48516s | train_loss=0.0087
501
+ step 4990000 | 48605s | train_loss=0.0086
502
+ step 5000000 | 48764s | train_loss=0.0090 | fid=46.5726
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d60/config.json ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "vae_checkpoint": "vae_checkpoints/celeba_resnet_modernloss_d60/vae.pt",
3
+ "out": "results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d60",
4
+ "subset_path": "diagnostics/celeba_diverse_1k_mean_plus_sd/subset_indices.json",
5
+ "train_tar": "/overflow/data/cerberus/celeba_hq_256_partitioned_ws/train/celeba_hq_256-train-0000.tar",
6
+ "n_train": 1000,
7
+ "total_steps": 5000000,
8
+ "eval_interval": 10000,
9
+ "mem_interval": 100000,
10
+ "fid_interval": 100000,
11
+ "fid_n_real": 1000,
12
+ "fid_n_gen": 10000,
13
+ "n_gen_samples": 10000,
14
+ "n_sde_steps": 500,
15
+ "batch_size": 512,
16
+ "hidden": 1024,
17
+ "depth": 5,
18
+ "optimizer": "sgd",
19
+ "lr": 0.001,
20
+ "momentum": 0.8,
21
+ "t_min": 0.01,
22
+ "t_max": 3.0,
23
+ "t_eval": 0.1,
24
+ "seed": 45,
25
+ "device": null
26
+ }
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d60/metrics.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d60/runner.log ADDED
@@ -0,0 +1,502 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ FID real-side stats cached over 1000 diverse train images
2
+ step 1 | 69s | train_loss=1.0004 | fid=302.2531
3
+ step 10000 | 156s | train_loss=0.8377
4
+ step 20000 | 249s | train_loss=0.7605
5
+ step 30000 | 342s | train_loss=0.6517
6
+ step 40000 | 435s | train_loss=0.4969
7
+ step 50000 | 528s | train_loss=0.3412
8
+ step 60000 | 620s | train_loss=0.2415
9
+ step 70000 | 713s | train_loss=0.1928
10
+ step 80000 | 805s | train_loss=0.1620
11
+ step 90000 | 898s | train_loss=0.1466
12
+ step 100000 | 1058s | train_loss=0.1340 | fid=55.4319
13
+ step 110000 | 1146s | train_loss=0.1246
14
+ step 120000 | 1239s | train_loss=0.1180
15
+ step 130000 | 1332s | train_loss=0.1132
16
+ step 140000 | 1424s | train_loss=0.1070
17
+ step 150000 | 1518s | train_loss=0.1007
18
+ step 160000 | 1611s | train_loss=0.0971
19
+ step 170000 | 1704s | train_loss=0.0928
20
+ step 180000 | 1797s | train_loss=0.0901
21
+ step 190000 | 1890s | train_loss=0.0867
22
+ step 200000 | 2052s | train_loss=0.0835 | fid=53.7858
23
+ step 210000 | 2141s | train_loss=0.0817
24
+ step 220000 | 2234s | train_loss=0.0791
25
+ step 230000 | 2327s | train_loss=0.0759
26
+ step 240000 | 2420s | train_loss=0.0741
27
+ step 250000 | 2514s | train_loss=0.0718
28
+ step 260000 | 2607s | train_loss=0.0708
29
+ step 270000 | 2700s | train_loss=0.0693
30
+ step 280000 | 2793s | train_loss=0.0674
31
+ step 290000 | 2884s | train_loss=0.0655
32
+ step 300000 | 3049s | train_loss=0.0633 | fid=51.3606
33
+ step 310000 | 3140s | train_loss=0.0640
34
+ step 320000 | 3232s | train_loss=0.0610
35
+ step 330000 | 3325s | train_loss=0.0607
36
+ step 340000 | 3418s | train_loss=0.0593
37
+ step 350000 | 3512s | train_loss=0.0583
38
+ step 360000 | 3605s | train_loss=0.0572
39
+ step 370000 | 3698s | train_loss=0.0559
40
+ step 380000 | 3791s | train_loss=0.0554
41
+ step 390000 | 3880s | train_loss=0.0547
42
+ step 400000 | 4047s | train_loss=0.0533 | fid=51.0974
43
+ step 410000 | 4138s | train_loss=0.0527
44
+ step 420000 | 4230s | train_loss=0.0530
45
+ step 430000 | 4323s | train_loss=0.0519
46
+ step 440000 | 4416s | train_loss=0.0506
47
+ step 450000 | 4509s | train_loss=0.0499
48
+ step 460000 | 4602s | train_loss=0.0492
49
+ step 470000 | 4695s | train_loss=0.0487
50
+ step 480000 | 4785s | train_loss=0.0480
51
+ step 490000 | 4876s | train_loss=0.0481
52
+ step 500000 | 5042s | train_loss=0.0480 | fid=50.6802
53
+ step 510000 | 5136s | train_loss=0.0466
54
+ step 520000 | 5226s | train_loss=0.0459
55
+ step 530000 | 5317s | train_loss=0.0458
56
+ step 540000 | 5410s | train_loss=0.0449
57
+ step 550000 | 5503s | train_loss=0.0459
58
+ step 560000 | 5596s | train_loss=0.0440
59
+ step 570000 | 5689s | train_loss=0.0438
60
+ step 580000 | 5777s | train_loss=0.0435
61
+ step 590000 | 5870s | train_loss=0.0443
62
+ step 600000 | 6037s | train_loss=0.0435 | fid=48.7920
63
+ step 610000 | 6130s | train_loss=0.0427
64
+ step 620000 | 6221s | train_loss=0.0425
65
+ step 630000 | 6312s | train_loss=0.0414
66
+ step 640000 | 6405s | train_loss=0.0416
67
+ step 650000 | 6498s | train_loss=0.0407
68
+ step 660000 | 6592s | train_loss=0.0402
69
+ step 670000 | 6680s | train_loss=0.0404
70
+ step 680000 | 6771s | train_loss=0.0398
71
+ step 690000 | 6864s | train_loss=0.0391
72
+ step 700000 | 7030s | train_loss=0.0396 | fid=48.2260
73
+ step 710000 | 7123s | train_loss=0.0397
74
+ step 720000 | 7213s | train_loss=0.0385
75
+ step 730000 | 7306s | train_loss=0.0383
76
+ step 740000 | 7397s | train_loss=0.0377
77
+ step 750000 | 7491s | train_loss=0.0381
78
+ step 760000 | 7581s | train_loss=0.0375
79
+ step 770000 | 7671s | train_loss=0.0375
80
+ step 780000 | 7764s | train_loss=0.0366
81
+ step 790000 | 7857s | train_loss=0.0364
82
+ step 800000 | 8023s | train_loss=0.0357 | fid=47.3284
83
+ step 810000 | 8117s | train_loss=0.0362
84
+ step 820000 | 8207s | train_loss=0.0363
85
+ step 830000 | 8300s | train_loss=0.0363
86
+ step 840000 | 8392s | train_loss=0.0354
87
+ step 850000 | 8485s | train_loss=0.0349
88
+ step 860000 | 8572s | train_loss=0.0356
89
+ step 870000 | 8664s | train_loss=0.0350
90
+ step 880000 | 8757s | train_loss=0.0347
91
+ step 890000 | 8850s | train_loss=0.0340
92
+ step 900000 | 9016s | train_loss=0.0341 | fid=46.7722
93
+ step 910000 | 9109s | train_loss=0.0343
94
+ step 920000 | 9202s | train_loss=0.0333
95
+ step 930000 | 9293s | train_loss=0.0337
96
+ step 940000 | 9386s | train_loss=0.0335
97
+ step 950000 | 9475s | train_loss=0.0328
98
+ step 960000 | 9565s | train_loss=0.0326
99
+ step 970000 | 9658s | train_loss=0.0332
100
+ step 980000 | 9752s | train_loss=0.0323
101
+ step 990000 | 9845s | train_loss=0.0319
102
+ step 1000000 | 10011s | train_loss=0.0319 | fid=46.9953
103
+ step 1010000 | 10104s | train_loss=0.0317
104
+ step 1020000 | 10198s | train_loss=0.0312
105
+ step 1030000 | 10288s | train_loss=0.0306
106
+ step 1040000 | 10381s | train_loss=0.0313
107
+ step 1050000 | 10466s | train_loss=0.0311
108
+ step 1060000 | 10559s | train_loss=0.0306
109
+ step 1070000 | 10652s | train_loss=0.0308
110
+ step 1080000 | 10745s | train_loss=0.0303
111
+ step 1090000 | 10838s | train_loss=0.0306
112
+ step 1100000 | 11005s | train_loss=0.0302 | fid=46.1267
113
+ step 1110000 | 11098s | train_loss=0.0296
114
+ step 1120000 | 11191s | train_loss=0.0299
115
+ step 1130000 | 11281s | train_loss=0.0295
116
+ step 1140000 | 11371s | train_loss=0.0304
117
+ step 1150000 | 11459s | train_loss=0.0291
118
+ step 1160000 | 11553s | train_loss=0.0291
119
+ step 1170000 | 11646s | train_loss=0.0283
120
+ step 1180000 | 11739s | train_loss=0.0283
121
+ step 1190000 | 11832s | train_loss=0.0282
122
+ step 1200000 | 11999s | train_loss=0.0285 | fid=45.1804
123
+ step 1210000 | 12092s | train_loss=0.0287
124
+ step 1220000 | 12185s | train_loss=0.0284
125
+ step 1230000 | 12273s | train_loss=0.0279
126
+ step 1240000 | 12363s | train_loss=0.0275
127
+ step 1250000 | 12456s | train_loss=0.0281
128
+ step 1260000 | 12547s | train_loss=0.0269
129
+ step 1270000 | 12640s | train_loss=0.0271
130
+ step 1280000 | 12733s | train_loss=0.0269
131
+ step 1290000 | 12827s | train_loss=0.0273
132
+ step 1300000 | 12993s | train_loss=0.0278 | fid=45.2893
133
+ step 1310000 | 13086s | train_loss=0.0268
134
+ step 1320000 | 13180s | train_loss=0.0267
135
+ step 1330000 | 13268s | train_loss=0.0261
136
+ step 1340000 | 13357s | train_loss=0.0267
137
+ step 1350000 | 13450s | train_loss=0.0261
138
+ step 1360000 | 13542s | train_loss=0.0267
139
+ step 1370000 | 13635s | train_loss=0.0255
140
+ step 1380000 | 13728s | train_loss=0.0257
141
+ step 1390000 | 13821s | train_loss=0.0260
142
+ step 1400000 | 13986s | train_loss=0.0258 | fid=45.0791
143
+ step 1410000 | 14079s | train_loss=0.0255
144
+ step 1420000 | 14168s | train_loss=0.0256
145
+ step 1430000 | 14258s | train_loss=0.0246
146
+ step 1440000 | 14348s | train_loss=0.0247
147
+ step 1450000 | 14441s | train_loss=0.0249
148
+ step 1460000 | 14534s | train_loss=0.0247
149
+ step 1470000 | 14625s | train_loss=0.0253
150
+ step 1480000 | 14718s | train_loss=0.0251
151
+ step 1490000 | 14811s | train_loss=0.0253
152
+ step 1500000 | 14977s | train_loss=0.0245 | fid=44.1431
153
+ step 1510000 | 15070s | train_loss=0.0250
154
+ step 1520000 | 15160s | train_loss=0.0242
155
+ step 1530000 | 15247s | train_loss=0.0242
156
+ step 1540000 | 15334s | train_loss=0.0236
157
+ step 1550000 | 15427s | train_loss=0.0244
158
+ step 1560000 | 15520s | train_loss=0.0240
159
+ step 1570000 | 15611s | train_loss=0.0244
160
+ step 1580000 | 15704s | train_loss=0.0243
161
+ step 1590000 | 15797s | train_loss=0.0230
162
+ step 1600000 | 15964s | train_loss=0.0234 | fid=44.6786
163
+ step 1610000 | 16048s | train_loss=0.0232
164
+ step 1620000 | 16140s | train_loss=0.0237
165
+ step 1630000 | 16227s | train_loss=0.0228
166
+ step 1640000 | 16315s | train_loss=0.0225
167
+ step 1650000 | 16403s | train_loss=0.0229
168
+ step 1660000 | 16496s | train_loss=0.0228
169
+ step 1670000 | 16590s | train_loss=0.0225
170
+ step 1680000 | 16681s | train_loss=0.0222
171
+ step 1690000 | 16775s | train_loss=0.0228
172
+ step 1700000 | 16941s | train_loss=0.0219 | fid=44.2597
173
+ step 1710000 | 17025s | train_loss=0.0223
174
+ step 1720000 | 17111s | train_loss=0.0217
175
+ step 1730000 | 17203s | train_loss=0.0221
176
+ step 1740000 | 17296s | train_loss=0.0218
177
+ step 1750000 | 17380s | train_loss=0.0218
178
+ step 1760000 | 17473s | train_loss=0.0216
179
+ step 1770000 | 17566s | train_loss=0.0221
180
+ step 1780000 | 17658s | train_loss=0.0211
181
+ step 1790000 | 17751s | train_loss=0.0208
182
+ step 1800000 | 17917s | train_loss=0.0214 | fid=44.5868
183
+ step 1810000 | 18004s | train_loss=0.0211
184
+ step 1820000 | 18087s | train_loss=0.0207
185
+ step 1830000 | 18180s | train_loss=0.0212
186
+ step 1840000 | 18273s | train_loss=0.0211
187
+ step 1850000 | 18365s | train_loss=0.0202
188
+ step 1860000 | 18452s | train_loss=0.0206
189
+ step 1870000 | 18545s | train_loss=0.0206
190
+ step 1880000 | 18637s | train_loss=0.0215
191
+ step 1890000 | 18730s | train_loss=0.0205
192
+ step 1900000 | 18891s | train_loss=0.0204 | fid=44.6742
193
+ step 1910000 | 18982s | train_loss=0.0206
194
+ step 1920000 | 19067s | train_loss=0.0201
195
+ step 1930000 | 19160s | train_loss=0.0200
196
+ step 1940000 | 19252s | train_loss=0.0195
197
+ step 1950000 | 19346s | train_loss=0.0202
198
+ step 1960000 | 19428s | train_loss=0.0207
199
+ step 1970000 | 19522s | train_loss=0.0203
200
+ step 1980000 | 19615s | train_loss=0.0197
201
+ step 1990000 | 19706s | train_loss=0.0200
202
+ step 2000000 | 19863s | train_loss=0.0194 | fid=44.2525
203
+ step 2010000 | 19956s | train_loss=0.0190
204
+ step 2020000 | 20039s | train_loss=0.0201
205
+ step 2030000 | 20131s | train_loss=0.0192
206
+ step 2040000 | 20225s | train_loss=0.0194
207
+ step 2050000 | 20318s | train_loss=0.0196
208
+ step 2060000 | 20411s | train_loss=0.0197
209
+ step 2070000 | 20492s | train_loss=0.0196
210
+ step 2080000 | 20585s | train_loss=0.0190
211
+ step 2090000 | 20677s | train_loss=0.0190
212
+ step 2100000 | 20837s | train_loss=0.0184 | fid=44.8157
213
+ step 2110000 | 20928s | train_loss=0.0182
214
+ step 2120000 | 21015s | train_loss=0.0194
215
+ step 2130000 | 21107s | train_loss=0.0193
216
+ step 2140000 | 21200s | train_loss=0.0183
217
+ step 2150000 | 21294s | train_loss=0.0186
218
+ step 2160000 | 21387s | train_loss=0.0184
219
+ step 2170000 | 21475s | train_loss=0.0179
220
+ step 2180000 | 21557s | train_loss=0.0191
221
+ step 2190000 | 21651s | train_loss=0.0183
222
+ step 2200000 | 21798s | train_loss=0.0184 | fid=44.1407
223
+ step 2210000 | 21886s | train_loss=0.0188
224
+ step 2220000 | 21965s | train_loss=0.0183
225
+ step 2230000 | 22058s | train_loss=0.0173
226
+ step 2240000 | 22151s | train_loss=0.0186
227
+ step 2250000 | 22244s | train_loss=0.0173
228
+ step 2260000 | 22338s | train_loss=0.0178
229
+ step 2270000 | 22431s | train_loss=0.0174
230
+ step 2280000 | 22518s | train_loss=0.0172
231
+ step 2290000 | 22604s | train_loss=0.0173
232
+ step 2300000 | 22755s | train_loss=0.0178 | fid=44.5804
233
+ step 2310000 | 22848s | train_loss=0.0168
234
+ step 2320000 | 22927s | train_loss=0.0170
235
+ step 2330000 | 23020s | train_loss=0.0171
236
+ step 2340000 | 23114s | train_loss=0.0168
237
+ step 2350000 | 23207s | train_loss=0.0170
238
+ step 2360000 | 23300s | train_loss=0.0169
239
+ step 2370000 | 23393s | train_loss=0.0177
240
+ step 2380000 | 23487s | train_loss=0.0168
241
+ step 2390000 | 23569s | train_loss=0.0178
242
+ step 2400000 | 23728s | train_loss=0.0168 | fid=44.1256
243
+ step 2410000 | 23821s | train_loss=0.0174
244
+ step 2420000 | 23905s | train_loss=0.0167
245
+ step 2430000 | 23998s | train_loss=0.0167
246
+ step 2440000 | 24090s | train_loss=0.0163
247
+ step 2450000 | 24183s | train_loss=0.0164
248
+ step 2460000 | 24276s | train_loss=0.0172
249
+ step 2470000 | 24369s | train_loss=0.0162
250
+ step 2480000 | 24462s | train_loss=0.0168
251
+ step 2490000 | 24556s | train_loss=0.0160
252
+ step 2500000 | 24701s | train_loss=0.0163 | fid=44.9638
253
+ step 2510000 | 24794s | train_loss=0.0167
254
+ step 2520000 | 24880s | train_loss=0.0156
255
+ step 2530000 | 24973s | train_loss=0.0157
256
+ step 2540000 | 25065s | train_loss=0.0157
257
+ step 2550000 | 25158s | train_loss=0.0160
258
+ step 2560000 | 25251s | train_loss=0.0157
259
+ step 2570000 | 25344s | train_loss=0.0156
260
+ step 2580000 | 25437s | train_loss=0.0160
261
+ step 2590000 | 25530s | train_loss=0.0158
262
+ step 2600000 | 25681s | train_loss=0.0153 | fid=44.4850
263
+ step 2610000 | 25767s | train_loss=0.0154
264
+ step 2620000 | 25853s | train_loss=0.0157
265
+ step 2630000 | 25946s | train_loss=0.0162
266
+ step 2640000 | 26039s | train_loss=0.0152
267
+ step 2650000 | 26132s | train_loss=0.0159
268
+ step 2660000 | 26225s | train_loss=0.0151
269
+ step 2670000 | 26317s | train_loss=0.0156
270
+ step 2680000 | 26410s | train_loss=0.0151
271
+ step 2690000 | 26497s | train_loss=0.0158
272
+ step 2700000 | 26650s | train_loss=0.0146 | fid=44.3605
273
+ step 2710000 | 26735s | train_loss=0.0149
274
+ step 2720000 | 26824s | train_loss=0.0151
275
+ step 2730000 | 26917s | train_loss=0.0143
276
+ step 2740000 | 27010s | train_loss=0.0146
277
+ step 2750000 | 27102s | train_loss=0.0146
278
+ step 2760000 | 27195s | train_loss=0.0147
279
+ step 2770000 | 27288s | train_loss=0.0149
280
+ step 2780000 | 27382s | train_loss=0.0149
281
+ step 2790000 | 27466s | train_loss=0.0144
282
+ step 2800000 | 27632s | train_loss=0.0142 | fid=44.0973
283
+ step 2810000 | 27706s | train_loss=0.0146
284
+ step 2820000 | 27796s | train_loss=0.0143
285
+ step 2830000 | 27889s | train_loss=0.0156
286
+ step 2840000 | 27982s | train_loss=0.0146
287
+ step 2850000 | 28074s | train_loss=0.0148
288
+ step 2860000 | 28167s | train_loss=0.0139
289
+ step 2870000 | 28260s | train_loss=0.0151
290
+ step 2880000 | 28353s | train_loss=0.0138
291
+ step 2890000 | 28436s | train_loss=0.0143
292
+ step 2900000 | 28602s | train_loss=0.0143 | fid=43.9178
293
+ step 2910000 | 28678s | train_loss=0.0137
294
+ step 2920000 | 28766s | train_loss=0.0137
295
+ step 2930000 | 28859s | train_loss=0.0142
296
+ step 2940000 | 28952s | train_loss=0.0142
297
+ step 2950000 | 29045s | train_loss=0.0144
298
+ step 2960000 | 29138s | train_loss=0.0139
299
+ step 2970000 | 29231s | train_loss=0.0140
300
+ step 2980000 | 29324s | train_loss=0.0137
301
+ step 2990000 | 29410s | train_loss=0.0140
302
+ step 3000000 | 29576s | train_loss=0.0139 | fid=44.5452
303
+ step 3010000 | 29660s | train_loss=0.0145
304
+ step 3020000 | 29738s | train_loss=0.0134
305
+ step 3030000 | 29828s | train_loss=0.0133
306
+ step 3040000 | 29921s | train_loss=0.0142
307
+ step 3050000 | 30015s | train_loss=0.0136
308
+ step 3060000 | 30108s | train_loss=0.0136
309
+ step 3070000 | 30202s | train_loss=0.0136
310
+ step 3080000 | 30291s | train_loss=0.0131
311
+ step 3090000 | 30376s | train_loss=0.0130
312
+ step 3100000 | 30541s | train_loss=0.0134 | fid=44.5030
313
+ step 3110000 | 30623s | train_loss=0.0134
314
+ step 3120000 | 30717s | train_loss=0.0129
315
+ step 3130000 | 30791s | train_loss=0.0132
316
+ step 3140000 | 30880s | train_loss=0.0132
317
+ step 3150000 | 30973s | train_loss=0.0127
318
+ step 3160000 | 31066s | train_loss=0.0126
319
+ step 3170000 | 31160s | train_loss=0.0133
320
+ step 3180000 | 31248s | train_loss=0.0133
321
+ step 3190000 | 31335s | train_loss=0.0129
322
+ step 3200000 | 31501s | train_loss=0.0128 | fid=44.4485
323
+ step 3210000 | 31585s | train_loss=0.0129
324
+ step 3220000 | 31679s | train_loss=0.0130
325
+ step 3230000 | 31771s | train_loss=0.0134
326
+ step 3240000 | 31853s | train_loss=0.0125
327
+ step 3250000 | 31946s | train_loss=0.0130
328
+ step 3260000 | 32039s | train_loss=0.0128
329
+ step 3270000 | 32133s | train_loss=0.0124
330
+ step 3280000 | 32218s | train_loss=0.0127
331
+ step 3290000 | 32307s | train_loss=0.0124
332
+ step 3300000 | 32471s | train_loss=0.0123 | fid=44.2625
333
+ step 3310000 | 32550s | train_loss=0.0122
334
+ step 3320000 | 32643s | train_loss=0.0125
335
+ step 3330000 | 32737s | train_loss=0.0126
336
+ step 3340000 | 32827s | train_loss=0.0125
337
+ step 3350000 | 32909s | train_loss=0.0123
338
+ step 3360000 | 33002s | train_loss=0.0121
339
+ step 3370000 | 33095s | train_loss=0.0126
340
+ step 3380000 | 33179s | train_loss=0.0120
341
+ step 3390000 | 33272s | train_loss=0.0121
342
+ step 3400000 | 33440s | train_loss=0.0120 | fid=45.0854
343
+ step 3410000 | 33522s | train_loss=0.0125
344
+ step 3420000 | 33615s | train_loss=0.0122
345
+ step 3430000 | 33709s | train_loss=0.0119
346
+ step 3440000 | 33800s | train_loss=0.0117
347
+ step 3450000 | 33887s | train_loss=0.0121
348
+ step 3460000 | 33967s | train_loss=0.0119
349
+ step 3470000 | 34060s | train_loss=0.0118
350
+ step 3480000 | 34142s | train_loss=0.0123
351
+ step 3490000 | 34236s | train_loss=0.0121
352
+ step 3500000 | 34401s | train_loss=0.0121 | fid=44.0338
353
+ step 3510000 | 34478s | train_loss=0.0117
354
+ step 3520000 | 34571s | train_loss=0.0118
355
+ step 3530000 | 34665s | train_loss=0.0119
356
+ step 3540000 | 34758s | train_loss=0.0120
357
+ step 3550000 | 34850s | train_loss=0.0114
358
+ step 3560000 | 34939s | train_loss=0.0115
359
+ step 3570000 | 35020s | train_loss=0.0114
360
+ step 3580000 | 35099s | train_loss=0.0113
361
+ step 3590000 | 35191s | train_loss=0.0112
362
+ step 3600000 | 35358s | train_loss=0.0118 | fid=44.3330
363
+ step 3610000 | 35438s | train_loss=0.0116
364
+ step 3620000 | 35528s | train_loss=0.0114
365
+ step 3630000 | 35622s | train_loss=0.0117
366
+ step 3640000 | 35715s | train_loss=0.0111
367
+ step 3650000 | 35807s | train_loss=0.0113
368
+ step 3660000 | 35899s | train_loss=0.0109
369
+ step 3670000 | 35987s | train_loss=0.0111
370
+ step 3680000 | 36059s | train_loss=0.0112
371
+ step 3690000 | 36152s | train_loss=0.0114
372
+ step 3700000 | 36318s | train_loss=0.0124 | fid=44.7685
373
+ step 3710000 | 36403s | train_loss=0.0117
374
+ step 3720000 | 36489s | train_loss=0.0112
375
+ step 3730000 | 36583s | train_loss=0.0114
376
+ step 3740000 | 36676s | train_loss=0.0108
377
+ step 3750000 | 36769s | train_loss=0.0112
378
+ step 3760000 | 36861s | train_loss=0.0111
379
+ step 3770000 | 36954s | train_loss=0.0112
380
+ step 3780000 | 37021s | train_loss=0.0116
381
+ step 3790000 | 37113s | train_loss=0.0109
382
+ step 3800000 | 37278s | train_loss=0.0111 | fid=44.5978
383
+ step 3810000 | 37366s | train_loss=0.0108
384
+ step 3820000 | 37453s | train_loss=0.0108
385
+ step 3830000 | 37547s | train_loss=0.0112
386
+ step 3840000 | 37640s | train_loss=0.0113
387
+ step 3850000 | 37733s | train_loss=0.0107
388
+ step 3860000 | 37824s | train_loss=0.0108
389
+ step 3870000 | 37917s | train_loss=0.0107
390
+ step 3880000 | 37998s | train_loss=0.0105
391
+ step 3890000 | 38070s | train_loss=0.0109
392
+ step 3900000 | 38235s | train_loss=0.0107 | fid=44.4200
393
+ step 3910000 | 38322s | train_loss=0.0111
394
+ step 3920000 | 38409s | train_loss=0.0109
395
+ step 3930000 | 38502s | train_loss=0.0109
396
+ step 3940000 | 38596s | train_loss=0.0105
397
+ step 3950000 | 38689s | train_loss=0.0101
398
+ step 3960000 | 38782s | train_loss=0.0103
399
+ step 3970000 | 38873s | train_loss=0.0105
400
+ step 3980000 | 38958s | train_loss=0.0104
401
+ step 3990000 | 39045s | train_loss=0.0112
402
+ step 4000000 | 39192s | train_loss=0.0104 | fid=44.2699
403
+ step 4010000 | 39279s | train_loss=0.0102
404
+ step 4020000 | 39369s | train_loss=0.0104
405
+ step 4030000 | 39462s | train_loss=0.0103
406
+ step 4040000 | 39556s | train_loss=0.0108
407
+ step 4050000 | 39649s | train_loss=0.0108
408
+ step 4060000 | 39742s | train_loss=0.0107
409
+ step 4070000 | 39834s | train_loss=0.0101
410
+ step 4080000 | 39920s | train_loss=0.0106
411
+ step 4090000 | 40000s | train_loss=0.0102
412
+ step 4100000 | 40156s | train_loss=0.0098 | fid=44.2713
413
+ step 4110000 | 40237s | train_loss=0.0101
414
+ step 4120000 | 40326s | train_loss=0.0102
415
+ step 4130000 | 40419s | train_loss=0.0104
416
+ step 4140000 | 40513s | train_loss=0.0098
417
+ step 4150000 | 40606s | train_loss=0.0105
418
+ step 4160000 | 40699s | train_loss=0.0103
419
+ step 4170000 | 40793s | train_loss=0.0098
420
+ step 4180000 | 40883s | train_loss=0.0104
421
+ step 4190000 | 40964s | train_loss=0.0100
422
+ step 4200000 | 41130s | train_loss=0.0100 | fid=44.9403
423
+ step 4210000 | 41202s | train_loss=0.0100
424
+ step 4220000 | 41289s | train_loss=0.0098
425
+ step 4230000 | 41382s | train_loss=0.0099
426
+ step 4240000 | 41475s | train_loss=0.0100
427
+ step 4250000 | 41568s | train_loss=0.0106
428
+ step 4260000 | 41661s | train_loss=0.0101
429
+ step 4270000 | 41754s | train_loss=0.0101
430
+ step 4280000 | 41845s | train_loss=0.0100
431
+ step 4290000 | 41922s | train_loss=0.0097
432
+ step 4300000 | 42087s | train_loss=0.0097 | fid=44.3572
433
+ step 4310000 | 42173s | train_loss=0.0102
434
+ step 4320000 | 42241s | train_loss=0.0102
435
+ step 4330000 | 42335s | train_loss=0.0099
436
+ step 4340000 | 42428s | train_loss=0.0104
437
+ step 4350000 | 42521s | train_loss=0.0094
438
+ step 4360000 | 42615s | train_loss=0.0112
439
+ step 4370000 | 42708s | train_loss=0.0098
440
+ step 4380000 | 42801s | train_loss=0.0097
441
+ step 4390000 | 42883s | train_loss=0.0103
442
+ step 4400000 | 43050s | train_loss=0.0096 | fid=44.5305
443
+ step 4410000 | 43137s | train_loss=0.0102
444
+ step 4420000 | 43221s | train_loss=0.0094
445
+ step 4430000 | 43299s | train_loss=0.0096
446
+ step 4440000 | 43392s | train_loss=0.0095
447
+ step 4450000 | 43485s | train_loss=0.0095
448
+ step 4460000 | 43579s | train_loss=0.0100
449
+ step 4470000 | 43672s | train_loss=0.0099
450
+ step 4480000 | 43764s | train_loss=0.0097
451
+ step 4490000 | 43845s | train_loss=0.0093
452
+ step 4500000 | 44010s | train_loss=0.0103 | fid=44.6985
453
+ step 4510000 | 44100s | train_loss=0.0096
454
+ step 4520000 | 44183s | train_loss=0.0095
455
+ step 4530000 | 44272s | train_loss=0.0094
456
+ step 4540000 | 44356s | train_loss=0.0094
457
+ step 4550000 | 44449s | train_loss=0.0096
458
+ step 4560000 | 44542s | train_loss=0.0092
459
+ step 4570000 | 44635s | train_loss=0.0095
460
+ step 4580000 | 44725s | train_loss=0.0100
461
+ step 4590000 | 44807s | train_loss=0.0094
462
+ step 4600000 | 44972s | train_loss=0.0095 | fid=44.3366
463
+ step 4610000 | 45064s | train_loss=0.0092
464
+ step 4620000 | 45144s | train_loss=0.0098
465
+ step 4630000 | 45237s | train_loss=0.0091
466
+ step 4640000 | 45322s | train_loss=0.0090
467
+ step 4650000 | 45406s | train_loss=0.0098
468
+ step 4660000 | 45500s | train_loss=0.0092
469
+ step 4670000 | 45593s | train_loss=0.0091
470
+ step 4680000 | 45684s | train_loss=0.0092
471
+ step 4690000 | 45766s | train_loss=0.0089
472
+ step 4700000 | 45931s | train_loss=0.0091 | fid=43.8924
473
+ step 4710000 | 46024s | train_loss=0.0090
474
+ step 4720000 | 46104s | train_loss=0.0096
475
+ step 4730000 | 46198s | train_loss=0.0090
476
+ step 4740000 | 46291s | train_loss=0.0098
477
+ step 4750000 | 46376s | train_loss=0.0093
478
+ step 4760000 | 46464s | train_loss=0.0093
479
+ step 4770000 | 46557s | train_loss=0.0088
480
+ step 4780000 | 46647s | train_loss=0.0092
481
+ step 4790000 | 46736s | train_loss=0.0093
482
+ step 4800000 | 46901s | train_loss=0.0090 | fid=43.8631
483
+ step 4810000 | 46994s | train_loss=0.0087
484
+ step 4820000 | 47072s | train_loss=0.0090
485
+ step 4830000 | 47166s | train_loss=0.0094
486
+ step 4840000 | 47259s | train_loss=0.0091
487
+ step 4850000 | 47352s | train_loss=0.0087
488
+ step 4860000 | 47433s | train_loss=0.0089
489
+ step 4870000 | 47524s | train_loss=0.0087
490
+ step 4880000 | 47608s | train_loss=0.0088
491
+ step 4890000 | 47683s | train_loss=0.0088
492
+ step 4900000 | 47825s | train_loss=0.0094 | fid=44.4404
493
+ step 4910000 | 47900s | train_loss=0.0085
494
+ step 4920000 | 47958s | train_loss=0.0086
495
+ step 4930000 | 48016s | train_loss=0.0084
496
+ step 4940000 | 48073s | train_loss=0.0088
497
+ step 4950000 | 48131s | train_loss=0.0089
498
+ step 4960000 | 48189s | train_loss=0.0086
499
+ step 4970000 | 48241s | train_loss=0.0087
500
+ step 4980000 | 48299s | train_loss=0.0085
501
+ step 4990000 | 48356s | train_loss=0.0087
502
+ step 5000000 | 48466s | train_loss=0.0087 | fid=43.9021
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d70/config.json ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "vae_checkpoint": "vae_checkpoints/celeba_resnet_modernloss_d70/vae.pt",
3
+ "out": "results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d70",
4
+ "subset_path": "diagnostics/celeba_diverse_1k_mean_plus_sd/subset_indices.json",
5
+ "train_tar": "/overflow/data/cerberus/celeba_hq_256_partitioned_ws/train/celeba_hq_256-train-0000.tar",
6
+ "n_train": 1000,
7
+ "total_steps": 5000000,
8
+ "eval_interval": 10000,
9
+ "mem_interval": 100000,
10
+ "fid_interval": 100000,
11
+ "fid_n_real": 1000,
12
+ "fid_n_gen": 10000,
13
+ "n_gen_samples": 10000,
14
+ "n_sde_steps": 500,
15
+ "batch_size": 512,
16
+ "hidden": 1024,
17
+ "depth": 5,
18
+ "optimizer": "sgd",
19
+ "lr": 0.001,
20
+ "momentum": 0.8,
21
+ "t_min": 0.01,
22
+ "t_max": 3.0,
23
+ "t_eval": 0.1,
24
+ "seed": 45,
25
+ "device": null
26
+ }
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d70/metrics.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d70/runner.log ADDED
@@ -0,0 +1,502 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ FID real-side stats cached over 1000 diverse train images
2
+ step 1 | 70s | train_loss=1.0024 | fid=282.3384
3
+ step 10000 | 157s | train_loss=0.8289
4
+ step 20000 | 245s | train_loss=0.7344
5
+ step 30000 | 333s | train_loss=0.6098
6
+ step 40000 | 421s | train_loss=0.4462
7
+ step 50000 | 509s | train_loss=0.3035
8
+ step 60000 | 596s | train_loss=0.2181
9
+ step 70000 | 684s | train_loss=0.1771
10
+ step 80000 | 772s | train_loss=0.1550
11
+ step 90000 | 860s | train_loss=0.1378
12
+ step 100000 | 1016s | train_loss=0.1291 | fid=55.5425
13
+ step 110000 | 1099s | train_loss=0.1208
14
+ step 120000 | 1187s | train_loss=0.1131
15
+ step 130000 | 1275s | train_loss=0.1096
16
+ step 140000 | 1363s | train_loss=0.1026
17
+ step 150000 | 1451s | train_loss=0.0996
18
+ step 160000 | 1539s | train_loss=0.0954
19
+ step 170000 | 1627s | train_loss=0.0918
20
+ step 180000 | 1715s | train_loss=0.0879
21
+ step 190000 | 1803s | train_loss=0.0854
22
+ step 200000 | 1958s | train_loss=0.0831 | fid=54.5343
23
+ step 210000 | 2046s | train_loss=0.0809
24
+ step 220000 | 2130s | train_loss=0.0785
25
+ step 230000 | 2218s | train_loss=0.0760
26
+ step 240000 | 2306s | train_loss=0.0737
27
+ step 250000 | 2394s | train_loss=0.0731
28
+ step 260000 | 2482s | train_loss=0.0715
29
+ step 270000 | 2571s | train_loss=0.0685
30
+ step 280000 | 2659s | train_loss=0.0675
31
+ step 290000 | 2747s | train_loss=0.0669
32
+ step 300000 | 2902s | train_loss=0.0656 | fid=50.8327
33
+ step 310000 | 2990s | train_loss=0.0632
34
+ step 320000 | 3077s | train_loss=0.0617
35
+ step 330000 | 3164s | train_loss=0.0609
36
+ step 340000 | 3250s | train_loss=0.0598
37
+ step 350000 | 3339s | train_loss=0.0592
38
+ step 360000 | 3427s | train_loss=0.0579
39
+ step 370000 | 3515s | train_loss=0.0569
40
+ step 380000 | 3603s | train_loss=0.0560
41
+ step 390000 | 3692s | train_loss=0.0560
42
+ step 400000 | 3847s | train_loss=0.0551 | fid=49.2680
43
+ step 410000 | 3935s | train_loss=0.0537
44
+ step 420000 | 4024s | train_loss=0.0527
45
+ step 430000 | 4111s | train_loss=0.0518
46
+ step 440000 | 4197s | train_loss=0.0524
47
+ step 450000 | 4283s | train_loss=0.0517
48
+ step 460000 | 4372s | train_loss=0.0496
49
+ step 470000 | 4460s | train_loss=0.0499
50
+ step 480000 | 4548s | train_loss=0.0493
51
+ step 490000 | 4636s | train_loss=0.0482
52
+ step 500000 | 4793s | train_loss=0.0487 | fid=49.1604
53
+ step 510000 | 4881s | train_loss=0.0472
54
+ step 520000 | 4970s | train_loss=0.0483
55
+ step 530000 | 5057s | train_loss=0.0461
56
+ step 540000 | 5143s | train_loss=0.0454
57
+ step 550000 | 5231s | train_loss=0.0458
58
+ step 560000 | 5317s | train_loss=0.0449
59
+ step 570000 | 5405s | train_loss=0.0450
60
+ step 580000 | 5493s | train_loss=0.0444
61
+ step 590000 | 5582s | train_loss=0.0446
62
+ step 600000 | 5739s | train_loss=0.0443 | fid=48.6509
63
+ step 610000 | 5827s | train_loss=0.0432
64
+ step 620000 | 5915s | train_loss=0.0423
65
+ step 630000 | 6004s | train_loss=0.0427
66
+ step 640000 | 6090s | train_loss=0.0417
67
+ step 650000 | 6176s | train_loss=0.0412
68
+ step 660000 | 6265s | train_loss=0.0413
69
+ step 670000 | 6351s | train_loss=0.0405
70
+ step 680000 | 6439s | train_loss=0.0403
71
+ step 690000 | 6527s | train_loss=0.0403
72
+ step 700000 | 6685s | train_loss=0.0397 | fid=47.4012
73
+ step 710000 | 6773s | train_loss=0.0396
74
+ step 720000 | 6861s | train_loss=0.0386
75
+ step 730000 | 6949s | train_loss=0.0400
76
+ step 740000 | 7037s | train_loss=0.0384
77
+ step 750000 | 7125s | train_loss=0.0389
78
+ step 760000 | 7210s | train_loss=0.0383
79
+ step 770000 | 7299s | train_loss=0.0380
80
+ step 780000 | 7385s | train_loss=0.0381
81
+ step 790000 | 7473s | train_loss=0.0372
82
+ step 800000 | 7629s | train_loss=0.0374 | fid=45.9399
83
+ step 810000 | 7717s | train_loss=0.0376
84
+ step 820000 | 7805s | train_loss=0.0376
85
+ step 830000 | 7893s | train_loss=0.0372
86
+ step 840000 | 7982s | train_loss=0.0361
87
+ step 850000 | 8069s | train_loss=0.0362
88
+ step 860000 | 8157s | train_loss=0.0363
89
+ step 870000 | 8242s | train_loss=0.0352
90
+ step 880000 | 8331s | train_loss=0.0359
91
+ step 890000 | 8417s | train_loss=0.0348
92
+ step 900000 | 8575s | train_loss=0.0364 | fid=46.4216
93
+ step 910000 | 8663s | train_loss=0.0342
94
+ step 920000 | 8751s | train_loss=0.0344
95
+ step 930000 | 8839s | train_loss=0.0348
96
+ step 940000 | 8928s | train_loss=0.0345
97
+ step 950000 | 9015s | train_loss=0.0339
98
+ step 960000 | 9102s | train_loss=0.0338
99
+ step 970000 | 9191s | train_loss=0.0325
100
+ step 980000 | 9277s | train_loss=0.0330
101
+ step 990000 | 9365s | train_loss=0.0327
102
+ step 1000000 | 9519s | train_loss=0.0339 | fid=44.6542
103
+ step 1010000 | 9607s | train_loss=0.0338
104
+ step 1020000 | 9696s | train_loss=0.0327
105
+ step 1030000 | 9784s | train_loss=0.0322
106
+ step 1040000 | 9872s | train_loss=0.0312
107
+ step 1050000 | 9960s | train_loss=0.0319
108
+ step 1060000 | 10047s | train_loss=0.0315
109
+ step 1070000 | 10135s | train_loss=0.0318
110
+ step 1080000 | 10221s | train_loss=0.0304
111
+ step 1090000 | 10309s | train_loss=0.0310
112
+ step 1100000 | 10466s | train_loss=0.0309 | fid=44.0043
113
+ step 1110000 | 10554s | train_loss=0.0306
114
+ step 1120000 | 10642s | train_loss=0.0313
115
+ step 1130000 | 10730s | train_loss=0.0301
116
+ step 1140000 | 10818s | train_loss=0.0310
117
+ step 1150000 | 10907s | train_loss=0.0297
118
+ step 1160000 | 10995s | train_loss=0.0300
119
+ step 1170000 | 11082s | train_loss=0.0292
120
+ step 1180000 | 11170s | train_loss=0.0298
121
+ step 1190000 | 11255s | train_loss=0.0297
122
+ step 1200000 | 11414s | train_loss=0.0289 | fid=43.4392
123
+ step 1210000 | 11500s | train_loss=0.0293
124
+ step 1220000 | 11588s | train_loss=0.0289
125
+ step 1230000 | 11677s | train_loss=0.0280
126
+ step 1240000 | 11765s | train_loss=0.0289
127
+ step 1250000 | 11853s | train_loss=0.0299
128
+ step 1260000 | 11941s | train_loss=0.0285
129
+ step 1270000 | 12028s | train_loss=0.0289
130
+ step 1280000 | 12116s | train_loss=0.0279
131
+ step 1290000 | 12205s | train_loss=0.0283
132
+ step 1300000 | 12357s | train_loss=0.0287 | fid=43.3973
133
+ step 1310000 | 12446s | train_loss=0.0278
134
+ step 1320000 | 12532s | train_loss=0.0273
135
+ step 1330000 | 12620s | train_loss=0.0269
136
+ step 1340000 | 12709s | train_loss=0.0280
137
+ step 1350000 | 12797s | train_loss=0.0280
138
+ step 1360000 | 12885s | train_loss=0.0282
139
+ step 1370000 | 12974s | train_loss=0.0269
140
+ step 1380000 | 13060s | train_loss=0.0263
141
+ step 1390000 | 13149s | train_loss=0.0265
142
+ step 1400000 | 13304s | train_loss=0.0264 | fid=42.6503
143
+ step 1410000 | 13393s | train_loss=0.0260
144
+ step 1420000 | 13481s | train_loss=0.0269
145
+ step 1430000 | 13567s | train_loss=0.0263
146
+ step 1440000 | 13656s | train_loss=0.0257
147
+ step 1450000 | 13744s | train_loss=0.0260
148
+ step 1460000 | 13832s | train_loss=0.0256
149
+ step 1470000 | 13920s | train_loss=0.0259
150
+ step 1480000 | 14007s | train_loss=0.0256
151
+ step 1490000 | 14095s | train_loss=0.0270
152
+ step 1500000 | 14253s | train_loss=0.0256 | fid=42.5217
153
+ step 1510000 | 14338s | train_loss=0.0252
154
+ step 1520000 | 14426s | train_loss=0.0252
155
+ step 1530000 | 14514s | train_loss=0.0245
156
+ step 1540000 | 14600s | train_loss=0.0247
157
+ step 1550000 | 14689s | train_loss=0.0252
158
+ step 1560000 | 14777s | train_loss=0.0249
159
+ step 1570000 | 14865s | train_loss=0.0246
160
+ step 1580000 | 14953s | train_loss=0.0255
161
+ step 1590000 | 15040s | train_loss=0.0241
162
+ step 1600000 | 15209s | train_loss=0.0241 | fid=43.1478
163
+ step 1610000 | 15293s | train_loss=0.0236
164
+ step 1620000 | 15379s | train_loss=0.0240
165
+ step 1630000 | 15467s | train_loss=0.0237
166
+ step 1640000 | 15556s | train_loss=0.0235
167
+ step 1650000 | 15642s | train_loss=0.0241
168
+ step 1660000 | 15730s | train_loss=0.0240
169
+ step 1670000 | 15818s | train_loss=0.0234
170
+ step 1680000 | 15907s | train_loss=0.0239
171
+ step 1690000 | 15993s | train_loss=0.0236
172
+ step 1700000 | 16163s | train_loss=0.0229 | fid=42.1207
173
+ step 1710000 | 16251s | train_loss=0.0229
174
+ step 1720000 | 16331s | train_loss=0.0236
175
+ step 1730000 | 16418s | train_loss=0.0232
176
+ step 1740000 | 16506s | train_loss=0.0229
177
+ step 1750000 | 16594s | train_loss=0.0224
178
+ step 1760000 | 16681s | train_loss=0.0229
179
+ step 1770000 | 16769s | train_loss=0.0228
180
+ step 1780000 | 16858s | train_loss=0.0221
181
+ step 1790000 | 16945s | train_loss=0.0223
182
+ step 1800000 | 17118s | train_loss=0.0224 | fid=43.0620
183
+ step 1810000 | 17206s | train_loss=0.0227
184
+ step 1820000 | 17294s | train_loss=0.0220
185
+ step 1830000 | 17373s | train_loss=0.0222
186
+ step 1840000 | 17462s | train_loss=0.0216
187
+ step 1850000 | 17550s | train_loss=0.0213
188
+ step 1860000 | 17637s | train_loss=0.0212
189
+ step 1870000 | 17725s | train_loss=0.0213
190
+ step 1880000 | 17813s | train_loss=0.0222
191
+ step 1890000 | 17902s | train_loss=0.0212
192
+ step 1900000 | 18080s | train_loss=0.0215 | fid=43.0240
193
+ step 1910000 | 18168s | train_loss=0.0217
194
+ step 1920000 | 18256s | train_loss=0.0220
195
+ step 1930000 | 18345s | train_loss=0.0208
196
+ step 1940000 | 18426s | train_loss=0.0204
197
+ step 1950000 | 18514s | train_loss=0.0209
198
+ step 1960000 | 18602s | train_loss=0.0212
199
+ step 1970000 | 18689s | train_loss=0.0215
200
+ step 1980000 | 18777s | train_loss=0.0208
201
+ step 1990000 | 18865s | train_loss=0.0204
202
+ step 2000000 | 19035s | train_loss=0.0210 | fid=42.7086
203
+ step 2010000 | 19123s | train_loss=0.0202
204
+ step 2020000 | 19211s | train_loss=0.0201
205
+ step 2030000 | 19300s | train_loss=0.0204
206
+ step 2040000 | 19387s | train_loss=0.0196
207
+ step 2050000 | 19466s | train_loss=0.0195
208
+ step 2060000 | 19554s | train_loss=0.0203
209
+ step 2070000 | 19643s | train_loss=0.0204
210
+ step 2080000 | 19729s | train_loss=0.0203
211
+ step 2090000 | 19817s | train_loss=0.0198
212
+ step 2100000 | 19998s | train_loss=0.0192 | fid=42.2036
213
+ step 2110000 | 20087s | train_loss=0.0192
214
+ step 2120000 | 20175s | train_loss=0.0199
215
+ step 2130000 | 20263s | train_loss=0.0196
216
+ step 2140000 | 20351s | train_loss=0.0190
217
+ step 2150000 | 20434s | train_loss=0.0189
218
+ step 2160000 | 20516s | train_loss=0.0193
219
+ step 2170000 | 20604s | train_loss=0.0192
220
+ step 2180000 | 20690s | train_loss=0.0196
221
+ step 2190000 | 20776s | train_loss=0.0187
222
+ step 2200000 | 20953s | train_loss=0.0185 | fid=43.2272
223
+ step 2210000 | 21041s | train_loss=0.0192
224
+ step 2220000 | 21130s | train_loss=0.0196
225
+ step 2230000 | 21218s | train_loss=0.0185
226
+ step 2240000 | 21306s | train_loss=0.0184
227
+ step 2250000 | 21394s | train_loss=0.0181
228
+ step 2260000 | 21477s | train_loss=0.0183
229
+ step 2270000 | 21554s | train_loss=0.0190
230
+ step 2280000 | 21643s | train_loss=0.0187
231
+ step 2290000 | 21723s | train_loss=0.0183
232
+ step 2300000 | 21943s | train_loss=0.0184 | fid=42.5980
233
+ step 2310000 | 22031s | train_loss=0.0179
234
+ step 2320000 | 22119s | train_loss=0.0178
235
+ step 2330000 | 22207s | train_loss=0.0184
236
+ step 2340000 | 22296s | train_loss=0.0177
237
+ step 2350000 | 22384s | train_loss=0.0182
238
+ step 2360000 | 22473s | train_loss=0.0178
239
+ step 2370000 | 22547s | train_loss=0.0192
240
+ step 2380000 | 22635s | train_loss=0.0177
241
+ step 2390000 | 22711s | train_loss=0.0186
242
+ step 2400000 | 22915s | train_loss=0.0173 | fid=42.3795
243
+ step 2410000 | 23003s | train_loss=0.0178
244
+ step 2420000 | 23091s | train_loss=0.0174
245
+ step 2430000 | 23180s | train_loss=0.0173
246
+ step 2440000 | 23268s | train_loss=0.0173
247
+ step 2450000 | 23356s | train_loss=0.0178
248
+ step 2460000 | 23444s | train_loss=0.0180
249
+ step 2470000 | 23531s | train_loss=0.0175
250
+ step 2480000 | 23609s | train_loss=0.0170
251
+ step 2490000 | 23691s | train_loss=0.0170
252
+ step 2500000 | 23873s | train_loss=0.0172 | fid=41.9078
253
+ step 2510000 | 23962s | train_loss=0.0178
254
+ step 2520000 | 24049s | train_loss=0.0163
255
+ step 2530000 | 24138s | train_loss=0.0170
256
+ step 2540000 | 24226s | train_loss=0.0165
257
+ step 2550000 | 24314s | train_loss=0.0163
258
+ step 2560000 | 24402s | train_loss=0.0165
259
+ step 2570000 | 24490s | train_loss=0.0163
260
+ step 2580000 | 24575s | train_loss=0.0172
261
+ step 2590000 | 24648s | train_loss=0.0167
262
+ step 2600000 | 24821s | train_loss=0.0165 | fid=41.9370
263
+ step 2610000 | 24910s | train_loss=0.0172
264
+ step 2620000 | 24997s | train_loss=0.0164
265
+ step 2630000 | 25085s | train_loss=0.0168
266
+ step 2640000 | 25173s | train_loss=0.0163
267
+ step 2650000 | 25262s | train_loss=0.0173
268
+ step 2660000 | 25350s | train_loss=0.0166
269
+ step 2670000 | 25438s | train_loss=0.0166
270
+ step 2680000 | 25526s | train_loss=0.0159
271
+ step 2690000 | 25603s | train_loss=0.0164
272
+ step 2700000 | 25792s | train_loss=0.0159 | fid=42.2106
273
+ step 2710000 | 25880s | train_loss=0.0155
274
+ step 2720000 | 25968s | train_loss=0.0161
275
+ step 2730000 | 26056s | train_loss=0.0154
276
+ step 2740000 | 26144s | train_loss=0.0157
277
+ step 2750000 | 26232s | train_loss=0.0156
278
+ step 2760000 | 26320s | train_loss=0.0156
279
+ step 2770000 | 26408s | train_loss=0.0158
280
+ step 2780000 | 26491s | train_loss=0.0154
281
+ step 2790000 | 26574s | train_loss=0.0165
282
+ step 2800000 | 26746s | train_loss=0.0152 | fid=41.9297
283
+ step 2810000 | 26834s | train_loss=0.0159
284
+ step 2820000 | 26922s | train_loss=0.0157
285
+ step 2830000 | 27010s | train_loss=0.0170
286
+ step 2840000 | 27098s | train_loss=0.0155
287
+ step 2850000 | 27186s | train_loss=0.0164
288
+ step 2860000 | 27274s | train_loss=0.0150
289
+ step 2870000 | 27362s | train_loss=0.0153
290
+ step 2880000 | 27446s | train_loss=0.0150
291
+ step 2890000 | 27530s | train_loss=0.0157
292
+ step 2900000 | 27708s | train_loss=0.0160 | fid=42.4080
293
+ step 2910000 | 27794s | train_loss=0.0149
294
+ step 2920000 | 27883s | train_loss=0.0148
295
+ step 2930000 | 27971s | train_loss=0.0157
296
+ step 2940000 | 28059s | train_loss=0.0151
297
+ step 2950000 | 28147s | train_loss=0.0159
298
+ step 2960000 | 28235s | train_loss=0.0153
299
+ step 2970000 | 28323s | train_loss=0.0147
300
+ step 2980000 | 28405s | train_loss=0.0151
301
+ step 2990000 | 28489s | train_loss=0.0158
302
+ step 3000000 | 28670s | train_loss=0.0159 | fid=41.9829
303
+ step 3010000 | 28752s | train_loss=0.0151
304
+ step 3020000 | 28840s | train_loss=0.0149
305
+ step 3030000 | 28928s | train_loss=0.0148
306
+ step 3040000 | 29016s | train_loss=0.0150
307
+ step 3050000 | 29104s | train_loss=0.0149
308
+ step 3060000 | 29192s | train_loss=0.0148
309
+ step 3070000 | 29281s | train_loss=0.0146
310
+ step 3080000 | 29362s | train_loss=0.0146
311
+ step 3090000 | 29450s | train_loss=0.0143
312
+ step 3100000 | 29636s | train_loss=0.0142 | fid=42.1751
313
+ step 3110000 | 29716s | train_loss=0.0146
314
+ step 3120000 | 29795s | train_loss=0.0143
315
+ step 3130000 | 29883s | train_loss=0.0142
316
+ step 3140000 | 29971s | train_loss=0.0145
317
+ step 3150000 | 30059s | train_loss=0.0141
318
+ step 3160000 | 30148s | train_loss=0.0142
319
+ step 3170000 | 30236s | train_loss=0.0143
320
+ step 3180000 | 30314s | train_loss=0.0150
321
+ step 3190000 | 30400s | train_loss=0.0146
322
+ step 3200000 | 30595s | train_loss=0.0140 | fid=42.4358
323
+ step 3210000 | 30684s | train_loss=0.0142
324
+ step 3220000 | 30761s | train_loss=0.0141
325
+ step 3230000 | 30837s | train_loss=0.0145
326
+ step 3240000 | 30925s | train_loss=0.0139
327
+ step 3250000 | 31014s | train_loss=0.0144
328
+ step 3260000 | 31102s | train_loss=0.0142
329
+ step 3270000 | 31190s | train_loss=0.0134
330
+ step 3280000 | 31268s | train_loss=0.0141
331
+ step 3290000 | 31357s | train_loss=0.0137
332
+ step 3300000 | 31543s | train_loss=0.0138 | fid=42.4920
333
+ step 3310000 | 31632s | train_loss=0.0135
334
+ step 3320000 | 31720s | train_loss=0.0142
335
+ step 3330000 | 31805s | train_loss=0.0137
336
+ step 3340000 | 31884s | train_loss=0.0141
337
+ step 3350000 | 31972s | train_loss=0.0135
338
+ step 3360000 | 32060s | train_loss=0.0138
339
+ step 3370000 | 32149s | train_loss=0.0140
340
+ step 3380000 | 32227s | train_loss=0.0136
341
+ step 3390000 | 32313s | train_loss=0.0136
342
+ step 3400000 | 32532s | train_loss=0.0140 | fid=41.5971
343
+ step 3410000 | 32621s | train_loss=0.0137
344
+ step 3420000 | 32709s | train_loss=0.0134
345
+ step 3430000 | 32796s | train_loss=0.0135
346
+ step 3440000 | 32871s | train_loss=0.0132
347
+ step 3450000 | 32960s | train_loss=0.0133
348
+ step 3460000 | 33048s | train_loss=0.0144
349
+ step 3470000 | 33136s | train_loss=0.0133
350
+ step 3480000 | 33215s | train_loss=0.0135
351
+ step 3490000 | 33303s | train_loss=0.0131
352
+ step 3500000 | 33493s | train_loss=0.0133 | fid=41.9488
353
+ step 3510000 | 33581s | train_loss=0.0133
354
+ step 3520000 | 33669s | train_loss=0.0134
355
+ step 3530000 | 33757s | train_loss=0.0136
356
+ step 3540000 | 33844s | train_loss=0.0128
357
+ step 3550000 | 33917s | train_loss=0.0130
358
+ step 3560000 | 34000s | train_loss=0.0126
359
+ step 3570000 | 34087s | train_loss=0.0125
360
+ step 3580000 | 34166s | train_loss=0.0125
361
+ step 3590000 | 34255s | train_loss=0.0128
362
+ step 3600000 | 34473s | train_loss=0.0132 | fid=42.0667
363
+ step 3610000 | 34561s | train_loss=0.0132
364
+ step 3620000 | 34649s | train_loss=0.0126
365
+ step 3630000 | 34738s | train_loss=0.0133
366
+ step 3640000 | 34824s | train_loss=0.0127
367
+ step 3650000 | 34913s | train_loss=0.0128
368
+ step 3660000 | 34985s | train_loss=0.0124
369
+ step 3670000 | 35064s | train_loss=0.0123
370
+ step 3680000 | 35146s | train_loss=0.0123
371
+ step 3690000 | 35235s | train_loss=0.0126
372
+ step 3700000 | 35450s | train_loss=0.0147 | fid=41.5839
373
+ step 3710000 | 35539s | train_loss=0.0124
374
+ step 3720000 | 35627s | train_loss=0.0131
375
+ step 3730000 | 35715s | train_loss=0.0128
376
+ step 3740000 | 35802s | train_loss=0.0122
377
+ step 3750000 | 35890s | train_loss=0.0127
378
+ step 3760000 | 35976s | train_loss=0.0125
379
+ step 3770000 | 36044s | train_loss=0.0127
380
+ step 3780000 | 36128s | train_loss=0.0125
381
+ step 3790000 | 36216s | train_loss=0.0129
382
+ step 3800000 | 36429s | train_loss=0.0127 | fid=41.8975
383
+ step 3810000 | 36517s | train_loss=0.0120
384
+ step 3820000 | 36606s | train_loss=0.0123
385
+ step 3830000 | 36694s | train_loss=0.0126
386
+ step 3840000 | 36782s | train_loss=0.0122
387
+ step 3850000 | 36869s | train_loss=0.0119
388
+ step 3860000 | 36957s | train_loss=0.0125
389
+ step 3870000 | 37020s | train_loss=0.0122
390
+ step 3880000 | 37106s | train_loss=0.0126
391
+ step 3890000 | 37195s | train_loss=0.0120
392
+ step 3900000 | 37390s | train_loss=0.0117 | fid=41.7928
393
+ step 3910000 | 37478s | train_loss=0.0118
394
+ step 3920000 | 37566s | train_loss=0.0122
395
+ step 3930000 | 37655s | train_loss=0.0122
396
+ step 3940000 | 37743s | train_loss=0.0117
397
+ step 3950000 | 37829s | train_loss=0.0117
398
+ step 3960000 | 37918s | train_loss=0.0120
399
+ step 3970000 | 37994s | train_loss=0.0118
400
+ step 3980000 | 38062s | train_loss=0.0115
401
+ step 3990000 | 38149s | train_loss=0.0121
402
+ step 4000000 | 38346s | train_loss=0.0117 | fid=42.1292
403
+ step 4010000 | 38434s | train_loss=0.0117
404
+ step 4020000 | 38523s | train_loss=0.0115
405
+ step 4030000 | 38611s | train_loss=0.0116
406
+ step 4040000 | 38699s | train_loss=0.0118
407
+ step 4050000 | 38787s | train_loss=0.0124
408
+ step 4060000 | 38874s | train_loss=0.0116
409
+ step 4070000 | 38954s | train_loss=0.0112
410
+ step 4080000 | 39035s | train_loss=0.0122
411
+ step 4090000 | 39109s | train_loss=0.0116
412
+ step 4100000 | 39291s | train_loss=0.0115 | fid=41.2561
413
+ step 4110000 | 39379s | train_loss=0.0113
414
+ step 4120000 | 39467s | train_loss=0.0115
415
+ step 4130000 | 39555s | train_loss=0.0119
416
+ step 4140000 | 39643s | train_loss=0.0114
417
+ step 4150000 | 39731s | train_loss=0.0116
418
+ step 4160000 | 39818s | train_loss=0.0116
419
+ step 4170000 | 39903s | train_loss=0.0115
420
+ step 4180000 | 39975s | train_loss=0.0120
421
+ step 4190000 | 40063s | train_loss=0.0116
422
+ step 4200000 | 40252s | train_loss=0.0115 | fid=41.1279
423
+ step 4210000 | 40340s | train_loss=0.0113
424
+ step 4220000 | 40428s | train_loss=0.0115
425
+ step 4230000 | 40517s | train_loss=0.0113
426
+ step 4240000 | 40605s | train_loss=0.0112
427
+ step 4250000 | 40693s | train_loss=0.0114
428
+ step 4260000 | 40781s | train_loss=0.0113
429
+ step 4270000 | 40868s | train_loss=0.0114
430
+ step 4280000 | 40942s | train_loss=0.0112
431
+ step 4290000 | 41030s | train_loss=0.0108
432
+ step 4300000 | 41226s | train_loss=0.0113 | fid=41.5679
433
+ step 4310000 | 41314s | train_loss=0.0111
434
+ step 4320000 | 41402s | train_loss=0.0118
435
+ step 4330000 | 41491s | train_loss=0.0111
436
+ step 4340000 | 41579s | train_loss=0.0119
437
+ step 4350000 | 41667s | train_loss=0.0109
438
+ step 4360000 | 41755s | train_loss=0.0119
439
+ step 4370000 | 41841s | train_loss=0.0115
440
+ step 4380000 | 41913s | train_loss=0.0111
441
+ step 4390000 | 42001s | train_loss=0.0114
442
+ step 4400000 | 42191s | train_loss=0.0113 | fid=41.7459
443
+ step 4410000 | 42268s | train_loss=0.0112
444
+ step 4420000 | 42357s | train_loss=0.0108
445
+ step 4430000 | 42445s | train_loss=0.0108
446
+ step 4440000 | 42533s | train_loss=0.0108
447
+ step 4450000 | 42621s | train_loss=0.0109
448
+ step 4460000 | 42709s | train_loss=0.0113
449
+ step 4470000 | 42798s | train_loss=0.0113
450
+ step 4480000 | 42875s | train_loss=0.0111
451
+ step 4490000 | 42963s | train_loss=0.0108
452
+ step 4500000 | 43169s | train_loss=0.0114 | fid=42.1991
453
+ step 4510000 | 43249s | train_loss=0.0109
454
+ step 4520000 | 43331s | train_loss=0.0105
455
+ step 4530000 | 43419s | train_loss=0.0109
456
+ step 4540000 | 43507s | train_loss=0.0105
457
+ step 4550000 | 43595s | train_loss=0.0110
458
+ step 4560000 | 43683s | train_loss=0.0106
459
+ step 4570000 | 43769s | train_loss=0.0105
460
+ step 4580000 | 43847s | train_loss=0.0110
461
+ step 4590000 | 43936s | train_loss=0.0104
462
+ step 4600000 | 44144s | train_loss=0.0107 | fid=41.8384
463
+ step 4610000 | 44232s | train_loss=0.0108
464
+ step 4620000 | 44308s | train_loss=0.0108
465
+ step 4630000 | 44395s | train_loss=0.0104
466
+ step 4640000 | 44483s | train_loss=0.0102
467
+ step 4650000 | 44571s | train_loss=0.0106
468
+ step 4660000 | 44659s | train_loss=0.0103
469
+ step 4670000 | 44741s | train_loss=0.0101
470
+ step 4680000 | 44820s | train_loss=0.0104
471
+ step 4690000 | 44908s | train_loss=0.0101
472
+ step 4700000 | 45116s | train_loss=0.0106 | fid=41.2624
473
+ step 4710000 | 45204s | train_loss=0.0104
474
+ step 4720000 | 45292s | train_loss=0.0106
475
+ step 4730000 | 45364s | train_loss=0.0101
476
+ step 4740000 | 45452s | train_loss=0.0109
477
+ step 4750000 | 45540s | train_loss=0.0107
478
+ step 4760000 | 45628s | train_loss=0.0105
479
+ step 4770000 | 45708s | train_loss=0.0102
480
+ step 4780000 | 45791s | train_loss=0.0105
481
+ step 4790000 | 45878s | train_loss=0.0105
482
+ step 4800000 | 46080s | train_loss=0.0101 | fid=42.1270
483
+ step 4810000 | 46168s | train_loss=0.0100
484
+ step 4820000 | 46256s | train_loss=0.0103
485
+ step 4830000 | 46344s | train_loss=0.0102
486
+ step 4840000 | 46419s | train_loss=0.0101
487
+ step 4850000 | 46507s | train_loss=0.0099
488
+ step 4860000 | 46595s | train_loss=0.0100
489
+ step 4870000 | 46677s | train_loss=0.0096
490
+ step 4880000 | 46765s | train_loss=0.0099
491
+ step 4890000 | 46852s | train_loss=0.0098
492
+ step 4900000 | 47062s | train_loss=0.0109 | fid=41.5480
493
+ step 4910000 | 47150s | train_loss=0.0097
494
+ step 4920000 | 47238s | train_loss=0.0098
495
+ step 4930000 | 47326s | train_loss=0.0096
496
+ step 4940000 | 47408s | train_loss=0.0100
497
+ step 4950000 | 47488s | train_loss=0.0098
498
+ step 4960000 | 47574s | train_loss=0.0100
499
+ step 4970000 | 47646s | train_loss=0.0098
500
+ step 4980000 | 47717s | train_loss=0.0097
501
+ step 4990000 | 47788s | train_loss=0.0099
502
+ step 5000000 | 47914s | train_loss=0.0104 | fid=41.1391
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d80/config.json ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "vae_checkpoint": "vae_checkpoints/celeba_resnet_modernloss_d80/vae.pt",
3
+ "out": "results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed45/d80",
4
+ "subset_path": "diagnostics/celeba_diverse_1k_mean_plus_sd/subset_indices.json",
5
+ "train_tar": "/overflow/data/cerberus/celeba_hq_256_partitioned_ws/train/celeba_hq_256-train-0000.tar",
6
+ "n_train": 1000,
7
+ "total_steps": 5000000,
8
+ "eval_interval": 10000,
9
+ "mem_interval": 100000,
10
+ "fid_interval": 100000,
11
+ "fid_n_real": 1000,
12
+ "fid_n_gen": 10000,
13
+ "n_gen_samples": 10000,
14
+ "n_sde_steps": 500,
15
+ "batch_size": 512,
16
+ "hidden": 1024,
17
+ "depth": 5,
18
+ "optimizer": "sgd",
19
+ "lr": 0.001,
20
+ "momentum": 0.8,
21
+ "t_min": 0.01,
22
+ "t_max": 3.0,
23
+ "t_eval": 0.1,
24
+ "seed": 45,
25
+ "device": null
26
+ }