Hugging Face's logo

trevorbchen
/
diffusion_memorization

Model card Files
xet
 Community
trevorbchen commited on
Commit
9bd1e58
·
verified ·
1 Parent(s): 542c2a7

Add files using upload-large-folder tool

Browse files
Files changed (50) hide show
  1. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/d10/config.json +26 -0
  2. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/d10/metrics.jsonl +0 -0
  3. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/d10/runner.log +502 -0
  4. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/d100/config.json +26 -0
  5. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/d100/metrics.jsonl +0 -0
  6. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/d100/runner.log +502 -0
  7. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/d120/config.json +26 -0
  8. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/d120/metrics.jsonl +0 -0
  9. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/d120/runner.log +502 -0
  10. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/d140/config.json +26 -0
  11. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/d140/metrics.jsonl +0 -0
  12. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/d140/runner.log +502 -0
  13. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/d160/config.json +26 -0
  14. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/d160/metrics.jsonl +0 -0
  15. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/d160/runner.log +502 -0
  16. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/d180/config.json +26 -0
  17. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/d180/metrics.jsonl +0 -0
  18. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/d180/runner.log +502 -0
  19. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/d20/config.json +26 -0
  20. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/d20/metrics.jsonl +0 -0
  21. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/d20/runner.log +502 -0
  22. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/d200/config.json +26 -0
  23. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/d200/metrics.jsonl +0 -0
  24. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/d200/runner.log +502 -0
  25. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/d30/config.json +26 -0
  26. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/d30/metrics.jsonl +0 -0
  27. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/d30/runner.log +502 -0
  28. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/d40/config.json +26 -0
  29. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/d40/metrics.jsonl +0 -0
  30. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/d40/runner.log +502 -0
  31. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/d50/config.json +26 -0
  32. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/d50/metrics.jsonl +0 -0
  33. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/d50/runner.log +502 -0
  34. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/d60/config.json +26 -0
  35. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/d60/metrics.jsonl +0 -0
  36. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/d60/runner.log +502 -0
  37. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/d70/config.json +26 -0
  38. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/d70/metrics.jsonl +0 -0
  39. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/d70/runner.log +502 -0
  40. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/d80/config.json +26 -0
  41. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/d80/metrics.jsonl +0 -0
  42. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/d80/runner.log +502 -0
  43. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/d90/config.json +26 -0
  44. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/d90/metrics.jsonl +0 -0
  45. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/d90/runner.log +502 -0
  46. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed43/d10/config.json +26 -0
  47. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed43/d10/metrics.jsonl +0 -0
  48. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed43/d10/runner.log +502 -0
  49. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed43/d100/config.json +26 -0
  50. results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed43/d100/metrics.jsonl +0 -0
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/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/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": 42,
25
+ "device": null
26
+ }
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/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/d10/runner.log ADDED
@@ -0,0 +1,502 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ FID real-side stats cached over 1000 diverse train images
2
+ step 1 | 68s | train_loss=0.9779 | fid=194.7757
3
+ step 10000 | 159s | train_loss=0.9101
4
+ step 20000 | 251s | train_loss=0.9070
5
+ step 30000 | 343s | train_loss=0.9057
6
+ step 40000 | 435s | train_loss=0.9049
7
+ step 50000 | 528s | train_loss=0.9041
8
+ step 60000 | 620s | train_loss=0.9035
9
+ step 70000 | 712s | train_loss=0.9028
10
+ step 80000 | 804s | train_loss=0.9023
11
+ step 90000 | 895s | train_loss=0.9016
12
+ step 100000 | 1043s | train_loss=0.9011 | fid=87.0800
13
+ step 110000 | 1134s | train_loss=0.9006
14
+ step 120000 | 1225s | train_loss=0.9000
15
+ step 130000 | 1315s | train_loss=0.8994
16
+ step 140000 | 1406s | train_loss=0.8989
17
+ step 150000 | 1497s | train_loss=0.8982
18
+ step 160000 | 1587s | train_loss=0.8977
19
+ step 170000 | 1677s | train_loss=0.8971
20
+ step 180000 | 1767s | train_loss=0.8964
21
+ step 190000 | 1858s | train_loss=0.8955
22
+ step 200000 | 2007s | train_loss=0.8945 | fid=86.7825
23
+ step 210000 | 2098s | train_loss=0.8936
24
+ step 220000 | 2189s | train_loss=0.8924
25
+ step 230000 | 2280s | train_loss=0.8911
26
+ step 240000 | 2371s | train_loss=0.8896
27
+ step 250000 | 2461s | train_loss=0.8877
28
+ step 260000 | 2553s | train_loss=0.8855
29
+ step 270000 | 2644s | train_loss=0.8830
30
+ step 280000 | 2735s | train_loss=0.8800
31
+ step 290000 | 2818s | train_loss=0.8756
32
+ step 300000 | 2975s | train_loss=0.8707 | fid=87.2808
33
+ step 310000 | 3066s | train_loss=0.8633
34
+ step 320000 | 3157s | train_loss=0.8526
35
+ step 330000 | 3249s | train_loss=0.8389
36
+ step 340000 | 3341s | train_loss=0.8204
37
+ step 350000 | 3434s | train_loss=0.8041
38
+ step 360000 | 3526s | train_loss=0.7725
39
+ step 370000 | 3618s | train_loss=0.7394
40
+ step 380000 | 3704s | train_loss=0.7098
41
+ step 390000 | 3794s | train_loss=0.6718
42
+ step 400000 | 3953s | train_loss=0.6177 | fid=84.8395
43
+ step 410000 | 4044s | train_loss=0.5892
44
+ step 420000 | 4136s | train_loss=0.5362
45
+ step 430000 | 4228s | train_loss=0.4988
46
+ step 440000 | 4319s | train_loss=0.4757
47
+ step 450000 | 4412s | train_loss=0.4390
48
+ step 460000 | 4503s | train_loss=0.4119
49
+ step 470000 | 4590s | train_loss=0.3879
50
+ step 480000 | 4679s | train_loss=0.3691
51
+ step 490000 | 4767s | train_loss=0.3533
52
+ step 500000 | 4930s | train_loss=0.3337 | fid=86.4675
53
+ step 510000 | 5021s | train_loss=0.3222
54
+ step 520000 | 5113s | train_loss=0.3132
55
+ step 530000 | 5204s | train_loss=0.2943
56
+ step 540000 | 5297s | train_loss=0.2852
57
+ step 550000 | 5389s | train_loss=0.2750
58
+ step 560000 | 5476s | train_loss=0.2604
59
+ step 570000 | 5564s | train_loss=0.2585
60
+ step 580000 | 5657s | train_loss=0.2471
61
+ step 590000 | 5745s | train_loss=0.2399
62
+ step 600000 | 5908s | train_loss=0.2320 | fid=87.2500
63
+ step 610000 | 6000s | train_loss=0.2216
64
+ step 620000 | 6091s | train_loss=0.2202
65
+ step 630000 | 6184s | train_loss=0.2175
66
+ step 640000 | 6275s | train_loss=0.2108
67
+ step 650000 | 6364s | train_loss=0.2149
68
+ step 660000 | 6451s | train_loss=0.2055
69
+ step 670000 | 6543s | train_loss=0.1987
70
+ step 680000 | 6631s | train_loss=0.1940
71
+ step 690000 | 6722s | train_loss=0.1929
72
+ step 700000 | 6885s | train_loss=0.1898 | fid=87.3466
73
+ step 710000 | 6977s | train_loss=0.1850
74
+ step 720000 | 7069s | train_loss=0.1817
75
+ step 730000 | 7161s | train_loss=0.1783
76
+ step 740000 | 7252s | train_loss=0.1776
77
+ step 750000 | 7338s | train_loss=0.1758
78
+ step 760000 | 7427s | train_loss=0.1723
79
+ step 770000 | 7519s | train_loss=0.1681
80
+ step 780000 | 7606s | train_loss=0.1654
81
+ step 790000 | 7698s | train_loss=0.1655
82
+ step 800000 | 7862s | train_loss=0.1623 | fid=85.6919
83
+ step 810000 | 7954s | train_loss=0.1604
84
+ step 820000 | 8045s | train_loss=0.1569
85
+ step 830000 | 8136s | train_loss=0.1537
86
+ step 840000 | 8222s | train_loss=0.1552
87
+ step 850000 | 8311s | train_loss=0.1518
88
+ step 860000 | 8402s | train_loss=0.1518
89
+ step 870000 | 8493s | train_loss=0.1516
90
+ step 880000 | 8582s | train_loss=0.1476
91
+ step 890000 | 8674s | train_loss=0.1459
92
+ step 900000 | 8836s | train_loss=0.1469 | fid=88.7175
93
+ step 910000 | 8928s | train_loss=0.1417
94
+ step 920000 | 9019s | train_loss=0.1426
95
+ step 930000 | 9105s | train_loss=0.1384
96
+ step 940000 | 9194s | train_loss=0.1400
97
+ step 950000 | 9285s | train_loss=0.1371
98
+ step 960000 | 9378s | train_loss=0.1384
99
+ step 970000 | 9465s | train_loss=0.1326
100
+ step 980000 | 9558s | train_loss=0.1345
101
+ step 990000 | 9649s | train_loss=0.1337
102
+ step 1000000 | 9812s | train_loss=0.1300 | fid=86.9683
103
+ step 1010000 | 9904s | train_loss=0.1285
104
+ step 1020000 | 9989s | train_loss=0.1286
105
+ step 1030000 | 10079s | train_loss=0.1268
106
+ step 1040000 | 10169s | train_loss=0.1289
107
+ step 1050000 | 10261s | train_loss=0.1315
108
+ step 1060000 | 10353s | train_loss=0.1267
109
+ step 1070000 | 10440s | train_loss=0.1265
110
+ step 1080000 | 10532s | train_loss=0.1258
111
+ step 1090000 | 10623s | train_loss=0.1204
112
+ step 1100000 | 10785s | train_loss=0.1215 | fid=86.8557
113
+ step 1110000 | 10873s | train_loss=0.1217
114
+ step 1120000 | 10963s | train_loss=0.1205
115
+ step 1130000 | 11051s | train_loss=0.1209
116
+ step 1140000 | 11142s | train_loss=0.1202
117
+ step 1150000 | 11233s | train_loss=0.1176
118
+ step 1160000 | 11322s | train_loss=0.1178
119
+ step 1170000 | 11413s | train_loss=0.1162
120
+ step 1180000 | 11504s | train_loss=0.1151
121
+ step 1190000 | 11596s | train_loss=0.1157
122
+ step 1200000 | 11758s | train_loss=0.1129 | fid=86.2311
123
+ step 1210000 | 11845s | train_loss=0.1133
124
+ step 1220000 | 11934s | train_loss=0.1093
125
+ step 1230000 | 12025s | train_loss=0.1088
126
+ step 1240000 | 12116s | train_loss=0.1126
127
+ step 1250000 | 12208s | train_loss=0.1092
128
+ step 1260000 | 12295s | train_loss=0.1086
129
+ step 1270000 | 12387s | train_loss=0.1078
130
+ step 1280000 | 12479s | train_loss=0.1077
131
+ step 1290000 | 12571s | train_loss=0.1084
132
+ step 1300000 | 12729s | train_loss=0.1093 | fid=86.7596
133
+ step 1310000 | 12818s | train_loss=0.1089
134
+ step 1320000 | 12909s | train_loss=0.1052
135
+ step 1330000 | 13000s | train_loss=0.1050
136
+ step 1340000 | 13091s | train_loss=0.1039
137
+ step 1350000 | 13181s | train_loss=0.1034
138
+ step 1360000 | 13269s | train_loss=0.1033
139
+ step 1370000 | 13360s | train_loss=0.1054
140
+ step 1380000 | 13451s | train_loss=0.1017
141
+ step 1390000 | 13542s | train_loss=0.1005
142
+ step 1400000 | 13699s | train_loss=0.0990 | fid=85.4309
143
+ step 1410000 | 13788s | train_loss=0.1013
144
+ step 1420000 | 13879s | train_loss=0.0992
145
+ step 1430000 | 13970s | train_loss=0.1008
146
+ step 1440000 | 14061s | train_loss=0.0975
147
+ step 1450000 | 14148s | train_loss=0.0982
148
+ step 1460000 | 14239s | train_loss=0.0971
149
+ step 1470000 | 14330s | train_loss=0.0962
150
+ step 1480000 | 14423s | train_loss=0.0966
151
+ step 1490000 | 14499s | train_loss=0.0957
152
+ step 1500000 | 14650s | train_loss=0.0953 | fid=85.9069
153
+ step 1510000 | 14737s | train_loss=0.0946
154
+ step 1520000 | 14829s | train_loss=0.0960
155
+ step 1530000 | 14921s | train_loss=0.0911
156
+ step 1540000 | 15013s | train_loss=0.0929
157
+ step 1550000 | 15095s | train_loss=0.0941
158
+ step 1560000 | 15186s | train_loss=0.0918
159
+ step 1570000 | 15277s | train_loss=0.0914
160
+ step 1580000 | 15367s | train_loss=0.0911
161
+ step 1590000 | 15454s | train_loss=0.0915
162
+ step 1600000 | 15614s | train_loss=0.0897 | fid=86.9494
163
+ step 1610000 | 15705s | train_loss=0.0885
164
+ step 1620000 | 15797s | train_loss=0.0887
165
+ step 1630000 | 15888s | train_loss=0.0904
166
+ step 1640000 | 15980s | train_loss=0.0886
167
+ step 1650000 | 16067s | train_loss=0.0891
168
+ step 1660000 | 16159s | train_loss=0.0878
169
+ step 1670000 | 16250s | train_loss=0.0850
170
+ step 1680000 | 16336s | train_loss=0.0863
171
+ step 1690000 | 16429s | train_loss=0.0864
172
+ step 1700000 | 16587s | train_loss=0.0855 | fid=86.1843
173
+ step 1710000 | 16679s | train_loss=0.0853
174
+ step 1720000 | 16771s | train_loss=0.0860
175
+ step 1730000 | 16863s | train_loss=0.0850
176
+ step 1740000 | 16950s | train_loss=0.0841
177
+ step 1750000 | 17040s | train_loss=0.0832
178
+ step 1760000 | 17132s | train_loss=0.0845
179
+ step 1770000 | 17217s | train_loss=0.0836
180
+ step 1780000 | 17309s | train_loss=0.0837
181
+ step 1790000 | 17396s | train_loss=0.0831
182
+ step 1800000 | 17558s | train_loss=0.0839 | fid=87.2725
183
+ step 1810000 | 17650s | train_loss=0.0840
184
+ step 1820000 | 17742s | train_loss=0.0811
185
+ step 1830000 | 17834s | train_loss=0.0817
186
+ step 1840000 | 17923s | train_loss=0.0812
187
+ step 1850000 | 18014s | train_loss=0.0814
188
+ step 1860000 | 18099s | train_loss=0.0803
189
+ step 1870000 | 18191s | train_loss=0.0808
190
+ step 1880000 | 18282s | train_loss=0.0815
191
+ step 1890000 | 18371s | train_loss=0.0807
192
+ step 1900000 | 18534s | train_loss=0.0800 | fid=86.4817
193
+ step 1910000 | 18625s | train_loss=0.0793
194
+ step 1920000 | 18717s | train_loss=0.0793
195
+ step 1930000 | 18806s | train_loss=0.0779
196
+ step 1940000 | 18896s | train_loss=0.0795
197
+ step 1950000 | 18983s | train_loss=0.0767
198
+ step 1960000 | 19073s | train_loss=0.0778
199
+ step 1970000 | 19164s | train_loss=0.0786
200
+ step 1980000 | 19253s | train_loss=0.0773
201
+ step 1990000 | 19345s | train_loss=0.0767
202
+ step 2000000 | 19506s | train_loss=0.0761 | fid=87.8481
203
+ step 2010000 | 19598s | train_loss=0.0756
204
+ step 2020000 | 19690s | train_loss=0.0750
205
+ step 2030000 | 19772s | train_loss=0.0757
206
+ step 2040000 | 19857s | train_loss=0.0754
207
+ step 2050000 | 19936s | train_loss=0.0753
208
+ step 2060000 | 20029s | train_loss=0.0749
209
+ step 2070000 | 20115s | train_loss=0.0753
210
+ step 2080000 | 20200s | train_loss=0.0724
211
+ step 2090000 | 20292s | train_loss=0.0730
212
+ step 2100000 | 20454s | train_loss=0.0735 | fid=86.2122
213
+ step 2110000 | 20545s | train_loss=0.0747
214
+ step 2120000 | 20638s | train_loss=0.0729
215
+ step 2130000 | 20721s | train_loss=0.0736
216
+ step 2140000 | 20803s | train_loss=0.0733
217
+ step 2150000 | 20889s | train_loss=0.0725
218
+ step 2160000 | 20980s | train_loss=0.0725
219
+ step 2170000 | 21062s | train_loss=0.0721
220
+ step 2180000 | 21149s | train_loss=0.0726
221
+ step 2190000 | 21240s | train_loss=0.0724
222
+ step 2200000 | 21401s | train_loss=0.0716 | fid=85.7371
223
+ step 2210000 | 21493s | train_loss=0.0712
224
+ step 2220000 | 21585s | train_loss=0.0698
225
+ step 2230000 | 21666s | train_loss=0.0715
226
+ step 2240000 | 21743s | train_loss=0.0702
227
+ step 2250000 | 21831s | train_loss=0.0710
228
+ step 2260000 | 21922s | train_loss=0.0714
229
+ step 2270000 | 22002s | train_loss=0.0718
230
+ step 2280000 | 22088s | train_loss=0.0695
231
+ step 2290000 | 22179s | train_loss=0.0706
232
+ step 2300000 | 22342s | train_loss=0.0691 | fid=87.2041
233
+ step 2310000 | 22433s | train_loss=0.0688
234
+ step 2320000 | 22522s | train_loss=0.0686
235
+ step 2330000 | 22605s | train_loss=0.0691
236
+ step 2340000 | 22683s | train_loss=0.0700
237
+ step 2350000 | 22773s | train_loss=0.0688
238
+ step 2360000 | 22865s | train_loss=0.0694
239
+ step 2370000 | 22947s | train_loss=0.0694
240
+ step 2380000 | 23032s | train_loss=0.0668
241
+ step 2390000 | 23124s | train_loss=0.0685
242
+ step 2400000 | 23287s | train_loss=0.0677 | fid=87.4835
243
+ step 2410000 | 23377s | train_loss=0.0676
244
+ step 2420000 | 23468s | train_loss=0.0666
245
+ step 2430000 | 23547s | train_loss=0.0682
246
+ step 2440000 | 23630s | train_loss=0.0666
247
+ step 2450000 | 23720s | train_loss=0.0648
248
+ step 2460000 | 23811s | train_loss=0.0668
249
+ step 2470000 | 23892s | train_loss=0.0665
250
+ step 2480000 | 23978s | train_loss=0.0657
251
+ step 2490000 | 24069s | train_loss=0.0656
252
+ step 2500000 | 24233s | train_loss=0.0650 | fid=86.4387
253
+ step 2510000 | 24322s | train_loss=0.0651
254
+ step 2520000 | 24413s | train_loss=0.0651
255
+ step 2530000 | 24486s | train_loss=0.0653
256
+ step 2540000 | 24568s | train_loss=0.0651
257
+ step 2550000 | 24660s | train_loss=0.0641
258
+ step 2560000 | 24750s | train_loss=0.0652
259
+ step 2570000 | 24830s | train_loss=0.0641
260
+ step 2580000 | 24916s | train_loss=0.0639
261
+ step 2590000 | 25008s | train_loss=0.0627
262
+ step 2600000 | 25169s | train_loss=0.0636 | fid=87.4759
263
+ step 2610000 | 25260s | train_loss=0.0633
264
+ step 2620000 | 25350s | train_loss=0.0633
265
+ step 2630000 | 25430s | train_loss=0.0636
266
+ step 2640000 | 25511s | train_loss=0.0631
267
+ step 2650000 | 25602s | train_loss=0.0634
268
+ step 2660000 | 25693s | train_loss=0.0622
269
+ step 2670000 | 25774s | train_loss=0.0617
270
+ step 2680000 | 25859s | train_loss=0.0622
271
+ step 2690000 | 25950s | train_loss=0.0626
272
+ step 2700000 | 26110s | train_loss=0.0619 | fid=87.9749
273
+ step 2710000 | 26202s | train_loss=0.0617
274
+ step 2720000 | 26284s | train_loss=0.0618
275
+ step 2730000 | 26369s | train_loss=0.0622
276
+ step 2740000 | 26448s | train_loss=0.0623
277
+ step 2750000 | 26539s | train_loss=0.0606
278
+ step 2760000 | 26630s | train_loss=0.0619
279
+ step 2770000 | 26711s | train_loss=0.0610
280
+ step 2780000 | 26799s | train_loss=0.0619
281
+ step 2790000 | 26888s | train_loss=0.0611
282
+ step 2800000 | 27051s | train_loss=0.0603 | fid=87.4593
283
+ step 2810000 | 27143s | train_loss=0.0596
284
+ step 2820000 | 27223s | train_loss=0.0602
285
+ step 2830000 | 27310s | train_loss=0.0595
286
+ step 2840000 | 27390s | train_loss=0.0593
287
+ step 2850000 | 27480s | train_loss=0.0598
288
+ step 2860000 | 27571s | train_loss=0.0599
289
+ step 2870000 | 27652s | train_loss=0.0592
290
+ step 2880000 | 27741s | train_loss=0.0588
291
+ step 2890000 | 27832s | train_loss=0.0577
292
+ step 2900000 | 27994s | train_loss=0.0579 | fid=87.7690
293
+ step 2910000 | 28086s | train_loss=0.0592
294
+ step 2920000 | 28166s | train_loss=0.0575
295
+ step 2930000 | 28258s | train_loss=0.0583
296
+ step 2940000 | 28333s | train_loss=0.0581
297
+ step 2950000 | 28422s | train_loss=0.0579
298
+ step 2960000 | 28513s | train_loss=0.0578
299
+ step 2970000 | 28595s | train_loss=0.0582
300
+ step 2980000 | 28685s | train_loss=0.0574
301
+ step 2990000 | 28776s | train_loss=0.0575
302
+ step 3000000 | 28939s | train_loss=0.0566 | fid=87.1931
303
+ step 3010000 | 29029s | train_loss=0.0572
304
+ step 3020000 | 29111s | train_loss=0.0574
305
+ step 3030000 | 29203s | train_loss=0.0575
306
+ step 3040000 | 29285s | train_loss=0.0566
307
+ step 3050000 | 29374s | train_loss=0.0576
308
+ step 3060000 | 29458s | train_loss=0.0566
309
+ step 3070000 | 29540s | train_loss=0.0565
310
+ step 3080000 | 29630s | train_loss=0.0571
311
+ step 3090000 | 29722s | train_loss=0.0557
312
+ step 3100000 | 29884s | train_loss=0.0555 | fid=86.7166
313
+ step 3110000 | 29972s | train_loss=0.0553
314
+ step 3120000 | 30064s | train_loss=0.0552
315
+ step 3130000 | 30155s | train_loss=0.0555
316
+ step 3140000 | 30240s | train_loss=0.0562
317
+ step 3150000 | 30331s | train_loss=0.0557
318
+ step 3160000 | 30410s | train_loss=0.0554
319
+ step 3170000 | 30494s | train_loss=0.0548
320
+ step 3180000 | 30585s | train_loss=0.0556
321
+ step 3190000 | 30677s | train_loss=0.0550
322
+ step 3200000 | 30836s | train_loss=0.0555 | fid=86.7308
323
+ step 3210000 | 30927s | train_loss=0.0544
324
+ step 3220000 | 31020s | train_loss=0.0547
325
+ step 3230000 | 31111s | train_loss=0.0536
326
+ step 3240000 | 31199s | train_loss=0.0556
327
+ step 3250000 | 31288s | train_loss=0.0547
328
+ step 3260000 | 31375s | train_loss=0.0553
329
+ step 3270000 | 31466s | train_loss=0.0535
330
+ step 3280000 | 31558s | train_loss=0.0539
331
+ step 3290000 | 31649s | train_loss=0.0552
332
+ step 3300000 | 31807s | train_loss=0.0524 | fid=87.0882
333
+ step 3310000 | 31898s | train_loss=0.0547
334
+ step 3320000 | 31990s | train_loss=0.0541
335
+ step 3330000 | 32081s | train_loss=0.0531
336
+ step 3340000 | 32169s | train_loss=0.0537
337
+ step 3350000 | 32255s | train_loss=0.0539
338
+ step 3360000 | 32347s | train_loss=0.0544
339
+ step 3370000 | 32437s | train_loss=0.0529
340
+ step 3380000 | 32529s | train_loss=0.0532
341
+ step 3390000 | 32617s | train_loss=0.0539
342
+ step 3400000 | 32768s | train_loss=0.0520 | fid=86.1955
343
+ step 3410000 | 32859s | train_loss=0.0540
344
+ step 3420000 | 32951s | train_loss=0.0537
345
+ step 3430000 | 33037s | train_loss=0.0523
346
+ step 3440000 | 33116s | train_loss=0.0541
347
+ step 3450000 | 33196s | train_loss=0.0532
348
+ step 3460000 | 33287s | train_loss=0.0522
349
+ step 3470000 | 33378s | train_loss=0.0528
350
+ step 3480000 | 33469s | train_loss=0.0519
351
+ step 3490000 | 33554s | train_loss=0.0519
352
+ step 3500000 | 33708s | train_loss=0.0518 | fid=86.4849
353
+ step 3510000 | 33800s | train_loss=0.0517
354
+ step 3520000 | 33892s | train_loss=0.0519
355
+ step 3530000 | 33981s | train_loss=0.0513
356
+ step 3540000 | 34059s | train_loss=0.0510
357
+ step 3550000 | 34137s | train_loss=0.0520
358
+ step 3560000 | 34229s | train_loss=0.0514
359
+ step 3570000 | 34321s | train_loss=0.0521
360
+ step 3580000 | 34411s | train_loss=0.0516
361
+ step 3590000 | 34492s | train_loss=0.0511
362
+ step 3600000 | 34654s | train_loss=0.0510 | fid=87.4082
363
+ step 3610000 | 34746s | train_loss=0.0509
364
+ step 3620000 | 34835s | train_loss=0.0511
365
+ step 3630000 | 34926s | train_loss=0.0508
366
+ step 3640000 | 35002s | train_loss=0.0513
367
+ step 3650000 | 35084s | train_loss=0.0521
368
+ step 3660000 | 35175s | train_loss=0.0530
369
+ step 3670000 | 35267s | train_loss=0.0509
370
+ step 3680000 | 35356s | train_loss=0.0513
371
+ step 3690000 | 35434s | train_loss=0.0506
372
+ step 3700000 | 35596s | train_loss=0.0507 | fid=87.0584
373
+ step 3710000 | 35686s | train_loss=0.0507
374
+ step 3720000 | 35776s | train_loss=0.0506
375
+ step 3730000 | 35868s | train_loss=0.0508
376
+ step 3740000 | 35941s | train_loss=0.0500
377
+ step 3750000 | 36026s | train_loss=0.0495
378
+ step 3760000 | 36117s | train_loss=0.0495
379
+ step 3770000 | 36208s | train_loss=0.0508
380
+ step 3780000 | 36296s | train_loss=0.0503
381
+ step 3790000 | 36376s | train_loss=0.0505
382
+ step 3800000 | 36538s | train_loss=0.0489 | fid=86.1904
383
+ step 3810000 | 36627s | train_loss=0.0496
384
+ step 3820000 | 36718s | train_loss=0.0494
385
+ step 3830000 | 36810s | train_loss=0.0489
386
+ step 3840000 | 36881s | train_loss=0.0489
387
+ step 3850000 | 36965s | train_loss=0.0497
388
+ step 3860000 | 37056s | train_loss=0.0497
389
+ step 3870000 | 37147s | train_loss=0.0481
390
+ step 3880000 | 37232s | train_loss=0.0485
391
+ step 3890000 | 37323s | train_loss=0.0496
392
+ step 3900000 | 37483s | train_loss=0.0485 | fid=86.8610
393
+ step 3910000 | 37574s | train_loss=0.0495
394
+ step 3920000 | 37666s | train_loss=0.0485
395
+ step 3930000 | 37757s | train_loss=0.0500
396
+ step 3940000 | 37842s | train_loss=0.0492
397
+ step 3950000 | 37933s | train_loss=0.0478
398
+ step 3960000 | 38024s | train_loss=0.0489
399
+ step 3970000 | 38112s | train_loss=0.0483
400
+ step 3980000 | 38203s | train_loss=0.0491
401
+ step 3990000 | 38294s | train_loss=0.0483
402
+ step 4000000 | 38454s | train_loss=0.0480 | fid=87.8268
403
+ step 4010000 | 38546s | train_loss=0.0480
404
+ step 4020000 | 38637s | train_loss=0.0490
405
+ step 4030000 | 38726s | train_loss=0.0477
406
+ step 4040000 | 38813s | train_loss=0.0483
407
+ step 4050000 | 38905s | train_loss=0.0480
408
+ step 4060000 | 38995s | train_loss=0.0482
409
+ step 4070000 | 39080s | train_loss=0.0469
410
+ step 4080000 | 39171s | train_loss=0.0489
411
+ step 4090000 | 39260s | train_loss=0.0474
412
+ step 4100000 | 39423s | train_loss=0.0469 | fid=87.0254
413
+ step 4110000 | 39515s | train_loss=0.0469
414
+ step 4120000 | 39607s | train_loss=0.0467
415
+ step 4130000 | 39691s | train_loss=0.0480
416
+ step 4140000 | 39783s | train_loss=0.0470
417
+ step 4150000 | 39875s | train_loss=0.0475
418
+ step 4160000 | 39962s | train_loss=0.0466
419
+ step 4170000 | 40055s | train_loss=0.0469
420
+ step 4180000 | 40143s | train_loss=0.0465
421
+ step 4190000 | 40235s | train_loss=0.0462
422
+ step 4200000 | 40398s | train_loss=0.0470 | fid=86.7064
423
+ step 4210000 | 40490s | train_loss=0.0475
424
+ step 4220000 | 40571s | train_loss=0.0467
425
+ step 4230000 | 40643s | train_loss=0.0466
426
+ step 4240000 | 40732s | train_loss=0.0465
427
+ step 4250000 | 40814s | train_loss=0.0464
428
+ step 4260000 | 40904s | train_loss=0.0460
429
+ step 4270000 | 40995s | train_loss=0.0457
430
+ step 4280000 | 41084s | train_loss=0.0464
431
+ step 4290000 | 41176s | train_loss=0.0460
432
+ step 4300000 | 41338s | train_loss=0.0461 | fid=86.5752
433
+ step 4310000 | 41431s | train_loss=0.0459
434
+ step 4320000 | 41512s | train_loss=0.0466
435
+ step 4330000 | 41595s | train_loss=0.0460
436
+ step 4340000 | 41676s | train_loss=0.0461
437
+ step 4350000 | 41756s | train_loss=0.0463
438
+ step 4360000 | 41847s | train_loss=0.0472
439
+ step 4370000 | 41937s | train_loss=0.0463
440
+ step 4380000 | 42027s | train_loss=0.0465
441
+ step 4390000 | 42118s | train_loss=0.0460
442
+ step 4400000 | 42280s | train_loss=0.0459 | fid=86.2542
443
+ step 4410000 | 42371s | train_loss=0.0450
444
+ step 4420000 | 42450s | train_loss=0.0457
445
+ step 4430000 | 42537s | train_loss=0.0455
446
+ step 4440000 | 42611s | train_loss=0.0450
447
+ step 4450000 | 42690s | train_loss=0.0445
448
+ step 4460000 | 42778s | train_loss=0.0456
449
+ step 4470000 | 42869s | train_loss=0.0437
450
+ step 4480000 | 42961s | train_loss=0.0454
451
+ step 4490000 | 43052s | train_loss=0.0449
452
+ step 4500000 | 43214s | train_loss=0.0456 | fid=86.1850
453
+ step 4510000 | 43306s | train_loss=0.0452
454
+ step 4520000 | 43385s | train_loss=0.0451
455
+ step 4530000 | 43473s | train_loss=0.0443
456
+ step 4540000 | 43545s | train_loss=0.0441
457
+ step 4550000 | 43625s | train_loss=0.0441
458
+ step 4560000 | 43716s | train_loss=0.0451
459
+ step 4570000 | 43807s | train_loss=0.0439
460
+ step 4580000 | 43897s | train_loss=0.0438
461
+ step 4590000 | 43988s | train_loss=0.0443
462
+ step 4600000 | 44149s | train_loss=0.0439 | fid=87.9320
463
+ step 4610000 | 44241s | train_loss=0.0441
464
+ step 4620000 | 44320s | train_loss=0.0438
465
+ step 4630000 | 44412s | train_loss=0.0438
466
+ step 4640000 | 44482s | train_loss=0.0440
467
+ step 4650000 | 44554s | train_loss=0.0441
468
+ step 4660000 | 44628s | train_loss=0.0442
469
+ step 4670000 | 44700s | train_loss=0.0442
470
+ step 4680000 | 44773s | train_loss=0.0442
471
+ step 4690000 | 44848s | train_loss=0.0432
472
+ step 4700000 | 44978s | train_loss=0.0437 | fid=86.9798
473
+ step 4710000 | 45051s | train_loss=0.0441
474
+ step 4720000 | 45109s | train_loss=0.0435
475
+ step 4730000 | 45182s | train_loss=0.0432
476
+ step 4740000 | 45232s | train_loss=0.0443
477
+ step 4750000 | 45287s | train_loss=0.0430
478
+ step 4760000 | 45342s | train_loss=0.0438
479
+ step 4770000 | 45398s | train_loss=0.0433
480
+ step 4780000 | 45453s | train_loss=0.0436
481
+ step 4790000 | 45509s | train_loss=0.0430
482
+ step 4800000 | 45612s | train_loss=0.0431 | fid=87.3757
483
+ step 4810000 | 45666s | train_loss=0.0426
484
+ step 4820000 | 45704s | train_loss=0.0437
485
+ step 4830000 | 45742s | train_loss=0.0424
486
+ step 4840000 | 45775s | train_loss=0.0436
487
+ step 4850000 | 45813s | train_loss=0.0434
488
+ step 4860000 | 45850s | train_loss=0.0430
489
+ step 4870000 | 45887s | train_loss=0.0422
490
+ step 4880000 | 45925s | train_loss=0.0433
491
+ step 4890000 | 45963s | train_loss=0.0423
492
+ step 4900000 | 46034s | train_loss=0.0431 | fid=87.4389
493
+ step 4910000 | 46072s | train_loss=0.0433
494
+ step 4920000 | 46110s | train_loss=0.0423
495
+ step 4930000 | 46143s | train_loss=0.0435
496
+ step 4940000 | 46166s | train_loss=0.0426
497
+ step 4950000 | 46191s | train_loss=0.0434
498
+ step 4960000 | 46215s | train_loss=0.0425
499
+ step 4970000 | 46239s | train_loss=0.0421
500
+ step 4980000 | 46263s | train_loss=0.0419
501
+ step 4990000 | 46286s | train_loss=0.0418
502
+ step 5000000 | 46328s | train_loss=0.0416 | fid=87.2397
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/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/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": 42,
25
+ "device": null
26
+ }
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/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/d100/runner.log ADDED
@@ -0,0 +1,502 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ FID real-side stats cached over 1000 diverse train images
2
+ step 1 | 72s | train_loss=0.9951 | fid=288.8943
3
+ step 10000 | 166s | train_loss=0.8008
4
+ step 20000 | 261s | train_loss=0.6759
5
+ step 30000 | 356s | train_loss=0.5366
6
+ step 40000 | 451s | train_loss=0.3950
7
+ step 50000 | 546s | train_loss=0.2822
8
+ step 60000 | 641s | train_loss=0.2168
9
+ step 70000 | 736s | train_loss=0.1752
10
+ step 80000 | 831s | train_loss=0.1511
11
+ step 90000 | 926s | train_loss=0.1364
12
+ step 100000 | 1086s | train_loss=0.1270 | fid=58.6037
13
+ step 110000 | 1181s | train_loss=0.1197
14
+ step 120000 | 1276s | train_loss=0.1140
15
+ step 130000 | 1371s | train_loss=0.1079
16
+ step 140000 | 1466s | train_loss=0.1025
17
+ step 150000 | 1561s | train_loss=0.0995
18
+ step 160000 | 1656s | train_loss=0.0967
19
+ step 170000 | 1752s | train_loss=0.0934
20
+ step 180000 | 1847s | train_loss=0.0886
21
+ step 190000 | 1938s | train_loss=0.0869
22
+ step 200000 | 2103s | train_loss=0.0839 | fid=59.1418
23
+ step 210000 | 2198s | train_loss=0.0838
24
+ step 220000 | 2293s | train_loss=0.0815
25
+ step 230000 | 2388s | train_loss=0.0769
26
+ step 240000 | 2483s | train_loss=0.0764
27
+ step 250000 | 2578s | train_loss=0.0743
28
+ step 260000 | 2673s | train_loss=0.0730
29
+ step 270000 | 2768s | train_loss=0.0716
30
+ step 280000 | 2863s | train_loss=0.0707
31
+ step 290000 | 2953s | train_loss=0.0681
32
+ step 300000 | 3119s | train_loss=0.0669 | fid=55.3515
33
+ step 310000 | 3214s | train_loss=0.0658
34
+ step 320000 | 3309s | train_loss=0.0647
35
+ step 330000 | 3404s | train_loss=0.0653
36
+ step 340000 | 3499s | train_loss=0.0641
37
+ step 350000 | 3594s | train_loss=0.0615
38
+ step 360000 | 3689s | train_loss=0.0606
39
+ step 370000 | 3784s | train_loss=0.0619
40
+ step 380000 | 3873s | train_loss=0.0615
41
+ step 390000 | 3968s | train_loss=0.0587
42
+ step 400000 | 4136s | train_loss=0.0579 | fid=54.2836
43
+ step 410000 | 4231s | train_loss=0.0570
44
+ step 420000 | 4326s | train_loss=0.0572
45
+ step 430000 | 4421s | train_loss=0.0565
46
+ step 440000 | 4516s | train_loss=0.0561
47
+ step 450000 | 4611s | train_loss=0.0554
48
+ step 460000 | 4706s | train_loss=0.0545
49
+ step 470000 | 4796s | train_loss=0.0531
50
+ step 480000 | 4891s | train_loss=0.0534
51
+ step 490000 | 4986s | train_loss=0.0534
52
+ step 500000 | 5151s | train_loss=0.0509 | fid=53.0690
53
+ step 510000 | 5246s | train_loss=0.0529
54
+ step 520000 | 5341s | train_loss=0.0518
55
+ step 530000 | 5436s | train_loss=0.0521
56
+ step 540000 | 5531s | train_loss=0.0508
57
+ step 550000 | 5626s | train_loss=0.0496
58
+ step 560000 | 5715s | train_loss=0.0489
59
+ step 570000 | 5810s | train_loss=0.0488
60
+ step 580000 | 5905s | train_loss=0.0474
61
+ step 590000 | 6000s | train_loss=0.0489
62
+ step 600000 | 6167s | train_loss=0.0484 | fid=53.1119
63
+ step 610000 | 6262s | train_loss=0.0481
64
+ step 620000 | 6357s | train_loss=0.0459
65
+ step 630000 | 6452s | train_loss=0.0455
66
+ step 640000 | 6547s | train_loss=0.0465
67
+ step 650000 | 6639s | train_loss=0.0454
68
+ step 660000 | 6732s | train_loss=0.0459
69
+ step 670000 | 6827s | train_loss=0.0469
70
+ step 680000 | 6922s | train_loss=0.0443
71
+ step 690000 | 7018s | train_loss=0.0456
72
+ step 700000 | 7184s | train_loss=0.0449 | fid=50.9561
73
+ step 710000 | 7279s | train_loss=0.0438
74
+ step 720000 | 7374s | train_loss=0.0449
75
+ step 730000 | 7469s | train_loss=0.0442
76
+ step 740000 | 7561s | train_loss=0.0444
77
+ step 750000 | 7653s | train_loss=0.0428
78
+ step 760000 | 7748s | train_loss=0.0423
79
+ step 770000 | 7843s | train_loss=0.0425
80
+ step 780000 | 7938s | train_loss=0.0421
81
+ step 790000 | 8034s | train_loss=0.0422
82
+ step 800000 | 8201s | train_loss=0.0431 | fid=51.7319
83
+ step 810000 | 8296s | train_loss=0.0419
84
+ step 820000 | 8391s | train_loss=0.0407
85
+ step 830000 | 8483s | train_loss=0.0417
86
+ step 840000 | 8575s | train_loss=0.0411
87
+ step 850000 | 8670s | train_loss=0.0407
88
+ step 860000 | 8765s | train_loss=0.0394
89
+ step 870000 | 8860s | train_loss=0.0399
90
+ step 880000 | 8955s | train_loss=0.0410
91
+ step 890000 | 9049s | train_loss=0.0396
92
+ step 900000 | 9215s | train_loss=0.0393 | fid=49.7603
93
+ step 910000 | 9310s | train_loss=0.0393
94
+ step 920000 | 9405s | train_loss=0.0394
95
+ step 930000 | 9494s | train_loss=0.0389
96
+ step 940000 | 9589s | train_loss=0.0382
97
+ step 950000 | 9684s | train_loss=0.0388
98
+ step 960000 | 9779s | train_loss=0.0387
99
+ step 970000 | 9874s | train_loss=0.0375
100
+ step 980000 | 9969s | train_loss=0.0379
101
+ step 990000 | 10064s | train_loss=0.0371
102
+ step 1000000 | 10231s | train_loss=0.0375 | fid=51.4076
103
+ step 1010000 | 10326s | train_loss=0.0373
104
+ step 1020000 | 10415s | train_loss=0.0370
105
+ step 1030000 | 10510s | train_loss=0.0374
106
+ step 1040000 | 10605s | train_loss=0.0375
107
+ step 1050000 | 10700s | train_loss=0.0367
108
+ step 1060000 | 10795s | train_loss=0.0378
109
+ step 1070000 | 10890s | train_loss=0.0364
110
+ step 1080000 | 10985s | train_loss=0.0365
111
+ step 1090000 | 11077s | train_loss=0.0361
112
+ step 1100000 | 11246s | train_loss=0.0368 | fid=47.7480
113
+ step 1110000 | 11337s | train_loss=0.0352
114
+ step 1120000 | 11430s | train_loss=0.0343
115
+ step 1130000 | 11525s | train_loss=0.0361
116
+ step 1140000 | 11620s | train_loss=0.0350
117
+ step 1150000 | 11715s | train_loss=0.0356
118
+ step 1160000 | 11810s | train_loss=0.0341
119
+ step 1170000 | 11905s | train_loss=0.0346
120
+ step 1180000 | 12001s | train_loss=0.0336
121
+ step 1190000 | 12093s | train_loss=0.0348
122
+ step 1200000 | 12259s | train_loss=0.0347 | fid=46.8326
123
+ step 1210000 | 12351s | train_loss=0.0340
124
+ step 1220000 | 12446s | train_loss=0.0343
125
+ step 1230000 | 12541s | train_loss=0.0338
126
+ step 1240000 | 12636s | train_loss=0.0355
127
+ step 1250000 | 12731s | train_loss=0.0327
128
+ step 1260000 | 12826s | train_loss=0.0335
129
+ step 1270000 | 12921s | train_loss=0.0325
130
+ step 1280000 | 13015s | train_loss=0.0323
131
+ step 1290000 | 13109s | train_loss=0.0325
132
+ step 1300000 | 13273s | train_loss=0.0333 | fid=46.1509
133
+ step 1310000 | 13368s | train_loss=0.0334
134
+ step 1320000 | 13463s | train_loss=0.0329
135
+ step 1330000 | 13558s | train_loss=0.0330
136
+ step 1340000 | 13653s | train_loss=0.0321
137
+ step 1350000 | 13748s | train_loss=0.0320
138
+ step 1360000 | 13843s | train_loss=0.0329
139
+ step 1370000 | 13938s | train_loss=0.0312
140
+ step 1380000 | 14032s | train_loss=0.0313
141
+ step 1390000 | 14123s | train_loss=0.0319
142
+ step 1400000 | 14277s | train_loss=0.0321 | fid=46.7062
143
+ step 1410000 | 14372s | train_loss=0.0302
144
+ step 1420000 | 14468s | train_loss=0.0311
145
+ step 1430000 | 14562s | train_loss=0.0307
146
+ step 1440000 | 14657s | train_loss=0.0311
147
+ step 1450000 | 14752s | train_loss=0.0305
148
+ step 1460000 | 14847s | train_loss=0.0305
149
+ step 1470000 | 14942s | train_loss=0.0302
150
+ step 1480000 | 15032s | train_loss=0.0302
151
+ step 1490000 | 15119s | train_loss=0.0297
152
+ step 1500000 | 15284s | train_loss=0.0297 | fid=46.5345
153
+ step 1510000 | 15379s | train_loss=0.0301
154
+ step 1520000 | 15474s | train_loss=0.0297
155
+ step 1530000 | 15569s | train_loss=0.0309
156
+ step 1540000 | 15664s | train_loss=0.0293
157
+ step 1550000 | 15759s | train_loss=0.0292
158
+ step 1560000 | 15854s | train_loss=0.0309
159
+ step 1570000 | 15950s | train_loss=0.0297
160
+ step 1580000 | 16040s | train_loss=0.0294
161
+ step 1590000 | 16130s | train_loss=0.0285
162
+ step 1600000 | 16298s | train_loss=0.0285 | fid=44.1141
163
+ step 1610000 | 16393s | train_loss=0.0300
164
+ step 1620000 | 16488s | train_loss=0.0287
165
+ step 1630000 | 16583s | train_loss=0.0290
166
+ step 1640000 | 16678s | train_loss=0.0283
167
+ step 1650000 | 16772s | train_loss=0.0286
168
+ step 1660000 | 16867s | train_loss=0.0291
169
+ step 1670000 | 16960s | train_loss=0.0274
170
+ step 1680000 | 17050s | train_loss=0.0276
171
+ step 1690000 | 17143s | train_loss=0.0278
172
+ step 1700000 | 17313s | train_loss=0.0277 | fid=42.4506
173
+ step 1710000 | 17407s | train_loss=0.0279
174
+ step 1720000 | 17502s | train_loss=0.0273
175
+ step 1730000 | 17597s | train_loss=0.0274
176
+ step 1740000 | 17692s | train_loss=0.0273
177
+ step 1750000 | 17787s | train_loss=0.0266
178
+ step 1760000 | 17878s | train_loss=0.0277
179
+ step 1770000 | 17968s | train_loss=0.0272
180
+ step 1780000 | 18062s | train_loss=0.0273
181
+ step 1790000 | 18157s | train_loss=0.0267
182
+ step 1800000 | 18325s | train_loss=0.0261 | fid=42.9106
183
+ step 1810000 | 18420s | train_loss=0.0265
184
+ step 1820000 | 18515s | train_loss=0.0262
185
+ step 1830000 | 18610s | train_loss=0.0262
186
+ step 1840000 | 18705s | train_loss=0.0267
187
+ step 1850000 | 18799s | train_loss=0.0275
188
+ step 1860000 | 18891s | train_loss=0.0257
189
+ step 1870000 | 18982s | train_loss=0.0261
190
+ step 1880000 | 19075s | train_loss=0.0255
191
+ step 1890000 | 19170s | train_loss=0.0256
192
+ step 1900000 | 19339s | train_loss=0.0265 | fid=42.0647
193
+ step 1910000 | 19434s | train_loss=0.0263
194
+ step 1920000 | 19529s | train_loss=0.0256
195
+ step 1930000 | 19624s | train_loss=0.0259
196
+ step 1940000 | 19719s | train_loss=0.0261
197
+ step 1950000 | 19805s | train_loss=0.0255
198
+ step 1960000 | 19888s | train_loss=0.0252
199
+ step 1970000 | 19982s | train_loss=0.0243
200
+ step 1980000 | 20075s | train_loss=0.0256
201
+ step 1990000 | 20170s | train_loss=0.0243
202
+ step 2000000 | 20338s | train_loss=0.0243 | fid=42.6207
203
+ step 2010000 | 20433s | train_loss=0.0251
204
+ step 2020000 | 20528s | train_loss=0.0242
205
+ step 2030000 | 20623s | train_loss=0.0244
206
+ step 2040000 | 20712s | train_loss=0.0253
207
+ step 2050000 | 20803s | train_loss=0.0247
208
+ step 2060000 | 20892s | train_loss=0.0241
209
+ step 2070000 | 20986s | train_loss=0.0245
210
+ step 2080000 | 21079s | train_loss=0.0235
211
+ step 2090000 | 21174s | train_loss=0.0236
212
+ step 2100000 | 21343s | train_loss=0.0241 | fid=42.4474
213
+ step 2110000 | 21438s | train_loss=0.0229
214
+ step 2120000 | 21533s | train_loss=0.0242
215
+ step 2130000 | 21628s | train_loss=0.0233
216
+ step 2140000 | 21711s | train_loss=0.0235
217
+ step 2150000 | 21794s | train_loss=0.0234
218
+ step 2160000 | 21888s | train_loss=0.0235
219
+ step 2170000 | 21982s | train_loss=0.0238
220
+ step 2180000 | 22075s | train_loss=0.0242
221
+ step 2190000 | 22170s | train_loss=0.0231
222
+ step 2200000 | 22340s | train_loss=0.0246 | fid=44.0293
223
+ step 2210000 | 22435s | train_loss=0.0227
224
+ step 2220000 | 22530s | train_loss=0.0226
225
+ step 2230000 | 22621s | train_loss=0.0231
226
+ step 2240000 | 22711s | train_loss=0.0238
227
+ step 2250000 | 22792s | train_loss=0.0233
228
+ step 2260000 | 22886s | train_loss=0.0216
229
+ step 2270000 | 22981s | train_loss=0.0220
230
+ step 2280000 | 23075s | train_loss=0.0229
231
+ step 2290000 | 23170s | train_loss=0.0223
232
+ step 2300000 | 23339s | train_loss=0.0226 | fid=42.9386
233
+ step 2310000 | 23434s | train_loss=0.0231
234
+ step 2320000 | 23529s | train_loss=0.0224
235
+ step 2330000 | 23613s | train_loss=0.0224
236
+ step 2340000 | 23707s | train_loss=0.0218
237
+ step 2350000 | 23789s | train_loss=0.0224
238
+ step 2360000 | 23882s | train_loss=0.0217
239
+ step 2370000 | 23976s | train_loss=0.0227
240
+ step 2380000 | 24071s | train_loss=0.0213
241
+ step 2390000 | 24167s | train_loss=0.0221
242
+ step 2400000 | 24336s | train_loss=0.0212 | fid=44.2544
243
+ step 2410000 | 24431s | train_loss=0.0215
244
+ step 2420000 | 24522s | train_loss=0.0218
245
+ step 2430000 | 24611s | train_loss=0.0220
246
+ step 2440000 | 24702s | train_loss=0.0206
247
+ step 2450000 | 24788s | train_loss=0.0223
248
+ step 2460000 | 24882s | train_loss=0.0212
249
+ step 2470000 | 24975s | train_loss=0.0215
250
+ step 2480000 | 25071s | train_loss=0.0206
251
+ step 2490000 | 25166s | train_loss=0.0212
252
+ step 2500000 | 25335s | train_loss=0.0214 | fid=43.5538
253
+ step 2510000 | 25430s | train_loss=0.0226
254
+ step 2520000 | 25513s | train_loss=0.0209
255
+ step 2530000 | 25608s | train_loss=0.0205
256
+ step 2540000 | 25694s | train_loss=0.0204
257
+ step 2550000 | 25783s | train_loss=0.0214
258
+ step 2560000 | 25877s | train_loss=0.0206
259
+ step 2570000 | 25971s | train_loss=0.0214
260
+ step 2580000 | 26066s | train_loss=0.0209
261
+ step 2590000 | 26161s | train_loss=0.0208
262
+ step 2600000 | 26331s | train_loss=0.0200 | fid=42.8874
263
+ step 2610000 | 26425s | train_loss=0.0207
264
+ step 2620000 | 26509s | train_loss=0.0208
265
+ step 2630000 | 26603s | train_loss=0.0205
266
+ step 2640000 | 26688s | train_loss=0.0197
267
+ step 2650000 | 26782s | train_loss=0.0212
268
+ step 2660000 | 26877s | train_loss=0.0196
269
+ step 2670000 | 26970s | train_loss=0.0202
270
+ step 2680000 | 27065s | train_loss=0.0197
271
+ step 2690000 | 27160s | train_loss=0.0204
272
+ step 2700000 | 27328s | train_loss=0.0198 | fid=42.5185
273
+ step 2710000 | 27416s | train_loss=0.0204
274
+ step 2720000 | 27502s | train_loss=0.0196
275
+ step 2730000 | 27597s | train_loss=0.0200
276
+ step 2740000 | 27683s | train_loss=0.0203
277
+ step 2750000 | 27777s | train_loss=0.0192
278
+ step 2760000 | 27873s | train_loss=0.0196
279
+ step 2770000 | 27966s | train_loss=0.0204
280
+ step 2780000 | 28061s | train_loss=0.0195
281
+ step 2790000 | 28156s | train_loss=0.0188
282
+ step 2800000 | 28325s | train_loss=0.0189 | fid=42.7153
283
+ step 2810000 | 28411s | train_loss=0.0190
284
+ step 2820000 | 28500s | train_loss=0.0198
285
+ step 2830000 | 28587s | train_loss=0.0193
286
+ step 2840000 | 28676s | train_loss=0.0196
287
+ step 2850000 | 28770s | train_loss=0.0190
288
+ step 2860000 | 28865s | train_loss=0.0190
289
+ step 2870000 | 28958s | train_loss=0.0191
290
+ step 2880000 | 29053s | train_loss=0.0185
291
+ step 2890000 | 29148s | train_loss=0.0189
292
+ step 2900000 | 29316s | train_loss=0.0201 | fid=42.4508
293
+ step 2910000 | 29400s | train_loss=0.0193
294
+ step 2920000 | 29495s | train_loss=0.0190
295
+ step 2930000 | 29581s | train_loss=0.0180
296
+ step 2940000 | 29677s | train_loss=0.0184
297
+ step 2950000 | 29770s | train_loss=0.0190
298
+ step 2960000 | 29864s | train_loss=0.0188
299
+ step 2970000 | 29959s | train_loss=0.0194
300
+ step 2980000 | 30054s | train_loss=0.0189
301
+ step 2990000 | 30149s | train_loss=0.0184
302
+ step 3000000 | 30318s | train_loss=0.0183 | fid=40.7919
303
+ step 3010000 | 30404s | train_loss=0.0196
304
+ step 3020000 | 30497s | train_loss=0.0181
305
+ step 3030000 | 30587s | train_loss=0.0200
306
+ step 3040000 | 30682s | train_loss=0.0184
307
+ step 3050000 | 30776s | train_loss=0.0182
308
+ step 3060000 | 30869s | train_loss=0.0177
309
+ step 3070000 | 30964s | train_loss=0.0175
310
+ step 3080000 | 31059s | train_loss=0.0184
311
+ step 3090000 | 31154s | train_loss=0.0179
312
+ step 3100000 | 31320s | train_loss=0.0182 | fid=40.5046
313
+ step 3110000 | 31415s | train_loss=0.0174
314
+ step 3120000 | 31507s | train_loss=0.0183
315
+ step 3130000 | 31602s | train_loss=0.0178
316
+ step 3140000 | 31697s | train_loss=0.0177
317
+ step 3150000 | 31792s | train_loss=0.0179
318
+ step 3160000 | 31885s | train_loss=0.0174
319
+ step 3170000 | 31981s | train_loss=0.0187
320
+ step 3180000 | 32076s | train_loss=0.0176
321
+ step 3190000 | 32171s | train_loss=0.0183
322
+ step 3200000 | 32337s | train_loss=0.0174 | fid=41.4193
323
+ step 3210000 | 32423s | train_loss=0.0178
324
+ step 3220000 | 32515s | train_loss=0.0170
325
+ step 3230000 | 32610s | train_loss=0.0175
326
+ step 3240000 | 32704s | train_loss=0.0175
327
+ step 3250000 | 32799s | train_loss=0.0182
328
+ step 3260000 | 32893s | train_loss=0.0172
329
+ step 3270000 | 32988s | train_loss=0.0180
330
+ step 3280000 | 33083s | train_loss=0.0166
331
+ step 3290000 | 33171s | train_loss=0.0173
332
+ step 3300000 | 33339s | train_loss=0.0169 | fid=40.6265
333
+ step 3310000 | 33421s | train_loss=0.0171
334
+ step 3320000 | 33516s | train_loss=0.0173
335
+ step 3330000 | 33610s | train_loss=0.0180
336
+ step 3340000 | 33704s | train_loss=0.0172
337
+ step 3350000 | 33800s | train_loss=0.0174
338
+ step 3360000 | 33893s | train_loss=0.0176
339
+ step 3370000 | 33988s | train_loss=0.0177
340
+ step 3380000 | 34083s | train_loss=0.0171
341
+ step 3390000 | 34166s | train_loss=0.0180
342
+ step 3400000 | 34335s | train_loss=0.0168 | fid=40.9162
343
+ step 3410000 | 34417s | train_loss=0.0167
344
+ step 3420000 | 34511s | train_loss=0.0163
345
+ step 3430000 | 34607s | train_loss=0.0171
346
+ step 3440000 | 34700s | train_loss=0.0165
347
+ step 3450000 | 34796s | train_loss=0.0164
348
+ step 3460000 | 34889s | train_loss=0.0169
349
+ step 3470000 | 34984s | train_loss=0.0165
350
+ step 3480000 | 35074s | train_loss=0.0168
351
+ step 3490000 | 35161s | train_loss=0.0162
352
+ step 3500000 | 35327s | train_loss=0.0165 | fid=39.6144
353
+ step 3510000 | 35412s | train_loss=0.0165
354
+ step 3520000 | 35508s | train_loss=0.0160
355
+ step 3530000 | 35603s | train_loss=0.0162
356
+ step 3540000 | 35697s | train_loss=0.0168
357
+ step 3550000 | 35790s | train_loss=0.0164
358
+ step 3560000 | 35885s | train_loss=0.0171
359
+ step 3570000 | 35980s | train_loss=0.0172
360
+ step 3580000 | 36066s | train_loss=0.0161
361
+ step 3590000 | 36161s | train_loss=0.0170
362
+ step 3600000 | 36321s | train_loss=0.0162 | fid=40.7724
363
+ step 3610000 | 36411s | train_loss=0.0162
364
+ step 3620000 | 36507s | train_loss=0.0170
365
+ step 3630000 | 36600s | train_loss=0.0165
366
+ step 3640000 | 36695s | train_loss=0.0175
367
+ step 3650000 | 36789s | train_loss=0.0158
368
+ step 3660000 | 36884s | train_loss=0.0163
369
+ step 3670000 | 36979s | train_loss=0.0161
370
+ step 3680000 | 37061s | train_loss=0.0172
371
+ step 3690000 | 37157s | train_loss=0.0165
372
+ step 3700000 | 37322s | train_loss=0.0157 | fid=41.2574
373
+ step 3710000 | 37417s | train_loss=0.0165
374
+ step 3720000 | 37513s | train_loss=0.0164
375
+ step 3730000 | 37607s | train_loss=0.0168
376
+ step 3740000 | 37702s | train_loss=0.0164
377
+ step 3750000 | 37796s | train_loss=0.0152
378
+ step 3760000 | 37890s | train_loss=0.0162
379
+ step 3770000 | 37982s | train_loss=0.0154
380
+ step 3780000 | 38077s | train_loss=0.0156
381
+ step 3790000 | 38172s | train_loss=0.0166
382
+ step 3800000 | 38338s | train_loss=0.0165 | fid=40.6002
383
+ step 3810000 | 38433s | train_loss=0.0157
384
+ step 3820000 | 38528s | train_loss=0.0156
385
+ step 3830000 | 38622s | train_loss=0.0159
386
+ step 3840000 | 38717s | train_loss=0.0153
387
+ step 3850000 | 38810s | train_loss=0.0156
388
+ step 3860000 | 38903s | train_loss=0.0160
389
+ step 3870000 | 38992s | train_loss=0.0151
390
+ step 3880000 | 39086s | train_loss=0.0155
391
+ step 3890000 | 39178s | train_loss=0.0149
392
+ step 3900000 | 39348s | train_loss=0.0162 | fid=40.7402
393
+ step 3910000 | 39443s | train_loss=0.0157
394
+ step 3920000 | 39539s | train_loss=0.0154
395
+ step 3930000 | 39632s | train_loss=0.0158
396
+ step 3940000 | 39727s | train_loss=0.0160
397
+ step 3950000 | 39820s | train_loss=0.0150
398
+ step 3960000 | 39912s | train_loss=0.0148
399
+ step 3970000 | 40007s | train_loss=0.0149
400
+ step 3980000 | 40099s | train_loss=0.0163
401
+ step 3990000 | 40194s | train_loss=0.0152
402
+ step 4000000 | 40363s | train_loss=0.0151 | fid=40.1443
403
+ step 4010000 | 40458s | train_loss=0.0150
404
+ step 4020000 | 40552s | train_loss=0.0153
405
+ step 4030000 | 40647s | train_loss=0.0158
406
+ step 4040000 | 40742s | train_loss=0.0147
407
+ step 4050000 | 40827s | train_loss=0.0153
408
+ step 4060000 | 40921s | train_loss=0.0149
409
+ step 4070000 | 41012s | train_loss=0.0148
410
+ step 4080000 | 41098s | train_loss=0.0151
411
+ step 4090000 | 41193s | train_loss=0.0144
412
+ step 4100000 | 41362s | train_loss=0.0152 | fid=40.5122
413
+ step 4110000 | 41457s | train_loss=0.0147
414
+ step 4120000 | 41551s | train_loss=0.0148
415
+ step 4130000 | 41646s | train_loss=0.0154
416
+ step 4140000 | 41739s | train_loss=0.0143
417
+ step 4150000 | 41821s | train_loss=0.0154
418
+ step 4160000 | 41916s | train_loss=0.0178
419
+ step 4170000 | 42003s | train_loss=0.0156
420
+ step 4180000 | 42091s | train_loss=0.0140
421
+ step 4190000 | 42186s | train_loss=0.0153
422
+ step 4200000 | 42356s | train_loss=0.0146 | fid=40.2646
423
+ step 4210000 | 42451s | train_loss=0.0143
424
+ step 4220000 | 42545s | train_loss=0.0147
425
+ step 4230000 | 42639s | train_loss=0.0146
426
+ step 4240000 | 42729s | train_loss=0.0147
427
+ step 4250000 | 42814s | train_loss=0.0145
428
+ step 4260000 | 42909s | train_loss=0.0144
429
+ step 4270000 | 42993s | train_loss=0.0137
430
+ step 4280000 | 43086s | train_loss=0.0144
431
+ step 4290000 | 43181s | train_loss=0.0143
432
+ step 4300000 | 43349s | train_loss=0.0144 | fid=40.6055
433
+ step 4310000 | 43445s | train_loss=0.0146
434
+ step 4320000 | 43538s | train_loss=0.0147
435
+ step 4330000 | 43633s | train_loss=0.0149
436
+ step 4340000 | 43719s | train_loss=0.0139
437
+ step 4350000 | 43808s | train_loss=0.0144
438
+ step 4360000 | 43902s | train_loss=0.0134
439
+ step 4370000 | 43986s | train_loss=0.0152
440
+ step 4380000 | 44081s | train_loss=0.0152
441
+ step 4390000 | 44176s | train_loss=0.0147
442
+ step 4400000 | 44344s | train_loss=0.0137 | fid=41.1455
443
+ step 4410000 | 44440s | train_loss=0.0147
444
+ step 4420000 | 44532s | train_loss=0.0140
445
+ step 4430000 | 44628s | train_loss=0.0152
446
+ step 4440000 | 44710s | train_loss=0.0144
447
+ step 4450000 | 44801s | train_loss=0.0140
448
+ step 4460000 | 44894s | train_loss=0.0144
449
+ step 4470000 | 44979s | train_loss=0.0144
450
+ step 4480000 | 45074s | train_loss=0.0135
451
+ step 4490000 | 45169s | train_loss=0.0147
452
+ step 4500000 | 45338s | train_loss=0.0142 | fid=39.2491
453
+ step 4510000 | 45432s | train_loss=0.0158
454
+ step 4520000 | 45526s | train_loss=0.0135
455
+ step 4530000 | 45622s | train_loss=0.0151
456
+ step 4540000 | 45712s | train_loss=0.0141
457
+ step 4550000 | 45807s | train_loss=0.0137
458
+ step 4560000 | 45899s | train_loss=0.0146
459
+ step 4570000 | 45995s | train_loss=0.0162
460
+ step 4580000 | 46090s | train_loss=0.0139
461
+ step 4590000 | 46185s | train_loss=0.0137
462
+ step 4600000 | 46354s | train_loss=0.0137 | fid=40.6116
463
+ step 4610000 | 46448s | train_loss=0.0140
464
+ step 4620000 | 46542s | train_loss=0.0143
465
+ step 4630000 | 46635s | train_loss=0.0136
466
+ step 4640000 | 46729s | train_loss=0.0130
467
+ step 4650000 | 46822s | train_loss=0.0134
468
+ step 4660000 | 46917s | train_loss=0.0140
469
+ step 4670000 | 47012s | train_loss=0.0134
470
+ step 4680000 | 47107s | train_loss=0.0152
471
+ step 4690000 | 47201s | train_loss=0.0132
472
+ step 4700000 | 47371s | train_loss=0.0141 | fid=39.5449
473
+ step 4710000 | 47464s | train_loss=0.0132
474
+ step 4720000 | 47551s | train_loss=0.0130
475
+ step 4730000 | 47626s | train_loss=0.0138
476
+ step 4740000 | 47701s | train_loss=0.0132
477
+ step 4750000 | 47759s | train_loss=0.0139
478
+ step 4760000 | 47818s | train_loss=0.0136
479
+ step 4770000 | 47876s | train_loss=0.0136
480
+ step 4780000 | 47935s | train_loss=0.0136
481
+ step 4790000 | 47993s | train_loss=0.0140
482
+ step 4800000 | 48101s | train_loss=0.0132 | fid=39.9997
483
+ step 4810000 | 48156s | train_loss=0.0140
484
+ step 4820000 | 48215s | train_loss=0.0134
485
+ step 4830000 | 48269s | train_loss=0.0132
486
+ step 4840000 | 48327s | train_loss=0.0138
487
+ step 4850000 | 48386s | train_loss=0.0133
488
+ step 4860000 | 48444s | train_loss=0.0133
489
+ step 4870000 | 48503s | train_loss=0.0130
490
+ step 4880000 | 48562s | train_loss=0.0133
491
+ step 4890000 | 48620s | train_loss=0.0127
492
+ step 4900000 | 48728s | train_loss=0.0136 | fid=39.6727
493
+ step 4910000 | 48767s | train_loss=0.0126
494
+ step 4920000 | 48806s | train_loss=0.0131
495
+ step 4930000 | 48839s | train_loss=0.0150
496
+ step 4940000 | 48864s | train_loss=0.0130
497
+ step 4950000 | 48888s | train_loss=0.0136
498
+ step 4960000 | 48912s | train_loss=0.0131
499
+ step 4970000 | 48937s | train_loss=0.0131
500
+ step 4980000 | 48962s | train_loss=0.0131
501
+ step 4990000 | 48986s | train_loss=0.0125
502
+ step 5000000 | 49026s | train_loss=0.0134 | fid=39.0622
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/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/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": 42,
25
+ "device": null
26
+ }
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/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/d120/runner.log ADDED
@@ -0,0 +1,502 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ FID real-side stats cached over 1000 diverse train images
2
+ step 1 | 67s | train_loss=0.9999 | fid=266.3316
3
+ step 10000 | 161s | train_loss=0.7931
4
+ step 20000 | 255s | train_loss=0.6574
5
+ step 30000 | 350s | train_loss=0.5265
6
+ step 40000 | 445s | train_loss=0.4067
7
+ step 50000 | 540s | train_loss=0.3105
8
+ step 60000 | 636s | train_loss=0.2431
9
+ step 70000 | 731s | train_loss=0.2001
10
+ step 80000 | 826s | train_loss=0.1689
11
+ step 90000 | 921s | train_loss=0.1505
12
+ step 100000 | 1078s | train_loss=0.1379 | fid=64.2899
13
+ step 110000 | 1173s | train_loss=0.1252
14
+ step 120000 | 1268s | train_loss=0.1171
15
+ step 130000 | 1361s | train_loss=0.1143
16
+ step 140000 | 1456s | train_loss=0.1096
17
+ step 150000 | 1551s | train_loss=0.1049
18
+ step 160000 | 1646s | train_loss=0.1008
19
+ step 170000 | 1741s | train_loss=0.0957
20
+ step 180000 | 1837s | train_loss=0.0949
21
+ step 190000 | 1931s | train_loss=0.0907
22
+ step 200000 | 2087s | train_loss=0.0890 | fid=67.8545
23
+ step 210000 | 2181s | train_loss=0.0863
24
+ step 220000 | 2275s | train_loss=0.0825
25
+ step 230000 | 2369s | train_loss=0.0835
26
+ step 240000 | 2464s | train_loss=0.0807
27
+ step 250000 | 2559s | train_loss=0.0776
28
+ step 260000 | 2653s | train_loss=0.0768
29
+ step 270000 | 2748s | train_loss=0.0750
30
+ step 280000 | 2843s | train_loss=0.0735
31
+ step 290000 | 2932s | train_loss=0.0733
32
+ step 300000 | 3096s | train_loss=0.0714 | fid=63.4003
33
+ step 310000 | 3190s | train_loss=0.0684
34
+ step 320000 | 3284s | train_loss=0.0692
35
+ step 330000 | 3378s | train_loss=0.0686
36
+ step 340000 | 3473s | train_loss=0.0660
37
+ step 350000 | 3568s | train_loss=0.0640
38
+ step 360000 | 3663s | train_loss=0.0636
39
+ step 370000 | 3757s | train_loss=0.0642
40
+ step 380000 | 3848s | train_loss=0.0627
41
+ step 390000 | 3937s | train_loss=0.0622
42
+ step 400000 | 4104s | train_loss=0.0613 | fid=61.9486
43
+ step 410000 | 4199s | train_loss=0.0598
44
+ step 420000 | 4292s | train_loss=0.0587
45
+ step 430000 | 4386s | train_loss=0.0594
46
+ step 440000 | 4481s | train_loss=0.0586
47
+ step 450000 | 4575s | train_loss=0.0575
48
+ step 460000 | 4670s | train_loss=0.0557
49
+ step 470000 | 4764s | train_loss=0.0563
50
+ step 480000 | 4854s | train_loss=0.0566
51
+ step 490000 | 4945s | train_loss=0.0570
52
+ step 500000 | 5113s | train_loss=0.0547 | fid=62.0234
53
+ step 510000 | 5208s | train_loss=0.0539
54
+ step 520000 | 5301s | train_loss=0.0533
55
+ step 530000 | 5395s | train_loss=0.0525
56
+ step 540000 | 5490s | train_loss=0.0525
57
+ step 550000 | 5585s | train_loss=0.0514
58
+ step 560000 | 5679s | train_loss=0.0508
59
+ step 570000 | 5768s | train_loss=0.0504
60
+ step 580000 | 5863s | train_loss=0.0505
61
+ step 590000 | 5954s | train_loss=0.0514
62
+ step 600000 | 6122s | train_loss=0.0500 | fid=61.0952
63
+ step 610000 | 6216s | train_loss=0.0492
64
+ step 620000 | 6310s | train_loss=0.0494
65
+ step 630000 | 6404s | train_loss=0.0493
66
+ step 640000 | 6498s | train_loss=0.0495
67
+ step 650000 | 6592s | train_loss=0.0481
68
+ step 660000 | 6685s | train_loss=0.0499
69
+ step 670000 | 6775s | train_loss=0.0499
70
+ step 680000 | 6867s | train_loss=0.0493
71
+ step 690000 | 6961s | train_loss=0.0478
72
+ step 700000 | 7128s | train_loss=0.0475 | fid=59.0958
73
+ step 710000 | 7222s | train_loss=0.0458
74
+ step 720000 | 7316s | train_loss=0.0467
75
+ step 730000 | 7409s | train_loss=0.0479
76
+ step 740000 | 7503s | train_loss=0.0461
77
+ step 750000 | 7597s | train_loss=0.0446
78
+ step 760000 | 7686s | train_loss=0.0445
79
+ step 770000 | 7780s | train_loss=0.0443
80
+ step 780000 | 7872s | train_loss=0.0445
81
+ step 790000 | 7966s | train_loss=0.0432
82
+ step 800000 | 8133s | train_loss=0.0449 | fid=58.4905
83
+ step 810000 | 8227s | train_loss=0.0430
84
+ step 820000 | 8321s | train_loss=0.0436
85
+ step 830000 | 8415s | train_loss=0.0438
86
+ step 840000 | 8509s | train_loss=0.0435
87
+ step 850000 | 8598s | train_loss=0.0437
88
+ step 860000 | 8693s | train_loss=0.0415
89
+ step 870000 | 8787s | train_loss=0.0433
90
+ step 880000 | 8879s | train_loss=0.0449
91
+ step 890000 | 8973s | train_loss=0.0415
92
+ step 900000 | 9141s | train_loss=0.0418 | fid=63.2618
93
+ step 910000 | 9234s | train_loss=0.0414
94
+ step 920000 | 9328s | train_loss=0.0421
95
+ step 930000 | 9421s | train_loss=0.0409
96
+ step 940000 | 9516s | train_loss=0.0409
97
+ step 950000 | 9606s | train_loss=0.0428
98
+ step 960000 | 9700s | train_loss=0.0418
99
+ step 970000 | 9792s | train_loss=0.0406
100
+ step 980000 | 9887s | train_loss=0.0416
101
+ step 990000 | 9981s | train_loss=0.0397
102
+ step 1000000 | 10150s | train_loss=0.0404 | fid=61.9216
103
+ step 1010000 | 10243s | train_loss=0.0404
104
+ step 1020000 | 10338s | train_loss=0.0398
105
+ step 1030000 | 10431s | train_loss=0.0405
106
+ step 1040000 | 10521s | train_loss=0.0398
107
+ step 1050000 | 10615s | train_loss=0.0405
108
+ step 1060000 | 10710s | train_loss=0.0386
109
+ step 1070000 | 10801s | train_loss=0.0393
110
+ step 1080000 | 10896s | train_loss=0.0381
111
+ step 1090000 | 10990s | train_loss=0.0383
112
+ step 1100000 | 11157s | train_loss=0.0385 | fid=61.1120
113
+ step 1110000 | 11250s | train_loss=0.0382
114
+ step 1120000 | 11344s | train_loss=0.0378
115
+ step 1130000 | 11433s | train_loss=0.0372
116
+ step 1140000 | 11528s | train_loss=0.0381
117
+ step 1150000 | 11623s | train_loss=0.0381
118
+ step 1160000 | 11715s | train_loss=0.0366
119
+ step 1170000 | 11809s | train_loss=0.0376
120
+ step 1180000 | 11904s | train_loss=0.0371
121
+ step 1190000 | 11999s | train_loss=0.0377
122
+ step 1200000 | 12167s | train_loss=0.0372 | fid=62.6498
123
+ step 1210000 | 12260s | train_loss=0.0378
124
+ step 1220000 | 12354s | train_loss=0.0376
125
+ step 1230000 | 12444s | train_loss=0.0372
126
+ step 1240000 | 12538s | train_loss=0.0377
127
+ step 1250000 | 12632s | train_loss=0.0353
128
+ step 1260000 | 12724s | train_loss=0.0350
129
+ step 1270000 | 12819s | train_loss=0.0361
130
+ step 1280000 | 12914s | train_loss=0.0362
131
+ step 1290000 | 13008s | train_loss=0.0363
132
+ step 1300000 | 13176s | train_loss=0.0369 | fid=61.1260
133
+ step 1310000 | 13269s | train_loss=0.0361
134
+ step 1320000 | 13359s | train_loss=0.0353
135
+ step 1330000 | 13453s | train_loss=0.0353
136
+ step 1340000 | 13547s | train_loss=0.0371
137
+ step 1350000 | 13642s | train_loss=0.0345
138
+ step 1360000 | 13734s | train_loss=0.0350
139
+ step 1370000 | 13829s | train_loss=0.0362
140
+ step 1380000 | 13923s | train_loss=0.0339
141
+ step 1390000 | 14019s | train_loss=0.0354
142
+ step 1400000 | 14186s | train_loss=0.0346 | fid=63.0369
143
+ step 1410000 | 14274s | train_loss=0.0361
144
+ step 1420000 | 14369s | train_loss=0.0336
145
+ step 1430000 | 14463s | train_loss=0.0336
146
+ step 1440000 | 14557s | train_loss=0.0338
147
+ step 1450000 | 14649s | train_loss=0.0346
148
+ step 1460000 | 14744s | train_loss=0.0328
149
+ step 1470000 | 14839s | train_loss=0.0334
150
+ step 1480000 | 14933s | train_loss=0.0334
151
+ step 1490000 | 15028s | train_loss=0.0337
152
+ step 1500000 | 15195s | train_loss=0.0333 | fid=64.1658
153
+ step 1510000 | 15282s | train_loss=0.0336
154
+ step 1520000 | 15377s | train_loss=0.0319
155
+ step 1530000 | 15471s | train_loss=0.0349
156
+ step 1540000 | 15566s | train_loss=0.0328
157
+ step 1550000 | 15657s | train_loss=0.0328
158
+ step 1560000 | 15752s | train_loss=0.0343
159
+ step 1570000 | 15847s | train_loss=0.0311
160
+ step 1580000 | 15942s | train_loss=0.0333
161
+ step 1590000 | 16037s | train_loss=0.0314
162
+ step 1600000 | 16199s | train_loss=0.0313 | fid=65.1102
163
+ step 1610000 | 16292s | train_loss=0.0329
164
+ step 1620000 | 16387s | train_loss=0.0306
165
+ step 1630000 | 16482s | train_loss=0.0322
166
+ step 1640000 | 16575s | train_loss=0.0311
167
+ step 1650000 | 16667s | train_loss=0.0303
168
+ step 1660000 | 16762s | train_loss=0.0297
169
+ step 1670000 | 16857s | train_loss=0.0313
170
+ step 1680000 | 16952s | train_loss=0.0318
171
+ step 1690000 | 17048s | train_loss=0.0303
172
+ step 1700000 | 17210s | train_loss=0.0301 | fid=64.3673
173
+ step 1710000 | 17303s | train_loss=0.0301
174
+ step 1720000 | 17398s | train_loss=0.0314
175
+ step 1730000 | 17492s | train_loss=0.0312
176
+ step 1740000 | 17584s | train_loss=0.0301
177
+ step 1750000 | 17678s | train_loss=0.0297
178
+ step 1760000 | 17772s | train_loss=0.0310
179
+ step 1770000 | 17866s | train_loss=0.0310
180
+ step 1780000 | 17961s | train_loss=0.0302
181
+ step 1790000 | 18053s | train_loss=0.0302
182
+ step 1800000 | 18217s | train_loss=0.0305 | fid=67.4975
183
+ step 1810000 | 18309s | train_loss=0.0293
184
+ step 1820000 | 18404s | train_loss=0.0288
185
+ step 1830000 | 18498s | train_loss=0.0288
186
+ step 1840000 | 18590s | train_loss=0.0287
187
+ step 1850000 | 18685s | train_loss=0.0304
188
+ step 1860000 | 18779s | train_loss=0.0298
189
+ step 1870000 | 18874s | train_loss=0.0304
190
+ step 1880000 | 18968s | train_loss=0.0274
191
+ step 1890000 | 19059s | train_loss=0.0282
192
+ step 1900000 | 19226s | train_loss=0.0299 | fid=63.5500
193
+ step 1910000 | 19318s | train_loss=0.0292
194
+ step 1920000 | 19413s | train_loss=0.0282
195
+ step 1930000 | 19505s | train_loss=0.0294
196
+ step 1940000 | 19599s | train_loss=0.0288
197
+ step 1950000 | 19693s | train_loss=0.0278
198
+ step 1960000 | 19787s | train_loss=0.0280
199
+ step 1970000 | 19881s | train_loss=0.0273
200
+ step 1980000 | 19971s | train_loss=0.0284
201
+ step 1990000 | 20066s | train_loss=0.0278
202
+ step 2000000 | 20233s | train_loss=0.0282 | fid=63.1967
203
+ step 2010000 | 20326s | train_loss=0.0273
204
+ step 2020000 | 20421s | train_loss=0.0263
205
+ step 2030000 | 20512s | train_loss=0.0268
206
+ step 2040000 | 20607s | train_loss=0.0280
207
+ step 2050000 | 20702s | train_loss=0.0270
208
+ step 2060000 | 20796s | train_loss=0.0262
209
+ step 2070000 | 20889s | train_loss=0.0267
210
+ step 2080000 | 20981s | train_loss=0.0278
211
+ step 2090000 | 21076s | train_loss=0.0257
212
+ step 2100000 | 21243s | train_loss=0.0280 | fid=55.8109
213
+ step 2110000 | 21336s | train_loss=0.0267
214
+ step 2120000 | 21431s | train_loss=0.0269
215
+ step 2130000 | 21522s | train_loss=0.0263
216
+ step 2140000 | 21617s | train_loss=0.0258
217
+ step 2150000 | 21712s | train_loss=0.0256
218
+ step 2160000 | 21806s | train_loss=0.0257
219
+ step 2170000 | 21895s | train_loss=0.0258
220
+ step 2180000 | 21990s | train_loss=0.0266
221
+ step 2190000 | 22084s | train_loss=0.0260
222
+ step 2200000 | 22252s | train_loss=0.0271 | fid=52.2180
223
+ step 2210000 | 22345s | train_loss=0.0267
224
+ step 2220000 | 22437s | train_loss=0.0255
225
+ step 2230000 | 22532s | train_loss=0.0262
226
+ step 2240000 | 22627s | train_loss=0.0273
227
+ step 2250000 | 22721s | train_loss=0.0251
228
+ step 2260000 | 22811s | train_loss=0.0247
229
+ step 2270000 | 22906s | train_loss=0.0249
230
+ step 2280000 | 23000s | train_loss=0.0259
231
+ step 2290000 | 23096s | train_loss=0.0250
232
+ step 2300000 | 23264s | train_loss=0.0245 | fid=51.5318
233
+ step 2310000 | 23356s | train_loss=0.0251
234
+ step 2320000 | 23448s | train_loss=0.0245
235
+ step 2330000 | 23543s | train_loss=0.0248
236
+ step 2340000 | 23637s | train_loss=0.0244
237
+ step 2350000 | 23730s | train_loss=0.0248
238
+ step 2360000 | 23822s | train_loss=0.0238
239
+ step 2370000 | 23916s | train_loss=0.0254
240
+ step 2380000 | 24010s | train_loss=0.0245
241
+ step 2390000 | 24105s | train_loss=0.0251
242
+ step 2400000 | 24273s | train_loss=0.0240 | fid=50.0185
243
+ step 2410000 | 24365s | train_loss=0.0244
244
+ step 2420000 | 24459s | train_loss=0.0236
245
+ step 2430000 | 24554s | train_loss=0.0231
246
+ step 2440000 | 24649s | train_loss=0.0229
247
+ step 2450000 | 24738s | train_loss=0.0253
248
+ step 2460000 | 24833s | train_loss=0.0241
249
+ step 2470000 | 24926s | train_loss=0.0231
250
+ step 2480000 | 25021s | train_loss=0.0241
251
+ step 2490000 | 25116s | train_loss=0.0229
252
+ step 2500000 | 25284s | train_loss=0.0234 | fid=51.0426
253
+ step 2510000 | 25375s | train_loss=0.0238
254
+ step 2520000 | 25470s | train_loss=0.0230
255
+ step 2530000 | 25565s | train_loss=0.0225
256
+ step 2540000 | 25655s | train_loss=0.0247
257
+ step 2550000 | 25749s | train_loss=0.0233
258
+ step 2560000 | 25844s | train_loss=0.0232
259
+ step 2570000 | 25939s | train_loss=0.0240
260
+ step 2580000 | 26033s | train_loss=0.0242
261
+ step 2590000 | 26128s | train_loss=0.0230
262
+ step 2600000 | 26296s | train_loss=0.0222 | fid=48.9415
263
+ step 2610000 | 26387s | train_loss=0.0230
264
+ step 2620000 | 26482s | train_loss=0.0228
265
+ step 2630000 | 26575s | train_loss=0.0224
266
+ step 2640000 | 26667s | train_loss=0.0225
267
+ step 2650000 | 26762s | train_loss=0.0229
268
+ step 2660000 | 26856s | train_loss=0.0217
269
+ step 2670000 | 26951s | train_loss=0.0223
270
+ step 2680000 | 27045s | train_loss=0.0225
271
+ step 2690000 | 27140s | train_loss=0.0223
272
+ step 2700000 | 27306s | train_loss=0.0221 | fid=49.6115
273
+ step 2710000 | 27399s | train_loss=0.0215
274
+ step 2720000 | 27494s | train_loss=0.0215
275
+ step 2730000 | 27584s | train_loss=0.0213
276
+ step 2740000 | 27678s | train_loss=0.0232
277
+ step 2750000 | 27773s | train_loss=0.0223
278
+ step 2760000 | 27868s | train_loss=0.0219
279
+ step 2770000 | 27963s | train_loss=0.0229
280
+ step 2780000 | 28058s | train_loss=0.0230
281
+ step 2790000 | 28153s | train_loss=0.0210
282
+ step 2800000 | 28319s | train_loss=0.0215 | fid=48.4126
283
+ step 2810000 | 28411s | train_loss=0.0216
284
+ step 2820000 | 28501s | train_loss=0.0214
285
+ step 2830000 | 28595s | train_loss=0.0217
286
+ step 2840000 | 28689s | train_loss=0.0220
287
+ step 2850000 | 28784s | train_loss=0.0211
288
+ step 2860000 | 28879s | train_loss=0.0223
289
+ step 2870000 | 28973s | train_loss=0.0210
290
+ step 2880000 | 29068s | train_loss=0.0213
291
+ step 2890000 | 29162s | train_loss=0.0211
292
+ step 2900000 | 29328s | train_loss=0.0222 | fid=47.7153
293
+ step 2910000 | 29419s | train_loss=0.0214
294
+ step 2920000 | 29511s | train_loss=0.0212
295
+ step 2930000 | 29605s | train_loss=0.0199
296
+ step 2940000 | 29700s | train_loss=0.0204
297
+ step 2950000 | 29795s | train_loss=0.0217
298
+ step 2960000 | 29890s | train_loss=0.0221
299
+ step 2970000 | 29985s | train_loss=0.0225
300
+ step 2980000 | 30080s | train_loss=0.0213
301
+ step 2990000 | 30175s | train_loss=0.0205
302
+ step 3000000 | 30341s | train_loss=0.0202 | fid=48.1862
303
+ step 3010000 | 30428s | train_loss=0.0212
304
+ step 3020000 | 30523s | train_loss=0.0212
305
+ step 3030000 | 30618s | train_loss=0.0215
306
+ step 3040000 | 30713s | train_loss=0.0209
307
+ step 3050000 | 30807s | train_loss=0.0206
308
+ step 3060000 | 30902s | train_loss=0.0206
309
+ step 3070000 | 30997s | train_loss=0.0201
310
+ step 3080000 | 31092s | train_loss=0.0213
311
+ step 3090000 | 31184s | train_loss=0.0211
312
+ step 3100000 | 31350s | train_loss=0.0204 | fid=46.8548
313
+ step 3110000 | 31442s | train_loss=0.0200
314
+ step 3120000 | 31536s | train_loss=0.0218
315
+ step 3130000 | 31631s | train_loss=0.0196
316
+ step 3140000 | 31726s | train_loss=0.0202
317
+ step 3150000 | 31821s | train_loss=0.0204
318
+ step 3160000 | 31916s | train_loss=0.0201
319
+ step 3170000 | 32010s | train_loss=0.0209
320
+ step 3180000 | 32105s | train_loss=0.0205
321
+ step 3190000 | 32197s | train_loss=0.0195
322
+ step 3200000 | 32360s | train_loss=0.0201 | fid=46.2062
323
+ step 3210000 | 32453s | train_loss=0.0200
324
+ step 3220000 | 32547s | train_loss=0.0193
325
+ step 3230000 | 32642s | train_loss=0.0203
326
+ step 3240000 | 32736s | train_loss=0.0209
327
+ step 3250000 | 32831s | train_loss=0.0201
328
+ step 3260000 | 32926s | train_loss=0.0205
329
+ step 3270000 | 33020s | train_loss=0.0196
330
+ step 3280000 | 33114s | train_loss=0.0187
331
+ step 3290000 | 33205s | train_loss=0.0201
332
+ step 3300000 | 33370s | train_loss=0.0190 | fid=46.4072
333
+ step 3310000 | 33463s | train_loss=0.0198
334
+ step 3320000 | 33557s | train_loss=0.0187
335
+ step 3330000 | 33652s | train_loss=0.0190
336
+ step 3340000 | 33747s | train_loss=0.0186
337
+ step 3350000 | 33841s | train_loss=0.0199
338
+ step 3360000 | 33936s | train_loss=0.0195
339
+ step 3370000 | 34031s | train_loss=0.0198
340
+ step 3380000 | 34124s | train_loss=0.0194
341
+ step 3390000 | 34213s | train_loss=0.0191
342
+ step 3400000 | 34381s | train_loss=0.0193 | fid=45.7442
343
+ step 3410000 | 34474s | train_loss=0.0193
344
+ step 3420000 | 34569s | train_loss=0.0191
345
+ step 3430000 | 34663s | train_loss=0.0192
346
+ step 3440000 | 34758s | train_loss=0.0190
347
+ step 3450000 | 34852s | train_loss=0.0191
348
+ step 3460000 | 34947s | train_loss=0.0190
349
+ step 3470000 | 35042s | train_loss=0.0188
350
+ step 3480000 | 35131s | train_loss=0.0191
351
+ step 3490000 | 35223s | train_loss=0.0192
352
+ step 3500000 | 35391s | train_loss=0.0187 | fid=44.7324
353
+ step 3510000 | 35485s | train_loss=0.0189
354
+ step 3520000 | 35580s | train_loss=0.0179
355
+ step 3530000 | 35674s | train_loss=0.0186
356
+ step 3540000 | 35769s | train_loss=0.0203
357
+ step 3550000 | 35863s | train_loss=0.0188
358
+ step 3560000 | 35958s | train_loss=0.0185
359
+ step 3570000 | 36047s | train_loss=0.0192
360
+ step 3580000 | 36139s | train_loss=0.0179
361
+ step 3590000 | 36233s | train_loss=0.0190
362
+ step 3600000 | 36403s | train_loss=0.0174 | fid=44.5252
363
+ step 3610000 | 36498s | train_loss=0.0187
364
+ step 3620000 | 36593s | train_loss=0.0186
365
+ step 3630000 | 36687s | train_loss=0.0181
366
+ step 3640000 | 36781s | train_loss=0.0202
367
+ step 3650000 | 36877s | train_loss=0.0183
368
+ step 3660000 | 36971s | train_loss=0.0186
369
+ step 3670000 | 37058s | train_loss=0.0187
370
+ step 3680000 | 37152s | train_loss=0.0197
371
+ step 3690000 | 37247s | train_loss=0.0190
372
+ step 3700000 | 37415s | train_loss=0.0181 | fid=44.1040
373
+ step 3710000 | 37510s | train_loss=0.0182
374
+ step 3720000 | 37604s | train_loss=0.0179
375
+ step 3730000 | 37698s | train_loss=0.0191
376
+ step 3740000 | 37794s | train_loss=0.0183
377
+ step 3750000 | 37888s | train_loss=0.0168
378
+ step 3760000 | 37978s | train_loss=0.0183
379
+ step 3770000 | 38070s | train_loss=0.0175
380
+ step 3780000 | 38165s | train_loss=0.0175
381
+ step 3790000 | 38260s | train_loss=0.0177
382
+ step 3800000 | 38429s | train_loss=0.0188 | fid=44.7914
383
+ step 3810000 | 38524s | train_loss=0.0172
384
+ step 3820000 | 38618s | train_loss=0.0173
385
+ step 3830000 | 38713s | train_loss=0.0178
386
+ step 3840000 | 38808s | train_loss=0.0174
387
+ step 3850000 | 38899s | train_loss=0.0173
388
+ step 3860000 | 38988s | train_loss=0.0177
389
+ step 3870000 | 39083s | train_loss=0.0171
390
+ step 3880000 | 39177s | train_loss=0.0175
391
+ step 3890000 | 39272s | train_loss=0.0171
392
+ step 3900000 | 39439s | train_loss=0.0178 | fid=44.8249
393
+ step 3910000 | 39533s | train_loss=0.0179
394
+ step 3920000 | 39628s | train_loss=0.0178
395
+ step 3930000 | 39722s | train_loss=0.0185
396
+ step 3940000 | 39816s | train_loss=0.0188
397
+ step 3950000 | 39906s | train_loss=0.0174
398
+ step 3960000 | 40000s | train_loss=0.0169
399
+ step 3970000 | 40094s | train_loss=0.0180
400
+ step 3980000 | 40188s | train_loss=0.0179
401
+ step 3990000 | 40283s | train_loss=0.0177
402
+ step 4000000 | 40452s | train_loss=0.0173 | fid=44.5869
403
+ step 4010000 | 40547s | train_loss=0.0177
404
+ step 4020000 | 40641s | train_loss=0.0173
405
+ step 4030000 | 40735s | train_loss=0.0174
406
+ step 4040000 | 40824s | train_loss=0.0178
407
+ step 4050000 | 40918s | train_loss=0.0176
408
+ step 4060000 | 41013s | train_loss=0.0169
409
+ step 4070000 | 41108s | train_loss=0.0173
410
+ step 4080000 | 41202s | train_loss=0.0179
411
+ step 4090000 | 41297s | train_loss=0.0178
412
+ step 4100000 | 41467s | train_loss=0.0169 | fid=42.6971
413
+ step 4110000 | 41561s | train_loss=0.0175
414
+ step 4120000 | 41655s | train_loss=0.0169
415
+ step 4130000 | 41746s | train_loss=0.0180
416
+ step 4140000 | 41839s | train_loss=0.0171
417
+ step 4150000 | 41933s | train_loss=0.0198
418
+ step 4160000 | 42027s | train_loss=0.0200
419
+ step 4170000 | 42122s | train_loss=0.0180
420
+ step 4180000 | 42217s | train_loss=0.0168
421
+ step 4190000 | 42312s | train_loss=0.0176
422
+ step 4200000 | 42479s | train_loss=0.0174 | fid=42.3971
423
+ step 4210000 | 42572s | train_loss=0.0175
424
+ step 4220000 | 42664s | train_loss=0.0165
425
+ step 4230000 | 42756s | train_loss=0.0173
426
+ step 4240000 | 42849s | train_loss=0.0166
427
+ step 4250000 | 42944s | train_loss=0.0164
428
+ step 4260000 | 43039s | train_loss=0.0162
429
+ step 4270000 | 43133s | train_loss=0.0172
430
+ step 4280000 | 43228s | train_loss=0.0165
431
+ step 4290000 | 43323s | train_loss=0.0165
432
+ step 4300000 | 43490s | train_loss=0.0163 | fid=41.8944
433
+ step 4310000 | 43584s | train_loss=0.0170
434
+ step 4320000 | 43674s | train_loss=0.0167
435
+ step 4330000 | 43768s | train_loss=0.0175
436
+ step 4340000 | 43861s | train_loss=0.0166
437
+ step 4350000 | 43955s | train_loss=0.0159
438
+ step 4360000 | 44050s | train_loss=0.0157
439
+ step 4370000 | 44145s | train_loss=0.0174
440
+ step 4380000 | 44239s | train_loss=0.0182
441
+ step 4390000 | 44334s | train_loss=0.0160
442
+ step 4400000 | 44499s | train_loss=0.0161 | fid=42.7789
443
+ step 4410000 | 44592s | train_loss=0.0164
444
+ step 4420000 | 44685s | train_loss=0.0160
445
+ step 4430000 | 44777s | train_loss=0.0167
446
+ step 4440000 | 44872s | train_loss=0.0162
447
+ step 4450000 | 44967s | train_loss=0.0175
448
+ step 4460000 | 45062s | train_loss=0.0166
449
+ step 4470000 | 45158s | train_loss=0.0167
450
+ step 4480000 | 45252s | train_loss=0.0157
451
+ step 4490000 | 45347s | train_loss=0.0159
452
+ step 4500000 | 45513s | train_loss=0.0165 | fid=44.0331
453
+ step 4510000 | 45605s | train_loss=0.0187
454
+ step 4520000 | 45700s | train_loss=0.0154
455
+ step 4530000 | 45792s | train_loss=0.0172
456
+ step 4540000 | 45887s | train_loss=0.0172
457
+ step 4550000 | 45982s | train_loss=0.0154
458
+ step 4560000 | 46077s | train_loss=0.0165
459
+ step 4570000 | 46172s | train_loss=0.0179
460
+ step 4580000 | 46266s | train_loss=0.0162
461
+ step 4590000 | 46362s | train_loss=0.0173
462
+ step 4600000 | 46523s | train_loss=0.0158 | fid=40.9611
463
+ step 4610000 | 46617s | train_loss=0.0161
464
+ step 4620000 | 46710s | train_loss=0.0163
465
+ step 4630000 | 46804s | train_loss=0.0157
466
+ step 4640000 | 46899s | train_loss=0.0155
467
+ step 4650000 | 46993s | train_loss=0.0167
468
+ step 4660000 | 47088s | train_loss=0.0168
469
+ step 4670000 | 47183s | train_loss=0.0163
470
+ step 4680000 | 47277s | train_loss=0.0175
471
+ step 4690000 | 47371s | train_loss=0.0159
472
+ step 4700000 | 47491s | train_loss=0.0157 | fid=41.5115
473
+ step 4710000 | 47548s | train_loss=0.0154
474
+ step 4720000 | 47602s | train_loss=0.0156
475
+ step 4730000 | 47660s | train_loss=0.0161
476
+ step 4740000 | 47718s | train_loss=0.0160
477
+ step 4750000 | 47775s | train_loss=0.0167
478
+ step 4760000 | 47833s | train_loss=0.0153
479
+ step 4770000 | 47891s | train_loss=0.0153
480
+ step 4780000 | 47949s | train_loss=0.0161
481
+ step 4790000 | 48007s | train_loss=0.0151
482
+ step 4800000 | 48111s | train_loss=0.0155 | fid=42.5904
483
+ step 4810000 | 48167s | train_loss=0.0156
484
+ step 4820000 | 48206s | train_loss=0.0158
485
+ step 4830000 | 48246s | train_loss=0.0158
486
+ step 4840000 | 48286s | train_loss=0.0158
487
+ step 4850000 | 48325s | train_loss=0.0164
488
+ step 4860000 | 48364s | train_loss=0.0151
489
+ step 4870000 | 48403s | train_loss=0.0157
490
+ step 4880000 | 48442s | train_loss=0.0153
491
+ step 4890000 | 48481s | train_loss=0.0150
492
+ step 4900000 | 48553s | train_loss=0.0157 | fid=40.9877
493
+ step 4910000 | 48593s | train_loss=0.0144
494
+ step 4920000 | 48632s | train_loss=0.0171
495
+ step 4930000 | 48672s | train_loss=0.0172
496
+ step 4940000 | 48711s | train_loss=0.0148
497
+ step 4950000 | 48751s | train_loss=0.0165
498
+ step 4960000 | 48791s | train_loss=0.0151
499
+ step 4970000 | 48830s | train_loss=0.0153
500
+ step 4980000 | 48869s | train_loss=0.0151
501
+ step 4990000 | 48909s | train_loss=0.0150
502
+ step 5000000 | 48983s | train_loss=0.0159 | fid=42.4787
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/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/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": 42,
25
+ "device": null
26
+ }
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/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/d140/runner.log ADDED
@@ -0,0 +1,502 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ FID real-side stats cached over 1000 diverse train images
2
+ step 1 | 65s | train_loss=0.9987 | fid=310.2640
3
+ step 10000 | 154s | train_loss=0.7602
4
+ step 20000 | 242s | train_loss=0.6316
5
+ step 30000 | 331s | train_loss=0.5189
6
+ step 40000 | 419s | train_loss=0.4167
7
+ step 50000 | 508s | train_loss=0.3335
8
+ step 60000 | 597s | train_loss=0.2694
9
+ step 70000 | 685s | train_loss=0.2268
10
+ step 80000 | 774s | train_loss=0.1883
11
+ step 90000 | 862s | train_loss=0.1648
12
+ step 100000 | 1017s | train_loss=0.1469 | fid=61.5436
13
+ step 110000 | 1102s | train_loss=0.1331
14
+ step 120000 | 1191s | train_loss=0.1250
15
+ step 130000 | 1279s | train_loss=0.1209
16
+ step 140000 | 1367s | train_loss=0.1104
17
+ step 150000 | 1456s | train_loss=0.1026
18
+ step 160000 | 1545s | train_loss=0.1023
19
+ step 170000 | 1633s | train_loss=0.0978
20
+ step 180000 | 1722s | train_loss=0.0929
21
+ step 190000 | 1810s | train_loss=0.0959
22
+ step 200000 | 1965s | train_loss=0.0903 | fid=68.6268
23
+ step 210000 | 2052s | train_loss=0.0851
24
+ step 220000 | 2137s | train_loss=0.0823
25
+ step 230000 | 2225s | train_loss=0.0821
26
+ step 240000 | 2314s | train_loss=0.0805
27
+ step 250000 | 2402s | train_loss=0.0805
28
+ step 260000 | 2491s | train_loss=0.0767
29
+ step 270000 | 2579s | train_loss=0.0737
30
+ step 280000 | 2668s | train_loss=0.0727
31
+ step 290000 | 2756s | train_loss=0.0699
32
+ step 300000 | 2912s | train_loss=0.0710 | fid=66.8740
33
+ step 310000 | 2998s | train_loss=0.0697
34
+ step 320000 | 3087s | train_loss=0.0679
35
+ step 330000 | 3171s | train_loss=0.0680
36
+ step 340000 | 3259s | train_loss=0.0661
37
+ step 350000 | 3348s | train_loss=0.0659
38
+ step 360000 | 3436s | train_loss=0.0632
39
+ step 370000 | 3525s | train_loss=0.0626
40
+ step 380000 | 3613s | train_loss=0.0605
41
+ step 390000 | 3702s | train_loss=0.0610
42
+ step 400000 | 3857s | train_loss=0.0606 | fid=68.4837
43
+ step 410000 | 3944s | train_loss=0.0594
44
+ step 420000 | 4032s | train_loss=0.0572
45
+ step 430000 | 4117s | train_loss=0.0578
46
+ step 440000 | 4205s | train_loss=0.0578
47
+ step 450000 | 4294s | train_loss=0.0555
48
+ step 460000 | 4382s | train_loss=0.0559
49
+ step 470000 | 4471s | train_loss=0.0572
50
+ step 480000 | 4559s | train_loss=0.0556
51
+ step 490000 | 4648s | train_loss=0.0530
52
+ step 500000 | 4802s | train_loss=0.0524 | fid=67.7389
53
+ step 510000 | 4890s | train_loss=0.0515
54
+ step 520000 | 4976s | train_loss=0.0516
55
+ step 530000 | 5065s | train_loss=0.0523
56
+ step 540000 | 5150s | train_loss=0.0526
57
+ step 550000 | 5238s | train_loss=0.0503
58
+ step 560000 | 5327s | train_loss=0.0509
59
+ step 570000 | 5415s | train_loss=0.0498
60
+ step 580000 | 5504s | train_loss=0.0500
61
+ step 590000 | 5592s | train_loss=0.0497
62
+ step 600000 | 5745s | train_loss=0.0498 | fid=66.2141
63
+ step 610000 | 5834s | train_loss=0.0489
64
+ step 620000 | 5920s | train_loss=0.0496
65
+ step 630000 | 6009s | train_loss=0.0474
66
+ step 640000 | 6098s | train_loss=0.0467
67
+ step 650000 | 6181s | train_loss=0.0461
68
+ step 660000 | 6270s | train_loss=0.0454
69
+ step 670000 | 6358s | train_loss=0.0465
70
+ step 680000 | 6447s | train_loss=0.0486
71
+ step 690000 | 6535s | train_loss=0.0466
72
+ step 700000 | 6690s | train_loss=0.0437 | fid=65.2047
73
+ step 710000 | 6779s | train_loss=0.0447
74
+ step 720000 | 6865s | train_loss=0.0450
75
+ step 730000 | 6954s | train_loss=0.0460
76
+ step 740000 | 7042s | train_loss=0.0454
77
+ step 750000 | 7127s | train_loss=0.0458
78
+ step 760000 | 7216s | train_loss=0.0441
79
+ step 770000 | 7304s | train_loss=0.0427
80
+ step 780000 | 7392s | train_loss=0.0423
81
+ step 790000 | 7481s | train_loss=0.0420
82
+ step 800000 | 7635s | train_loss=0.0425 | fid=63.8116
83
+ step 810000 | 7724s | train_loss=0.0418
84
+ step 820000 | 7812s | train_loss=0.0429
85
+ step 830000 | 7899s | train_loss=0.0432
86
+ step 840000 | 7987s | train_loss=0.0410
87
+ step 850000 | 8076s | train_loss=0.0429
88
+ step 860000 | 8161s | train_loss=0.0394
89
+ step 870000 | 8250s | train_loss=0.0412
90
+ step 880000 | 8338s | train_loss=0.0424
91
+ step 890000 | 8427s | train_loss=0.0418
92
+ step 900000 | 8582s | train_loss=0.0407 | fid=62.2961
93
+ step 910000 | 8670s | train_loss=0.0418
94
+ step 920000 | 8758s | train_loss=0.0396
95
+ step 930000 | 8844s | train_loss=0.0383
96
+ step 940000 | 8933s | train_loss=0.0410
97
+ step 950000 | 9021s | train_loss=0.0389
98
+ step 960000 | 9110s | train_loss=0.0384
99
+ step 970000 | 9195s | train_loss=0.0388
100
+ step 980000 | 9284s | train_loss=0.0378
101
+ step 990000 | 9372s | train_loss=0.0383
102
+ step 1000000 | 9527s | train_loss=0.0383 | fid=62.0278
103
+ step 1010000 | 9615s | train_loss=0.0381
104
+ step 1020000 | 9703s | train_loss=0.0375
105
+ step 1030000 | 9790s | train_loss=0.0381
106
+ step 1040000 | 9878s | train_loss=0.0369
107
+ step 1050000 | 9967s | train_loss=0.0386
108
+ step 1060000 | 10055s | train_loss=0.0379
109
+ step 1070000 | 10144s | train_loss=0.0381
110
+ step 1080000 | 10229s | train_loss=0.0367
111
+ step 1090000 | 10318s | train_loss=0.0380
112
+ step 1100000 | 10474s | train_loss=0.0358 | fid=60.5712
113
+ step 1110000 | 10561s | train_loss=0.0347
114
+ step 1120000 | 10649s | train_loss=0.0360
115
+ step 1130000 | 10736s | train_loss=0.0354
116
+ step 1140000 | 10824s | train_loss=0.0357
117
+ step 1150000 | 10913s | train_loss=0.0351
118
+ step 1160000 | 11001s | train_loss=0.0349
119
+ step 1170000 | 11090s | train_loss=0.0338
120
+ step 1180000 | 11176s | train_loss=0.0348
121
+ step 1190000 | 11264s | train_loss=0.0346
122
+ step 1200000 | 11420s | train_loss=0.0350 | fid=60.5484
123
+ step 1210000 | 11507s | train_loss=0.0371
124
+ step 1220000 | 11596s | train_loss=0.0354
125
+ step 1230000 | 11684s | train_loss=0.0348
126
+ step 1240000 | 11770s | train_loss=0.0340
127
+ step 1250000 | 11859s | train_loss=0.0346
128
+ step 1260000 | 11948s | train_loss=0.0329
129
+ step 1270000 | 12036s | train_loss=0.0320
130
+ step 1280000 | 12125s | train_loss=0.0342
131
+ step 1290000 | 12210s | train_loss=0.0339
132
+ step 1300000 | 12365s | train_loss=0.0347 | fid=60.1411
133
+ step 1310000 | 12452s | train_loss=0.0346
134
+ step 1320000 | 12540s | train_loss=0.0325
135
+ step 1330000 | 12629s | train_loss=0.0331
136
+ step 1340000 | 12715s | train_loss=0.0330
137
+ step 1350000 | 12803s | train_loss=0.0326
138
+ step 1360000 | 12892s | train_loss=0.0337
139
+ step 1370000 | 12980s | train_loss=0.0314
140
+ step 1380000 | 13069s | train_loss=0.0309
141
+ step 1390000 | 13158s | train_loss=0.0324
142
+ step 1400000 | 13311s | train_loss=0.0320 | fid=61.5831
143
+ step 1410000 | 13397s | train_loss=0.0309
144
+ step 1420000 | 13486s | train_loss=0.0305
145
+ step 1430000 | 13574s | train_loss=0.0301
146
+ step 1440000 | 13660s | train_loss=0.0307
147
+ step 1450000 | 13749s | train_loss=0.0295
148
+ step 1460000 | 13838s | train_loss=0.0308
149
+ step 1470000 | 13926s | train_loss=0.0309
150
+ step 1480000 | 14015s | train_loss=0.0293
151
+ step 1490000 | 14103s | train_loss=0.0304
152
+ step 1500000 | 14257s | train_loss=0.0298 | fid=62.2280
153
+ step 1510000 | 14343s | train_loss=0.0302
154
+ step 1520000 | 14432s | train_loss=0.0303
155
+ step 1530000 | 14520s | train_loss=0.0293
156
+ step 1540000 | 14609s | train_loss=0.0302
157
+ step 1550000 | 14695s | train_loss=0.0308
158
+ step 1560000 | 14783s | train_loss=0.0312
159
+ step 1570000 | 14872s | train_loss=0.0304
160
+ step 1580000 | 14960s | train_loss=0.0315
161
+ step 1590000 | 15049s | train_loss=0.0286
162
+ step 1600000 | 15204s | train_loss=0.0293 | fid=61.4583
163
+ step 1610000 | 15291s | train_loss=0.0299
164
+ step 1620000 | 15379s | train_loss=0.0283
165
+ step 1630000 | 15468s | train_loss=0.0297
166
+ step 1640000 | 15556s | train_loss=0.0299
167
+ step 1650000 | 15642s | train_loss=0.0283
168
+ step 1660000 | 15731s | train_loss=0.0299
169
+ step 1670000 | 15820s | train_loss=0.0297
170
+ step 1680000 | 15908s | train_loss=0.0286
171
+ step 1690000 | 15997s | train_loss=0.0281
172
+ step 1700000 | 16153s | train_loss=0.0277 | fid=59.4070
173
+ step 1710000 | 16237s | train_loss=0.0282
174
+ step 1720000 | 16325s | train_loss=0.0284
175
+ step 1730000 | 16414s | train_loss=0.0293
176
+ step 1740000 | 16502s | train_loss=0.0296
177
+ step 1750000 | 16588s | train_loss=0.0297
178
+ step 1760000 | 16677s | train_loss=0.0287
179
+ step 1770000 | 16765s | train_loss=0.0277
180
+ step 1780000 | 16854s | train_loss=0.0272
181
+ step 1790000 | 16943s | train_loss=0.0273
182
+ step 1800000 | 17098s | train_loss=0.0282 | fid=60.9388
183
+ step 1810000 | 17185s | train_loss=0.0272
184
+ step 1820000 | 17270s | train_loss=0.0263
185
+ step 1830000 | 17359s | train_loss=0.0260
186
+ step 1840000 | 17447s | train_loss=0.0284
187
+ step 1850000 | 17535s | train_loss=0.0284
188
+ step 1860000 | 17622s | train_loss=0.0273
189
+ step 1870000 | 17710s | train_loss=0.0277
190
+ step 1880000 | 17799s | train_loss=0.0274
191
+ step 1890000 | 17887s | train_loss=0.0270
192
+ step 1900000 | 18044s | train_loss=0.0287 | fid=58.5706
193
+ step 1910000 | 18130s | train_loss=0.0280
194
+ step 1920000 | 18218s | train_loss=0.0273
195
+ step 1930000 | 18304s | train_loss=0.0280
196
+ step 1940000 | 18393s | train_loss=0.0260
197
+ step 1950000 | 18481s | train_loss=0.0274
198
+ step 1960000 | 18568s | train_loss=0.0257
199
+ step 1970000 | 18656s | train_loss=0.0262
200
+ step 1980000 | 18745s | train_loss=0.0259
201
+ step 1990000 | 18833s | train_loss=0.0253
202
+ step 2000000 | 18990s | train_loss=0.0254 | fid=58.3315
203
+ step 2010000 | 19076s | train_loss=0.0259
204
+ step 2020000 | 19165s | train_loss=0.0249
205
+ step 2030000 | 19251s | train_loss=0.0261
206
+ step 2040000 | 19339s | train_loss=0.0265
207
+ step 2050000 | 19428s | train_loss=0.0250
208
+ step 2060000 | 19514s | train_loss=0.0246
209
+ step 2070000 | 19602s | train_loss=0.0246
210
+ step 2080000 | 19691s | train_loss=0.0246
211
+ step 2090000 | 19779s | train_loss=0.0244
212
+ step 2100000 | 19936s | train_loss=0.0262 | fid=53.0892
213
+ step 2110000 | 20022s | train_loss=0.0252
214
+ step 2120000 | 20111s | train_loss=0.0249
215
+ step 2130000 | 20200s | train_loss=0.0241
216
+ step 2140000 | 20285s | train_loss=0.0239
217
+ step 2150000 | 20374s | train_loss=0.0246
218
+ step 2160000 | 20462s | train_loss=0.0241
219
+ step 2170000 | 20548s | train_loss=0.0236
220
+ step 2180000 | 20637s | train_loss=0.0238
221
+ step 2190000 | 20725s | train_loss=0.0235
222
+ step 2200000 | 20881s | train_loss=0.0248 | fid=53.3792
223
+ step 2210000 | 20967s | train_loss=0.0236
224
+ step 2220000 | 21056s | train_loss=0.0235
225
+ step 2230000 | 21144s | train_loss=0.0242
226
+ step 2240000 | 21234s | train_loss=0.0271
227
+ step 2250000 | 21319s | train_loss=0.0241
228
+ step 2260000 | 21408s | train_loss=0.0235
229
+ step 2270000 | 21494s | train_loss=0.0230
230
+ step 2280000 | 21582s | train_loss=0.0246
231
+ step 2290000 | 21671s | train_loss=0.0236
232
+ step 2300000 | 21827s | train_loss=0.0236 | fid=52.2370
233
+ step 2310000 | 21913s | train_loss=0.0226
234
+ step 2320000 | 22002s | train_loss=0.0236
235
+ step 2330000 | 22090s | train_loss=0.0233
236
+ step 2340000 | 22179s | train_loss=0.0226
237
+ step 2350000 | 22265s | train_loss=0.0233
238
+ step 2360000 | 22354s | train_loss=0.0240
239
+ step 2370000 | 22440s | train_loss=0.0233
240
+ step 2380000 | 22529s | train_loss=0.0235
241
+ step 2390000 | 22617s | train_loss=0.0231
242
+ step 2400000 | 22773s | train_loss=0.0219 | fid=51.1923
243
+ step 2410000 | 22859s | train_loss=0.0227
244
+ step 2420000 | 22947s | train_loss=0.0221
245
+ step 2430000 | 23036s | train_loss=0.0219
246
+ step 2440000 | 23125s | train_loss=0.0211
247
+ step 2450000 | 23213s | train_loss=0.0248
248
+ step 2460000 | 23299s | train_loss=0.0224
249
+ step 2470000 | 23387s | train_loss=0.0213
250
+ step 2480000 | 23474s | train_loss=0.0227
251
+ step 2490000 | 23562s | train_loss=0.0212
252
+ step 2500000 | 23719s | train_loss=0.0224 | fid=49.5937
253
+ step 2510000 | 23805s | train_loss=0.0223
254
+ step 2520000 | 23893s | train_loss=0.0212
255
+ step 2530000 | 23982s | train_loss=0.0227
256
+ step 2540000 | 24070s | train_loss=0.0239
257
+ step 2550000 | 24159s | train_loss=0.0224
258
+ step 2560000 | 24248s | train_loss=0.0226
259
+ step 2570000 | 24333s | train_loss=0.0220
260
+ step 2580000 | 24420s | train_loss=0.0215
261
+ step 2590000 | 24508s | train_loss=0.0215
262
+ step 2600000 | 24665s | train_loss=0.0215 | fid=48.7963
263
+ step 2610000 | 24751s | train_loss=0.0223
264
+ step 2620000 | 24840s | train_loss=0.0215
265
+ step 2630000 | 24928s | train_loss=0.0216
266
+ step 2640000 | 25017s | train_loss=0.0201
267
+ step 2650000 | 25105s | train_loss=0.0215
268
+ step 2660000 | 25194s | train_loss=0.0215
269
+ step 2670000 | 25282s | train_loss=0.0215
270
+ step 2680000 | 25367s | train_loss=0.0212
271
+ step 2690000 | 25456s | train_loss=0.0204
272
+ step 2700000 | 25612s | train_loss=0.0204 | fid=46.9669
273
+ step 2710000 | 25698s | train_loss=0.0211
274
+ step 2720000 | 25787s | train_loss=0.0212
275
+ step 2730000 | 25875s | train_loss=0.0216
276
+ step 2740000 | 25964s | train_loss=0.0208
277
+ step 2750000 | 26052s | train_loss=0.0215
278
+ step 2760000 | 26141s | train_loss=0.0206
279
+ step 2770000 | 26230s | train_loss=0.0221
280
+ step 2780000 | 26314s | train_loss=0.0218
281
+ step 2790000 | 26402s | train_loss=0.0199
282
+ step 2800000 | 26557s | train_loss=0.0210 | fid=46.5901
283
+ step 2810000 | 26644s | train_loss=0.0227
284
+ step 2820000 | 26732s | train_loss=0.0217
285
+ step 2830000 | 26821s | train_loss=0.0200
286
+ step 2840000 | 26909s | train_loss=0.0203
287
+ step 2850000 | 26998s | train_loss=0.0202
288
+ step 2860000 | 27086s | train_loss=0.0211
289
+ step 2870000 | 27175s | train_loss=0.0202
290
+ step 2880000 | 27264s | train_loss=0.0194
291
+ step 2890000 | 27347s | train_loss=0.0209
292
+ step 2900000 | 27503s | train_loss=0.0220 | fid=47.3844
293
+ step 2910000 | 27589s | train_loss=0.0206
294
+ step 2920000 | 27678s | train_loss=0.0208
295
+ step 2930000 | 27766s | train_loss=0.0190
296
+ step 2940000 | 27855s | train_loss=0.0194
297
+ step 2950000 | 27943s | train_loss=0.0205
298
+ step 2960000 | 28032s | train_loss=0.0194
299
+ step 2970000 | 28120s | train_loss=0.0209
300
+ step 2980000 | 28209s | train_loss=0.0201
301
+ step 2990000 | 28296s | train_loss=0.0198
302
+ step 3000000 | 28449s | train_loss=0.0190 | fid=45.2746
303
+ step 3010000 | 28536s | train_loss=0.0197
304
+ step 3020000 | 28624s | train_loss=0.0201
305
+ step 3030000 | 28713s | train_loss=0.0193
306
+ step 3040000 | 28801s | train_loss=0.0201
307
+ step 3050000 | 28890s | train_loss=0.0196
308
+ step 3060000 | 28978s | train_loss=0.0194
309
+ step 3070000 | 29067s | train_loss=0.0184
310
+ step 3080000 | 29155s | train_loss=0.0197
311
+ step 3090000 | 29242s | train_loss=0.0200
312
+ step 3100000 | 29395s | train_loss=0.0192 | fid=44.8147
313
+ step 3110000 | 29481s | train_loss=0.0194
314
+ step 3120000 | 29570s | train_loss=0.0204
315
+ step 3130000 | 29658s | train_loss=0.0185
316
+ step 3140000 | 29747s | train_loss=0.0189
317
+ step 3150000 | 29835s | train_loss=0.0202
318
+ step 3160000 | 29924s | train_loss=0.0178
319
+ step 3170000 | 30012s | train_loss=0.0194
320
+ step 3180000 | 30101s | train_loss=0.0194
321
+ step 3190000 | 30190s | train_loss=0.0199
322
+ step 3200000 | 30344s | train_loss=0.0196 | fid=44.6607
323
+ step 3210000 | 30429s | train_loss=0.0188
324
+ step 3220000 | 30518s | train_loss=0.0191
325
+ step 3230000 | 30607s | train_loss=0.0192
326
+ step 3240000 | 30695s | train_loss=0.0196
327
+ step 3250000 | 30784s | train_loss=0.0197
328
+ step 3260000 | 30872s | train_loss=0.0184
329
+ step 3270000 | 30961s | train_loss=0.0181
330
+ step 3280000 | 31049s | train_loss=0.0182
331
+ step 3290000 | 31138s | train_loss=0.0186
332
+ step 3300000 | 31293s | train_loss=0.0184 | fid=45.0984
333
+ step 3310000 | 31374s | train_loss=0.0196
334
+ step 3320000 | 31463s | train_loss=0.0182
335
+ step 3330000 | 31551s | train_loss=0.0197
336
+ step 3340000 | 31640s | train_loss=0.0179
337
+ step 3350000 | 31728s | train_loss=0.0190
338
+ step 3360000 | 31817s | train_loss=0.0190
339
+ step 3370000 | 31905s | train_loss=0.0199
340
+ step 3380000 | 31994s | train_loss=0.0183
341
+ step 3390000 | 32082s | train_loss=0.0196
342
+ step 3400000 | 32238s | train_loss=0.0185 | fid=44.2625
343
+ step 3410000 | 32324s | train_loss=0.0193
344
+ step 3420000 | 32409s | train_loss=0.0178
345
+ step 3430000 | 32498s | train_loss=0.0182
346
+ step 3440000 | 32586s | train_loss=0.0184
347
+ step 3450000 | 32675s | train_loss=0.0194
348
+ step 3460000 | 32764s | train_loss=0.0185
349
+ step 3470000 | 32852s | train_loss=0.0178
350
+ step 3480000 | 32941s | train_loss=0.0197
351
+ step 3490000 | 33029s | train_loss=0.0175
352
+ step 3500000 | 33184s | train_loss=0.0186 | fid=43.8484
353
+ step 3510000 | 33270s | train_loss=0.0182
354
+ step 3520000 | 33360s | train_loss=0.0175
355
+ step 3530000 | 33445s | train_loss=0.0169
356
+ step 3540000 | 33533s | train_loss=0.0182
357
+ step 3550000 | 33622s | train_loss=0.0175
358
+ step 3560000 | 33710s | train_loss=0.0176
359
+ step 3570000 | 33799s | train_loss=0.0177
360
+ step 3580000 | 33887s | train_loss=0.0169
361
+ step 3590000 | 33976s | train_loss=0.0180
362
+ step 3600000 | 34132s | train_loss=0.0176 | fid=42.9946
363
+ step 3610000 | 34218s | train_loss=0.0178
364
+ step 3620000 | 34306s | train_loss=0.0179
365
+ step 3630000 | 34391s | train_loss=0.0174
366
+ step 3640000 | 34480s | train_loss=0.0201
367
+ step 3650000 | 34569s | train_loss=0.0173
368
+ step 3660000 | 34657s | train_loss=0.0176
369
+ step 3670000 | 34746s | train_loss=0.0183
370
+ step 3680000 | 34834s | train_loss=0.0182
371
+ step 3690000 | 34923s | train_loss=0.0185
372
+ step 3700000 | 35079s | train_loss=0.0168 | fid=41.1138
373
+ step 3710000 | 35166s | train_loss=0.0183
374
+ step 3720000 | 35254s | train_loss=0.0176
375
+ step 3730000 | 35343s | train_loss=0.0183
376
+ step 3740000 | 35427s | train_loss=0.0177
377
+ step 3750000 | 35516s | train_loss=0.0159
378
+ step 3760000 | 35604s | train_loss=0.0163
379
+ step 3770000 | 35693s | train_loss=0.0169
380
+ step 3780000 | 35781s | train_loss=0.0165
381
+ step 3790000 | 35870s | train_loss=0.0170
382
+ step 3800000 | 36029s | train_loss=0.0184 | fid=40.9244
383
+ step 3810000 | 36115s | train_loss=0.0168
384
+ step 3820000 | 36203s | train_loss=0.0177
385
+ step 3830000 | 36292s | train_loss=0.0173
386
+ step 3840000 | 36381s | train_loss=0.0172
387
+ step 3850000 | 36464s | train_loss=0.0175
388
+ step 3860000 | 36553s | train_loss=0.0168
389
+ step 3870000 | 36641s | train_loss=0.0164
390
+ step 3880000 | 36730s | train_loss=0.0178
391
+ step 3890000 | 36818s | train_loss=0.0170
392
+ step 3900000 | 36978s | train_loss=0.0163 | fid=42.9471
393
+ step 3910000 | 37062s | train_loss=0.0177
394
+ step 3920000 | 37151s | train_loss=0.0171
395
+ step 3930000 | 37239s | train_loss=0.0178
396
+ step 3940000 | 37328s | train_loss=0.0170
397
+ step 3950000 | 37412s | train_loss=0.0162
398
+ step 3960000 | 37501s | train_loss=0.0163
399
+ step 3970000 | 37589s | train_loss=0.0178
400
+ step 3980000 | 37678s | train_loss=0.0171
401
+ step 3990000 | 37766s | train_loss=0.0169
402
+ step 4000000 | 37924s | train_loss=0.0165 | fid=41.1020
403
+ step 4010000 | 38008s | train_loss=0.0168
404
+ step 4020000 | 38097s | train_loss=0.0169
405
+ step 4030000 | 38185s | train_loss=0.0171
406
+ step 4040000 | 38274s | train_loss=0.0170
407
+ step 4050000 | 38363s | train_loss=0.0164
408
+ step 4060000 | 38446s | train_loss=0.0163
409
+ step 4070000 | 38535s | train_loss=0.0167
410
+ step 4080000 | 38623s | train_loss=0.0169
411
+ step 4090000 | 38712s | train_loss=0.0165
412
+ step 4100000 | 38872s | train_loss=0.0156 | fid=40.4764
413
+ step 4110000 | 38956s | train_loss=0.0162
414
+ step 4120000 | 39044s | train_loss=0.0167
415
+ step 4130000 | 39133s | train_loss=0.0163
416
+ step 4140000 | 39221s | train_loss=0.0169
417
+ step 4150000 | 39310s | train_loss=0.0176
418
+ step 4160000 | 39399s | train_loss=0.0193
419
+ step 4170000 | 39483s | train_loss=0.0181
420
+ step 4180000 | 39571s | train_loss=0.0158
421
+ step 4190000 | 39660s | train_loss=0.0175
422
+ step 4200000 | 39818s | train_loss=0.0169 | fid=41.4063
423
+ step 4210000 | 39904s | train_loss=0.0172
424
+ step 4220000 | 39991s | train_loss=0.0156
425
+ step 4230000 | 40080s | train_loss=0.0165
426
+ step 4240000 | 40168s | train_loss=0.0164
427
+ step 4250000 | 40257s | train_loss=0.0155
428
+ step 4260000 | 40345s | train_loss=0.0163
429
+ step 4270000 | 40434s | train_loss=0.0159
430
+ step 4280000 | 40518s | train_loss=0.0158
431
+ step 4290000 | 40607s | train_loss=0.0160
432
+ step 4300000 | 40764s | train_loss=0.0159 | fid=40.6833
433
+ step 4310000 | 40851s | train_loss=0.0170
434
+ step 4320000 | 40938s | train_loss=0.0158
435
+ step 4330000 | 41026s | train_loss=0.0172
436
+ step 4340000 | 41115s | train_loss=0.0160
437
+ step 4350000 | 41203s | train_loss=0.0155
438
+ step 4360000 | 41292s | train_loss=0.0149
439
+ step 4370000 | 41381s | train_loss=0.0162
440
+ step 4380000 | 41465s | train_loss=0.0162
441
+ step 4390000 | 41554s | train_loss=0.0159
442
+ step 4400000 | 41712s | train_loss=0.0156 | fid=39.2172
443
+ step 4410000 | 41799s | train_loss=0.0167
444
+ step 4420000 | 41885s | train_loss=0.0154
445
+ step 4430000 | 41974s | train_loss=0.0161
446
+ step 4440000 | 42063s | train_loss=0.0160
447
+ step 4450000 | 42151s | train_loss=0.0176
448
+ step 4460000 | 42240s | train_loss=0.0160
449
+ step 4470000 | 42328s | train_loss=0.0153
450
+ step 4480000 | 42417s | train_loss=0.0158
451
+ step 4490000 | 42502s | train_loss=0.0151
452
+ step 4500000 | 42661s | train_loss=0.0161 | fid=39.6576
453
+ step 4510000 | 42748s | train_loss=0.0177
454
+ step 4520000 | 42835s | train_loss=0.0149
455
+ step 4530000 | 42923s | train_loss=0.0181
456
+ step 4540000 | 43012s | train_loss=0.0158
457
+ step 4550000 | 43100s | train_loss=0.0154
458
+ step 4560000 | 43189s | train_loss=0.0166
459
+ step 4570000 | 43277s | train_loss=0.0193
460
+ step 4580000 | 43366s | train_loss=0.0158
461
+ step 4590000 | 43455s | train_loss=0.0168
462
+ step 4600000 | 43608s | train_loss=0.0153 | fid=38.4480
463
+ step 4610000 | 43695s | train_loss=0.0158
464
+ step 4620000 | 43783s | train_loss=0.0163
465
+ step 4630000 | 43870s | train_loss=0.0153
466
+ step 4640000 | 43958s | train_loss=0.0150
467
+ step 4650000 | 44047s | train_loss=0.0150
468
+ step 4660000 | 44136s | train_loss=0.0161
469
+ step 4670000 | 44224s | train_loss=0.0162
470
+ step 4680000 | 44313s | train_loss=0.0156
471
+ step 4690000 | 44401s | train_loss=0.0151
472
+ step 4700000 | 44555s | train_loss=0.0165 | fid=39.6950
473
+ step 4710000 | 44642s | train_loss=0.0151
474
+ step 4720000 | 44730s | train_loss=0.0146
475
+ step 4730000 | 44817s | train_loss=0.0167
476
+ step 4740000 | 44905s | train_loss=0.0163
477
+ step 4750000 | 44994s | train_loss=0.0176
478
+ step 4760000 | 45083s | train_loss=0.0146
479
+ step 4770000 | 45171s | train_loss=0.0147
480
+ step 4780000 | 45260s | train_loss=0.0158
481
+ step 4790000 | 45348s | train_loss=0.0149
482
+ step 4800000 | 45504s | train_loss=0.0150 | fid=39.7873
483
+ step 4810000 | 45588s | train_loss=0.0158
484
+ step 4820000 | 45677s | train_loss=0.0143
485
+ step 4830000 | 45763s | train_loss=0.0157
486
+ step 4840000 | 45852s | train_loss=0.0156
487
+ step 4850000 | 45940s | train_loss=0.0158
488
+ step 4860000 | 46029s | train_loss=0.0143
489
+ step 4870000 | 46118s | train_loss=0.0155
490
+ step 4880000 | 46206s | train_loss=0.0155
491
+ step 4890000 | 46295s | train_loss=0.0145
492
+ step 4900000 | 46451s | train_loss=0.0151 | fid=40.0185
493
+ step 4910000 | 46533s | train_loss=0.0151
494
+ step 4920000 | 46622s | train_loss=0.0155
495
+ step 4930000 | 46709s | train_loss=0.0177
496
+ step 4940000 | 46798s | train_loss=0.0144
497
+ step 4950000 | 46886s | train_loss=0.0171
498
+ step 4960000 | 46975s | train_loss=0.0146
499
+ step 4970000 | 47064s | train_loss=0.0153
500
+ step 4980000 | 47152s | train_loss=0.0145
501
+ step 4990000 | 47241s | train_loss=0.0143
502
+ step 5000000 | 47400s | train_loss=0.0154 | fid=39.0545
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/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/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": 42,
25
+ "device": null
26
+ }
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/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/d160/runner.log ADDED
@@ -0,0 +1,502 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ FID real-side stats cached over 1000 diverse train images
2
+ step 1 | 67s | train_loss=0.9994 | fid=288.7424
3
+ step 10000 | 148s | train_loss=0.7238
4
+ step 20000 | 236s | train_loss=0.6126
5
+ step 30000 | 325s | train_loss=0.5318
6
+ step 40000 | 414s | train_loss=0.4539
7
+ step 50000 | 502s | train_loss=0.3803
8
+ step 60000 | 591s | train_loss=0.3190
9
+ step 70000 | 680s | train_loss=0.2726
10
+ step 80000 | 768s | train_loss=0.2292
11
+ step 90000 | 857s | train_loss=0.1969
12
+ step 100000 | 1015s | train_loss=0.1721 | fid=62.0849
13
+ step 110000 | 1094s | train_loss=0.1555
14
+ step 120000 | 1183s | train_loss=0.1442
15
+ step 130000 | 1271s | train_loss=0.1400
16
+ step 140000 | 1360s | train_loss=0.1276
17
+ step 150000 | 1448s | train_loss=0.1183
18
+ step 160000 | 1537s | train_loss=0.1123
19
+ step 170000 | 1626s | train_loss=0.1109
20
+ step 180000 | 1714s | train_loss=0.1012
21
+ step 190000 | 1803s | train_loss=0.1064
22
+ step 200000 | 1961s | train_loss=0.0941 | fid=81.2820
23
+ step 210000 | 2045s | train_loss=0.0936
24
+ step 220000 | 2130s | train_loss=0.0859
25
+ step 230000 | 2219s | train_loss=0.0938
26
+ step 240000 | 2307s | train_loss=0.0852
27
+ step 250000 | 2396s | train_loss=0.0795
28
+ step 260000 | 2484s | train_loss=0.0833
29
+ step 270000 | 2573s | train_loss=0.0770
30
+ step 280000 | 2661s | train_loss=0.0800
31
+ step 290000 | 2750s | train_loss=0.0723
32
+ step 300000 | 2908s | train_loss=0.0735 | fid=87.3153
33
+ step 310000 | 2992s | train_loss=0.0727
34
+ step 320000 | 3081s | train_loss=0.0687
35
+ step 330000 | 3165s | train_loss=0.0727
36
+ step 340000 | 3254s | train_loss=0.0664
37
+ step 350000 | 3342s | train_loss=0.0666
38
+ step 360000 | 3431s | train_loss=0.0643
39
+ step 370000 | 3519s | train_loss=0.0628
40
+ step 380000 | 3608s | train_loss=0.0620
41
+ step 390000 | 3697s | train_loss=0.0613
42
+ step 400000 | 3854s | train_loss=0.0586 | fid=90.5996
43
+ step 410000 | 3938s | train_loss=0.0595
44
+ step 420000 | 4027s | train_loss=0.0564
45
+ step 430000 | 4112s | train_loss=0.0625
46
+ step 440000 | 4200s | train_loss=0.0586
47
+ step 450000 | 4289s | train_loss=0.0560
48
+ step 460000 | 4377s | train_loss=0.0564
49
+ step 470000 | 4466s | train_loss=0.0612
50
+ step 480000 | 4554s | train_loss=0.0566
51
+ step 490000 | 4643s | train_loss=0.0554
52
+ step 500000 | 4800s | train_loss=0.0541 | fid=85.8195
53
+ step 510000 | 4887s | train_loss=0.0506
54
+ step 520000 | 4973s | train_loss=0.0539
55
+ step 530000 | 5062s | train_loss=0.0515
56
+ step 540000 | 5147s | train_loss=0.0542
57
+ step 550000 | 5235s | train_loss=0.0527
58
+ step 560000 | 5324s | train_loss=0.0519
59
+ step 570000 | 5412s | train_loss=0.0516
60
+ step 580000 | 5501s | train_loss=0.0517
61
+ step 590000 | 5589s | train_loss=0.0485
62
+ step 600000 | 5747s | train_loss=0.0500 | fid=80.3299
63
+ step 610000 | 5835s | train_loss=0.0498
64
+ step 620000 | 5921s | train_loss=0.0488
65
+ step 630000 | 6009s | train_loss=0.0471
66
+ step 640000 | 6099s | train_loss=0.0474
67
+ step 650000 | 6182s | train_loss=0.0477
68
+ step 660000 | 6271s | train_loss=0.0465
69
+ step 670000 | 6359s | train_loss=0.0479
70
+ step 680000 | 6448s | train_loss=0.0445
71
+ step 690000 | 6536s | train_loss=0.0443
72
+ step 700000 | 6694s | train_loss=0.0425 | fid=78.4211
73
+ step 710000 | 6782s | train_loss=0.0428
74
+ step 720000 | 6868s | train_loss=0.0466
75
+ step 730000 | 6957s | train_loss=0.0469
76
+ step 740000 | 7045s | train_loss=0.0437
77
+ step 750000 | 7131s | train_loss=0.0464
78
+ step 760000 | 7219s | train_loss=0.0416
79
+ step 770000 | 7307s | train_loss=0.0410
80
+ step 780000 | 7396s | train_loss=0.0436
81
+ step 790000 | 7484s | train_loss=0.0402
82
+ step 800000 | 7642s | train_loss=0.0417 | fid=74.8254
83
+ step 810000 | 7730s | train_loss=0.0409
84
+ step 820000 | 7819s | train_loss=0.0410
85
+ step 830000 | 7905s | train_loss=0.0419
86
+ step 840000 | 7994s | train_loss=0.0399
87
+ step 850000 | 8083s | train_loss=0.0422
88
+ step 860000 | 8168s | train_loss=0.0396
89
+ step 870000 | 8256s | train_loss=0.0402
90
+ step 880000 | 8345s | train_loss=0.0416
91
+ step 890000 | 8433s | train_loss=0.0410
92
+ step 900000 | 8591s | train_loss=0.0383 | fid=69.6789
93
+ step 910000 | 8680s | train_loss=0.0423
94
+ step 920000 | 8768s | train_loss=0.0403
95
+ step 930000 | 8855s | train_loss=0.0388
96
+ step 940000 | 8943s | train_loss=0.0377
97
+ step 950000 | 9032s | train_loss=0.0382
98
+ step 960000 | 9121s | train_loss=0.0387
99
+ step 970000 | 9206s | train_loss=0.0391
100
+ step 980000 | 9294s | train_loss=0.0382
101
+ step 990000 | 9383s | train_loss=0.0384
102
+ step 1000000 | 9539s | train_loss=0.0403 | fid=71.5882
103
+ step 1010000 | 9628s | train_loss=0.0372
104
+ step 1020000 | 9716s | train_loss=0.0357
105
+ step 1030000 | 9802s | train_loss=0.0389
106
+ step 1040000 | 9891s | train_loss=0.0367
107
+ step 1050000 | 9980s | train_loss=0.0403
108
+ step 1060000 | 10068s | train_loss=0.0369
109
+ step 1070000 | 10155s | train_loss=0.0394
110
+ step 1080000 | 10242s | train_loss=0.0347
111
+ step 1090000 | 10331s | train_loss=0.0359
112
+ step 1100000 | 10487s | train_loss=0.0347 | fid=65.8554
113
+ step 1110000 | 10576s | train_loss=0.0344
114
+ step 1120000 | 10664s | train_loss=0.0350
115
+ step 1130000 | 10750s | train_loss=0.0372
116
+ step 1140000 | 10839s | train_loss=0.0363
117
+ step 1150000 | 10928s | train_loss=0.0357
118
+ step 1160000 | 11016s | train_loss=0.0357
119
+ step 1170000 | 11105s | train_loss=0.0338
120
+ step 1180000 | 11191s | train_loss=0.0354
121
+ step 1190000 | 11279s | train_loss=0.0367
122
+ step 1200000 | 11435s | train_loss=0.0345 | fid=66.0838
123
+ step 1210000 | 11524s | train_loss=0.0358
124
+ step 1220000 | 11612s | train_loss=0.0343
125
+ step 1230000 | 11701s | train_loss=0.0343
126
+ step 1240000 | 11787s | train_loss=0.0345
127
+ step 1250000 | 11876s | train_loss=0.0333
128
+ step 1260000 | 11964s | train_loss=0.0336
129
+ step 1270000 | 12053s | train_loss=0.0323
130
+ step 1280000 | 12142s | train_loss=0.0337
131
+ step 1290000 | 12227s | train_loss=0.0340
132
+ step 1300000 | 12383s | train_loss=0.0338 | fid=65.2165
133
+ step 1310000 | 12472s | train_loss=0.0364
134
+ step 1320000 | 12560s | train_loss=0.0320
135
+ step 1330000 | 12649s | train_loss=0.0335
136
+ step 1340000 | 12735s | train_loss=0.0356
137
+ step 1350000 | 12824s | train_loss=0.0315
138
+ step 1360000 | 12912s | train_loss=0.0330
139
+ step 1370000 | 13001s | train_loss=0.0325
140
+ step 1380000 | 13089s | train_loss=0.0320
141
+ step 1390000 | 13178s | train_loss=0.0333
142
+ step 1400000 | 13331s | train_loss=0.0303 | fid=61.5521
143
+ step 1410000 | 13420s | train_loss=0.0319
144
+ step 1420000 | 13508s | train_loss=0.0320
145
+ step 1430000 | 13597s | train_loss=0.0310
146
+ step 1440000 | 13683s | train_loss=0.0297
147
+ step 1450000 | 13772s | train_loss=0.0299
148
+ step 1460000 | 13860s | train_loss=0.0304
149
+ step 1470000 | 13949s | train_loss=0.0308
150
+ step 1480000 | 14038s | train_loss=0.0294
151
+ step 1490000 | 14126s | train_loss=0.0320
152
+ step 1500000 | 14279s | train_loss=0.0305 | fid=61.4227
153
+ step 1510000 | 14368s | train_loss=0.0316
154
+ step 1520000 | 14456s | train_loss=0.0303
155
+ step 1530000 | 14545s | train_loss=0.0304
156
+ step 1540000 | 14632s | train_loss=0.0308
157
+ step 1550000 | 14720s | train_loss=0.0303
158
+ step 1560000 | 14808s | train_loss=0.0335
159
+ step 1570000 | 14897s | train_loss=0.0297
160
+ step 1580000 | 14985s | train_loss=0.0306
161
+ step 1590000 | 15074s | train_loss=0.0294
162
+ step 1600000 | 15227s | train_loss=0.0281 | fid=59.4362
163
+ step 1610000 | 15315s | train_loss=0.0309
164
+ step 1620000 | 15404s | train_loss=0.0303
165
+ step 1630000 | 15493s | train_loss=0.0313
166
+ step 1640000 | 15581s | train_loss=0.0300
167
+ step 1650000 | 15667s | train_loss=0.0297
168
+ step 1660000 | 15756s | train_loss=0.0286
169
+ step 1670000 | 15845s | train_loss=0.0292
170
+ step 1680000 | 15933s | train_loss=0.0278
171
+ step 1690000 | 16022s | train_loss=0.0300
172
+ step 1700000 | 16178s | train_loss=0.0287 | fid=61.4846
173
+ step 1710000 | 16264s | train_loss=0.0289
174
+ step 1720000 | 16352s | train_loss=0.0289
175
+ step 1730000 | 16441s | train_loss=0.0294
176
+ step 1740000 | 16529s | train_loss=0.0290
177
+ step 1750000 | 16616s | train_loss=0.0298
178
+ step 1760000 | 16704s | train_loss=0.0288
179
+ step 1770000 | 16793s | train_loss=0.0292
180
+ step 1780000 | 16881s | train_loss=0.0297
181
+ step 1790000 | 16970s | train_loss=0.0288
182
+ step 1800000 | 17126s | train_loss=0.0272 | fid=63.7984
183
+ step 1810000 | 17214s | train_loss=0.0291
184
+ step 1820000 | 17301s | train_loss=0.0270
185
+ step 1830000 | 17389s | train_loss=0.0271
186
+ step 1840000 | 17478s | train_loss=0.0286
187
+ step 1850000 | 17564s | train_loss=0.0309
188
+ step 1860000 | 17653s | train_loss=0.0296
189
+ step 1870000 | 17741s | train_loss=0.0277
190
+ step 1880000 | 17830s | train_loss=0.0263
191
+ step 1890000 | 17918s | train_loss=0.0264
192
+ step 1900000 | 18074s | train_loss=0.0286 | fid=62.1022
193
+ step 1910000 | 18163s | train_loss=0.0296
194
+ step 1920000 | 18249s | train_loss=0.0274
195
+ step 1930000 | 18337s | train_loss=0.0283
196
+ step 1940000 | 18426s | train_loss=0.0265
197
+ step 1950000 | 18514s | train_loss=0.0276
198
+ step 1960000 | 18601s | train_loss=0.0274
199
+ step 1970000 | 18689s | train_loss=0.0262
200
+ step 1980000 | 18778s | train_loss=0.0260
201
+ step 1990000 | 18866s | train_loss=0.0254
202
+ step 2000000 | 19022s | train_loss=0.0267 | fid=62.5400
203
+ step 2010000 | 19111s | train_loss=0.0273
204
+ step 2020000 | 19200s | train_loss=0.0270
205
+ step 2030000 | 19285s | train_loss=0.0245
206
+ step 2040000 | 19374s | train_loss=0.0282
207
+ step 2050000 | 19463s | train_loss=0.0261
208
+ step 2060000 | 19549s | train_loss=0.0262
209
+ step 2070000 | 19637s | train_loss=0.0253
210
+ step 2080000 | 19726s | train_loss=0.0251
211
+ step 2090000 | 19814s | train_loss=0.0246
212
+ step 2100000 | 19970s | train_loss=0.0270 | fid=58.8063
213
+ step 2110000 | 20059s | train_loss=0.0263
214
+ step 2120000 | 20147s | train_loss=0.0266
215
+ step 2130000 | 20235s | train_loss=0.0252
216
+ step 2140000 | 20322s | train_loss=0.0247
217
+ step 2150000 | 20410s | train_loss=0.0243
218
+ step 2160000 | 20497s | train_loss=0.0254
219
+ step 2170000 | 20585s | train_loss=0.0240
220
+ step 2180000 | 20674s | train_loss=0.0265
221
+ step 2190000 | 20762s | train_loss=0.0258
222
+ step 2200000 | 20918s | train_loss=0.0265 | fid=58.1897
223
+ step 2210000 | 21007s | train_loss=0.0260
224
+ step 2220000 | 21096s | train_loss=0.0233
225
+ step 2230000 | 21184s | train_loss=0.0251
226
+ step 2240000 | 21270s | train_loss=0.0263
227
+ step 2250000 | 21359s | train_loss=0.0253
228
+ step 2260000 | 21447s | train_loss=0.0221
229
+ step 2270000 | 21534s | train_loss=0.0241
230
+ step 2280000 | 21622s | train_loss=0.0246
231
+ step 2290000 | 21711s | train_loss=0.0233
232
+ step 2300000 | 21867s | train_loss=0.0252 | fid=59.6530
233
+ step 2310000 | 21955s | train_loss=0.0242
234
+ step 2320000 | 22044s | train_loss=0.0243
235
+ step 2330000 | 22132s | train_loss=0.0241
236
+ step 2340000 | 22222s | train_loss=0.0238
237
+ step 2350000 | 22307s | train_loss=0.0234
238
+ step 2360000 | 22396s | train_loss=0.0235
239
+ step 2370000 | 22482s | train_loss=0.0245
240
+ step 2380000 | 22571s | train_loss=0.0248
241
+ step 2390000 | 22660s | train_loss=0.0243
242
+ step 2400000 | 22816s | train_loss=0.0232 | fid=59.9169
243
+ step 2410000 | 22904s | train_loss=0.0241
244
+ step 2420000 | 22993s | train_loss=0.0237
245
+ step 2430000 | 23082s | train_loss=0.0215
246
+ step 2440000 | 23170s | train_loss=0.0210
247
+ step 2450000 | 23260s | train_loss=0.0230
248
+ step 2460000 | 23345s | train_loss=0.0230
249
+ step 2470000 | 23431s | train_loss=0.0228
250
+ step 2480000 | 23519s | train_loss=0.0227
251
+ step 2490000 | 23608s | train_loss=0.0218
252
+ step 2500000 | 23764s | train_loss=0.0221 | fid=61.1406
253
+ step 2510000 | 23852s | train_loss=0.0237
254
+ step 2520000 | 23941s | train_loss=0.0221
255
+ step 2530000 | 24029s | train_loss=0.0239
256
+ step 2540000 | 24118s | train_loss=0.0244
257
+ step 2550000 | 24207s | train_loss=0.0242
258
+ step 2560000 | 24292s | train_loss=0.0225
259
+ step 2570000 | 24379s | train_loss=0.0243
260
+ step 2580000 | 24468s | train_loss=0.0220
261
+ step 2590000 | 24556s | train_loss=0.0217
262
+ step 2600000 | 24712s | train_loss=0.0221 | fid=60.5832
263
+ step 2610000 | 24801s | train_loss=0.0236
264
+ step 2620000 | 24889s | train_loss=0.0232
265
+ step 2630000 | 24978s | train_loss=0.0222
266
+ step 2640000 | 25066s | train_loss=0.0215
267
+ step 2650000 | 25155s | train_loss=0.0228
268
+ step 2660000 | 25244s | train_loss=0.0212
269
+ step 2670000 | 25330s | train_loss=0.0216
270
+ step 2680000 | 25417s | train_loss=0.0226
271
+ step 2690000 | 25505s | train_loss=0.0220
272
+ step 2700000 | 25661s | train_loss=0.0217 | fid=57.6142
273
+ step 2710000 | 25750s | train_loss=0.0216
274
+ step 2720000 | 25838s | train_loss=0.0215
275
+ step 2730000 | 25927s | train_loss=0.0236
276
+ step 2740000 | 26015s | train_loss=0.0230
277
+ step 2750000 | 26104s | train_loss=0.0217
278
+ step 2760000 | 26193s | train_loss=0.0231
279
+ step 2770000 | 26282s | train_loss=0.0220
280
+ step 2780000 | 26365s | train_loss=0.0222
281
+ step 2790000 | 26454s | train_loss=0.0206
282
+ step 2800000 | 26610s | train_loss=0.0226 | fid=62.1038
283
+ step 2810000 | 26699s | train_loss=0.0229
284
+ step 2820000 | 26787s | train_loss=0.0232
285
+ step 2830000 | 26876s | train_loss=0.0210
286
+ step 2840000 | 26964s | train_loss=0.0217
287
+ step 2850000 | 27053s | train_loss=0.0198
288
+ step 2860000 | 27142s | train_loss=0.0228
289
+ step 2870000 | 27230s | train_loss=0.0212
290
+ step 2880000 | 27316s | train_loss=0.0207
291
+ step 2890000 | 27403s | train_loss=0.0205
292
+ step 2900000 | 27559s | train_loss=0.0239 | fid=61.8681
293
+ step 2910000 | 27648s | train_loss=0.0226
294
+ step 2920000 | 27736s | train_loss=0.0218
295
+ step 2930000 | 27825s | train_loss=0.0193
296
+ step 2940000 | 27913s | train_loss=0.0193
297
+ step 2950000 | 28002s | train_loss=0.0218
298
+ step 2960000 | 28091s | train_loss=0.0202
299
+ step 2970000 | 28179s | train_loss=0.0244
300
+ step 2980000 | 28266s | train_loss=0.0223
301
+ step 2990000 | 28352s | train_loss=0.0206
302
+ step 3000000 | 28508s | train_loss=0.0200 | fid=57.8924
303
+ step 3010000 | 28596s | train_loss=0.0222
304
+ step 3020000 | 28685s | train_loss=0.0207
305
+ step 3030000 | 28774s | train_loss=0.0209
306
+ step 3040000 | 28862s | train_loss=0.0211
307
+ step 3050000 | 28951s | train_loss=0.0218
308
+ step 3060000 | 29039s | train_loss=0.0220
309
+ step 3070000 | 29128s | train_loss=0.0193
310
+ step 3080000 | 29216s | train_loss=0.0197
311
+ step 3090000 | 29304s | train_loss=0.0226
312
+ step 3100000 | 29457s | train_loss=0.0199 | fid=60.4680
313
+ step 3110000 | 29546s | train_loss=0.0195
314
+ step 3120000 | 29634s | train_loss=0.0213
315
+ step 3130000 | 29723s | train_loss=0.0194
316
+ step 3140000 | 29811s | train_loss=0.0202
317
+ step 3150000 | 29900s | train_loss=0.0205
318
+ step 3160000 | 29989s | train_loss=0.0195
319
+ step 3170000 | 30077s | train_loss=0.0203
320
+ step 3180000 | 30166s | train_loss=0.0199
321
+ step 3190000 | 30252s | train_loss=0.0217
322
+ step 3200000 | 30404s | train_loss=0.0209 | fid=53.0846
323
+ step 3210000 | 30493s | train_loss=0.0200
324
+ step 3220000 | 30582s | train_loss=0.0191
325
+ step 3230000 | 30670s | train_loss=0.0195
326
+ step 3240000 | 30759s | train_loss=0.0202
327
+ step 3250000 | 30847s | train_loss=0.0197
328
+ step 3260000 | 30936s | train_loss=0.0197
329
+ step 3270000 | 31025s | train_loss=0.0188
330
+ step 3280000 | 31113s | train_loss=0.0190
331
+ step 3290000 | 31200s | train_loss=0.0206
332
+ step 3300000 | 31357s | train_loss=0.0193 | fid=56.3711
333
+ step 3310000 | 31444s | train_loss=0.0203
334
+ step 3320000 | 31533s | train_loss=0.0197
335
+ step 3330000 | 31622s | train_loss=0.0211
336
+ step 3340000 | 31710s | train_loss=0.0189
337
+ step 3350000 | 31799s | train_loss=0.0221
338
+ step 3360000 | 31887s | train_loss=0.0206
339
+ step 3370000 | 31976s | train_loss=0.0211
340
+ step 3380000 | 32065s | train_loss=0.0192
341
+ step 3390000 | 32152s | train_loss=0.0202
342
+ step 3400000 | 32307s | train_loss=0.0189 | fid=51.8910
343
+ step 3410000 | 32393s | train_loss=0.0207
344
+ step 3420000 | 32481s | train_loss=0.0209
345
+ step 3430000 | 32570s | train_loss=0.0192
346
+ step 3440000 | 32658s | train_loss=0.0192
347
+ step 3450000 | 32747s | train_loss=0.0206
348
+ step 3460000 | 32836s | train_loss=0.0202
349
+ step 3470000 | 32924s | train_loss=0.0186
350
+ step 3480000 | 33013s | train_loss=0.0213
351
+ step 3490000 | 33101s | train_loss=0.0189
352
+ step 3500000 | 33255s | train_loss=0.0210 | fid=54.9984
353
+ step 3510000 | 33344s | train_loss=0.0192
354
+ step 3520000 | 33430s | train_loss=0.0190
355
+ step 3530000 | 33518s | train_loss=0.0184
356
+ step 3540000 | 33607s | train_loss=0.0199
357
+ step 3550000 | 33696s | train_loss=0.0212
358
+ step 3560000 | 33784s | train_loss=0.0190
359
+ step 3570000 | 33873s | train_loss=0.0193
360
+ step 3580000 | 33961s | train_loss=0.0174
361
+ step 3590000 | 34050s | train_loss=0.0197
362
+ step 3600000 | 34204s | train_loss=0.0188 | fid=61.8987
363
+ step 3610000 | 34292s | train_loss=0.0200
364
+ step 3620000 | 34381s | train_loss=0.0194
365
+ step 3630000 | 34466s | train_loss=0.0182
366
+ step 3640000 | 34554s | train_loss=0.0227
367
+ step 3650000 | 34643s | train_loss=0.0180
368
+ step 3660000 | 34732s | train_loss=0.0187
369
+ step 3670000 | 34820s | train_loss=0.0193
370
+ step 3680000 | 34909s | train_loss=0.0198
371
+ step 3690000 | 34997s | train_loss=0.0227
372
+ step 3700000 | 35151s | train_loss=0.0186 | fid=55.1708
373
+ step 3710000 | 35239s | train_loss=0.0209
374
+ step 3720000 | 35328s | train_loss=0.0189
375
+ step 3730000 | 35413s | train_loss=0.0209
376
+ step 3740000 | 35501s | train_loss=0.0189
377
+ step 3750000 | 35590s | train_loss=0.0188
378
+ step 3760000 | 35678s | train_loss=0.0177
379
+ step 3770000 | 35767s | train_loss=0.0177
380
+ step 3780000 | 35856s | train_loss=0.0174
381
+ step 3790000 | 35944s | train_loss=0.0190
382
+ step 3800000 | 36097s | train_loss=0.0197 | fid=57.5905
383
+ step 3810000 | 36185s | train_loss=0.0176
384
+ step 3820000 | 36274s | train_loss=0.0196
385
+ step 3830000 | 36363s | train_loss=0.0184
386
+ step 3840000 | 36446s | train_loss=0.0188
387
+ step 3850000 | 36535s | train_loss=0.0180
388
+ step 3860000 | 36624s | train_loss=0.0184
389
+ step 3870000 | 36712s | train_loss=0.0181
390
+ step 3880000 | 36801s | train_loss=0.0187
391
+ step 3890000 | 36889s | train_loss=0.0175
392
+ step 3900000 | 37043s | train_loss=0.0183 | fid=53.0515
393
+ step 3910000 | 37131s | train_loss=0.0201
394
+ step 3920000 | 37220s | train_loss=0.0174
395
+ step 3930000 | 37309s | train_loss=0.0193
396
+ step 3940000 | 37398s | train_loss=0.0180
397
+ step 3950000 | 37481s | train_loss=0.0165
398
+ step 3960000 | 37570s | train_loss=0.0173
399
+ step 3970000 | 37659s | train_loss=0.0189
400
+ step 3980000 | 37747s | train_loss=0.0185
401
+ step 3990000 | 37836s | train_loss=0.0191
402
+ step 4000000 | 37990s | train_loss=0.0184 | fid=61.4458
403
+ step 4010000 | 38078s | train_loss=0.0176
404
+ step 4020000 | 38167s | train_loss=0.0169
405
+ step 4030000 | 38255s | train_loss=0.0190
406
+ step 4040000 | 38344s | train_loss=0.0191
407
+ step 4050000 | 38429s | train_loss=0.0177
408
+ step 4060000 | 38516s | train_loss=0.0184
409
+ step 4070000 | 38605s | train_loss=0.0171
410
+ step 4080000 | 38694s | train_loss=0.0183
411
+ step 4090000 | 38782s | train_loss=0.0169
412
+ step 4100000 | 38937s | train_loss=0.0178 | fid=55.4380
413
+ step 4110000 | 39024s | train_loss=0.0173
414
+ step 4120000 | 39113s | train_loss=0.0180
415
+ step 4130000 | 39201s | train_loss=0.0193
416
+ step 4140000 | 39290s | train_loss=0.0183
417
+ step 4150000 | 39379s | train_loss=0.0189
418
+ step 4160000 | 39463s | train_loss=0.0200
419
+ step 4170000 | 39552s | train_loss=0.0180
420
+ step 4180000 | 39641s | train_loss=0.0173
421
+ step 4190000 | 39729s | train_loss=0.0187
422
+ step 4200000 | 39883s | train_loss=0.0197 | fid=51.9560
423
+ step 4210000 | 39970s | train_loss=0.0185
424
+ step 4220000 | 40059s | train_loss=0.0168
425
+ step 4230000 | 40147s | train_loss=0.0181
426
+ step 4240000 | 40236s | train_loss=0.0177
427
+ step 4250000 | 40325s | train_loss=0.0174
428
+ step 4260000 | 40413s | train_loss=0.0178
429
+ step 4270000 | 40498s | train_loss=0.0174
430
+ step 4280000 | 40586s | train_loss=0.0178
431
+ step 4290000 | 40675s | train_loss=0.0174
432
+ step 4300000 | 40829s | train_loss=0.0180 | fid=51.2181
433
+ step 4310000 | 40916s | train_loss=0.0186
434
+ step 4320000 | 41004s | train_loss=0.0186
435
+ step 4330000 | 41093s | train_loss=0.0189
436
+ step 4340000 | 41182s | train_loss=0.0167
437
+ step 4350000 | 41270s | train_loss=0.0166
438
+ step 4360000 | 41359s | train_loss=0.0162
439
+ step 4370000 | 41448s | train_loss=0.0178
440
+ step 4380000 | 41532s | train_loss=0.0173
441
+ step 4390000 | 41621s | train_loss=0.0169
442
+ step 4400000 | 41773s | train_loss=0.0181 | fid=51.2979
443
+ step 4410000 | 41860s | train_loss=0.0167
444
+ step 4420000 | 41949s | train_loss=0.0166
445
+ step 4430000 | 42037s | train_loss=0.0168
446
+ step 4440000 | 42126s | train_loss=0.0189
447
+ step 4450000 | 42215s | train_loss=0.0192
448
+ step 4460000 | 42303s | train_loss=0.0178
449
+ step 4470000 | 42392s | train_loss=0.0178
450
+ step 4480000 | 42479s | train_loss=0.0172
451
+ step 4490000 | 42566s | train_loss=0.0167
452
+ step 4500000 | 42720s | train_loss=0.0182 | fid=51.7305
453
+ step 4510000 | 42808s | train_loss=0.0206
454
+ step 4520000 | 42895s | train_loss=0.0162
455
+ step 4530000 | 42983s | train_loss=0.0202
456
+ step 4540000 | 43072s | train_loss=0.0185
457
+ step 4550000 | 43161s | train_loss=0.0164
458
+ step 4560000 | 43249s | train_loss=0.0171
459
+ step 4570000 | 43338s | train_loss=0.0219
460
+ step 4580000 | 43427s | train_loss=0.0178
461
+ step 4590000 | 43512s | train_loss=0.0183
462
+ step 4600000 | 43666s | train_loss=0.0163 | fid=49.4397
463
+ step 4610000 | 43755s | train_loss=0.0173
464
+ step 4620000 | 43841s | train_loss=0.0193
465
+ step 4630000 | 43930s | train_loss=0.0192
466
+ step 4640000 | 44018s | train_loss=0.0164
467
+ step 4650000 | 44107s | train_loss=0.0166
468
+ step 4660000 | 44196s | train_loss=0.0175
469
+ step 4670000 | 44284s | train_loss=0.0166
470
+ step 4680000 | 44373s | train_loss=0.0174
471
+ step 4690000 | 44462s | train_loss=0.0175
472
+ step 4700000 | 44613s | train_loss=0.0185 | fid=48.5901
473
+ step 4710000 | 44702s | train_loss=0.0163
474
+ step 4720000 | 44789s | train_loss=0.0161
475
+ step 4730000 | 44877s | train_loss=0.0196
476
+ step 4740000 | 44966s | train_loss=0.0169
477
+ step 4750000 | 45054s | train_loss=0.0193
478
+ step 4760000 | 45143s | train_loss=0.0171
479
+ step 4770000 | 45232s | train_loss=0.0162
480
+ step 4780000 | 45320s | train_loss=0.0164
481
+ step 4790000 | 45409s | train_loss=0.0167
482
+ step 4800000 | 45559s | train_loss=0.0182 | fid=51.2203
483
+ step 4810000 | 45648s | train_loss=0.0161
484
+ step 4820000 | 45736s | train_loss=0.0157
485
+ step 4830000 | 45823s | train_loss=0.0169
486
+ step 4840000 | 45911s | train_loss=0.0165
487
+ step 4850000 | 46000s | train_loss=0.0165
488
+ step 4860000 | 46089s | train_loss=0.0162
489
+ step 4870000 | 46177s | train_loss=0.0163
490
+ step 4880000 | 46266s | train_loss=0.0165
491
+ step 4890000 | 46354s | train_loss=0.0158
492
+ step 4900000 | 46508s | train_loss=0.0167 | fid=47.9707
493
+ step 4910000 | 46595s | train_loss=0.0159
494
+ step 4920000 | 46683s | train_loss=0.0196
495
+ step 4930000 | 46771s | train_loss=0.0234
496
+ step 4940000 | 46859s | train_loss=0.0169
497
+ step 4950000 | 46948s | train_loss=0.0182
498
+ step 4960000 | 47037s | train_loss=0.0162
499
+ step 4970000 | 47126s | train_loss=0.0161
500
+ step 4980000 | 47214s | train_loss=0.0153
501
+ step 4990000 | 47303s | train_loss=0.0157
502
+ step 5000000 | 47449s | train_loss=0.0173 | fid=52.7509
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/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/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": 42,
25
+ "device": null
26
+ }
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/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/d180/runner.log ADDED
@@ -0,0 +1,502 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ FID real-side stats cached over 1000 diverse train images
2
+ step 1 | 68s | train_loss=0.9991 | fid=346.9662
3
+ step 10000 | 162s | train_loss=0.6684
4
+ step 20000 | 256s | train_loss=0.5598
5
+ step 30000 | 351s | train_loss=0.4965
6
+ step 40000 | 445s | train_loss=0.4387
7
+ step 50000 | 540s | train_loss=0.3660
8
+ step 60000 | 634s | train_loss=0.2967
9
+ step 70000 | 729s | train_loss=0.2468
10
+ step 80000 | 823s | train_loss=0.2180
11
+ step 90000 | 917s | train_loss=0.1853
12
+ step 100000 | 1076s | train_loss=0.1653 | fid=67.8639
13
+ step 110000 | 1169s | train_loss=0.1534
14
+ step 120000 | 1264s | train_loss=0.1453
15
+ step 130000 | 1358s | train_loss=0.1462
16
+ step 140000 | 1452s | train_loss=0.1231
17
+ step 150000 | 1546s | train_loss=0.1222
18
+ step 160000 | 1641s | train_loss=0.1141
19
+ step 170000 | 1735s | train_loss=0.1171
20
+ step 180000 | 1830s | train_loss=0.1042
21
+ step 190000 | 1924s | train_loss=0.1119
22
+ step 200000 | 2085s | train_loss=0.1009 | fid=80.2556
23
+ step 210000 | 2179s | train_loss=0.0966
24
+ step 220000 | 2273s | train_loss=0.0970
25
+ step 230000 | 2367s | train_loss=0.0929
26
+ step 240000 | 2461s | train_loss=0.0895
27
+ step 250000 | 2555s | train_loss=0.0975
28
+ step 260000 | 2650s | train_loss=0.0897
29
+ step 270000 | 2744s | train_loss=0.0875
30
+ step 280000 | 2839s | train_loss=0.0866
31
+ step 290000 | 2928s | train_loss=0.0864
32
+ step 300000 | 3095s | train_loss=0.0792 | fid=92.3924
33
+ step 310000 | 3188s | train_loss=0.0782
34
+ step 320000 | 3283s | train_loss=0.0820
35
+ step 330000 | 3376s | train_loss=0.0754
36
+ step 340000 | 3471s | train_loss=0.0731
37
+ step 350000 | 3565s | train_loss=0.0723
38
+ step 360000 | 3660s | train_loss=0.0760
39
+ step 370000 | 3754s | train_loss=0.0708
40
+ step 380000 | 3847s | train_loss=0.0769
41
+ step 390000 | 3935s | train_loss=0.0707
42
+ step 400000 | 4103s | train_loss=0.0689 | fid=107.0084
43
+ step 410000 | 4197s | train_loss=0.0669
44
+ step 420000 | 4291s | train_loss=0.0669
45
+ step 430000 | 4385s | train_loss=0.0682
46
+ step 440000 | 4479s | train_loss=0.0726
47
+ step 450000 | 4573s | train_loss=0.0656
48
+ step 460000 | 4668s | train_loss=0.0655
49
+ step 470000 | 4762s | train_loss=0.0632
50
+ step 480000 | 4852s | train_loss=0.0629
51
+ step 490000 | 4944s | train_loss=0.0636
52
+ step 500000 | 5113s | train_loss=0.0632 | fid=111.8782
53
+ step 510000 | 5207s | train_loss=0.0617
54
+ step 520000 | 5301s | train_loss=0.0580
55
+ step 530000 | 5395s | train_loss=0.0628
56
+ step 540000 | 5490s | train_loss=0.0599
57
+ step 550000 | 5584s | train_loss=0.0609
58
+ step 560000 | 5678s | train_loss=0.0605
59
+ step 570000 | 5768s | train_loss=0.0579
60
+ step 580000 | 5862s | train_loss=0.0578
61
+ step 590000 | 5954s | train_loss=0.0578
62
+ step 600000 | 6122s | train_loss=0.0556 | fid=117.2252
63
+ step 610000 | 6216s | train_loss=0.0555
64
+ step 620000 | 6310s | train_loss=0.0565
65
+ step 630000 | 6404s | train_loss=0.0536
66
+ step 640000 | 6498s | train_loss=0.0536
67
+ step 650000 | 6593s | train_loss=0.0550
68
+ step 660000 | 6686s | train_loss=0.0562
69
+ step 670000 | 6776s | train_loss=0.0600
70
+ step 680000 | 6868s | train_loss=0.0547
71
+ step 690000 | 6962s | train_loss=0.0563
72
+ step 700000 | 7129s | train_loss=0.0504 | fid=109.8627
73
+ step 710000 | 7224s | train_loss=0.0537
74
+ step 720000 | 7318s | train_loss=0.0525
75
+ step 730000 | 7411s | train_loss=0.0496
76
+ step 740000 | 7506s | train_loss=0.0494
77
+ step 750000 | 7600s | train_loss=0.0535
78
+ step 760000 | 7689s | train_loss=0.0504
79
+ step 770000 | 7783s | train_loss=0.0513
80
+ step 780000 | 7875s | train_loss=0.0489
81
+ step 790000 | 7969s | train_loss=0.0469
82
+ step 800000 | 8138s | train_loss=0.0489 | fid=108.7618
83
+ step 810000 | 8232s | train_loss=0.0480
84
+ step 820000 | 8326s | train_loss=0.0507
85
+ step 830000 | 8420s | train_loss=0.0464
86
+ step 840000 | 8515s | train_loss=0.0465
87
+ step 850000 | 8604s | train_loss=0.0469
88
+ step 860000 | 8699s | train_loss=0.0492
89
+ step 870000 | 8791s | train_loss=0.0481
90
+ step 880000 | 8884s | train_loss=0.0459
91
+ step 890000 | 8979s | train_loss=0.0483
92
+ step 900000 | 9145s | train_loss=0.0496 | fid=102.1423
93
+ step 910000 | 9239s | train_loss=0.0442
94
+ step 920000 | 9333s | train_loss=0.0541
95
+ step 930000 | 9427s | train_loss=0.0460
96
+ step 940000 | 9521s | train_loss=0.0486
97
+ step 950000 | 9611s | train_loss=0.0456
98
+ step 960000 | 9706s | train_loss=0.0497
99
+ step 970000 | 9798s | train_loss=0.0448
100
+ step 980000 | 9892s | train_loss=0.0420
101
+ step 990000 | 9987s | train_loss=0.0468
102
+ step 1000000 | 10155s | train_loss=0.0417 | fid=95.9305
103
+ step 1010000 | 10249s | train_loss=0.0434
104
+ step 1020000 | 10343s | train_loss=0.0478
105
+ step 1030000 | 10437s | train_loss=0.0487
106
+ step 1040000 | 10527s | train_loss=0.0417
107
+ step 1050000 | 10621s | train_loss=0.0440
108
+ step 1060000 | 10716s | train_loss=0.0443
109
+ step 1070000 | 10808s | train_loss=0.0426
110
+ step 1080000 | 10902s | train_loss=0.0445
111
+ step 1090000 | 10996s | train_loss=0.0404
112
+ step 1100000 | 11164s | train_loss=0.0405 | fid=99.3407
113
+ step 1110000 | 11258s | train_loss=0.0422
114
+ step 1120000 | 11352s | train_loss=0.0408
115
+ step 1130000 | 11442s | train_loss=0.0416
116
+ step 1140000 | 11536s | train_loss=0.0418
117
+ step 1150000 | 11630s | train_loss=0.0418
118
+ step 1160000 | 11722s | train_loss=0.0423
119
+ step 1170000 | 11817s | train_loss=0.0392
120
+ step 1180000 | 11911s | train_loss=0.0375
121
+ step 1190000 | 12006s | train_loss=0.0399
122
+ step 1200000 | 12174s | train_loss=0.0384 | fid=92.2979
123
+ step 1210000 | 12268s | train_loss=0.0399
124
+ step 1220000 | 12362s | train_loss=0.0387
125
+ step 1230000 | 12453s | train_loss=0.0409
126
+ step 1240000 | 12547s | train_loss=0.0375
127
+ step 1250000 | 12641s | train_loss=0.0390
128
+ step 1260000 | 12733s | train_loss=0.0385
129
+ step 1270000 | 12828s | train_loss=0.0391
130
+ step 1280000 | 12922s | train_loss=0.0380
131
+ step 1290000 | 13016s | train_loss=0.0389
132
+ step 1300000 | 13184s | train_loss=0.0384 | fid=83.3211
133
+ step 1310000 | 13279s | train_loss=0.0393
134
+ step 1320000 | 13368s | train_loss=0.0381
135
+ step 1330000 | 13463s | train_loss=0.0363
136
+ step 1340000 | 13557s | train_loss=0.0384
137
+ step 1350000 | 13651s | train_loss=0.0365
138
+ step 1360000 | 13743s | train_loss=0.0368
139
+ step 1370000 | 13837s | train_loss=0.0370
140
+ step 1380000 | 13932s | train_loss=0.0426
141
+ step 1390000 | 14027s | train_loss=0.0369
142
+ step 1400000 | 14194s | train_loss=0.0371 | fid=87.9211
143
+ step 1410000 | 14284s | train_loss=0.0357
144
+ step 1420000 | 14378s | train_loss=0.0375
145
+ step 1430000 | 14473s | train_loss=0.0342
146
+ step 1440000 | 14567s | train_loss=0.0378
147
+ step 1450000 | 14658s | train_loss=0.0385
148
+ step 1460000 | 14753s | train_loss=0.0348
149
+ step 1470000 | 14847s | train_loss=0.0341
150
+ step 1480000 | 14942s | train_loss=0.0362
151
+ step 1490000 | 15036s | train_loss=0.0429
152
+ step 1500000 | 15204s | train_loss=0.0349 | fid=81.1181
153
+ step 1510000 | 15295s | train_loss=0.0344
154
+ step 1520000 | 15390s | train_loss=0.0385
155
+ step 1530000 | 15484s | train_loss=0.0338
156
+ step 1540000 | 15578s | train_loss=0.0352
157
+ step 1550000 | 15670s | train_loss=0.0384
158
+ step 1560000 | 15765s | train_loss=0.0369
159
+ step 1570000 | 15859s | train_loss=0.0333
160
+ step 1580000 | 15954s | train_loss=0.0327
161
+ step 1590000 | 16048s | train_loss=0.0451
162
+ step 1600000 | 16211s | train_loss=0.0334 | fid=78.4777
163
+ step 1610000 | 16305s | train_loss=0.0337
164
+ step 1620000 | 16400s | train_loss=0.0342
165
+ step 1630000 | 16494s | train_loss=0.0347
166
+ step 1640000 | 16586s | train_loss=0.0326
167
+ step 1650000 | 16680s | train_loss=0.0343
168
+ step 1660000 | 16774s | train_loss=0.0336
169
+ step 1670000 | 16868s | train_loss=0.0310
170
+ step 1680000 | 16963s | train_loss=0.0331
171
+ step 1690000 | 17057s | train_loss=0.0331
172
+ step 1700000 | 17220s | train_loss=0.0325 | fid=80.6583
173
+ step 1710000 | 17315s | train_loss=0.0336
174
+ step 1720000 | 17409s | train_loss=0.0305
175
+ step 1730000 | 17503s | train_loss=0.0304
176
+ step 1740000 | 17595s | train_loss=0.0315
177
+ step 1750000 | 17689s | train_loss=0.0323
178
+ step 1760000 | 17783s | train_loss=0.0323
179
+ step 1770000 | 17878s | train_loss=0.0307
180
+ step 1780000 | 17972s | train_loss=0.0320
181
+ step 1790000 | 18064s | train_loss=0.0334
182
+ step 1800000 | 18229s | train_loss=0.0304 | fid=73.3772
183
+ step 1810000 | 18323s | train_loss=0.0310
184
+ step 1820000 | 18417s | train_loss=0.0308
185
+ step 1830000 | 18512s | train_loss=0.0371
186
+ step 1840000 | 18604s | train_loss=0.0336
187
+ step 1850000 | 18698s | train_loss=0.0376
188
+ step 1860000 | 18792s | train_loss=0.0306
189
+ step 1870000 | 18886s | train_loss=0.0324
190
+ step 1880000 | 18981s | train_loss=0.0367
191
+ step 1890000 | 19071s | train_loss=0.0301
192
+ step 1900000 | 19239s | train_loss=0.0376 | fid=69.5976
193
+ step 1910000 | 19333s | train_loss=0.0323
194
+ step 1920000 | 19428s | train_loss=0.0320
195
+ step 1930000 | 19520s | train_loss=0.0308
196
+ step 1940000 | 19614s | train_loss=0.0329
197
+ step 1950000 | 19708s | train_loss=0.0327
198
+ step 1960000 | 19802s | train_loss=0.0305
199
+ step 1970000 | 19896s | train_loss=0.0303
200
+ step 1980000 | 19987s | train_loss=0.0327
201
+ step 1990000 | 20081s | train_loss=0.0297
202
+ step 2000000 | 20248s | train_loss=0.0343 | fid=73.0184
203
+ step 2010000 | 20342s | train_loss=0.0281
204
+ step 2020000 | 20437s | train_loss=0.0273
205
+ step 2030000 | 20528s | train_loss=0.0273
206
+ step 2040000 | 20623s | train_loss=0.0295
207
+ step 2050000 | 20717s | train_loss=0.0287
208
+ step 2060000 | 20812s | train_loss=0.0319
209
+ step 2070000 | 20904s | train_loss=0.0315
210
+ step 2080000 | 20996s | train_loss=0.0273
211
+ step 2090000 | 21091s | train_loss=0.0316
212
+ step 2100000 | 21258s | train_loss=0.0270 | fid=67.3365
213
+ step 2110000 | 21352s | train_loss=0.0353
214
+ step 2120000 | 21446s | train_loss=0.0280
215
+ step 2130000 | 21538s | train_loss=0.0287
216
+ step 2140000 | 21633s | train_loss=0.0266
217
+ step 2150000 | 21727s | train_loss=0.0323
218
+ step 2160000 | 21821s | train_loss=0.0279
219
+ step 2170000 | 21911s | train_loss=0.0276
220
+ step 2180000 | 22005s | train_loss=0.0284
221
+ step 2190000 | 22099s | train_loss=0.0287
222
+ step 2200000 | 22267s | train_loss=0.0286 | fid=67.8138
223
+ step 2210000 | 22361s | train_loss=0.0292
224
+ step 2220000 | 22453s | train_loss=0.0299
225
+ step 2230000 | 22548s | train_loss=0.0303
226
+ step 2240000 | 22643s | train_loss=0.0278
227
+ step 2250000 | 22737s | train_loss=0.0275
228
+ step 2260000 | 22827s | train_loss=0.0282
229
+ step 2270000 | 22921s | train_loss=0.0303
230
+ step 2280000 | 23016s | train_loss=0.0275
231
+ step 2290000 | 23110s | train_loss=0.0265
232
+ step 2300000 | 23277s | train_loss=0.0267 | fid=71.4255
233
+ step 2310000 | 23372s | train_loss=0.0274
234
+ step 2320000 | 23464s | train_loss=0.0274
235
+ step 2330000 | 23558s | train_loss=0.0268
236
+ step 2340000 | 23653s | train_loss=0.0263
237
+ step 2350000 | 23745s | train_loss=0.0291
238
+ step 2360000 | 23837s | train_loss=0.0266
239
+ step 2370000 | 23931s | train_loss=0.0276
240
+ step 2380000 | 24026s | train_loss=0.0280
241
+ step 2390000 | 24120s | train_loss=0.0265
242
+ step 2400000 | 24287s | train_loss=0.0266 | fid=62.5476
243
+ step 2410000 | 24380s | train_loss=0.0268
244
+ step 2420000 | 24475s | train_loss=0.0251
245
+ step 2430000 | 24569s | train_loss=0.0292
246
+ step 2440000 | 24663s | train_loss=0.0251
247
+ step 2450000 | 24754s | train_loss=0.0272
248
+ step 2460000 | 24848s | train_loss=0.0279
249
+ step 2470000 | 24942s | train_loss=0.0273
250
+ step 2480000 | 25037s | train_loss=0.0272
251
+ step 2490000 | 25131s | train_loss=0.0278
252
+ step 2500000 | 25297s | train_loss=0.0245 | fid=66.9150
253
+ step 2510000 | 25390s | train_loss=0.0259
254
+ step 2520000 | 25485s | train_loss=0.0265
255
+ step 2530000 | 25579s | train_loss=0.0315
256
+ step 2540000 | 25669s | train_loss=0.0267
257
+ step 2550000 | 25764s | train_loss=0.0253
258
+ step 2560000 | 25858s | train_loss=0.0240
259
+ step 2570000 | 25952s | train_loss=0.0277
260
+ step 2580000 | 26047s | train_loss=0.0286
261
+ step 2590000 | 26141s | train_loss=0.0284
262
+ step 2600000 | 26308s | train_loss=0.0271 | fid=64.8537
263
+ step 2610000 | 26401s | train_loss=0.0262
264
+ step 2620000 | 26496s | train_loss=0.0294
265
+ step 2630000 | 26588s | train_loss=0.0247
266
+ step 2640000 | 26680s | train_loss=0.0248
267
+ step 2650000 | 26775s | train_loss=0.0242
268
+ step 2660000 | 26869s | train_loss=0.0262
269
+ step 2670000 | 26963s | train_loss=0.0242
270
+ step 2680000 | 27058s | train_loss=0.0262
271
+ step 2690000 | 27152s | train_loss=0.0282
272
+ step 2700000 | 27318s | train_loss=0.0256 | fid=67.1950
273
+ step 2710000 | 27412s | train_loss=0.0243
274
+ step 2720000 | 27505s | train_loss=0.0276
275
+ step 2730000 | 27597s | train_loss=0.0230
276
+ step 2740000 | 27691s | train_loss=0.0244
277
+ step 2750000 | 27785s | train_loss=0.0257
278
+ step 2760000 | 27880s | train_loss=0.0261
279
+ step 2770000 | 27975s | train_loss=0.0238
280
+ step 2780000 | 28069s | train_loss=0.0251
281
+ step 2790000 | 28163s | train_loss=0.0249
282
+ step 2800000 | 28329s | train_loss=0.0285 | fid=57.9465
283
+ step 2810000 | 28423s | train_loss=0.0254
284
+ step 2820000 | 28513s | train_loss=0.0258
285
+ step 2830000 | 28608s | train_loss=0.0254
286
+ step 2840000 | 28702s | train_loss=0.0243
287
+ step 2850000 | 28797s | train_loss=0.0248
288
+ step 2860000 | 28891s | train_loss=0.0245
289
+ step 2870000 | 28986s | train_loss=0.0235
290
+ step 2880000 | 29080s | train_loss=0.0255
291
+ step 2890000 | 29174s | train_loss=0.0228
292
+ step 2900000 | 29339s | train_loss=0.0244 | fid=59.6216
293
+ step 2910000 | 29432s | train_loss=0.0269
294
+ step 2920000 | 29523s | train_loss=0.0238
295
+ step 2930000 | 29618s | train_loss=0.0292
296
+ step 2940000 | 29712s | train_loss=0.0227
297
+ step 2950000 | 29807s | train_loss=0.0239
298
+ step 2960000 | 29901s | train_loss=0.0226
299
+ step 2970000 | 29996s | train_loss=0.0254
300
+ step 2980000 | 30090s | train_loss=0.0238
301
+ step 2990000 | 30184s | train_loss=0.0239
302
+ step 3000000 | 30349s | train_loss=0.0235 | fid=55.4758
303
+ step 3010000 | 30441s | train_loss=0.0234
304
+ step 3020000 | 30535s | train_loss=0.0250
305
+ step 3030000 | 30630s | train_loss=0.0223
306
+ step 3040000 | 30724s | train_loss=0.0277
307
+ step 3050000 | 30819s | train_loss=0.0217
308
+ step 3060000 | 30913s | train_loss=0.0268
309
+ step 3070000 | 31007s | train_loss=0.0230
310
+ step 3080000 | 31102s | train_loss=0.0241
311
+ step 3090000 | 31194s | train_loss=0.0229
312
+ step 3100000 | 31357s | train_loss=0.0263 | fid=58.0714
313
+ step 3110000 | 31452s | train_loss=0.0237
314
+ step 3120000 | 31546s | train_loss=0.0220
315
+ step 3130000 | 31641s | train_loss=0.0240
316
+ step 3140000 | 31735s | train_loss=0.0249
317
+ step 3150000 | 31829s | train_loss=0.0237
318
+ step 3160000 | 31924s | train_loss=0.0234
319
+ step 3170000 | 32018s | train_loss=0.0261
320
+ step 3180000 | 32112s | train_loss=0.0228
321
+ step 3190000 | 32205s | train_loss=0.0229
322
+ step 3200000 | 32367s | train_loss=0.0258 | fid=53.9922
323
+ step 3210000 | 32462s | train_loss=0.0222
324
+ step 3220000 | 32557s | train_loss=0.0234
325
+ step 3230000 | 32650s | train_loss=0.0238
326
+ step 3240000 | 32745s | train_loss=0.0247
327
+ step 3250000 | 32840s | train_loss=0.0243
328
+ step 3260000 | 32934s | train_loss=0.0231
329
+ step 3270000 | 33028s | train_loss=0.0246
330
+ step 3280000 | 33121s | train_loss=0.0225
331
+ step 3290000 | 33213s | train_loss=0.0236
332
+ step 3300000 | 33376s | train_loss=0.0228 | fid=60.5897
333
+ step 3310000 | 33471s | train_loss=0.0219
334
+ step 3320000 | 33565s | train_loss=0.0229
335
+ step 3330000 | 33660s | train_loss=0.0226
336
+ step 3340000 | 33754s | train_loss=0.0201
337
+ step 3350000 | 33849s | train_loss=0.0240
338
+ step 3360000 | 33943s | train_loss=0.0254
339
+ step 3370000 | 34037s | train_loss=0.0240
340
+ step 3380000 | 34129s | train_loss=0.0238
341
+ step 3390000 | 34219s | train_loss=0.0210
342
+ step 3400000 | 34386s | train_loss=0.0252 | fid=57.8412
343
+ step 3410000 | 34481s | train_loss=0.0232
344
+ step 3420000 | 34575s | train_loss=0.0219
345
+ step 3430000 | 34670s | train_loss=0.0227
346
+ step 3440000 | 34764s | train_loss=0.0212
347
+ step 3450000 | 34858s | train_loss=0.0227
348
+ step 3460000 | 34952s | train_loss=0.0244
349
+ step 3470000 | 35047s | train_loss=0.0230
350
+ step 3480000 | 35137s | train_loss=0.0244
351
+ step 3490000 | 35229s | train_loss=0.0227
352
+ step 3500000 | 35396s | train_loss=0.0213 | fid=71.7081
353
+ step 3510000 | 35490s | train_loss=0.0241
354
+ step 3520000 | 35585s | train_loss=0.0247
355
+ step 3530000 | 35679s | train_loss=0.0238
356
+ step 3540000 | 35773s | train_loss=0.0230
357
+ step 3550000 | 35868s | train_loss=0.0219
358
+ step 3560000 | 35963s | train_loss=0.0236
359
+ step 3570000 | 36052s | train_loss=0.0234
360
+ step 3580000 | 36143s | train_loss=0.0266
361
+ step 3590000 | 36237s | train_loss=0.0244
362
+ step 3600000 | 36405s | train_loss=0.0210 | fid=52.3586
363
+ step 3610000 | 36500s | train_loss=0.0252
364
+ step 3620000 | 36594s | train_loss=0.0214
365
+ step 3630000 | 36688s | train_loss=0.0216
366
+ step 3640000 | 36782s | train_loss=0.0224
367
+ step 3650000 | 36876s | train_loss=0.0205
368
+ step 3660000 | 36971s | train_loss=0.0215
369
+ step 3670000 | 37058s | train_loss=0.0225
370
+ step 3680000 | 37152s | train_loss=0.0212
371
+ step 3690000 | 37247s | train_loss=0.0212
372
+ step 3700000 | 37415s | train_loss=0.0223 | fid=56.4406
373
+ step 3710000 | 37509s | train_loss=0.0209
374
+ step 3720000 | 37603s | train_loss=0.0233
375
+ step 3730000 | 37697s | train_loss=0.0214
376
+ step 3740000 | 37791s | train_loss=0.0202
377
+ step 3750000 | 37886s | train_loss=0.0243
378
+ step 3760000 | 37976s | train_loss=0.0214
379
+ step 3770000 | 38067s | train_loss=0.0224
380
+ step 3780000 | 38162s | train_loss=0.0230
381
+ step 3790000 | 38256s | train_loss=0.0242
382
+ step 3800000 | 38425s | train_loss=0.0215 | fid=47.7908
383
+ step 3810000 | 38518s | train_loss=0.0206
384
+ step 3820000 | 38612s | train_loss=0.0247
385
+ step 3830000 | 38706s | train_loss=0.0200
386
+ step 3840000 | 38801s | train_loss=0.0218
387
+ step 3850000 | 38892s | train_loss=0.0220
388
+ step 3860000 | 38981s | train_loss=0.0209
389
+ step 3870000 | 39075s | train_loss=0.0195
390
+ step 3880000 | 39170s | train_loss=0.0207
391
+ step 3890000 | 39264s | train_loss=0.0195
392
+ step 3900000 | 39432s | train_loss=0.0207 | fid=51.2018
393
+ step 3910000 | 39524s | train_loss=0.0201
394
+ step 3920000 | 39619s | train_loss=0.0204
395
+ step 3930000 | 39713s | train_loss=0.0198
396
+ step 3940000 | 39808s | train_loss=0.0193
397
+ step 3950000 | 39898s | train_loss=0.0193
398
+ step 3960000 | 39990s | train_loss=0.0203
399
+ step 3970000 | 40084s | train_loss=0.0195
400
+ step 3980000 | 40179s | train_loss=0.0194
401
+ step 3990000 | 40273s | train_loss=0.0219
402
+ step 4000000 | 40442s | train_loss=0.0221 | fid=49.8883
403
+ step 4010000 | 40533s | train_loss=0.0246
404
+ step 4020000 | 40627s | train_loss=0.0218
405
+ step 4030000 | 40721s | train_loss=0.0200
406
+ step 4040000 | 40813s | train_loss=0.0194
407
+ step 4050000 | 40904s | train_loss=0.0213
408
+ step 4060000 | 40998s | train_loss=0.0230
409
+ step 4070000 | 41092s | train_loss=0.0210
410
+ step 4080000 | 41186s | train_loss=0.0194
411
+ step 4090000 | 41281s | train_loss=0.0200
412
+ step 4100000 | 41450s | train_loss=0.0200 | fid=56.0015
413
+ step 4110000 | 41541s | train_loss=0.0217
414
+ step 4120000 | 41635s | train_loss=0.0201
415
+ step 4130000 | 41727s | train_loss=0.0206
416
+ step 4140000 | 41820s | train_loss=0.0232
417
+ step 4150000 | 41912s | train_loss=0.0199
418
+ step 4160000 | 42007s | train_loss=0.0211
419
+ step 4170000 | 42101s | train_loss=0.0200
420
+ step 4180000 | 42196s | train_loss=0.0216
421
+ step 4190000 | 42290s | train_loss=0.0196
422
+ step 4200000 | 42460s | train_loss=0.0205 | fid=49.2510
423
+ step 4210000 | 42551s | train_loss=0.0190
424
+ step 4220000 | 42645s | train_loss=0.0202
425
+ step 4230000 | 42735s | train_loss=0.0225
426
+ step 4240000 | 42827s | train_loss=0.0195
427
+ step 4250000 | 42922s | train_loss=0.0190
428
+ step 4260000 | 43016s | train_loss=0.0199
429
+ step 4270000 | 43110s | train_loss=0.0210
430
+ step 4280000 | 43205s | train_loss=0.0206
431
+ step 4290000 | 43299s | train_loss=0.0198
432
+ step 4300000 | 43468s | train_loss=0.0209 | fid=50.0264
433
+ step 4310000 | 43560s | train_loss=0.0195
434
+ step 4320000 | 43651s | train_loss=0.0212
435
+ step 4330000 | 43744s | train_loss=0.0197
436
+ step 4340000 | 43836s | train_loss=0.0188
437
+ step 4350000 | 43930s | train_loss=0.0204
438
+ step 4360000 | 44025s | train_loss=0.0206
439
+ step 4370000 | 44119s | train_loss=0.0200
440
+ step 4380000 | 44214s | train_loss=0.0200
441
+ step 4390000 | 44308s | train_loss=0.0184
442
+ step 4400000 | 44477s | train_loss=0.0191 | fid=49.2860
443
+ step 4410000 | 44568s | train_loss=0.0199
444
+ step 4420000 | 44661s | train_loss=0.0219
445
+ step 4430000 | 44755s | train_loss=0.0192
446
+ step 4440000 | 44848s | train_loss=0.0186
447
+ step 4450000 | 44942s | train_loss=0.0214
448
+ step 4460000 | 45036s | train_loss=0.0186
449
+ step 4470000 | 45131s | train_loss=0.0203
450
+ step 4480000 | 45225s | train_loss=0.0204
451
+ step 4490000 | 45320s | train_loss=0.0189
452
+ step 4500000 | 45490s | train_loss=0.0201 | fid=52.4096
453
+ step 4510000 | 45577s | train_loss=0.0209
454
+ step 4520000 | 45671s | train_loss=0.0194
455
+ step 4530000 | 45764s | train_loss=0.0199
456
+ step 4540000 | 45858s | train_loss=0.0207
457
+ step 4550000 | 45952s | train_loss=0.0192
458
+ step 4560000 | 46047s | train_loss=0.0185
459
+ step 4570000 | 46141s | train_loss=0.0169
460
+ step 4580000 | 46236s | train_loss=0.0193
461
+ step 4590000 | 46330s | train_loss=0.0199
462
+ step 4600000 | 46500s | train_loss=0.0191 | fid=51.1784
463
+ step 4610000 | 46591s | train_loss=0.0194
464
+ step 4620000 | 46686s | train_loss=0.0199
465
+ step 4630000 | 46779s | train_loss=0.0219
466
+ step 4640000 | 46873s | train_loss=0.0219
467
+ step 4650000 | 46968s | train_loss=0.0203
468
+ step 4660000 | 47062s | train_loss=0.0205
469
+ step 4670000 | 47157s | train_loss=0.0205
470
+ step 4680000 | 47251s | train_loss=0.0184
471
+ step 4690000 | 47346s | train_loss=0.0190
472
+ step 4700000 | 47483s | train_loss=0.0197 | fid=51.5955
473
+ step 4710000 | 47537s | train_loss=0.0180
474
+ step 4720000 | 47591s | train_loss=0.0189
475
+ step 4730000 | 47649s | train_loss=0.0187
476
+ step 4740000 | 47707s | train_loss=0.0186
477
+ step 4750000 | 47764s | train_loss=0.0189
478
+ step 4760000 | 47822s | train_loss=0.0182
479
+ step 4770000 | 47880s | train_loss=0.0176
480
+ step 4780000 | 47938s | train_loss=0.0193
481
+ step 4790000 | 47996s | train_loss=0.0201
482
+ step 4800000 | 48105s | train_loss=0.0185 | fid=50.1269
483
+ step 4810000 | 48158s | train_loss=0.0173
484
+ step 4820000 | 48199s | train_loss=0.0195
485
+ step 4830000 | 48239s | train_loss=0.0175
486
+ step 4840000 | 48279s | train_loss=0.0176
487
+ step 4850000 | 48318s | train_loss=0.0177
488
+ step 4860000 | 48357s | train_loss=0.0194
489
+ step 4870000 | 48397s | train_loss=0.0184
490
+ step 4880000 | 48435s | train_loss=0.0166
491
+ step 4890000 | 48475s | train_loss=0.0183
492
+ step 4900000 | 48552s | train_loss=0.0185 | fid=52.4162
493
+ step 4910000 | 48591s | train_loss=0.0182
494
+ step 4920000 | 48630s | train_loss=0.0211
495
+ step 4930000 | 48670s | train_loss=0.0182
496
+ step 4940000 | 48709s | train_loss=0.0216
497
+ step 4950000 | 48749s | train_loss=0.0195
498
+ step 4960000 | 48788s | train_loss=0.0201
499
+ step 4970000 | 48828s | train_loss=0.0184
500
+ step 4980000 | 48867s | train_loss=0.0188
501
+ step 4990000 | 48907s | train_loss=0.0182
502
+ step 5000000 | 48985s | train_loss=0.0192 | fid=52.1886
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/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/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": 42,
25
+ "device": null
26
+ }
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/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/d20/runner.log ADDED
@@ -0,0 +1,502 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ FID real-side stats cached over 1000 diverse train images
2
+ step 1 | 67s | train_loss=0.9905 | fid=225.9377
3
+ step 10000 | 151s | train_loss=0.8705
4
+ step 20000 | 236s | train_loss=0.8647
5
+ step 30000 | 320s | train_loss=0.8600
6
+ step 40000 | 404s | train_loss=0.8544
7
+ step 50000 | 488s | train_loss=0.8485
8
+ step 60000 | 572s | train_loss=0.8418
9
+ step 70000 | 656s | train_loss=0.8304
10
+ step 80000 | 740s | train_loss=0.8097
11
+ step 90000 | 824s | train_loss=0.7759
12
+ step 100000 | 974s | train_loss=0.7306 | fid=72.1051
13
+ step 110000 | 1050s | train_loss=0.6690
14
+ step 120000 | 1134s | train_loss=0.5884
15
+ step 130000 | 1218s | train_loss=0.5027
16
+ step 140000 | 1302s | train_loss=0.4233
17
+ step 150000 | 1386s | train_loss=0.3553
18
+ step 160000 | 1470s | train_loss=0.3036
19
+ step 170000 | 1554s | train_loss=0.2686
20
+ step 180000 | 1637s | train_loss=0.2449
21
+ step 190000 | 1721s | train_loss=0.2231
22
+ step 200000 | 1871s | train_loss=0.2118 | fid=71.1903
23
+ step 210000 | 1948s | train_loss=0.1937
24
+ step 220000 | 2030s | train_loss=0.1848
25
+ step 230000 | 2114s | train_loss=0.1746
26
+ step 240000 | 2198s | train_loss=0.1694
27
+ step 250000 | 2282s | train_loss=0.1602
28
+ step 260000 | 2366s | train_loss=0.1546
29
+ step 270000 | 2450s | train_loss=0.1485
30
+ step 280000 | 2534s | train_loss=0.1436
31
+ step 290000 | 2618s | train_loss=0.1364
32
+ step 300000 | 2768s | train_loss=0.1323 | fid=69.8701
33
+ step 310000 | 2847s | train_loss=0.1282
34
+ step 320000 | 2928s | train_loss=0.1239
35
+ step 330000 | 3011s | train_loss=0.1211
36
+ step 340000 | 3095s | train_loss=0.1173
37
+ step 350000 | 3179s | train_loss=0.1143
38
+ step 360000 | 3263s | train_loss=0.1109
39
+ step 370000 | 3347s | train_loss=0.1077
40
+ step 380000 | 3431s | train_loss=0.1057
41
+ step 390000 | 3515s | train_loss=0.1030
42
+ step 400000 | 3666s | train_loss=0.1013 | fid=70.6623
43
+ step 410000 | 3745s | train_loss=0.0986
44
+ step 420000 | 3826s | train_loss=0.0976
45
+ step 430000 | 3910s | train_loss=0.0947
46
+ step 440000 | 3993s | train_loss=0.0919
47
+ step 450000 | 4077s | train_loss=0.0909
48
+ step 460000 | 4161s | train_loss=0.0884
49
+ step 470000 | 4245s | train_loss=0.0865
50
+ step 480000 | 4329s | train_loss=0.0854
51
+ step 490000 | 4413s | train_loss=0.0841
52
+ step 500000 | 4562s | train_loss=0.0816 | fid=70.3333
53
+ step 510000 | 4642s | train_loss=0.0806
54
+ step 520000 | 4726s | train_loss=0.0799
55
+ step 530000 | 4807s | train_loss=0.0779
56
+ step 540000 | 4892s | train_loss=0.0767
57
+ step 550000 | 4974s | train_loss=0.0756
58
+ step 560000 | 5058s | train_loss=0.0741
59
+ step 570000 | 5142s | train_loss=0.0731
60
+ step 580000 | 5226s | train_loss=0.0728
61
+ step 590000 | 5310s | train_loss=0.0711
62
+ step 600000 | 5462s | train_loss=0.0695 | fid=70.1528
63
+ step 610000 | 5542s | train_loss=0.0698
64
+ step 620000 | 5626s | train_loss=0.0689
65
+ step 630000 | 5707s | train_loss=0.0671
66
+ step 640000 | 5791s | train_loss=0.0676
67
+ step 650000 | 5876s | train_loss=0.0654
68
+ step 660000 | 5958s | train_loss=0.0653
69
+ step 670000 | 6042s | train_loss=0.0645
70
+ step 680000 | 6126s | train_loss=0.0630
71
+ step 690000 | 6210s | train_loss=0.0627
72
+ step 700000 | 6362s | train_loss=0.0612 | fid=69.1181
73
+ step 710000 | 6442s | train_loss=0.0608
74
+ step 720000 | 6526s | train_loss=0.0607
75
+ step 730000 | 6610s | train_loss=0.0598
76
+ step 740000 | 6691s | train_loss=0.0601
77
+ step 750000 | 6775s | train_loss=0.0588
78
+ step 760000 | 6859s | train_loss=0.0584
79
+ step 770000 | 6942s | train_loss=0.0577
80
+ step 780000 | 7026s | train_loss=0.0564
81
+ step 790000 | 7110s | train_loss=0.0563
82
+ step 800000 | 7262s | train_loss=0.0559 | fid=68.8551
83
+ step 810000 | 7344s | train_loss=0.0557
84
+ step 820000 | 7426s | train_loss=0.0545
85
+ step 830000 | 7510s | train_loss=0.0535
86
+ step 840000 | 7591s | train_loss=0.0532
87
+ step 850000 | 7675s | train_loss=0.0528
88
+ step 860000 | 7760s | train_loss=0.0522
89
+ step 870000 | 7844s | train_loss=0.0517
90
+ step 880000 | 7926s | train_loss=0.0514
91
+ step 890000 | 8010s | train_loss=0.0513
92
+ step 900000 | 8163s | train_loss=0.0504 | fid=67.2350
93
+ step 910000 | 8245s | train_loss=0.0509
94
+ step 920000 | 8326s | train_loss=0.0495
95
+ step 930000 | 8410s | train_loss=0.0493
96
+ step 940000 | 8493s | train_loss=0.0488
97
+ step 950000 | 8575s | train_loss=0.0489
98
+ step 960000 | 8659s | train_loss=0.0478
99
+ step 970000 | 8743s | train_loss=0.0478
100
+ step 980000 | 8828s | train_loss=0.0474
101
+ step 990000 | 8910s | train_loss=0.0472
102
+ step 1000000 | 9063s | train_loss=0.0465 | fid=69.0239
103
+ step 1010000 | 9146s | train_loss=0.0463
104
+ step 1020000 | 9227s | train_loss=0.0457
105
+ step 1030000 | 9311s | train_loss=0.0461
106
+ step 1040000 | 9395s | train_loss=0.0451
107
+ step 1050000 | 9476s | train_loss=0.0449
108
+ step 1060000 | 9560s | train_loss=0.0449
109
+ step 1070000 | 9644s | train_loss=0.0444
110
+ step 1080000 | 9728s | train_loss=0.0439
111
+ step 1090000 | 9811s | train_loss=0.0435
112
+ step 1100000 | 9963s | train_loss=0.0431 | fid=68.2317
113
+ step 1110000 | 10046s | train_loss=0.0433
114
+ step 1120000 | 10127s | train_loss=0.0430
115
+ step 1130000 | 10211s | train_loss=0.0431
116
+ step 1140000 | 10295s | train_loss=0.0419
117
+ step 1150000 | 10376s | train_loss=0.0415
118
+ step 1160000 | 10460s | train_loss=0.0419
119
+ step 1170000 | 10544s | train_loss=0.0411
120
+ step 1180000 | 10628s | train_loss=0.0410
121
+ step 1190000 | 10712s | train_loss=0.0413
122
+ step 1200000 | 10864s | train_loss=0.0405 | fid=67.5463
123
+ step 1210000 | 10946s | train_loss=0.0399
124
+ step 1220000 | 11027s | train_loss=0.0399
125
+ step 1230000 | 11111s | train_loss=0.0401
126
+ step 1240000 | 11195s | train_loss=0.0398
127
+ step 1250000 | 11279s | train_loss=0.0399
128
+ step 1260000 | 11361s | train_loss=0.0388
129
+ step 1270000 | 11445s | train_loss=0.0389
130
+ step 1280000 | 11529s | train_loss=0.0386
131
+ step 1290000 | 11613s | train_loss=0.0392
132
+ step 1300000 | 11764s | train_loss=0.0385 | fid=68.4357
133
+ step 1310000 | 11847s | train_loss=0.0386
134
+ step 1320000 | 11928s | train_loss=0.0377
135
+ step 1330000 | 12012s | train_loss=0.0376
136
+ step 1340000 | 12096s | train_loss=0.0373
137
+ step 1350000 | 12180s | train_loss=0.0373
138
+ step 1360000 | 12262s | train_loss=0.0376
139
+ step 1370000 | 12346s | train_loss=0.0369
140
+ step 1380000 | 12430s | train_loss=0.0365
141
+ step 1390000 | 12514s | train_loss=0.0361
142
+ step 1400000 | 12665s | train_loss=0.0362 | fid=67.1393
143
+ step 1410000 | 12746s | train_loss=0.0364
144
+ step 1420000 | 12828s | train_loss=0.0355
145
+ step 1430000 | 12912s | train_loss=0.0362
146
+ step 1440000 | 12996s | train_loss=0.0352
147
+ step 1450000 | 13080s | train_loss=0.0361
148
+ step 1460000 | 13164s | train_loss=0.0351
149
+ step 1470000 | 13245s | train_loss=0.0350
150
+ step 1480000 | 13329s | train_loss=0.0347
151
+ step 1490000 | 13413s | train_loss=0.0349
152
+ step 1500000 | 13565s | train_loss=0.0348 | fid=67.3283
153
+ step 1510000 | 13647s | train_loss=0.0347
154
+ step 1520000 | 13727s | train_loss=0.0343
155
+ step 1530000 | 13811s | train_loss=0.0338
156
+ step 1540000 | 13895s | train_loss=0.0339
157
+ step 1550000 | 13979s | train_loss=0.0339
158
+ step 1560000 | 14063s | train_loss=0.0339
159
+ step 1570000 | 14144s | train_loss=0.0333
160
+ step 1580000 | 14228s | train_loss=0.0333
161
+ step 1590000 | 14312s | train_loss=0.0331
162
+ step 1600000 | 14507s | train_loss=0.0329 | fid=67.4677
163
+ step 1610000 | 14591s | train_loss=0.0328
164
+ step 1620000 | 14659s | train_loss=0.0329
165
+ step 1630000 | 14741s | train_loss=0.0330
166
+ step 1640000 | 14826s | train_loss=0.0324
167
+ step 1650000 | 14910s | train_loss=0.0326
168
+ step 1660000 | 14994s | train_loss=0.0319
169
+ step 1670000 | 15069s | train_loss=0.0318
170
+ step 1680000 | 15153s | train_loss=0.0325
171
+ step 1690000 | 15237s | train_loss=0.0318
172
+ step 1700000 | 15390s | train_loss=0.0314 | fid=68.2917
173
+ step 1710000 | 15474s | train_loss=0.0316
174
+ step 1720000 | 15558s | train_loss=0.0313
175
+ step 1730000 | 15637s | train_loss=0.0312
176
+ step 1740000 | 15721s | train_loss=0.0309
177
+ step 1750000 | 15805s | train_loss=0.0312
178
+ step 1760000 | 15889s | train_loss=0.0309
179
+ step 1770000 | 15973s | train_loss=0.0302
180
+ step 1780000 | 16054s | train_loss=0.0307
181
+ step 1790000 | 16138s | train_loss=0.0305
182
+ step 1800000 | 16288s | train_loss=0.0303 | fid=68.5644
183
+ step 1810000 | 16372s | train_loss=0.0300
184
+ step 1820000 | 16456s | train_loss=0.0302
185
+ step 1830000 | 16537s | train_loss=0.0297
186
+ step 1840000 | 16619s | train_loss=0.0298
187
+ step 1850000 | 16703s | train_loss=0.0298
188
+ step 1860000 | 16788s | train_loss=0.0293
189
+ step 1870000 | 16872s | train_loss=0.0298
190
+ step 1880000 | 16952s | train_loss=0.0293
191
+ step 1890000 | 17036s | train_loss=0.0298
192
+ step 1900000 | 17189s | train_loss=0.0295 | fid=68.0634
193
+ step 1910000 | 17273s | train_loss=0.0289
194
+ step 1920000 | 17357s | train_loss=0.0293
195
+ step 1930000 | 17438s | train_loss=0.0293
196
+ step 1940000 | 17522s | train_loss=0.0288
197
+ step 1950000 | 17604s | train_loss=0.0286
198
+ step 1960000 | 17688s | train_loss=0.0286
199
+ step 1970000 | 17772s | train_loss=0.0286
200
+ step 1980000 | 17857s | train_loss=0.0282
201
+ step 1990000 | 17938s | train_loss=0.0281
202
+ step 2000000 | 18090s | train_loss=0.0281 | fid=67.1998
203
+ step 2010000 | 18174s | train_loss=0.0280
204
+ step 2020000 | 18258s | train_loss=0.0280
205
+ step 2030000 | 18339s | train_loss=0.0278
206
+ step 2040000 | 18423s | train_loss=0.0277
207
+ step 2050000 | 18507s | train_loss=0.0274
208
+ step 2060000 | 18590s | train_loss=0.0278
209
+ step 2070000 | 18673s | train_loss=0.0277
210
+ step 2080000 | 18758s | train_loss=0.0275
211
+ step 2090000 | 18839s | train_loss=0.0273
212
+ step 2100000 | 18988s | train_loss=0.0273 | fid=68.1058
213
+ step 2110000 | 19072s | train_loss=0.0272
214
+ step 2120000 | 19156s | train_loss=0.0273
215
+ step 2130000 | 19238s | train_loss=0.0269
216
+ step 2140000 | 19322s | train_loss=0.0267
217
+ step 2150000 | 19406s | train_loss=0.0269
218
+ step 2160000 | 19490s | train_loss=0.0267
219
+ step 2170000 | 19573s | train_loss=0.0264
220
+ step 2180000 | 19657s | train_loss=0.0262
221
+ step 2190000 | 19740s | train_loss=0.0268
222
+ step 2200000 | 19913s | train_loss=0.0263 | fid=66.9639
223
+ step 2210000 | 19997s | train_loss=0.0265
224
+ step 2220000 | 20081s | train_loss=0.0262
225
+ step 2230000 | 20153s | train_loss=0.0258
226
+ step 2240000 | 20238s | train_loss=0.0258
227
+ step 2250000 | 20322s | train_loss=0.0263
228
+ step 2260000 | 20406s | train_loss=0.0258
229
+ step 2270000 | 20488s | train_loss=0.0259
230
+ step 2280000 | 20573s | train_loss=0.0260
231
+ step 2290000 | 20657s | train_loss=0.0255
232
+ step 2300000 | 20823s | train_loss=0.0255 | fid=68.1763
233
+ step 2310000 | 20907s | train_loss=0.0256
234
+ step 2320000 | 20992s | train_loss=0.0250
235
+ step 2330000 | 21066s | train_loss=0.0255
236
+ step 2340000 | 21147s | train_loss=0.0254
237
+ step 2350000 | 21231s | train_loss=0.0251
238
+ step 2360000 | 21315s | train_loss=0.0251
239
+ step 2370000 | 21398s | train_loss=0.0251
240
+ step 2380000 | 21481s | train_loss=0.0249
241
+ step 2390000 | 21566s | train_loss=0.0252
242
+ step 2400000 | 21751s | train_loss=0.0249 | fid=67.3247
243
+ step 2410000 | 21835s | train_loss=0.0242
244
+ step 2420000 | 21919s | train_loss=0.0245
245
+ step 2430000 | 21995s | train_loss=0.0247
246
+ step 2440000 | 22071s | train_loss=0.0246
247
+ step 2450000 | 22155s | train_loss=0.0243
248
+ step 2460000 | 22239s | train_loss=0.0243
249
+ step 2470000 | 22323s | train_loss=0.0243
250
+ step 2480000 | 22405s | train_loss=0.0240
251
+ step 2490000 | 22489s | train_loss=0.0239
252
+ step 2500000 | 22646s | train_loss=0.0237 | fid=67.4470
253
+ step 2510000 | 22730s | train_loss=0.0242
254
+ step 2520000 | 22814s | train_loss=0.0237
255
+ step 2530000 | 22897s | train_loss=0.0239
256
+ step 2540000 | 22968s | train_loss=0.0237
257
+ step 2550000 | 23052s | train_loss=0.0237
258
+ step 2560000 | 23136s | train_loss=0.0237
259
+ step 2570000 | 23220s | train_loss=0.0237
260
+ step 2580000 | 23302s | train_loss=0.0232
261
+ step 2590000 | 23384s | train_loss=0.0232
262
+ step 2600000 | 23556s | train_loss=0.0231 | fid=67.3083
263
+ step 2610000 | 23640s | train_loss=0.0229
264
+ step 2620000 | 23724s | train_loss=0.0230
265
+ step 2630000 | 23808s | train_loss=0.0230
266
+ step 2640000 | 23884s | train_loss=0.0233
267
+ step 2650000 | 23961s | train_loss=0.0228
268
+ step 2660000 | 24045s | train_loss=0.0227
269
+ step 2670000 | 24129s | train_loss=0.0233
270
+ step 2680000 | 24214s | train_loss=0.0228
271
+ step 2690000 | 24294s | train_loss=0.0227
272
+ step 2700000 | 24476s | train_loss=0.0226 | fid=68.3175
273
+ step 2710000 | 24550s | train_loss=0.0223
274
+ step 2720000 | 24634s | train_loss=0.0228
275
+ step 2730000 | 24717s | train_loss=0.0223
276
+ step 2740000 | 24799s | train_loss=0.0225
277
+ step 2750000 | 24870s | train_loss=0.0225
278
+ step 2760000 | 24955s | train_loss=0.0224
279
+ step 2770000 | 25039s | train_loss=0.0219
280
+ step 2780000 | 25121s | train_loss=0.0221
281
+ step 2790000 | 25203s | train_loss=0.0223
282
+ step 2800000 | 25383s | train_loss=0.0216 | fid=67.3541
283
+ step 2810000 | 25459s | train_loss=0.0219
284
+ step 2820000 | 25541s | train_loss=0.0219
285
+ step 2830000 | 25625s | train_loss=0.0215
286
+ step 2840000 | 25708s | train_loss=0.0218
287
+ step 2850000 | 25780s | train_loss=0.0220
288
+ step 2860000 | 25861s | train_loss=0.0218
289
+ step 2870000 | 25945s | train_loss=0.0213
290
+ step 2880000 | 26027s | train_loss=0.0216
291
+ step 2890000 | 26109s | train_loss=0.0217
292
+ step 2900000 | 26308s | train_loss=0.0214 | fid=67.0265
293
+ step 2910000 | 26387s | train_loss=0.0213
294
+ step 2920000 | 26465s | train_loss=0.0210
295
+ step 2930000 | 26549s | train_loss=0.0211
296
+ step 2940000 | 26633s | train_loss=0.0212
297
+ step 2950000 | 26708s | train_loss=0.0212
298
+ step 2960000 | 26788s | train_loss=0.0210
299
+ step 2970000 | 26870s | train_loss=0.0211
300
+ step 2980000 | 26954s | train_loss=0.0210
301
+ step 2990000 | 27039s | train_loss=0.0208
302
+ step 3000000 | 27237s | train_loss=0.0208 | fid=68.2062
303
+ step 3010000 | 27320s | train_loss=0.0210
304
+ step 3020000 | 27394s | train_loss=0.0204
305
+ step 3030000 | 27478s | train_loss=0.0205
306
+ step 3040000 | 27562s | train_loss=0.0207
307
+ step 3050000 | 27637s | train_loss=0.0206
308
+ step 3060000 | 27721s | train_loss=0.0205
309
+ step 3070000 | 27803s | train_loss=0.0208
310
+ step 3080000 | 27887s | train_loss=0.0205
311
+ step 3090000 | 27972s | train_loss=0.0201
312
+ step 3100000 | 28153s | train_loss=0.0205 | fid=67.8910
313
+ step 3110000 | 28237s | train_loss=0.0201
314
+ step 3120000 | 28311s | train_loss=0.0199
315
+ step 3130000 | 28390s | train_loss=0.0200
316
+ step 3140000 | 28474s | train_loss=0.0199
317
+ step 3150000 | 28550s | train_loss=0.0201
318
+ step 3160000 | 28632s | train_loss=0.0199
319
+ step 3170000 | 28716s | train_loss=0.0200
320
+ step 3180000 | 28800s | train_loss=0.0197
321
+ step 3190000 | 28884s | train_loss=0.0201
322
+ step 3200000 | 29060s | train_loss=0.0197 | fid=68.0076
323
+ step 3210000 | 29144s | train_loss=0.0197
324
+ step 3220000 | 29228s | train_loss=0.0199
325
+ step 3230000 | 29301s | train_loss=0.0194
326
+ step 3240000 | 29385s | train_loss=0.0195
327
+ step 3250000 | 29461s | train_loss=0.0198
328
+ step 3260000 | 29537s | train_loss=0.0195
329
+ step 3270000 | 29621s | train_loss=0.0196
330
+ step 3280000 | 29705s | train_loss=0.0194
331
+ step 3290000 | 29789s | train_loss=0.0196
332
+ step 3300000 | 29938s | train_loss=0.0195 | fid=68.5422
333
+ step 3310000 | 30022s | train_loss=0.0193
334
+ step 3320000 | 30106s | train_loss=0.0197
335
+ step 3330000 | 30190s | train_loss=0.0193
336
+ step 3340000 | 30268s | train_loss=0.0192
337
+ step 3350000 | 30352s | train_loss=0.0190
338
+ step 3360000 | 30421s | train_loss=0.0193
339
+ step 3370000 | 30502s | train_loss=0.0189
340
+ step 3380000 | 30586s | train_loss=0.0189
341
+ step 3390000 | 30671s | train_loss=0.0192
342
+ step 3400000 | 30822s | train_loss=0.0188 | fid=67.2079
343
+ step 3410000 | 30904s | train_loss=0.0190
344
+ step 3420000 | 30988s | train_loss=0.0189
345
+ step 3430000 | 31072s | train_loss=0.0188
346
+ step 3440000 | 31155s | train_loss=0.0189
347
+ step 3450000 | 31237s | train_loss=0.0189
348
+ step 3460000 | 31319s | train_loss=0.0187
349
+ step 3470000 | 31400s | train_loss=0.0186
350
+ step 3480000 | 31484s | train_loss=0.0184
351
+ step 3490000 | 31568s | train_loss=0.0183
352
+ step 3500000 | 31719s | train_loss=0.0185 | fid=66.0084
353
+ step 3510000 | 31803s | train_loss=0.0181
354
+ step 3520000 | 31885s | train_loss=0.0184
355
+ step 3530000 | 31970s | train_loss=0.0183
356
+ step 3540000 | 32054s | train_loss=0.0180
357
+ step 3550000 | 32135s | train_loss=0.0182
358
+ step 3560000 | 32217s | train_loss=0.0181
359
+ step 3570000 | 32298s | train_loss=0.0179
360
+ step 3580000 | 32382s | train_loss=0.0181
361
+ step 3590000 | 32465s | train_loss=0.0181
362
+ step 3600000 | 32660s | train_loss=0.0180 | fid=67.1706
363
+ step 3610000 | 32744s | train_loss=0.0180
364
+ step 3620000 | 32827s | train_loss=0.0181
365
+ step 3630000 | 32911s | train_loss=0.0181
366
+ step 3640000 | 32995s | train_loss=0.0179
367
+ step 3650000 | 33064s | train_loss=0.0180
368
+ step 3660000 | 33144s | train_loss=0.0178
369
+ step 3670000 | 33220s | train_loss=0.0178
370
+ step 3680000 | 33304s | train_loss=0.0179
371
+ step 3690000 | 33388s | train_loss=0.0177
372
+ step 3700000 | 33581s | train_loss=0.0173 | fid=67.1518
373
+ step 3710000 | 33665s | train_loss=0.0179
374
+ step 3720000 | 33748s | train_loss=0.0176
375
+ step 3730000 | 33832s | train_loss=0.0177
376
+ step 3740000 | 33916s | train_loss=0.0175
377
+ step 3750000 | 33994s | train_loss=0.0177
378
+ step 3760000 | 34067s | train_loss=0.0174
379
+ step 3770000 | 34141s | train_loss=0.0177
380
+ step 3780000 | 34226s | train_loss=0.0176
381
+ step 3790000 | 34310s | train_loss=0.0177
382
+ step 3800000 | 34488s | train_loss=0.0174 | fid=66.8800
383
+ step 3810000 | 34572s | train_loss=0.0173
384
+ step 3820000 | 34655s | train_loss=0.0175
385
+ step 3830000 | 34739s | train_loss=0.0171
386
+ step 3840000 | 34821s | train_loss=0.0171
387
+ step 3850000 | 34905s | train_loss=0.0171
388
+ step 3860000 | 34978s | train_loss=0.0171
389
+ step 3870000 | 35051s | train_loss=0.0171
390
+ step 3880000 | 35135s | train_loss=0.0170
391
+ step 3890000 | 35219s | train_loss=0.0171
392
+ step 3900000 | 35411s | train_loss=0.0171 | fid=68.3678
393
+ step 3910000 | 35495s | train_loss=0.0169
394
+ step 3920000 | 35579s | train_loss=0.0168
395
+ step 3930000 | 35662s | train_loss=0.0171
396
+ step 3940000 | 35744s | train_loss=0.0168
397
+ step 3950000 | 35828s | train_loss=0.0170
398
+ step 3960000 | 35907s | train_loss=0.0169
399
+ step 3970000 | 35972s | train_loss=0.0167
400
+ step 3980000 | 36056s | train_loss=0.0164
401
+ step 3990000 | 36140s | train_loss=0.0164
402
+ step 4000000 | 36324s | train_loss=0.0166 | fid=67.2159
403
+ step 4010000 | 36408s | train_loss=0.0169
404
+ step 4020000 | 36492s | train_loss=0.0167
405
+ step 4030000 | 36573s | train_loss=0.0163
406
+ step 4040000 | 36657s | train_loss=0.0166
407
+ step 4050000 | 36741s | train_loss=0.0165
408
+ step 4060000 | 36825s | train_loss=0.0162
409
+ step 4070000 | 36888s | train_loss=0.0163
410
+ step 4080000 | 36968s | train_loss=0.0164
411
+ step 4090000 | 37052s | train_loss=0.0161
412
+ step 4100000 | 37213s | train_loss=0.0163 | fid=66.8550
413
+ step 4110000 | 37297s | train_loss=0.0163
414
+ step 4120000 | 37381s | train_loss=0.0160
415
+ step 4130000 | 37464s | train_loss=0.0164
416
+ step 4140000 | 37546s | train_loss=0.0165
417
+ step 4150000 | 37630s | train_loss=0.0162
418
+ step 4160000 | 37714s | train_loss=0.0160
419
+ step 4170000 | 37798s | train_loss=0.0160
420
+ step 4180000 | 37875s | train_loss=0.0157
421
+ step 4190000 | 37959s | train_loss=0.0161
422
+ step 4200000 | 38115s | train_loss=0.0158 | fid=66.8536
423
+ step 4210000 | 38199s | train_loss=0.0163
424
+ step 4220000 | 38283s | train_loss=0.0158
425
+ step 4230000 | 38365s | train_loss=0.0158
426
+ step 4240000 | 38449s | train_loss=0.0159
427
+ step 4250000 | 38531s | train_loss=0.0159
428
+ step 4260000 | 38615s | train_loss=0.0159
429
+ step 4270000 | 38699s | train_loss=0.0160
430
+ step 4280000 | 38778s | train_loss=0.0160
431
+ step 4290000 | 38862s | train_loss=0.0156
432
+ step 4300000 | 39022s | train_loss=0.0158 | fid=67.9134
433
+ step 4310000 | 39106s | train_loss=0.0157
434
+ step 4320000 | 39190s | train_loss=0.0158
435
+ step 4330000 | 39272s | train_loss=0.0156
436
+ step 4340000 | 39356s | train_loss=0.0154
437
+ step 4350000 | 39439s | train_loss=0.0154
438
+ step 4360000 | 39523s | train_loss=0.0155
439
+ step 4370000 | 39607s | train_loss=0.0155
440
+ step 4380000 | 39685s | train_loss=0.0154
441
+ step 4390000 | 39769s | train_loss=0.0155
442
+ step 4400000 | 39922s | train_loss=0.0154 | fid=66.9424
443
+ step 4410000 | 40007s | train_loss=0.0153
444
+ step 4420000 | 40090s | train_loss=0.0152
445
+ step 4430000 | 40173s | train_loss=0.0154
446
+ step 4440000 | 40256s | train_loss=0.0153
447
+ step 4450000 | 40341s | train_loss=0.0153
448
+ step 4460000 | 40423s | train_loss=0.0154
449
+ step 4470000 | 40507s | train_loss=0.0151
450
+ step 4480000 | 40580s | train_loss=0.0151
451
+ step 4490000 | 40647s | train_loss=0.0152
452
+ step 4500000 | 40818s | train_loss=0.0152 | fid=67.5459
453
+ step 4510000 | 40902s | train_loss=0.0148
454
+ step 4520000 | 40986s | train_loss=0.0152
455
+ step 4530000 | 41068s | train_loss=0.0149
456
+ step 4540000 | 41152s | train_loss=0.0149
457
+ step 4550000 | 41236s | train_loss=0.0153
458
+ step 4560000 | 41320s | train_loss=0.0153
459
+ step 4570000 | 41403s | train_loss=0.0150
460
+ step 4580000 | 41484s | train_loss=0.0154
461
+ step 4590000 | 41559s | train_loss=0.0148
462
+ step 4600000 | 41724s | train_loss=0.0147 | fid=67.4269
463
+ step 4610000 | 41808s | train_loss=0.0146
464
+ step 4620000 | 41890s | train_loss=0.0147
465
+ step 4630000 | 41974s | train_loss=0.0146
466
+ step 4640000 | 42058s | train_loss=0.0149
467
+ step 4650000 | 42142s | train_loss=0.0149
468
+ step 4660000 | 42227s | train_loss=0.0147
469
+ step 4670000 | 42309s | train_loss=0.0145
470
+ step 4680000 | 42393s | train_loss=0.0146
471
+ step 4690000 | 42462s | train_loss=0.0147
472
+ step 4700000 | 42654s | train_loss=0.0142 | fid=66.6967
473
+ step 4710000 | 42737s | train_loss=0.0146
474
+ step 4720000 | 42819s | train_loss=0.0146
475
+ step 4730000 | 42903s | train_loss=0.0149
476
+ step 4740000 | 42987s | train_loss=0.0146
477
+ step 4750000 | 43071s | train_loss=0.0144
478
+ step 4760000 | 43156s | train_loss=0.0147
479
+ step 4770000 | 43238s | train_loss=0.0144
480
+ step 4780000 | 43322s | train_loss=0.0147
481
+ step 4790000 | 43392s | train_loss=0.0144
482
+ step 4800000 | 43580s | train_loss=0.0144 | fid=67.0134
483
+ step 4810000 | 43662s | train_loss=0.0146
484
+ step 4820000 | 43746s | train_loss=0.0141
485
+ step 4830000 | 43830s | train_loss=0.0148
486
+ step 4840000 | 43914s | train_loss=0.0141
487
+ step 4850000 | 43998s | train_loss=0.0141
488
+ step 4860000 | 44082s | train_loss=0.0142
489
+ step 4870000 | 44164s | train_loss=0.0143
490
+ step 4880000 | 44248s | train_loss=0.0139
491
+ step 4890000 | 44321s | train_loss=0.0143
492
+ step 4900000 | 44492s | train_loss=0.0139 | fid=67.5552
493
+ step 4910000 | 44560s | train_loss=0.0140
494
+ step 4920000 | 44627s | train_loss=0.0139
495
+ step 4930000 | 44694s | train_loss=0.0140
496
+ step 4940000 | 44762s | train_loss=0.0141
497
+ step 4950000 | 44830s | train_loss=0.0140
498
+ step 4960000 | 44897s | train_loss=0.0144
499
+ step 4970000 | 44965s | train_loss=0.0139
500
+ step 4980000 | 45029s | train_loss=0.0140
501
+ step 4990000 | 45087s | train_loss=0.0138
502
+ step 5000000 | 45216s | train_loss=0.0137 | fid=67.7716
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/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/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": 42,
25
+ "device": null
26
+ }
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/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/d200/runner.log ADDED
@@ -0,0 +1,502 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ FID real-side stats cached over 1000 diverse train images
2
+ step 1 | 65s | train_loss=0.9991 | fid=309.2330
3
+ step 10000 | 137s | train_loss=0.6180
4
+ step 20000 | 228s | train_loss=0.5131
5
+ step 30000 | 319s | train_loss=0.4610
6
+ step 40000 | 410s | train_loss=0.4152
7
+ step 50000 | 501s | train_loss=0.3575
8
+ step 60000 | 592s | train_loss=0.3030
9
+ step 70000 | 684s | train_loss=0.2603
10
+ step 80000 | 775s | train_loss=0.2228
11
+ step 90000 | 866s | train_loss=0.2000
12
+ step 100000 | 1024s | train_loss=0.1809 | fid=66.5456
13
+ step 110000 | 1111s | train_loss=0.1639
14
+ step 120000 | 1203s | train_loss=0.1546
15
+ step 130000 | 1293s | train_loss=0.1419
16
+ step 140000 | 1384s | train_loss=0.1371
17
+ step 150000 | 1476s | train_loss=0.1286
18
+ step 160000 | 1567s | train_loss=0.1245
19
+ step 170000 | 1658s | train_loss=0.1254
20
+ step 180000 | 1749s | train_loss=0.1124
21
+ step 190000 | 1840s | train_loss=0.1124
22
+ step 200000 | 1996s | train_loss=0.1096 | fid=75.2916
23
+ step 210000 | 2087s | train_loss=0.1039
24
+ step 220000 | 2175s | train_loss=0.1027
25
+ step 230000 | 2266s | train_loss=0.1001
26
+ step 240000 | 2357s | train_loss=0.0968
27
+ step 250000 | 2448s | train_loss=0.0973
28
+ step 260000 | 2539s | train_loss=0.0951
29
+ step 270000 | 2630s | train_loss=0.0912
30
+ step 280000 | 2721s | train_loss=0.0872
31
+ step 290000 | 2812s | train_loss=0.0933
32
+ step 300000 | 2969s | train_loss=0.0861 | fid=85.1393
33
+ step 310000 | 3060s | train_loss=0.0873
34
+ step 320000 | 3147s | train_loss=0.0838
35
+ step 330000 | 3238s | train_loss=0.0808
36
+ step 340000 | 3329s | train_loss=0.0805
37
+ step 350000 | 3420s | train_loss=0.0778
38
+ step 360000 | 3511s | train_loss=0.0786
39
+ step 370000 | 3602s | train_loss=0.0758
40
+ step 380000 | 3693s | train_loss=0.0802
41
+ step 390000 | 3784s | train_loss=0.0759
42
+ step 400000 | 3940s | train_loss=0.0739 | fid=90.5859
43
+ step 410000 | 4032s | train_loss=0.0713
44
+ step 420000 | 4119s | train_loss=0.0698
45
+ step 430000 | 4210s | train_loss=0.0720
46
+ step 440000 | 4301s | train_loss=0.0717
47
+ step 450000 | 4392s | train_loss=0.0715
48
+ step 460000 | 4483s | train_loss=0.0727
49
+ step 470000 | 4574s | train_loss=0.0670
50
+ step 480000 | 4665s | train_loss=0.0672
51
+ step 490000 | 4756s | train_loss=0.0696
52
+ step 500000 | 4913s | train_loss=0.0684 | fid=96.5825
53
+ step 510000 | 5004s | train_loss=0.0682
54
+ step 520000 | 5096s | train_loss=0.0634
55
+ step 530000 | 5182s | train_loss=0.0678
56
+ step 540000 | 5273s | train_loss=0.0647
57
+ step 550000 | 5364s | train_loss=0.0638
58
+ step 560000 | 5455s | train_loss=0.0656
59
+ step 570000 | 5546s | train_loss=0.0604
60
+ step 580000 | 5637s | train_loss=0.0635
61
+ step 590000 | 5728s | train_loss=0.0614
62
+ step 600000 | 5885s | train_loss=0.0580 | fid=103.1025
63
+ step 610000 | 5976s | train_loss=0.0604
64
+ step 620000 | 6067s | train_loss=0.0628
65
+ step 630000 | 6154s | train_loss=0.0588
66
+ step 640000 | 6245s | train_loss=0.0580
67
+ step 650000 | 6336s | train_loss=0.0586
68
+ step 660000 | 6427s | train_loss=0.0577
69
+ step 670000 | 6518s | train_loss=0.0650
70
+ step 680000 | 6609s | train_loss=0.0570
71
+ step 690000 | 6695s | train_loss=0.0590
72
+ step 700000 | 6856s | train_loss=0.0544 | fid=95.1304
73
+ step 710000 | 6947s | train_loss=0.0607
74
+ step 720000 | 7038s | train_loss=0.0561
75
+ step 730000 | 7129s | train_loss=0.0552
76
+ step 740000 | 7216s | train_loss=0.0545
77
+ step 750000 | 7307s | train_loss=0.0562
78
+ step 760000 | 7398s | train_loss=0.0538
79
+ step 770000 | 7489s | train_loss=0.0529
80
+ step 780000 | 7580s | train_loss=0.0523
81
+ step 790000 | 7666s | train_loss=0.0512
82
+ step 800000 | 7827s | train_loss=0.0520 | fid=93.9043
83
+ step 810000 | 7918s | train_loss=0.0519
84
+ step 820000 | 8009s | train_loss=0.0546
85
+ step 830000 | 8101s | train_loss=0.0501
86
+ step 840000 | 8188s | train_loss=0.0494
87
+ step 850000 | 8279s | train_loss=0.0502
88
+ step 860000 | 8370s | train_loss=0.0520
89
+ step 870000 | 8461s | train_loss=0.0525
90
+ step 880000 | 8552s | train_loss=0.0483
91
+ step 890000 | 8639s | train_loss=0.0504
92
+ step 900000 | 8799s | train_loss=0.0516 | fid=86.9620
93
+ step 910000 | 8890s | train_loss=0.0471
94
+ step 920000 | 8981s | train_loss=0.0536
95
+ step 930000 | 9072s | train_loss=0.0476
96
+ step 940000 | 9161s | train_loss=0.0513
97
+ step 950000 | 9252s | train_loss=0.0491
98
+ step 960000 | 9343s | train_loss=0.0483
99
+ step 970000 | 9433s | train_loss=0.0499
100
+ step 980000 | 9524s | train_loss=0.0460
101
+ step 990000 | 9611s | train_loss=0.0502
102
+ step 1000000 | 9772s | train_loss=0.0454 | fid=85.9270
103
+ step 1010000 | 9864s | train_loss=0.0469
104
+ step 1020000 | 9955s | train_loss=0.0474
105
+ step 1030000 | 10046s | train_loss=0.0491
106
+ step 1040000 | 10138s | train_loss=0.0446
107
+ step 1050000 | 10225s | train_loss=0.0471
108
+ step 1060000 | 10316s | train_loss=0.0476
109
+ step 1070000 | 10407s | train_loss=0.0442
110
+ step 1080000 | 10493s | train_loss=0.0469
111
+ step 1090000 | 10584s | train_loss=0.0432
112
+ step 1100000 | 10745s | train_loss=0.0436 | fid=82.3192
113
+ step 1110000 | 10836s | train_loss=0.0443
114
+ step 1120000 | 10927s | train_loss=0.0435
115
+ step 1130000 | 11018s | train_loss=0.0450
116
+ step 1140000 | 11109s | train_loss=0.0470
117
+ step 1150000 | 11198s | train_loss=0.0474
118
+ step 1160000 | 11289s | train_loss=0.0447
119
+ step 1170000 | 11380s | train_loss=0.0438
120
+ step 1180000 | 11466s | train_loss=0.0408
121
+ step 1190000 | 11557s | train_loss=0.0444
122
+ step 1200000 | 11718s | train_loss=0.0416 | fid=75.1311
123
+ step 1210000 | 11809s | train_loss=0.0459
124
+ step 1220000 | 11900s | train_loss=0.0407
125
+ step 1230000 | 11991s | train_loss=0.0435
126
+ step 1240000 | 12083s | train_loss=0.0422
127
+ step 1250000 | 12172s | train_loss=0.0418
128
+ step 1260000 | 12262s | train_loss=0.0432
129
+ step 1270000 | 12353s | train_loss=0.0397
130
+ step 1280000 | 12440s | train_loss=0.0407
131
+ step 1290000 | 12531s | train_loss=0.0424
132
+ step 1300000 | 12692s | train_loss=0.0381 | fid=75.9222
133
+ step 1310000 | 12783s | train_loss=0.0408
134
+ step 1320000 | 12874s | train_loss=0.0428
135
+ step 1330000 | 12965s | train_loss=0.0398
136
+ step 1340000 | 13056s | train_loss=0.0415
137
+ step 1350000 | 13148s | train_loss=0.0406
138
+ step 1360000 | 13236s | train_loss=0.0397
139
+ step 1370000 | 13325s | train_loss=0.0400
140
+ step 1380000 | 13413s | train_loss=0.0425
141
+ step 1390000 | 13504s | train_loss=0.0398
142
+ step 1400000 | 13665s | train_loss=0.0410 | fid=70.9013
143
+ step 1410000 | 13757s | train_loss=0.0383
144
+ step 1420000 | 13848s | train_loss=0.0375
145
+ step 1430000 | 13939s | train_loss=0.0366
146
+ step 1440000 | 14030s | train_loss=0.0408
147
+ step 1450000 | 14121s | train_loss=0.0400
148
+ step 1460000 | 14209s | train_loss=0.0386
149
+ step 1470000 | 14296s | train_loss=0.0376
150
+ step 1480000 | 14387s | train_loss=0.0398
151
+ step 1490000 | 14478s | train_loss=0.0459
152
+ step 1500000 | 14639s | train_loss=0.0385 | fid=69.9036
153
+ step 1510000 | 14730s | train_loss=0.0365
154
+ step 1520000 | 14822s | train_loss=0.0413
155
+ step 1530000 | 14913s | train_loss=0.0363
156
+ step 1540000 | 15004s | train_loss=0.0362
157
+ step 1550000 | 15095s | train_loss=0.0426
158
+ step 1560000 | 15187s | train_loss=0.0376
159
+ step 1570000 | 15269s | train_loss=0.0376
160
+ step 1580000 | 15361s | train_loss=0.0366
161
+ step 1590000 | 15452s | train_loss=0.0409
162
+ step 1600000 | 15613s | train_loss=0.0393 | fid=60.4878
163
+ step 1610000 | 15704s | train_loss=0.0365
164
+ step 1620000 | 15795s | train_loss=0.0387
165
+ step 1630000 | 15886s | train_loss=0.0376
166
+ step 1640000 | 15977s | train_loss=0.0366
167
+ step 1650000 | 16068s | train_loss=0.0378
168
+ step 1660000 | 16158s | train_loss=0.0360
169
+ step 1670000 | 16243s | train_loss=0.0347
170
+ step 1680000 | 16334s | train_loss=0.0364
171
+ step 1690000 | 16425s | train_loss=0.0366
172
+ step 1700000 | 16587s | train_loss=0.0351 | fid=63.7436
173
+ step 1710000 | 16678s | train_loss=0.0346
174
+ step 1720000 | 16769s | train_loss=0.0343
175
+ step 1730000 | 16860s | train_loss=0.0334
176
+ step 1740000 | 16951s | train_loss=0.0335
177
+ step 1750000 | 17042s | train_loss=0.0349
178
+ step 1760000 | 17129s | train_loss=0.0344
179
+ step 1770000 | 17219s | train_loss=0.0341
180
+ step 1780000 | 17308s | train_loss=0.0348
181
+ step 1790000 | 17400s | train_loss=0.0393
182
+ step 1800000 | 17561s | train_loss=0.0340 | fid=63.5764
183
+ step 1810000 | 17652s | train_loss=0.0324
184
+ step 1820000 | 17743s | train_loss=0.0333
185
+ step 1830000 | 17834s | train_loss=0.0385
186
+ step 1840000 | 17925s | train_loss=0.0365
187
+ step 1850000 | 18016s | train_loss=0.0400
188
+ step 1860000 | 18103s | train_loss=0.0353
189
+ step 1870000 | 18194s | train_loss=0.0349
190
+ step 1880000 | 18282s | train_loss=0.0375
191
+ step 1890000 | 18373s | train_loss=0.0328
192
+ step 1900000 | 18534s | train_loss=0.0374 | fid=58.5621
193
+ step 1910000 | 18625s | train_loss=0.0339
194
+ step 1920000 | 18716s | train_loss=0.0347
195
+ step 1930000 | 18807s | train_loss=0.0344
196
+ step 1940000 | 18898s | train_loss=0.0340
197
+ step 1950000 | 18989s | train_loss=0.0330
198
+ step 1960000 | 19077s | train_loss=0.0335
199
+ step 1970000 | 19168s | train_loss=0.0342
200
+ step 1980000 | 19256s | train_loss=0.0349
201
+ step 1990000 | 19347s | train_loss=0.0336
202
+ step 2000000 | 19509s | train_loss=0.0372 | fid=55.2117
203
+ step 2010000 | 19600s | train_loss=0.0316
204
+ step 2020000 | 19691s | train_loss=0.0304
205
+ step 2030000 | 19782s | train_loss=0.0304
206
+ step 2040000 | 19873s | train_loss=0.0337
207
+ step 2050000 | 19962s | train_loss=0.0310
208
+ step 2060000 | 20051s | train_loss=0.0338
209
+ step 2070000 | 20142s | train_loss=0.0320
210
+ step 2080000 | 20233s | train_loss=0.0301
211
+ step 2090000 | 20321s | train_loss=0.0347
212
+ step 2100000 | 20482s | train_loss=0.0295 | fid=54.1778
213
+ step 2110000 | 20573s | train_loss=0.0372
214
+ step 2120000 | 20664s | train_loss=0.0323
215
+ step 2130000 | 20756s | train_loss=0.0307
216
+ step 2140000 | 20846s | train_loss=0.0302
217
+ step 2150000 | 20934s | train_loss=0.0336
218
+ step 2160000 | 21025s | train_loss=0.0307
219
+ step 2170000 | 21116s | train_loss=0.0307
220
+ step 2180000 | 21207s | train_loss=0.0297
221
+ step 2190000 | 21296s | train_loss=0.0304
222
+ step 2200000 | 21457s | train_loss=0.0319 | fid=56.9937
223
+ step 2210000 | 21548s | train_loss=0.0331
224
+ step 2220000 | 21639s | train_loss=0.0318
225
+ step 2230000 | 21730s | train_loss=0.0310
226
+ step 2240000 | 21821s | train_loss=0.0314
227
+ step 2250000 | 21908s | train_loss=0.0303
228
+ step 2260000 | 21999s | train_loss=0.0304
229
+ step 2270000 | 22090s | train_loss=0.0320
230
+ step 2280000 | 22181s | train_loss=0.0306
231
+ step 2290000 | 22270s | train_loss=0.0296
232
+ step 2300000 | 22430s | train_loss=0.0308 | fid=55.7033
233
+ step 2310000 | 22521s | train_loss=0.0296
234
+ step 2320000 | 22612s | train_loss=0.0298
235
+ step 2330000 | 22703s | train_loss=0.0296
236
+ step 2340000 | 22792s | train_loss=0.0282
237
+ step 2350000 | 22881s | train_loss=0.0316
238
+ step 2360000 | 22972s | train_loss=0.0300
239
+ step 2370000 | 23063s | train_loss=0.0305
240
+ step 2380000 | 23154s | train_loss=0.0293
241
+ step 2390000 | 23246s | train_loss=0.0281
242
+ step 2400000 | 23404s | train_loss=0.0289 | fid=53.8030
243
+ step 2410000 | 23495s | train_loss=0.0306
244
+ step 2420000 | 23586s | train_loss=0.0285
245
+ step 2430000 | 23677s | train_loss=0.0307
246
+ step 2440000 | 23764s | train_loss=0.0285
247
+ step 2450000 | 23855s | train_loss=0.0302
248
+ step 2460000 | 23946s | train_loss=0.0299
249
+ step 2470000 | 24037s | train_loss=0.0308
250
+ step 2480000 | 24128s | train_loss=0.0299
251
+ step 2490000 | 24219s | train_loss=0.0285
252
+ step 2500000 | 24376s | train_loss=0.0279 | fid=52.3568
253
+ step 2510000 | 24467s | train_loss=0.0282
254
+ step 2520000 | 24558s | train_loss=0.0285
255
+ step 2530000 | 24650s | train_loss=0.0319
256
+ step 2540000 | 24736s | train_loss=0.0291
257
+ step 2550000 | 24827s | train_loss=0.0287
258
+ step 2560000 | 24918s | train_loss=0.0263
259
+ step 2570000 | 25009s | train_loss=0.0330
260
+ step 2580000 | 25100s | train_loss=0.0322
261
+ step 2590000 | 25192s | train_loss=0.0299
262
+ step 2600000 | 25349s | train_loss=0.0311 | fid=52.4883
263
+ step 2610000 | 25440s | train_loss=0.0285
264
+ step 2620000 | 25531s | train_loss=0.0318
265
+ step 2630000 | 25620s | train_loss=0.0280
266
+ step 2640000 | 25709s | train_loss=0.0281
267
+ step 2650000 | 25800s | train_loss=0.0273
268
+ step 2660000 | 25891s | train_loss=0.0285
269
+ step 2670000 | 25982s | train_loss=0.0269
270
+ step 2680000 | 26073s | train_loss=0.0282
271
+ step 2690000 | 26164s | train_loss=0.0301
272
+ step 2700000 | 26321s | train_loss=0.0298 | fid=51.2828
273
+ step 2710000 | 26412s | train_loss=0.0266
274
+ step 2720000 | 26503s | train_loss=0.0283
275
+ step 2730000 | 26593s | train_loss=0.0260
276
+ step 2740000 | 26682s | train_loss=0.0275
277
+ step 2750000 | 26773s | train_loss=0.0327
278
+ step 2760000 | 26864s | train_loss=0.0273
279
+ step 2770000 | 26955s | train_loss=0.0271
280
+ step 2780000 | 27046s | train_loss=0.0271
281
+ step 2790000 | 27137s | train_loss=0.0283
282
+ step 2800000 | 27297s | train_loss=0.0302 | fid=51.5236
283
+ step 2810000 | 27385s | train_loss=0.0278
284
+ step 2820000 | 27476s | train_loss=0.0275
285
+ step 2830000 | 27563s | train_loss=0.0277
286
+ step 2840000 | 27654s | train_loss=0.0273
287
+ step 2850000 | 27745s | train_loss=0.0269
288
+ step 2860000 | 27836s | train_loss=0.0277
289
+ step 2870000 | 27927s | train_loss=0.0257
290
+ step 2880000 | 28018s | train_loss=0.0288
291
+ step 2890000 | 28109s | train_loss=0.0266
292
+ step 2900000 | 28269s | train_loss=0.0272 | fid=53.5902
293
+ step 2910000 | 28358s | train_loss=0.0291
294
+ step 2920000 | 28449s | train_loss=0.0263
295
+ step 2930000 | 28535s | train_loss=0.0298
296
+ step 2940000 | 28626s | train_loss=0.0256
297
+ step 2950000 | 28717s | train_loss=0.0266
298
+ step 2960000 | 28809s | train_loss=0.0256
299
+ step 2970000 | 28900s | train_loss=0.0285
300
+ step 2980000 | 28991s | train_loss=0.0258
301
+ step 2990000 | 29082s | train_loss=0.0256
302
+ step 3000000 | 29242s | train_loss=0.0256 | fid=50.9749
303
+ step 3010000 | 29332s | train_loss=0.0266
304
+ step 3020000 | 29420s | train_loss=0.0276
305
+ step 3030000 | 29508s | train_loss=0.0238
306
+ step 3040000 | 29599s | train_loss=0.0294
307
+ step 3050000 | 29690s | train_loss=0.0243
308
+ step 3060000 | 29782s | train_loss=0.0281
309
+ step 3070000 | 29873s | train_loss=0.0255
310
+ step 3080000 | 29964s | train_loss=0.0268
311
+ step 3090000 | 30055s | train_loss=0.0265
312
+ step 3100000 | 30215s | train_loss=0.0274 | fid=50.4457
313
+ step 3110000 | 30306s | train_loss=0.0254
314
+ step 3120000 | 30391s | train_loss=0.0252
315
+ step 3130000 | 30480s | train_loss=0.0275
316
+ step 3140000 | 30571s | train_loss=0.0285
317
+ step 3150000 | 30662s | train_loss=0.0258
318
+ step 3160000 | 30753s | train_loss=0.0259
319
+ step 3170000 | 30844s | train_loss=0.0288
320
+ step 3180000 | 30935s | train_loss=0.0247
321
+ step 3190000 | 31026s | train_loss=0.0251
322
+ step 3200000 | 31185s | train_loss=0.0271 | fid=47.0454
323
+ step 3210000 | 31276s | train_loss=0.0246
324
+ step 3220000 | 31359s | train_loss=0.0270
325
+ step 3230000 | 31449s | train_loss=0.0250
326
+ step 3240000 | 31540s | train_loss=0.0273
327
+ step 3250000 | 31631s | train_loss=0.0266
328
+ step 3260000 | 31722s | train_loss=0.0250
329
+ step 3270000 | 31813s | train_loss=0.0283
330
+ step 3280000 | 31904s | train_loss=0.0252
331
+ step 3290000 | 31995s | train_loss=0.0245
332
+ step 3300000 | 32154s | train_loss=0.0264 | fid=56.0010
333
+ step 3310000 | 32243s | train_loss=0.0246
334
+ step 3320000 | 32332s | train_loss=0.0253
335
+ step 3330000 | 32420s | train_loss=0.0244
336
+ step 3340000 | 32511s | train_loss=0.0222
337
+ step 3350000 | 32602s | train_loss=0.0257
338
+ step 3360000 | 32693s | train_loss=0.0276
339
+ step 3370000 | 32784s | train_loss=0.0263
340
+ step 3380000 | 32875s | train_loss=0.0258
341
+ step 3390000 | 32966s | train_loss=0.0237
342
+ step 3400000 | 33126s | train_loss=0.0269 | fid=55.4919
343
+ step 3410000 | 33215s | train_loss=0.0259
344
+ step 3420000 | 33304s | train_loss=0.0229
345
+ step 3430000 | 33392s | train_loss=0.0269
346
+ step 3440000 | 33483s | train_loss=0.0240
347
+ step 3450000 | 33574s | train_loss=0.0249
348
+ step 3460000 | 33665s | train_loss=0.0272
349
+ step 3470000 | 33756s | train_loss=0.0250
350
+ step 3480000 | 33847s | train_loss=0.0263
351
+ step 3490000 | 33939s | train_loss=0.0274
352
+ step 3500000 | 34098s | train_loss=0.0228 | fid=56.4175
353
+ step 3510000 | 34187s | train_loss=0.0259
354
+ step 3520000 | 34276s | train_loss=0.0277
355
+ step 3530000 | 34368s | train_loss=0.0238
356
+ step 3540000 | 34454s | train_loss=0.0253
357
+ step 3550000 | 34545s | train_loss=0.0250
358
+ step 3560000 | 34637s | train_loss=0.0254
359
+ step 3570000 | 34728s | train_loss=0.0241
360
+ step 3580000 | 34819s | train_loss=0.0284
361
+ step 3590000 | 34910s | train_loss=0.0243
362
+ step 3600000 | 35069s | train_loss=0.0228 | fid=51.8542
363
+ step 3610000 | 35156s | train_loss=0.0268
364
+ step 3620000 | 35247s | train_loss=0.0235
365
+ step 3630000 | 35338s | train_loss=0.0233
366
+ step 3640000 | 35425s | train_loss=0.0224
367
+ step 3650000 | 35516s | train_loss=0.0227
368
+ step 3660000 | 35607s | train_loss=0.0229
369
+ step 3670000 | 35698s | train_loss=0.0251
370
+ step 3680000 | 35789s | train_loss=0.0230
371
+ step 3690000 | 35880s | train_loss=0.0240
372
+ step 3700000 | 36039s | train_loss=0.0233 | fid=51.6590
373
+ step 3710000 | 36128s | train_loss=0.0236
374
+ step 3720000 | 36219s | train_loss=0.0237
375
+ step 3730000 | 36309s | train_loss=0.0242
376
+ step 3740000 | 36401s | train_loss=0.0223
377
+ step 3750000 | 36487s | train_loss=0.0246
378
+ step 3760000 | 36578s | train_loss=0.0229
379
+ step 3770000 | 36669s | train_loss=0.0240
380
+ step 3780000 | 36760s | train_loss=0.0239
381
+ step 3790000 | 36851s | train_loss=0.0252
382
+ step 3800000 | 37008s | train_loss=0.0230 | fid=53.8361
383
+ step 3810000 | 37097s | train_loss=0.0222
384
+ step 3820000 | 37188s | train_loss=0.0253
385
+ step 3830000 | 37279s | train_loss=0.0223
386
+ step 3840000 | 37370s | train_loss=0.0237
387
+ step 3850000 | 37457s | train_loss=0.0232
388
+ step 3860000 | 37548s | train_loss=0.0230
389
+ step 3870000 | 37639s | train_loss=0.0219
390
+ step 3880000 | 37730s | train_loss=0.0240
391
+ step 3890000 | 37821s | train_loss=0.0219
392
+ step 3900000 | 37977s | train_loss=0.0226 | fid=51.5403
393
+ step 3910000 | 38066s | train_loss=0.0223
394
+ step 3920000 | 38157s | train_loss=0.0224
395
+ step 3930000 | 38248s | train_loss=0.0215
396
+ step 3940000 | 38339s | train_loss=0.0218
397
+ step 3950000 | 38428s | train_loss=0.0223
398
+ step 3960000 | 38517s | train_loss=0.0210
399
+ step 3970000 | 38608s | train_loss=0.0217
400
+ step 3980000 | 38699s | train_loss=0.0212
401
+ step 3990000 | 38790s | train_loss=0.0220
402
+ step 4000000 | 38945s | train_loss=0.0238 | fid=53.5927
403
+ step 4010000 | 39036s | train_loss=0.0272
404
+ step 4020000 | 39127s | train_loss=0.0238
405
+ step 4030000 | 39218s | train_loss=0.0222
406
+ step 4040000 | 39309s | train_loss=0.0221
407
+ step 4050000 | 39400s | train_loss=0.0235
408
+ step 4060000 | 39487s | train_loss=0.0239
409
+ step 4070000 | 39578s | train_loss=0.0229
410
+ step 4080000 | 39669s | train_loss=0.0219
411
+ step 4090000 | 39760s | train_loss=0.0220
412
+ step 4100000 | 39916s | train_loss=0.0216 | fid=49.7605
413
+ step 4110000 | 40007s | train_loss=0.0230
414
+ step 4120000 | 40098s | train_loss=0.0217
415
+ step 4130000 | 40189s | train_loss=0.0218
416
+ step 4140000 | 40280s | train_loss=0.0234
417
+ step 4150000 | 40371s | train_loss=0.0218
418
+ step 4160000 | 40458s | train_loss=0.0238
419
+ step 4170000 | 40549s | train_loss=0.0224
420
+ step 4180000 | 40640s | train_loss=0.0235
421
+ step 4190000 | 40731s | train_loss=0.0207
422
+ step 4200000 | 40886s | train_loss=0.0229 | fid=54.3718
423
+ step 4210000 | 40977s | train_loss=0.0216
424
+ step 4220000 | 41068s | train_loss=0.0227
425
+ step 4230000 | 41159s | train_loss=0.0242
426
+ step 4240000 | 41250s | train_loss=0.0212
427
+ step 4250000 | 41341s | train_loss=0.0211
428
+ step 4260000 | 41433s | train_loss=0.0223
429
+ step 4270000 | 41519s | train_loss=0.0224
430
+ step 4280000 | 41610s | train_loss=0.0223
431
+ step 4290000 | 41701s | train_loss=0.0210
432
+ step 4300000 | 41857s | train_loss=0.0222 | fid=50.5967
433
+ step 4310000 | 41948s | train_loss=0.0215
434
+ step 4320000 | 42039s | train_loss=0.0230
435
+ step 4330000 | 42130s | train_loss=0.0207
436
+ step 4340000 | 42221s | train_loss=0.0217
437
+ step 4350000 | 42313s | train_loss=0.0214
438
+ step 4360000 | 42404s | train_loss=0.0214
439
+ step 4370000 | 42492s | train_loss=0.0222
440
+ step 4380000 | 42583s | train_loss=0.0217
441
+ step 4390000 | 42671s | train_loss=0.0210
442
+ step 4400000 | 42829s | train_loss=0.0213 | fid=53.6237
443
+ step 4410000 | 42920s | train_loss=0.0206
444
+ step 4420000 | 43011s | train_loss=0.0231
445
+ step 4430000 | 43102s | train_loss=0.0210
446
+ step 4440000 | 43193s | train_loss=0.0206
447
+ step 4450000 | 43284s | train_loss=0.0232
448
+ step 4460000 | 43375s | train_loss=0.0206
449
+ step 4470000 | 43467s | train_loss=0.0226
450
+ step 4480000 | 43554s | train_loss=0.0221
451
+ step 4490000 | 43643s | train_loss=0.0209
452
+ step 4500000 | 43801s | train_loss=0.0216 | fid=52.9946
453
+ step 4510000 | 43892s | train_loss=0.0210
454
+ step 4520000 | 43983s | train_loss=0.0209
455
+ step 4530000 | 44074s | train_loss=0.0214
456
+ step 4540000 | 44165s | train_loss=0.0217
457
+ step 4550000 | 44256s | train_loss=0.0210
458
+ step 4560000 | 44347s | train_loss=0.0198
459
+ step 4570000 | 44439s | train_loss=0.0194
460
+ step 4580000 | 44527s | train_loss=0.0214
461
+ step 4590000 | 44614s | train_loss=0.0204
462
+ step 4600000 | 44774s | train_loss=0.0206 | fid=51.8797
463
+ step 4610000 | 44865s | train_loss=0.0209
464
+ step 4620000 | 44956s | train_loss=0.0199
465
+ step 4630000 | 45047s | train_loss=0.0218
466
+ step 4640000 | 45138s | train_loss=0.0240
467
+ step 4650000 | 45229s | train_loss=0.0225
468
+ step 4660000 | 45321s | train_loss=0.0222
469
+ step 4670000 | 45412s | train_loss=0.0220
470
+ step 4680000 | 45501s | train_loss=0.0197
471
+ step 4690000 | 45585s | train_loss=0.0200
472
+ step 4700000 | 45745s | train_loss=0.0212 | fid=53.3237
473
+ step 4710000 | 45836s | train_loss=0.0203
474
+ step 4720000 | 45927s | train_loss=0.0201
475
+ step 4730000 | 46018s | train_loss=0.0214
476
+ step 4740000 | 46109s | train_loss=0.0198
477
+ step 4750000 | 46201s | train_loss=0.0193
478
+ step 4760000 | 46292s | train_loss=0.0203
479
+ step 4770000 | 46383s | train_loss=0.0193
480
+ step 4780000 | 46473s | train_loss=0.0215
481
+ step 4790000 | 46557s | train_loss=0.0214
482
+ step 4800000 | 46716s | train_loss=0.0200 | fid=54.3551
483
+ step 4810000 | 46807s | train_loss=0.0203
484
+ step 4820000 | 46898s | train_loss=0.0205
485
+ step 4830000 | 46989s | train_loss=0.0185
486
+ step 4840000 | 47080s | train_loss=0.0187
487
+ step 4850000 | 47172s | train_loss=0.0199
488
+ step 4860000 | 47263s | train_loss=0.0220
489
+ step 4870000 | 47354s | train_loss=0.0211
490
+ step 4880000 | 47436s | train_loss=0.0200
491
+ step 4890000 | 47489s | train_loss=0.0200
492
+ step 4900000 | 47586s | train_loss=0.0207 | fid=53.3152
493
+ step 4910000 | 47642s | train_loss=0.0206
494
+ step 4920000 | 47697s | train_loss=0.0222
495
+ step 4930000 | 47753s | train_loss=0.0199
496
+ step 4940000 | 47809s | train_loss=0.0217
497
+ step 4950000 | 47865s | train_loss=0.0206
498
+ step 4960000 | 47920s | train_loss=0.0217
499
+ step 4970000 | 47976s | train_loss=0.0197
500
+ step 4980000 | 48032s | train_loss=0.0199
501
+ step 4990000 | 48088s | train_loss=0.0193
502
+ step 5000000 | 48176s | train_loss=0.0219 | fid=57.4913
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/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/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": 42,
25
+ "device": null
26
+ }
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/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/d30/runner.log ADDED
@@ -0,0 +1,502 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ FID real-side stats cached over 1000 diverse train images
2
+ step 1 | 67s | train_loss=0.9884 | fid=247.6490
3
+ step 10000 | 154s | train_loss=0.8574
4
+ step 20000 | 242s | train_loss=0.8460
5
+ step 30000 | 329s | train_loss=0.8296
6
+ step 40000 | 417s | train_loss=0.7968
7
+ step 50000 | 505s | train_loss=0.7418
8
+ step 60000 | 593s | train_loss=0.6621
9
+ step 70000 | 681s | train_loss=0.5516
10
+ step 80000 | 769s | train_loss=0.4341
11
+ step 90000 | 857s | train_loss=0.3379
12
+ step 100000 | 1004s | train_loss=0.2674 | fid=62.6431
13
+ step 110000 | 1090s | train_loss=0.2273
14
+ step 120000 | 1178s | train_loss=0.1988
15
+ step 130000 | 1265s | train_loss=0.1792
16
+ step 140000 | 1353s | train_loss=0.1634
17
+ step 150000 | 1440s | train_loss=0.1523
18
+ step 160000 | 1528s | train_loss=0.1424
19
+ step 170000 | 1615s | train_loss=0.1357
20
+ step 180000 | 1703s | train_loss=0.1275
21
+ step 190000 | 1790s | train_loss=0.1205
22
+ step 200000 | 1939s | train_loss=0.1153 | fid=60.8282
23
+ step 210000 | 2026s | train_loss=0.1100
24
+ step 220000 | 2113s | train_loss=0.1045
25
+ step 230000 | 2201s | train_loss=0.1008
26
+ step 240000 | 2289s | train_loss=0.0972
27
+ step 250000 | 2376s | train_loss=0.0935
28
+ step 260000 | 2464s | train_loss=0.0905
29
+ step 270000 | 2552s | train_loss=0.0886
30
+ step 280000 | 2639s | train_loss=0.0850
31
+ step 290000 | 2727s | train_loss=0.0824
32
+ step 300000 | 2877s | train_loss=0.0802 | fid=60.0789
33
+ step 310000 | 2965s | train_loss=0.0789
34
+ step 320000 | 3051s | train_loss=0.0759
35
+ step 330000 | 3138s | train_loss=0.0750
36
+ step 340000 | 3226s | train_loss=0.0718
37
+ step 350000 | 3314s | train_loss=0.0708
38
+ step 360000 | 3402s | train_loss=0.0688
39
+ step 370000 | 3490s | train_loss=0.0679
40
+ step 380000 | 3578s | train_loss=0.0662
41
+ step 390000 | 3664s | train_loss=0.0648
42
+ step 400000 | 3814s | train_loss=0.0627 | fid=59.2028
43
+ step 410000 | 3902s | train_loss=0.0619
44
+ step 420000 | 3988s | train_loss=0.0609
45
+ step 430000 | 4076s | train_loss=0.0594
46
+ step 440000 | 4164s | train_loss=0.0584
47
+ step 450000 | 4252s | train_loss=0.0584
48
+ step 460000 | 4339s | train_loss=0.0569
49
+ step 470000 | 4427s | train_loss=0.0556
50
+ step 480000 | 4515s | train_loss=0.0548
51
+ step 490000 | 4599s | train_loss=0.0538
52
+ step 500000 | 4751s | train_loss=0.0520 | fid=59.2954
53
+ step 510000 | 4839s | train_loss=0.0518
54
+ step 520000 | 4926s | train_loss=0.0513
55
+ step 530000 | 5013s | train_loss=0.0508
56
+ step 540000 | 5101s | train_loss=0.0503
57
+ step 550000 | 5189s | train_loss=0.0497
58
+ step 560000 | 5277s | train_loss=0.0489
59
+ step 570000 | 5364s | train_loss=0.0482
60
+ step 580000 | 5452s | train_loss=0.0472
61
+ step 590000 | 5533s | train_loss=0.0468
62
+ step 600000 | 5688s | train_loss=0.0467 | fid=58.4510
63
+ step 610000 | 5776s | train_loss=0.0461
64
+ step 620000 | 5864s | train_loss=0.0455
65
+ step 630000 | 5950s | train_loss=0.0451
66
+ step 640000 | 6038s | train_loss=0.0447
67
+ step 650000 | 6125s | train_loss=0.0442
68
+ step 660000 | 6213s | train_loss=0.0439
69
+ step 670000 | 6301s | train_loss=0.0431
70
+ step 680000 | 6384s | train_loss=0.0427
71
+ step 690000 | 6470s | train_loss=0.0419
72
+ step 700000 | 6626s | train_loss=0.0416 | fid=58.0972
73
+ step 710000 | 6714s | train_loss=0.0411
74
+ step 720000 | 6801s | train_loss=0.0414
75
+ step 730000 | 6888s | train_loss=0.0406
76
+ step 740000 | 6976s | train_loss=0.0408
77
+ step 750000 | 7063s | train_loss=0.0393
78
+ step 760000 | 7151s | train_loss=0.0397
79
+ step 770000 | 7239s | train_loss=0.0392
80
+ step 780000 | 7322s | train_loss=0.0388
81
+ step 790000 | 7407s | train_loss=0.0384
82
+ step 800000 | 7564s | train_loss=0.0379 | fid=57.9697
83
+ step 810000 | 7651s | train_loss=0.0379
84
+ step 820000 | 7739s | train_loss=0.0374
85
+ step 830000 | 7827s | train_loss=0.0369
86
+ step 840000 | 7914s | train_loss=0.0363
87
+ step 850000 | 8001s | train_loss=0.0365
88
+ step 860000 | 8089s | train_loss=0.0366
89
+ step 870000 | 8174s | train_loss=0.0363
90
+ step 880000 | 8258s | train_loss=0.0361
91
+ step 890000 | 8345s | train_loss=0.0355
92
+ step 900000 | 8501s | train_loss=0.0349 | fid=57.4109
93
+ step 910000 | 8589s | train_loss=0.0346
94
+ step 920000 | 8677s | train_loss=0.0346
95
+ step 930000 | 8765s | train_loss=0.0346
96
+ step 940000 | 8851s | train_loss=0.0339
97
+ step 950000 | 8939s | train_loss=0.0333
98
+ step 960000 | 9027s | train_loss=0.0337
99
+ step 970000 | 9110s | train_loss=0.0332
100
+ step 980000 | 9195s | train_loss=0.0327
101
+ step 990000 | 9283s | train_loss=0.0335
102
+ step 1000000 | 9440s | train_loss=0.0326 | fid=56.2507
103
+ step 1010000 | 9528s | train_loss=0.0328
104
+ step 1020000 | 9616s | train_loss=0.0319
105
+ step 1030000 | 9703s | train_loss=0.0322
106
+ step 1040000 | 9792s | train_loss=0.0311
107
+ step 1050000 | 9877s | train_loss=0.0321
108
+ step 1060000 | 9963s | train_loss=0.0312
109
+ step 1070000 | 10048s | train_loss=0.0306
110
+ step 1080000 | 10133s | train_loss=0.0313
111
+ step 1090000 | 10221s | train_loss=0.0308
112
+ step 1100000 | 10378s | train_loss=0.0305 | fid=57.6960
113
+ step 1110000 | 10465s | train_loss=0.0307
114
+ step 1120000 | 10553s | train_loss=0.0299
115
+ step 1130000 | 10641s | train_loss=0.0296
116
+ step 1140000 | 10729s | train_loss=0.0298
117
+ step 1150000 | 10815s | train_loss=0.0292
118
+ step 1160000 | 10898s | train_loss=0.0290
119
+ step 1170000 | 10985s | train_loss=0.0293
120
+ step 1180000 | 11071s | train_loss=0.0292
121
+ step 1190000 | 11159s | train_loss=0.0288
122
+ step 1200000 | 11313s | train_loss=0.0291 | fid=57.0255
123
+ step 1210000 | 11400s | train_loss=0.0288
124
+ step 1220000 | 11488s | train_loss=0.0283
125
+ step 1230000 | 11576s | train_loss=0.0286
126
+ step 1240000 | 11664s | train_loss=0.0280
127
+ step 1250000 | 11752s | train_loss=0.0279
128
+ step 1260000 | 11832s | train_loss=0.0273
129
+ step 1270000 | 11918s | train_loss=0.0272
130
+ step 1280000 | 12006s | train_loss=0.0273
131
+ step 1290000 | 12093s | train_loss=0.0274
132
+ step 1300000 | 12248s | train_loss=0.0271 | fid=57.8118
133
+ step 1310000 | 12336s | train_loss=0.0269
134
+ step 1320000 | 12424s | train_loss=0.0270
135
+ step 1330000 | 12512s | train_loss=0.0269
136
+ step 1340000 | 12600s | train_loss=0.0268
137
+ step 1350000 | 12684s | train_loss=0.0259
138
+ step 1360000 | 12769s | train_loss=0.0264
139
+ step 1370000 | 12855s | train_loss=0.0262
140
+ step 1380000 | 12943s | train_loss=0.0261
141
+ step 1390000 | 13030s | train_loss=0.0260
142
+ step 1400000 | 13187s | train_loss=0.0254 | fid=56.4424
143
+ step 1410000 | 13274s | train_loss=0.0255
144
+ step 1420000 | 13362s | train_loss=0.0253
145
+ step 1430000 | 13450s | train_loss=0.0257
146
+ step 1440000 | 13538s | train_loss=0.0256
147
+ step 1450000 | 13620s | train_loss=0.0253
148
+ step 1460000 | 13707s | train_loss=0.0263
149
+ step 1470000 | 13792s | train_loss=0.0250
150
+ step 1480000 | 13880s | train_loss=0.0249
151
+ step 1490000 | 13968s | train_loss=0.0245
152
+ step 1500000 | 14125s | train_loss=0.0244 | fid=57.1871
153
+ step 1510000 | 14213s | train_loss=0.0249
154
+ step 1520000 | 14301s | train_loss=0.0242
155
+ step 1530000 | 14389s | train_loss=0.0241
156
+ step 1540000 | 14472s | train_loss=0.0242
157
+ step 1550000 | 14548s | train_loss=0.0240
158
+ step 1560000 | 14629s | train_loss=0.0241
159
+ step 1570000 | 14706s | train_loss=0.0237
160
+ step 1580000 | 14793s | train_loss=0.0231
161
+ step 1590000 | 14881s | train_loss=0.0237
162
+ step 1600000 | 15060s | train_loss=0.0232 | fid=57.8615
163
+ step 1610000 | 15147s | train_loss=0.0231
164
+ step 1620000 | 15235s | train_loss=0.0235
165
+ step 1630000 | 15323s | train_loss=0.0227
166
+ step 1640000 | 15405s | train_loss=0.0231
167
+ step 1650000 | 15493s | train_loss=0.0230
168
+ step 1660000 | 15578s | train_loss=0.0230
169
+ step 1670000 | 15664s | train_loss=0.0224
170
+ step 1680000 | 15752s | train_loss=0.0225
171
+ step 1690000 | 15840s | train_loss=0.0225
172
+ step 1700000 | 15996s | train_loss=0.0224 | fid=56.7030
173
+ step 1710000 | 16084s | train_loss=0.0220
174
+ step 1720000 | 16172s | train_loss=0.0222
175
+ step 1730000 | 16260s | train_loss=0.0221
176
+ step 1740000 | 16343s | train_loss=0.0218
177
+ step 1750000 | 16431s | train_loss=0.0221
178
+ step 1760000 | 16515s | train_loss=0.0221
179
+ step 1770000 | 16602s | train_loss=0.0218
180
+ step 1780000 | 16689s | train_loss=0.0218
181
+ step 1790000 | 16777s | train_loss=0.0216
182
+ step 1800000 | 16935s | train_loss=0.0213 | fid=57.8741
183
+ step 1810000 | 17023s | train_loss=0.0213
184
+ step 1820000 | 17110s | train_loss=0.0214
185
+ step 1830000 | 17193s | train_loss=0.0214
186
+ step 1840000 | 17281s | train_loss=0.0208
187
+ step 1850000 | 17368s | train_loss=0.0211
188
+ step 1860000 | 17453s | train_loss=0.0210
189
+ step 1870000 | 17541s | train_loss=0.0213
190
+ step 1880000 | 17627s | train_loss=0.0207
191
+ step 1890000 | 17714s | train_loss=0.0205
192
+ step 1900000 | 17872s | train_loss=0.0203 | fid=57.1356
193
+ step 1910000 | 17960s | train_loss=0.0204
194
+ step 1920000 | 18047s | train_loss=0.0203
195
+ step 1930000 | 18130s | train_loss=0.0203
196
+ step 1940000 | 18218s | train_loss=0.0205
197
+ step 1950000 | 18303s | train_loss=0.0200
198
+ step 1960000 | 18391s | train_loss=0.0203
199
+ step 1970000 | 18479s | train_loss=0.0203
200
+ step 1980000 | 18565s | train_loss=0.0200
201
+ step 1990000 | 18652s | train_loss=0.0199
202
+ step 2000000 | 18809s | train_loss=0.0198 | fid=56.4199
203
+ step 2010000 | 18897s | train_loss=0.0195
204
+ step 2020000 | 18982s | train_loss=0.0195
205
+ step 2030000 | 19068s | train_loss=0.0200
206
+ step 2040000 | 19156s | train_loss=0.0202
207
+ step 2050000 | 19241s | train_loss=0.0202
208
+ step 2060000 | 19329s | train_loss=0.0201
209
+ step 2070000 | 19416s | train_loss=0.0196
210
+ step 2080000 | 19504s | train_loss=0.0192
211
+ step 2090000 | 19591s | train_loss=0.0191
212
+ step 2100000 | 19782s | train_loss=0.0192 | fid=56.5452
213
+ step 2110000 | 19867s | train_loss=0.0194
214
+ step 2120000 | 19944s | train_loss=0.0188
215
+ step 2130000 | 20032s | train_loss=0.0189
216
+ step 2140000 | 20115s | train_loss=0.0187
217
+ step 2150000 | 20196s | train_loss=0.0189
218
+ step 2160000 | 20284s | train_loss=0.0186
219
+ step 2170000 | 20372s | train_loss=0.0191
220
+ step 2180000 | 20458s | train_loss=0.0185
221
+ step 2190000 | 20546s | train_loss=0.0187
222
+ step 2200000 | 20728s | train_loss=0.0183 | fid=55.9710
223
+ step 2210000 | 20808s | train_loss=0.0186
224
+ step 2220000 | 20892s | train_loss=0.0184
225
+ step 2230000 | 20980s | train_loss=0.0183
226
+ step 2240000 | 21060s | train_loss=0.0182
227
+ step 2250000 | 21142s | train_loss=0.0183
228
+ step 2260000 | 21230s | train_loss=0.0185
229
+ step 2270000 | 21318s | train_loss=0.0180
230
+ step 2280000 | 21404s | train_loss=0.0181
231
+ step 2290000 | 21492s | train_loss=0.0181
232
+ step 2300000 | 21671s | train_loss=0.0179 | fid=56.6094
233
+ step 2310000 | 21746s | train_loss=0.0181
234
+ step 2320000 | 21832s | train_loss=0.0181
235
+ step 2330000 | 21920s | train_loss=0.0176
236
+ step 2340000 | 21998s | train_loss=0.0176
237
+ step 2350000 | 22079s | train_loss=0.0179
238
+ step 2360000 | 22167s | train_loss=0.0173
239
+ step 2370000 | 22254s | train_loss=0.0179
240
+ step 2380000 | 22341s | train_loss=0.0178
241
+ step 2390000 | 22428s | train_loss=0.0177
242
+ step 2400000 | 22620s | train_loss=0.0174 | fid=56.5607
243
+ step 2410000 | 22702s | train_loss=0.0172
244
+ step 2420000 | 22789s | train_loss=0.0173
245
+ step 2430000 | 22877s | train_loss=0.0170
246
+ step 2440000 | 22954s | train_loss=0.0169
247
+ step 2450000 | 23038s | train_loss=0.0169
248
+ step 2460000 | 23126s | train_loss=0.0170
249
+ step 2470000 | 23214s | train_loss=0.0170
250
+ step 2480000 | 23300s | train_loss=0.0171
251
+ step 2490000 | 23386s | train_loss=0.0171
252
+ step 2500000 | 23557s | train_loss=0.0169 | fid=57.2594
253
+ step 2510000 | 23644s | train_loss=0.0172
254
+ step 2520000 | 23732s | train_loss=0.0171
255
+ step 2530000 | 23820s | train_loss=0.0168
256
+ step 2540000 | 23896s | train_loss=0.0168
257
+ step 2550000 | 23981s | train_loss=0.0170
258
+ step 2560000 | 24068s | train_loss=0.0165
259
+ step 2570000 | 24156s | train_loss=0.0165
260
+ step 2580000 | 24241s | train_loss=0.0165
261
+ step 2590000 | 24328s | train_loss=0.0163
262
+ step 2600000 | 24517s | train_loss=0.0164 | fid=57.2136
263
+ step 2610000 | 24605s | train_loss=0.0162
264
+ step 2620000 | 24693s | train_loss=0.0165
265
+ step 2630000 | 24780s | train_loss=0.0164
266
+ step 2640000 | 24853s | train_loss=0.0160
267
+ step 2650000 | 24941s | train_loss=0.0161
268
+ step 2660000 | 25028s | train_loss=0.0166
269
+ step 2670000 | 25115s | train_loss=0.0163
270
+ step 2680000 | 25200s | train_loss=0.0158
271
+ step 2690000 | 25288s | train_loss=0.0158
272
+ step 2700000 | 25466s | train_loss=0.0158 | fid=56.3963
273
+ step 2710000 | 25554s | train_loss=0.0160
274
+ step 2720000 | 25642s | train_loss=0.0158
275
+ step 2730000 | 25729s | train_loss=0.0159
276
+ step 2740000 | 25801s | train_loss=0.0159
277
+ step 2750000 | 25889s | train_loss=0.0156
278
+ step 2760000 | 25976s | train_loss=0.0156
279
+ step 2770000 | 26062s | train_loss=0.0161
280
+ step 2780000 | 26148s | train_loss=0.0157
281
+ step 2790000 | 26236s | train_loss=0.0158
282
+ step 2800000 | 26413s | train_loss=0.0154 | fid=57.2994
283
+ step 2810000 | 26500s | train_loss=0.0153
284
+ step 2820000 | 26588s | train_loss=0.0155
285
+ step 2830000 | 26675s | train_loss=0.0155
286
+ step 2840000 | 26751s | train_loss=0.0151
287
+ step 2850000 | 26839s | train_loss=0.0153
288
+ step 2860000 | 26925s | train_loss=0.0152
289
+ step 2870000 | 27013s | train_loss=0.0152
290
+ step 2880000 | 27099s | train_loss=0.0153
291
+ step 2890000 | 27185s | train_loss=0.0152
292
+ step 2900000 | 27360s | train_loss=0.0155 | fid=56.5144
293
+ step 2910000 | 27447s | train_loss=0.0155
294
+ step 2920000 | 27535s | train_loss=0.0152
295
+ step 2930000 | 27618s | train_loss=0.0152
296
+ step 2940000 | 27702s | train_loss=0.0149
297
+ step 2950000 | 27787s | train_loss=0.0146
298
+ step 2960000 | 27875s | train_loss=0.0147
299
+ step 2970000 | 27963s | train_loss=0.0148
300
+ step 2980000 | 28049s | train_loss=0.0151
301
+ step 2990000 | 28132s | train_loss=0.0147
302
+ step 3000000 | 28334s | train_loss=0.0152 | fid=57.1313
303
+ step 3010000 | 28421s | train_loss=0.0149
304
+ step 3020000 | 28509s | train_loss=0.0151
305
+ step 3030000 | 28588s | train_loss=0.0148
306
+ step 3040000 | 28674s | train_loss=0.0145
307
+ step 3050000 | 28762s | train_loss=0.0149
308
+ step 3060000 | 28850s | train_loss=0.0148
309
+ step 3070000 | 28936s | train_loss=0.0150
310
+ step 3080000 | 29023s | train_loss=0.0145
311
+ step 3090000 | 29102s | train_loss=0.0147
312
+ step 3100000 | 29294s | train_loss=0.0147 | fid=57.1800
313
+ step 3110000 | 29381s | train_loss=0.0147
314
+ step 3120000 | 29461s | train_loss=0.0143
315
+ step 3130000 | 29541s | train_loss=0.0145
316
+ step 3140000 | 29629s | train_loss=0.0144
317
+ step 3150000 | 29717s | train_loss=0.0143
318
+ step 3160000 | 29804s | train_loss=0.0144
319
+ step 3170000 | 29891s | train_loss=0.0141
320
+ step 3180000 | 29976s | train_loss=0.0140
321
+ step 3190000 | 30064s | train_loss=0.0139
322
+ step 3200000 | 30236s | train_loss=0.0141 | fid=56.4665
323
+ step 3210000 | 30324s | train_loss=0.0144
324
+ step 3220000 | 30402s | train_loss=0.0140
325
+ step 3230000 | 30482s | train_loss=0.0144
326
+ step 3240000 | 30569s | train_loss=0.0139
327
+ step 3250000 | 30657s | train_loss=0.0139
328
+ step 3260000 | 30745s | train_loss=0.0140
329
+ step 3270000 | 30830s | train_loss=0.0142
330
+ step 3280000 | 30916s | train_loss=0.0136
331
+ step 3290000 | 31004s | train_loss=0.0142
332
+ step 3300000 | 31162s | train_loss=0.0139 | fid=56.3160
333
+ step 3310000 | 31249s | train_loss=0.0141
334
+ step 3320000 | 31334s | train_loss=0.0137
335
+ step 3330000 | 31420s | train_loss=0.0144
336
+ step 3340000 | 31507s | train_loss=0.0134
337
+ step 3350000 | 31595s | train_loss=0.0138
338
+ step 3360000 | 31683s | train_loss=0.0137
339
+ step 3370000 | 31768s | train_loss=0.0139
340
+ step 3380000 | 31855s | train_loss=0.0133
341
+ step 3390000 | 31942s | train_loss=0.0137
342
+ step 3400000 | 32100s | train_loss=0.0141 | fid=56.9810
343
+ step 3410000 | 32186s | train_loss=0.0137
344
+ step 3420000 | 32270s | train_loss=0.0134
345
+ step 3430000 | 32358s | train_loss=0.0134
346
+ step 3440000 | 32446s | train_loss=0.0134
347
+ step 3450000 | 32534s | train_loss=0.0136
348
+ step 3460000 | 32619s | train_loss=0.0137
349
+ step 3470000 | 32696s | train_loss=0.0136
350
+ step 3480000 | 32782s | train_loss=0.0132
351
+ step 3490000 | 32870s | train_loss=0.0131
352
+ step 3500000 | 33069s | train_loss=0.0135 | fid=56.7191
353
+ step 3510000 | 33152s | train_loss=0.0130
354
+ step 3520000 | 33234s | train_loss=0.0134
355
+ step 3530000 | 33322s | train_loss=0.0132
356
+ step 3540000 | 33409s | train_loss=0.0127
357
+ step 3550000 | 33497s | train_loss=0.0132
358
+ step 3560000 | 33574s | train_loss=0.0132
359
+ step 3570000 | 33660s | train_loss=0.0130
360
+ step 3580000 | 33746s | train_loss=0.0129
361
+ step 3590000 | 33834s | train_loss=0.0128
362
+ step 3600000 | 34023s | train_loss=0.0129 | fid=56.8008
363
+ step 3610000 | 34100s | train_loss=0.0133
364
+ step 3620000 | 34186s | train_loss=0.0130
365
+ step 3630000 | 34273s | train_loss=0.0128
366
+ step 3640000 | 34361s | train_loss=0.0125
367
+ step 3650000 | 34449s | train_loss=0.0126
368
+ step 3660000 | 34527s | train_loss=0.0126
369
+ step 3670000 | 34615s | train_loss=0.0128
370
+ step 3680000 | 34701s | train_loss=0.0130
371
+ step 3690000 | 34789s | train_loss=0.0129
372
+ step 3700000 | 34980s | train_loss=0.0126 | fid=57.0308
373
+ step 3710000 | 35057s | train_loss=0.0127
374
+ step 3720000 | 35144s | train_loss=0.0125
375
+ step 3730000 | 35232s | train_loss=0.0123
376
+ step 3740000 | 35320s | train_loss=0.0123
377
+ step 3750000 | 35398s | train_loss=0.0123
378
+ step 3760000 | 35483s | train_loss=0.0126
379
+ step 3770000 | 35571s | train_loss=0.0128
380
+ step 3780000 | 35657s | train_loss=0.0127
381
+ step 3790000 | 35743s | train_loss=0.0125
382
+ step 3800000 | 35929s | train_loss=0.0123 | fid=56.8571
383
+ step 3810000 | 36008s | train_loss=0.0126
384
+ step 3820000 | 36095s | train_loss=0.0122
385
+ step 3830000 | 36183s | train_loss=0.0125
386
+ step 3840000 | 36271s | train_loss=0.0121
387
+ step 3850000 | 36347s | train_loss=0.0121
388
+ step 3860000 | 36435s | train_loss=0.0125
389
+ step 3870000 | 36523s | train_loss=0.0124
390
+ step 3880000 | 36608s | train_loss=0.0123
391
+ step 3890000 | 36695s | train_loss=0.0124
392
+ step 3900000 | 36878s | train_loss=0.0122 | fid=56.9751
393
+ step 3910000 | 36959s | train_loss=0.0124
394
+ step 3920000 | 37047s | train_loss=0.0125
395
+ step 3930000 | 37134s | train_loss=0.0123
396
+ step 3940000 | 37217s | train_loss=0.0121
397
+ step 3950000 | 37305s | train_loss=0.0121
398
+ step 3960000 | 37393s | train_loss=0.0119
399
+ step 3970000 | 37478s | train_loss=0.0125
400
+ step 3980000 | 37565s | train_loss=0.0118
401
+ step 3990000 | 37653s | train_loss=0.0118
402
+ step 4000000 | 37811s | train_loss=0.0120 | fid=56.0406
403
+ step 4010000 | 37898s | train_loss=0.0120
404
+ step 4020000 | 37986s | train_loss=0.0119
405
+ step 4030000 | 38073s | train_loss=0.0122
406
+ step 4040000 | 38158s | train_loss=0.0118
407
+ step 4050000 | 38245s | train_loss=0.0119
408
+ step 4060000 | 38333s | train_loss=0.0117
409
+ step 4070000 | 38419s | train_loss=0.0121
410
+ step 4080000 | 38505s | train_loss=0.0116
411
+ step 4090000 | 38593s | train_loss=0.0120
412
+ step 4100000 | 38745s | train_loss=0.0117 | fid=56.5438
413
+ step 4110000 | 38833s | train_loss=0.0115
414
+ step 4120000 | 38920s | train_loss=0.0119
415
+ step 4130000 | 39007s | train_loss=0.0118
416
+ step 4140000 | 39091s | train_loss=0.0119
417
+ step 4150000 | 39178s | train_loss=0.0121
418
+ step 4160000 | 39264s | train_loss=0.0116
419
+ step 4170000 | 39352s | train_loss=0.0118
420
+ step 4180000 | 39438s | train_loss=0.0114
421
+ step 4190000 | 39526s | train_loss=0.0118
422
+ step 4200000 | 39681s | train_loss=0.0113 | fid=56.2503
423
+ step 4210000 | 39768s | train_loss=0.0115
424
+ step 4220000 | 39856s | train_loss=0.0117
425
+ step 4230000 | 39941s | train_loss=0.0115
426
+ step 4240000 | 40029s | train_loss=0.0116
427
+ step 4250000 | 40116s | train_loss=0.0116
428
+ step 4260000 | 40202s | train_loss=0.0116
429
+ step 4270000 | 40290s | train_loss=0.0115
430
+ step 4280000 | 40378s | train_loss=0.0111
431
+ step 4290000 | 40464s | train_loss=0.0113
432
+ step 4300000 | 40657s | train_loss=0.0116 | fid=56.1360
433
+ step 4310000 | 40745s | train_loss=0.0114
434
+ step 4320000 | 40825s | train_loss=0.0114
435
+ step 4330000 | 40912s | train_loss=0.0110
436
+ step 4340000 | 41000s | train_loss=0.0114
437
+ step 4350000 | 41086s | train_loss=0.0118
438
+ step 4360000 | 41173s | train_loss=0.0111
439
+ step 4370000 | 41261s | train_loss=0.0113
440
+ step 4380000 | 41348s | train_loss=0.0116
441
+ step 4390000 | 41435s | train_loss=0.0110
442
+ step 4400000 | 41633s | train_loss=0.0113 | fid=56.4201
443
+ step 4410000 | 41713s | train_loss=0.0110
444
+ step 4420000 | 41798s | train_loss=0.0115
445
+ step 4430000 | 41884s | train_loss=0.0112
446
+ step 4440000 | 41972s | train_loss=0.0110
447
+ step 4450000 | 42059s | train_loss=0.0112
448
+ step 4460000 | 42147s | train_loss=0.0111
449
+ step 4470000 | 42235s | train_loss=0.0113
450
+ step 4480000 | 42321s | train_loss=0.0112
451
+ step 4490000 | 42407s | train_loss=0.0111
452
+ step 4500000 | 42585s | train_loss=0.0109 | fid=56.4538
453
+ step 4510000 | 42658s | train_loss=0.0109
454
+ step 4520000 | 42745s | train_loss=0.0109
455
+ step 4530000 | 42831s | train_loss=0.0109
456
+ step 4540000 | 42919s | train_loss=0.0106
457
+ step 4550000 | 43006s | train_loss=0.0110
458
+ step 4560000 | 43094s | train_loss=0.0112
459
+ step 4570000 | 43182s | train_loss=0.0108
460
+ step 4580000 | 43268s | train_loss=0.0109
461
+ step 4590000 | 43352s | train_loss=0.0106
462
+ step 4600000 | 43532s | train_loss=0.0108 | fid=56.8092
463
+ step 4610000 | 43607s | train_loss=0.0106
464
+ step 4620000 | 43693s | train_loss=0.0106
465
+ step 4630000 | 43781s | train_loss=0.0105
466
+ step 4640000 | 43869s | train_loss=0.0108
467
+ step 4650000 | 43956s | train_loss=0.0107
468
+ step 4660000 | 44044s | train_loss=0.0103
469
+ step 4670000 | 44132s | train_loss=0.0109
470
+ step 4680000 | 44218s | train_loss=0.0109
471
+ step 4690000 | 44298s | train_loss=0.0108
472
+ step 4700000 | 44477s | train_loss=0.0106 | fid=56.2313
473
+ step 4710000 | 44547s | train_loss=0.0106
474
+ step 4720000 | 44617s | train_loss=0.0107
475
+ step 4730000 | 44688s | train_loss=0.0105
476
+ step 4740000 | 44758s | train_loss=0.0108
477
+ step 4750000 | 44829s | train_loss=0.0103
478
+ step 4760000 | 44899s | train_loss=0.0103
479
+ step 4770000 | 44970s | train_loss=0.0105
480
+ step 4780000 | 45037s | train_loss=0.0106
481
+ step 4790000 | 45096s | train_loss=0.0103
482
+ step 4800000 | 45223s | train_loss=0.0104 | fid=56.9091
483
+ step 4810000 | 45276s | train_loss=0.0104
484
+ step 4820000 | 45329s | train_loss=0.0105
485
+ step 4830000 | 45383s | train_loss=0.0102
486
+ step 4840000 | 45436s | train_loss=0.0108
487
+ step 4850000 | 45489s | train_loss=0.0100
488
+ step 4860000 | 45543s | train_loss=0.0101
489
+ step 4870000 | 45596s | train_loss=0.0103
490
+ step 4880000 | 45645s | train_loss=0.0102
491
+ step 4890000 | 45688s | train_loss=0.0100
492
+ step 4900000 | 45756s | train_loss=0.0102 | fid=56.5227
493
+ step 4910000 | 45793s | train_loss=0.0110
494
+ step 4920000 | 45829s | train_loss=0.0103
495
+ step 4930000 | 45865s | train_loss=0.0103
496
+ step 4940000 | 45901s | train_loss=0.0100
497
+ step 4950000 | 45937s | train_loss=0.0101
498
+ step 4960000 | 45973s | train_loss=0.0103
499
+ step 4970000 | 46009s | train_loss=0.0101
500
+ step 4980000 | 46041s | train_loss=0.0106
501
+ step 4990000 | 46077s | train_loss=0.0101
502
+ step 5000000 | 46145s | train_loss=0.0098 | fid=57.4873
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/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/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": 42,
25
+ "device": null
26
+ }
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/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/d40/runner.log ADDED
@@ -0,0 +1,502 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ FID real-side stats cached over 1000 diverse train images
2
+ step 1 | 66s | train_loss=0.9905 | fid=276.9747
3
+ step 10000 | 149s | train_loss=0.8507
4
+ step 20000 | 234s | train_loss=0.8227
5
+ step 30000 | 318s | train_loss=0.7673
6
+ step 40000 | 403s | train_loss=0.6815
7
+ step 50000 | 487s | train_loss=0.5504
8
+ step 60000 | 572s | train_loss=0.4021
9
+ step 70000 | 657s | train_loss=0.2963
10
+ step 80000 | 741s | train_loss=0.2316
11
+ step 90000 | 826s | train_loss=0.1958
12
+ step 100000 | 975s | train_loss=0.1728 | fid=59.5879
13
+ step 110000 | 1054s | train_loss=0.1560
14
+ step 120000 | 1138s | train_loss=0.1435
15
+ step 130000 | 1222s | train_loss=0.1332
16
+ step 140000 | 1307s | train_loss=0.1237
17
+ step 150000 | 1391s | train_loss=0.1167
18
+ step 160000 | 1475s | train_loss=0.1123
19
+ step 170000 | 1560s | train_loss=0.1053
20
+ step 180000 | 1644s | train_loss=0.1014
21
+ step 190000 | 1728s | train_loss=0.0970
22
+ step 200000 | 1876s | train_loss=0.0931 | fid=58.1435
23
+ step 210000 | 1955s | train_loss=0.0890
24
+ step 220000 | 2038s | train_loss=0.0865
25
+ step 230000 | 2123s | train_loss=0.0839
26
+ step 240000 | 2207s | train_loss=0.0806
27
+ step 250000 | 2292s | train_loss=0.0789
28
+ step 260000 | 2376s | train_loss=0.0761
29
+ step 270000 | 2460s | train_loss=0.0738
30
+ step 280000 | 2545s | train_loss=0.0728
31
+ step 290000 | 2629s | train_loss=0.0699
32
+ step 300000 | 2777s | train_loss=0.0691 | fid=58.4042
33
+ step 310000 | 2858s | train_loss=0.0672
34
+ step 320000 | 2940s | train_loss=0.0652
35
+ step 330000 | 3023s | train_loss=0.0644
36
+ step 340000 | 3107s | train_loss=0.0619
37
+ step 350000 | 3192s | train_loss=0.0612
38
+ step 360000 | 3276s | train_loss=0.0612
39
+ step 370000 | 3361s | train_loss=0.0590
40
+ step 380000 | 3446s | train_loss=0.0579
41
+ step 390000 | 3530s | train_loss=0.0567
42
+ step 400000 | 3677s | train_loss=0.0545 | fid=56.0015
43
+ step 410000 | 3759s | train_loss=0.0544
44
+ step 420000 | 3840s | train_loss=0.0533
45
+ step 430000 | 3925s | train_loss=0.0528
46
+ step 440000 | 4008s | train_loss=0.0523
47
+ step 450000 | 4092s | train_loss=0.0509
48
+ step 460000 | 4177s | train_loss=0.0500
49
+ step 470000 | 4261s | train_loss=0.0493
50
+ step 480000 | 4346s | train_loss=0.0494
51
+ step 490000 | 4430s | train_loss=0.0479
52
+ step 500000 | 4579s | train_loss=0.0475 | fid=55.8934
53
+ step 510000 | 4660s | train_loss=0.0470
54
+ step 520000 | 4744s | train_loss=0.0461
55
+ step 530000 | 4827s | train_loss=0.0459
56
+ step 540000 | 4912s | train_loss=0.0450
57
+ step 550000 | 4995s | train_loss=0.0449
58
+ step 560000 | 5079s | train_loss=0.0441
59
+ step 570000 | 5164s | train_loss=0.0442
60
+ step 580000 | 5248s | train_loss=0.0435
61
+ step 590000 | 5333s | train_loss=0.0432
62
+ step 600000 | 5481s | train_loss=0.0417 | fid=54.6378
63
+ step 610000 | 5563s | train_loss=0.0421
64
+ step 620000 | 5647s | train_loss=0.0415
65
+ step 630000 | 5729s | train_loss=0.0408
66
+ step 640000 | 5814s | train_loss=0.0410
67
+ step 650000 | 5899s | train_loss=0.0406
68
+ step 660000 | 5981s | train_loss=0.0395
69
+ step 670000 | 6066s | train_loss=0.0400
70
+ step 680000 | 6150s | train_loss=0.0389
71
+ step 690000 | 6235s | train_loss=0.0385
72
+ step 700000 | 6383s | train_loss=0.0379 | fid=54.6835
73
+ step 710000 | 6465s | train_loss=0.0380
74
+ step 720000 | 6550s | train_loss=0.0375
75
+ step 730000 | 6631s | train_loss=0.0371
76
+ step 740000 | 6716s | train_loss=0.0372
77
+ step 750000 | 6800s | train_loss=0.0366
78
+ step 760000 | 6884s | train_loss=0.0377
79
+ step 770000 | 6968s | train_loss=0.0365
80
+ step 780000 | 7052s | train_loss=0.0354
81
+ step 790000 | 7137s | train_loss=0.0357
82
+ step 800000 | 7283s | train_loss=0.0358 | fid=53.1323
83
+ step 810000 | 7365s | train_loss=0.0349
84
+ step 820000 | 7450s | train_loss=0.0345
85
+ step 830000 | 7534s | train_loss=0.0350
86
+ step 840000 | 7616s | train_loss=0.0337
87
+ step 850000 | 7700s | train_loss=0.0339
88
+ step 860000 | 7785s | train_loss=0.0341
89
+ step 870000 | 7868s | train_loss=0.0342
90
+ step 880000 | 7952s | train_loss=0.0333
91
+ step 890000 | 8037s | train_loss=0.0328
92
+ step 900000 | 8184s | train_loss=0.0330 | fid=52.5083
93
+ step 910000 | 8266s | train_loss=0.0322
94
+ step 920000 | 8350s | train_loss=0.0317
95
+ step 930000 | 8435s | train_loss=0.0322
96
+ step 940000 | 8517s | train_loss=0.0317
97
+ step 950000 | 8601s | train_loss=0.0317
98
+ step 960000 | 8686s | train_loss=0.0312
99
+ step 970000 | 8770s | train_loss=0.0315
100
+ step 980000 | 8853s | train_loss=0.0308
101
+ step 990000 | 8938s | train_loss=0.0311
102
+ step 1000000 | 9084s | train_loss=0.0310 | fid=52.2989
103
+ step 1010000 | 9167s | train_loss=0.0304
104
+ step 1020000 | 9250s | train_loss=0.0303
105
+ step 1030000 | 9335s | train_loss=0.0305
106
+ step 1040000 | 9419s | train_loss=0.0296
107
+ step 1050000 | 9501s | train_loss=0.0300
108
+ step 1060000 | 9586s | train_loss=0.0297
109
+ step 1070000 | 9670s | train_loss=0.0295
110
+ step 1080000 | 9755s | train_loss=0.0292
111
+ step 1090000 | 9837s | train_loss=0.0290
112
+ step 1100000 | 9985s | train_loss=0.0292 | fid=52.2728
113
+ step 1110000 | 10069s | train_loss=0.0286
114
+ step 1120000 | 10151s | train_loss=0.0284
115
+ step 1130000 | 10235s | train_loss=0.0279
116
+ step 1140000 | 10320s | train_loss=0.0281
117
+ step 1150000 | 10401s | train_loss=0.0277
118
+ step 1160000 | 10486s | train_loss=0.0276
119
+ step 1170000 | 10570s | train_loss=0.0272
120
+ step 1180000 | 10655s | train_loss=0.0272
121
+ step 1190000 | 10740s | train_loss=0.0275
122
+ step 1200000 | 10886s | train_loss=0.0270 | fid=52.2834
123
+ step 1210000 | 10970s | train_loss=0.0272
124
+ step 1220000 | 11052s | train_loss=0.0270
125
+ step 1230000 | 11136s | train_loss=0.0266
126
+ step 1240000 | 11221s | train_loss=0.0266
127
+ step 1250000 | 11305s | train_loss=0.0268
128
+ step 1260000 | 11387s | train_loss=0.0260
129
+ step 1270000 | 11472s | train_loss=0.0255
130
+ step 1280000 | 11556s | train_loss=0.0262
131
+ step 1290000 | 11641s | train_loss=0.0259
132
+ step 1300000 | 11786s | train_loss=0.0265 | fid=52.5566
133
+ step 1310000 | 11871s | train_loss=0.0254
134
+ step 1320000 | 11953s | train_loss=0.0254
135
+ step 1330000 | 12037s | train_loss=0.0252
136
+ step 1340000 | 12122s | train_loss=0.0251
137
+ step 1350000 | 12206s | train_loss=0.0250
138
+ step 1360000 | 12288s | train_loss=0.0256
139
+ step 1370000 | 12373s | train_loss=0.0243
140
+ step 1380000 | 12457s | train_loss=0.0247
141
+ step 1390000 | 12542s | train_loss=0.0245
142
+ step 1400000 | 12690s | train_loss=0.0237 | fid=52.6737
143
+ step 1410000 | 12772s | train_loss=0.0241
144
+ step 1420000 | 12855s | train_loss=0.0241
145
+ step 1430000 | 12939s | train_loss=0.0240
146
+ step 1440000 | 13024s | train_loss=0.0241
147
+ step 1450000 | 13108s | train_loss=0.0244
148
+ step 1460000 | 13190s | train_loss=0.0237
149
+ step 1470000 | 13274s | train_loss=0.0233
150
+ step 1480000 | 13358s | train_loss=0.0233
151
+ step 1490000 | 13443s | train_loss=0.0235
152
+ step 1500000 | 13590s | train_loss=0.0231 | fid=52.4340
153
+ step 1510000 | 13675s | train_loss=0.0235
154
+ step 1520000 | 13755s | train_loss=0.0238
155
+ step 1530000 | 13840s | train_loss=0.0232
156
+ step 1540000 | 13924s | train_loss=0.0235
157
+ step 1550000 | 14008s | train_loss=0.0231
158
+ step 1560000 | 14093s | train_loss=0.0234
159
+ step 1570000 | 14174s | train_loss=0.0226
160
+ step 1580000 | 14258s | train_loss=0.0228
161
+ step 1590000 | 14343s | train_loss=0.0221
162
+ step 1600000 | 14487s | train_loss=0.0218 | fid=52.1806
163
+ step 1610000 | 14566s | train_loss=0.0226
164
+ step 1620000 | 14642s | train_loss=0.0220
165
+ step 1630000 | 14718s | train_loss=0.0217
166
+ step 1640000 | 14802s | train_loss=0.0216
167
+ step 1650000 | 14887s | train_loss=0.0221
168
+ step 1660000 | 14971s | train_loss=0.0211
169
+ step 1670000 | 15050s | train_loss=0.0217
170
+ step 1680000 | 15132s | train_loss=0.0214
171
+ step 1690000 | 15216s | train_loss=0.0213
172
+ step 1700000 | 15367s | train_loss=0.0215 | fid=52.3332
173
+ step 1710000 | 15448s | train_loss=0.0221
174
+ step 1720000 | 15533s | train_loss=0.0213
175
+ step 1730000 | 15612s | train_loss=0.0210
176
+ step 1740000 | 15697s | train_loss=0.0213
177
+ step 1750000 | 15781s | train_loss=0.0213
178
+ step 1760000 | 15866s | train_loss=0.0214
179
+ step 1770000 | 15950s | train_loss=0.0209
180
+ step 1780000 | 16032s | train_loss=0.0209
181
+ step 1790000 | 16116s | train_loss=0.0210
182
+ step 1800000 | 16266s | train_loss=0.0209 | fid=52.5116
183
+ step 1810000 | 16349s | train_loss=0.0202
184
+ step 1820000 | 16433s | train_loss=0.0206
185
+ step 1830000 | 16514s | train_loss=0.0202
186
+ step 1840000 | 16597s | train_loss=0.0207
187
+ step 1850000 | 16682s | train_loss=0.0201
188
+ step 1860000 | 16766s | train_loss=0.0199
189
+ step 1870000 | 16851s | train_loss=0.0201
190
+ step 1880000 | 16933s | train_loss=0.0199
191
+ step 1890000 | 17016s | train_loss=0.0200
192
+ step 1900000 | 17166s | train_loss=0.0196 | fid=51.7250
193
+ step 1910000 | 17248s | train_loss=0.0195
194
+ step 1920000 | 17332s | train_loss=0.0195
195
+ step 1930000 | 17413s | train_loss=0.0190
196
+ step 1940000 | 17498s | train_loss=0.0201
197
+ step 1950000 | 17581s | train_loss=0.0197
198
+ step 1960000 | 17665s | train_loss=0.0196
199
+ step 1970000 | 17750s | train_loss=0.0199
200
+ step 1980000 | 17834s | train_loss=0.0195
201
+ step 1990000 | 17916s | train_loss=0.0189
202
+ step 2000000 | 18065s | train_loss=0.0190 | fid=51.8618
203
+ step 2010000 | 18146s | train_loss=0.0186
204
+ step 2020000 | 18231s | train_loss=0.0193
205
+ step 2030000 | 18312s | train_loss=0.0189
206
+ step 2040000 | 18397s | train_loss=0.0197
207
+ step 2050000 | 18481s | train_loss=0.0192
208
+ step 2060000 | 18564s | train_loss=0.0189
209
+ step 2070000 | 18649s | train_loss=0.0186
210
+ step 2080000 | 18733s | train_loss=0.0184
211
+ step 2090000 | 18815s | train_loss=0.0184
212
+ step 2100000 | 18965s | train_loss=0.0187 | fid=51.7487
213
+ step 2110000 | 19047s | train_loss=0.0180
214
+ step 2120000 | 19131s | train_loss=0.0185
215
+ step 2130000 | 19213s | train_loss=0.0181
216
+ step 2140000 | 19298s | train_loss=0.0181
217
+ step 2150000 | 19382s | train_loss=0.0178
218
+ step 2160000 | 19467s | train_loss=0.0175
219
+ step 2170000 | 19550s | train_loss=0.0175
220
+ step 2180000 | 19634s | train_loss=0.0179
221
+ step 2190000 | 19719s | train_loss=0.0189
222
+ step 2200000 | 19858s | train_loss=0.0177 | fid=51.1069
223
+ step 2210000 | 19931s | train_loss=0.0173
224
+ step 2220000 | 20016s | train_loss=0.0177
225
+ step 2230000 | 20099s | train_loss=0.0177
226
+ step 2240000 | 20173s | train_loss=0.0172
227
+ step 2250000 | 20258s | train_loss=0.0174
228
+ step 2260000 | 20342s | train_loss=0.0178
229
+ step 2270000 | 20427s | train_loss=0.0169
230
+ step 2280000 | 20510s | train_loss=0.0171
231
+ step 2290000 | 20595s | train_loss=0.0173
232
+ step 2300000 | 20736s | train_loss=0.0172 | fid=52.1871
233
+ step 2310000 | 20814s | train_loss=0.0170
234
+ step 2320000 | 20896s | train_loss=0.0176
235
+ step 2330000 | 20980s | train_loss=0.0169
236
+ step 2340000 | 21057s | train_loss=0.0169
237
+ step 2350000 | 21136s | train_loss=0.0169
238
+ step 2360000 | 21221s | train_loss=0.0171
239
+ step 2370000 | 21305s | train_loss=0.0167
240
+ step 2380000 | 21390s | train_loss=0.0165
241
+ step 2390000 | 21473s | train_loss=0.0168
242
+ step 2400000 | 21623s | train_loss=0.0160 | fid=51.8897
243
+ step 2410000 | 21699s | train_loss=0.0168
244
+ step 2420000 | 21770s | train_loss=0.0163
245
+ step 2430000 | 21854s | train_loss=0.0166
246
+ step 2440000 | 21939s | train_loss=0.0162
247
+ step 2450000 | 22011s | train_loss=0.0165
248
+ step 2460000 | 22091s | train_loss=0.0164
249
+ step 2470000 | 22176s | train_loss=0.0165
250
+ step 2480000 | 22260s | train_loss=0.0168
251
+ step 2490000 | 22343s | train_loss=0.0165
252
+ step 2500000 | 22493s | train_loss=0.0166 | fid=52.2889
253
+ step 2510000 | 22577s | train_loss=0.0158
254
+ step 2520000 | 22643s | train_loss=0.0154
255
+ step 2530000 | 22727s | train_loss=0.0157
256
+ step 2540000 | 22812s | train_loss=0.0161
257
+ step 2550000 | 22896s | train_loss=0.0159
258
+ step 2560000 | 22967s | train_loss=0.0154
259
+ step 2570000 | 23051s | train_loss=0.0152
260
+ step 2580000 | 23136s | train_loss=0.0155
261
+ step 2590000 | 23220s | train_loss=0.0168
262
+ step 2600000 | 23369s | train_loss=0.0152 | fid=51.4974
263
+ step 2610000 | 23454s | train_loss=0.0157
264
+ step 2620000 | 23534s | train_loss=0.0151
265
+ step 2630000 | 23604s | train_loss=0.0150
266
+ step 2640000 | 23688s | train_loss=0.0157
267
+ step 2650000 | 23773s | train_loss=0.0160
268
+ step 2660000 | 23856s | train_loss=0.0156
269
+ step 2670000 | 23927s | train_loss=0.0155
270
+ step 2680000 | 24012s | train_loss=0.0146
271
+ step 2690000 | 24096s | train_loss=0.0151
272
+ step 2700000 | 24245s | train_loss=0.0147 | fid=51.0738
273
+ step 2710000 | 24329s | train_loss=0.0147
274
+ step 2720000 | 24414s | train_loss=0.0145
275
+ step 2730000 | 24482s | train_loss=0.0147
276
+ step 2740000 | 24558s | train_loss=0.0145
277
+ step 2750000 | 24643s | train_loss=0.0143
278
+ step 2760000 | 24727s | train_loss=0.0149
279
+ step 2770000 | 24807s | train_loss=0.0155
280
+ step 2780000 | 24881s | train_loss=0.0145
281
+ step 2790000 | 24965s | train_loss=0.0143
282
+ step 2800000 | 25116s | train_loss=0.0143 | fid=52.6279
283
+ step 2810000 | 25199s | train_loss=0.0146
284
+ step 2820000 | 25283s | train_loss=0.0147
285
+ step 2830000 | 25363s | train_loss=0.0145
286
+ step 2840000 | 25439s | train_loss=0.0141
287
+ step 2850000 | 25516s | train_loss=0.0141
288
+ step 2860000 | 25601s | train_loss=0.0143
289
+ step 2870000 | 25685s | train_loss=0.0144
290
+ step 2880000 | 25762s | train_loss=0.0146
291
+ step 2890000 | 25838s | train_loss=0.0144
292
+ step 2900000 | 25989s | train_loss=0.0140 | fid=51.8458
293
+ step 2910000 | 26073s | train_loss=0.0141
294
+ step 2920000 | 26156s | train_loss=0.0137
295
+ step 2930000 | 26240s | train_loss=0.0141
296
+ step 2940000 | 26311s | train_loss=0.0137
297
+ step 2950000 | 26390s | train_loss=0.0134
298
+ step 2960000 | 26468s | train_loss=0.0137
299
+ step 2970000 | 26553s | train_loss=0.0141
300
+ step 2980000 | 26637s | train_loss=0.0135
301
+ step 2990000 | 26712s | train_loss=0.0144
302
+ step 3000000 | 26860s | train_loss=0.0136 | fid=52.0663
303
+ step 3010000 | 26944s | train_loss=0.0138
304
+ step 3020000 | 27029s | train_loss=0.0134
305
+ step 3030000 | 27111s | train_loss=0.0135
306
+ step 3040000 | 27193s | train_loss=0.0139
307
+ step 3050000 | 27266s | train_loss=0.0136
308
+ step 3060000 | 27344s | train_loss=0.0144
309
+ step 3070000 | 27424s | train_loss=0.0135
310
+ step 3080000 | 27508s | train_loss=0.0133
311
+ step 3090000 | 27593s | train_loss=0.0138
312
+ step 3100000 | 27734s | train_loss=0.0135 | fid=52.3483
313
+ step 3110000 | 27819s | train_loss=0.0131
314
+ step 3120000 | 27903s | train_loss=0.0137
315
+ step 3130000 | 27988s | train_loss=0.0135
316
+ step 3140000 | 28070s | train_loss=0.0130
317
+ step 3150000 | 28146s | train_loss=0.0132
318
+ step 3160000 | 28229s | train_loss=0.0140
319
+ step 3170000 | 28305s | train_loss=0.0127
320
+ step 3180000 | 28383s | train_loss=0.0126
321
+ step 3190000 | 28467s | train_loss=0.0129
322
+ step 3200000 | 28609s | train_loss=0.0129 | fid=51.2528
323
+ step 3210000 | 28693s | train_loss=0.0129
324
+ step 3220000 | 28778s | train_loss=0.0126
325
+ step 3230000 | 28862s | train_loss=0.0124
326
+ step 3240000 | 28945s | train_loss=0.0133
327
+ step 3250000 | 29029s | train_loss=0.0128
328
+ step 3260000 | 29105s | train_loss=0.0131
329
+ step 3270000 | 29190s | train_loss=0.0126
330
+ step 3280000 | 29269s | train_loss=0.0125
331
+ step 3290000 | 29347s | train_loss=0.0128
332
+ step 3300000 | 29485s | train_loss=0.0126 | fid=51.7602
333
+ step 3310000 | 29570s | train_loss=0.0129
334
+ step 3320000 | 29654s | train_loss=0.0132
335
+ step 3330000 | 29739s | train_loss=0.0123
336
+ step 3340000 | 29823s | train_loss=0.0120
337
+ step 3350000 | 29906s | train_loss=0.0128
338
+ step 3360000 | 29988s | train_loss=0.0128
339
+ step 3370000 | 30073s | train_loss=0.0131
340
+ step 3380000 | 30157s | train_loss=0.0117
341
+ step 3390000 | 30235s | train_loss=0.0123
342
+ step 3400000 | 30385s | train_loss=0.0122 | fid=51.7564
343
+ step 3410000 | 30457s | train_loss=0.0120
344
+ step 3420000 | 30542s | train_loss=0.0119
345
+ step 3430000 | 30626s | train_loss=0.0122
346
+ step 3440000 | 30711s | train_loss=0.0124
347
+ step 3450000 | 30795s | train_loss=0.0120
348
+ step 3460000 | 30876s | train_loss=0.0122
349
+ step 3470000 | 30960s | train_loss=0.0122
350
+ step 3480000 | 31045s | train_loss=0.0119
351
+ step 3490000 | 31129s | train_loss=0.0127
352
+ step 3500000 | 31277s | train_loss=0.0130 | fid=51.6045
353
+ step 3510000 | 31358s | train_loss=0.0122
354
+ step 3520000 | 31442s | train_loss=0.0117
355
+ step 3530000 | 31527s | train_loss=0.0124
356
+ step 3540000 | 31611s | train_loss=0.0119
357
+ step 3550000 | 31696s | train_loss=0.0114
358
+ step 3560000 | 31778s | train_loss=0.0118
359
+ step 3570000 | 31861s | train_loss=0.0115
360
+ step 3580000 | 31946s | train_loss=0.0114
361
+ step 3590000 | 32030s | train_loss=0.0113
362
+ step 3600000 | 32177s | train_loss=0.0119 | fid=51.8009
363
+ step 3610000 | 32259s | train_loss=0.0114
364
+ step 3620000 | 32343s | train_loss=0.0115
365
+ step 3630000 | 32428s | train_loss=0.0116
366
+ step 3640000 | 32512s | train_loss=0.0113
367
+ step 3650000 | 32597s | train_loss=0.0116
368
+ step 3660000 | 32668s | train_loss=0.0112
369
+ step 3670000 | 32753s | train_loss=0.0115
370
+ step 3680000 | 32835s | train_loss=0.0117
371
+ step 3690000 | 32920s | train_loss=0.0110
372
+ step 3700000 | 33059s | train_loss=0.0111 | fid=52.4919
373
+ step 3710000 | 33139s | train_loss=0.0111
374
+ step 3720000 | 33215s | train_loss=0.0112
375
+ step 3730000 | 33300s | train_loss=0.0115
376
+ step 3740000 | 33384s | train_loss=0.0111
377
+ step 3750000 | 33469s | train_loss=0.0112
378
+ step 3760000 | 33549s | train_loss=0.0120
379
+ step 3770000 | 33625s | train_loss=0.0116
380
+ step 3780000 | 33708s | train_loss=0.0111
381
+ step 3790000 | 33792s | train_loss=0.0110
382
+ step 3800000 | 33942s | train_loss=0.0108 | fid=51.3082
383
+ step 3810000 | 34016s | train_loss=0.0113
384
+ step 3820000 | 34090s | train_loss=0.0114
385
+ step 3830000 | 34170s | train_loss=0.0112
386
+ step 3840000 | 34255s | train_loss=0.0108
387
+ step 3850000 | 34339s | train_loss=0.0112
388
+ step 3860000 | 34424s | train_loss=0.0110
389
+ step 3870000 | 34499s | train_loss=0.0107
390
+ step 3880000 | 34583s | train_loss=0.0105
391
+ step 3890000 | 34667s | train_loss=0.0113
392
+ step 3900000 | 34817s | train_loss=0.0111 | fid=51.3211
393
+ step 3910000 | 34902s | train_loss=0.0110
394
+ step 3920000 | 34975s | train_loss=0.0108
395
+ step 3930000 | 35048s | train_loss=0.0112
396
+ step 3940000 | 35132s | train_loss=0.0107
397
+ step 3950000 | 35217s | train_loss=0.0111
398
+ step 3960000 | 35301s | train_loss=0.0108
399
+ step 3970000 | 35380s | train_loss=0.0109
400
+ step 3980000 | 35457s | train_loss=0.0111
401
+ step 3990000 | 35542s | train_loss=0.0103
402
+ step 4000000 | 35691s | train_loss=0.0105 | fid=51.1100
403
+ step 4010000 | 35775s | train_loss=0.0106
404
+ step 4020000 | 35860s | train_loss=0.0103
405
+ step 4030000 | 35929s | train_loss=0.0107
406
+ step 4040000 | 36005s | train_loss=0.0108
407
+ step 4050000 | 36089s | train_loss=0.0105
408
+ step 4060000 | 36174s | train_loss=0.0107
409
+ step 4070000 | 36258s | train_loss=0.0105
410
+ step 4080000 | 36332s | train_loss=0.0103
411
+ step 4090000 | 36416s | train_loss=0.0102
412
+ step 4100000 | 36564s | train_loss=0.0108 | fid=51.4610
413
+ step 4110000 | 36648s | train_loss=0.0102
414
+ step 4120000 | 36733s | train_loss=0.0102
415
+ step 4130000 | 36817s | train_loss=0.0100
416
+ step 4140000 | 36882s | train_loss=0.0101
417
+ step 4150000 | 36961s | train_loss=0.0104
418
+ step 4160000 | 37045s | train_loss=0.0104
419
+ step 4170000 | 37130s | train_loss=0.0106
420
+ step 4180000 | 37210s | train_loss=0.0101
421
+ step 4190000 | 37294s | train_loss=0.0099
422
+ step 4200000 | 37444s | train_loss=0.0100 | fid=52.0313
423
+ step 4210000 | 37527s | train_loss=0.0101
424
+ step 4220000 | 37612s | train_loss=0.0107
425
+ step 4230000 | 37696s | train_loss=0.0101
426
+ step 4240000 | 37781s | train_loss=0.0106
427
+ step 4250000 | 37858s | train_loss=0.0104
428
+ step 4260000 | 37943s | train_loss=0.0100
429
+ step 4270000 | 38027s | train_loss=0.0102
430
+ step 4280000 | 38109s | train_loss=0.0103
431
+ step 4290000 | 38193s | train_loss=0.0102
432
+ step 4300000 | 38343s | train_loss=0.0098 | fid=51.4589
433
+ step 4310000 | 38428s | train_loss=0.0109
434
+ step 4320000 | 38510s | train_loss=0.0098
435
+ step 4330000 | 38595s | train_loss=0.0096
436
+ step 4340000 | 38679s | train_loss=0.0097
437
+ step 4350000 | 38758s | train_loss=0.0102
438
+ step 4360000 | 38842s | train_loss=0.0102
439
+ step 4370000 | 38927s | train_loss=0.0097
440
+ step 4380000 | 39009s | train_loss=0.0100
441
+ step 4390000 | 39091s | train_loss=0.0098
442
+ step 4400000 | 39241s | train_loss=0.0099 | fid=52.2776
443
+ step 4410000 | 39325s | train_loss=0.0096
444
+ step 4420000 | 39410s | train_loss=0.0094
445
+ step 4430000 | 39493s | train_loss=0.0104
446
+ step 4440000 | 39577s | train_loss=0.0096
447
+ step 4450000 | 39659s | train_loss=0.0102
448
+ step 4460000 | 39740s | train_loss=0.0098
449
+ step 4470000 | 39825s | train_loss=0.0097
450
+ step 4480000 | 39909s | train_loss=0.0097
451
+ step 4490000 | 39991s | train_loss=0.0092
452
+ step 4500000 | 40141s | train_loss=0.0092 | fid=51.5826
453
+ step 4510000 | 40225s | train_loss=0.0101
454
+ step 4520000 | 40310s | train_loss=0.0099
455
+ step 4530000 | 40394s | train_loss=0.0091
456
+ step 4540000 | 40477s | train_loss=0.0094
457
+ step 4550000 | 40556s | train_loss=0.0097
458
+ step 4560000 | 40624s | train_loss=0.0100
459
+ step 4570000 | 40700s | train_loss=0.0094
460
+ step 4580000 | 40785s | train_loss=0.0092
461
+ step 4590000 | 40861s | train_loss=0.0093
462
+ step 4600000 | 41011s | train_loss=0.0095 | fid=52.7279
463
+ step 4610000 | 41096s | train_loss=0.0095
464
+ step 4620000 | 41180s | train_loss=0.0094
465
+ step 4630000 | 41265s | train_loss=0.0097
466
+ step 4640000 | 41348s | train_loss=0.0100
467
+ step 4650000 | 41432s | train_loss=0.0096
468
+ step 4660000 | 41509s | train_loss=0.0093
469
+ step 4670000 | 41586s | train_loss=0.0092
470
+ step 4680000 | 41659s | train_loss=0.0090
471
+ step 4690000 | 41733s | train_loss=0.0091
472
+ step 4700000 | 41884s | train_loss=0.0089 | fid=51.1982
473
+ step 4710000 | 41968s | train_loss=0.0089
474
+ step 4720000 | 42053s | train_loss=0.0089
475
+ step 4730000 | 42137s | train_loss=0.0090
476
+ step 4740000 | 42222s | train_loss=0.0091
477
+ step 4750000 | 42305s | train_loss=0.0091
478
+ step 4760000 | 42389s | train_loss=0.0089
479
+ step 4770000 | 42459s | train_loss=0.0095
480
+ step 4780000 | 42543s | train_loss=0.0091
481
+ step 4790000 | 42612s | train_loss=0.0091
482
+ step 4800000 | 42751s | train_loss=0.0092 | fid=52.3092
483
+ step 4810000 | 42836s | train_loss=0.0091
484
+ step 4820000 | 42920s | train_loss=0.0092
485
+ step 4830000 | 43005s | train_loss=0.0094
486
+ step 4840000 | 43089s | train_loss=0.0091
487
+ step 4850000 | 43174s | train_loss=0.0090
488
+ step 4860000 | 43257s | train_loss=0.0095
489
+ step 4870000 | 43340s | train_loss=0.0087
490
+ step 4880000 | 43412s | train_loss=0.0090
491
+ step 4890000 | 43494s | train_loss=0.0086
492
+ step 4900000 | 43620s | train_loss=0.0087 | fid=52.0171
493
+ step 4910000 | 43704s | train_loss=0.0093
494
+ step 4920000 | 43789s | train_loss=0.0088
495
+ step 4930000 | 43873s | train_loss=0.0088
496
+ step 4940000 | 43958s | train_loss=0.0087
497
+ step 4950000 | 44042s | train_loss=0.0090
498
+ step 4960000 | 44127s | train_loss=0.0089
499
+ step 4970000 | 44209s | train_loss=0.0086
500
+ step 4980000 | 44289s | train_loss=0.0093
501
+ step 4990000 | 44367s | train_loss=0.0089
502
+ step 5000000 | 44496s | train_loss=0.0088 | fid=52.3520
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/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/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": 42,
25
+ "device": null
26
+ }
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/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/d50/runner.log ADDED
@@ -0,0 +1,502 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ FID real-side stats cached over 1000 diverse train images
2
+ step 1 | 68s | train_loss=0.9919 | fid=246.7101
3
+ step 10000 | 153s | train_loss=0.8439
4
+ step 20000 | 238s | train_loss=0.7862
5
+ step 30000 | 323s | train_loss=0.6993
6
+ step 40000 | 408s | train_loss=0.5671
7
+ step 50000 | 493s | train_loss=0.4057
8
+ step 60000 | 578s | train_loss=0.2844
9
+ step 70000 | 663s | train_loss=0.2167
10
+ step 80000 | 748s | train_loss=0.1825
11
+ step 90000 | 833s | train_loss=0.1609
12
+ step 100000 | 983s | train_loss=0.1459 | fid=59.2101
13
+ step 110000 | 1063s | train_loss=0.1355
14
+ step 120000 | 1148s | train_loss=0.1256
15
+ step 130000 | 1233s | train_loss=0.1185
16
+ step 140000 | 1318s | train_loss=0.1137
17
+ step 150000 | 1402s | train_loss=0.1077
18
+ step 160000 | 1487s | train_loss=0.1032
19
+ step 170000 | 1572s | train_loss=0.0988
20
+ step 180000 | 1657s | train_loss=0.0936
21
+ step 190000 | 1741s | train_loss=0.0903
22
+ step 200000 | 1891s | train_loss=0.0880 | fid=56.6036
23
+ step 210000 | 1973s | train_loss=0.0834
24
+ step 220000 | 2056s | train_loss=0.0811
25
+ step 230000 | 2141s | train_loss=0.0780
26
+ step 240000 | 2226s | train_loss=0.0763
27
+ step 250000 | 2311s | train_loss=0.0739
28
+ step 260000 | 2396s | train_loss=0.0743
29
+ step 270000 | 2480s | train_loss=0.0707
30
+ step 280000 | 2565s | train_loss=0.0689
31
+ step 290000 | 2650s | train_loss=0.0672
32
+ step 300000 | 2801s | train_loss=0.0656 | fid=54.9854
33
+ step 310000 | 2882s | train_loss=0.0651
34
+ step 320000 | 2968s | train_loss=0.0624
35
+ step 330000 | 3051s | train_loss=0.0622
36
+ step 340000 | 3136s | train_loss=0.0596
37
+ step 350000 | 3221s | train_loss=0.0601
38
+ step 360000 | 3306s | train_loss=0.0579
39
+ step 370000 | 3391s | train_loss=0.0562
40
+ step 380000 | 3476s | train_loss=0.0562
41
+ step 390000 | 3561s | train_loss=0.0555
42
+ step 400000 | 3712s | train_loss=0.0542 | fid=52.9958
43
+ step 410000 | 3797s | train_loss=0.0539
44
+ step 420000 | 3880s | train_loss=0.0532
45
+ step 430000 | 3963s | train_loss=0.0513
46
+ step 440000 | 4048s | train_loss=0.0514
47
+ step 450000 | 4133s | train_loss=0.0510
48
+ step 460000 | 4218s | train_loss=0.0497
49
+ step 470000 | 4303s | train_loss=0.0494
50
+ step 480000 | 4388s | train_loss=0.0479
51
+ step 490000 | 4473s | train_loss=0.0478
52
+ step 500000 | 4624s | train_loss=0.0476 | fid=52.3497
53
+ step 510000 | 4709s | train_loss=0.0484
54
+ step 520000 | 4792s | train_loss=0.0460
55
+ step 530000 | 4877s | train_loss=0.0456
56
+ step 540000 | 4960s | train_loss=0.0446
57
+ step 550000 | 5045s | train_loss=0.0447
58
+ step 560000 | 5130s | train_loss=0.0441
59
+ step 570000 | 5215s | train_loss=0.0432
60
+ step 580000 | 5300s | train_loss=0.0430
61
+ step 590000 | 5385s | train_loss=0.0427
62
+ step 600000 | 5534s | train_loss=0.0416 | fid=51.1385
63
+ step 610000 | 5619s | train_loss=0.0417
64
+ step 620000 | 5702s | train_loss=0.0413
65
+ step 630000 | 5787s | train_loss=0.0413
66
+ step 640000 | 5872s | train_loss=0.0413
67
+ step 650000 | 5955s | train_loss=0.0400
68
+ step 660000 | 6040s | train_loss=0.0402
69
+ step 670000 | 6125s | train_loss=0.0407
70
+ step 680000 | 6210s | train_loss=0.0393
71
+ step 690000 | 6295s | train_loss=0.0392
72
+ step 700000 | 6442s | train_loss=0.0379 | fid=49.2325
73
+ step 710000 | 6527s | train_loss=0.0381
74
+ step 720000 | 6612s | train_loss=0.0379
75
+ step 730000 | 6694s | train_loss=0.0373
76
+ step 740000 | 6779s | train_loss=0.0371
77
+ step 750000 | 6865s | train_loss=0.0373
78
+ step 760000 | 6948s | train_loss=0.0370
79
+ step 770000 | 7033s | train_loss=0.0369
80
+ step 780000 | 7118s | train_loss=0.0368
81
+ step 790000 | 7203s | train_loss=0.0358
82
+ step 800000 | 7353s | train_loss=0.0364 | fid=48.6373
83
+ step 810000 | 7438s | train_loss=0.0347
84
+ step 820000 | 7523s | train_loss=0.0356
85
+ step 830000 | 7605s | train_loss=0.0345
86
+ step 840000 | 7690s | train_loss=0.0348
87
+ step 850000 | 7775s | train_loss=0.0342
88
+ step 860000 | 7859s | train_loss=0.0345
89
+ step 870000 | 7943s | train_loss=0.0344
90
+ step 880000 | 8028s | train_loss=0.0336
91
+ step 890000 | 8113s | train_loss=0.0328
92
+ step 900000 | 8262s | train_loss=0.0339 | fid=49.2059
93
+ step 910000 | 8347s | train_loss=0.0331
94
+ step 920000 | 8432s | train_loss=0.0330
95
+ step 930000 | 8515s | train_loss=0.0337
96
+ step 940000 | 8599s | train_loss=0.0324
97
+ step 950000 | 8684s | train_loss=0.0323
98
+ step 960000 | 8770s | train_loss=0.0315
99
+ step 970000 | 8853s | train_loss=0.0318
100
+ step 980000 | 8938s | train_loss=0.0317
101
+ step 990000 | 9023s | train_loss=0.0314
102
+ step 1000000 | 9174s | train_loss=0.0306 | fid=48.3376
103
+ step 1010000 | 9259s | train_loss=0.0310
104
+ step 1020000 | 9344s | train_loss=0.0303
105
+ step 1030000 | 9428s | train_loss=0.0306
106
+ step 1040000 | 9511s | train_loss=0.0300
107
+ step 1050000 | 9596s | train_loss=0.0303
108
+ step 1060000 | 9680s | train_loss=0.0299
109
+ step 1070000 | 9766s | train_loss=0.0293
110
+ step 1080000 | 9849s | train_loss=0.0295
111
+ step 1090000 | 9934s | train_loss=0.0292
112
+ step 1100000 | 10084s | train_loss=0.0299 | fid=48.4520
113
+ step 1110000 | 10169s | train_loss=0.0289
114
+ step 1120000 | 10254s | train_loss=0.0291
115
+ step 1130000 | 10339s | train_loss=0.0286
116
+ step 1140000 | 10421s | train_loss=0.0284
117
+ step 1150000 | 10505s | train_loss=0.0291
118
+ step 1160000 | 10590s | train_loss=0.0277
119
+ step 1170000 | 10676s | train_loss=0.0280
120
+ step 1180000 | 10761s | train_loss=0.0280
121
+ step 1190000 | 10844s | train_loss=0.0287
122
+ step 1200000 | 10994s | train_loss=0.0285 | fid=48.8923
123
+ step 1210000 | 11079s | train_loss=0.0276
124
+ step 1220000 | 11164s | train_loss=0.0273
125
+ step 1230000 | 11249s | train_loss=0.0271
126
+ step 1240000 | 11331s | train_loss=0.0272
127
+ step 1250000 | 11416s | train_loss=0.0268
128
+ step 1260000 | 11501s | train_loss=0.0273
129
+ step 1270000 | 11586s | train_loss=0.0265
130
+ step 1280000 | 11671s | train_loss=0.0270
131
+ step 1290000 | 11755s | train_loss=0.0270
132
+ step 1300000 | 11901s | train_loss=0.0265 | fid=47.9189
133
+ step 1310000 | 11986s | train_loss=0.0263
134
+ step 1320000 | 12071s | train_loss=0.0253
135
+ step 1330000 | 12155s | train_loss=0.0257
136
+ step 1340000 | 12240s | train_loss=0.0249
137
+ step 1350000 | 12323s | train_loss=0.0263
138
+ step 1360000 | 12408s | train_loss=0.0268
139
+ step 1370000 | 12493s | train_loss=0.0253
140
+ step 1380000 | 12578s | train_loss=0.0248
141
+ step 1390000 | 12661s | train_loss=0.0252
142
+ step 1400000 | 12808s | train_loss=0.0252 | fid=47.5287
143
+ step 1410000 | 12893s | train_loss=0.0243
144
+ step 1420000 | 12978s | train_loss=0.0250
145
+ step 1430000 | 13063s | train_loss=0.0246
146
+ step 1440000 | 13148s | train_loss=0.0244
147
+ step 1450000 | 13230s | train_loss=0.0246
148
+ step 1460000 | 13315s | train_loss=0.0257
149
+ step 1470000 | 13400s | train_loss=0.0240
150
+ step 1480000 | 13485s | train_loss=0.0238
151
+ step 1490000 | 13567s | train_loss=0.0243
152
+ step 1500000 | 13715s | train_loss=0.0242 | fid=48.1922
153
+ step 1510000 | 13800s | train_loss=0.0235
154
+ step 1520000 | 13885s | train_loss=0.0235
155
+ step 1530000 | 13970s | train_loss=0.0235
156
+ step 1540000 | 14055s | train_loss=0.0239
157
+ step 1550000 | 14137s | train_loss=0.0224
158
+ step 1560000 | 14222s | train_loss=0.0240
159
+ step 1570000 | 14307s | train_loss=0.0228
160
+ step 1580000 | 14392s | train_loss=0.0236
161
+ step 1590000 | 14472s | train_loss=0.0224
162
+ step 1600000 | 14664s | train_loss=0.0222 | fid=47.6388
163
+ step 1610000 | 14749s | train_loss=0.0235
164
+ step 1620000 | 14834s | train_loss=0.0221
165
+ step 1630000 | 14919s | train_loss=0.0221
166
+ step 1640000 | 15004s | train_loss=0.0220
167
+ step 1650000 | 15080s | train_loss=0.0217
168
+ step 1660000 | 15165s | train_loss=0.0220
169
+ step 1670000 | 15250s | train_loss=0.0220
170
+ step 1680000 | 15335s | train_loss=0.0226
171
+ step 1690000 | 15415s | train_loss=0.0215
172
+ step 1700000 | 15571s | train_loss=0.0224 | fid=47.4528
173
+ step 1710000 | 15654s | train_loss=0.0225
174
+ step 1720000 | 15739s | train_loss=0.0212
175
+ step 1730000 | 15824s | train_loss=0.0217
176
+ step 1740000 | 15909s | train_loss=0.0212
177
+ step 1750000 | 15992s | train_loss=0.0224
178
+ step 1760000 | 16076s | train_loss=0.0214
179
+ step 1770000 | 16161s | train_loss=0.0210
180
+ step 1780000 | 16246s | train_loss=0.0210
181
+ step 1790000 | 16326s | train_loss=0.0204
182
+ step 1800000 | 16482s | train_loss=0.0209 | fid=47.9258
183
+ step 1810000 | 16568s | train_loss=0.0205
184
+ step 1820000 | 16650s | train_loss=0.0207
185
+ step 1830000 | 16736s | train_loss=0.0217
186
+ step 1840000 | 16821s | train_loss=0.0208
187
+ step 1850000 | 16905s | train_loss=0.0210
188
+ step 1860000 | 16987s | train_loss=0.0205
189
+ step 1870000 | 17072s | train_loss=0.0198
190
+ step 1880000 | 17157s | train_loss=0.0200
191
+ step 1890000 | 17236s | train_loss=0.0199
192
+ step 1900000 | 17393s | train_loss=0.0207 | fid=47.4666
193
+ step 1910000 | 17478s | train_loss=0.0196
194
+ step 1920000 | 17561s | train_loss=0.0203
195
+ step 1930000 | 17646s | train_loss=0.0192
196
+ step 1940000 | 17731s | train_loss=0.0195
197
+ step 1950000 | 17816s | train_loss=0.0194
198
+ step 1960000 | 17898s | train_loss=0.0191
199
+ step 1970000 | 17983s | train_loss=0.0205
200
+ step 1980000 | 18066s | train_loss=0.0198
201
+ step 1990000 | 18148s | train_loss=0.0187
202
+ step 2000000 | 18304s | train_loss=0.0192 | fid=47.4670
203
+ step 2010000 | 18389s | train_loss=0.0187
204
+ step 2020000 | 18474s | train_loss=0.0201
205
+ step 2030000 | 18558s | train_loss=0.0192
206
+ step 2040000 | 18643s | train_loss=0.0193
207
+ step 2050000 | 18728s | train_loss=0.0190
208
+ step 2060000 | 18810s | train_loss=0.0187
209
+ step 2070000 | 18895s | train_loss=0.0189
210
+ step 2080000 | 18977s | train_loss=0.0181
211
+ step 2090000 | 19060s | train_loss=0.0182
212
+ step 2100000 | 19213s | train_loss=0.0188 | fid=47.6967
213
+ step 2110000 | 19298s | train_loss=0.0180
214
+ step 2120000 | 19383s | train_loss=0.0182
215
+ step 2130000 | 19468s | train_loss=0.0185
216
+ step 2140000 | 19552s | train_loss=0.0180
217
+ step 2150000 | 19637s | train_loss=0.0176
218
+ step 2160000 | 19722s | train_loss=0.0180
219
+ step 2170000 | 19795s | train_loss=0.0179
220
+ step 2180000 | 19876s | train_loss=0.0176
221
+ step 2190000 | 19952s | train_loss=0.0187
222
+ step 2200000 | 20141s | train_loss=0.0173 | fid=46.2277
223
+ step 2210000 | 20226s | train_loss=0.0177
224
+ step 2220000 | 20311s | train_loss=0.0179
225
+ step 2230000 | 20396s | train_loss=0.0182
226
+ step 2240000 | 20480s | train_loss=0.0174
227
+ step 2250000 | 20565s | train_loss=0.0177
228
+ step 2260000 | 20650s | train_loss=0.0178
229
+ step 2270000 | 20726s | train_loss=0.0166
230
+ step 2280000 | 20804s | train_loss=0.0174
231
+ step 2290000 | 20884s | train_loss=0.0173
232
+ step 2300000 | 21080s | train_loss=0.0178 | fid=47.5105
233
+ step 2310000 | 21164s | train_loss=0.0172
234
+ step 2320000 | 21249s | train_loss=0.0178
235
+ step 2330000 | 21334s | train_loss=0.0170
236
+ step 2340000 | 21418s | train_loss=0.0171
237
+ step 2350000 | 21503s | train_loss=0.0168
238
+ step 2360000 | 21588s | train_loss=0.0174
239
+ step 2370000 | 21664s | train_loss=0.0162
240
+ step 2380000 | 21737s | train_loss=0.0167
241
+ step 2390000 | 21818s | train_loss=0.0166
242
+ step 2400000 | 22027s | train_loss=0.0164 | fid=47.7332
243
+ step 2410000 | 22112s | train_loss=0.0160
244
+ step 2420000 | 22197s | train_loss=0.0163
245
+ step 2430000 | 22282s | train_loss=0.0168
246
+ step 2440000 | 22365s | train_loss=0.0158
247
+ step 2450000 | 22450s | train_loss=0.0162
248
+ step 2460000 | 22533s | train_loss=0.0169
249
+ step 2470000 | 22609s | train_loss=0.0162
250
+ step 2480000 | 22683s | train_loss=0.0162
251
+ step 2490000 | 22768s | train_loss=0.0160
252
+ step 2500000 | 22966s | train_loss=0.0161 | fid=47.1614
253
+ step 2510000 | 23051s | train_loss=0.0165
254
+ step 2520000 | 23136s | train_loss=0.0156
255
+ step 2530000 | 23221s | train_loss=0.0154
256
+ step 2540000 | 23305s | train_loss=0.0161
257
+ step 2550000 | 23387s | train_loss=0.0162
258
+ step 2560000 | 23472s | train_loss=0.0159
259
+ step 2570000 | 23546s | train_loss=0.0156
260
+ step 2580000 | 23624s | train_loss=0.0153
261
+ step 2590000 | 23708s | train_loss=0.0165
262
+ step 2600000 | 23909s | train_loss=0.0154 | fid=47.2053
263
+ step 2610000 | 23994s | train_loss=0.0163
264
+ step 2620000 | 24079s | train_loss=0.0150
265
+ step 2630000 | 24164s | train_loss=0.0156
266
+ step 2640000 | 24245s | train_loss=0.0151
267
+ step 2650000 | 24330s | train_loss=0.0150
268
+ step 2660000 | 24415s | train_loss=0.0150
269
+ step 2670000 | 24484s | train_loss=0.0157
270
+ step 2680000 | 24560s | train_loss=0.0150
271
+ step 2690000 | 24645s | train_loss=0.0155
272
+ step 2700000 | 24848s | train_loss=0.0147 | fid=46.6519
273
+ step 2710000 | 24933s | train_loss=0.0153
274
+ step 2720000 | 25018s | train_loss=0.0151
275
+ step 2730000 | 25103s | train_loss=0.0146
276
+ step 2740000 | 25184s | train_loss=0.0152
277
+ step 2750000 | 25269s | train_loss=0.0147
278
+ step 2760000 | 25352s | train_loss=0.0146
279
+ step 2770000 | 25428s | train_loss=0.0152
280
+ step 2780000 | 25504s | train_loss=0.0142
281
+ step 2790000 | 25589s | train_loss=0.0144
282
+ step 2800000 | 25795s | train_loss=0.0148 | fid=46.9532
283
+ step 2810000 | 25880s | train_loss=0.0141
284
+ step 2820000 | 25965s | train_loss=0.0142
285
+ step 2830000 | 26047s | train_loss=0.0147
286
+ step 2840000 | 26131s | train_loss=0.0144
287
+ step 2850000 | 26216s | train_loss=0.0138
288
+ step 2860000 | 26291s | train_loss=0.0143
289
+ step 2870000 | 26371s | train_loss=0.0144
290
+ step 2880000 | 26445s | train_loss=0.0137
291
+ step 2890000 | 26530s | train_loss=0.0141
292
+ step 2900000 | 26722s | train_loss=0.0144 | fid=47.3817
293
+ step 2910000 | 26807s | train_loss=0.0141
294
+ step 2920000 | 26890s | train_loss=0.0140
295
+ step 2930000 | 26975s | train_loss=0.0133
296
+ step 2940000 | 27058s | train_loss=0.0134
297
+ step 2950000 | 27143s | train_loss=0.0135
298
+ step 2960000 | 27219s | train_loss=0.0132
299
+ step 2970000 | 27299s | train_loss=0.0140
300
+ step 2980000 | 27373s | train_loss=0.0131
301
+ step 2990000 | 27458s | train_loss=0.0140
302
+ step 3000000 | 27637s | train_loss=0.0131 | fid=47.5070
303
+ step 3010000 | 27721s | train_loss=0.0133
304
+ step 3020000 | 27804s | train_loss=0.0133
305
+ step 3030000 | 27889s | train_loss=0.0133
306
+ step 3040000 | 27975s | train_loss=0.0131
307
+ step 3050000 | 28058s | train_loss=0.0129
308
+ step 3060000 | 28137s | train_loss=0.0131
309
+ step 3070000 | 28217s | train_loss=0.0136
310
+ step 3080000 | 28296s | train_loss=0.0129
311
+ step 3090000 | 28372s | train_loss=0.0134
312
+ step 3100000 | 28546s | train_loss=0.0130 | fid=47.1746
313
+ step 3110000 | 28629s | train_loss=0.0125
314
+ step 3120000 | 28714s | train_loss=0.0134
315
+ step 3130000 | 28799s | train_loss=0.0131
316
+ step 3140000 | 28884s | train_loss=0.0131
317
+ step 3150000 | 28967s | train_loss=0.0123
318
+ step 3160000 | 29046s | train_loss=0.0130
319
+ step 3170000 | 29128s | train_loss=0.0123
320
+ step 3180000 | 29213s | train_loss=0.0128
321
+ step 3190000 | 29288s | train_loss=0.0130
322
+ step 3200000 | 29481s | train_loss=0.0121 | fid=47.2685
323
+ step 3210000 | 29565s | train_loss=0.0124
324
+ step 3220000 | 29650s | train_loss=0.0125
325
+ step 3230000 | 29735s | train_loss=0.0121
326
+ step 3240000 | 29820s | train_loss=0.0124
327
+ step 3250000 | 29903s | train_loss=0.0120
328
+ step 3260000 | 29986s | train_loss=0.0130
329
+ step 3270000 | 30071s | train_loss=0.0127
330
+ step 3280000 | 30156s | train_loss=0.0125
331
+ step 3290000 | 30234s | train_loss=0.0123
332
+ step 3300000 | 30433s | train_loss=0.0124 | fid=47.4537
333
+ step 3310000 | 30518s | train_loss=0.0132
334
+ step 3320000 | 30603s | train_loss=0.0124
335
+ step 3330000 | 30688s | train_loss=0.0126
336
+ step 3340000 | 30773s | train_loss=0.0118
337
+ step 3350000 | 30854s | train_loss=0.0123
338
+ step 3360000 | 30939s | train_loss=0.0122
339
+ step 3370000 | 31024s | train_loss=0.0125
340
+ step 3380000 | 31109s | train_loss=0.0119
341
+ step 3390000 | 31191s | train_loss=0.0119
342
+ step 3400000 | 31344s | train_loss=0.0121 | fid=47.0477
343
+ step 3410000 | 31429s | train_loss=0.0116
344
+ step 3420000 | 31514s | train_loss=0.0117
345
+ step 3430000 | 31599s | train_loss=0.0116
346
+ step 3440000 | 31684s | train_loss=0.0117
347
+ step 3450000 | 31767s | train_loss=0.0120
348
+ step 3460000 | 31850s | train_loss=0.0120
349
+ step 3470000 | 31935s | train_loss=0.0117
350
+ step 3480000 | 32020s | train_loss=0.0118
351
+ step 3490000 | 32102s | train_loss=0.0117
352
+ step 3500000 | 32256s | train_loss=0.0122 | fid=47.1960
353
+ step 3510000 | 32341s | train_loss=0.0113
354
+ step 3520000 | 32425s | train_loss=0.0119
355
+ step 3530000 | 32510s | train_loss=0.0119
356
+ step 3540000 | 32595s | train_loss=0.0110
357
+ step 3550000 | 32667s | train_loss=0.0112
358
+ step 3560000 | 32752s | train_loss=0.0111
359
+ step 3570000 | 32835s | train_loss=0.0110
360
+ step 3580000 | 32920s | train_loss=0.0117
361
+ step 3590000 | 33005s | train_loss=0.0114
362
+ step 3600000 | 33173s | train_loss=0.0114 | fid=47.1739
363
+ step 3610000 | 33258s | train_loss=0.0109
364
+ step 3620000 | 33343s | train_loss=0.0116
365
+ step 3630000 | 33428s | train_loss=0.0125
366
+ step 3640000 | 33513s | train_loss=0.0110
367
+ step 3650000 | 33585s | train_loss=0.0113
368
+ step 3660000 | 33670s | train_loss=0.0112
369
+ step 3670000 | 33753s | train_loss=0.0112
370
+ step 3680000 | 33838s | train_loss=0.0112
371
+ step 3690000 | 33923s | train_loss=0.0116
372
+ step 3700000 | 34107s | train_loss=0.0106 | fid=47.1999
373
+ step 3710000 | 34192s | train_loss=0.0111
374
+ step 3720000 | 34277s | train_loss=0.0110
375
+ step 3730000 | 34361s | train_loss=0.0111
376
+ step 3740000 | 34446s | train_loss=0.0108
377
+ step 3750000 | 34522s | train_loss=0.0109
378
+ step 3760000 | 34607s | train_loss=0.0114
379
+ step 3770000 | 34691s | train_loss=0.0113
380
+ step 3780000 | 34776s | train_loss=0.0107
381
+ step 3790000 | 34859s | train_loss=0.0109
382
+ step 3800000 | 35034s | train_loss=0.0105 | fid=46.7863
383
+ step 3810000 | 35119s | train_loss=0.0109
384
+ step 3820000 | 35204s | train_loss=0.0107
385
+ step 3830000 | 35289s | train_loss=0.0112
386
+ step 3840000 | 35371s | train_loss=0.0104
387
+ step 3850000 | 35446s | train_loss=0.0106
388
+ step 3860000 | 35531s | train_loss=0.0108
389
+ step 3870000 | 35614s | train_loss=0.0102
390
+ step 3880000 | 35697s | train_loss=0.0107
391
+ step 3890000 | 35782s | train_loss=0.0105
392
+ step 3900000 | 35967s | train_loss=0.0104 | fid=46.5991
393
+ step 3910000 | 36052s | train_loss=0.0102
394
+ step 3920000 | 36137s | train_loss=0.0110
395
+ step 3930000 | 36222s | train_loss=0.0103
396
+ step 3940000 | 36302s | train_loss=0.0107
397
+ step 3950000 | 36380s | train_loss=0.0106
398
+ step 3960000 | 36465s | train_loss=0.0101
399
+ step 3970000 | 36549s | train_loss=0.0105
400
+ step 3980000 | 36632s | train_loss=0.0101
401
+ step 3990000 | 36717s | train_loss=0.0100
402
+ step 4000000 | 36906s | train_loss=0.0101 | fid=46.6595
403
+ step 4010000 | 36991s | train_loss=0.0104
404
+ step 4020000 | 37076s | train_loss=0.0100
405
+ step 4030000 | 37161s | train_loss=0.0102
406
+ step 4040000 | 37241s | train_loss=0.0104
407
+ step 4050000 | 37326s | train_loss=0.0107
408
+ step 4060000 | 37411s | train_loss=0.0101
409
+ step 4070000 | 37493s | train_loss=0.0099
410
+ step 4080000 | 37577s | train_loss=0.0102
411
+ step 4090000 | 37663s | train_loss=0.0097
412
+ step 4100000 | 37815s | train_loss=0.0104 | fid=47.4107
413
+ step 4110000 | 37900s | train_loss=0.0104
414
+ step 4120000 | 37985s | train_loss=0.0104
415
+ step 4130000 | 38070s | train_loss=0.0100
416
+ step 4140000 | 38151s | train_loss=0.0098
417
+ step 4150000 | 38236s | train_loss=0.0103
418
+ step 4160000 | 38321s | train_loss=0.0105
419
+ step 4170000 | 38404s | train_loss=0.0102
420
+ step 4180000 | 38487s | train_loss=0.0096
421
+ step 4190000 | 38572s | train_loss=0.0097
422
+ step 4200000 | 38728s | train_loss=0.0096 | fid=46.6825
423
+ step 4210000 | 38811s | train_loss=0.0099
424
+ step 4220000 | 38896s | train_loss=0.0102
425
+ step 4230000 | 38981s | train_loss=0.0096
426
+ step 4240000 | 39061s | train_loss=0.0098
427
+ step 4250000 | 39145s | train_loss=0.0105
428
+ step 4260000 | 39230s | train_loss=0.0096
429
+ step 4270000 | 39313s | train_loss=0.0100
430
+ step 4280000 | 39399s | train_loss=0.0096
431
+ step 4290000 | 39482s | train_loss=0.0097
432
+ step 4300000 | 39638s | train_loss=0.0099 | fid=47.0023
433
+ step 4310000 | 39719s | train_loss=0.0099
434
+ step 4320000 | 39804s | train_loss=0.0095
435
+ step 4330000 | 39889s | train_loss=0.0095
436
+ step 4340000 | 39971s | train_loss=0.0095
437
+ step 4350000 | 40056s | train_loss=0.0094
438
+ step 4360000 | 40140s | train_loss=0.0094
439
+ step 4370000 | 40224s | train_loss=0.0093
440
+ step 4380000 | 40309s | train_loss=0.0097
441
+ step 4390000 | 40394s | train_loss=0.0093
442
+ step 4400000 | 40595s | train_loss=0.0101 | fid=47.4748
443
+ step 4410000 | 40664s | train_loss=0.0092
444
+ step 4420000 | 40749s | train_loss=0.0094
445
+ step 4430000 | 40826s | train_loss=0.0094
446
+ step 4440000 | 40911s | train_loss=0.0092
447
+ step 4450000 | 40996s | train_loss=0.0098
448
+ step 4460000 | 41079s | train_loss=0.0093
449
+ step 4470000 | 41164s | train_loss=0.0094
450
+ step 4480000 | 41249s | train_loss=0.0091
451
+ step 4490000 | 41334s | train_loss=0.0092
452
+ step 4500000 | 41522s | train_loss=0.0092 | fid=47.0908
453
+ step 4510000 | 41600s | train_loss=0.0097
454
+ step 4520000 | 41676s | train_loss=0.0092
455
+ step 4530000 | 41751s | train_loss=0.0092
456
+ step 4540000 | 41836s | train_loss=0.0094
457
+ step 4550000 | 41919s | train_loss=0.0094
458
+ step 4560000 | 42004s | train_loss=0.0091
459
+ step 4570000 | 42089s | train_loss=0.0095
460
+ step 4580000 | 42174s | train_loss=0.0091
461
+ step 4590000 | 42259s | train_loss=0.0090
462
+ step 4600000 | 42461s | train_loss=0.0096 | fid=47.4540
463
+ step 4610000 | 42544s | train_loss=0.0095
464
+ step 4620000 | 42614s | train_loss=0.0094
465
+ step 4630000 | 42689s | train_loss=0.0089
466
+ step 4640000 | 42771s | train_loss=0.0095
467
+ step 4650000 | 42856s | train_loss=0.0093
468
+ step 4660000 | 42941s | train_loss=0.0094
469
+ step 4670000 | 43026s | train_loss=0.0088
470
+ step 4680000 | 43111s | train_loss=0.0091
471
+ step 4690000 | 43196s | train_loss=0.0091
472
+ step 4700000 | 43393s | train_loss=0.0090 | fid=47.3429
473
+ step 4710000 | 43478s | train_loss=0.0089
474
+ step 4720000 | 43546s | train_loss=0.0092
475
+ step 4730000 | 43620s | train_loss=0.0089
476
+ step 4740000 | 43705s | train_loss=0.0085
477
+ step 4750000 | 43789s | train_loss=0.0087
478
+ step 4760000 | 43874s | train_loss=0.0089
479
+ step 4770000 | 43959s | train_loss=0.0089
480
+ step 4780000 | 44044s | train_loss=0.0090
481
+ step 4790000 | 44129s | train_loss=0.0090
482
+ step 4800000 | 44315s | train_loss=0.0086 | fid=47.4511
483
+ step 4810000 | 44400s | train_loss=0.0088
484
+ step 4820000 | 44471s | train_loss=0.0089
485
+ step 4830000 | 44537s | train_loss=0.0087
486
+ step 4840000 | 44605s | train_loss=0.0086
487
+ step 4850000 | 44673s | train_loss=0.0087
488
+ step 4860000 | 44741s | train_loss=0.0094
489
+ step 4870000 | 44809s | train_loss=0.0087
490
+ step 4880000 | 44878s | train_loss=0.0086
491
+ step 4890000 | 44946s | train_loss=0.0082
492
+ step 4900000 | 45106s | train_loss=0.0089 | fid=46.7630
493
+ step 4910000 | 45174s | train_loss=0.0089
494
+ step 4920000 | 45223s | train_loss=0.0087
495
+ step 4930000 | 45274s | train_loss=0.0085
496
+ step 4940000 | 45326s | train_loss=0.0087
497
+ step 4950000 | 45377s | train_loss=0.0088
498
+ step 4960000 | 45429s | train_loss=0.0086
499
+ step 4970000 | 45480s | train_loss=0.0085
500
+ step 4980000 | 45532s | train_loss=0.0089
501
+ step 4990000 | 45584s | train_loss=0.0085
502
+ step 5000000 | 45676s | train_loss=0.0087 | fid=46.8034
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/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/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": 42,
25
+ "device": null
26
+ }
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/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/d60/runner.log ADDED
@@ -0,0 +1,502 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ FID real-side stats cached over 1000 diverse train images
2
+ step 1 | 73s | train_loss=0.9958 | fid=302.5537
3
+ step 10000 | 166s | train_loss=0.8357
4
+ step 20000 | 260s | train_loss=0.7573
5
+ step 30000 | 353s | train_loss=0.6506
6
+ step 40000 | 447s | train_loss=0.4970
7
+ step 50000 | 540s | train_loss=0.3423
8
+ step 60000 | 634s | train_loss=0.2424
9
+ step 70000 | 727s | train_loss=0.1903
10
+ step 80000 | 821s | train_loss=0.1640
11
+ step 90000 | 914s | train_loss=0.1470
12
+ step 100000 | 1075s | train_loss=0.1350 | fid=56.3475
13
+ step 110000 | 1167s | train_loss=0.1267
14
+ step 120000 | 1260s | train_loss=0.1167
15
+ step 130000 | 1354s | train_loss=0.1126
16
+ step 140000 | 1447s | train_loss=0.1074
17
+ step 150000 | 1541s | train_loss=0.1021
18
+ step 160000 | 1634s | train_loss=0.0975
19
+ step 170000 | 1728s | train_loss=0.0946
20
+ step 180000 | 1821s | train_loss=0.0902
21
+ step 190000 | 1914s | train_loss=0.0865
22
+ step 200000 | 2075s | train_loss=0.0855 | fid=55.2112
23
+ step 210000 | 2167s | train_loss=0.0817
24
+ step 220000 | 2260s | train_loss=0.0796
25
+ step 230000 | 2354s | train_loss=0.0762
26
+ step 240000 | 2447s | train_loss=0.0751
27
+ step 250000 | 2541s | train_loss=0.0717
28
+ step 260000 | 2634s | train_loss=0.0716
29
+ step 270000 | 2728s | train_loss=0.0695
30
+ step 280000 | 2821s | train_loss=0.0681
31
+ step 290000 | 2909s | train_loss=0.0667
32
+ step 300000 | 3075s | train_loss=0.0648 | fid=53.2192
33
+ step 310000 | 3167s | train_loss=0.0630
34
+ step 320000 | 3261s | train_loss=0.0625
35
+ step 330000 | 3354s | train_loss=0.0612
36
+ step 340000 | 3448s | train_loss=0.0598
37
+ step 350000 | 3541s | train_loss=0.0589
38
+ step 360000 | 3635s | train_loss=0.0578
39
+ step 370000 | 3728s | train_loss=0.0564
40
+ step 380000 | 3818s | train_loss=0.0559
41
+ step 390000 | 3910s | train_loss=0.0560
42
+ step 400000 | 4074s | train_loss=0.0541 | fid=51.1340
43
+ step 410000 | 4166s | train_loss=0.0531
44
+ step 420000 | 4260s | train_loss=0.0526
45
+ step 430000 | 4353s | train_loss=0.0511
46
+ step 440000 | 4447s | train_loss=0.0519
47
+ step 450000 | 4541s | train_loss=0.0504
48
+ step 460000 | 4634s | train_loss=0.0498
49
+ step 470000 | 4728s | train_loss=0.0498
50
+ step 480000 | 4816s | train_loss=0.0487
51
+ step 490000 | 4910s | train_loss=0.0494
52
+ step 500000 | 5076s | train_loss=0.0468 | fid=49.9115
53
+ step 510000 | 5168s | train_loss=0.0467
54
+ step 520000 | 5261s | train_loss=0.0463
55
+ step 530000 | 5355s | train_loss=0.0461
56
+ step 540000 | 5449s | train_loss=0.0450
57
+ step 550000 | 5542s | train_loss=0.0448
58
+ step 560000 | 5636s | train_loss=0.0454
59
+ step 570000 | 5723s | train_loss=0.0442
60
+ step 580000 | 5817s | train_loss=0.0443
61
+ step 590000 | 5911s | train_loss=0.0435
62
+ step 600000 | 6076s | train_loss=0.0430 | fid=49.5294
63
+ step 610000 | 6168s | train_loss=0.0423
64
+ step 620000 | 6262s | train_loss=0.0431
65
+ step 630000 | 6355s | train_loss=0.0420
66
+ step 640000 | 6448s | train_loss=0.0415
67
+ step 650000 | 6542s | train_loss=0.0411
68
+ step 660000 | 6633s | train_loss=0.0409
69
+ step 670000 | 6724s | train_loss=0.0410
70
+ step 680000 | 6817s | train_loss=0.0405
71
+ step 690000 | 6910s | train_loss=0.0391
72
+ step 700000 | 7076s | train_loss=0.0392 | fid=48.5091
73
+ step 710000 | 7170s | train_loss=0.0386
74
+ step 720000 | 7262s | train_loss=0.0385
75
+ step 730000 | 7355s | train_loss=0.0389
76
+ step 740000 | 7448s | train_loss=0.0380
77
+ step 750000 | 7541s | train_loss=0.0380
78
+ step 760000 | 7630s | train_loss=0.0373
79
+ step 770000 | 7724s | train_loss=0.0375
80
+ step 780000 | 7817s | train_loss=0.0373
81
+ step 790000 | 7911s | train_loss=0.0361
82
+ step 800000 | 8077s | train_loss=0.0364 | fid=46.9388
83
+ step 810000 | 8171s | train_loss=0.0360
84
+ step 820000 | 8263s | train_loss=0.0357
85
+ step 830000 | 8356s | train_loss=0.0353
86
+ step 840000 | 8450s | train_loss=0.0351
87
+ step 850000 | 8538s | train_loss=0.0349
88
+ step 860000 | 8631s | train_loss=0.0348
89
+ step 870000 | 8724s | train_loss=0.0352
90
+ step 880000 | 8818s | train_loss=0.0339
91
+ step 890000 | 8912s | train_loss=0.0341
92
+ step 900000 | 9077s | train_loss=0.0342 | fid=46.0678
93
+ step 910000 | 9170s | train_loss=0.0338
94
+ step 920000 | 9263s | train_loss=0.0336
95
+ step 930000 | 9357s | train_loss=0.0335
96
+ step 940000 | 9447s | train_loss=0.0330
97
+ step 950000 | 9538s | train_loss=0.0331
98
+ step 960000 | 9631s | train_loss=0.0332
99
+ step 970000 | 9725s | train_loss=0.0321
100
+ step 980000 | 9818s | train_loss=0.0320
101
+ step 990000 | 9912s | train_loss=0.0327
102
+ step 1000000 | 10078s | train_loss=0.0315 | fid=47.0841
103
+ step 1010000 | 10171s | train_loss=0.0318
104
+ step 1020000 | 10263s | train_loss=0.0309
105
+ step 1030000 | 10354s | train_loss=0.0318
106
+ step 1040000 | 10444s | train_loss=0.0305
107
+ step 1050000 | 10538s | train_loss=0.0311
108
+ step 1060000 | 10631s | train_loss=0.0313
109
+ step 1070000 | 10725s | train_loss=0.0309
110
+ step 1080000 | 10819s | train_loss=0.0302
111
+ step 1090000 | 10913s | train_loss=0.0308
112
+ step 1100000 | 11078s | train_loss=0.0303 | fid=46.0613
113
+ step 1110000 | 11172s | train_loss=0.0302
114
+ step 1120000 | 11265s | train_loss=0.0294
115
+ step 1130000 | 11352s | train_loss=0.0286
116
+ step 1140000 | 11446s | train_loss=0.0295
117
+ step 1150000 | 11539s | train_loss=0.0297
118
+ step 1160000 | 11633s | train_loss=0.0286
119
+ step 1170000 | 11727s | train_loss=0.0289
120
+ step 1180000 | 11820s | train_loss=0.0292
121
+ step 1190000 | 11913s | train_loss=0.0283
122
+ step 1200000 | 12079s | train_loss=0.0285 | fid=45.3341
123
+ step 1210000 | 12172s | train_loss=0.0283
124
+ step 1220000 | 12262s | train_loss=0.0283
125
+ step 1230000 | 12352s | train_loss=0.0279
126
+ step 1240000 | 12446s | train_loss=0.0274
127
+ step 1250000 | 12539s | train_loss=0.0282
128
+ step 1260000 | 12633s | train_loss=0.0275
129
+ step 1270000 | 12726s | train_loss=0.0278
130
+ step 1280000 | 12820s | train_loss=0.0274
131
+ step 1290000 | 12913s | train_loss=0.0274
132
+ step 1300000 | 13079s | train_loss=0.0276 | fid=45.3483
133
+ step 1310000 | 13171s | train_loss=0.0265
134
+ step 1320000 | 13261s | train_loss=0.0269
135
+ step 1330000 | 13353s | train_loss=0.0264
136
+ step 1340000 | 13447s | train_loss=0.0261
137
+ step 1350000 | 13540s | train_loss=0.0263
138
+ step 1360000 | 13633s | train_loss=0.0264
139
+ step 1370000 | 13727s | train_loss=0.0257
140
+ step 1380000 | 13820s | train_loss=0.0253
141
+ step 1390000 | 13913s | train_loss=0.0264
142
+ step 1400000 | 14079s | train_loss=0.0255 | fid=44.5628
143
+ step 1410000 | 14166s | train_loss=0.0250
144
+ step 1420000 | 14250s | train_loss=0.0253
145
+ step 1430000 | 14342s | train_loss=0.0252
146
+ step 1440000 | 14436s | train_loss=0.0249
147
+ step 1450000 | 14530s | train_loss=0.0256
148
+ step 1460000 | 14623s | train_loss=0.0250
149
+ step 1470000 | 14716s | train_loss=0.0245
150
+ step 1480000 | 14810s | train_loss=0.0248
151
+ step 1490000 | 14904s | train_loss=0.0248
152
+ step 1500000 | 15060s | train_loss=0.0241 | fid=45.4586
153
+ step 1510000 | 15150s | train_loss=0.0238
154
+ step 1520000 | 15244s | train_loss=0.0238
155
+ step 1530000 | 15336s | train_loss=0.0240
156
+ step 1540000 | 15430s | train_loss=0.0235
157
+ step 1550000 | 15523s | train_loss=0.0233
158
+ step 1560000 | 15617s | train_loss=0.0239
159
+ step 1570000 | 15710s | train_loss=0.0231
160
+ step 1580000 | 15804s | train_loss=0.0231
161
+ step 1590000 | 15898s | train_loss=0.0230
162
+ step 1600000 | 16060s | train_loss=0.0236 | fid=44.3242
163
+ step 1610000 | 16150s | train_loss=0.0242
164
+ step 1620000 | 16244s | train_loss=0.0236
165
+ step 1630000 | 16336s | train_loss=0.0231
166
+ step 1640000 | 16430s | train_loss=0.0224
167
+ step 1650000 | 16523s | train_loss=0.0230
168
+ step 1660000 | 16617s | train_loss=0.0226
169
+ step 1670000 | 16710s | train_loss=0.0226
170
+ step 1680000 | 16803s | train_loss=0.0224
171
+ step 1690000 | 16897s | train_loss=0.0219
172
+ step 1700000 | 17056s | train_loss=0.0218 | fid=43.9918
173
+ step 1710000 | 17149s | train_loss=0.0228
174
+ step 1720000 | 17243s | train_loss=0.0218
175
+ step 1730000 | 17335s | train_loss=0.0220
176
+ step 1740000 | 17428s | train_loss=0.0215
177
+ step 1750000 | 17522s | train_loss=0.0225
178
+ step 1760000 | 17615s | train_loss=0.0216
179
+ step 1770000 | 17709s | train_loss=0.0217
180
+ step 1780000 | 17802s | train_loss=0.0214
181
+ step 1790000 | 17892s | train_loss=0.0210
182
+ step 1800000 | 18055s | train_loss=0.0213 | fid=44.8402
183
+ step 1810000 | 18148s | train_loss=0.0208
184
+ step 1820000 | 18242s | train_loss=0.0212
185
+ step 1830000 | 18334s | train_loss=0.0217
186
+ step 1840000 | 18428s | train_loss=0.0207
187
+ step 1850000 | 18521s | train_loss=0.0207
188
+ step 1860000 | 18615s | train_loss=0.0207
189
+ step 1870000 | 18709s | train_loss=0.0206
190
+ step 1880000 | 18801s | train_loss=0.0205
191
+ step 1890000 | 18892s | train_loss=0.0212
192
+ step 1900000 | 19055s | train_loss=0.0205 | fid=44.4206
193
+ step 1910000 | 19148s | train_loss=0.0202
194
+ step 1920000 | 19242s | train_loss=0.0210
195
+ step 1930000 | 19335s | train_loss=0.0195
196
+ step 1940000 | 19428s | train_loss=0.0196
197
+ step 1950000 | 19521s | train_loss=0.0201
198
+ step 1960000 | 19615s | train_loss=0.0200
199
+ step 1970000 | 19708s | train_loss=0.0198
200
+ step 1980000 | 19792s | train_loss=0.0196
201
+ step 1990000 | 19874s | train_loss=0.0197
202
+ step 2000000 | 20040s | train_loss=0.0195 | fid=43.9117
203
+ step 2010000 | 20133s | train_loss=0.0191
204
+ step 2020000 | 20227s | train_loss=0.0198
205
+ step 2030000 | 20321s | train_loss=0.0199
206
+ step 2040000 | 20413s | train_loss=0.0195
207
+ step 2050000 | 20506s | train_loss=0.0196
208
+ step 2060000 | 20600s | train_loss=0.0195
209
+ step 2070000 | 20692s | train_loss=0.0190
210
+ step 2080000 | 20778s | train_loss=0.0187
211
+ step 2090000 | 20864s | train_loss=0.0193
212
+ step 2100000 | 21030s | train_loss=0.0189 | fid=44.5271
213
+ step 2110000 | 21123s | train_loss=0.0189
214
+ step 2120000 | 21216s | train_loss=0.0190
215
+ step 2130000 | 21310s | train_loss=0.0184
216
+ step 2140000 | 21402s | train_loss=0.0187
217
+ step 2150000 | 21496s | train_loss=0.0185
218
+ step 2160000 | 21589s | train_loss=0.0189
219
+ step 2170000 | 21675s | train_loss=0.0182
220
+ step 2180000 | 21760s | train_loss=0.0185
221
+ step 2190000 | 21844s | train_loss=0.0191
222
+ step 2200000 | 22009s | train_loss=0.0179 | fid=44.6302
223
+ step 2210000 | 22103s | train_loss=0.0178
224
+ step 2220000 | 22197s | train_loss=0.0184
225
+ step 2230000 | 22290s | train_loss=0.0188
226
+ step 2240000 | 22382s | train_loss=0.0179
227
+ step 2250000 | 22476s | train_loss=0.0176
228
+ step 2260000 | 22569s | train_loss=0.0181
229
+ step 2270000 | 22654s | train_loss=0.0172
230
+ step 2280000 | 22743s | train_loss=0.0180
231
+ step 2290000 | 22827s | train_loss=0.0173
232
+ step 2300000 | 22991s | train_loss=0.0181 | fid=44.4038
233
+ step 2310000 | 23085s | train_loss=0.0175
234
+ step 2320000 | 23178s | train_loss=0.0179
235
+ step 2330000 | 23272s | train_loss=0.0174
236
+ step 2340000 | 23364s | train_loss=0.0171
237
+ step 2350000 | 23458s | train_loss=0.0184
238
+ step 2360000 | 23551s | train_loss=0.0173
239
+ step 2370000 | 23634s | train_loss=0.0168
240
+ step 2380000 | 23723s | train_loss=0.0171
241
+ step 2390000 | 23807s | train_loss=0.0165
242
+ step 2400000 | 23972s | train_loss=0.0166 | fid=43.7581
243
+ step 2410000 | 24066s | train_loss=0.0163
244
+ step 2420000 | 24159s | train_loss=0.0168
245
+ step 2430000 | 24253s | train_loss=0.0167
246
+ step 2440000 | 24345s | train_loss=0.0167
247
+ step 2450000 | 24439s | train_loss=0.0163
248
+ step 2460000 | 24527s | train_loss=0.0175
249
+ step 2470000 | 24616s | train_loss=0.0162
250
+ step 2480000 | 24705s | train_loss=0.0169
251
+ step 2490000 | 24790s | train_loss=0.0160
252
+ step 2500000 | 24956s | train_loss=0.0161 | fid=44.2116
253
+ step 2510000 | 25050s | train_loss=0.0164
254
+ step 2520000 | 25143s | train_loss=0.0157
255
+ step 2530000 | 25237s | train_loss=0.0156
256
+ step 2540000 | 25330s | train_loss=0.0163
257
+ step 2550000 | 25422s | train_loss=0.0161
258
+ step 2560000 | 25505s | train_loss=0.0164
259
+ step 2570000 | 25597s | train_loss=0.0165
260
+ step 2580000 | 25684s | train_loss=0.0157
261
+ step 2590000 | 25769s | train_loss=0.0162
262
+ step 2600000 | 25933s | train_loss=0.0158 | fid=44.1417
263
+ step 2610000 | 26027s | train_loss=0.0156
264
+ step 2620000 | 26120s | train_loss=0.0154
265
+ step 2630000 | 26214s | train_loss=0.0153
266
+ step 2640000 | 26308s | train_loss=0.0156
267
+ step 2650000 | 26399s | train_loss=0.0153
268
+ step 2660000 | 26481s | train_loss=0.0158
269
+ step 2670000 | 26574s | train_loss=0.0157
270
+ step 2680000 | 26663s | train_loss=0.0151
271
+ step 2690000 | 26751s | train_loss=0.0153
272
+ step 2700000 | 26917s | train_loss=0.0152 | fid=44.2630
273
+ step 2710000 | 27010s | train_loss=0.0156
274
+ step 2720000 | 27103s | train_loss=0.0147
275
+ step 2730000 | 27197s | train_loss=0.0150
276
+ step 2740000 | 27290s | train_loss=0.0150
277
+ step 2750000 | 27382s | train_loss=0.0148
278
+ step 2760000 | 27460s | train_loss=0.0143
279
+ step 2770000 | 27554s | train_loss=0.0154
280
+ step 2780000 | 27638s | train_loss=0.0145
281
+ step 2790000 | 27732s | train_loss=0.0146
282
+ step 2800000 | 27897s | train_loss=0.0151 | fid=44.4994
283
+ step 2810000 | 27990s | train_loss=0.0141
284
+ step 2820000 | 28084s | train_loss=0.0146
285
+ step 2830000 | 28177s | train_loss=0.0139
286
+ step 2840000 | 28271s | train_loss=0.0144
287
+ step 2850000 | 28363s | train_loss=0.0139
288
+ step 2860000 | 28442s | train_loss=0.0141
289
+ step 2870000 | 28535s | train_loss=0.0142
290
+ step 2880000 | 28614s | train_loss=0.0143
291
+ step 2890000 | 28708s | train_loss=0.0137
292
+ step 2900000 | 28873s | train_loss=0.0143 | fid=45.0040
293
+ step 2910000 | 28967s | train_loss=0.0139
294
+ step 2920000 | 29061s | train_loss=0.0138
295
+ step 2930000 | 29154s | train_loss=0.0141
296
+ step 2940000 | 29248s | train_loss=0.0139
297
+ step 2950000 | 29340s | train_loss=0.0131
298
+ step 2960000 | 29422s | train_loss=0.0131
299
+ step 2970000 | 29515s | train_loss=0.0133
300
+ step 2980000 | 29600s | train_loss=0.0138
301
+ step 2990000 | 29694s | train_loss=0.0142
302
+ step 3000000 | 29860s | train_loss=0.0134 | fid=43.5674
303
+ step 3010000 | 29953s | train_loss=0.0136
304
+ step 3020000 | 30046s | train_loss=0.0132
305
+ step 3030000 | 30140s | train_loss=0.0137
306
+ step 3040000 | 30234s | train_loss=0.0131
307
+ step 3050000 | 30324s | train_loss=0.0129
308
+ step 3060000 | 30410s | train_loss=0.0133
309
+ step 3070000 | 30501s | train_loss=0.0133
310
+ step 3080000 | 30590s | train_loss=0.0133
311
+ step 3090000 | 30684s | train_loss=0.0130
312
+ step 3100000 | 30849s | train_loss=0.0133 | fid=44.0411
313
+ step 3110000 | 30942s | train_loss=0.0130
314
+ step 3120000 | 31036s | train_loss=0.0131
315
+ step 3130000 | 31130s | train_loss=0.0132
316
+ step 3140000 | 31223s | train_loss=0.0127
317
+ step 3150000 | 31314s | train_loss=0.0130
318
+ step 3160000 | 31405s | train_loss=0.0132
319
+ step 3170000 | 31497s | train_loss=0.0123
320
+ step 3180000 | 31590s | train_loss=0.0127
321
+ step 3190000 | 31684s | train_loss=0.0127
322
+ step 3200000 | 31849s | train_loss=0.0126 | fid=44.0437
323
+ step 3210000 | 31943s | train_loss=0.0125
324
+ step 3220000 | 32036s | train_loss=0.0130
325
+ step 3230000 | 32130s | train_loss=0.0120
326
+ step 3240000 | 32222s | train_loss=0.0126
327
+ step 3250000 | 32314s | train_loss=0.0124
328
+ step 3260000 | 32402s | train_loss=0.0125
329
+ step 3270000 | 32487s | train_loss=0.0125
330
+ step 3280000 | 32581s | train_loss=0.0126
331
+ step 3290000 | 32674s | train_loss=0.0127
332
+ step 3300000 | 32839s | train_loss=0.0121 | fid=44.2864
333
+ step 3310000 | 32933s | train_loss=0.0130
334
+ step 3320000 | 33026s | train_loss=0.0128
335
+ step 3330000 | 33120s | train_loss=0.0127
336
+ step 3340000 | 33205s | train_loss=0.0122
337
+ step 3350000 | 33298s | train_loss=0.0125
338
+ step 3360000 | 33385s | train_loss=0.0120
339
+ step 3370000 | 33471s | train_loss=0.0126
340
+ step 3380000 | 33564s | train_loss=0.0119
341
+ step 3390000 | 33656s | train_loss=0.0120
342
+ step 3400000 | 33823s | train_loss=0.0123 | fid=44.6804
343
+ step 3410000 | 33917s | train_loss=0.0115
344
+ step 3420000 | 34010s | train_loss=0.0122
345
+ step 3430000 | 34101s | train_loss=0.0124
346
+ step 3440000 | 34186s | train_loss=0.0114
347
+ step 3450000 | 34279s | train_loss=0.0118
348
+ step 3460000 | 34364s | train_loss=0.0119
349
+ step 3470000 | 34451s | train_loss=0.0118
350
+ step 3480000 | 34544s | train_loss=0.0118
351
+ step 3490000 | 34636s | train_loss=0.0120
352
+ step 3500000 | 34803s | train_loss=0.0122 | fid=44.4324
353
+ step 3510000 | 34897s | train_loss=0.0118
354
+ step 3520000 | 34990s | train_loss=0.0115
355
+ step 3530000 | 35078s | train_loss=0.0118
356
+ step 3540000 | 35164s | train_loss=0.0111
357
+ step 3550000 | 35257s | train_loss=0.0113
358
+ step 3560000 | 35342s | train_loss=0.0116
359
+ step 3570000 | 35430s | train_loss=0.0110
360
+ step 3580000 | 35523s | train_loss=0.0111
361
+ step 3590000 | 35616s | train_loss=0.0114
362
+ step 3600000 | 35783s | train_loss=0.0120 | fid=44.5137
363
+ step 3610000 | 35876s | train_loss=0.0110
364
+ step 3620000 | 35969s | train_loss=0.0118
365
+ step 3630000 | 36055s | train_loss=0.0111
366
+ step 3640000 | 36147s | train_loss=0.0116
367
+ step 3650000 | 36241s | train_loss=0.0112
368
+ step 3660000 | 36321s | train_loss=0.0110
369
+ step 3670000 | 36410s | train_loss=0.0113
370
+ step 3680000 | 36504s | train_loss=0.0116
371
+ step 3690000 | 36596s | train_loss=0.0111
372
+ step 3700000 | 36762s | train_loss=0.0114 | fid=44.3437
373
+ step 3710000 | 36855s | train_loss=0.0109
374
+ step 3720000 | 36949s | train_loss=0.0110
375
+ step 3730000 | 37029s | train_loss=0.0115
376
+ step 3740000 | 37123s | train_loss=0.0112
377
+ step 3750000 | 37216s | train_loss=0.0109
378
+ step 3760000 | 37308s | train_loss=0.0120
379
+ step 3770000 | 37400s | train_loss=0.0114
380
+ step 3780000 | 37493s | train_loss=0.0109
381
+ step 3790000 | 37585s | train_loss=0.0109
382
+ step 3800000 | 37752s | train_loss=0.0108 | fid=43.9566
383
+ step 3810000 | 37845s | train_loss=0.0113
384
+ step 3820000 | 37939s | train_loss=0.0105
385
+ step 3830000 | 38029s | train_loss=0.0109
386
+ step 3840000 | 38123s | train_loss=0.0109
387
+ step 3850000 | 38214s | train_loss=0.0104
388
+ step 3860000 | 38308s | train_loss=0.0106
389
+ step 3870000 | 38400s | train_loss=0.0105
390
+ step 3880000 | 38493s | train_loss=0.0105
391
+ step 3890000 | 38585s | train_loss=0.0107
392
+ step 3900000 | 38752s | train_loss=0.0107 | fid=44.4928
393
+ step 3910000 | 38846s | train_loss=0.0103
394
+ step 3920000 | 38931s | train_loss=0.0120
395
+ step 3930000 | 39024s | train_loss=0.0108
396
+ step 3940000 | 39118s | train_loss=0.0104
397
+ step 3950000 | 39209s | train_loss=0.0105
398
+ step 3960000 | 39302s | train_loss=0.0103
399
+ step 3970000 | 39394s | train_loss=0.0109
400
+ step 3980000 | 39488s | train_loss=0.0105
401
+ step 3990000 | 39580s | train_loss=0.0101
402
+ step 4000000 | 39746s | train_loss=0.0101 | fid=43.8741
403
+ step 4010000 | 39839s | train_loss=0.0100
404
+ step 4020000 | 39929s | train_loss=0.0099
405
+ step 4030000 | 40023s | train_loss=0.0104
406
+ step 4040000 | 40114s | train_loss=0.0104
407
+ step 4050000 | 40207s | train_loss=0.0102
408
+ step 4060000 | 40301s | train_loss=0.0100
409
+ step 4070000 | 40393s | train_loss=0.0100
410
+ step 4080000 | 40486s | train_loss=0.0100
411
+ step 4090000 | 40578s | train_loss=0.0100
412
+ step 4100000 | 40746s | train_loss=0.0109 | fid=44.5205
413
+ step 4110000 | 40830s | train_loss=0.0103
414
+ step 4120000 | 40923s | train_loss=0.0099
415
+ step 4130000 | 41013s | train_loss=0.0106
416
+ step 4140000 | 41096s | train_loss=0.0099
417
+ step 4150000 | 41190s | train_loss=0.0104
418
+ step 4160000 | 41283s | train_loss=0.0103
419
+ step 4170000 | 41376s | train_loss=0.0098
420
+ step 4180000 | 41469s | train_loss=0.0098
421
+ step 4190000 | 41561s | train_loss=0.0096
422
+ step 4200000 | 41728s | train_loss=0.0099 | fid=44.7562
423
+ step 4210000 | 41808s | train_loss=0.0099
424
+ step 4220000 | 41901s | train_loss=0.0101
425
+ step 4230000 | 41990s | train_loss=0.0096
426
+ step 4240000 | 42073s | train_loss=0.0101
427
+ step 4250000 | 42167s | train_loss=0.0100
428
+ step 4260000 | 42261s | train_loss=0.0100
429
+ step 4270000 | 42354s | train_loss=0.0101
430
+ step 4280000 | 42447s | train_loss=0.0099
431
+ step 4290000 | 42538s | train_loss=0.0098
432
+ step 4300000 | 42704s | train_loss=0.0098 | fid=43.8175
433
+ step 4310000 | 42785s | train_loss=0.0102
434
+ step 4320000 | 42878s | train_loss=0.0099
435
+ step 4330000 | 42967s | train_loss=0.0098
436
+ step 4340000 | 43052s | train_loss=0.0097
437
+ step 4350000 | 43146s | train_loss=0.0101
438
+ step 4360000 | 43239s | train_loss=0.0093
439
+ step 4370000 | 43333s | train_loss=0.0098
440
+ step 4380000 | 43424s | train_loss=0.0096
441
+ step 4390000 | 43516s | train_loss=0.0098
442
+ step 4400000 | 43683s | train_loss=0.0098 | fid=44.5436
443
+ step 4410000 | 43763s | train_loss=0.0093
444
+ step 4420000 | 43856s | train_loss=0.0096
445
+ step 4430000 | 43944s | train_loss=0.0098
446
+ step 4440000 | 44031s | train_loss=0.0095
447
+ step 4450000 | 44125s | train_loss=0.0099
448
+ step 4460000 | 44218s | train_loss=0.0097
449
+ step 4470000 | 44311s | train_loss=0.0100
450
+ step 4480000 | 44403s | train_loss=0.0099
451
+ step 4490000 | 44494s | train_loss=0.0093
452
+ step 4500000 | 44661s | train_loss=0.0093 | fid=44.2214
453
+ step 4510000 | 44740s | train_loss=0.0100
454
+ step 4520000 | 44833s | train_loss=0.0095
455
+ step 4530000 | 44918s | train_loss=0.0092
456
+ step 4540000 | 45009s | train_loss=0.0098
457
+ step 4550000 | 45102s | train_loss=0.0091
458
+ step 4560000 | 45195s | train_loss=0.0092
459
+ step 4570000 | 45289s | train_loss=0.0093
460
+ step 4580000 | 45381s | train_loss=0.0091
461
+ step 4590000 | 45473s | train_loss=0.0092
462
+ step 4600000 | 45639s | train_loss=0.0100 | fid=43.9772
463
+ step 4610000 | 45729s | train_loss=0.0091
464
+ step 4620000 | 45822s | train_loss=0.0093
465
+ step 4630000 | 45914s | train_loss=0.0093
466
+ step 4640000 | 46007s | train_loss=0.0092
467
+ step 4650000 | 46101s | train_loss=0.0092
468
+ step 4660000 | 46195s | train_loss=0.0092
469
+ step 4670000 | 46289s | train_loss=0.0092
470
+ step 4680000 | 46382s | train_loss=0.0088
471
+ step 4690000 | 46474s | train_loss=0.0094
472
+ step 4700000 | 46639s | train_loss=0.0089 | fid=44.3874
473
+ step 4710000 | 46733s | train_loss=0.0086
474
+ step 4720000 | 46825s | train_loss=0.0094
475
+ step 4730000 | 46919s | train_loss=0.0094
476
+ step 4740000 | 47013s | train_loss=0.0090
477
+ step 4750000 | 47107s | train_loss=0.0088
478
+ step 4760000 | 47200s | train_loss=0.0086
479
+ step 4770000 | 47294s | train_loss=0.0089
480
+ step 4780000 | 47387s | train_loss=0.0088
481
+ step 4790000 | 47479s | train_loss=0.0090
482
+ step 4800000 | 47623s | train_loss=0.0089 | fid=44.0962
483
+ step 4810000 | 47698s | train_loss=0.0088
484
+ step 4820000 | 47756s | train_loss=0.0089
485
+ step 4830000 | 47814s | train_loss=0.0088
486
+ step 4840000 | 47872s | train_loss=0.0093
487
+ step 4850000 | 47930s | train_loss=0.0094
488
+ step 4860000 | 47988s | train_loss=0.0094
489
+ step 4870000 | 48046s | train_loss=0.0090
490
+ step 4880000 | 48102s | train_loss=0.0084
491
+ step 4890000 | 48156s | train_loss=0.0084
492
+ step 4900000 | 48265s | train_loss=0.0090 | fid=44.1033
493
+ step 4910000 | 48323s | train_loss=0.0088
494
+ step 4920000 | 48381s | train_loss=0.0090
495
+ step 4930000 | 48439s | train_loss=0.0086
496
+ step 4940000 | 48497s | train_loss=0.0089
497
+ step 4950000 | 48555s | train_loss=0.0094
498
+ step 4960000 | 48613s | train_loss=0.0088
499
+ step 4970000 | 48671s | train_loss=0.0082
500
+ step 4980000 | 48726s | train_loss=0.0093
501
+ step 4990000 | 48764s | train_loss=0.0089
502
+ step 5000000 | 48841s | train_loss=0.0089 | fid=44.3534
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/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/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": 42,
25
+ "device": null
26
+ }
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/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/d70/runner.log ADDED
@@ -0,0 +1,502 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ FID real-side stats cached over 1000 diverse train images
2
+ step 1 | 68s | train_loss=0.9935 | fid=281.8394
3
+ step 10000 | 147s | train_loss=0.8209
4
+ step 20000 | 234s | train_loss=0.7272
5
+ step 30000 | 321s | train_loss=0.6008
6
+ step 40000 | 409s | train_loss=0.4381
7
+ step 50000 | 496s | train_loss=0.2978
8
+ step 60000 | 583s | train_loss=0.2169
9
+ step 70000 | 670s | train_loss=0.1770
10
+ step 80000 | 757s | train_loss=0.1536
11
+ step 90000 | 845s | train_loss=0.1406
12
+ step 100000 | 1000s | train_loss=0.1293 | fid=55.6227
13
+ step 110000 | 1082s | train_loss=0.1192
14
+ step 120000 | 1169s | train_loss=0.1134
15
+ step 130000 | 1256s | train_loss=0.1075
16
+ step 140000 | 1343s | train_loss=0.1030
17
+ step 150000 | 1431s | train_loss=0.0984
18
+ step 160000 | 1518s | train_loss=0.0945
19
+ step 170000 | 1605s | train_loss=0.0904
20
+ step 180000 | 1692s | train_loss=0.0886
21
+ step 190000 | 1780s | train_loss=0.0843
22
+ step 200000 | 1935s | train_loss=0.0827 | fid=54.4891
23
+ step 210000 | 2020s | train_loss=0.0796
24
+ step 220000 | 2103s | train_loss=0.0773
25
+ step 230000 | 2191s | train_loss=0.0746
26
+ step 240000 | 2278s | train_loss=0.0757
27
+ step 250000 | 2365s | train_loss=0.0714
28
+ step 260000 | 2452s | train_loss=0.0698
29
+ step 270000 | 2540s | train_loss=0.0674
30
+ step 280000 | 2627s | train_loss=0.0664
31
+ step 290000 | 2714s | train_loss=0.0656
32
+ step 300000 | 2869s | train_loss=0.0636 | fid=54.2569
33
+ step 310000 | 2954s | train_loss=0.0622
34
+ step 320000 | 3042s | train_loss=0.0624
35
+ step 330000 | 3125s | train_loss=0.0594
36
+ step 340000 | 3212s | train_loss=0.0596
37
+ step 350000 | 3299s | train_loss=0.0580
38
+ step 360000 | 3387s | train_loss=0.0583
39
+ step 370000 | 3474s | train_loss=0.0558
40
+ step 380000 | 3561s | train_loss=0.0557
41
+ step 390000 | 3648s | train_loss=0.0545
42
+ step 400000 | 3804s | train_loss=0.0536 | fid=50.5963
43
+ step 410000 | 3889s | train_loss=0.0526
44
+ step 420000 | 3976s | train_loss=0.0531
45
+ step 430000 | 4063s | train_loss=0.0514
46
+ step 440000 | 4147s | train_loss=0.0522
47
+ step 450000 | 4234s | train_loss=0.0508
48
+ step 460000 | 4322s | train_loss=0.0507
49
+ step 470000 | 4409s | train_loss=0.0492
50
+ step 480000 | 4496s | train_loss=0.0478
51
+ step 490000 | 4583s | train_loss=0.0483
52
+ step 500000 | 4740s | train_loss=0.0479 | fid=49.9973
53
+ step 510000 | 4825s | train_loss=0.0470
54
+ step 520000 | 4912s | train_loss=0.0470
55
+ step 530000 | 4999s | train_loss=0.0461
56
+ step 540000 | 5084s | train_loss=0.0452
57
+ step 550000 | 5170s | train_loss=0.0453
58
+ step 560000 | 5257s | train_loss=0.0456
59
+ step 570000 | 5344s | train_loss=0.0444
60
+ step 580000 | 5432s | train_loss=0.0438
61
+ step 590000 | 5519s | train_loss=0.0443
62
+ step 600000 | 5676s | train_loss=0.0429 | fid=48.6129
63
+ step 610000 | 5760s | train_loss=0.0433
64
+ step 620000 | 5847s | train_loss=0.0430
65
+ step 630000 | 5935s | train_loss=0.0426
66
+ step 640000 | 6022s | train_loss=0.0420
67
+ step 650000 | 6106s | train_loss=0.0418
68
+ step 660000 | 6192s | train_loss=0.0418
69
+ step 670000 | 6280s | train_loss=0.0420
70
+ step 680000 | 6367s | train_loss=0.0409
71
+ step 690000 | 6454s | train_loss=0.0407
72
+ step 700000 | 6610s | train_loss=0.0405 | fid=47.2086
73
+ step 710000 | 6694s | train_loss=0.0386
74
+ step 720000 | 6781s | train_loss=0.0391
75
+ step 730000 | 6869s | train_loss=0.0396
76
+ step 740000 | 6956s | train_loss=0.0396
77
+ step 750000 | 7042s | train_loss=0.0384
78
+ step 760000 | 7128s | train_loss=0.0385
79
+ step 770000 | 7214s | train_loss=0.0383
80
+ step 780000 | 7301s | train_loss=0.0380
81
+ step 790000 | 7388s | train_loss=0.0379
82
+ step 800000 | 7544s | train_loss=0.0370 | fid=46.5227
83
+ step 810000 | 7628s | train_loss=0.0368
84
+ step 820000 | 7716s | train_loss=0.0364
85
+ step 830000 | 7803s | train_loss=0.0364
86
+ step 840000 | 7890s | train_loss=0.0369
87
+ step 850000 | 7978s | train_loss=0.0357
88
+ step 860000 | 8064s | train_loss=0.0357
89
+ step 870000 | 8150s | train_loss=0.0366
90
+ step 880000 | 8236s | train_loss=0.0353
91
+ step 890000 | 8323s | train_loss=0.0350
92
+ step 900000 | 8481s | train_loss=0.0351 | fid=46.4474
93
+ step 910000 | 8565s | train_loss=0.0344
94
+ step 920000 | 8653s | train_loss=0.0341
95
+ step 930000 | 8740s | train_loss=0.0337
96
+ step 940000 | 8827s | train_loss=0.0342
97
+ step 950000 | 8914s | train_loss=0.0335
98
+ step 960000 | 9002s | train_loss=0.0330
99
+ step 970000 | 9086s | train_loss=0.0336
100
+ step 980000 | 9173s | train_loss=0.0329
101
+ step 990000 | 9259s | train_loss=0.0327
102
+ step 1000000 | 9418s | train_loss=0.0323 | fid=45.3521
103
+ step 1010000 | 9502s | train_loss=0.0332
104
+ step 1020000 | 9589s | train_loss=0.0318
105
+ step 1030000 | 9677s | train_loss=0.0328
106
+ step 1040000 | 9764s | train_loss=0.0314
107
+ step 1050000 | 9851s | train_loss=0.0325
108
+ step 1060000 | 9938s | train_loss=0.0313
109
+ step 1070000 | 10025s | train_loss=0.0321
110
+ step 1080000 | 10110s | train_loss=0.0317
111
+ step 1090000 | 10198s | train_loss=0.0313
112
+ step 1100000 | 10355s | train_loss=0.0306 | fid=44.6679
113
+ step 1110000 | 10439s | train_loss=0.0304
114
+ step 1120000 | 10527s | train_loss=0.0303
115
+ step 1130000 | 10614s | train_loss=0.0305
116
+ step 1140000 | 10701s | train_loss=0.0300
117
+ step 1150000 | 10788s | train_loss=0.0304
118
+ step 1160000 | 10876s | train_loss=0.0296
119
+ step 1170000 | 10963s | train_loss=0.0292
120
+ step 1180000 | 11049s | train_loss=0.0295
121
+ step 1190000 | 11135s | train_loss=0.0295
122
+ step 1200000 | 11290s | train_loss=0.0289 | fid=45.5062
123
+ step 1210000 | 11374s | train_loss=0.0290
124
+ step 1220000 | 11462s | train_loss=0.0288
125
+ step 1230000 | 11549s | train_loss=0.0287
126
+ step 1240000 | 11636s | train_loss=0.0294
127
+ step 1250000 | 11723s | train_loss=0.0286
128
+ step 1260000 | 11811s | train_loss=0.0286
129
+ step 1270000 | 11898s | train_loss=0.0283
130
+ step 1280000 | 11985s | train_loss=0.0281
131
+ step 1290000 | 12072s | train_loss=0.0281
132
+ step 1300000 | 12226s | train_loss=0.0283 | fid=44.5253
133
+ step 1310000 | 12309s | train_loss=0.0269
134
+ step 1320000 | 12396s | train_loss=0.0273
135
+ step 1330000 | 12483s | train_loss=0.0268
136
+ step 1340000 | 12570s | train_loss=0.0270
137
+ step 1350000 | 12658s | train_loss=0.0268
138
+ step 1360000 | 12745s | train_loss=0.0279
139
+ step 1370000 | 12832s | train_loss=0.0271
140
+ step 1380000 | 12919s | train_loss=0.0263
141
+ step 1390000 | 13007s | train_loss=0.0269
142
+ step 1400000 | 13161s | train_loss=0.0261 | fid=43.0239
143
+ step 1410000 | 13246s | train_loss=0.0255
144
+ step 1420000 | 13331s | train_loss=0.0265
145
+ step 1430000 | 13419s | train_loss=0.0266
146
+ step 1440000 | 13506s | train_loss=0.0258
147
+ step 1450000 | 13593s | train_loss=0.0257
148
+ step 1460000 | 13680s | train_loss=0.0256
149
+ step 1470000 | 13768s | train_loss=0.0246
150
+ step 1480000 | 13855s | train_loss=0.0248
151
+ step 1490000 | 13942s | train_loss=0.0249
152
+ step 1500000 | 14099s | train_loss=0.0244 | fid=42.7577
153
+ step 1510000 | 14179s | train_loss=0.0256
154
+ step 1520000 | 14261s | train_loss=0.0250
155
+ step 1530000 | 14346s | train_loss=0.0243
156
+ step 1540000 | 14434s | train_loss=0.0242
157
+ step 1550000 | 14521s | train_loss=0.0241
158
+ step 1560000 | 14608s | train_loss=0.0251
159
+ step 1570000 | 14695s | train_loss=0.0234
160
+ step 1580000 | 14783s | train_loss=0.0248
161
+ step 1590000 | 14870s | train_loss=0.0238
162
+ step 1600000 | 15058s | train_loss=0.0236 | fid=43.0851
163
+ step 1610000 | 15141s | train_loss=0.0239
164
+ step 1620000 | 15228s | train_loss=0.0244
165
+ step 1630000 | 15314s | train_loss=0.0236
166
+ step 1640000 | 15402s | train_loss=0.0241
167
+ step 1650000 | 15489s | train_loss=0.0230
168
+ step 1660000 | 15576s | train_loss=0.0230
169
+ step 1670000 | 15663s | train_loss=0.0229
170
+ step 1680000 | 15750s | train_loss=0.0229
171
+ step 1690000 | 15838s | train_loss=0.0225
172
+ step 1700000 | 15995s | train_loss=0.0225 | fid=42.5724
173
+ step 1710000 | 16079s | train_loss=0.0238
174
+ step 1720000 | 16165s | train_loss=0.0226
175
+ step 1730000 | 16253s | train_loss=0.0220
176
+ step 1740000 | 16339s | train_loss=0.0228
177
+ step 1750000 | 16426s | train_loss=0.0228
178
+ step 1760000 | 16513s | train_loss=0.0227
179
+ step 1770000 | 16600s | train_loss=0.0222
180
+ step 1780000 | 16688s | train_loss=0.0213
181
+ step 1790000 | 16775s | train_loss=0.0217
182
+ step 1800000 | 16935s | train_loss=0.0217 | fid=42.9757
183
+ step 1810000 | 17018s | train_loss=0.0214
184
+ step 1820000 | 17103s | train_loss=0.0216
185
+ step 1830000 | 17190s | train_loss=0.0215
186
+ step 1840000 | 17278s | train_loss=0.0212
187
+ step 1850000 | 17364s | train_loss=0.0210
188
+ step 1860000 | 17451s | train_loss=0.0213
189
+ step 1870000 | 17538s | train_loss=0.0210
190
+ step 1880000 | 17625s | train_loss=0.0215
191
+ step 1890000 | 17712s | train_loss=0.0212
192
+ step 1900000 | 17872s | train_loss=0.0212 | fid=41.9301
193
+ step 1910000 | 17957s | train_loss=0.0206
194
+ step 1920000 | 18042s | train_loss=0.0219
195
+ step 1930000 | 18127s | train_loss=0.0200
196
+ step 1940000 | 18214s | train_loss=0.0206
197
+ step 1950000 | 18302s | train_loss=0.0207
198
+ step 1960000 | 18388s | train_loss=0.0201
199
+ step 1970000 | 18475s | train_loss=0.0202
200
+ step 1980000 | 18562s | train_loss=0.0204
201
+ step 1990000 | 18649s | train_loss=0.0200
202
+ step 2000000 | 18808s | train_loss=0.0201 | fid=42.6749
203
+ step 2010000 | 18895s | train_loss=0.0199
204
+ step 2020000 | 18978s | train_loss=0.0202
205
+ step 2030000 | 19064s | train_loss=0.0201
206
+ step 2040000 | 19152s | train_loss=0.0194
207
+ step 2050000 | 19239s | train_loss=0.0202
208
+ step 2060000 | 19327s | train_loss=0.0203
209
+ step 2070000 | 19412s | train_loss=0.0201
210
+ step 2080000 | 19499s | train_loss=0.0192
211
+ step 2090000 | 19586s | train_loss=0.0194
212
+ step 2100000 | 19771s | train_loss=0.0192 | fid=42.5819
213
+ step 2110000 | 19852s | train_loss=0.0192
214
+ step 2120000 | 19933s | train_loss=0.0195
215
+ step 2130000 | 20020s | train_loss=0.0187
216
+ step 2140000 | 20106s | train_loss=0.0185
217
+ step 2150000 | 20193s | train_loss=0.0189
218
+ step 2160000 | 20280s | train_loss=0.0192
219
+ step 2170000 | 20366s | train_loss=0.0185
220
+ step 2180000 | 20454s | train_loss=0.0187
221
+ step 2190000 | 20541s | train_loss=0.0192
222
+ step 2200000 | 20723s | train_loss=0.0180 | fid=41.5569
223
+ step 2210000 | 20807s | train_loss=0.0181
224
+ step 2220000 | 20890s | train_loss=0.0182
225
+ step 2230000 | 20976s | train_loss=0.0188
226
+ step 2240000 | 21062s | train_loss=0.0183
227
+ step 2250000 | 21149s | train_loss=0.0186
228
+ step 2260000 | 21236s | train_loss=0.0189
229
+ step 2270000 | 21324s | train_loss=0.0176
230
+ step 2280000 | 21410s | train_loss=0.0180
231
+ step 2290000 | 21497s | train_loss=0.0180
232
+ step 2300000 | 21693s | train_loss=0.0182 | fid=42.3320
233
+ step 2310000 | 21773s | train_loss=0.0186
234
+ step 2320000 | 21854s | train_loss=0.0178
235
+ step 2330000 | 21940s | train_loss=0.0180
236
+ step 2340000 | 22026s | train_loss=0.0179
237
+ step 2350000 | 22113s | train_loss=0.0187
238
+ step 2360000 | 22201s | train_loss=0.0178
239
+ step 2370000 | 22288s | train_loss=0.0169
240
+ step 2380000 | 22374s | train_loss=0.0176
241
+ step 2390000 | 22461s | train_loss=0.0174
242
+ step 2400000 | 22639s | train_loss=0.0176 | fid=41.5288
243
+ step 2410000 | 22726s | train_loss=0.0170
244
+ step 2420000 | 22800s | train_loss=0.0173
245
+ step 2430000 | 22886s | train_loss=0.0171
246
+ step 2440000 | 22973s | train_loss=0.0166
247
+ step 2450000 | 23059s | train_loss=0.0170
248
+ step 2460000 | 23146s | train_loss=0.0175
249
+ step 2470000 | 23233s | train_loss=0.0171
250
+ step 2480000 | 23321s | train_loss=0.0173
251
+ step 2490000 | 23407s | train_loss=0.0169
252
+ step 2500000 | 23595s | train_loss=0.0170 | fid=42.0099
253
+ step 2510000 | 23682s | train_loss=0.0170
254
+ step 2520000 | 23756s | train_loss=0.0164
255
+ step 2530000 | 23842s | train_loss=0.0164
256
+ step 2540000 | 23928s | train_loss=0.0167
257
+ step 2550000 | 24013s | train_loss=0.0164
258
+ step 2560000 | 24101s | train_loss=0.0169
259
+ step 2570000 | 24188s | train_loss=0.0165
260
+ step 2580000 | 24275s | train_loss=0.0166
261
+ step 2590000 | 24361s | train_loss=0.0169
262
+ step 2600000 | 24546s | train_loss=0.0161 | fid=41.7307
263
+ step 2610000 | 24633s | train_loss=0.0166
264
+ step 2620000 | 24714s | train_loss=0.0161
265
+ step 2630000 | 24796s | train_loss=0.0155
266
+ step 2640000 | 24881s | train_loss=0.0164
267
+ step 2650000 | 24967s | train_loss=0.0165
268
+ step 2660000 | 25055s | train_loss=0.0163
269
+ step 2670000 | 25142s | train_loss=0.0165
270
+ step 2680000 | 25229s | train_loss=0.0157
271
+ step 2690000 | 25317s | train_loss=0.0156
272
+ step 2700000 | 25513s | train_loss=0.0158 | fid=41.9250
273
+ step 2710000 | 25601s | train_loss=0.0156
274
+ step 2720000 | 25682s | train_loss=0.0154
275
+ step 2730000 | 25760s | train_loss=0.0158
276
+ step 2740000 | 25846s | train_loss=0.0155
277
+ step 2750000 | 25932s | train_loss=0.0157
278
+ step 2760000 | 26020s | train_loss=0.0153
279
+ step 2770000 | 26107s | train_loss=0.0157
280
+ step 2780000 | 26194s | train_loss=0.0154
281
+ step 2790000 | 26282s | train_loss=0.0152
282
+ step 2800000 | 26474s | train_loss=0.0152 | fid=41.3415
283
+ step 2810000 | 26561s | train_loss=0.0147
284
+ step 2820000 | 26647s | train_loss=0.0159
285
+ step 2830000 | 26725s | train_loss=0.0150
286
+ step 2840000 | 26811s | train_loss=0.0152
287
+ step 2850000 | 26899s | train_loss=0.0146
288
+ step 2860000 | 26984s | train_loss=0.0152
289
+ step 2870000 | 27071s | train_loss=0.0158
290
+ step 2880000 | 27158s | train_loss=0.0146
291
+ step 2890000 | 27245s | train_loss=0.0150
292
+ step 2900000 | 27453s | train_loss=0.0147 | fid=42.3971
293
+ step 2910000 | 27540s | train_loss=0.0145
294
+ step 2920000 | 27622s | train_loss=0.0145
295
+ step 2930000 | 27706s | train_loss=0.0152
296
+ step 2940000 | 27792s | train_loss=0.0144
297
+ step 2950000 | 27880s | train_loss=0.0148
298
+ step 2960000 | 27965s | train_loss=0.0142
299
+ step 2970000 | 28052s | train_loss=0.0147
300
+ step 2980000 | 28140s | train_loss=0.0141
301
+ step 2990000 | 28227s | train_loss=0.0149
302
+ step 3000000 | 28434s | train_loss=0.0142 | fid=41.4917
303
+ step 3010000 | 28521s | train_loss=0.0143
304
+ step 3020000 | 28597s | train_loss=0.0140
305
+ step 3030000 | 28681s | train_loss=0.0145
306
+ step 3040000 | 28767s | train_loss=0.0142
307
+ step 3050000 | 28855s | train_loss=0.0142
308
+ step 3060000 | 28940s | train_loss=0.0148
309
+ step 3070000 | 29028s | train_loss=0.0139
310
+ step 3080000 | 29115s | train_loss=0.0142
311
+ step 3090000 | 29202s | train_loss=0.0141
312
+ step 3100000 | 29397s | train_loss=0.0143 | fid=42.2806
313
+ step 3110000 | 29485s | train_loss=0.0137
314
+ step 3120000 | 29564s | train_loss=0.0138
315
+ step 3130000 | 29651s | train_loss=0.0138
316
+ step 3140000 | 29737s | train_loss=0.0138
317
+ step 3150000 | 29824s | train_loss=0.0134
318
+ step 3160000 | 29910s | train_loss=0.0137
319
+ step 3170000 | 29997s | train_loss=0.0133
320
+ step 3180000 | 30084s | train_loss=0.0135
321
+ step 3190000 | 30172s | train_loss=0.0135
322
+ step 3200000 | 30355s | train_loss=0.0139 | fid=41.6479
323
+ step 3210000 | 30442s | train_loss=0.0137
324
+ step 3220000 | 30522s | train_loss=0.0135
325
+ step 3230000 | 30610s | train_loss=0.0133
326
+ step 3240000 | 30696s | train_loss=0.0135
327
+ step 3250000 | 30783s | train_loss=0.0138
328
+ step 3260000 | 30869s | train_loss=0.0137
329
+ step 3270000 | 30956s | train_loss=0.0135
330
+ step 3280000 | 31043s | train_loss=0.0134
331
+ step 3290000 | 31130s | train_loss=0.0137
332
+ step 3300000 | 31289s | train_loss=0.0137 | fid=41.9527
333
+ step 3310000 | 31375s | train_loss=0.0136
334
+ step 3320000 | 31460s | train_loss=0.0129
335
+ step 3330000 | 31547s | train_loss=0.0134
336
+ step 3340000 | 31634s | train_loss=0.0133
337
+ step 3350000 | 31720s | train_loss=0.0138
338
+ step 3360000 | 31808s | train_loss=0.0131
339
+ step 3370000 | 31894s | train_loss=0.0142
340
+ step 3380000 | 31981s | train_loss=0.0126
341
+ step 3390000 | 32068s | train_loss=0.0134
342
+ step 3400000 | 32227s | train_loss=0.0129 | fid=41.9957
343
+ step 3410000 | 32315s | train_loss=0.0127
344
+ step 3420000 | 32398s | train_loss=0.0125
345
+ step 3430000 | 32476s | train_loss=0.0131
346
+ step 3440000 | 32563s | train_loss=0.0125
347
+ step 3450000 | 32651s | train_loss=0.0124
348
+ step 3460000 | 32736s | train_loss=0.0128
349
+ step 3470000 | 32824s | train_loss=0.0129
350
+ step 3480000 | 32910s | train_loss=0.0129
351
+ step 3490000 | 32997s | train_loss=0.0127
352
+ step 3500000 | 33176s | train_loss=0.0127 | fid=41.7986
353
+ step 3510000 | 33263s | train_loss=0.0124
354
+ step 3520000 | 33349s | train_loss=0.0124
355
+ step 3530000 | 33423s | train_loss=0.0128
356
+ step 3540000 | 33511s | train_loss=0.0121
357
+ step 3550000 | 33598s | train_loss=0.0121
358
+ step 3560000 | 33684s | train_loss=0.0124
359
+ step 3570000 | 33771s | train_loss=0.0123
360
+ step 3580000 | 33857s | train_loss=0.0124
361
+ step 3590000 | 33944s | train_loss=0.0124
362
+ step 3600000 | 34139s | train_loss=0.0132 | fid=41.6852
363
+ step 3610000 | 34226s | train_loss=0.0117
364
+ step 3620000 | 34314s | train_loss=0.0125
365
+ step 3630000 | 34387s | train_loss=0.0125
366
+ step 3640000 | 34472s | train_loss=0.0122
367
+ step 3650000 | 34560s | train_loss=0.0126
368
+ step 3660000 | 34646s | train_loss=0.0123
369
+ step 3670000 | 34733s | train_loss=0.0119
370
+ step 3680000 | 34819s | train_loss=0.0123
371
+ step 3690000 | 34906s | train_loss=0.0122
372
+ step 3700000 | 35110s | train_loss=0.0121 | fid=42.5351
373
+ step 3710000 | 35198s | train_loss=0.0117
374
+ step 3720000 | 35285s | train_loss=0.0115
375
+ step 3730000 | 35359s | train_loss=0.0123
376
+ step 3740000 | 35445s | train_loss=0.0118
377
+ step 3750000 | 35532s | train_loss=0.0122
378
+ step 3760000 | 35618s | train_loss=0.0123
379
+ step 3770000 | 35706s | train_loss=0.0124
380
+ step 3780000 | 35791s | train_loss=0.0121
381
+ step 3790000 | 35879s | train_loss=0.0118
382
+ step 3800000 | 36062s | train_loss=0.0117 | fid=41.2504
383
+ step 3810000 | 36149s | train_loss=0.0119
384
+ step 3820000 | 36236s | train_loss=0.0116
385
+ step 3830000 | 36315s | train_loss=0.0118
386
+ step 3840000 | 36396s | train_loss=0.0112
387
+ step 3850000 | 36483s | train_loss=0.0115
388
+ step 3860000 | 36571s | train_loss=0.0115
389
+ step 3870000 | 36656s | train_loss=0.0115
390
+ step 3880000 | 36744s | train_loss=0.0112
391
+ step 3890000 | 36830s | train_loss=0.0116
392
+ step 3900000 | 37031s | train_loss=0.0117 | fid=41.4824
393
+ step 3910000 | 37118s | train_loss=0.0110
394
+ step 3920000 | 37205s | train_loss=0.0122
395
+ step 3930000 | 37290s | train_loss=0.0112
396
+ step 3940000 | 37376s | train_loss=0.0111
397
+ step 3950000 | 37463s | train_loss=0.0119
398
+ step 3960000 | 37551s | train_loss=0.0110
399
+ step 3970000 | 37637s | train_loss=0.0129
400
+ step 3980000 | 37724s | train_loss=0.0114
401
+ step 3990000 | 37810s | train_loss=0.0111
402
+ step 4000000 | 37970s | train_loss=0.0110 | fid=41.1628
403
+ step 4010000 | 38057s | train_loss=0.0111
404
+ step 4020000 | 38145s | train_loss=0.0112
405
+ step 4030000 | 38229s | train_loss=0.0113
406
+ step 4040000 | 38317s | train_loss=0.0114
407
+ step 4050000 | 38402s | train_loss=0.0112
408
+ step 4060000 | 38489s | train_loss=0.0111
409
+ step 4070000 | 38575s | train_loss=0.0110
410
+ step 4080000 | 38662s | train_loss=0.0111
411
+ step 4090000 | 38749s | train_loss=0.0107
412
+ step 4100000 | 38931s | train_loss=0.0112 | fid=41.6378
413
+ step 4110000 | 39018s | train_loss=0.0109
414
+ step 4120000 | 39105s | train_loss=0.0113
415
+ step 4130000 | 39190s | train_loss=0.0116
416
+ step 4140000 | 39277s | train_loss=0.0109
417
+ step 4150000 | 39363s | train_loss=0.0109
418
+ step 4160000 | 39450s | train_loss=0.0109
419
+ step 4170000 | 39538s | train_loss=0.0104
420
+ step 4180000 | 39623s | train_loss=0.0106
421
+ step 4190000 | 39711s | train_loss=0.0107
422
+ step 4200000 | 39868s | train_loss=0.0108 | fid=41.9838
423
+ step 4210000 | 39956s | train_loss=0.0107
424
+ step 4220000 | 40043s | train_loss=0.0111
425
+ step 4230000 | 40127s | train_loss=0.0107
426
+ step 4240000 | 40214s | train_loss=0.0108
427
+ step 4250000 | 40302s | train_loss=0.0110
428
+ step 4260000 | 40387s | train_loss=0.0111
429
+ step 4270000 | 40475s | train_loss=0.0108
430
+ step 4280000 | 40561s | train_loss=0.0106
431
+ step 4290000 | 40648s | train_loss=0.0107
432
+ step 4300000 | 40831s | train_loss=0.0109 | fid=42.1153
433
+ step 4310000 | 40919s | train_loss=0.0112
434
+ step 4320000 | 41004s | train_loss=0.0105
435
+ step 4330000 | 41080s | train_loss=0.0102
436
+ step 4340000 | 41167s | train_loss=0.0112
437
+ step 4350000 | 41255s | train_loss=0.0106
438
+ step 4360000 | 41342s | train_loss=0.0103
439
+ step 4370000 | 41428s | train_loss=0.0105
440
+ step 4380000 | 41515s | train_loss=0.0106
441
+ step 4390000 | 41601s | train_loss=0.0104
442
+ step 4400000 | 41803s | train_loss=0.0109 | fid=42.1812
443
+ step 4410000 | 41891s | train_loss=0.0102
444
+ step 4420000 | 41976s | train_loss=0.0108
445
+ step 4430000 | 42050s | train_loss=0.0105
446
+ step 4440000 | 42138s | train_loss=0.0099
447
+ step 4450000 | 42225s | train_loss=0.0109
448
+ step 4460000 | 42312s | train_loss=0.0102
449
+ step 4470000 | 42398s | train_loss=0.0101
450
+ step 4480000 | 42486s | train_loss=0.0104
451
+ step 4490000 | 42571s | train_loss=0.0098
452
+ step 4500000 | 42772s | train_loss=0.0101 | fid=42.0289
453
+ step 4510000 | 42859s | train_loss=0.0110
454
+ step 4520000 | 42946s | train_loss=0.0103
455
+ step 4530000 | 43021s | train_loss=0.0099
456
+ step 4540000 | 43108s | train_loss=0.0104
457
+ step 4550000 | 43195s | train_loss=0.0099
458
+ step 4560000 | 43282s | train_loss=0.0103
459
+ step 4570000 | 43368s | train_loss=0.0101
460
+ step 4580000 | 43456s | train_loss=0.0095
461
+ step 4590000 | 43541s | train_loss=0.0102
462
+ step 4600000 | 43746s | train_loss=0.0105 | fid=42.3099
463
+ step 4610000 | 43833s | train_loss=0.0106
464
+ step 4620000 | 43921s | train_loss=0.0099
465
+ step 4630000 | 43996s | train_loss=0.0101
466
+ step 4640000 | 44083s | train_loss=0.0104
467
+ step 4650000 | 44170s | train_loss=0.0104
468
+ step 4660000 | 44258s | train_loss=0.0102
469
+ step 4670000 | 44343s | train_loss=0.0100
470
+ step 4680000 | 44431s | train_loss=0.0095
471
+ step 4690000 | 44516s | train_loss=0.0099
472
+ step 4700000 | 44725s | train_loss=0.0097 | fid=41.8264
473
+ step 4710000 | 44813s | train_loss=0.0099
474
+ step 4720000 | 44897s | train_loss=0.0105
475
+ step 4730000 | 44976s | train_loss=0.0097
476
+ step 4740000 | 45063s | train_loss=0.0099
477
+ step 4750000 | 45150s | train_loss=0.0095
478
+ step 4760000 | 45238s | train_loss=0.0098
479
+ step 4770000 | 45325s | train_loss=0.0092
480
+ step 4780000 | 45411s | train_loss=0.0097
481
+ step 4790000 | 45497s | train_loss=0.0095
482
+ step 4800000 | 45655s | train_loss=0.0099 | fid=42.4347
483
+ step 4810000 | 45742s | train_loss=0.0098
484
+ step 4820000 | 45830s | train_loss=0.0099
485
+ step 4830000 | 45915s | train_loss=0.0096
486
+ step 4840000 | 46002s | train_loss=0.0096
487
+ step 4850000 | 46089s | train_loss=0.0094
488
+ step 4860000 | 46176s | train_loss=0.0102
489
+ step 4870000 | 46264s | train_loss=0.0098
490
+ step 4880000 | 46351s | train_loss=0.0094
491
+ step 4890000 | 46436s | train_loss=0.0092
492
+ step 4900000 | 46591s | train_loss=0.0096 | fid=42.3681
493
+ step 4910000 | 46677s | train_loss=0.0097
494
+ step 4920000 | 46765s | train_loss=0.0101
495
+ step 4930000 | 46850s | train_loss=0.0101
496
+ step 4940000 | 46937s | train_loss=0.0097
497
+ step 4950000 | 47024s | train_loss=0.0094
498
+ step 4960000 | 47112s | train_loss=0.0097
499
+ step 4970000 | 47199s | train_loss=0.0092
500
+ step 4980000 | 47286s | train_loss=0.0097
501
+ step 4990000 | 47372s | train_loss=0.0098
502
+ step 5000000 | 47527s | train_loss=0.0094 | fid=41.6137
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/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/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": 42,
25
+ "device": null
26
+ }
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/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/d80/runner.log ADDED
@@ -0,0 +1,502 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ FID real-side stats cached over 1000 diverse train images
2
+ step 1 | 72s | train_loss=0.9937 | fid=273.5214
3
+ step 10000 | 165s | train_loss=0.8201
4
+ step 20000 | 258s | train_loss=0.7163
5
+ step 30000 | 350s | train_loss=0.5850
6
+ step 40000 | 443s | train_loss=0.4292
7
+ step 50000 | 536s | train_loss=0.2958
8
+ step 60000 | 629s | train_loss=0.2170
9
+ step 70000 | 722s | train_loss=0.1758
10
+ step 80000 | 815s | train_loss=0.1524
11
+ step 90000 | 908s | train_loss=0.1384
12
+ step 100000 | 1069s | train_loss=0.1251 | fid=55.9376
13
+ step 110000 | 1158s | train_loss=0.1197
14
+ step 120000 | 1251s | train_loss=0.1130
15
+ step 130000 | 1344s | train_loss=0.1094
16
+ step 140000 | 1437s | train_loss=0.1047
17
+ step 150000 | 1530s | train_loss=0.0991
18
+ step 160000 | 1623s | train_loss=0.0965
19
+ step 170000 | 1716s | train_loss=0.0918
20
+ step 180000 | 1809s | train_loss=0.0887
21
+ step 190000 | 1902s | train_loss=0.0848
22
+ step 200000 | 2062s | train_loss=0.0831 | fid=54.4494
23
+ step 210000 | 2152s | train_loss=0.0803
24
+ step 220000 | 2245s | train_loss=0.0808
25
+ step 230000 | 2338s | train_loss=0.0767
26
+ step 240000 | 2431s | train_loss=0.0759
27
+ step 250000 | 2525s | train_loss=0.0733
28
+ step 260000 | 2618s | train_loss=0.0726
29
+ step 270000 | 2711s | train_loss=0.0704
30
+ step 280000 | 2804s | train_loss=0.0691
31
+ step 290000 | 2892s | train_loss=0.0665
32
+ step 300000 | 3058s | train_loss=0.0653 | fid=52.6953
33
+ step 310000 | 3148s | train_loss=0.0649
34
+ step 320000 | 3242s | train_loss=0.0631
35
+ step 330000 | 3335s | train_loss=0.0619
36
+ step 340000 | 3428s | train_loss=0.0608
37
+ step 350000 | 3521s | train_loss=0.0610
38
+ step 360000 | 3614s | train_loss=0.0599
39
+ step 370000 | 3707s | train_loss=0.0576
40
+ step 380000 | 3799s | train_loss=0.0571
41
+ step 390000 | 3888s | train_loss=0.0570
42
+ step 400000 | 4055s | train_loss=0.0562 | fid=51.2644
43
+ step 410000 | 4145s | train_loss=0.0562
44
+ step 420000 | 4238s | train_loss=0.0550
45
+ step 430000 | 4332s | train_loss=0.0543
46
+ step 440000 | 4425s | train_loss=0.0535
47
+ step 450000 | 4518s | train_loss=0.0536
48
+ step 460000 | 4611s | train_loss=0.0518
49
+ step 470000 | 4703s | train_loss=0.0511
50
+ step 480000 | 4791s | train_loss=0.0507
51
+ step 490000 | 4884s | train_loss=0.0505
52
+ step 500000 | 5051s | train_loss=0.0505 | fid=49.5847
53
+ step 510000 | 5143s | train_loss=0.0491
54
+ step 520000 | 5234s | train_loss=0.0487
55
+ step 530000 | 5327s | train_loss=0.0491
56
+ step 540000 | 5420s | train_loss=0.0486
57
+ step 550000 | 5513s | train_loss=0.0488
58
+ step 560000 | 5606s | train_loss=0.0473
59
+ step 570000 | 5696s | train_loss=0.0479
60
+ step 580000 | 5786s | train_loss=0.0467
61
+ step 590000 | 5880s | train_loss=0.0460
62
+ step 600000 | 6047s | train_loss=0.0454 | fid=48.3462
63
+ step 610000 | 6138s | train_loss=0.0455
64
+ step 620000 | 6230s | train_loss=0.0456
65
+ step 630000 | 6323s | train_loss=0.0435
66
+ step 640000 | 6416s | train_loss=0.0452
67
+ step 650000 | 6509s | train_loss=0.0445
68
+ step 660000 | 6602s | train_loss=0.0433
69
+ step 670000 | 6689s | train_loss=0.0433
70
+ step 680000 | 6783s | train_loss=0.0439
71
+ step 690000 | 6876s | train_loss=0.0422
72
+ step 700000 | 7043s | train_loss=0.0426 | fid=47.0411
73
+ step 710000 | 7134s | train_loss=0.0408
74
+ step 720000 | 7226s | train_loss=0.0413
75
+ step 730000 | 7319s | train_loss=0.0415
76
+ step 740000 | 7412s | train_loss=0.0416
77
+ step 750000 | 7505s | train_loss=0.0408
78
+ step 760000 | 7592s | train_loss=0.0401
79
+ step 770000 | 7685s | train_loss=0.0400
80
+ step 780000 | 7778s | train_loss=0.0403
81
+ step 790000 | 7871s | train_loss=0.0400
82
+ step 800000 | 8037s | train_loss=0.0388 | fid=46.2259
83
+ step 810000 | 8128s | train_loss=0.0386
84
+ step 820000 | 8220s | train_loss=0.0389
85
+ step 830000 | 8313s | train_loss=0.0387
86
+ step 840000 | 8406s | train_loss=0.0382
87
+ step 850000 | 8497s | train_loss=0.0387
88
+ step 860000 | 8586s | train_loss=0.0378
89
+ step 870000 | 8679s | train_loss=0.0382
90
+ step 880000 | 8772s | train_loss=0.0375
91
+ step 890000 | 8865s | train_loss=0.0369
92
+ step 900000 | 9032s | train_loss=0.0369 | fid=45.9883
93
+ step 910000 | 9123s | train_loss=0.0362
94
+ step 920000 | 9215s | train_loss=0.0359
95
+ step 930000 | 9308s | train_loss=0.0373
96
+ step 940000 | 9401s | train_loss=0.0364
97
+ step 950000 | 9489s | train_loss=0.0356
98
+ step 960000 | 9582s | train_loss=0.0352
99
+ step 970000 | 9675s | train_loss=0.0356
100
+ step 980000 | 9768s | train_loss=0.0341
101
+ step 990000 | 9861s | train_loss=0.0344
102
+ step 1000000 | 10026s | train_loss=0.0345 | fid=44.5569
103
+ step 1010000 | 10118s | train_loss=0.0352
104
+ step 1020000 | 10211s | train_loss=0.0336
105
+ step 1030000 | 10303s | train_loss=0.0339
106
+ step 1040000 | 10393s | train_loss=0.0330
107
+ step 1050000 | 10483s | train_loss=0.0333
108
+ step 1060000 | 10576s | train_loss=0.0338
109
+ step 1070000 | 10669s | train_loss=0.0333
110
+ step 1080000 | 10762s | train_loss=0.0340
111
+ step 1090000 | 10855s | train_loss=0.0323
112
+ step 1100000 | 11022s | train_loss=0.0323 | fid=44.1248
113
+ step 1110000 | 11113s | train_loss=0.0322
114
+ step 1120000 | 11207s | train_loss=0.0322
115
+ step 1130000 | 11296s | train_loss=0.0323
116
+ step 1140000 | 11386s | train_loss=0.0315
117
+ step 1150000 | 11479s | train_loss=0.0319
118
+ step 1160000 | 11572s | train_loss=0.0314
119
+ step 1170000 | 11665s | train_loss=0.0309
120
+ step 1180000 | 11758s | train_loss=0.0310
121
+ step 1190000 | 11851s | train_loss=0.0315
122
+ step 1200000 | 12017s | train_loss=0.0306 | fid=43.3078
123
+ step 1210000 | 12109s | train_loss=0.0310
124
+ step 1220000 | 12202s | train_loss=0.0310
125
+ step 1230000 | 12288s | train_loss=0.0301
126
+ step 1240000 | 12381s | train_loss=0.0306
127
+ step 1250000 | 12474s | train_loss=0.0299
128
+ step 1260000 | 12567s | train_loss=0.0302
129
+ step 1270000 | 12660s | train_loss=0.0308
130
+ step 1280000 | 12753s | train_loss=0.0294
131
+ step 1290000 | 12846s | train_loss=0.0302
132
+ step 1300000 | 13011s | train_loss=0.0297 | fid=42.5265
133
+ step 1310000 | 13103s | train_loss=0.0291
134
+ step 1320000 | 13193s | train_loss=0.0298
135
+ step 1330000 | 13282s | train_loss=0.0286
136
+ step 1340000 | 13375s | train_loss=0.0284
137
+ step 1350000 | 13468s | train_loss=0.0283
138
+ step 1360000 | 13561s | train_loss=0.0284
139
+ step 1370000 | 13654s | train_loss=0.0288
140
+ step 1380000 | 13748s | train_loss=0.0285
141
+ step 1390000 | 13841s | train_loss=0.0281
142
+ step 1400000 | 14008s | train_loss=0.0278 | fid=42.1456
143
+ step 1410000 | 14097s | train_loss=0.0270
144
+ step 1420000 | 14182s | train_loss=0.0285
145
+ step 1430000 | 14269s | train_loss=0.0280
146
+ step 1440000 | 14361s | train_loss=0.0278
147
+ step 1450000 | 14454s | train_loss=0.0273
148
+ step 1460000 | 14547s | train_loss=0.0284
149
+ step 1470000 | 14640s | train_loss=0.0270
150
+ step 1480000 | 14733s | train_loss=0.0263
151
+ step 1490000 | 14826s | train_loss=0.0272
152
+ step 1500000 | 14992s | train_loss=0.0269 | fid=42.7701
153
+ step 1510000 | 15073s | train_loss=0.0274
154
+ step 1520000 | 15162s | train_loss=0.0261
155
+ step 1530000 | 15255s | train_loss=0.0268
156
+ step 1540000 | 15347s | train_loss=0.0259
157
+ step 1550000 | 15440s | train_loss=0.0256
158
+ step 1560000 | 15533s | train_loss=0.0272
159
+ step 1570000 | 15626s | train_loss=0.0255
160
+ step 1580000 | 15719s | train_loss=0.0259
161
+ step 1590000 | 15813s | train_loss=0.0249
162
+ step 1600000 | 15979s | train_loss=0.0261 | fid=42.3588
163
+ step 1610000 | 16067s | train_loss=0.0261
164
+ step 1620000 | 16157s | train_loss=0.0257
165
+ step 1630000 | 16250s | train_loss=0.0245
166
+ step 1640000 | 16342s | train_loss=0.0249
167
+ step 1650000 | 16434s | train_loss=0.0246
168
+ step 1660000 | 16527s | train_loss=0.0243
169
+ step 1670000 | 16620s | train_loss=0.0245
170
+ step 1680000 | 16713s | train_loss=0.0249
171
+ step 1690000 | 16806s | train_loss=0.0241
172
+ step 1700000 | 16969s | train_loss=0.0243 | fid=41.5411
173
+ step 1710000 | 17058s | train_loss=0.0248
174
+ step 1720000 | 17151s | train_loss=0.0236
175
+ step 1730000 | 17244s | train_loss=0.0235
176
+ step 1740000 | 17336s | train_loss=0.0237
177
+ step 1750000 | 17429s | train_loss=0.0241
178
+ step 1760000 | 17522s | train_loss=0.0245
179
+ step 1770000 | 17615s | train_loss=0.0237
180
+ step 1780000 | 17708s | train_loss=0.0229
181
+ step 1790000 | 17801s | train_loss=0.0228
182
+ step 1800000 | 17962s | train_loss=0.0230 | fid=42.0455
183
+ step 1810000 | 18054s | train_loss=0.0226
184
+ step 1820000 | 18147s | train_loss=0.0232
185
+ step 1830000 | 18240s | train_loss=0.0238
186
+ step 1840000 | 18332s | train_loss=0.0222
187
+ step 1850000 | 18425s | train_loss=0.0221
188
+ step 1860000 | 18518s | train_loss=0.0230
189
+ step 1870000 | 18611s | train_loss=0.0224
190
+ step 1880000 | 18704s | train_loss=0.0228
191
+ step 1890000 | 18797s | train_loss=0.0225
192
+ step 1900000 | 18957s | train_loss=0.0218 | fid=41.6948
193
+ step 1910000 | 19050s | train_loss=0.0215
194
+ step 1920000 | 19142s | train_loss=0.0227
195
+ step 1930000 | 19235s | train_loss=0.0211
196
+ step 1940000 | 19328s | train_loss=0.0224
197
+ step 1950000 | 19419s | train_loss=0.0219
198
+ step 1960000 | 19512s | train_loss=0.0209
199
+ step 1970000 | 19605s | train_loss=0.0223
200
+ step 1980000 | 19698s | train_loss=0.0216
201
+ step 1990000 | 19782s | train_loss=0.0214
202
+ step 2000000 | 19936s | train_loss=0.0215 | fid=40.5635
203
+ step 2010000 | 20030s | train_loss=0.0206
204
+ step 2020000 | 20121s | train_loss=0.0214
205
+ step 2030000 | 20214s | train_loss=0.0209
206
+ step 2040000 | 20308s | train_loss=0.0216
207
+ step 2050000 | 20399s | train_loss=0.0213
208
+ step 2060000 | 20493s | train_loss=0.0211
209
+ step 2070000 | 20586s | train_loss=0.0205
210
+ step 2080000 | 20678s | train_loss=0.0206
211
+ step 2090000 | 20763s | train_loss=0.0209
212
+ step 2100000 | 20922s | train_loss=0.0200 | fid=41.6852
213
+ step 2110000 | 21015s | train_loss=0.0207
214
+ step 2120000 | 21106s | train_loss=0.0199
215
+ step 2130000 | 21199s | train_loss=0.0198
216
+ step 2140000 | 21292s | train_loss=0.0197
217
+ step 2150000 | 21384s | train_loss=0.0206
218
+ step 2160000 | 21477s | train_loss=0.0209
219
+ step 2170000 | 21570s | train_loss=0.0198
220
+ step 2180000 | 21659s | train_loss=0.0196
221
+ step 2190000 | 21744s | train_loss=0.0213
222
+ step 2200000 | 21897s | train_loss=0.0189 | fid=40.7775
223
+ step 2210000 | 21990s | train_loss=0.0197
224
+ step 2220000 | 22081s | train_loss=0.0199
225
+ step 2230000 | 22175s | train_loss=0.0200
226
+ step 2240000 | 22267s | train_loss=0.0192
227
+ step 2250000 | 22359s | train_loss=0.0194
228
+ step 2260000 | 22452s | train_loss=0.0197
229
+ step 2270000 | 22546s | train_loss=0.0189
230
+ step 2280000 | 22632s | train_loss=0.0195
231
+ step 2290000 | 22723s | train_loss=0.0191
232
+ step 2300000 | 22875s | train_loss=0.0195 | fid=40.5685
233
+ step 2310000 | 22969s | train_loss=0.0190
234
+ step 2320000 | 23060s | train_loss=0.0194
235
+ step 2330000 | 23153s | train_loss=0.0199
236
+ step 2340000 | 23246s | train_loss=0.0187
237
+ step 2350000 | 23338s | train_loss=0.0194
238
+ step 2360000 | 23431s | train_loss=0.0190
239
+ step 2370000 | 23524s | train_loss=0.0183
240
+ step 2380000 | 23607s | train_loss=0.0182
241
+ step 2390000 | 23700s | train_loss=0.0187
242
+ step 2400000 | 23851s | train_loss=0.0182 | fid=40.7630
243
+ step 2410000 | 23945s | train_loss=0.0183
244
+ step 2420000 | 24036s | train_loss=0.0184
245
+ step 2430000 | 24129s | train_loss=0.0179
246
+ step 2440000 | 24222s | train_loss=0.0179
247
+ step 2450000 | 24316s | train_loss=0.0182
248
+ step 2460000 | 24408s | train_loss=0.0195
249
+ step 2470000 | 24500s | train_loss=0.0177
250
+ step 2480000 | 24584s | train_loss=0.0185
251
+ step 2490000 | 24676s | train_loss=0.0184
252
+ step 2500000 | 24831s | train_loss=0.0178 | fid=40.5069
253
+ step 2510000 | 24924s | train_loss=0.0179
254
+ step 2520000 | 25016s | train_loss=0.0176
255
+ step 2530000 | 25109s | train_loss=0.0176
256
+ step 2540000 | 25202s | train_loss=0.0183
257
+ step 2550000 | 25295s | train_loss=0.0173
258
+ step 2560000 | 25386s | train_loss=0.0181
259
+ step 2570000 | 25476s | train_loss=0.0179
260
+ step 2580000 | 25560s | train_loss=0.0174
261
+ step 2590000 | 25653s | train_loss=0.0183
262
+ step 2600000 | 25804s | train_loss=0.0179 | fid=40.2529
263
+ step 2610000 | 25897s | train_loss=0.0181
264
+ step 2620000 | 25989s | train_loss=0.0168
265
+ step 2630000 | 26082s | train_loss=0.0171
266
+ step 2640000 | 26175s | train_loss=0.0174
267
+ step 2650000 | 26268s | train_loss=0.0175
268
+ step 2660000 | 26360s | train_loss=0.0173
269
+ step 2670000 | 26447s | train_loss=0.0176
270
+ step 2680000 | 26534s | train_loss=0.0173
271
+ step 2690000 | 26627s | train_loss=0.0178
272
+ step 2700000 | 26782s | train_loss=0.0170 | fid=40.4910
273
+ step 2710000 | 26876s | train_loss=0.0164
274
+ step 2720000 | 26967s | train_loss=0.0167
275
+ step 2730000 | 27060s | train_loss=0.0170
276
+ step 2740000 | 27153s | train_loss=0.0166
277
+ step 2750000 | 27246s | train_loss=0.0167
278
+ step 2760000 | 27338s | train_loss=0.0173
279
+ step 2770000 | 27423s | train_loss=0.0175
280
+ step 2780000 | 27509s | train_loss=0.0161
281
+ step 2790000 | 27602s | train_loss=0.0163
282
+ step 2800000 | 27759s | train_loss=0.0163 | fid=40.7030
283
+ step 2810000 | 27853s | train_loss=0.0164
284
+ step 2820000 | 27944s | train_loss=0.0171
285
+ step 2830000 | 28037s | train_loss=0.0163
286
+ step 2840000 | 28130s | train_loss=0.0167
287
+ step 2850000 | 28223s | train_loss=0.0158
288
+ step 2860000 | 28317s | train_loss=0.0167
289
+ step 2870000 | 28401s | train_loss=0.0167
290
+ step 2880000 | 28486s | train_loss=0.0164
291
+ step 2890000 | 28574s | train_loss=0.0163
292
+ step 2900000 | 28731s | train_loss=0.0166 | fid=40.0594
293
+ step 2910000 | 28824s | train_loss=0.0162
294
+ step 2920000 | 28916s | train_loss=0.0163
295
+ step 2930000 | 29009s | train_loss=0.0162
296
+ step 2940000 | 29102s | train_loss=0.0157
297
+ step 2950000 | 29195s | train_loss=0.0157
298
+ step 2960000 | 29289s | train_loss=0.0157
299
+ step 2970000 | 29374s | train_loss=0.0155
300
+ step 2980000 | 29461s | train_loss=0.0153
301
+ step 2990000 | 29550s | train_loss=0.0160
302
+ step 3000000 | 29712s | train_loss=0.0155 | fid=39.6790
303
+ step 3010000 | 29806s | train_loss=0.0153
304
+ step 3020000 | 29897s | train_loss=0.0154
305
+ step 3030000 | 29990s | train_loss=0.0158
306
+ step 3040000 | 30083s | train_loss=0.0158
307
+ step 3050000 | 30176s | train_loss=0.0149
308
+ step 3060000 | 30269s | train_loss=0.0159
309
+ step 3070000 | 30353s | train_loss=0.0153
310
+ step 3080000 | 30445s | train_loss=0.0156
311
+ step 3090000 | 30531s | train_loss=0.0150
312
+ step 3100000 | 30698s | train_loss=0.0155 | fid=39.8621
313
+ step 3110000 | 30791s | train_loss=0.0147
314
+ step 3120000 | 30882s | train_loss=0.0151
315
+ step 3130000 | 30975s | train_loss=0.0148
316
+ step 3140000 | 31068s | train_loss=0.0149
317
+ step 3150000 | 31161s | train_loss=0.0156
318
+ step 3160000 | 31254s | train_loss=0.0154
319
+ step 3170000 | 31343s | train_loss=0.0148
320
+ step 3180000 | 31436s | train_loss=0.0147
321
+ step 3190000 | 31526s | train_loss=0.0151
322
+ step 3200000 | 31691s | train_loss=0.0148 | fid=40.5878
323
+ step 3210000 | 31784s | train_loss=0.0148
324
+ step 3220000 | 31876s | train_loss=0.0150
325
+ step 3230000 | 31970s | train_loss=0.0142
326
+ step 3240000 | 32063s | train_loss=0.0150
327
+ step 3250000 | 32156s | train_loss=0.0143
328
+ step 3260000 | 32245s | train_loss=0.0149
329
+ step 3270000 | 32337s | train_loss=0.0151
330
+ step 3280000 | 32422s | train_loss=0.0152
331
+ step 3290000 | 32511s | train_loss=0.0144
332
+ step 3300000 | 32678s | train_loss=0.0145 | fid=39.6467
333
+ step 3310000 | 32771s | train_loss=0.0150
334
+ step 3320000 | 32863s | train_loss=0.0144
335
+ step 3330000 | 32956s | train_loss=0.0149
336
+ step 3340000 | 33049s | train_loss=0.0139
337
+ step 3350000 | 33142s | train_loss=0.0148
338
+ step 3360000 | 33227s | train_loss=0.0146
339
+ step 3370000 | 33320s | train_loss=0.0145
340
+ step 3380000 | 33402s | train_loss=0.0136
341
+ step 3390000 | 33491s | train_loss=0.0141
342
+ step 3400000 | 33657s | train_loss=0.0147 | fid=39.3933
343
+ step 3410000 | 33750s | train_loss=0.0137
344
+ step 3420000 | 33842s | train_loss=0.0135
345
+ step 3430000 | 33935s | train_loss=0.0147
346
+ step 3440000 | 34028s | train_loss=0.0142
347
+ step 3450000 | 34115s | train_loss=0.0143
348
+ step 3460000 | 34203s | train_loss=0.0139
349
+ step 3470000 | 34296s | train_loss=0.0139
350
+ step 3480000 | 34377s | train_loss=0.0143
351
+ step 3490000 | 34466s | train_loss=0.0143
352
+ step 3500000 | 34633s | train_loss=0.0146 | fid=39.1876
353
+ step 3510000 | 34726s | train_loss=0.0140
354
+ step 3520000 | 34818s | train_loss=0.0135
355
+ step 3530000 | 34911s | train_loss=0.0146
356
+ step 3540000 | 35004s | train_loss=0.0134
357
+ step 3550000 | 35089s | train_loss=0.0136
358
+ step 3560000 | 35177s | train_loss=0.0133
359
+ step 3570000 | 35269s | train_loss=0.0137
360
+ step 3580000 | 35351s | train_loss=0.0139
361
+ step 3590000 | 35441s | train_loss=0.0137
362
+ step 3600000 | 35607s | train_loss=0.0144 | fid=39.7412
363
+ step 3610000 | 35700s | train_loss=0.0129
364
+ step 3620000 | 35791s | train_loss=0.0142
365
+ step 3630000 | 35884s | train_loss=0.0135
366
+ step 3640000 | 35977s | train_loss=0.0133
367
+ step 3650000 | 36061s | train_loss=0.0135
368
+ step 3660000 | 36155s | train_loss=0.0135
369
+ step 3670000 | 36248s | train_loss=0.0134
370
+ step 3680000 | 36326s | train_loss=0.0137
371
+ step 3690000 | 36416s | train_loss=0.0138
372
+ step 3700000 | 36583s | train_loss=0.0129 | fid=39.8542
373
+ step 3710000 | 36676s | train_loss=0.0135
374
+ step 3720000 | 36768s | train_loss=0.0130
375
+ step 3730000 | 36861s | train_loss=0.0134
376
+ step 3740000 | 36954s | train_loss=0.0125
377
+ step 3750000 | 37034s | train_loss=0.0126
378
+ step 3760000 | 37128s | train_loss=0.0136
379
+ step 3770000 | 37221s | train_loss=0.0140
380
+ step 3780000 | 37312s | train_loss=0.0132
381
+ step 3790000 | 37404s | train_loss=0.0134
382
+ step 3800000 | 37569s | train_loss=0.0126 | fid=38.8542
383
+ step 3810000 | 37662s | train_loss=0.0132
384
+ step 3820000 | 37753s | train_loss=0.0133
385
+ step 3830000 | 37846s | train_loss=0.0130
386
+ step 3840000 | 37939s | train_loss=0.0127
387
+ step 3850000 | 38029s | train_loss=0.0129
388
+ step 3860000 | 38123s | train_loss=0.0127
389
+ step 3870000 | 38213s | train_loss=0.0126
390
+ step 3880000 | 38306s | train_loss=0.0131
391
+ step 3890000 | 38397s | train_loss=0.0127
392
+ step 3900000 | 38564s | train_loss=0.0127 | fid=39.8657
393
+ step 3910000 | 38657s | train_loss=0.0126
394
+ step 3920000 | 38749s | train_loss=0.0125
395
+ step 3930000 | 38842s | train_loss=0.0124
396
+ step 3940000 | 38927s | train_loss=0.0122
397
+ step 3950000 | 39019s | train_loss=0.0122
398
+ step 3960000 | 39112s | train_loss=0.0122
399
+ step 3970000 | 39202s | train_loss=0.0134
400
+ step 3980000 | 39296s | train_loss=0.0125
401
+ step 3990000 | 39387s | train_loss=0.0127
402
+ step 4000000 | 39554s | train_loss=0.0125 | fid=39.1661
403
+ step 4010000 | 39647s | train_loss=0.0122
404
+ step 4020000 | 39740s | train_loss=0.0122
405
+ step 4030000 | 39831s | train_loss=0.0126
406
+ step 4040000 | 39920s | train_loss=0.0128
407
+ step 4050000 | 40013s | train_loss=0.0123
408
+ step 4060000 | 40103s | train_loss=0.0121
409
+ step 4070000 | 40196s | train_loss=0.0125
410
+ step 4080000 | 40289s | train_loss=0.0120
411
+ step 4090000 | 40381s | train_loss=0.0122
412
+ step 4100000 | 40548s | train_loss=0.0128 | fid=38.7626
413
+ step 4110000 | 40641s | train_loss=0.0121
414
+ step 4120000 | 40735s | train_loss=0.0119
415
+ step 4130000 | 40819s | train_loss=0.0123
416
+ step 4140000 | 40909s | train_loss=0.0121
417
+ step 4150000 | 41001s | train_loss=0.0126
418
+ step 4160000 | 41082s | train_loss=0.0119
419
+ step 4170000 | 41175s | train_loss=0.0120
420
+ step 4180000 | 41268s | train_loss=0.0117
421
+ step 4190000 | 41360s | train_loss=0.0118
422
+ step 4200000 | 41526s | train_loss=0.0121 | fid=39.6039
423
+ step 4210000 | 41619s | train_loss=0.0117
424
+ step 4220000 | 41712s | train_loss=0.0115
425
+ step 4230000 | 41793s | train_loss=0.0118
426
+ step 4240000 | 41883s | train_loss=0.0120
427
+ step 4250000 | 41975s | train_loss=0.0127
428
+ step 4260000 | 42054s | train_loss=0.0122
429
+ step 4270000 | 42147s | train_loss=0.0119
430
+ step 4280000 | 42240s | train_loss=0.0114
431
+ step 4290000 | 42334s | train_loss=0.0123
432
+ step 4300000 | 42498s | train_loss=0.0120 | fid=39.5039
433
+ step 4310000 | 42591s | train_loss=0.0126
434
+ step 4320000 | 42684s | train_loss=0.0112
435
+ step 4330000 | 42764s | train_loss=0.0110
436
+ step 4340000 | 42855s | train_loss=0.0121
437
+ step 4350000 | 42948s | train_loss=0.0125
438
+ step 4360000 | 43028s | train_loss=0.0113
439
+ step 4370000 | 43121s | train_loss=0.0119
440
+ step 4380000 | 43214s | train_loss=0.0118
441
+ step 4390000 | 43307s | train_loss=0.0117
442
+ step 4400000 | 43472s | train_loss=0.0117 | fid=39.1011
443
+ step 4410000 | 43565s | train_loss=0.0111
444
+ step 4420000 | 43658s | train_loss=0.0117
445
+ step 4430000 | 43738s | train_loss=0.0114
446
+ step 4440000 | 43829s | train_loss=0.0114
447
+ step 4450000 | 43922s | train_loss=0.0121
448
+ step 4460000 | 44003s | train_loss=0.0116
449
+ step 4470000 | 44096s | train_loss=0.0112
450
+ step 4480000 | 44189s | train_loss=0.0111
451
+ step 4490000 | 44282s | train_loss=0.0111
452
+ step 4500000 | 44446s | train_loss=0.0114 | fid=39.4032
453
+ step 4510000 | 44539s | train_loss=0.0119
454
+ step 4520000 | 44632s | train_loss=0.0118
455
+ step 4530000 | 44712s | train_loss=0.0111
456
+ step 4540000 | 44802s | train_loss=0.0113
457
+ step 4550000 | 44893s | train_loss=0.0109
458
+ step 4560000 | 44976s | train_loss=0.0121
459
+ step 4570000 | 45069s | train_loss=0.0117
460
+ step 4580000 | 45162s | train_loss=0.0112
461
+ step 4590000 | 45254s | train_loss=0.0112
462
+ step 4600000 | 45420s | train_loss=0.0117 | fid=39.7570
463
+ step 4610000 | 45513s | train_loss=0.0115
464
+ step 4620000 | 45606s | train_loss=0.0114
465
+ step 4630000 | 45695s | train_loss=0.0112
466
+ step 4640000 | 45788s | train_loss=0.0114
467
+ step 4650000 | 45879s | train_loss=0.0111
468
+ step 4660000 | 45972s | train_loss=0.0114
469
+ step 4670000 | 46065s | train_loss=0.0106
470
+ step 4680000 | 46158s | train_loss=0.0107
471
+ step 4690000 | 46251s | train_loss=0.0110
472
+ step 4700000 | 46416s | train_loss=0.0110 | fid=39.7827
473
+ step 4710000 | 46509s | train_loss=0.0108
474
+ step 4720000 | 46601s | train_loss=0.0119
475
+ step 4730000 | 46692s | train_loss=0.0115
476
+ step 4740000 | 46785s | train_loss=0.0108
477
+ step 4750000 | 46877s | train_loss=0.0106
478
+ step 4760000 | 46970s | train_loss=0.0109
479
+ step 4770000 | 47063s | train_loss=0.0107
480
+ step 4780000 | 47155s | train_loss=0.0110
481
+ step 4790000 | 47249s | train_loss=0.0105
482
+ step 4800000 | 47414s | train_loss=0.0107 | fid=40.0854
483
+ step 4810000 | 47508s | train_loss=0.0108
484
+ step 4820000 | 47585s | train_loss=0.0108
485
+ step 4830000 | 47658s | train_loss=0.0106
486
+ step 4840000 | 47724s | train_loss=0.0114
487
+ step 4850000 | 47782s | train_loss=0.0108
488
+ step 4860000 | 47839s | train_loss=0.0113
489
+ step 4870000 | 47897s | train_loss=0.0114
490
+ step 4880000 | 47954s | train_loss=0.0108
491
+ step 4890000 | 48012s | train_loss=0.0100
492
+ step 4900000 | 48115s | train_loss=0.0105 | fid=38.8438
493
+ step 4910000 | 48172s | train_loss=0.0109
494
+ step 4920000 | 48230s | train_loss=0.0107
495
+ step 4930000 | 48283s | train_loss=0.0108
496
+ step 4940000 | 48341s | train_loss=0.0103
497
+ step 4950000 | 48398s | train_loss=0.0109
498
+ step 4960000 | 48456s | train_loss=0.0109
499
+ step 4970000 | 48513s | train_loss=0.0104
500
+ step 4980000 | 48571s | train_loss=0.0107
501
+ step 4990000 | 48628s | train_loss=0.0106
502
+ step 5000000 | 48732s | train_loss=0.0110 | fid=39.6805
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/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/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": 42,
25
+ "device": null
26
+ }
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/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/d90/runner.log ADDED
@@ -0,0 +1,502 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ FID real-side stats cached over 1000 diverse train images
2
+ step 1 | 70s | train_loss=0.9941 | fid=306.3547
3
+ step 10000 | 154s | train_loss=0.8074
4
+ step 20000 | 241s | train_loss=0.6939
5
+ step 30000 | 329s | train_loss=0.5612
6
+ step 40000 | 416s | train_loss=0.4170
7
+ step 50000 | 504s | train_loss=0.2966
8
+ step 60000 | 591s | train_loss=0.2207
9
+ step 70000 | 678s | train_loss=0.1790
10
+ step 80000 | 766s | train_loss=0.1536
11
+ step 90000 | 853s | train_loss=0.1378
12
+ step 100000 | 1009s | train_loss=0.1298 | fid=60.4654
13
+ step 110000 | 1092s | train_loss=0.1201
14
+ step 120000 | 1180s | train_loss=0.1148
15
+ step 130000 | 1267s | train_loss=0.1088
16
+ step 140000 | 1354s | train_loss=0.1029
17
+ step 150000 | 1442s | train_loss=0.0988
18
+ step 160000 | 1529s | train_loss=0.0947
19
+ step 170000 | 1617s | train_loss=0.0933
20
+ step 180000 | 1704s | train_loss=0.0897
21
+ step 190000 | 1792s | train_loss=0.0862
22
+ step 200000 | 1948s | train_loss=0.0833 | fid=59.8516
23
+ step 210000 | 2036s | train_loss=0.0817
24
+ step 220000 | 2120s | train_loss=0.0797
25
+ step 230000 | 2207s | train_loss=0.0766
26
+ step 240000 | 2295s | train_loss=0.0749
27
+ step 250000 | 2382s | train_loss=0.0735
28
+ step 260000 | 2470s | train_loss=0.0719
29
+ step 270000 | 2557s | train_loss=0.0699
30
+ step 280000 | 2644s | train_loss=0.0689
31
+ step 290000 | 2732s | train_loss=0.0683
32
+ step 300000 | 2888s | train_loss=0.0658 | fid=56.0142
33
+ step 310000 | 2975s | train_loss=0.0648
34
+ step 320000 | 3062s | train_loss=0.0643
35
+ step 330000 | 3146s | train_loss=0.0632
36
+ step 340000 | 3234s | train_loss=0.0616
37
+ step 350000 | 3321s | train_loss=0.0602
38
+ step 360000 | 3408s | train_loss=0.0600
39
+ step 370000 | 3496s | train_loss=0.0596
40
+ step 380000 | 3583s | train_loss=0.0585
41
+ step 390000 | 3671s | train_loss=0.0577
42
+ step 400000 | 3827s | train_loss=0.0559 | fid=54.6207
43
+ step 410000 | 3914s | train_loss=0.0554
44
+ step 420000 | 4002s | train_loss=0.0538
45
+ step 430000 | 4087s | train_loss=0.0551
46
+ step 440000 | 4173s | train_loss=0.0533
47
+ step 450000 | 4260s | train_loss=0.0531
48
+ step 460000 | 4348s | train_loss=0.0528
49
+ step 470000 | 4435s | train_loss=0.0521
50
+ step 480000 | 4522s | train_loss=0.0514
51
+ step 490000 | 4610s | train_loss=0.0511
52
+ step 500000 | 4766s | train_loss=0.0504 | fid=52.4342
53
+ step 510000 | 4854s | train_loss=0.0497
54
+ step 520000 | 4941s | train_loss=0.0496
55
+ step 530000 | 5029s | train_loss=0.0498
56
+ step 540000 | 5114s | train_loss=0.0480
57
+ step 550000 | 5200s | train_loss=0.0481
58
+ step 560000 | 5287s | train_loss=0.0475
59
+ step 570000 | 5375s | train_loss=0.0467
60
+ step 580000 | 5462s | train_loss=0.0466
61
+ step 590000 | 5549s | train_loss=0.0465
62
+ step 600000 | 5705s | train_loss=0.0452 | fid=51.0297
63
+ step 610000 | 5793s | train_loss=0.0454
64
+ step 620000 | 5880s | train_loss=0.0443
65
+ step 630000 | 5967s | train_loss=0.0446
66
+ step 640000 | 6054s | train_loss=0.0436
67
+ step 650000 | 6140s | train_loss=0.0436
68
+ step 660000 | 6226s | train_loss=0.0434
69
+ step 670000 | 6313s | train_loss=0.0445
70
+ step 680000 | 6401s | train_loss=0.0433
71
+ step 690000 | 6488s | train_loss=0.0431
72
+ step 700000 | 6646s | train_loss=0.0425 | fid=49.8288
73
+ step 710000 | 6734s | train_loss=0.0416
74
+ step 720000 | 6821s | train_loss=0.0433
75
+ step 730000 | 6908s | train_loss=0.0417
76
+ step 740000 | 6996s | train_loss=0.0420
77
+ step 750000 | 7081s | train_loss=0.0419
78
+ step 760000 | 7169s | train_loss=0.0412
79
+ step 770000 | 7254s | train_loss=0.0401
80
+ step 780000 | 7342s | train_loss=0.0402
81
+ step 790000 | 7429s | train_loss=0.0400
82
+ step 800000 | 7585s | train_loss=0.0412 | fid=47.2304
83
+ step 810000 | 7672s | train_loss=0.0392
84
+ step 820000 | 7760s | train_loss=0.0391
85
+ step 830000 | 7847s | train_loss=0.0398
86
+ step 840000 | 7935s | train_loss=0.0384
87
+ step 850000 | 8023s | train_loss=0.0394
88
+ step 860000 | 8107s | train_loss=0.0378
89
+ step 870000 | 8195s | train_loss=0.0384
90
+ step 880000 | 8281s | train_loss=0.0377
91
+ step 890000 | 8368s | train_loss=0.0377
92
+ step 900000 | 8525s | train_loss=0.0374 | fid=47.4502
93
+ step 910000 | 8613s | train_loss=0.0371
94
+ step 920000 | 8700s | train_loss=0.0380
95
+ step 930000 | 8788s | train_loss=0.0378
96
+ step 940000 | 8875s | train_loss=0.0364
97
+ step 950000 | 8962s | train_loss=0.0365
98
+ step 960000 | 9049s | train_loss=0.0372
99
+ step 970000 | 9134s | train_loss=0.0362
100
+ step 980000 | 9221s | train_loss=0.0372
101
+ step 990000 | 9308s | train_loss=0.0358
102
+ step 1000000 | 9465s | train_loss=0.0356 | fid=46.3100
103
+ step 1010000 | 9553s | train_loss=0.0354
104
+ step 1020000 | 9640s | train_loss=0.0361
105
+ step 1030000 | 9728s | train_loss=0.0361
106
+ step 1040000 | 9815s | train_loss=0.0353
107
+ step 1050000 | 9902s | train_loss=0.0343
108
+ step 1060000 | 9990s | train_loss=0.0353
109
+ step 1070000 | 10075s | train_loss=0.0345
110
+ step 1080000 | 10162s | train_loss=0.0349
111
+ step 1090000 | 10248s | train_loss=0.0340
112
+ step 1100000 | 10404s | train_loss=0.0335 | fid=46.3599
113
+ step 1110000 | 10492s | train_loss=0.0327
114
+ step 1120000 | 10579s | train_loss=0.0332
115
+ step 1130000 | 10667s | train_loss=0.0333
116
+ step 1140000 | 10754s | train_loss=0.0326
117
+ step 1150000 | 10842s | train_loss=0.0330
118
+ step 1160000 | 10929s | train_loss=0.0328
119
+ step 1170000 | 11017s | train_loss=0.0321
120
+ step 1180000 | 11101s | train_loss=0.0323
121
+ step 1190000 | 11189s | train_loss=0.0329
122
+ step 1200000 | 11344s | train_loss=0.0326 | fid=45.6827
123
+ step 1210000 | 11432s | train_loss=0.0331
124
+ step 1220000 | 11519s | train_loss=0.0324
125
+ step 1230000 | 11606s | train_loss=0.0317
126
+ step 1240000 | 11694s | train_loss=0.0324
127
+ step 1250000 | 11781s | train_loss=0.0310
128
+ step 1260000 | 11869s | train_loss=0.0302
129
+ step 1270000 | 11956s | train_loss=0.0311
130
+ step 1280000 | 12042s | train_loss=0.0301
131
+ step 1290000 | 12128s | train_loss=0.0301
132
+ step 1300000 | 12283s | train_loss=0.0319 | fid=46.9911
133
+ step 1310000 | 12370s | train_loss=0.0321
134
+ step 1320000 | 12458s | train_loss=0.0307
135
+ step 1330000 | 12545s | train_loss=0.0304
136
+ step 1340000 | 12633s | train_loss=0.0313
137
+ step 1350000 | 12720s | train_loss=0.0296
138
+ step 1360000 | 12808s | train_loss=0.0308
139
+ step 1370000 | 12895s | train_loss=0.0293
140
+ step 1380000 | 12983s | train_loss=0.0290
141
+ step 1390000 | 13069s | train_loss=0.0295
142
+ step 1400000 | 13224s | train_loss=0.0302 | fid=47.4044
143
+ step 1410000 | 13310s | train_loss=0.0293
144
+ step 1420000 | 13397s | train_loss=0.0297
145
+ step 1430000 | 13485s | train_loss=0.0282
146
+ step 1440000 | 13572s | train_loss=0.0296
147
+ step 1450000 | 13660s | train_loss=0.0285
148
+ step 1460000 | 13747s | train_loss=0.0278
149
+ step 1470000 | 13834s | train_loss=0.0286
150
+ step 1480000 | 13922s | train_loss=0.0276
151
+ step 1490000 | 14008s | train_loss=0.0280
152
+ step 1500000 | 14207s | train_loss=0.0281 | fid=48.4357
153
+ step 1510000 | 14293s | train_loss=0.0278
154
+ step 1520000 | 14381s | train_loss=0.0268
155
+ step 1530000 | 14468s | train_loss=0.0275
156
+ step 1540000 | 14556s | train_loss=0.0265
157
+ step 1550000 | 14643s | train_loss=0.0279
158
+ step 1560000 | 14730s | train_loss=0.0279
159
+ step 1570000 | 14818s | train_loss=0.0280
160
+ step 1580000 | 14905s | train_loss=0.0270
161
+ step 1590000 | 14992s | train_loss=0.0266
162
+ step 1600000 | 15140s | train_loss=0.0264 | fid=46.9537
163
+ step 1610000 | 15228s | train_loss=0.0270
164
+ step 1620000 | 15314s | train_loss=0.0264
165
+ step 1630000 | 15401s | train_loss=0.0266
166
+ step 1640000 | 15489s | train_loss=0.0269
167
+ step 1650000 | 15576s | train_loss=0.0263
168
+ step 1660000 | 15663s | train_loss=0.0253
169
+ step 1670000 | 15751s | train_loss=0.0255
170
+ step 1680000 | 15838s | train_loss=0.0256
171
+ step 1690000 | 15926s | train_loss=0.0252
172
+ step 1700000 | 16080s | train_loss=0.0249 | fid=46.7741
173
+ step 1710000 | 16168s | train_loss=0.0255
174
+ step 1720000 | 16255s | train_loss=0.0262
175
+ step 1730000 | 16342s | train_loss=0.0262
176
+ step 1740000 | 16429s | train_loss=0.0257
177
+ step 1750000 | 16516s | train_loss=0.0242
178
+ step 1760000 | 16604s | train_loss=0.0254
179
+ step 1770000 | 16691s | train_loss=0.0252
180
+ step 1780000 | 16778s | train_loss=0.0254
181
+ step 1790000 | 16866s | train_loss=0.0246
182
+ step 1800000 | 17022s | train_loss=0.0244 | fid=45.7961
183
+ step 1810000 | 17108s | train_loss=0.0245
184
+ step 1820000 | 17195s | train_loss=0.0241
185
+ step 1830000 | 17283s | train_loss=0.0238
186
+ step 1840000 | 17369s | train_loss=0.0250
187
+ step 1850000 | 17456s | train_loss=0.0239
188
+ step 1860000 | 17544s | train_loss=0.0245
189
+ step 1870000 | 17631s | train_loss=0.0237
190
+ step 1880000 | 17718s | train_loss=0.0230
191
+ step 1890000 | 17806s | train_loss=0.0234
192
+ step 1900000 | 17963s | train_loss=0.0246 | fid=44.4650
193
+ step 1910000 | 18051s | train_loss=0.0242
194
+ step 1920000 | 18137s | train_loss=0.0234
195
+ step 1930000 | 18224s | train_loss=0.0248
196
+ step 1940000 | 18312s | train_loss=0.0236
197
+ step 1950000 | 18398s | train_loss=0.0236
198
+ step 1960000 | 18485s | train_loss=0.0235
199
+ step 1970000 | 18573s | train_loss=0.0232
200
+ step 1980000 | 18660s | train_loss=0.0231
201
+ step 1990000 | 18748s | train_loss=0.0225
202
+ step 2000000 | 18905s | train_loss=0.0219 | fid=44.1513
203
+ step 2010000 | 18991s | train_loss=0.0230
204
+ step 2020000 | 19077s | train_loss=0.0221
205
+ step 2030000 | 19164s | train_loss=0.0225
206
+ step 2040000 | 19251s | train_loss=0.0239
207
+ step 2050000 | 19338s | train_loss=0.0227
208
+ step 2060000 | 19425s | train_loss=0.0220
209
+ step 2070000 | 19512s | train_loss=0.0225
210
+ step 2080000 | 19600s | train_loss=0.0221
211
+ step 2090000 | 19687s | train_loss=0.0217
212
+ step 2100000 | 19874s | train_loss=0.0223 | fid=42.0061
213
+ step 2110000 | 19960s | train_loss=0.0211
214
+ step 2120000 | 20046s | train_loss=0.0225
215
+ step 2130000 | 20133s | train_loss=0.0217
216
+ step 2140000 | 20221s | train_loss=0.0213
217
+ step 2150000 | 20309s | train_loss=0.0218
218
+ step 2160000 | 20395s | train_loss=0.0215
219
+ step 2170000 | 20482s | train_loss=0.0208
220
+ step 2180000 | 20570s | train_loss=0.0216
221
+ step 2190000 | 20657s | train_loss=0.0216
222
+ step 2200000 | 20826s | train_loss=0.0218 | fid=41.0657
223
+ step 2210000 | 20914s | train_loss=0.0207
224
+ step 2220000 | 21000s | train_loss=0.0209
225
+ step 2230000 | 21086s | train_loss=0.0215
226
+ step 2240000 | 21173s | train_loss=0.0211
227
+ step 2250000 | 21260s | train_loss=0.0215
228
+ step 2260000 | 21347s | train_loss=0.0196
229
+ step 2270000 | 21434s | train_loss=0.0206
230
+ step 2280000 | 21521s | train_loss=0.0212
231
+ step 2290000 | 21609s | train_loss=0.0202
232
+ step 2300000 | 21800s | train_loss=0.0204 | fid=41.3404
233
+ step 2310000 | 21888s | train_loss=0.0204
234
+ step 2320000 | 21973s | train_loss=0.0208
235
+ step 2330000 | 22059s | train_loss=0.0201
236
+ step 2340000 | 22147s | train_loss=0.0198
237
+ step 2350000 | 22234s | train_loss=0.0202
238
+ step 2360000 | 22322s | train_loss=0.0201
239
+ step 2370000 | 22408s | train_loss=0.0206
240
+ step 2380000 | 22495s | train_loss=0.0200
241
+ step 2390000 | 22583s | train_loss=0.0202
242
+ step 2400000 | 22781s | train_loss=0.0197 | fid=41.2987
243
+ step 2410000 | 22869s | train_loss=0.0198
244
+ step 2420000 | 22955s | train_loss=0.0197
245
+ step 2430000 | 23041s | train_loss=0.0198
246
+ step 2440000 | 23128s | train_loss=0.0190
247
+ step 2450000 | 23216s | train_loss=0.0199
248
+ step 2460000 | 23303s | train_loss=0.0196
249
+ step 2470000 | 23389s | train_loss=0.0193
250
+ step 2480000 | 23477s | train_loss=0.0193
251
+ step 2490000 | 23562s | train_loss=0.0192
252
+ step 2500000 | 23766s | train_loss=0.0197 | fid=41.0434
253
+ step 2510000 | 23852s | train_loss=0.0203
254
+ step 2520000 | 23939s | train_loss=0.0201
255
+ step 2530000 | 24025s | train_loss=0.0189
256
+ step 2540000 | 24113s | train_loss=0.0192
257
+ step 2550000 | 24200s | train_loss=0.0193
258
+ step 2560000 | 24288s | train_loss=0.0193
259
+ step 2570000 | 24374s | train_loss=0.0193
260
+ step 2580000 | 24461s | train_loss=0.0184
261
+ step 2590000 | 24541s | train_loss=0.0191
262
+ step 2600000 | 24739s | train_loss=0.0194 | fid=40.6276
263
+ step 2610000 | 24827s | train_loss=0.0194
264
+ step 2620000 | 24913s | train_loss=0.0191
265
+ step 2630000 | 24999s | train_loss=0.0193
266
+ step 2640000 | 25087s | train_loss=0.0191
267
+ step 2650000 | 25174s | train_loss=0.0195
268
+ step 2660000 | 25261s | train_loss=0.0190
269
+ step 2670000 | 25347s | train_loss=0.0186
270
+ step 2680000 | 25435s | train_loss=0.0177
271
+ step 2690000 | 25511s | train_loss=0.0177
272
+ step 2700000 | 25721s | train_loss=0.0186 | fid=39.4785
273
+ step 2710000 | 25807s | train_loss=0.0185
274
+ step 2720000 | 25895s | train_loss=0.0182
275
+ step 2730000 | 25981s | train_loss=0.0188
276
+ step 2740000 | 26069s | train_loss=0.0192
277
+ step 2750000 | 26156s | train_loss=0.0180
278
+ step 2760000 | 26244s | train_loss=0.0187
279
+ step 2770000 | 26330s | train_loss=0.0188
280
+ step 2780000 | 26417s | train_loss=0.0184
281
+ step 2790000 | 26493s | train_loss=0.0174
282
+ step 2800000 | 26686s | train_loss=0.0184 | fid=39.7828
283
+ step 2810000 | 26774s | train_loss=0.0189
284
+ step 2820000 | 26860s | train_loss=0.0179
285
+ step 2830000 | 26946s | train_loss=0.0180
286
+ step 2840000 | 27033s | train_loss=0.0185
287
+ step 2850000 | 27120s | train_loss=0.0176
288
+ step 2860000 | 27208s | train_loss=0.0176
289
+ step 2870000 | 27295s | train_loss=0.0179
290
+ step 2880000 | 27381s | train_loss=0.0168
291
+ step 2890000 | 27454s | train_loss=0.0180
292
+ step 2900000 | 27637s | train_loss=0.0181 | fid=39.8042
293
+ step 2910000 | 27725s | train_loss=0.0184
294
+ step 2920000 | 27811s | train_loss=0.0171
295
+ step 2930000 | 27897s | train_loss=0.0173
296
+ step 2940000 | 27984s | train_loss=0.0163
297
+ step 2950000 | 28071s | train_loss=0.0178
298
+ step 2960000 | 28159s | train_loss=0.0179
299
+ step 2970000 | 28246s | train_loss=0.0180
300
+ step 2980000 | 28332s | train_loss=0.0174
301
+ step 2990000 | 28411s | train_loss=0.0178
302
+ step 3000000 | 28612s | train_loss=0.0169 | fid=39.7709
303
+ step 3010000 | 28700s | train_loss=0.0171
304
+ step 3020000 | 28786s | train_loss=0.0164
305
+ step 3030000 | 28872s | train_loss=0.0174
306
+ step 3040000 | 28959s | train_loss=0.0174
307
+ step 3050000 | 29047s | train_loss=0.0172
308
+ step 3060000 | 29134s | train_loss=0.0172
309
+ step 3070000 | 29222s | train_loss=0.0165
310
+ step 3080000 | 29310s | train_loss=0.0166
311
+ step 3090000 | 29387s | train_loss=0.0172
312
+ step 3100000 | 29567s | train_loss=0.0167 | fid=40.4726
313
+ step 3110000 | 29655s | train_loss=0.0165
314
+ step 3120000 | 29741s | train_loss=0.0172
315
+ step 3130000 | 29829s | train_loss=0.0162
316
+ step 3140000 | 29915s | train_loss=0.0169
317
+ step 3150000 | 30002s | train_loss=0.0166
318
+ step 3160000 | 30089s | train_loss=0.0163
319
+ step 3170000 | 30177s | train_loss=0.0167
320
+ step 3180000 | 30264s | train_loss=0.0160
321
+ step 3190000 | 30344s | train_loss=0.0165
322
+ step 3200000 | 30520s | train_loss=0.0158 | fid=40.6156
323
+ step 3210000 | 30607s | train_loss=0.0160
324
+ step 3220000 | 30694s | train_loss=0.0168
325
+ step 3230000 | 30781s | train_loss=0.0157
326
+ step 3240000 | 30867s | train_loss=0.0163
327
+ step 3250000 | 30954s | train_loss=0.0164
328
+ step 3260000 | 31042s | train_loss=0.0162
329
+ step 3270000 | 31129s | train_loss=0.0163
330
+ step 3280000 | 31216s | train_loss=0.0158
331
+ step 3290000 | 31302s | train_loss=0.0167
332
+ step 3300000 | 31457s | train_loss=0.0156 | fid=40.3151
333
+ step 3310000 | 31545s | train_loss=0.0158
334
+ step 3320000 | 31632s | train_loss=0.0157
335
+ step 3330000 | 31718s | train_loss=0.0160
336
+ step 3340000 | 31806s | train_loss=0.0154
337
+ step 3350000 | 31892s | train_loss=0.0163
338
+ step 3360000 | 31979s | train_loss=0.0168
339
+ step 3370000 | 32067s | train_loss=0.0165
340
+ step 3380000 | 32154s | train_loss=0.0153
341
+ step 3390000 | 32239s | train_loss=0.0154
342
+ step 3400000 | 32437s | train_loss=0.0153 | fid=39.4946
343
+ step 3410000 | 32524s | train_loss=0.0164
344
+ step 3420000 | 32612s | train_loss=0.0150
345
+ step 3430000 | 32698s | train_loss=0.0155
346
+ step 3440000 | 32785s | train_loss=0.0160
347
+ step 3450000 | 32872s | train_loss=0.0158
348
+ step 3460000 | 32959s | train_loss=0.0152
349
+ step 3470000 | 33046s | train_loss=0.0149
350
+ step 3480000 | 33134s | train_loss=0.0152
351
+ step 3490000 | 33213s | train_loss=0.0152
352
+ step 3500000 | 33415s | train_loss=0.0153 | fid=39.6567
353
+ step 3510000 | 33503s | train_loss=0.0155
354
+ step 3520000 | 33590s | train_loss=0.0151
355
+ step 3530000 | 33676s | train_loss=0.0147
356
+ step 3540000 | 33764s | train_loss=0.0151
357
+ step 3550000 | 33850s | train_loss=0.0152
358
+ step 3560000 | 33937s | train_loss=0.0150
359
+ step 3570000 | 34024s | train_loss=0.0166
360
+ step 3580000 | 34108s | train_loss=0.0147
361
+ step 3590000 | 34188s | train_loss=0.0152
362
+ step 3600000 | 34393s | train_loss=0.0145 | fid=40.3809
363
+ step 3610000 | 34480s | train_loss=0.0153
364
+ step 3620000 | 34568s | train_loss=0.0152
365
+ step 3630000 | 34654s | train_loss=0.0149
366
+ step 3640000 | 34741s | train_loss=0.0158
367
+ step 3650000 | 34827s | train_loss=0.0143
368
+ step 3660000 | 34915s | train_loss=0.0147
369
+ step 3670000 | 35002s | train_loss=0.0145
370
+ step 3680000 | 35083s | train_loss=0.0151
371
+ step 3690000 | 35164s | train_loss=0.0154
372
+ step 3700000 | 35364s | train_loss=0.0149 | fid=40.1025
373
+ step 3710000 | 35451s | train_loss=0.0147
374
+ step 3720000 | 35539s | train_loss=0.0149
375
+ step 3730000 | 35625s | train_loss=0.0155
376
+ step 3740000 | 35712s | train_loss=0.0145
377
+ step 3750000 | 35798s | train_loss=0.0141
378
+ step 3760000 | 35886s | train_loss=0.0145
379
+ step 3770000 | 35973s | train_loss=0.0144
380
+ step 3780000 | 36054s | train_loss=0.0143
381
+ step 3790000 | 36139s | train_loss=0.0140
382
+ step 3800000 | 36334s | train_loss=0.0147 | fid=39.8744
383
+ step 3810000 | 36421s | train_loss=0.0137
384
+ step 3820000 | 36509s | train_loss=0.0147
385
+ step 3830000 | 36595s | train_loss=0.0144
386
+ step 3840000 | 36682s | train_loss=0.0144
387
+ step 3850000 | 36769s | train_loss=0.0142
388
+ step 3860000 | 36856s | train_loss=0.0138
389
+ step 3870000 | 36944s | train_loss=0.0134
390
+ step 3880000 | 37021s | train_loss=0.0140
391
+ step 3890000 | 37106s | train_loss=0.0133
392
+ step 3900000 | 37264s | train_loss=0.0139 | fid=39.6010
393
+ step 3910000 | 37350s | train_loss=0.0139
394
+ step 3920000 | 37438s | train_loss=0.0140
395
+ step 3930000 | 37525s | train_loss=0.0145
396
+ step 3940000 | 37612s | train_loss=0.0145
397
+ step 3950000 | 37699s | train_loss=0.0135
398
+ step 3960000 | 37785s | train_loss=0.0145
399
+ step 3970000 | 37872s | train_loss=0.0137
400
+ step 3980000 | 37959s | train_loss=0.0147
401
+ step 3990000 | 38044s | train_loss=0.0139
402
+ step 4000000 | 38204s | train_loss=0.0135 | fid=39.6441
403
+ step 4010000 | 38292s | train_loss=0.0138
404
+ step 4020000 | 38378s | train_loss=0.0134
405
+ step 4030000 | 38465s | train_loss=0.0139
406
+ step 4040000 | 38554s | train_loss=0.0134
407
+ step 4050000 | 38639s | train_loss=0.0137
408
+ step 4060000 | 38727s | train_loss=0.0138
409
+ step 4070000 | 38812s | train_loss=0.0138
410
+ step 4080000 | 38899s | train_loss=0.0136
411
+ step 4090000 | 38979s | train_loss=0.0135
412
+ step 4100000 | 39136s | train_loss=0.0133 | fid=39.9201
413
+ step 4110000 | 39224s | train_loss=0.0136
414
+ step 4120000 | 39312s | train_loss=0.0141
415
+ step 4130000 | 39397s | train_loss=0.0144
416
+ step 4140000 | 39485s | train_loss=0.0136
417
+ step 4150000 | 39571s | train_loss=0.0138
418
+ step 4160000 | 39658s | train_loss=0.0146
419
+ step 4170000 | 39745s | train_loss=0.0150
420
+ step 4180000 | 39832s | train_loss=0.0129
421
+ step 4190000 | 39916s | train_loss=0.0142
422
+ step 4200000 | 40076s | train_loss=0.0131 | fid=38.7803
423
+ step 4210000 | 40164s | train_loss=0.0136
424
+ step 4220000 | 40251s | train_loss=0.0127
425
+ step 4230000 | 40339s | train_loss=0.0140
426
+ step 4240000 | 40425s | train_loss=0.0137
427
+ step 4250000 | 40512s | train_loss=0.0129
428
+ step 4260000 | 40598s | train_loss=0.0130
429
+ step 4270000 | 40686s | train_loss=0.0129
430
+ step 4280000 | 40772s | train_loss=0.0131
431
+ step 4290000 | 40850s | train_loss=0.0134
432
+ step 4300000 | 41053s | train_loss=0.0129 | fid=40.4070
433
+ step 4310000 | 41140s | train_loss=0.0138
434
+ step 4320000 | 41228s | train_loss=0.0132
435
+ step 4330000 | 41315s | train_loss=0.0141
436
+ step 4340000 | 41402s | train_loss=0.0126
437
+ step 4350000 | 41489s | train_loss=0.0127
438
+ step 4360000 | 41575s | train_loss=0.0122
439
+ step 4370000 | 41662s | train_loss=0.0135
440
+ step 4380000 | 41748s | train_loss=0.0137
441
+ step 4390000 | 41822s | train_loss=0.0126
442
+ step 4400000 | 42031s | train_loss=0.0128 | fid=40.0581
443
+ step 4410000 | 42119s | train_loss=0.0127
444
+ step 4420000 | 42206s | train_loss=0.0122
445
+ step 4430000 | 42294s | train_loss=0.0128
446
+ step 4440000 | 42380s | train_loss=0.0132
447
+ step 4450000 | 42468s | train_loss=0.0145
448
+ step 4460000 | 42553s | train_loss=0.0136
449
+ step 4470000 | 42641s | train_loss=0.0134
450
+ step 4480000 | 42725s | train_loss=0.0123
451
+ step 4490000 | 42801s | train_loss=0.0127
452
+ step 4500000 | 43004s | train_loss=0.0136 | fid=39.8249
453
+ step 4510000 | 43091s | train_loss=0.0145
454
+ step 4520000 | 43178s | train_loss=0.0119
455
+ step 4530000 | 43266s | train_loss=0.0140
456
+ step 4540000 | 43352s | train_loss=0.0123
457
+ step 4550000 | 43440s | train_loss=0.0123
458
+ step 4560000 | 43526s | train_loss=0.0131
459
+ step 4570000 | 43613s | train_loss=0.0127
460
+ step 4580000 | 43697s | train_loss=0.0125
461
+ step 4590000 | 43773s | train_loss=0.0125
462
+ step 4600000 | 43971s | train_loss=0.0119 | fid=39.5651
463
+ step 4610000 | 44058s | train_loss=0.0128
464
+ step 4620000 | 44145s | train_loss=0.0127
465
+ step 4630000 | 44233s | train_loss=0.0126
466
+ step 4640000 | 44320s | train_loss=0.0122
467
+ step 4650000 | 44406s | train_loss=0.0123
468
+ step 4660000 | 44492s | train_loss=0.0125
469
+ step 4670000 | 44579s | train_loss=0.0124
470
+ step 4680000 | 44665s | train_loss=0.0128
471
+ step 4690000 | 44738s | train_loss=0.0126
472
+ step 4700000 | 44931s | train_loss=0.0131 | fid=39.6921
473
+ step 4710000 | 45019s | train_loss=0.0121
474
+ step 4720000 | 45106s | train_loss=0.0124
475
+ step 4730000 | 45194s | train_loss=0.0127
476
+ step 4740000 | 45281s | train_loss=0.0119
477
+ step 4750000 | 45367s | train_loss=0.0128
478
+ step 4760000 | 45453s | train_loss=0.0120
479
+ step 4770000 | 45541s | train_loss=0.0124
480
+ step 4780000 | 45629s | train_loss=0.0131
481
+ step 4790000 | 45711s | train_loss=0.0118
482
+ step 4800000 | 45868s | train_loss=0.0123 | fid=40.6984
483
+ step 4810000 | 45955s | train_loss=0.0119
484
+ step 4820000 | 46043s | train_loss=0.0122
485
+ step 4830000 | 46130s | train_loss=0.0127
486
+ step 4840000 | 46218s | train_loss=0.0123
487
+ step 4850000 | 46305s | train_loss=0.0120
488
+ step 4860000 | 46392s | train_loss=0.0118
489
+ step 4870000 | 46478s | train_loss=0.0122
490
+ step 4880000 | 46565s | train_loss=0.0117
491
+ step 4890000 | 46649s | train_loss=0.0116
492
+ step 4900000 | 46805s | train_loss=0.0123 | fid=40.1795
493
+ step 4910000 | 46893s | train_loss=0.0116
494
+ step 4920000 | 46980s | train_loss=0.0115
495
+ step 4930000 | 47068s | train_loss=0.0128
496
+ step 4940000 | 47155s | train_loss=0.0118
497
+ step 4950000 | 47242s | train_loss=0.0117
498
+ step 4960000 | 47330s | train_loss=0.0123
499
+ step 4970000 | 47415s | train_loss=0.0117
500
+ step 4980000 | 47502s | train_loss=0.0120
501
+ step 4990000 | 47576s | train_loss=0.0118
502
+ step 5000000 | 47701s | train_loss=0.0121 | fid=39.8906
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed43/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/seed43/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": 43,
25
+ "device": null
26
+ }
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed43/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/seed43/d10/runner.log ADDED
@@ -0,0 +1,502 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ FID real-side stats cached over 1000 diverse train images
2
+ step 1 | 65s | train_loss=0.9877 | fid=194.6918
3
+ step 10000 | 145s | train_loss=0.9080
4
+ step 20000 | 228s | train_loss=0.9057
5
+ step 30000 | 311s | train_loss=0.9045
6
+ step 40000 | 394s | train_loss=0.9036
7
+ step 50000 | 477s | train_loss=0.9027
8
+ step 60000 | 560s | train_loss=0.9017
9
+ step 70000 | 644s | train_loss=0.9012
10
+ step 80000 | 727s | train_loss=0.9002
11
+ step 90000 | 810s | train_loss=0.8993
12
+ step 100000 | 959s | train_loss=0.8988 | fid=87.1300
13
+ step 110000 | 1036s | train_loss=0.8979
14
+ step 120000 | 1119s | train_loss=0.8972
15
+ step 130000 | 1202s | train_loss=0.8965
16
+ step 140000 | 1285s | train_loss=0.8958
17
+ step 150000 | 1368s | train_loss=0.8949
18
+ step 160000 | 1451s | train_loss=0.8943
19
+ step 170000 | 1534s | train_loss=0.8935
20
+ step 180000 | 1617s | train_loss=0.8926
21
+ step 190000 | 1700s | train_loss=0.8915
22
+ step 200000 | 1852s | train_loss=0.8906 | fid=87.3145
23
+ step 210000 | 1930s | train_loss=0.8895
24
+ step 220000 | 2010s | train_loss=0.8884
25
+ step 230000 | 2093s | train_loss=0.8870
26
+ step 240000 | 2176s | train_loss=0.8855
27
+ step 250000 | 2260s | train_loss=0.8838
28
+ step 260000 | 2343s | train_loss=0.8819
29
+ step 270000 | 2426s | train_loss=0.8795
30
+ step 280000 | 2509s | train_loss=0.8764
31
+ step 290000 | 2592s | train_loss=0.8725
32
+ step 300000 | 2740s | train_loss=0.8664 | fid=86.5259
33
+ step 310000 | 2819s | train_loss=0.8593
34
+ step 320000 | 2900s | train_loss=0.8486
35
+ step 330000 | 2982s | train_loss=0.8341
36
+ step 340000 | 3065s | train_loss=0.8167
37
+ step 350000 | 3148s | train_loss=0.7949
38
+ step 360000 | 3231s | train_loss=0.7678
39
+ step 370000 | 3314s | train_loss=0.7318
40
+ step 380000 | 3397s | train_loss=0.6978
41
+ step 390000 | 3481s | train_loss=0.6550
42
+ step 400000 | 3630s | train_loss=0.6148 | fid=87.3717
43
+ step 410000 | 3708s | train_loss=0.5683
44
+ step 420000 | 3791s | train_loss=0.5355
45
+ step 430000 | 3872s | train_loss=0.5063
46
+ step 440000 | 3954s | train_loss=0.4576
47
+ step 450000 | 4037s | train_loss=0.4292
48
+ step 460000 | 4120s | train_loss=0.4036
49
+ step 470000 | 4203s | train_loss=0.3781
50
+ step 480000 | 4286s | train_loss=0.3535
51
+ step 490000 | 4369s | train_loss=0.3461
52
+ step 500000 | 4521s | train_loss=0.3292 | fid=87.7379
53
+ step 510000 | 4599s | train_loss=0.3082
54
+ step 520000 | 4682s | train_loss=0.3069
55
+ step 530000 | 4762s | train_loss=0.2853
56
+ step 540000 | 4846s | train_loss=0.2759
57
+ step 550000 | 4927s | train_loss=0.2695
58
+ step 560000 | 5011s | train_loss=0.2517
59
+ step 570000 | 5094s | train_loss=0.2414
60
+ step 580000 | 5177s | train_loss=0.2373
61
+ step 590000 | 5260s | train_loss=0.2291
62
+ step 600000 | 5409s | train_loss=0.2292 | fid=87.1496
63
+ step 610000 | 5489s | train_loss=0.2171
64
+ step 620000 | 5570s | train_loss=0.2172
65
+ step 630000 | 5653s | train_loss=0.2124
66
+ step 640000 | 5734s | train_loss=0.2039
67
+ step 650000 | 5817s | train_loss=0.2009
68
+ step 660000 | 5898s | train_loss=0.1945
69
+ step 670000 | 5982s | train_loss=0.1880
70
+ step 680000 | 6065s | train_loss=0.1849
71
+ step 690000 | 6148s | train_loss=0.1831
72
+ step 700000 | 6297s | train_loss=0.1802 | fid=86.1498
73
+ step 710000 | 6377s | train_loss=0.1793
74
+ step 720000 | 6457s | train_loss=0.1737
75
+ step 730000 | 6541s | train_loss=0.1698
76
+ step 740000 | 6622s | train_loss=0.1663
77
+ step 750000 | 6705s | train_loss=0.1660
78
+ step 760000 | 6788s | train_loss=0.1659
79
+ step 770000 | 6869s | train_loss=0.1636
80
+ step 780000 | 6953s | train_loss=0.1570
81
+ step 790000 | 7036s | train_loss=0.1546
82
+ step 800000 | 7187s | train_loss=0.1542 | fid=87.6051
83
+ step 810000 | 7268s | train_loss=0.1522
84
+ step 820000 | 7348s | train_loss=0.1525
85
+ step 830000 | 7431s | train_loss=0.1497
86
+ step 840000 | 7514s | train_loss=0.1474
87
+ step 850000 | 7594s | train_loss=0.1482
88
+ step 860000 | 7677s | train_loss=0.1437
89
+ step 870000 | 7760s | train_loss=0.1430
90
+ step 880000 | 7842s | train_loss=0.1394
91
+ step 890000 | 7925s | train_loss=0.1374
92
+ step 900000 | 8077s | train_loss=0.1389 | fid=86.6205
93
+ step 910000 | 8158s | train_loss=0.1354
94
+ step 920000 | 8238s | train_loss=0.1320
95
+ step 930000 | 8321s | train_loss=0.1320
96
+ step 940000 | 8404s | train_loss=0.1294
97
+ step 950000 | 8484s | train_loss=0.1317
98
+ step 960000 | 8567s | train_loss=0.1254
99
+ step 970000 | 8650s | train_loss=0.1264
100
+ step 980000 | 8734s | train_loss=0.1249
101
+ step 990000 | 8815s | train_loss=0.1271
102
+ step 1000000 | 8965s | train_loss=0.1242 | fid=86.5478
103
+ step 1010000 | 9048s | train_loss=0.1214
104
+ step 1020000 | 9129s | train_loss=0.1181
105
+ step 1030000 | 9209s | train_loss=0.1196
106
+ step 1040000 | 9292s | train_loss=0.1214
107
+ step 1050000 | 9375s | train_loss=0.1176
108
+ step 1060000 | 9455s | train_loss=0.1181
109
+ step 1070000 | 9538s | train_loss=0.1185
110
+ step 1080000 | 9621s | train_loss=0.1145
111
+ step 1090000 | 9705s | train_loss=0.1151
112
+ step 1100000 | 9853s | train_loss=0.1140 | fid=86.4944
113
+ step 1110000 | 9936s | train_loss=0.1099
114
+ step 1120000 | 10016s | train_loss=0.1139
115
+ step 1130000 | 10096s | train_loss=0.1098
116
+ step 1140000 | 10179s | train_loss=0.1115
117
+ step 1150000 | 10262s | train_loss=0.1092
118
+ step 1160000 | 10345s | train_loss=0.1071
119
+ step 1170000 | 10425s | train_loss=0.1081
120
+ step 1180000 | 10508s | train_loss=0.1070
121
+ step 1190000 | 10592s | train_loss=0.1040
122
+ step 1200000 | 10743s | train_loss=0.1074 | fid=88.0667
123
+ step 1210000 | 10826s | train_loss=0.1077
124
+ step 1220000 | 10906s | train_loss=0.1046
125
+ step 1230000 | 10986s | train_loss=0.1034
126
+ step 1240000 | 11069s | train_loss=0.1036
127
+ step 1250000 | 11153s | train_loss=0.1016
128
+ step 1260000 | 11236s | train_loss=0.1036
129
+ step 1270000 | 11316s | train_loss=0.1036
130
+ step 1280000 | 11399s | train_loss=0.0999
131
+ step 1290000 | 11482s | train_loss=0.0988
132
+ step 1300000 | 11634s | train_loss=0.0998 | fid=86.7001
133
+ step 1310000 | 11715s | train_loss=0.0970
134
+ step 1320000 | 11795s | train_loss=0.0980
135
+ step 1330000 | 11876s | train_loss=0.0982
136
+ step 1340000 | 11959s | train_loss=0.0979
137
+ step 1350000 | 12042s | train_loss=0.0967
138
+ step 1360000 | 12125s | train_loss=0.0993
139
+ step 1370000 | 12208s | train_loss=0.0946
140
+ step 1380000 | 12289s | train_loss=0.0944
141
+ step 1390000 | 12372s | train_loss=0.0958
142
+ step 1400000 | 12522s | train_loss=0.0939 | fid=86.9982
143
+ step 1410000 | 12604s | train_loss=0.0936
144
+ step 1420000 | 12684s | train_loss=0.0917
145
+ step 1430000 | 12765s | train_loss=0.0928
146
+ step 1440000 | 12848s | train_loss=0.0919
147
+ step 1450000 | 12931s | train_loss=0.0918
148
+ step 1460000 | 13014s | train_loss=0.0914
149
+ step 1470000 | 13097s | train_loss=0.0911
150
+ step 1480000 | 13177s | train_loss=0.0899
151
+ step 1490000 | 13260s | train_loss=0.0897
152
+ step 1500000 | 13413s | train_loss=0.0890 | fid=86.8134
153
+ step 1510000 | 13496s | train_loss=0.0882
154
+ step 1520000 | 13574s | train_loss=0.0883
155
+ step 1530000 | 13657s | train_loss=0.0890
156
+ step 1540000 | 13738s | train_loss=0.0867
157
+ step 1550000 | 13821s | train_loss=0.0866
158
+ step 1560000 | 13904s | train_loss=0.0862
159
+ step 1570000 | 13987s | train_loss=0.0848
160
+ step 1580000 | 14070s | train_loss=0.0851
161
+ step 1590000 | 14150s | train_loss=0.0843
162
+ step 1600000 | 14302s | train_loss=0.0861 | fid=87.1846
163
+ step 1610000 | 14385s | train_loss=0.0858
164
+ step 1620000 | 14465s | train_loss=0.0817
165
+ step 1630000 | 14547s | train_loss=0.0834
166
+ step 1640000 | 14627s | train_loss=0.0821
167
+ step 1650000 | 14710s | train_loss=0.0838
168
+ step 1660000 | 14793s | train_loss=0.0814
169
+ step 1670000 | 14876s | train_loss=0.0824
170
+ step 1680000 | 14959s | train_loss=0.0817
171
+ step 1690000 | 15039s | train_loss=0.0804
172
+ step 1700000 | 15202s | train_loss=0.0790 | fid=87.5335
173
+ step 1710000 | 15285s | train_loss=0.0789
174
+ step 1720000 | 15362s | train_loss=0.0794
175
+ step 1730000 | 15441s | train_loss=0.0784
176
+ step 1740000 | 15510s | train_loss=0.0782
177
+ step 1750000 | 15590s | train_loss=0.0784
178
+ step 1760000 | 15673s | train_loss=0.0769
179
+ step 1770000 | 15756s | train_loss=0.0789
180
+ step 1780000 | 15839s | train_loss=0.0776
181
+ step 1790000 | 15922s | train_loss=0.0783
182
+ step 1800000 | 16075s | train_loss=0.0795 | fid=86.8615
183
+ step 1810000 | 16159s | train_loss=0.0758
184
+ step 1820000 | 16241s | train_loss=0.0775
185
+ step 1830000 | 16322s | train_loss=0.0774
186
+ step 1840000 | 16404s | train_loss=0.0762
187
+ step 1850000 | 16482s | train_loss=0.0769
188
+ step 1860000 | 16565s | train_loss=0.0761
189
+ step 1870000 | 16648s | train_loss=0.0763
190
+ step 1880000 | 16731s | train_loss=0.0755
191
+ step 1890000 | 16814s | train_loss=0.0759
192
+ step 1900000 | 16965s | train_loss=0.0733 | fid=87.0565
193
+ step 1910000 | 17048s | train_loss=0.0746
194
+ step 1920000 | 17131s | train_loss=0.0734
195
+ step 1930000 | 17211s | train_loss=0.0739
196
+ step 1940000 | 17294s | train_loss=0.0757
197
+ step 1950000 | 17372s | train_loss=0.0757
198
+ step 1960000 | 17455s | train_loss=0.0730
199
+ step 1970000 | 17538s | train_loss=0.0719
200
+ step 1980000 | 17621s | train_loss=0.0729
201
+ step 1990000 | 17704s | train_loss=0.0714
202
+ step 2000000 | 17855s | train_loss=0.0728 | fid=88.2613
203
+ step 2010000 | 17938s | train_loss=0.0721
204
+ step 2020000 | 18021s | train_loss=0.0708
205
+ step 2030000 | 18101s | train_loss=0.0725
206
+ step 2040000 | 18185s | train_loss=0.0734
207
+ step 2050000 | 18264s | train_loss=0.0705
208
+ step 2060000 | 18346s | train_loss=0.0711
209
+ step 2070000 | 18429s | train_loss=0.0702
210
+ step 2080000 | 18512s | train_loss=0.0720
211
+ step 2090000 | 18595s | train_loss=0.0693
212
+ step 2100000 | 18748s | train_loss=0.0692 | fid=87.9777
213
+ step 2110000 | 18828s | train_loss=0.0701
214
+ step 2120000 | 18911s | train_loss=0.0685
215
+ step 2130000 | 18991s | train_loss=0.0684
216
+ step 2140000 | 19074s | train_loss=0.0677
217
+ step 2150000 | 19155s | train_loss=0.0678
218
+ step 2160000 | 19238s | train_loss=0.0676
219
+ step 2170000 | 19319s | train_loss=0.0678
220
+ step 2180000 | 19402s | train_loss=0.0677
221
+ step 2190000 | 19485s | train_loss=0.0689
222
+ step 2200000 | 19648s | train_loss=0.0688 | fid=87.4676
223
+ step 2210000 | 19728s | train_loss=0.0691
224
+ step 2220000 | 19811s | train_loss=0.0668
225
+ step 2230000 | 19890s | train_loss=0.0667
226
+ step 2240000 | 19973s | train_loss=0.0658
227
+ step 2250000 | 20053s | train_loss=0.0659
228
+ step 2260000 | 20136s | train_loss=0.0661
229
+ step 2270000 | 20219s | train_loss=0.0646
230
+ step 2280000 | 20301s | train_loss=0.0655
231
+ step 2290000 | 20384s | train_loss=0.0646
232
+ step 2300000 | 20537s | train_loss=0.0659 | fid=86.6900
233
+ step 2310000 | 20620s | train_loss=0.0656
234
+ step 2320000 | 20700s | train_loss=0.0650
235
+ step 2330000 | 20780s | train_loss=0.0657
236
+ step 2340000 | 20863s | train_loss=0.0626
237
+ step 2350000 | 20943s | train_loss=0.0638
238
+ step 2360000 | 21026s | train_loss=0.0631
239
+ step 2370000 | 21109s | train_loss=0.0629
240
+ step 2380000 | 21191s | train_loss=0.0618
241
+ step 2390000 | 21273s | train_loss=0.0615
242
+ step 2400000 | 21426s | train_loss=0.0632 | fid=86.5871
243
+ step 2410000 | 21509s | train_loss=0.0622
244
+ step 2420000 | 21589s | train_loss=0.0629
245
+ step 2430000 | 21670s | train_loss=0.0618
246
+ step 2440000 | 21753s | train_loss=0.0618
247
+ step 2450000 | 21836s | train_loss=0.0628
248
+ step 2460000 | 21916s | train_loss=0.0622
249
+ step 2470000 | 21999s | train_loss=0.0632
250
+ step 2480000 | 22082s | train_loss=0.0617
251
+ step 2490000 | 22163s | train_loss=0.0624
252
+ step 2500000 | 22316s | train_loss=0.0630 | fid=87.5387
253
+ step 2510000 | 22399s | train_loss=0.0608
254
+ step 2520000 | 22482s | train_loss=0.0615
255
+ step 2530000 | 22559s | train_loss=0.0616
256
+ step 2540000 | 22642s | train_loss=0.0612
257
+ step 2550000 | 22725s | train_loss=0.0613
258
+ step 2560000 | 22805s | train_loss=0.0600
259
+ step 2570000 | 22888s | train_loss=0.0599
260
+ step 2580000 | 22971s | train_loss=0.0604
261
+ step 2590000 | 23054s | train_loss=0.0604
262
+ step 2600000 | 23200s | train_loss=0.0597 | fid=86.8934
263
+ step 2610000 | 23283s | train_loss=0.0586
264
+ step 2620000 | 23366s | train_loss=0.0610
265
+ step 2630000 | 23443s | train_loss=0.0586
266
+ step 2640000 | 23525s | train_loss=0.0603
267
+ step 2650000 | 23608s | train_loss=0.0586
268
+ step 2660000 | 23688s | train_loss=0.0588
269
+ step 2670000 | 23771s | train_loss=0.0590
270
+ step 2680000 | 23854s | train_loss=0.0596
271
+ step 2690000 | 23937s | train_loss=0.0577
272
+ step 2700000 | 24083s | train_loss=0.0579 | fid=86.6740
273
+ step 2710000 | 24166s | train_loss=0.0584
274
+ step 2720000 | 24249s | train_loss=0.0576
275
+ step 2730000 | 24330s | train_loss=0.0585
276
+ step 2740000 | 24408s | train_loss=0.0580
277
+ step 2750000 | 24491s | train_loss=0.0581
278
+ step 2760000 | 24571s | train_loss=0.0564
279
+ step 2770000 | 24654s | train_loss=0.0595
280
+ step 2780000 | 24737s | train_loss=0.0585
281
+ step 2790000 | 24820s | train_loss=0.0577
282
+ step 2800000 | 24968s | train_loss=0.0561 | fid=87.1146
283
+ step 2810000 | 25049s | train_loss=0.0580
284
+ step 2820000 | 25132s | train_loss=0.0571
285
+ step 2830000 | 25215s | train_loss=0.0573
286
+ step 2840000 | 25295s | train_loss=0.0561
287
+ step 2850000 | 25375s | train_loss=0.0565
288
+ step 2860000 | 25458s | train_loss=0.0566
289
+ step 2870000 | 25538s | train_loss=0.0566
290
+ step 2880000 | 25621s | train_loss=0.0569
291
+ step 2890000 | 25704s | train_loss=0.0556
292
+ step 2900000 | 25851s | train_loss=0.0548 | fid=86.2398
293
+ step 2910000 | 25934s | train_loss=0.0566
294
+ step 2920000 | 26016s | train_loss=0.0554
295
+ step 2930000 | 26099s | train_loss=0.0559
296
+ step 2940000 | 26179s | train_loss=0.0549
297
+ step 2950000 | 26258s | train_loss=0.0561
298
+ step 2960000 | 26341s | train_loss=0.0553
299
+ step 2970000 | 26422s | train_loss=0.0556
300
+ step 2980000 | 26505s | train_loss=0.0552
301
+ step 2990000 | 26588s | train_loss=0.0557
302
+ step 3000000 | 26735s | train_loss=0.0547 | fid=86.7772
303
+ step 3010000 | 26818s | train_loss=0.0540
304
+ step 3020000 | 26902s | train_loss=0.0541
305
+ step 3030000 | 26983s | train_loss=0.0536
306
+ step 3040000 | 27062s | train_loss=0.0550
307
+ step 3050000 | 27145s | train_loss=0.0542
308
+ step 3060000 | 27225s | train_loss=0.0532
309
+ step 3070000 | 27305s | train_loss=0.0536
310
+ step 3080000 | 27388s | train_loss=0.0543
311
+ step 3090000 | 27471s | train_loss=0.0527
312
+ step 3100000 | 27618s | train_loss=0.0519 | fid=86.8592
313
+ step 3110000 | 27701s | train_loss=0.0530
314
+ step 3120000 | 27784s | train_loss=0.0548
315
+ step 3130000 | 27867s | train_loss=0.0534
316
+ step 3140000 | 27945s | train_loss=0.0534
317
+ step 3150000 | 28028s | train_loss=0.0536
318
+ step 3160000 | 28111s | train_loss=0.0530
319
+ step 3170000 | 28190s | train_loss=0.0517
320
+ step 3180000 | 28272s | train_loss=0.0518
321
+ step 3190000 | 28355s | train_loss=0.0523
322
+ step 3200000 | 28503s | train_loss=0.0511 | fid=86.0674
323
+ step 3210000 | 28585s | train_loss=0.0505
324
+ step 3220000 | 28668s | train_loss=0.0515
325
+ step 3230000 | 28750s | train_loss=0.0515
326
+ step 3240000 | 28828s | train_loss=0.0516
327
+ step 3250000 | 28911s | train_loss=0.0521
328
+ step 3260000 | 28994s | train_loss=0.0516
329
+ step 3270000 | 29075s | train_loss=0.0509
330
+ step 3280000 | 29156s | train_loss=0.0519
331
+ step 3290000 | 29239s | train_loss=0.0526
332
+ step 3300000 | 29386s | train_loss=0.0513 | fid=86.5214
333
+ step 3310000 | 29469s | train_loss=0.0496
334
+ step 3320000 | 29551s | train_loss=0.0506
335
+ step 3330000 | 29633s | train_loss=0.0508
336
+ step 3340000 | 29715s | train_loss=0.0518
337
+ step 3350000 | 29794s | train_loss=0.0511
338
+ step 3360000 | 29877s | train_loss=0.0505
339
+ step 3370000 | 29960s | train_loss=0.0503
340
+ step 3380000 | 30038s | train_loss=0.0509
341
+ step 3390000 | 30121s | train_loss=0.0502
342
+ step 3400000 | 30270s | train_loss=0.0501 | fid=87.9564
343
+ step 3410000 | 30353s | train_loss=0.0504
344
+ step 3420000 | 30436s | train_loss=0.0496
345
+ step 3430000 | 30519s | train_loss=0.0492
346
+ step 3440000 | 30602s | train_loss=0.0490
347
+ step 3450000 | 30683s | train_loss=0.0495
348
+ step 3460000 | 30764s | train_loss=0.0494
349
+ step 3470000 | 30848s | train_loss=0.0489
350
+ step 3480000 | 30920s | train_loss=0.0490
351
+ step 3490000 | 30989s | train_loss=0.0491
352
+ step 3500000 | 31136s | train_loss=0.0479 | fid=87.7294
353
+ step 3510000 | 31220s | train_loss=0.0487
354
+ step 3520000 | 31303s | train_loss=0.0484
355
+ step 3530000 | 31386s | train_loss=0.0487
356
+ step 3540000 | 31469s | train_loss=0.0491
357
+ step 3550000 | 31540s | train_loss=0.0505
358
+ step 3560000 | 31621s | train_loss=0.0496
359
+ step 3570000 | 31702s | train_loss=0.0487
360
+ step 3580000 | 31785s | train_loss=0.0485
361
+ step 3590000 | 31860s | train_loss=0.0480
362
+ step 3600000 | 32002s | train_loss=0.0479 | fid=86.2252
363
+ step 3610000 | 32085s | train_loss=0.0472
364
+ step 3620000 | 32168s | train_loss=0.0482
365
+ step 3630000 | 32251s | train_loss=0.0487
366
+ step 3640000 | 32334s | train_loss=0.0471
367
+ step 3650000 | 32417s | train_loss=0.0466
368
+ step 3660000 | 32497s | train_loss=0.0478
369
+ step 3670000 | 32580s | train_loss=0.0478
370
+ step 3680000 | 32662s | train_loss=0.0481
371
+ step 3690000 | 32744s | train_loss=0.0469
372
+ step 3700000 | 32888s | train_loss=0.0472 | fid=87.6140
373
+ step 3710000 | 32971s | train_loss=0.0473
374
+ step 3720000 | 33054s | train_loss=0.0478
375
+ step 3730000 | 33136s | train_loss=0.0479
376
+ step 3740000 | 33219s | train_loss=0.0469
377
+ step 3750000 | 33302s | train_loss=0.0463
378
+ step 3760000 | 33383s | train_loss=0.0464
379
+ step 3770000 | 33466s | train_loss=0.0462
380
+ step 3780000 | 33549s | train_loss=0.0468
381
+ step 3790000 | 33631s | train_loss=0.0461
382
+ step 3800000 | 33771s | train_loss=0.0475 | fid=86.9839
383
+ step 3810000 | 33851s | train_loss=0.0470
384
+ step 3820000 | 33934s | train_loss=0.0467
385
+ step 3830000 | 34017s | train_loss=0.0466
386
+ step 3840000 | 34100s | train_loss=0.0462
387
+ step 3850000 | 34183s | train_loss=0.0464
388
+ step 3860000 | 34258s | train_loss=0.0460
389
+ step 3870000 | 34341s | train_loss=0.0458
390
+ step 3880000 | 34424s | train_loss=0.0465
391
+ step 3890000 | 34508s | train_loss=0.0462
392
+ step 3900000 | 34644s | train_loss=0.0456 | fid=87.1263
393
+ step 3910000 | 34721s | train_loss=0.0455
394
+ step 3920000 | 34795s | train_loss=0.0459
395
+ step 3930000 | 34878s | train_loss=0.0452
396
+ step 3940000 | 34961s | train_loss=0.0463
397
+ step 3950000 | 35044s | train_loss=0.0459
398
+ step 3960000 | 35126s | train_loss=0.0456
399
+ step 3970000 | 35196s | train_loss=0.0454
400
+ step 3980000 | 35279s | train_loss=0.0448
401
+ step 3990000 | 35362s | train_loss=0.0445
402
+ step 4000000 | 35500s | train_loss=0.0458 | fid=87.5912
403
+ step 4010000 | 35583s | train_loss=0.0449
404
+ step 4020000 | 35666s | train_loss=0.0446
405
+ step 4030000 | 35739s | train_loss=0.0452
406
+ step 4040000 | 35822s | train_loss=0.0459
407
+ step 4050000 | 35905s | train_loss=0.0452
408
+ step 4060000 | 35988s | train_loss=0.0445
409
+ step 4070000 | 36068s | train_loss=0.0450
410
+ step 4080000 | 36142s | train_loss=0.0449
411
+ step 4090000 | 36225s | train_loss=0.0444
412
+ step 4100000 | 36373s | train_loss=0.0436 | fid=87.8189
413
+ step 4110000 | 36441s | train_loss=0.0457
414
+ step 4120000 | 36524s | train_loss=0.0443
415
+ step 4130000 | 36607s | train_loss=0.0440
416
+ step 4140000 | 36676s | train_loss=0.0437
417
+ step 4150000 | 36759s | train_loss=0.0443
418
+ step 4160000 | 36842s | train_loss=0.0432
419
+ step 4170000 | 36925s | train_loss=0.0434
420
+ step 4180000 | 37001s | train_loss=0.0442
421
+ step 4190000 | 37078s | train_loss=0.0434
422
+ step 4200000 | 37226s | train_loss=0.0442 | fid=87.8985
423
+ step 4210000 | 37309s | train_loss=0.0443
424
+ step 4220000 | 37377s | train_loss=0.0445
425
+ step 4230000 | 37460s | train_loss=0.0430
426
+ step 4240000 | 37543s | train_loss=0.0433
427
+ step 4250000 | 37615s | train_loss=0.0433
428
+ step 4260000 | 37695s | train_loss=0.0428
429
+ step 4270000 | 37778s | train_loss=0.0422
430
+ step 4280000 | 37862s | train_loss=0.0434
431
+ step 4290000 | 37934s | train_loss=0.0440
432
+ step 4300000 | 38080s | train_loss=0.0423 | fid=87.2587
433
+ step 4310000 | 38163s | train_loss=0.0432
434
+ step 4320000 | 38244s | train_loss=0.0429
435
+ step 4330000 | 38312s | train_loss=0.0425
436
+ step 4340000 | 38395s | train_loss=0.0420
437
+ step 4350000 | 38478s | train_loss=0.0432
438
+ step 4360000 | 38555s | train_loss=0.0435
439
+ step 4370000 | 38634s | train_loss=0.0430
440
+ step 4380000 | 38717s | train_loss=0.0427
441
+ step 4390000 | 38800s | train_loss=0.0428
442
+ step 4400000 | 38934s | train_loss=0.0415 | fid=86.4329
443
+ step 4410000 | 39017s | train_loss=0.0427
444
+ step 4420000 | 39100s | train_loss=0.0425
445
+ step 4430000 | 39181s | train_loss=0.0425
446
+ step 4440000 | 39249s | train_loss=0.0427
447
+ step 4450000 | 39333s | train_loss=0.0428
448
+ step 4460000 | 39416s | train_loss=0.0425
449
+ step 4470000 | 39493s | train_loss=0.0425
450
+ step 4480000 | 39573s | train_loss=0.0418
451
+ step 4490000 | 39657s | train_loss=0.0426
452
+ step 4500000 | 39794s | train_loss=0.0423 | fid=86.6099
453
+ step 4510000 | 39878s | train_loss=0.0421
454
+ step 4520000 | 39961s | train_loss=0.0419
455
+ step 4530000 | 40044s | train_loss=0.0418
456
+ step 4540000 | 40125s | train_loss=0.0411
457
+ step 4550000 | 40194s | train_loss=0.0408
458
+ step 4560000 | 40277s | train_loss=0.0410
459
+ step 4570000 | 40360s | train_loss=0.0412
460
+ step 4580000 | 40437s | train_loss=0.0400
461
+ step 4590000 | 40516s | train_loss=0.0413
462
+ step 4600000 | 40662s | train_loss=0.0420 | fid=88.1404
463
+ step 4610000 | 40738s | train_loss=0.0412
464
+ step 4620000 | 40821s | train_loss=0.0420
465
+ step 4630000 | 40904s | train_loss=0.0418
466
+ step 4640000 | 40987s | train_loss=0.0412
467
+ step 4650000 | 41066s | train_loss=0.0405
468
+ step 4660000 | 41138s | train_loss=0.0402
469
+ step 4670000 | 41220s | train_loss=0.0412
470
+ step 4680000 | 41303s | train_loss=0.0413
471
+ step 4690000 | 41380s | train_loss=0.0410
472
+ step 4700000 | 41526s | train_loss=0.0406 | fid=87.3257
473
+ step 4710000 | 41603s | train_loss=0.0404
474
+ step 4720000 | 41686s | train_loss=0.0419
475
+ step 4730000 | 41768s | train_loss=0.0407
476
+ step 4740000 | 41851s | train_loss=0.0406
477
+ step 4750000 | 41934s | train_loss=0.0398
478
+ step 4760000 | 42011s | train_loss=0.0411
479
+ step 4770000 | 42094s | train_loss=0.0406
480
+ step 4780000 | 42177s | train_loss=0.0397
481
+ step 4790000 | 42259s | train_loss=0.0407
482
+ step 4800000 | 42403s | train_loss=0.0395 | fid=87.6172
483
+ step 4810000 | 42482s | train_loss=0.0402
484
+ step 4820000 | 42565s | train_loss=0.0404
485
+ step 4830000 | 42647s | train_loss=0.0405
486
+ step 4840000 | 42730s | train_loss=0.0407
487
+ step 4850000 | 42813s | train_loss=0.0395
488
+ step 4860000 | 42894s | train_loss=0.0403
489
+ step 4870000 | 42975s | train_loss=0.0399
490
+ step 4880000 | 43057s | train_loss=0.0396
491
+ step 4890000 | 43140s | train_loss=0.0397
492
+ step 4900000 | 43284s | train_loss=0.0392 | fid=86.9381
493
+ step 4910000 | 43366s | train_loss=0.0399
494
+ step 4920000 | 43446s | train_loss=0.0395
495
+ step 4930000 | 43528s | train_loss=0.0397
496
+ step 4940000 | 43611s | train_loss=0.0400
497
+ step 4950000 | 43694s | train_loss=0.0394
498
+ step 4960000 | 43776s | train_loss=0.0401
499
+ step 4970000 | 43857s | train_loss=0.0402
500
+ step 4980000 | 43938s | train_loss=0.0400
501
+ step 4990000 | 44021s | train_loss=0.0393
502
+ step 5000000 | 44168s | train_loss=0.0396 | fid=86.9596
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed43/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/seed43/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": 43,
25
+ "device": null
26
+ }
results/celeba_diverse1k_bigmlp_sgd_lr001_m08_10k_5m/seed43/d100/metrics.jsonl ADDED
The diff for this file is too large to render. See raw diff