Upload folder using huggingface_hub
Browse files- dt0_1_continuous/log.txt +562 -0
dt0_1_continuous/log.txt
ADDED
|
@@ -0,0 +1,562 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
fter patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 2 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 3 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 4 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [10:32<00:00, 3.25s/it]
|
| 5 |
+
FID is 35.09709548950195
|
| 6 |
+
Calc FID for CFG 1.0 and denoise_timesteps 64
|
| 7 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 8 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 9 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 10 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 11 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 12 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [05:16<00:00, 1.62s/it]
|
| 13 |
+
FID is 31.059368133544922
|
| 14 |
+
Calc FID for CFG 1.0 and denoise_timesteps 32
|
| 15 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 16 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 17 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 18 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 19 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 20 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [03:19<00:00, 1.02s/it]
|
| 21 |
+
FID is 28.118967056274414
|
| 22 |
+
Calc FID for CFG 1.0 and denoise_timesteps 16
|
| 23 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 24 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 25 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 26 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 27 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 28 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [02:20<00:00, 1.39it/s]
|
| 29 |
+
FID is 26.224519729614258
|
| 30 |
+
Calc FID for CFG 1.0 and denoise_timesteps 8
|
| 31 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 32 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 33 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 34 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 35 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 36 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [01:51<00:00, 1.76it/s]
|
| 37 |
+
FID is 24.95024871826172
|
| 38 |
+
Calc FID for CFG 1.0 and denoise_timesteps 4
|
| 39 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 40 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 41 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 42 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 43 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 44 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββοΏ½οΏ½βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [01:35<00:00, 2.04it/s]
|
| 45 |
+
FID is 24.247482299804688
|
| 46 |
+
Calc FID for CFG 1.0 and denoise_timesteps 2
|
| 47 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 48 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 49 |
+
embed, shape is (256, 256, 768) dtype bfloat16
|
| 50 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 51 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 52 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [01:27<00:00, 2.22it/s]
|
| 53 |
+
FID is 26.227968215942383
|
| 54 |
+
Calc FID for CFG 1.0 and denoise_timesteps 1
|
| 55 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 56 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 57 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 58 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 59 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 60 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [01:24<00:00, 2.30it/s]
|
| 61 |
+
FID is 35.41545867919922
|
| 62 |
+
Calc FID for CFG 1.25 and denoise_timesteps 128
|
| 63 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 64 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 65 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 66 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 67 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 68 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [16:53<00:00, 5.20s/it]
|
| 69 |
+
FID is 20.689212799072266
|
| 70 |
+
Calc FID for CFG 1.25 and denoise_timesteps 64
|
| 71 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 72 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 73 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 74 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 75 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 76 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [09:06<00:00, 2.80s/it]
|
| 77 |
+
FID is 18.268905639648438
|
| 78 |
+
Calc FID for CFG 1.25 and denoise_timesteps 32
|
| 79 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 80 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 81 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 82 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 83 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 84 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [05:13<00:00, 1.61s/it]
|
| 85 |
+
FID is 16.920024871826172
|
| 86 |
+
Calc FID for CFG 1.25 and denoise_timesteps 16
|
| 87 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 88 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 89 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 90 |
+
patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 91 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 92 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 93 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [03:17<00:00, 1.01s/it]
|
| 94 |
+
FID is 16.27607536315918
|
| 95 |
+
Calc FID for CFG 1.25 and denoise_timesteps 8
|
| 96 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 97 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 98 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 99 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 100 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 101 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [02:18<00:00, 1.40it/s]
|
| 102 |
+
FID is 16.279489517211914
|
| 103 |
+
Calc FID for CFG 1.25 and denoise_timesteps 4
|
| 104 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 105 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 106 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 107 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 108 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 109 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [01:50<00:00, 1.77it/s]
|
| 110 |
+
FID is 16.93604278564453
|
| 111 |
+
Calc FID for CFG 1.25 and denoise_timesteps 2
|
| 112 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 113 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 114 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 115 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 116 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 117 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [01:35<00:00, 2.05it/s]
|
| 118 |
+
FID is 20.04778289794922
|
| 119 |
+
Calc FID for CFG 1.25 and denoise_timesteps 1
|
| 120 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 121 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 122 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 123 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 124 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 125 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [01:28<00:00, 2.20it/s]
|
| 126 |
+
FID is 28.32234764099121
|
| 127 |
+
Calc FID for CFG 1.5 and denoise_timesteps 128
|
| 128 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 129 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 130 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 131 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 132 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 133 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [16:52<00:00, 5.19s/it]
|
| 134 |
+
FID is 12.754450798034668
|
| 135 |
+
Calc FID for CFG 1.5 and denoise_timesteps 64
|
| 136 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 137 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 138 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 139 |
+
G 1.5 and denoise_timesteps 64
|
| 140 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 141 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 142 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 143 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 144 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 145 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [09:07<00:00, 2.81s/it]
|
| 146 |
+
FID is 11.823137283325195
|
| 147 |
+
Calc FID for CFG 1.5 and denoise_timesteps 32
|
| 148 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 149 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 150 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 151 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 152 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 153 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [05:13<00:00, 1.61s/it]
|
| 154 |
+
FID is 11.53042984008789
|
| 155 |
+
Calc FID for CFG 1.5 and denoise_timesteps 16
|
| 156 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 157 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 158 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 159 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 160 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 161 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [03:17<00:00, 1.01s/it]
|
| 162 |
+
FID is 11.795125961303711
|
| 163 |
+
Calc FID for CFG 1.5 and denoise_timesteps 8
|
| 164 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 165 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 166 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 167 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 168 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 169 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [02:18<00:00, 1.40it/s]
|
| 170 |
+
FID is 12.550365447998047
|
| 171 |
+
Calc FID for CFG 1.5 and denoise_timesteps 4
|
| 172 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 173 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 174 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 175 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 176 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 177 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [01:49<00:00, 1.78it/s]
|
| 178 |
+
FID is 14.204133987426758
|
| 179 |
+
Calc FID for CFG 1.5 and denoise_timesteps 2
|
| 180 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 181 |
+
DiT: Input of shape (256, 32, 32, 4) dtype flo
|
| 182 |
+
embed, shape is (256, 256, 768) dtype bfloat16
|
| 183 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 184 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 185 |
+
100%|ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββοΏ½οΏ½οΏ½ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [01:35<00:00, 2.04it/s]
|
| 186 |
+
FID is 18.149208068847656
|
| 187 |
+
Calc FID for CFG 1.5 and denoise_timesteps 1
|
| 188 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 189 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 190 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 191 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 192 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 193 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [01:28<00:00, 2.21it/s]
|
| 194 |
+
FID is 28.582168579101562
|
| 195 |
+
Calc FID for CFG 1.75 and denoise_timesteps 128
|
| 196 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 197 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 198 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 199 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 200 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 201 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [16:53<00:00, 5.20s/it]
|
| 202 |
+
FID is 9.033015251159668
|
| 203 |
+
Calc FID for CFG 1.75 and denoise_timesteps 64
|
| 204 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 205 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 206 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 207 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 208 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 209 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [09:07<00:00, 2.81s/it]
|
| 210 |
+
FID is 8.93721866607666
|
| 211 |
+
Calc FID for CFG 1.75 and denoise_timesteps 32
|
| 212 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 213 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 214 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 215 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 216 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 217 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [05:14<00:00, 1.61s/it]
|
| 218 |
+
FID is 9.311577796936035
|
| 219 |
+
Calc FID for CFG 1.75 and denoise_timesteps 16e (256, 256, 768) dtype bfloat16
|
| 220 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 221 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [03:17<00:00, 1.01s/it]
|
| 222 |
+
FID is 10.116938591003418
|
| 223 |
+
Calc FID for CFG 1.75 and denoise_timesteps 8
|
| 224 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 225 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 226 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 227 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 228 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 229 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [02:19<00:00, 1.40it/s]
|
| 230 |
+
FID is 11.253520965576172
|
| 231 |
+
Calc FID for CFG 1.75 and denoise_timesteps 4
|
| 232 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 233 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 234 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 235 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 236 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 237 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [01:50<00:00, 1.77it/s]
|
| 238 |
+
FID is 13.326236724853516
|
| 239 |
+
Calc FID for CFG 1.75 and denoise_timesteps 2
|
| 240 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 241 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 242 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 243 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 244 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 245 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [01:35<00:00, 2.05it/s]
|
| 246 |
+
FID is 18.064693450927734
|
| 247 |
+
Calc FID for CFG 1.75 and denoise_timesteps 1
|
| 248 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 249 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 250 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 251 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 252 |
+
ditioning of shape (256, 768) dtype float32
|
| 253 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [01:28<00:00, 2.21it/s]
|
| 254 |
+
FID is 32.303932189941406
|
| 255 |
+
Calc FID for CFG 2.0 and denoise_timesteps 128
|
| 256 |
+
0%| 256, 32, 32, 4) dtype float32
|
| 257 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 258 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 259 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 260 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [16:54<00:00, 5.20s/it]
|
| 261 |
+
FID is 7.695302963256836
|
| 262 |
+
Calc FID for CFG 2.0 and denoise_timesteps 64
|
| 263 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 264 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 265 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 266 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 267 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 268 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [09:07<00:00, 2.81s/it]
|
| 269 |
+
FID is 8.14946460723877
|
| 270 |
+
Calc FID for CFG 2.0 and denoise_timesteps 32
|
| 271 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 272 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 273 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 274 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 275 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 276 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [05:14<00:00, 1.61s/it]
|
| 277 |
+
FID is 8.860971450805664
|
| 278 |
+
Calc FID for CFG 2.0 and denoise_timesteps 16
|
| 279 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 280 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 281 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 282 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 283 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 284 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [03:17<00:00, 1.01s/it]
|
| 285 |
+
FID is 9.84195327758789
|
| 286 |
+
Calc FID for CFG 2.0 and denoise_timesteps 8
|
| 287 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 288 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 289 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 290 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 291 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 292 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [02:19<00:00, 1.40it/s]
|
| 293 |
+
FID is 11.091032028198242
|
| 294 |
+
Calc FID for CFG 2.0 and denoise_timesteps 4
|
| 295 |
+
0%|
|
| 296 |
+
, shape is (256, 256, 768) dtype bfloat16
|
| 297 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 298 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 299 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [01:49<00:00, 1.77it/s]
|
| 300 |
+
FID is 13.3704833984375
|
| 301 |
+
Calc FID for CFG 2.0 and denoise_timesteps 2
|
| 302 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 303 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 304 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 305 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 306 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 307 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [01:35<00:00, 2.04it/s]
|
| 308 |
+
FID is 18.816028594970703
|
| 309 |
+
Calc FID for CFG 2.0 and denoise_timesteps 1
|
| 310 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 311 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 312 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 313 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 314 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 315 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [01:28<00:00, 2.21it/s]
|
| 316 |
+
FID is 37.690826416015625
|
| 317 |
+
Calc FID for CFG 2.25 and denoise_timesteps 128
|
| 318 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 319 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 320 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 321 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 322 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 323 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [16:54<00:00, 5.20s/it]
|
| 324 |
+
FID is 7.674609661102295
|
| 325 |
+
Calc FID for CFG 2.25 and denoise_timesteps 64
|
| 326 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 327 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 328 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 329 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 330 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 331 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [09:07<00:00, 2.81s/it]
|
| 332 |
+
FID is 8.370298385620117
|
| 333 |
+
Calc FID for CFG 2.25 and denoise_timesteps 32
|
| 334 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 335 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 336 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 337 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 338 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 339 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [05:14<00:00, 1.61s/it]
|
| 340 |
+
FID is 9.211350440979004
|
| 341 |
+
Calc FID for CFG 2.25 and denoise_timesteps 16
|
| 342 |
+
0%| embed, shape is (256, 256, 768) dtype bfloat16
|
| 343 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 344 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 345 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [03:17<00:00, 1.01s/it]
|
| 346 |
+
FID is 10.200955390930176
|
| 347 |
+
Calc FID for CFG 2.25 and denoise_timesteps 8
|
| 348 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 349 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 350 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 351 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 352 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 353 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [02:19<00:00, 1.40it/s]
|
| 354 |
+
FID is 11.429386138916016
|
| 355 |
+
Calc FID for CFG 2.25 and denoise_timesteps 4
|
| 356 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 357 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 358 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 359 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 360 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 361 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [01:50<00:00, 1.77it/s]
|
| 362 |
+
FID is 13.825882911682129
|
| 363 |
+
Calc FID for CFG 2.25 and denoise_timesteps 2
|
| 364 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 365 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 366 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 367 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 368 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 369 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [01:35<00:00, 2.04it/s]
|
| 370 |
+
FID is 20.189815521240234
|
| 371 |
+
Calc FID for CFG 2.25 and denoise_timesteps 1
|
| 372 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 373 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 374 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 375 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 376 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 377 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [01:28<00:00, 2.21it/s]
|
| 378 |
+
FID is 43.617523193359375
|
| 379 |
+
Calc FID for CFG 2.5 and denoise_timest23193359375 [137/1984]
|
| 380 |
+
Calc FID for CFG 2.5 and denoise_timesteps 128
|
| 381 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 382 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 383 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 384 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 385 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 386 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [16:55<00:00, 5.21s/it]
|
| 387 |
+
FID is 8.270273208618164
|
| 388 |
+
Calc FID for CFG 2.5 and denoise_timesteps 64
|
| 389 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 390 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 391 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 392 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 393 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 394 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [09:07<00:00, 2.81s/it]
|
| 395 |
+
FID is 9.068354606628418
|
| 396 |
+
Calc FID for CFG 2.5 and denoise_timesteps 32
|
| 397 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 398 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 399 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 400 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 401 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 402 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [05:14<00:00, 1.61s/it]
|
| 403 |
+
FID is 9.953636169433594
|
| 404 |
+
Calc FID for CFG 2.5 and denoise_timesteps 16
|
| 405 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 406 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 407 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 408 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 409 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 410 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [03:17<00:00, 1.02s/it]
|
| 411 |
+
FID is 10.890295028686523
|
| 412 |
+
Calc FID for CFG 2.5 and denoise_timesteps 8
|
| 413 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 414 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 415 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 416 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 417 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 418 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [02:19<00:00, 1.40it/s]
|
| 419 |
+
FID is 12.007729530334473
|
| 420 |
+
Calc FID for CFG 2.5 and denoise_timesteps 4
|
| 421 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 422 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 423 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 424 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 425 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 426 |
+
100%|ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββοΏ½οΏ½ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [01:50<00:00, 1.77it/s]
|
| 427 |
+
FID is 14.545646667480469
|
| 428 |
+
Calc FID for CFG 2.5 and denoise_timesteps 2
|
| 429 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 430 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 431 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 432 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 433 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 434 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [01:35<00:00, 2.03it/s$
|
| 435 |
+
FID is 22.039596557617188
|
| 436 |
+
Calc FID for CFG 2.5 and denoise_timestepsatch embed, shape is (256, 256, 768) dtype bfloat16
|
| 437 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 438 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 439 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [01:28<00:00, 2.21it/s]
|
| 440 |
+
FID is 49.47637939453125
|
| 441 |
+
Calc FID for CFG 2.75 and shape (256, 256, 768) dtype bfloat16
|
| 442 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 443 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [16:53<00:00, 5.20s/it]
|
| 444 |
+
FID is 9.153895378112793
|
| 445 |
+
Calc FID for CFG 2.75 and denoise_timesteps 64
|
| 446 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 447 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 448 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 449 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 450 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 451 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [09:07<00:00, 2.81s/it]
|
| 452 |
+
FID is 9.968490600585938
|
| 453 |
+
Calc FID for CFG 2.75 and denoise_timesteps 32
|
| 454 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 455 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 456 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 457 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 458 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 459 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [05:13<00:00, 1.61s/it]
|
| 460 |
+
FID is 10.834774017333984
|
| 461 |
+
Calc FID for CFG 2.75 and denoise_timesteps 16
|
| 462 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 463 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 464 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 465 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 466 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 467 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [03:17<00:00, 1.01s/it]
|
| 468 |
+
FID is 11.649170875549316
|
| 469 |
+
Calc FID for CFG 2.75 and denoise_timesteps 8
|
| 470 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 471 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 472 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 473 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 474 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 475 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [02:19<00:00, 1.40it/s]
|
| 476 |
+
FID is 12.659404754638672
|
| 477 |
+
Calc FID for CFG 2.75 and denoise_timesteps 4
|
| 478 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 479 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 480 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 481 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 482 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 483 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [01:50<00:00, 1.77it/s]
|
| 484 |
+
FID is 15.346792221069336
|
| 485 |
+
Calc FID for CFG 2.75 and denoise_timesteps 2
|
| 486 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 487 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 488 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 489 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 490 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 491 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [01:35<00:00, 2.04it/s]
|
| 492 |
+
FID is 24.261432647705078
|
| 493 |
+
Calc FID for CFG 2.75 and denoise_timesteps 1
|
| 494 |
+
nditioning of shape (256, 768) dtype float32
|
| 495 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [01:28<00:00, 2.21it/s]
|
| 496 |
+
FID is 54.95307922363281
|
| 497 |
+
Calc FID for CFG 3.0 and denoise_timesteps 128
|
| 498 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 499 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 500 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 501 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 502 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 503 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [16:52<00:00, 5.19s/it]
|
| 504 |
+
FID is 10.141067504882812
|
| 505 |
+
Calc FID for CFG 3.0 and denoise_timesteps 64
|
| 506 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 507 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 508 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 509 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 510 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 511 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [09:07<00:00, 2.81s/it]
|
| 512 |
+
FID is 10.961753845214844
|
| 513 |
+
Calc FID for CFG 3.0 and denoise_timesteps 32
|
| 514 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 515 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 516 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 517 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 518 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 519 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [05:13<00:00, 1.61s/it]
|
| 520 |
+
FID is 11.73600959777832
|
| 521 |
+
Calc FID for CFG 3.0 and denoise_timesteps 16
|
| 522 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 523 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 524 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 525 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 526 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 527 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [03:17<00:00, 1.01s/it]
|
| 528 |
+
FID is 12.452347755432129
|
| 529 |
+
Calc FID for CFG 3.0 and denoise_timesteps 8
|
| 530 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 531 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 532 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 533 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 534 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 535 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [02:19<00:00, 1.40it/s]
|
| 536 |
+
FID is 13.289637565612793
|
| 537 |
+
Calc FID for CFG 3.0 and denoise_timesteps 4
|
| 538 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 539 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 540 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 541 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 542 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 543 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [01:50<00:00, 1.77it/s]
|
| 544 |
+
FID is 16.239124298095703
|
| 545 |
+
CFG 2
|
| 546 |
+
ape (256, 32, 32, 4) dtype float32
|
| 547 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 548 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 549 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 550 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [01:35<00:00, 2.04it/s]
|
| 551 |
+
FID is 26.80874252319336
|
| 552 |
+
Calc FID for CFG 3.0 and denoise_timesteps 1
|
| 553 |
+
0%|
|
| 554 |
+
0%| | 0/195 [00:00<?, ?it/s]
|
| 555 |
+
DiT: Input of shape (256, 32, 32, 4) dtype float32
|
| 556 |
+
DiT: After patch embed, shape is (256, 256, 768) 2, 32, 4) dtype float32
|
| 557 |
+
DiT: After patch embed, shape is (256, 256, 768) dtype bfloat16
|
| 558 |
+
DiT: Patch Embed of shape (256, 256, 768) dtype bfloat16
|
| 559 |
+
DiT: Conditioning of shape (256, 768) dtype float32
|
| 560 |
+
100%|βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ| 195/195 [01:28<00:00, 2.21it/s]
|
| 561 |
+
FID is 60.0399055480957
|
| 562 |
+
|