Hugging Face's logo

MongolLabs
/
Shortcuts

Model card Files
xet
 Community
KublaiKhan1 commited on
Commit
7094714
·
verified ·
1 Parent(s): 7c33145

Upload folder using huggingface_hub

Browse files
Files changed (18) hide show
  1. .gitattributes +14 -0
  2. covariance_pixel/100001/100001.tmp +3 -0
  3. covariance_pixel/150001/150001.tmp +3 -0
  4. covariance_pixel/200001/200001.tmp +3 -0
  5. covariance_pixel/250001/250001.tmp +3 -0
  6. covariance_pixel/300001/300001.tmp +3 -0
  7. covariance_pixel/350001/350001.tmp +3 -0
  8. covariance_pixel/400001/400001.tmp +3 -0
  9. covariance_pixel/450001/450001.tmp +3 -0
  10. covariance_pixel/600001/600001.tmp +3 -0
  11. covariance_pixel/650001/650001.tmp +3 -0
  12. covariance_pixel/700001/700001.tmp +3 -0
  13. covariance_pixel/750001/750001.tmp +3 -0
  14. covariance_pixel/800001/800001.tmp +3 -0
  15. covariance_pixel/810001/810001.tmp +3 -0
  16. covariance_pixel/log.txt +516 -0
  17. covariance_pixel/mean_covar_pixel.tmp +0 -0
  18. covariance_pixel/std_covar_pixel.tmp +0 -0
.gitattributes CHANGED
@@ -236,3 +236,17 @@ f16_c16_large_patch1/650001/650001.tmp filter=lfs diff=lfs merge=lfs -text
236
  f16_c16_large_patch1/700001/700001.tmp filter=lfs diff=lfs merge=lfs -text
237
  f16_c16_large_patch1/750001/750001.tmp filter=lfs diff=lfs merge=lfs -text
238
  f16_c16_large_patch1/800001/800001.tmp filter=lfs diff=lfs merge=lfs -text
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
236
  f16_c16_large_patch1/700001/700001.tmp filter=lfs diff=lfs merge=lfs -text
237
  f16_c16_large_patch1/750001/750001.tmp filter=lfs diff=lfs merge=lfs -text
238
  f16_c16_large_patch1/800001/800001.tmp filter=lfs diff=lfs merge=lfs -text
239
+ covariance_pixel/100001/100001.tmp filter=lfs diff=lfs merge=lfs -text
240
+ covariance_pixel/150001/150001.tmp filter=lfs diff=lfs merge=lfs -text
241
+ covariance_pixel/200001/200001.tmp filter=lfs diff=lfs merge=lfs -text
242
+ covariance_pixel/250001/250001.tmp filter=lfs diff=lfs merge=lfs -text
243
+ covariance_pixel/300001/300001.tmp filter=lfs diff=lfs merge=lfs -text
244
+ covariance_pixel/350001/350001.tmp filter=lfs diff=lfs merge=lfs -text
245
+ covariance_pixel/400001/400001.tmp filter=lfs diff=lfs merge=lfs -text
246
+ covariance_pixel/450001/450001.tmp filter=lfs diff=lfs merge=lfs -text
247
+ covariance_pixel/600001/600001.tmp filter=lfs diff=lfs merge=lfs -text
248
+ covariance_pixel/650001/650001.tmp filter=lfs diff=lfs merge=lfs -text
249
+ covariance_pixel/700001/700001.tmp filter=lfs diff=lfs merge=lfs -text
250
+ covariance_pixel/750001/750001.tmp filter=lfs diff=lfs merge=lfs -text
251
+ covariance_pixel/800001/800001.tmp filter=lfs diff=lfs merge=lfs -text
252
+ covariance_pixel/810001/810001.tmp filter=lfs diff=lfs merge=lfs -text
covariance_pixel/100001/100001.tmp ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:53c930792720e6a84c28ff21abdc9e7d8a5e49256db9254cd1f24a61a5ed2c60
3
+ size 2097505391
covariance_pixel/150001/150001.tmp ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d7e54e5289f0a3270d07472d3edc8fb75299b0d0977dbaefe02234d70e375cb1
3
+ size 2097505391
covariance_pixel/200001/200001.tmp ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5967588664677a62f14fa504ec9635ee0696b85c959574611d8d9a99a6fba859
3
+ size 2097505391
covariance_pixel/250001/250001.tmp ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d5fac6c5386cc517075054a8d0fcb3153c61f02e92ce289730ed4731c8ec6738
3
+ size 2097505391
covariance_pixel/300001/300001.tmp ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:780cf35ceea56b716d94b604791154e18e8c8bb2cc0e1639bdb4f4457d6bacbf
3
+ size 2097505391
covariance_pixel/350001/350001.tmp ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:06b4b8e3bdf899f256659f4f8d61fcad5fc8f16d968e0e487cc895d91973eadb
3
+ size 2097505391
covariance_pixel/400001/400001.tmp ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:83fa8b78288d6e09d8781979197f11232ca3aecd3ca52d010886182246cbfaec
3
+ size 2097505391
covariance_pixel/450001/450001.tmp ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f4f13926c9fba108a5b26107375bfa8eb991242253ca2079e4301f965327cc30
3
+ size 2097505391
covariance_pixel/600001/600001.tmp ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:889c33f7e822baff921470016132b219a2dc5920138407b7366a7ef5b7d80303
3
+ size 2097505391
covariance_pixel/650001/650001.tmp ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:695a7054981bf5a26d940cc59f7cfada967f1fc09727d711c4ab14b12972fded
3
+ size 2097505391
covariance_pixel/700001/700001.tmp ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2227c6c641c5b6d9b5832d37d9456b99e54939eca999f351af6941aecce036f0
3
+ size 2097505391
covariance_pixel/750001/750001.tmp ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6a12cf6a8c8a6c935da8d5ead49d8e30e2d150c6ec95970124a978962090ab8d
3
+ size 2097505391
covariance_pixel/800001/800001.tmp ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dc3839ccea3d040029295c4cc61069c9158f1ba0a355b1cd789da56f7a25f12e
3
+ size 2097505391
covariance_pixel/810001/810001.tmp ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0d07f5bacaa189c533bee1689dda52b19a8518a36a9455900aaf484755a54558
3
+ size 2097505391
covariance_pixel/log.txt ADDED
@@ -0,0 +1,516 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ Calc FID for CFG 1.0 and denoise_timesteps 128
2
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
3
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
4
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
5
+ DiT: Conditioning of shape (512, 768) dtype float32
6
+ z_vectors shape (128, 32, 32, 4)
7
+ Decoder incoming shape (128, 32, 32, 4)
8
+ Decoder input (128, 32, 32, 512)
9
+ Mid Block Decoder layer (128, 32, 32, 512)
10
+ Mid Block Decoder layer (128, 32, 32, 512)
11
+ Decoder layer (128, 64, 64, 512)
12
+ Decoder layer (128, 128, 128, 512)
13
+ Decoder layer (128, 256, 256, 256)
14
+ Decoder layer (128, 256, 256, 128)
15
+ FID is 30.309669494628906
16
+ (512, 256, 256, 3)
17
+ Calc FID for CFG 1.0 and denoise_timesteps 64
18
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
19
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
20
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
21
+ DiT: Conditioning of shape (512, 768) dtype float32
22
+ FID is 31.03696060180664
23
+ (512, 256, 256, 3)
24
+ Calc FID for CFG 1.0 and denoise_timesteps 32
25
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
26
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
27
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
28
+ DiT: Conditioning of shape (512, 768) dtype float32
29
+ FID is 32.76511764526367
30
+ (512, 256, 256, 3)
31
+ Calc FID for CFG 1.0 and denoise_timesteps 16
32
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
33
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
34
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
35
+ DiT: Conditioning of shape (512, 768) dtype float32
36
+ FID is 37.20755386352539
37
+ (512, 256, 256, 3)
38
+ Calc FID for CFG 1.0 and denoise_timesteps 8
39
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
40
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
41
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
42
+ DiT: Conditioning of shape (512, 768) dtype float32
43
+ FID is 49.54936218261719
44
+ (512, 256, 256, 3)
45
+ Calc FID for CFG 1.0 and denoise_timesteps 4
46
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
47
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
48
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
49
+ DiT: Conditioning of shape (512, 768) dtype float32
50
+ FID is 93.10977172851562
51
+ (512, 256, 256, 3)
52
+ Calc FID for CFG 1.0 and denoise_timesteps 2
53
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
54
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
55
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
56
+ DiT: Conditioning of shape (512, 768) dtype float32
57
+ FID is 291.5469970703125
58
+ (512, 256, 256, 3)
59
+ Calc FID for CFG 1.0 and denoise_timesteps 1
60
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
61
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
62
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
63
+ DiT: Conditioning of shape (512, 768) dtype float32
64
+ FID is 300.5982971191406
65
+ (512, 256, 256, 3)
66
+ Calc FID for CFG 1.25 and denoise_timesteps 128
67
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
68
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
69
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
70
+ DiT: Conditioning of shape (512, 768) dtype float32
71
+ FID is 16.994489669799805
72
+ (512, 256, 256, 3)
73
+ Calc FID for CFG 1.25 and denoise_timesteps 64
74
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
75
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
76
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
77
+ DiT: Conditioning of shape (512, 768) dtype float32
78
+ FID is 17.397092819213867
79
+ (512, 256, 256, 3)
80
+ Calc FID for CFG 1.25 and denoise_timesteps 32
81
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
82
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
83
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
84
+ DiT: Conditioning of shape (512, 768) dtype float32
85
+ FID is 18.59233856201172
86
+ (512, 256, 256, 3)
87
+ Calc FID for CFG 1.25 and denoise_timesteps 16
88
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
89
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
90
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
91
+ DiT: Conditioning of shape (512, 768) dtype float32
92
+ FID is 21.81684112548828
93
+ (512, 256, 256, 3)
94
+ Calc FID for CFG 1.25 and denoise_timesteps 8
95
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
96
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
97
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
98
+ DiT: Conditioning of shape (512, 768) dtype float32
99
+ FID is 31.326385498046875
100
+ (512, 256, 256, 3)
101
+ Calc FID for CFG 1.25 and denoise_timesteps 4
102
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
103
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
104
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
105
+ DiT: Conditioning of shape (512, 768) dtype float32
106
+ FID is 68.97760009765625
107
+ (512, 256, 256, 3)
108
+ Calc FID for CFG 1.25 and denoise_timesteps 2
109
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
110
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
111
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
112
+ DiT: Conditioning of shape (512, 768) dtype float32
113
+ FID is 268.62054443359375
114
+ (512, 256, 256, 3)
115
+ Calc FID for CFG 1.25 and denoise_timesteps 1
116
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
117
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
118
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
119
+ DiT: Conditioning of shape (512, 768) dtype float32
120
+ FID is 287.94525146484375
121
+ (512, 256, 256, 3)
122
+ Calc FID for CFG 1.5 and denoise_timesteps 128
123
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
124
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
125
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
126
+ DiT: Conditioning of shape (512, 768) dtype float32
127
+ FID is 10.23225212097168
128
+ (512, 256, 256, 3)
129
+ Calc FID for CFG 1.5 and denoise_timesteps 64
130
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
131
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
132
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
133
+ DiT: Conditioning of shape (512, 768) dtype float32
134
+ FID is 10.469178199768066
135
+ (512, 256, 256, 3)
136
+ Calc FID for CFG 1.5 and denoise_timesteps 32
137
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
138
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
139
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
140
+ DiT: Conditioning of shape (512, 768) dtype float32
141
+ FID is 11.209946632385254
142
+ (512, 256, 256, 3)
143
+ Calc FID for CFG 1.5 and denoise_timesteps 16
144
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
145
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
146
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
147
+ DiT: Conditioning of shape (512, 768) dtype float32
148
+ FID is 13.52448844909668
149
+ (512, 256, 256, 3)
150
+ Calc FID for CFG 1.5 and denoise_timesteps 8
151
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
152
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
153
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
154
+ DiT: Conditioning of shape (512, 768) dtype float32
155
+ FID is 20.530975341796875
156
+ (512, 256, 256, 3)
157
+ Calc FID for CFG 1.5 and denoise_timesteps 4
158
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
159
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
160
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
161
+ DiT: Conditioning of shape (512, 768) dtype float32
162
+ FID is 51.010704040527344
163
+ (512, 256, 256, 3)
164
+ Calc FID for CFG 1.5 and denoise_timesteps 2
165
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
166
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
167
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
168
+ DiT: Conditioning of shape (512, 768) dtype float32
169
+ FID is 246.5587158203125
170
+ (512, 256, 256, 3)
171
+ Calc FID for CFG 1.5 and denoise_timesteps 1
172
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
173
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
174
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
175
+ DiT: Conditioning of shape (512, 768) dtype float32
176
+ FID is 280.76104736328125
177
+ (512, 256, 256, 3)
178
+ Calc FID for CFG 1.75 and denoise_timesteps 128
179
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
180
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
181
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
182
+ DiT: Conditioning of shape (512, 768) dtype float32
183
+ FID is 7.648271083831787
184
+ (512, 256, 256, 3)
185
+ Calc FID for CFG 1.75 and denoise_timesteps 64
186
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
187
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
188
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
189
+ DiT: Conditioning of shape (512, 768) dtype float32
190
+ FID is 7.717658042907715
191
+ (512, 256, 256, 3)
192
+ Calc FID for CFG 1.75 and denoise_timesteps 32
193
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
194
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
195
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
196
+ DiT: Conditioning of shape (512, 768) dtype float32
197
+ FID is 8.119889259338379
198
+ (512, 256, 256, 3)
199
+ Calc FID for CFG 1.75 and denoise_timesteps 16
200
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
201
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
202
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
203
+ DiT: Conditioning of shape (512, 768) dtype float32
204
+ FID is 9.593037605285645
205
+ (512, 256, 256, 3)
206
+ Calc FID for CFG 1.75 and denoise_timesteps 8
207
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
208
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
209
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
210
+ DiT: Conditioning of shape (512, 768) dtype float32
211
+ FID is 14.669331550598145
212
+ (512, 256, 256, 3)
213
+ Calc FID for CFG 1.75 and denoise_timesteps 4
214
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
215
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
216
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
217
+ DiT: Conditioning of shape (512, 768) dtype float32
218
+ FID is 38.7032356262207
219
+ (512, 256, 256, 3)
220
+ Calc FID for CFG 1.75 and denoise_timesteps 2
221
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
222
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
223
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
224
+ DiT: Conditioning of shape (512, 768) dtype float32
225
+ FID is 226.81658935546875
226
+ (512, 256, 256, 3)
227
+ Calc FID for CFG 1.75 and denoise_timesteps 1
228
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
229
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
230
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
231
+ DiT: Conditioning of shape (512, 768) dtype float32
232
+ FID is 277.3670654296875
233
+ (512, 256, 256, 3)
234
+ Calc FID for CFG 2.0 and denoise_timesteps 128
235
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
236
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
237
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
238
+ DiT: Conditioning of shape (512, 768) dtype float32
239
+ FID is 7.227836608886719
240
+ (512, 256, 256, 3)
241
+ Calc FID for CFG 2.0 and denoise_timesteps 64
242
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
243
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
244
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
245
+ DiT: Conditioning of shape (512, 768) dtype float32
246
+ FID is 7.215945720672607
247
+ (512, 256, 256, 3)
248
+ Calc FID for CFG 2.0 and denoise_timesteps 32
249
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
250
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
251
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
252
+ DiT: Conditioning of shape (512, 768) dtype float32
253
+ FID is 7.3633646965026855
254
+ (512, 256, 256, 3)
255
+ Calc FID for CFG 2.0 and denoise_timesteps 16
256
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
257
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
258
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
259
+ DiT: Conditioning of shape (512, 768) dtype float32
260
+ FID is 8.247041702270508
261
+ (512, 256, 256, 3)
262
+ Calc FID for CFG 2.0 and denoise_timesteps 8
263
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
264
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
265
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
266
+ DiT: Conditioning of shape (512, 768) dtype float32
267
+ FID is 11.816750526428223
268
+ (512, 256, 256, 3)
269
+ Calc FID for CFG 2.0 and denoise_timesteps 4
270
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
271
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
272
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
273
+ DiT: Conditioning of shape (512, 768) dtype float32
274
+ FID is 30.799434661865234
275
+ (512, 256, 256, 3)
276
+ Calc FID for CFG 2.0 and denoise_timesteps 2
277
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
278
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
279
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
280
+ DiT: Conditioning of shape (512, 768) dtype float32
281
+ FID is 210.01071166992188
282
+ (512, 256, 256, 3)
283
+ Calc FID for CFG 2.0 and denoise_timesteps 1
284
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
285
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
286
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
287
+ DiT: Conditioning of shape (512, 768) dtype float32
288
+ FID is 276.3877258300781
289
+ (512, 256, 256, 3)
290
+ Calc FID for CFG 2.25 and denoise_timesteps 128
291
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
292
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
293
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
294
+ DiT: Conditioning of shape (512, 768) dtype float32
295
+ FID is 7.888427734375
296
+ (512, 256, 256, 3)
297
+ Calc FID for CFG 2.25 and denoise_timesteps 64
298
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
299
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
300
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
301
+ DiT: Conditioning of shape (512, 768) dtype float32
302
+ FID is 7.806488990783691
303
+ (512, 256, 256, 3)
304
+ Calc FID for CFG 2.25 and denoise_timesteps 32
305
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
306
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
307
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
308
+ DiT: Conditioning of shape (512, 768) dtype float32
309
+ FID is 7.790701389312744
310
+ (512, 256, 256, 3)
311
+ Calc FID for CFG 2.25 and denoise_timesteps 16
312
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
313
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
314
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
315
+ DiT: Conditioning of shape (512, 768) dtype float32
316
+ FID is 8.26872444152832
317
+ (512, 256, 256, 3)
318
+ Calc FID for CFG 2.25 and denoise_timesteps 8
319
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
320
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
321
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
322
+ DiT: Conditioning of shape (512, 768) dtype float32
323
+ FID is 10.704044342041016
324
+ (512, 256, 256, 3)
325
+ Calc FID for CFG 2.25 and denoise_timesteps 4
326
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
327
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
328
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
329
+ DiT: Conditioning of shape (512, 768) dtype float32
330
+ FID is 25.77396583557129
331
+ (512, 256, 256, 3)
332
+ Calc FID for CFG 2.25 and denoise_timesteps 2
333
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
334
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
335
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
336
+ DiT: Conditioning of shape (512, 768) dtype float32
337
+ FID is 196.39703369140625
338
+ (512, 256, 256, 3)
339
+ Calc FID for CFG 2.25 and denoise_timesteps 1
340
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
341
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
342
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
343
+ DiT: Conditioning of shape (512, 768) dtype float32
344
+ FID is 276.3304138183594
345
+ (512, 256, 256, 3)
346
+ Calc FID for CFG 2.5 and denoise_timesteps 128
347
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
348
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
349
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
350
+ DiT: Conditioning of shape (512, 768) dtype float32
351
+ FID is 8.977917671203613
352
+ (512, 256, 256, 3)
353
+ Calc FID for CFG 2.5 and denoise_timesteps 64
354
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
355
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
356
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
357
+ DiT: Conditioning of shape (512, 768) dtype float32
358
+ FID is 8.903911590576172
359
+ (512, 256, 256, 3)
360
+ Calc FID for CFG 2.5 and denoise_timesteps 32
361
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
362
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
363
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
364
+ DiT: Conditioning of shape (512, 768) dtype float32
365
+ FID is 8.812511444091797
366
+ (512, 256, 256, 3)
367
+ Calc FID for CFG 2.5 and denoise_timesteps 16
368
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
369
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
370
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
371
+ DiT: Conditioning of shape (512, 768) dtype float32
372
+ FID is 8.980517387390137
373
+ (512, 256, 256, 3)
374
+ Calc FID for CFG 2.5 and denoise_timesteps 8
375
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
376
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
377
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
378
+ DiT: Conditioning of shape (512, 768) dtype float32
379
+ FID is 10.5935697555542
380
+ (512, 256, 256, 3)
381
+ Calc FID for CFG 2.5 and denoise_timesteps 4
382
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
383
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
384
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
385
+ DiT: Conditioning of shape (512, 768) dtype float32
386
+ FID is 22.700056076049805
387
+ (512, 256, 256, 3)
388
+ Calc FID for CFG 2.5 and denoise_timesteps 2
389
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
390
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
391
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
392
+ DiT: Conditioning of shape (512, 768) dtype float32
393
+ FID is 185.91275024414062
394
+ (512, 256, 256, 3)
395
+ Calc FID for CFG 2.5 and denoise_timesteps 1
396
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
397
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
398
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
399
+ DiT: Conditioning of shape (512, 768) dtype float32
400
+ FID is 275.87420654296875
401
+ (512, 256, 256, 3)
402
+ Calc FID for CFG 2.75 and denoise_timesteps 128
403
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
404
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
405
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
406
+ DiT: Conditioning of shape (512, 768) dtype float32
407
+ FID is 10.315759658813477
408
+ (512, 256, 256, 3)
409
+ Calc FID for CFG 2.75 and denoise_timesteps 64
410
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
411
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
412
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
413
+ DiT: Conditioning of shape (512, 768) dtype float32
414
+ FID is 10.183219909667969
415
+ (512, 256, 256, 3)
416
+ Calc FID for CFG 2.75 and denoise_timesteps 32
417
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
418
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
419
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
420
+ DiT: Conditioning of shape (512, 768) dtype float32
421
+ FID is 10.025999069213867
422
+ (512, 256, 256, 3)
423
+ Calc FID for CFG 2.75 and denoise_timesteps 16
424
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
425
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
426
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
427
+ DiT: Conditioning of shape (512, 768) dtype float32
428
+ FID is 10.00025463104248
429
+ (512, 256, 256, 3)
430
+ Calc FID for CFG 2.75 and denoise_timesteps 8
431
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
432
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
433
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
434
+ DiT: Conditioning of shape (512, 768) dtype float32
435
+ FID is 10.976386070251465
436
+ (512, 256, 256, 3)
437
+ Calc FID for CFG 2.75 and denoise_timesteps 4
438
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
439
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
440
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
441
+ DiT: Conditioning of shape (512, 768) dtype float32
442
+ FID is 20.871625900268555
443
+ (512, 256, 256, 3)
444
+ Calc FID for CFG 2.75 and denoise_timesteps 2
445
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
446
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
447
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
448
+ DiT: Conditioning of shape (512, 768) dtype float32
449
+ FID is 178.37503051757812
450
+ (512, 256, 256, 3)
451
+ Calc FID for CFG 2.75 and denoise_timesteps 1
452
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
453
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
454
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
455
+ DiT: Conditioning of shape (512, 768) dtype float32
456
+ FID is 275.13055419921875
457
+ (512, 256, 256, 3)
458
+ Calc FID for CFG 3.0 and denoise_timesteps 128
459
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
460
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
461
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
462
+ DiT: Conditioning of shape (512, 768) dtype float32
463
+ FID is 11.711248397827148
464
+ (512, 256, 256, 3)
465
+ Calc FID for CFG 3.0 and denoise_timesteps 64
466
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
467
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
468
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
469
+ DiT: Conditioning of shape (512, 768) dtype float32
470
+ FID is 11.547887802124023
471
+ (512, 256, 256, 3)
472
+ Calc FID for CFG 3.0 and denoise_timesteps 32
473
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
474
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
475
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
476
+ DiT: Conditioning of shape (512, 768) dtype float32
477
+ FID is 11.348922729492188
478
+ (512, 256, 256, 3)
479
+ Calc FID for CFG 3.0 and denoise_timesteps 16
480
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
481
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
482
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
483
+ DiT: Conditioning of shape (512, 768) dtype float32
484
+ FID is 11.138352394104004
485
+ (512, 256, 256, 3)
486
+ Calc FID for CFG 3.0 and denoise_timesteps 8
487
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
488
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
489
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
490
+ DiT: Conditioning of shape (512, 768) dtype float32
491
+ FID is 11.587258338928223
492
+ (512, 256, 256, 3)
493
+ Calc FID for CFG 3.0 and denoise_timesteps 4
494
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
495
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
496
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
497
+ DiT: Conditioning of shape (512, 768) dtype float32
498
+ FID is 19.907915115356445
499
+ (512, 256, 256, 3)
500
+ Calc FID for CFG 3.0 and denoise_timesteps 2
501
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
502
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
503
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
504
+ DiT: Conditioning of shape (512, 768) dtype float32
505
+ FID is 173.16567993164062
506
+ (512, 256, 256, 3)
507
+ Calc FID for CFG 3.0 and denoise_timesteps 1
508
+ DiT: Input of shape (512, 32, 32, 4) dtype float32
509
+ DiT: After patch embed, shape is (512, 256, 768) dtype bfloat16
510
+ DiT: Patch Embed of shape (512, 256, 768) dtype bfloat16
511
+ DiT: Conditioning of shape (512, 768) dtype float32
512
+ FID is 273.95343017578125
513
+ wandb:
514
+ wandb: 🚀 View run shortcut_imagenet256 at: https://wandb.ai/daniel-z-kaplan/shortcut/runs/shortcut_imagenet256_20250907_205404_345353_10
515
+ wandb: Find logs at: ../../../tmp/tmppqwdkpje/wandb/run-20250907_205404-shortcut_imagenet256_20250907_205404_345353_10/logs
516
+
covariance_pixel/mean_covar_pixel.tmp ADDED
Binary file (144 Bytes). View file
 
covariance_pixel/std_covar_pixel.tmp ADDED
Binary file (144 Bytes). View file