Hugging Face's logo

infosave
/
MiniMax-H3-Image-cmf

Text-to-Image
cortiq
English
Chinese
Russian
cmf
quantized
q4tp
cuda
nvidia
vulkan
metal
image-to-image
image-editing
Model card Files
xet
 Community
infosave commited on
Commit
8d8161d
·
verified ·
1 Parent(s): 08163fb

Card: 1024 x 1024 on CUDA, CUDA against Vulkan

Browse files
Files changed (1) hide show
  1. README.md +4 -4
README.md CHANGED
@@ -90,7 +90,7 @@ NVIDIA RTX 2000 Ada, 16 GB, one image, seed 42, single runs (2048 x 2048: the me
90
 
91
  | run | prompt, s | six steps, s | a step, s | decode, s | total, s | peak memory of the card, MiB |
92
  |---|---:|---:|---:|---:|---:|---:|
93
- | CUDA, 1024 x 1024 (`--min-area 1024`) | 3.4 | 13.4 | 1.92 | 0.6 | **19.0** | 12 116 |
94
  | CUDA, 2048 x 2048 | 2.6 | 44.3 | 7.38 | 1.7 | **50.3** | 12 654 |
95
  | CUDA, 4096 x 4096 | 9.2 | 264.5 | 43.90 | 8.3 | **283.7** | 13 414 |
96
  | CUDA, edit at 2048 x 2048, one 1024 x 1024 picture | 9.8 | 67.9 | 11.37 | 1.7 | **80.8** | 12 854 |
@@ -134,7 +134,7 @@ An 8-bit encoder, an 8-bit transformer and the f32 path give the same result; th
134
  originals; nothing was retrained.
135
 
136
  Checked: the CUDA chain against a per-operation f32 path (54.0 dB PSNR on a 2048 x 2048 edit), CUDA against
137
- Vulkan (44.6 dB), the 4 GB ceiling against none (49.6 dB). Not checked: the reference implementation was
138
  not run on the same card, so there is no picture-to-picture comparison with it.
139
 
140
  ## License
@@ -181,7 +181,7 @@ cortiq imagine minimax-h3-image-q4tp.cmf --prompt @prompt.txt --out out.png
181
 
182
  | прогон | промпт, с | шесть шагов, с | шаг, с | декодер, с | всего, с | пик памяти карты, МиБ |
183
  |---|---:|---:|---:|---:|---:|---:|
184
- | CUDA, 1024 x 1024 (`--min-area 1024`) | 3,4 | 13,4 | 1,92 | 0,6 | **19,0** | 12 116 |
185
  | CUDA, 2048 x 2048 | 2,6 | 44,3 | 7,38 | 1,7 | **50,3** | 12 654 |
186
  | CUDA, 4096 x 4096 | 9,2 | 264,5 | 43,90 | 8,3 | **283,7** | 13 414 |
187
  | CUDA, правка 2048 x 2048, одна картинка 1024 x 1024 | 9,8 | 67,9 | 11,37 | 1,7 | **80,8** | 12 854 |
@@ -241,7 +241,7 @@ CUDA 需要 NVIDIA 驱动和 CUDA 12.x 库（`libnvrtc.so.12`、`libcublas.so.12
241
 
242
  | 运行 | 提示词，秒 | 六步，秒 | 每步，秒 | 解码，秒 | 总计，秒 | 显存峰值，MiB |
243
  |---|---:|---:|---:|---:|---:|---:|
244
- | CUDA，1024 x 1024（`--min-area 1024`） | 3.4 | 13.4 | 1.92 | 0.6 | **19.0** | 12 116 |
245
  | CUDA，2048 x 2048 | 2.6 | 44.3 | 7.38 | 1.7 | **50.3** | 12 654 |
246
  | CUDA，4096 x 4096 | 9.2 | 264.5 | 43.90 | 8.3 | **283.7** | 13 414 |
247
  | CUDA，2048 x 2048 编辑，一张 1024 x 1024 参考图 | 9.8 | 67.9 | 11.37 | 1.7 | **80.8** | 12 854 |
 
90
 
91
  | run | prompt, s | six steps, s | a step, s | decode, s | total, s | peak memory of the card, MiB |
92
  |---|---:|---:|---:|---:|---:|---:|
93
+ | CUDA, 1024 x 1024 (`--min-area 1024`) | 2.4 | 11.7 | 1.95 | 0.5 | **16.1** | 12 116 |
94
  | CUDA, 2048 x 2048 | 2.6 | 44.3 | 7.38 | 1.7 | **50.3** | 12 654 |
95
  | CUDA, 4096 x 4096 | 9.2 | 264.5 | 43.90 | 8.3 | **283.7** | 13 414 |
96
  | CUDA, edit at 2048 x 2048, one 1024 x 1024 picture | 9.8 | 67.9 | 11.37 | 1.7 | **80.8** | 12 854 |
 
134
  originals; nothing was retrained.
135
 
136
  Checked: the CUDA chain against a per-operation f32 path (54.0 dB PSNR on a 2048 x 2048 edit), CUDA against
137
+ Vulkan (47.0 dB), the 4 GB ceiling against none (49.6 dB). Not checked: the reference implementation was
138
  not run on the same card, so there is no picture-to-picture comparison with it.
139
 
140
  ## License
 
181
 
182
  | прогон | промпт, с | шесть шагов, с | шаг, с | декодер, с | всего, с | пик памяти карты, МиБ |
183
  |---|---:|---:|---:|---:|---:|---:|
184
+ | CUDA, 1024 x 1024 (`--min-area 1024`) | 2,4 | 11,7 | 1,95 | 0,5 | **16,1** | 12 116 |
185
  | CUDA, 2048 x 2048 | 2,6 | 44,3 | 7,38 | 1,7 | **50,3** | 12 654 |
186
  | CUDA, 4096 x 4096 | 9,2 | 264,5 | 43,90 | 8,3 | **283,7** | 13 414 |
187
  | CUDA, правка 2048 x 2048, одна картинка 1024 x 1024 | 9,8 | 67,9 | 11,37 | 1,7 | **80,8** | 12 854 |
 
241
 
242
  | 运行 | 提示词，秒 | 六步，秒 | 每步，秒 | 解码，秒 | 总计，秒 | 显存峰值，MiB |
243
  |---|---:|---:|---:|---:|---:|---:|
244
+ | CUDA，1024 x 1024（`--min-area 1024`） | 2.4 | 11.7 | 1.95 | 0.5 | **16.1** | 12 116 |
245
  | CUDA，2048 x 2048 | 2.6 | 44.3 | 7.38 | 1.7 | **50.3** | 12 654 |
246
  | CUDA，4096 x 4096 | 9.2 | 264.5 | 43.90 | 8.3 | **283.7** | 13 414 |
247
  | CUDA，2048 x 2048 编辑，一张 1024 x 1024 参考图 | 9.8 | 67.9 | 11.37 | 1.7 | **80.8** | 12 854 |