Clybius
da2d4d2b5e
feat: initial Scaled FP8 support
2025-03-19 19:51:20 -05:00
Lodestone
a258aba7af
Merge pull request #26 from silveroxides/patch-1
...
Change name and add tooltip to loaders
2025-03-16 13:52:14 +07:00
Silver
b156ac6739
Change name and add tooltip to loaders
...
This will make it more obvious that the loader can load both .safetensors and .gguf. Many users apparently end up trying to use the loader from ComfyUI-GGUF.
2025-03-15 15:07:08 +01:00
Clybius
0431d094f9
chore: enable Chroma Style Model
2025-03-15 03:28:21 -05:00
Clybius
3521ff2cd5
feat: Chroma Style Model
...
Differences:
* Works with Chroma
* When using truncate_percent greater than 0, will take the first (100 * truncate_percent)% blocks. When truncate_percent is less than 0, take the last blocks instead.
2025-03-15 03:27:26 -05:00
blepping
a85346861e
Load pickle models to CPU first
2025-02-28 22:07:29 -07:00
lodestone-rock
63a42ffc43
add direct pth file read support
2025-02-27 13:01:07 +07:00
lodestone-rock
774df492cc
add padding removal node, also replace tab with space
2025-02-26 13:33:46 +07:00
silveroxides
6196af4276
Add None option for guidance_name and lite_patch_unet_name
2025-01-22 19:33:19 +01:00
silveroxides
3a10347167
Changed loaders to use diffusion_models directory and gave nodes proper names and changed the mappings assignment
2025-01-22 12:11:46 +01:00
blepping
f58064ba0a
Ensure guidance is set properly
2025-01-21 13:14:10 -07:00
blepping
cb046c28db
Fix Chroma guidance layer counting
2025-01-21 12:45:00 -07:00
blepping
22f69e642b
Add Chroma model support
2025-01-21 12:38:22 -07:00
Lodestone
8c4b5f8321
Merge pull request #15 from blepping/feat_gguf
...
Support applying FluxMod to GGUF models
2025-01-22 01:13:37 +07:00
lodestone-rock
907f2c22b1
add flex model and remove lite support
2025-01-21 03:53:26 +07:00
blepping
84165ad8fa
No need to replicate comfy.ldm.flux.math
...
Set Flux memory usage factor
2025-01-07 06:26:25 -07:00
blepping
dfd1ad5723
Support applying FluxMod to GGUF models
2025-01-07 00:11:16 -07:00
blepping
92725ace19
Add FluxModSamplerWrapper node to allow use with SamplerCustom, etc
...
Support Flux versions other than the original models
Refactoring to reduce code duplication
Support ComfyUI --fast option
Refactor to use ComfyUI's normal operations system
2025-01-05 07:04:40 -07:00
lodestone-rock
8f7407d4ab
add 6.4B version
2024-11-26 03:55:58 +07:00
owu-1
d28cbbe966
fix oom
2024-11-25 13:54:43 +10:00
lodestone-rock
2da24a8f7e
add v2 approximator
2024-11-24 11:11:09 +07:00
lodestone-rock
5260d2c37e
forgot the if statement
2024-11-21 13:39:00 +07:00
lodestone-rock
3e413e9f94
add skip layer feature to accelerate inference a bit
2024-11-21 11:50:46 +07:00
lodestone-rock
17105634b6
bug fix for true W8A16 inference, add custom sampler for switching fp16 and bf16 on the fly
2024-11-21 09:53:55 +07:00
lodestone-rock
031f5b37c0
add 8bit weight casting support (W8A16)
2024-11-20 02:46:27 +07:00
lodestone-rock
70acedfb10
move stuff around
2024-11-18 16:47:09 +07:00
Clybius