comfyanonymous
|
05676942b7
|
Add some more transformer hooks and move tomesd to comfy_extras.
Tomesd now uses q instead of x to decide which tokens to merge because
it seems to give better results.
|
1 year ago |
|
comfyanonymous
|
fa28d7334b
|
Remove useless code.
|
1 year ago |
|
comfyanonymous
|
8607c2d42d
|
Move latent scale factor from VAE to model.
|
1 year ago |
|
comfyanonymous
|
30a3861946
|
Fix bug when yaml config has no clip params.
|
1 year ago |
|
comfyanonymous
|
9e37f4c7d5
|
Fix error with ClipVision loader node.
|
1 year ago |
|
comfyanonymous
|
9f83b098c9
|
Don't merge weights when shapes don't match and print a warning.
|
1 year ago |
|
comfyanonymous
|
f87ec10a97
|
Support base SDXL and SDXL refiner models.
Large refactor of the model detection and loading code.
|
1 year ago |
|
comfyanonymous
|
9fccf4aa03
|
Add original_shape parameter to transformer patch extra_options.
|
1 year ago |
|
comfyanonymous
|
51581dbfa9
|
Fix last commits causing an issue with the text encoder lora.
|
1 year ago |
|
comfyanonymous
|
8125b51a62
|
Keep a set of model_keys for faster add_patches.
|
1 year ago |
|
comfyanonymous
|
45beebd33c
|
Add a type of model patch useful for model merging.
|
1 year ago |
|
comfyanonymous
|
036a22077c
|
Fix k_diffusion math being off by a tiny bit during txt2img.
|
1 year ago |
|
comfyanonymous
|
8883cb0f67
|
Add a way to set patches that modify the attn2 output.
Change the transformer patches function format to be more future proof.
|
1 year ago |
|
comfyanonymous
|
cd930d4e7f
|
pop clip vision keys after loading them.
|
1 year ago |
|
comfyanonymous
|
c9e4a8c9e5
|
Not needed anymore.
|
1 year ago |
|
comfyanonymous
|
fb4bf7f591
|
This is not needed anymore and causes issues with alphas_cumprod.
|
1 year ago |
|
comfyanonymous
|
45be2e92c1
|
Fix DDIM v-prediction.
|
1 year ago |
|
comfyanonymous
|
e6e50ab2dd
|
Fix an issue when alphas_comprod are half floats.
|
1 year ago |
|
comfyanonymous
|
ae43f09ef7
|
All the unet weights should now be initialized with the right dtype.
|
1 year ago |
|
comfyanonymous
|
f7edcfd927
|
Add a --gpu-only argument to keep and run everything on the GPU.
Make the CLIP model work on the GPU.
|
1 year ago |
|
comfyanonymous
|
7bf89ba923
|
Initialize more unet weights as the right dtype.
|
1 year ago |
|
comfyanonymous
|
e21d9ad445
|
Initialize transformer unet block weights in right dtype at the start.
|
1 year ago |
|
comfyanonymous
|
bb1f45d6e8
|
Properly disable weight initialization in clip models.
|
1 year ago |
|
comfyanonymous
|
21f04fe632
|
Disable default weight values in unet conv2d for faster loading.
|
1 year ago |
|
comfyanonymous
|
9d54066ebc
|
This isn't needed for inference.
|
1 year ago |
|
comfyanonymous
|
fa2cca056c
|
Don't initialize CLIPVision weights to default values.
|
1 year ago |
|
comfyanonymous
|
6b774589a5
|
Set model to fp16 before loading the state dict to lower ram bump.
|
1 year ago |
|
comfyanonymous
|
0c7cad404c
|
Don't initialize clip weights to default values.
|
1 year ago |
|
comfyanonymous
|
6971646b8b
|
Speed up model loading a bit.
Default pytorch Linear initializes the weights which is useless and slow.
|
1 year ago |
|
comfyanonymous
|
388567f20b
|
sampler_cfg_function now uses a dict for the argument.
This means arguments can be added without issues.
|
1 year ago |
|
comfyanonymous
|
ff9b22d79e
|
Turn on safe load for a few models.
|
1 year ago |
|
comfyanonymous
|
735ac4cf81
|
Remove pytorch_lightning dependency.
|
1 year ago |
|
comfyanonymous
|
2b14041d4b
|
Remove useless code.
|
1 year ago |
|
comfyanonymous
|
274dff3257
|
Remove more useless files.
|
1 year ago |
|
comfyanonymous
|
f0a2b81cd0
|
Cleanup: Remove a bunch of useless files.
|
1 year ago |
|
comfyanonymous
|
f8c5931053
|
Split the batch in VAEEncode if there's not enough memory.
|
1 year ago |
|
comfyanonymous
|
c069fc0730
|
Auto switch to tiled VAE encode if regular one runs out of memory.
|
1 year ago |
|
comfyanonymous
|
c64ca8c0b2
|
Refactor unCLIP noise augment out of samplers.py
|
1 year ago |
|
comfyanonymous
|
de142eaad5
|
Simpler base model code.
|
1 year ago |
|
comfyanonymous
|
23cf8ca7c5
|
Fix bug when embedding gets ignored because of mismatched size.
|
1 year ago |
|
comfyanonymous
|
0e425603fb
|
Small refactor.
|
1 year ago |
|
comfyanonymous
|
a3a713b6c5
|
Refactor previews into one command line argument.
Clean up a few things.
|
1 year ago |
space-nuko
|
3e17971acb
|
preview method autodetection
|
1 year ago |
|
space-nuko
|
d5a28fadaa
|
Add latent2rgb preview
|
1 year ago |
|
space-nuko
|
48f7ec750c
|
Make previews into cli option
|
1 year ago |
|
space-nuko
|
b4f434ee66
|
Preview sampled images with TAESD
|
1 year ago |
|
comfyanonymous
|
fed0a4dd29
|
Some comments to say what the vram state options mean.
|
1 year ago |
|
comfyanonymous
|
0a5fefd621
|
Cleanups and fixes for model_management.py
Hopefully fix regression on MPS and CPU.
|
1 year ago |
|
comfyanonymous
|
700491d81a
|
Implement global average pooling for controlnet.
|
1 year ago |
|
comfyanonymous
|
67892b5ac5
|
Refactor and improve model_management code related to free memory.
|
1 year ago |
