Adrien Toupet
a900af96fd
BlockSwap now uses minimal, pressure-based cache clearing instead of forced deep cleans: - Clear only when VRAM < 5% free (was 15%, more aggressive threshold) - Use minimal GPU cache clear (deep=False) for reduced overhead - Applied after each block/IO swap when pressure detected Also fixes peft dependency: Add explicit peft>=0.17.0 in requirements for diffusers>=0.33.1 compatibility in ComfyUI ecosystem where other plugins may install outdated peft versions.
14 lines
162 B
Plaintext
14 lines
162 B
Plaintext
torch
|
|
torchvision
|
|
safetensors
|
|
numpy
|
|
tqdm
|
|
psutil
|
|
einops
|
|
omegaconf>=2.3.0
|
|
diffusers>=0.33.1
|
|
peft>=0.17.0
|
|
rotary_embedding_torch>=0.5.3
|
|
opencv-python
|
|
gguf
|
|
matplotlib |