Skip to content

The most powerful and modular diffusion model GUI, api and backend with a graph/nodes interface. The fastest local inference engine in the world.

Commits on master

Support controlnet in diffusers format.
comfyanonymous committed 3 years ago
78e7958
58b2364
fix: ⚡️ small type fix
melMass committed 3 years ago
5190aa2
0115018
Disable cuda malloc on GTX 750 Ti.
comfyanonymous committed 3 years ago
39c58b2
d5c0765
Auto disable cuda malloc on some GPUs on windows.
comfyanonymous committed 3 years ago
799c08a
Fix typo.
comfyanonymous committed 3 years ago
0b284f6
Fix ddim issue with older torch versions.
comfyanonymous committed 3 years ago
e032ca6
Add MX450 and MX550 to list of cards with broken fp16.
comfyanonymous committed 3 years ago
18885f8
It's actually possible to torch.compile the unet now.
comfyanonymous committed 3 years ago
9ba4409
51d5477
Fix device print on old torch version.
comfyanonymous committed 3 years ago
ff6b047
Enable --cuda-malloc by default on torch 2.0 and up.
comfyanonymous committed 3 years ago
9871a15
--windows-standalone-build now enables --cuda-malloc
comfyanonymous committed 3 years ago
55d0fca
1679abd
ee8f8ee
Lower lora ram usage when in normal vram mode.
comfyanonymous committed 3 years ago
5f57362
Speed up lora loading a bit.
comfyanonymous committed 3 years ago
490771b
Fix CLIPSetLastLayer not reverting when removed.
comfyanonymous committed 3 years ago
50b1180
Reduce floating point rounding errors in loras.
comfyanonymous committed 3 years ago
6fb084f
Add a node to merge CLIP models.
comfyanonymous committed 3 years ago
91ed281
Refactor to make it easier to set the api path.
comfyanonymous committed 3 years ago
907c9fb