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

3dcad78
Remove some useless code.
comfyanonymous committed 3 years ago
2b13939
Faster VAE loading.
comfyanonymous committed 3 years ago
95d796f
Initialize the unet directly on the target device.
comfyanonymous committed 3 years ago
4b957a0
Fix ROCm nightly install command.
comfyanonymous committed 3 years ago
ad5866b
added slider and toggle widget
Guillaume Faguet committed 3 years ago
d3d9ad0
Remove unused code and torchdiffeq dependency.
comfyanonymous committed 3 years ago
c910b4a
Remove some prints.
comfyanonymous committed 3 years ago
68be24e
Fix timestep ranges when batch_size > 1
asagi4 committed 3 years ago
1ea4d84
Update colab notebook with SDXL links.
comfyanonymous committed 3 years ago
4ab75d9
Fix diffusers VAE loading.
comfyanonymous committed 3 years ago
5379051
00da9b3
Fix some new loras.
comfyanonymous committed 3 years ago
727588d
Update nightly ROCm pytorch command in readme to 5.6
comfyanonymous committed 3 years ago
315ba30
Fix potential issue with Save Checkpoint.
comfyanonymous committed 3 years ago
4f9b6f3
Disable cuda malloc on a bunch of quadro cards.
comfyanonymous committed 3 years ago
7c0a5a3
a51f33e
5f75d78
Add a ControlNetApplyAdvanced node.
comfyanonymous committed 3 years ago
d191c4f
Disable cuda malloc on all the 9xx series.
comfyanonymous committed 3 years ago
30de083
Try to fix memory issue with lora.
comfyanonymous committed 3 years ago
22f29d6
Nodes can now patch the unet function.
comfyanonymous committed 3 years ago
67be7eb
Del the right object when applying lora.
comfyanonymous committed 3 years ago
12a6e93
Disable cuda malloc on regular GTX 960.
comfyanonymous committed 3 years ago
85a8900