The most powerful and modular diffusion model GUI, api and backend with a graph/nodes interface. The fastest local inference engine in the world.
Flux tweak memory usage.
comfyanonymous2d75df45e6eb
diff --git a/comfy/supported_models.py b/comfy/supported_models.pyindex 94fdcc0d2..6cecb9a02 100644--- a/comfy/supported_models.py+++ b/comfy/supported_models.py@@ -640,7 +640,7 @@ class Flux(supported_models_base.BASE):unet_extra_config = {}latent_format = latent_formats.Flux- memory_usage_factor = 2.6+ memory_usage_factor = 2.8supported_inference_dtypes = [torch.bfloat16, torch.float32]