The most powerful and modular diffusion model GUI, api and backend with a graph/nodes interface. The fastest local inference engine in the world.
Add a try except block so if torch version is weird it won't crash.
comfyanonymous15564688edb9
diff --git a/comfy/model_management.py b/comfy/model_management.pyindex 244fa5c73..33891b929 100644--- a/comfy/model_management.py+++ b/comfy/model_management.py@@ -224,8 +224,11 @@ if ENABLE_PYTORCH_ATTENTION:torch.backends.cuda.enable_flash_sdp(True)torch.backends.cuda.enable_mem_efficient_sdp(True)-if int(torch_version[0]) == 2 and int(torch_version[2]) >= 5:- torch.backends.cuda.allow_fp16_bf16_reduction_math_sdp(True)+try:+ if int(torch_version[0]) == 2 and int(torch_version[2]) >= 5:+ torch.backends.cuda.allow_fp16_bf16_reduction_math_sdp(True)+except:+ logging.warning("Warning, could not set allow_fp16_bf16_reduction_math_sdp")if args.lowvram:set_vram_to = VRAMState.LOW_VRAM