Deep Mehta 8e4bc0edac fix: prevent autocast crash with fp8 weights in WAN model addcmul ops ... PyTorch 2.8's autocast `prioritize` function doesn't handle FP8 ScalarTypes, causing sporadic "Unexpected floating ScalarType in at::autocast::prioritize" RuntimeError when fp8_e4m3fn_fast weights are used with the WAN 2.1 model. This wraps all torch.addcmul calls in the WAN attention blocks with autocast-disabled context when autocast is active, matching the existing pattern used in sub_quadratic_attention.py and other ComfyUI modules. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>		2026-04-09 17:31:08 -07:00
..
model_animate.py	[BlockInfo] Wan (#10845)	2025-12-15 17:59:16 -08:00
model_multitalk.py	Support Multi/InfiniteTalk (#10179)	2026-01-21 23:09:48 -05:00
model.py	fix: prevent autocast crash with fp8 weights in WAN model addcmul ops	2026-04-09 17:31:08 -07:00
vae2_2.py	WAN2.2: Fix cache VRAM leak on error (#10308)	2025-10-13 15:23:11 -04:00
vae.py	wan: vae: Fix light/color change (#13101)	2026-03-21 18:44:35 -04:00