9 Commits

Author	SHA1	Message	Date
kijai	67a4e7a650	Allow larger inputs with fast model ... Not sure if useful, but at least it doesn't error out unlike the normal model did with larger inputs than 530x530	2025-02-04 18:05:07 +02:00
kijai	17d964aa37	Support CublasOps for decoding ... https://github.com/aredden/torch-cublas-hgemm on a 4090 this is ~25% faster	2025-02-04 16:28:35 +02:00
kijai	046bb97fac	Support fast model ... https://huggingface.co/tencent/Hunyuan3D-2/tree/main/hunyuan3d-dit-v2-0-fast	2025-02-03 14:39:28 +02:00
kijai	d563280607	Separate VAE (old workflows need updating)	2025-01-26 02:48:37 +02:00
kijai	a86f12236e	support sageattn for mesh gen	2025-01-25 19:56:54 +02:00
kijai	75ad5e41eb	Add torch compile for main model	2025-01-22 10:27:21 +02:00
kijai	703edf2051	Use Accelerate to load the main model faster	2025-01-22 09:35:57 +02:00
kijai	7242e7e22e	more offloading for first stage	2025-01-22 08:54:10 +02:00
Jukka Seppänen	edadfb1b13	init	2025-01-21 20:01:16 +02:00