37 Commits

Author	SHA1	Message	Date
kijai	a7646c0d6f	refactor ... - unify all pipelines into one - unify transformer model into one - separate VAE - add single file model loading	2024-11-19 03:04:22 +02:00
kijai	15aa68c95d	update	2024-11-17 00:48:01 +02:00
kijai	0bd3da569e	code cleanup ... codebase getting too bloated: drop PAB support in favor of FasterCache drop temporal tilling in favor of FreeNoise	2024-11-14 19:54:52 +02:00
kijai	34b650c785	fun fixes	2024-11-12 19:29:55 +02:00
kijai	f7f999adbc	Update pipeline_cogvideox_inpaint.py	2024-11-07 14:19:25 +02:00
kijai	666f7832f9	latent preview tests	2024-11-07 11:26:19 +02:00
kijai	0e8f8140e4	clear FasterCache after done	2024-10-31 02:03:59 +02:00
kijai	d314ddbe05	check that Tora trajectory length matches	2024-10-30 23:11:49 +02:00
kijai	5ba9b1d634	code cleanup	2024-10-30 21:30:03 +02:00
kijai	dccc8bdcb7	Tora for fun context scheduling	2024-10-29 19:51:36 +02:00
kijai	3f97f07275	Deprecate CogVideoXFunVid2VidSampler and move it's functionality to CogVideoXFunSampler ... too many nodes	2024-10-29 19:18:19 +02:00
kijai	5b4819ba65	support Tora for Fun -models	2024-10-29 10:44:09 +02:00
kijai	4c3fcd7b01	Support multiple LoRAs	2024-10-21 13:24:21 +03:00
kijai	a995247624	possibly fix LoRA scaling	2024-10-12 22:23:54 +03:00
kijai	e047e6f07f	controlnet with context windowing	2024-10-08 17:56:30 +03:00
Jukka Seppänen	49451c4a22	support sageattn ... https://github.com/thu-ml/SageAttention	2024-10-06 20:26:03 +03:00
kijai	5e2e11dd0a	Update pipeline_cogvideox_inpaint.py	2024-10-06 12:58:07 +03:00
kijai	d8a9f31de9	lora loading	2024-10-06 02:48:39 +03:00
kijai	5fde34468e	update workflows	2024-10-06 02:22:35 +03:00
kijai	00d38f9a22	context schedule and all samplers for the non-Fun sampler as well	2024-10-06 01:28:20 +03:00
kijai	2be8f694b0	use proper length RoPE embeds with context windowing	2024-10-05 22:04:51 +03:00
kijai	1801c65e97	FreeNoise noise shuffling for context windows	2024-10-05 19:29:09 +03:00
kijai	453ee1606e	Add context schedules for the control pipeline	2024-10-05 16:15:34 +03:00
kijai	7af6666c67	latent input for the control sampler, allows vid2vid along with pose input	2024-10-04 13:14:54 +03:00
kijai	3c8183ac65	Support lora loading with fp8, noise augment for control input	2024-10-04 12:48:32 +03:00
kijai	3efe90ba35	Support CogVideoX-Fun lora loading	2024-10-04 01:53:40 +03:00
kijai	4fb602cad5	remove prints	2024-10-02 20:27:38 +03:00
kijai	f9f06d595e	temporal tiling for the control pipe	2024-10-02 20:14:45 +03:00
kijai	627af9341c	separate control image encoder and add tiled encode support for it ... tiled encoding code thanks to MinusZoneML: https://github.com/MinusZoneAI/ComfyUI-CogVideoX-MZ/blob/main/mz_enable_vae_encode_tiling.py	2024-10-02 18:08:11 +03:00
Jukka Seppänen	f3a1ff933e	Add control strenght, start and end percent adjustments	2024-09-30 17:50:48 +03:00
kijai	55325f97a7	initial CogVideoX-Fun 1.1 support ... every possible model/workflow combination not tested yet	2024-09-30 13:57:37 +03:00
kijai	299cea851f	PAB for Fun	2024-09-23 01:11:58 +03:00
kijai	21675b296b	Add 5b I2V GGUF ... https://huggingface.co/Kijai/CogVideoX_GGUF/blob/main/CogVideoX_5b_I2V_GGUF_Q4_0.safetensors	2024-09-22 16:07:32 +03:00
kijai	bc31982707	fix for sequential_cpu_offload in "fun" pipeline	2024-09-21 18:43:26 +03:00
kijai	73fa4be48f	Cleanup unused dependencies	2024-09-21 16:37:54 +03:00
kijai	a125f61aad	fun text2vid	2024-09-19 01:07:18 +03:00
kijai	dd60660fee	initial CogVideoX-Fun support	2024-09-18 17:00:03 +03:00