2210 Commits

Include renames

Author	SHA1	Message	Date
rattusandGitHub	4e6a1b66a9	speed up and reduce VRAM of QWEN VAE and WAN (less so) (#12036) ... * ops: introduce autopad for conv3d This works around pytorch missing ability to causal pad as part of the kernel and avoids massive weight duplications for padding. * wan-vae: rework causal padding This currently uses F.pad which takes a full deep copy and is liable to be the VRAM peak. Instead, kick spatial padding back to the op and consolidate the temporal padding with the cat for the cache. * wan-vae: implement zero pad fast path The WAN VAE is also QWEN where it is used single-image. These convolutions are however zero padded 3d convolutions, which means the VAE is actually just 2D down the last element of the conv weight in the temporal dimension. Fast path this, to avoid adding zeros that then just evaporate in convoluton math but cost computation.	2026-01-23 19:56:14 -05:00
comfyanonymousandGitHub	9cf299a9f9	Make regular empty latent node work properly on flux 2 variants. (#12050)	2026-01-23 19:50:48 -05:00
ComfyUI WikiandGitHub	e89b22993a	Support ModelScope-Trainer/DiffSynth LoRA format for Flux.2 Klein models (#12042)	2026-01-23 15:27:49 -05:00
Jukka SeppänenandGitHub	55bd606e92	LTX2: Refactor forward function for better VRAM efficiency and fix spatial inpainting (#12046) ... * Disable timestep embed compression when inpainting Spatial inpainting not compatible with the compression * Reduce crossattn peak VRAM * LTX2: Refactor forward function for better VRAM efficiency	2026-01-23 15:26:38 -05:00
Omri MaromandGitHub	d7f3241bf6	qwen_image: propagate attention mask. (#11966)	2026-01-22 20:02:31 -05:00
comfyanonymousandGitHub	09a2e67151	Support loading flux 2 klein checkpoints saved with SaveCheckpoint. (#12033)	2026-01-22 18:20:48 -05:00
rattusandGitHub	0fd1b78736	Reduce LTX2 VAE VRAM consumption (#12028) ... * causal_video_ae: Remove attention ResNet This attention_head_dim argument does not exist on this constructor so this is dead code. Remove as generic attention mid VAE conflicts with temporal roll. * ltx-vae: consoldate causal/non-causal code paths * ltx-vae: add cache rolling adder * ltx-vae: use cached adder for resnet * ltx-vae: Implement rolling VAE Implement a temporal rolling VAE for the LTX2 VAE. Usually when doing temporal rolling VAEs you can just chunk on time relying on causality and cache behind you as you go. The LTX VAE is however non-causal. So go whole hog and implement per layer run ahead and backpressure between the decoder layers using recursive state beween the layers. Operations are ammended with temporal_cache_state{} which they can use to hold any state then need for partial execution. Convolutions cache their inputs behind the up to N-1 frames, and skip connections need to cache the mismatch between convolution input and output that happens due to missing future (non-causal) input. Each call to run_up() processes a layer accross a range on input that may or may not be complete. It goes depth first to process as much as possible to try and digest frames to the final output ASAP. If layers run out of input due to convolution losses, they simply return without action effectively applying back-pressure to the earlier layers. As the earlier layers do more work and caller deeper, the partial states are reconciled and output continues to digest depth first as much as possible. Chunking is done using a size quota rather than a fixed frame length and any layer can initiate chunking, and multiple layers can chunk at different granulatiries. This remove the old limitation of always having to process 1 latent frame to entirety and having to hold 8 full decoded frames as the VRAM peak.	2026-01-22 16:54:18 -05:00
Jukka SeppänenandGitHub	16b9aabd52	Support Multi/InfiniteTalk (#10179) ... * re-init * Update model_multitalk.py * whitespace... * Update model_multitalk.py * remove print * this is redundant * remove import * Restore preview functionality * Move block_idx to transformer_options * Remove LoopingSamplerCustomAdvanced * Remove looping functionality, keep extension functionality * Update model_multitalk.py * Handle ref_attn_mask with separate patch to avoid having to always return q and k from self_attn * Chunk attention map calculation for multiple speakers to reduce peak VRAM usage * Update model_multitalk.py * Add ModelPatch type back * Fix for latest upstream * Use DynamicCombo for cleaner node Basically just so that single_speaker mode hides mask inputs and 2nd audio input * Update nodes_wan.py	2026-01-21 23:09:48 -05:00
Jukka SeppänenandGitHub	245f6139b6	More targeted embedding_connector loading for LTX2 text encoder (#11992) ... Reduces errors	2026-01-21 23:05:06 -05:00
Jukka SeppänenandGitHub	3365ad18a5	Support LTX2 tiny vae (taeltx_2) (#11929)	2026-01-21 23:03:51 -05:00
comfyanonymousandGitHub	abe2ec26a6	Support the Anima model. (#12012)	2026-01-21 19:44:28 -05:00
MarkuryandGitHub	0fc15700be	Add LyCoris LoKr MLP layer support for Flux2 (#11997)	2026-01-20 23:18:33 -05:00
comfyanonymousandGitHub	e755268e7b	Config for Qwen 3 0.6B model. (#11998)	2026-01-20 23:08:31 -05:00
MyloandGitHub	c4a14df9a3	Dynamically detect chroma radiance patch size (#11991)	2026-01-20 18:46:11 -05:00
Ivan ZorinandGitHub	965d0ed509	fix: remove normalization of audio in LTX Mel spectrogram creation (#11990) ... For LTX Audio VAE, remove normalization of audio during MEL spectrogram creation. This aligs inference with training and prevents loud audio from being attenuated.	2026-01-20 18:44:28 -05:00
comfyanonymousandGitHub	8ccc0c94fa	Make omni stuff work on regular z image for easier testing. (#11985)	2026-01-20 00:32:00 -05:00
comfyanonymousandGitHub	2108167f9f	Support zimage omni base model. (#11979)	2026-01-19 23:17:38 -05:00
comfyanonymousandGitHub	70c91b8248	Fix #11963 (#11982)	2026-01-19 22:32:40 -05:00
rkfgandGitHub	0da5a0fe58	Convert mono audio to fake stereo for LTXV VAE encoding (#11965)	2026-01-19 22:12:02 -05:00
comfyanonymousandGitHub	e0eacb0688	Simpler way to implement the #11980 loras. (#11981)	2026-01-19 22:00:36 -05:00
comfyanonymousandGitHub	7ac999bf30	Add image sizes to clip vision outputs. (#11923)	2026-01-16 23:02:28 -05:00
comfyanonymousandGitHub	4c816d5c69	Adjust memory usage factor calculation for flux2 klein. (#11900)	2026-01-15 20:06:40 -05:00
comfyanonymousandGitHub	3b832231bb	Flux2 Klein support. (#11890)	2026-01-15 10:33:15 -05:00
Jukka SeppänenandGitHub	be518db5a7	Remove extraneous clip missing warnings when loading LTX2 embeddings_connector weights (#11874)	2026-01-14 17:54:04 -05:00
rattusandGitHub	80441eb15e	utils: fix lanczos grayscale upscaling (#11873)	2026-01-14 17:53:16 -05:00
comfyanonymousandGitHub	6165c38cb5	Optimize nvfp4 lora applying. (#11866) ... This changes results a bit but it also speeds up things a lot.	2026-01-14 00:49:38 -05:00
SilverandGitHub	712cca36a1	feat: throttle ProgressBar updates to reduce WebSocket flooding (#11504)	2026-01-13 22:41:44 -05:00
comfyanonymousandGitHub	eff2b9d412	Optimize nvfp4 lora applying. (#11856)	2026-01-13 19:37:19 -05:00
comfyanonymousandGitHub	15b312de7a	Optimize nvfp4 lora applying. (#11854)	2026-01-13 19:23:58 -05:00
comfyanonymousandGitHub	1dcbd9efaf	Bump ltxav mem estimation a bit. (#11842)	2026-01-13 01:42:07 -05:00
comfyanonymousandGitHub	117e7a5853	Refactor to try to lower mem usage. (#11840)	2026-01-12 21:01:52 -08:00
comfyanonymousandGitHub	b3c0e4de57	Make loras work on nvfp4 models. (#11837) ... The initial applying is a bit slow but will probably be sped up in the future.	2026-01-12 22:33:54 -05:00
Jukka SeppänenandGitHub	fd5c0755af	Reduce LTX2 VRAM use by more efficient timestep embed handling (#11829)	2026-01-12 17:28:59 -05:00
comfyanonymousandGitHub	c881a1d689	Support the siglip 2 naflex model as a clip vision model. (#11831) ... Not useful yet.	2026-01-12 17:05:54 -05:00
kelseyeeandGitHub	a3b5d4996a	Support ModelScope-Trainer DiffSynth lora for Z Image. (#11805)	2026-01-12 15:38:46 -05:00
comfyanonymousandGitHub	2f642d5d9b	Fix chroma fp8 te being treated as fp16. (#11795)	2026-01-10 14:40:42 -08:00
comfyanonymousandGitHub	cd912963f1	Fix issue with t5 text encoder in fp4. (#11794)	2026-01-10 17:31:31 -05:00
DELUXAandGitHub	6e4b1f9d00	pythorch_attn_by_def_on_gfx1200 (#11793)	2026-01-10 16:51:05 -05:00
comfyanonymousandGitHub	dc202a2e51	Properly save mixed ops. (#11772)	2026-01-10 02:03:57 -05:00
comfyanonymousandGitHub	bd0e6825e8	Be less strict when loading mixed ops weights. (#11769)	2026-01-09 14:21:06 -05:00
Jedrzej KosinskiandGitHub	1dc3da6314	Add most basic Asset support for models (#11315) ... * Brought over minimal elements from PR 10045 to reproduce seed_assets and register_assets_system without adding anything to the DB or server routes yet, for now making everything sync (can introduce async once everything is cleaned up and brought over) * Added db script to insert assets stuff, cleaned up some code; assets (models) now get added/rescanned * Added support for 5 http endpoints for assets * Replaced Optional with | None in schemas_in.py and schemas_out.py * Remove two routes that will not be relevant yet in this PR: HEAD /api/assets/hash/<hash> and PUT /api/assets/<id>/preview * Remove some functions the two deleted endpoints were using * Don't show assets scan message upon calling /object_info endpoint * removed unsued import to satisfy ruff * Simplified hashing function tpye hint and _hash_file_obj * Satisfied ruff	2026-01-08 22:21:51 -05:00
comfyanonymousandGitHub	1a20656448	Fix import issue. (#11746)	2026-01-08 17:23:59 -05:00
comfyanonymousandGitHub	0f11869d55	Better detection if AMD torch compiled with efficient attention. (#11745)	2026-01-08 17:16:58 -05:00
comfyanonymousandGitHub	50d6e1caf4	Tweak ltxv vae mem estimation. (#11722)	2026-01-07 23:07:05 -05:00
comfyanonymousandGitHub	21e8425087	Add warning for old pytorch. (#11718)	2026-01-07 21:07:26 -05:00
rattusandGitHub	b6c79a648a	ops: Fix offloading with FP8MM performance (#11697) ... This logic was checking comfy_cast_weights, and going straight to to the forward_comfy_cast_weights implementation without attempting to downscale input to fp8 in the event comfy_cast_weights is set. The main reason comfy_cast_weights would be set would be for async offload, which is not a good reason to nix FP8MM. So instead, and together the underlying exclusions for FP8MM which are: * having a weight_function (usually LowVramPatch) * force_cast_weights (compute dtype override) * the weight is not Quantized * the input is already quantized * the model or layer has MM explictily disabled. If you get past all of those exclusions, quantize the input tensor. Then hand the new input, quantized or not off to forward_comfy_cast_weights to handle it. If the weight is offloaded but input is quantized you will get an offloaded MM8.	2026-01-07 21:01:16 -05:00
comfyanonymousandGitHub	25bc1b5b57	Add memory estimation function to ltxav text encoder. (#11716)	2026-01-07 20:11:22 -05:00
comfyanonymousandGitHub	3cd19e99c1	Increase ltxav mem estimation by a bit. (#11715)	2026-01-07 20:04:56 -05:00
comfyanonymousandGitHub	34751fe9f9	Lower ltxv text encoder vram use. (#11713)	2026-01-07 19:12:15 -05:00
rattusandGitHub	48e5ea1dfd	model_patcher: Remove confusing load stat (#11710) ... If the loader passes 1e32 as the usable memory size, it means force the full load. This happens with CPU loads and a few other misc cases. Removing the confusing number and just leave the other details.	2026-01-07 18:39:20 -05:00