nodes/kijai-ComfyUI-WanVideoWrapper
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
1,304 Commits 8 Branches 0 Tags
001bc0e24a3f3448de3c48eee2c75d2f6ed158fb
Commit Graph
369 Commits
Author SHA1 Message Date
kijai 001bc0e24a Merge branch 'main' of https://github.com/kijai/ComfyUI-WanVideoWrapper 2025-12-04 16:44:29 +02:00
kijai 7189922f36 Update __init__.py 2025-12-04 16:44:18 +02:00
kijai faf9c5927d Merge branch 'pr/1510' 2025-12-04 16:44:02 +02:00
Jukka Seppänen 6c5dc0ba2a Merge pull request #1546 from wzxysf/main 
Enable vae tiling with end frame
 2025-12-04 16:40:29 +02:00
kijai b06c7d2d6d Add UltraViCo -sage attention mode and refactor some attention code 
https://github.com/thu-ml/DiT-Extrapolation/
 2025-12-03 13:32:45 +02:00
kijai 014e711972 cleanup 2025-12-03 11:09:09 +02:00
kijai b9f6c9aa50 Update __init__.py 2025-12-02 00:45:22 +02:00
kijai c1fbc93521 Add ViBTScheduler to use ViBT models 
https://github.com/Yuanshi9815/ViBT/tree/main
 2025-12-02 00:26:50 +02:00
kijai c4db00609a Use torch.chunk for chunking 2025-12-01 17:39:50 +02:00
kijai 5a52d6b92f Add VAE feat_cache offloading, VAE tqdm progress bar and memory usage report 2025-12-01 15:05:20 +02:00
kijai 196d39695f Fix sageattn_varlen 2025-12-01 01:11:06 +02:00
kijai e5be3e5263 Use torch custom_ops to avoid graph breaks with torch.compile 
Hopefully finally fixes the torch.compile VRAM issues...
 2025-12-01 00:29:27 +02:00
kijai a6071c7be5 Merge branch 'main' into steadydancer 2025-11-30 17:56:50 +02:00
kijai a9cd073f29 Remove unnecessary recompile when using cfg 2025-11-30 17:52:56 +02:00
kijai 1e9e2be622 Avoid recompile here 2025-11-30 17:32:24 +02:00
kijai 99c3978da4 Reduce peak VRAM usage when not using torch.compile (and some even with it) 
Found some intermediates that weren't freed which should reduce VRAM usage overall, and modified RoPE application outside torch compile for similar gains than when using torch.compile.
 2025-11-30 17:14:53 +02:00
kijai 66d44ec8db This doesn't really do anything useful 2025-11-28 21:15:25 +02:00
kijai 394c7c13d2 Add strength controls 2025-11-28 21:11:20 +02:00
kijai e54fa5d059 Init 2025-11-28 20:32:16 +02:00
kijai fa7a967ee7 Fix stand-in 2025-11-17 11:56:02 +02:00
kijai f872460285 Fix TTM for dual sampler setups 2025-11-16 19:23:19 +02:00
kijai e3c2a1431b Squashed commit of the following: 
commit f685ee33ac
Merge: bb5707f 4e31081
Author: kijai <40791699+kijai@users.noreply.github.com>
Date:   Thu Nov 13 16:37:38 2025 +0200

    Merge branch 'main' into bindweave

commit bb5707f601
Merge: acb662b ff26836
Author: kijai <40791699+kijai@users.noreply.github.com>
Date:   Tue Nov 11 18:53:19 2025 +0200

    Merge branch 'main' into bindweave

commit acb662b5af
Merge: 907c9e1 e926f7a
Author: kijai <40791699+kijai@users.noreply.github.com>
Date:   Tue Nov 11 11:44:26 2025 +0200

    Merge branch 'main' into bindweave

commit 907c9e1cdd
Author: kijai <40791699+kijai@users.noreply.github.com>
Date:   Mon Nov 10 21:02:58 2025 +0200

    Update nodes.py

commit e4a4d22537
Author: kijai <40791699+kijai@users.noreply.github.com>
Date:   Sat Nov 8 16:04:55 2025 +0200

    Update nodes_sampler.py

commit a3b2f67337
Author: kijai <40791699+kijai@users.noreply.github.com>
Date:   Sat Nov 8 16:03:00 2025 +0200

    Pad clip vision embeds like in original code

commit 1e00c8fb28
Author: kijai <40791699+kijai@users.noreply.github.com>
Date:   Sat Nov 8 12:21:11 2025 +0200

    Update nodes.py

commit ff16dce5c0
Author: kijai <40791699+kijai@users.noreply.github.com>
Date:   Fri Nov 7 01:15:13 2025 +0200

    Update nodes_sampler.py

commit f972b31bf2
Author: kijai <40791699+kijai@users.noreply.github.com>
Date:   Fri Nov 7 01:09:22 2025 +0200

    Update nodes.py

commit 3dacd6a719
Author: kijai <40791699+kijai@users.noreply.github.com>
Date:   Fri Nov 7 00:45:06 2025 +0200

    Update nodes_sampler.py

commit 7bf99791ad
Author: kijai <40791699+kijai@users.noreply.github.com>
Date:   Fri Nov 7 00:44:13 2025 +0200

    Update nodes.py

commit 7a5587b5af
Author: kijai <40791699+kijai@users.noreply.github.com>
Date:   Fri Nov 7 00:39:10 2025 +0200

    Let the user resize for QwenVL

    Seems to need smaller resolutions

commit d6cf172846
Author: kijai <40791699+kijai@users.noreply.github.com>
Date:   Thu Nov 6 23:41:24 2025 +0200

    Update nodes.py

commit cf86f4f0a4
Author: kijai <40791699+kijai@users.noreply.github.com>
Date:   Thu Nov 6 21:01:03 2025 +0200

    Update model.py

commit b1f8309a20
Author: kijai <40791699+kijai@users.noreply.github.com>
Date:   Thu Nov 6 19:55:54 2025 +0200

    Update nodes_model_loading.py

commit 8992c6af64
Author: kijai <40791699+kijai@users.noreply.github.com>
Date:   Thu Nov 6 19:22:56 2025 +0200

    Don't include padding for scheduler

commit e4084a961b
Author: kijai <40791699+kijai@users.noreply.github.com>
Date:   Thu Nov 6 18:32:53 2025 +0200

    Update nodes.py

commit 3ec1edefbe
Author: kijai <40791699+kijai@users.noreply.github.com>
Date:   Thu Nov 6 17:35:48 2025 +0200

    init

    For testing, no idea if it works yet
 2025-11-13 17:10:57 +02:00
kijai 1d0516a2a9 Avoid graph break for LongCat 2025-11-04 10:26:37 +02:00
kijai 8ce6916d72 Fix for some cases of using comfy_chunked rope 2025-11-03 10:36:55 +02:00
kijai 0d0d28569a Fix cases where text encoder isn't used (eg. Minimax remover) 2025-11-03 10:29:10 +02:00
kijai 75109fdb79 Fix custom sigmas with euler 2025-11-03 10:12:05 +02:00
kijai 5eae7087fa Fix for S2V 2025-11-02 01:24:50 +02:00
kijai 393fe78ec2 Update model.py 2025-10-31 23:39:55 +02:00
kijai cc9bf1e4f5 Store lora diffs in buffers for GGUF as well 2025-10-30 16:44:03 +02:00
kijai d2614a9a49 Merge branch 'main' into longcat 2025-10-29 02:33:37 +02:00
kijai 9d45b9f0de Use comfy core Conv3D workaround for VAE rather than the fp32 cast 2025-10-29 02:23:49 +02:00
chengzeyiandClaude d15cf3001f Fix dtype mismatch in ref_conv forward pass 
This commit fixes a RuntimeError that occurs when using Fun-Control
reference images: "Input type (float) and bias type (c10::Half)
should be the same"

Root cause:
- Commit 1ba1a16 changed the dtype handling strategy to convert
  the main latent `x` to `base_dtype` instead of converting
  embeddings to match `x.dtype`
- This caused `fun_ref` input to be in a different dtype than
  the `ref_conv` layer's weights and bias
- Line 2324 already handles this correctly for `attn_cond` by
  converting to `self.attn_conv_in.weight.dtype`

Solution:
- Convert `fun_ref` to match `self.ref_conv.weight.dtype` before
  passing through the convolution layer
- This follows the same pattern used for `attn_cond` on line 2324

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
 2025-10-28 12:08:23 +00:00
kijai eebbcd5ee0 Update model.py 2025-10-28 02:04:50 +02:00
kijai 90908df260 Update model.py 2025-10-28 01:54:42 +02:00
kijai c80a488f70 Use fp32 norms for other models too and other fixes 2025-10-28 01:52:48 +02:00
kijai e560366600 Update model.py 2025-10-27 18:55:30 +02:00
kijai c59e52ca44 Precision adjustments 2025-10-27 00:23:32 +02:00
kijai a0bdf20817 Some cleanup and allow full block swap 2025-10-26 23:05:17 +02:00
kijai 8ad7e50f33 Fix cross attention split point 2025-10-26 22:07:19 +02:00
kijai d504c96174 Separate attention for input images like in original 2025-10-26 19:14:47 +02:00
kijai 43acf83adb Update model.py 2025-10-26 16:57:25 +02:00
kijai fb00932cad Init 
https://huggingface.co/Kijai/LongCat-Video_comfy/tree/main
 2025-10-26 16:36:20 +02:00
wzxysf 6a37c0b2d6 Enable vae tiling with end frame 2025-10-24 22:23:31 +08:00
kijai 41168b1e82 Support light VAE 
https://huggingface.co/lightx2v/Autoencoders/tree/main
 2025-10-23 12:31:55 +03:00
kijai 67fcf0ba52 Reduce needless torch.compile recompiles 2025-10-22 13:29:13 +03:00
cmeka 425035d810 Fix custom sigmas for supported schedulers 
Previously, only unipc, dpm++, and dpm++_sde schedulers preserved custom input sigmas exactly. Other schedulers such as (euler, lcm, deis,
  etc.) would transform or modify the sigmas through their set_timesteps() methods, causing inconsistent behavior.
 2025-10-21 12:46:58 -04:00
kijai 1f0861b649 MoCha: modify RoPE function to be more torch.compile friendly 2025-10-21 17:54:25 +03:00
unrealMJ 88defbfdd1 add MoCha 2025-10-21 09:40:54 +08:00
kijai 200f6943e3 Add sageattn mode that allows torch.compile 
Latest wheel from woct0rdho includes the torch.compile fix:

https://github.com/woct0rdho/SageAttention/releases

Based on my quick testing this reduces peak VRAM usage a bit when running sageattn + torch.compile
 2025-10-20 15:16:43 +03:00
kijai 9cd79d3d4a Add experimental rCM scheduler 
Based on the original code, works but doesn't feel better than dpm++_sde so far
 2025-10-19 19:34:17 +03:00