John Pollock and Copilot
b6ab751f4c
Update example_workflows/wan2_2 t2i lightx2v lora distorch2.json
...
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com >
2025-10-13 03:13:44 -05:00
John Pollock and Copilot
eeed32ba1b
Update example_workflows/qwen_image_edit_2509 unet clip distorch2.json
...
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com >
2025-10-13 03:13:23 -05:00
John Pollock
ef07c0f41e
Documentation clean-up
2025-10-13 02:41:07 -05:00
John Pollock
5310d4a28d
Update documentation and example workflows to match Comfy guidelines
2025-10-13 02:16:52 -05:00
John Pollock
fd60ae6f65
Add MultiGPU support for various model loaders and encoders
2025-10-13 02:10:41 -05:00
John Pollock
01b41f4514
Update workflow examples to match Comfy Core guidelines
2025-10-13 01:52:20 -05:00
John Pollock
161174266c
Add UNet offload device support, and enhance Florence2 model loading with safetensors conversion option.
2025-10-12 04:35:43 -05:00
John Pollock
0d6dfc7524
rename directory to comfy expected
2025-10-11 07:42:09 -05:00
John Pollock
d6137f39d0
Merge pull request #122 from pollockjj/wvw
...
Pull Request: Refresh WanVideoWrapper MultiGPU support to latest upstream
2025-10-10 15:21:11 -05:00
John Pollock
c77a4b2063
Add support for secure connections in ComfyWorkflowRunner
2025-10-10 15:20:07 -05:00
John Pollock
830342453c
Update from Pull Request #122 review
2025-10-10 15:11:25 -05:00
John Pollock
e4c2006c44
Merge commit 'da6b605' into wvw
2025-10-10 14:31:18 -05:00
John Pollock
e31ba04f20
WanVideoWrapper updates to reflect reported changes ( #118 , #90 ) in the ComfyUI-WanVideoWrapper code base. Breaking changes. See example workflows.
2025-10-10 14:28:11 -05:00
John Pollock
55096eef4b
Pre-release cleanup
2025-10-10 14:24:21 -05:00
John Pollock
40d40ef9b6
Finished implementation of wanvideowrapper nodes, updated example workflows, added some benchmarking files for reference.
2025-10-10 13:40:37 -05:00
John Pollock
db931d66c3
Add WanVideoControlnetLoader class for loading ControlNet models
2025-10-10 09:30:46 -05:00
John Pollock
faef324975
correcting implementation of salvagable nodes and obiliterating one abomination
2025-10-10 09:28:58 -05:00
John Pollock
610c2102dd
Add multi-GPU support for remaining planned WanVideo model loaders
2025-10-10 04:35:16 -05:00
John Pollock
375cbd093c
Add LoadWanVideoClipTextEncoder and WanVideoClipVisionEncode for multi-GPU support
2025-10-09 19:05:28 -05:00
John Pollock
2f605e0db3
Add WanVideoVACEEncode and enhance WanVideoEncode for multi-GPU support
...
- Implemented WanVideoVACEEncode class for encoding with VACE, including input parameters for width, height, number of frames, and strength.
- Enhanced WanVideoEncode class to support multi-GPU encoding, with additional parameters for noise augmentation and latent strength.
- Updated device management to ensure proper device context during encoding processes.
2025-10-09 14:41:53 -05:00
John Pollock
ba0a3a6e21
improve documentation on supporting ROCm cuda-enumerated GPUs.
2025-10-09 13:01:20 -05:00
John Pollock
b04ae1e378
Coverted WanVideoImageToVideoEncode to shim
2025-10-09 12:32:29 -05:00
John Pollock
f0313e49aa
WanVideoDecode cleanup
2025-10-09 08:12:26 -05:00
John Pollock
15ff29c63b
methodological convergence and cleanup
2025-10-09 03:06:25 -05:00
John Pollock
7c87983d72
WIP
2025-10-08 23:06:54 -05:00
John Pollock
04013c3ee5
feat: add WanVideoTextEncodeCached and WanVideoTextEncodeSingle classes for enhanced text encoding functionality
2025-10-08 06:08:16 -05:00
John Pollock
b0930e284a
Merge pull request #120 from xmodo99/patch-1
...
fix dead link
2025-10-08 02:30:03 -05:00
xmodo99
da6b605de6
fix dead link
2025-10-07 18:57:44 -07:00
John Pollock
d300c12e9f
feat: add WanVideoBlockSwap class and integrate into MultiGPU node mappings
2025-10-07 14:22:01 -05:00
John Pollock
27cfb83733
feat: add MultiGPU logging enhancements and new CI scripts for workflow execution and log summarization, WanVidwoWrapper Model Loader/Sampler re-implemented
2025-10-07 13:19:09 -05:00
John Pollock
b034901a05
Housecleaning
2025-10-07 04:00:12 -05:00
John Pollock
616a89c23f
feat: add multi-GPU WanVideo VAE loader, encode, and decode nodes
...
- Added imports and registrations for WanVideoVAELoader, WanVideoTinyVAELoader,
WanVideoImageToVideoEncode, and WanVideoDecode in multi-GPU versions.
2025-10-06 05:53:23 -05:00
John Pollock
cfe6ebc70f
Re-coded T5 loader and WanVideo Text Encoder 2/24 complete
2025-10-06 00:55:34 -05:00
John Pollock
ddfc95d4a6
Deprecate HunyuanVideoWrapper support
...
Re-synchronize first node for WanVideoWrapper - LoadWanVideoT5TextEncoder
2025-10-05 23:54:41 -05:00
John Pollock
dde3480a82
Merge pull request #117 - refactor: Replace keep_loaded with eject_models Boolean for improved memory management
2025-10-04 18:30:03 -05:00
John Pollock
0919fd4ccb
Minor cleanup related to 2.5.1 release
2025-10-04 18:08:22 -05:00
John Pollock
f60aa6a9a7
refactor: keep_loaded --> eject_models Boolean switch. Use it to eject all other models prior to loading model for inference; helpful to maximize available latent space on device prior to UNet inference, for example
...
So this is a change from something just newly-released in 2.5.0, but most should either see an improvement or no change to behavior. This was the weakest, and jankiest part of 2.5.0 and my decision to manage a CPU memory leak turned into a too-aggressive solution with unwanted side effects.
This solution should provide a better way to manage `compute` VRAM as the most asked-for feature is a way to remove everything else from VRAM prior to main UNet inference, which this accomplishes nicely, as well as reporting back accurate information DisTorch2 on-device shard sizes.
2025-10-04 17:42:52 -05:00
John Pollock
72a20338ef
feat: patch load_models_gpu for accurate memory calculations; unpatch load_models_gpu
...
Refactor memory management in distorch_2.py to patch load_models_gpu instead of LoadedModel.model_memory_required. Implement correct memory reporting based on model flags (eject_models and is_distorch_model), ensuring proper eviction logic and improved handling of virtual VRAM. This drives behavior purely by either comfy core matching or DisTorch flag, fixing potential issues in multi-GPU setups.
2025-10-04 17:16:51 -05:00
John Pollock
e6d19951d7
Paranoia before cleanup
2025-10-04 12:32:37 -05:00
John Pollock
e3750fd737
minor changes
2025-10-04 09:27:33 -05:00
John Pollock
ee41f46beb
revert most changes
2025-10-04 06:47:08 -05:00
John Pollock
d8616acd5e
investigation
2025-10-04 05:29:32 -05:00
John Pollock
a1b7b1fdcf
Merge pull request #114 from pollockjj/d2_clip
...
Major Refactor
2025-09-30 20:38:39 -05:00
John Pollock
2a6a8f4c2b
workflow cleanup
2025-09-30 20:33:48 -05:00
John Pollock
a19e31c915
corrected instructions
2025-09-30 18:38:08 -05:00
John Pollock
24b27c4c83
docs: add node documentation guide to README
...
Add comprehensive guide for accessing and using documentation on core MultiGPU and DisTorch2 nodes, covering 36+ nodes with detailed parameters, outputs, and usage examples. This enhances user experience by providing easy reference for standard ComfyUI loaders and DisTorch2 features, while clarifying coverage excludes third-party nodes.
2025-09-30 18:32:57 -05:00
John Pollock
989b3dc4c3
additional standard loaders documentation
2025-09-30 18:10:51 -05:00
John Pollock
7e0d484b17
docs: Add documentation for standard, gguf, and DisTorch nodes/wrappers
2025-09-30 18:05:48 -05:00
John Pollock
a8a5a6f1fd
feat(__init__): add WEB_DIRECTORY constant for web assets path
...
Add a new constant WEB_DIRECTORY set to "./web" to define the directory path for web-related assets during package initialization. This improves organization by centralizing the path configuration. Additionally, removed trailing newline at file end to maintain consistent code formatting.
2025-09-30 17:02:59 -05:00
John Pollock
64d8ede091
prep for final release candidate
2025-09-30 10:41:14 -05:00
