c5e3e6a215
build: update to 2.5.11 based on bug fixes for latest comfy changes
John Pollock2026-01-02 22:54:25 -06:00
87bde131c4
fix: Ensure model patching fallback logic only applies when 'model_patches_models' attribute is absent.
John Pollock2026-01-02 22:48:43 -06:00
03e5368f8e
pdate Distorch for ComfyUI 0.6.0+ load list with device assignment caching.
John Pollock2026-01-02 22:20:55 -06:00
5912141732
Update distorch_2.py: Add compatibility for GGUFModelPatcher and ModelPatcher lacking model_patches_models attribute
Maksym Solomyanov2025-10-14 13:28:27 +02:00
a24f0a6e87
hotfix: Corrects for corner case when DisTorch VirtualVRAM=0.0 GB (previous refactor shunted to standard loader. This replicates that required logic across all nodes using DisTorch2 for allocations. Next time I will wait for the final test work flow to finish VAE conversion (where is the only place I test this.)
John Pollock2025-10-13 22:08:05 -05:00
1951513e92
refactor: migrate DisTorch2 allocation tracking to per-model metadata
John Pollock2025-10-13 19:00:59 -05:00
bdd612c619
hotfix: #124 - fix for CLIPTextEncode error: ‘GGUFModelPatcher’ object has no attribute ‘model_patches_models’.
John Pollockandmax-solo232025-10-13 13:28:32 -05:00
5310d4a28d
Update documentation and example workflows to match Comfy guidelines
John Pollock2025-10-13 02:16:52 -05:00
fd60ae6f65
Add MultiGPU support for various model loaders and encoders
John Pollock2025-10-13 02:10:41 -05:00
01b41f4514
Update workflow examples to match Comfy Core guidelines
John Pollock2025-10-13 01:52:20 -05:00
161174266c
Add UNet offload device support, and enhance Florence2 model loading with safetensors conversion option.
John Pollock2025-10-12 04:35:43 -05:00
e31ba04f20
WanVideoWrapper updates to reflect reported changes (#118, #90) in the ComfyUI-WanVideoWrapper code base. Breaking changes. See example workflows.
John Pollock2025-10-10 14:28:11 -05:00
40d40ef9b6
Finished implementation of wanvideowrapper nodes, updated example workflows, added some benchmarking files for reference.
John Pollock2025-10-10 13:40:37 -05:00
db931d66c3
Add WanVideoControlnetLoader class for loading ControlNet models
John Pollock2025-10-10 09:30:46 -05:00
faef324975
correcting implementation of salvagable nodes and obiliterating one abomination
John Pollock2025-10-10 09:28:58 -05:00
610c2102dd
Add multi-GPU support for remaining planned WanVideo model loaders
John Pollock2025-10-10 04:35:16 -05:00
375cbd093c
Add LoadWanVideoClipTextEncoder and WanVideoClipVisionEncode for multi-GPU support
John Pollock2025-10-09 19:05:28 -05:00
2f605e0db3
Add WanVideoVACEEncode and enhance WanVideoEncode for multi-GPU support
John Pollock2025-10-09 14:41:53 -05:00
ba0a3a6e21
improve documentation on supporting ROCm cuda-enumerated GPUs.
John Pollock2025-10-09 13:01:20 -05:00
b04ae1e378
Coverted WanVideoImageToVideoEncode to shim
John Pollock2025-10-09 12:32:29 -05:00
04013c3ee5
feat: add WanVideoTextEncodeCached and WanVideoTextEncodeSingle classes for enhanced text encoding functionality
John Pollock2025-10-08 06:08:16 -05:00
d300c12e9f
feat: add WanVideoBlockSwap class and integrate into MultiGPU node mappings
John Pollock2025-10-07 14:22:01 -05:00
27cfb83733
feat: add MultiGPU logging enhancements and new CI scripts for workflow execution and log summarization, WanVidwoWrapper Model Loader/Sampler re-implemented
John Pollock2025-10-07 13:19:09 -05:00
616a89c23f
feat: add multi-GPU WanVideo VAE loader, encode, and decode nodes
John Pollock2025-10-06 05:53:23 -05:00
cfe6ebc70f
Re-coded T5 loader and WanVideo Text Encoder 2/24 complete
John Pollock2025-10-06 00:55:34 -05:00
ddfc95d4a6
Deprecate HunyuanVideoWrapper support Re-synchronize first node for WanVideoWrapper - LoadWanVideoT5TextEncoder
John Pollock2025-10-05 23:54:41 -05:00
dde3480a82
Merge pull request #117 - refactor: Replace keep_loaded with eject_models Boolean for improved memory management
John Pollock2025-10-04 18:30:03 -05:00
0919fd4ccb
Minor cleanup related to 2.5.1 release
John Pollock2025-10-04 18:08:22 -05:00
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
John Pollock2025-10-04 17:42:52 -05:00
72a20338ef
feat: patch load_models_gpu for accurate memory calculations; unpatch load_models_gpu
John Pollock2025-10-04 17:16:51 -05:00