This website requires JavaScript.
b83a1c30b0
Add pyproject.toml and publish.yml from main
Adrien Toupet
2025-10-15 14:10:17 -04:00
f1ac7d786e
Add perceptually-accurate color correction methods
Adrien Toupet
2025-10-15 13:14:48 -04:00
964d30c212
Update pyproject.toml
Adrien Toupet
2025-10-15 09:57:19 -04:00
e166d3ada0
Merge pull request #205 from ComfyNodePRs/publish
Adrien Toupet
2025-10-15 09:20:33 -04:00
45b0f39d52
Update publish.yml
Adrien Toupet
2025-10-15 09:19:59 -04:00
f4b53f763c
Merge pull request #206 from ComfyNodePRs/pyproject
Adrien Toupet
2025-10-15 09:10:25 -04:00
d0b37e781c
chore(pyproject): Add pyproject.toml for Custom Node Registry
async () = process.env.GIT_USERNAME || (await ghUser()).email && (await ghUser()).name || DIE("Missing env.GIT_USERNAME")
2025-10-15 06:58:18 +00:00
ad7177332b
chore(publish): Add Github Action for Publishing to Comfy Registry
async () = process.env.GIT_USERNAME || (await ghUser()).email && (await ghUser()).name || DIE("Missing env.GIT_USERNAME")
2025-10-15 06:56:18 +00:00
e48c2da776
feat: global model cache, alpha upscaling rewrite, and quality improvements
Adrien Toupet
2025-10-15 00:51:56 -04:00
d6e04aa108
refactor: modularize node architecture and enhance BlockSwap flexibility
Adrien Toupet
2025-10-11 17:37:58 -04:00
bf46dfd674
Simplify RGBA handling and improve logging consistency
Adrien Toupet
2025-10-11 11:06:15 -04:00
1b00665757
feat: Add RGBA support, improve config caching, and enhance logging
Adrien Toupet
2025-10-11 02:26:59 -04:00
fb2b6c7d4c
Fix BlockSwap not applying on cached models and rename offload_io_components to swap_io_components
Adrien Toupet
2025-10-10 17:48:12 -04:00
9ee244d71b
feat: implement torch.compile optimization with BlockSwap compatibility
Adrien Toupet
2025-10-10 12:34:49 -04:00
5567f232af
feat!: independent DiT/VAE management with separate loader nodes
Adrien Toupet
2025-10-08 00:49:48 -04:00
3a90290c03
feat: implement deferred model loading for better memory management - VAE loads at encoding phase, DiT loads at upscaling phase - BlockSwap config changes for cached models deferred to DiT phase - Add comprehensive cleanup of temporary loading attributes
Adrien Toupet
2025-10-03 17:29:51 -04:00
a9b6e4201c
Fix BlockSwap GPU memory leak and restore performance metrics
Adrien Toupet
2025-10-03 16:04:48 -04:00
2fab3a1caf
refactor: split decode and post-processing into separate phases for cleaner architecture
Adrien Toupet
2025-10-03 09:53:13 -04:00
984f9af478
Merge pull request #171 from AInVFX/nightly
Adrien Toupet
2025-09-29 14:36:02 -04:00
9284332a1f
fix(regression): resolve tensor dimension mismatch in color correction when padding is applied (https://github.com/numz/ComfyUI-SeedVR2_VideoUpscaler/issues/169#issuecomment-3348373882 )
Adrien Toupet
2025-09-29 14:33:51 -04:00
c345b52966
Merge pull request #170 from AInVFX/nightly
Adrien Toupet
2025-09-29 12:04:15 -04:00
3e99ec72d0
feat: add type hints to generation.py for improved type safety and IDE support
Adrien Toupet
2025-09-29 11:55:42 -04:00
4a3efc7883
feat: add CPU offloading for intermediate data to reduce VRAM usage
Adrien Toupet
2025-09-25 11:30:32 -04:00
547ccc749b
Merge pull request #163 from AInVFX/nightly
Adrien Toupet
2025-09-25 00:34:54 -04:00
9b4a7dfa3e
fix: unify model registry for CLI and improve GGUF implementation
Adrien Toupet
2025-09-25 00:09:56 -04:00
1fa603899f
Fix GGUF implementation: centralize warning suppression, improve type hints, enhance error handling
Adrien Toupet
2025-09-24 15:47:18 -04:00
0b0c87ed4a
Add GGUF quantized model support (based on PR #121 from @cmeka / @lihaoyun6)
Adrien Toupet
2025-09-24 11:25:31 -04:00
953524b13b
feat: Add support for extra_model_paths.yaml configuration
Adrien Toupet
2025-09-23 00:56:26 -04:00
0967bb153c
fix: update dependencies to include all required packages for CLI and ComfyUI
Adrien Toupet
2025-09-23 00:01:02 -04:00
efa851009f
Fix corrupted model downloads with hash validation and resume support
Adrien Toupet
2025-09-22 23:45:43 -04:00
2b00252a49
Simplify __init__.py files + fix dependencies + add pyproject.toml
Adrien Toupet
2025-09-22 15:47:04 -04:00
8f46898073
Fix OOM when loading models without preserve_vram by using meta device initialization
Adrien Toupet
2025-09-22 14:48:15 -04:00
9927f45f02
Fix BlockSwap config detection and progress bar cleanup
Adrien Toupet
2025-09-22 13:46:50 -04:00
22f540a7d6
feat: Add input noise parameter and rename latent noise for clarity
Adrien Toupet
2025-09-22 11:47:49 -04:00
d0d24a18cb
Refactor: Replace dynamic relative imports with model class registry for cross-platform compatibility. Minor: Adjust ASCII logo width for better screen fit
Adrien Toupet
2025-09-20 01:53:25 -04:00
e89a8b00a5
feat: promote cond_noise_scale to UI for controlling conditioning noise level helping mitigate noise artifacts in high-res upscaling
Adrien Toupet
2025-09-18 23:19:35 -04:00
e6da542b7b
VAE tiling defaults + Fix BlockSwap + preserve_vram compatibility and device management
Adrien Toupet
2025-09-17 23:39:03 -04:00
ee95a3a673
feat: Add configurable color correction methods and improve memory management
Adrien Toupet
2025-09-16 22:21:45 -04:00
1518ecc1a5
refactor: Fix ComfyUI node conflicts via relative imports and clearer model structure
Adrien Toupet
2025-09-16 16:43:01 -04:00
b41998c5e1
docs: Clean up README formatting
Adrien Toupet
2025-09-16 15:34:48 -04:00
2c570105b1
docs: Update project documentation and credits
Adrien Toupet
2025-09-16 15:29:39 -04:00
477f57fd5a
Refactor: Three-phase batch processing pipeline for improved performance
Adrien Toupet
2025-09-11 15:57:09 -04:00
d504eeafa6
fix(regression): Restore MPS compatibility for VAE models
Adrien Toupet
2025-08-29 18:15:56 +02:00
cf274a990b
fix(regression): Restore MPS compatibility for VAE models
Adrien Toupet
2025-08-29 12:14:23 -04:00
36f4d80e6f
Merge pull request #137 from AInVFX/nightly - disable detailed tensor inspection
Adrien Toupet
2025-08-29 15:07:37 +02:00
fc27097045
fix: disable detailed tensor inspection to avoid deprecation warnings and indirect dependency calls
Adrien Toupet
2025-08-29 09:05:29 -04:00
fe61f1866e
Memory efficiency, performance improvements,and architecture refactoring
Adrien Toupet
2025-08-29 06:18:44 +02:00
66bcf436ab
perf: optimize debug logging to reduce performance overhead
Adrien Toupet
2025-08-28 23:50:49 -04:00
22ffc86af7
perf: Optimize VAE operations and add missing memory cleanup to decode
Adrien Toupet
2025-08-28 16:18:12 -04:00
5e2fb76414
fix: optimize RoPE frequency computation and remove initialization bottleneck
Adrien Toupet
2025-08-28 14:20:41 -04:00
feabb94e5a
perf: optimize memory usage and streamline generation pipeline
Adrien Toupet
2025-08-27 14:15:41 -04:00
267bc099e6
Fix BlockSwap compatibility with preserve_vram mode
Adrien Toupet
2025-08-26 17:31:04 -04:00
89bde29cbe
perf: optimize model initialization with meta device for 90% speedup
Adrien Toupet
2025-08-26 13:23:30 -04:00
1ee1ee0fb5
fix: add OOM retry mechanism and fix FP16 black frames regression
Adrien Toupet
2025-08-26 00:03:02 -04:00
0358198804
refactor: Remove redundant dtype conversions and autocast contexts
Adrien Toupet
2025-08-25 20:37:52 -04:00
5595d58597
refactor: streamline generation pipeline and improve dtype handling
Adrien Toupet
2025-08-25 17:11:29 -04:00
47552c0ed5
refactor: unify model configuration and improve logging precision
Adrien Toupet
2025-08-25 13:45:13 -04:00
53f41770c1
fix: use memory-mapped loading for VAE via assign=True
Adrien Toupet
2025-08-25 11:03:35 -04:00
8c48aa59ea
Fix VRAM memory leak in VAE decoder
Adrien Toupet
2025-08-25 10:12:40 -04:00
7e7853df54
Blockswap regression with cache model
Adrien Toupet
2025-08-24 18:15:31 +02:00
627f215b57
Fix BlockSwap memory configuration being lost during model caching
Adrien Toupet
2025-08-24 11:38:02 -04:00
a2b546c0e7
Fix BlockSwap regression: preserve block positions when caching
Adrien Toupet
2025-08-23 22:58:53 -04:00
5bf439ee3d
refactor: Memory management overhaul with enhanced debugging and performance optimizations
Adrien Toupet
2025-08-23 08:25:20 +02:00
8333fb856e
refactor(WIP): complete memory management overhaul with proper error handling
Adrien Toupet
2025-08-23 01:13:45 -04:00
faea9397d9
Fix: Remove non_blocking from BlockSwap to prevent CPU-GPU pinned memory leaks
Adrien Toupet
2025-08-22 16:57:48 -04:00
f51b0b3991
Fix debug logging: propagate debug instance to sampler and VAE submodules
Adrien Toupet
2025-08-22 16:29:44 -04:00
2053a80f38
refactor(WIP): Improve debug logging consistency and reduce redundancy - Remove "Force move weights to device" in FP8CompatibleDiT forward() to avoid clash with blockswap - This was used for preserve_vram but will refactor preserve_vram in a separate commit - Remove duplicate dtype detection in generation_step (now passed from generation_loop) - Add device checks before CPU moves to avoid redundant operations - Improve debug.log and debug.log_memory_state() formatting, content, and categories for better visibility - Remove unused imports and excessive clear_memory() calls - Clean up non-essential logs from always display
Adrien Toupet
2025-08-22 15:30:02 -04:00
9b7c113681
refactor(WIP): Memory Management Overhaul This is a work-in-progress commit that consolidates a series of changes to fix memory leaks, optimize VRAM/RAM usage, improve performance, and enhance code maintainability.
Adrien Toupet
2025-08-22 09:20:41 -04:00
0373a81d2b
Merge pull request #126 from benjaminherb/fix-prepend-frames
Adrien Toupet
2025-08-21 14:17:58 +02:00
22aa85abb7
fix: Updates --prepend_frames logic
Benjamin Herb
2025-08-21 08:12:07 +02:00
1a12465588
Merge pull request #125 from JohnAlcatraz/nightly
Adrien Toupet
2025-08-20 22:53:06 +02:00
2eaebf7709
Update comfyui_node.py
JohnAlcatraz
2025-08-20 22:07:37 +02:00
ba699ceb3d
Update generation.py
JohnAlcatraz
2025-08-20 22:06:55 +02:00
8cd5f981f4
Update infer.py
JohnAlcatraz
2025-08-20 22:06:37 +02:00
7f4a1585c3
Merge pull request #124 from AInVFX/nightly
Adrien Toupet
2025-08-19 21:10:09 +02:00
7f2d47e363
Fix NameError: undefined 'devices' in SeedVR2 execute method
Adrien Toupet
2025-08-19 15:05:41 -04:00
9d4d8a94a5
refactor: clarify batch merging terminology and fix debug categories
Adrien Toupet
2025-08-19 13:48:37 -04:00
c4bbf6f27b
Merge pull request #123 from AInVFX/nightly
Adrien Toupet
2025-08-19 17:53:10 +02:00
7ee3b33627
feat: improve VAE tiling flexibility and fix cached model params Remove tile size/overlap max limits, set min=64/step=32 for finer control, update tooltips for clarity, and fix runtime parameter updates when cache_model=True
Adrien Toupet
2025-08-19 11:43:45 -04:00
e23259df75
Removed deprecated files
Adrien Toupet
2025-08-19 09:49:53 -04:00
883ca638d6
Merge pull request #118 from lihaoyun6/nightly
Adrien Toupet
2025-08-19 15:06:33 +02:00
ed205f84b3
Set the batch_size of the image upscaling workflow to 1
lihaoyun6
2025-08-19 13:25:51 +08:00
0dd3831512
Added example workflows
lihaoyun6
2025-08-19 13:05:53 +08:00
12aefc08bb
Merge pull request #107 from lihaoyun6/main
Adrien Toupet
2025-08-18 21:27:23 +02:00
facbc1a824
Fix FPS calculation
lihaoyun6
2025-08-19 00:58:02 +08:00
e723e176cc
Added FPS calculation
lihaoyun6
2025-08-19 00:23:27 +08:00
286a0186de
Set use_non_blocking to False by default to prevent memory leaks
lihaoyun6
2025-08-18 23:18:36 +08:00
3ea159d763
Fix incorrect variable names
lihaoyun6
2025-08-18 22:14:01 +08:00
add941dc6e
Fix incorrect variable names
lihaoyun6
2025-08-18 20:53:18 +08:00
1cfde3d1c7
Added multi-gpu support for ComfyUI nodes
lihaoyun6
2025-08-15 21:06:16 +08:00
176562f829
Restored compatibility with FP8 safetensors for MPS backend; Changed the way to detect MPS device
lihaoyun6
2025-08-13 00:53:05 +08:00
8ae90c3af2
Keep weights and arguments on the same device when 'preserve_vram' is ON
lihaoyun6
2025-08-12 23:08:30 +08:00
4eef2ca85c
Merge pull request #103 from lihaoyun6/main
Adrien Toupet
2025-08-12 16:44:24 +02:00
3db50c3849
update .gitignore
lihaoyun6
2025-08-12 22:32:36 +08:00
e328545582
Delete .DS_Store file
lihaoyun6
2025-08-12 22:31:40 +08:00
3b0699ead9
Merge remote-tracking branch 'upstream/nightly'
lihaoyun6
2025-08-12 22:28:36 +08:00
6412f03484
Merge pull request #92 & #106
Adrien Toupet
2025-08-12 14:59:37 +02:00
58fbd215c3
Merge pull request #85 from AInVFX/requirements
Adrien Toupet
2025-08-12 14:41:50 +02:00
0029dd6283
Added nightly disclaimer
Adrien Toupet
2025-08-12 08:26:25 -04:00
54cd9083ea
Added tiled VAE options to Comfy node
not-ski
2025-08-11 17:02:49 -06:00
Adrien Toupet