This extension adds CUDA and CPU device selection to supported loader nodes in ComfyUI. By monkey-patching ComfyUI’s memory management, each model component (like UNet, Clip, or VAE) can be loaded on a specific GPU or the CPU. Examples included are multi-GPU workflows for SDXL, FLUX, LTXVideo, and Hunyuan Video loader nodes
Python · active 2024-12-18 → 2026-06-09 (UTC)
Activity over time
Daily event counts in the loaded window
Line chart, 539 days from 2024-12-18 to 2026-06-09. Pushes: 232 total, peak 11 in a day. Pull requests: 19 total, peak 2 in a day. Issues: 149 total, peak 7 in a day. Comments: 329 total, peak 26 in a day. Stars: 387 total, peak 7 in a day.
- Pushes
- Pull requests
- Issues
- Comments
- Stars
Stars, PRs, issues and forks are under-captured in the later part of this window. GH Archive progressively stopped capturing non-push events during 2026 — −95% or worse by the end of the window. Every series here except Pushes fades for that reason, so a decline above reflects the archive, not this repository. Pushes stay reliable throughout, so read them, and the contributor counts derived from them, as the real signal. Data health has the measurements.
Top contributors
Pushes, PRs, issues, reviews and comments — stars and forks excluded, so this is contribution rather than popularity
| Contributor | Contributions | Pushes | PRs | Comments |
|---|---|---|---|---|
| pollockjj | 442 | 232 | 12 | 125 |
| Copilot | 19 | 0 | 0 | 16 |
| wsbagnsv1 | 16 | 0 | 0 | 16 |
| rkfg | 10 | 0 | 0 | 7 |
| sourcery-ai[bot] | 9 | 0 | 0 | 6 |
| Trojaner | 8 | 0 | 0 | 7 |
| sinand99 | 8 | 0 | 0 | 7 |
| Weroxig | 7 | 0 | 0 | 6 |
| x4080 | 7 | 0 | 0 | 6 |
| an80sPWNstar | 7 | 0 | 0 | 7 |
| suheming | 6 | 0 | 0 | 4 |
| KyleMaas | 6 | 0 | 0 | 6 |
| entauri | 6 | 0 | 0 | 4 |
| doogyhatts | 5 | 0 | 0 | 2 |
| toxyy | 4 | 0 | 0 | 4 |
| AshD | 4 | 0 | 0 | 3 |
| komikndr | 4 | 0 | 0 | 4 |
| Mu5hr00moO | 4 | 0 | 0 | 1 |
| Symbiomatrix | 4 | 0 | 0 | 3 |
| SurpassHR | 4 | 0 | 0 | 3 |
Recent activity
Latest issues, pull requests and releases
- Issue comment#199qa-agent-seveneves[bot]2026-05-08 18:35[codex] Fix aimdo device fallback for MultiGPU
- Issue comment#186Duemellon2026-04-16 00:50New ComfyUI DynamicVRam
- Issue#189lifeishard2026-04-02 19:16CLIPLoaderMultiGPU to cuda:1 gives blank black video
- Issue comment#184Bigfield772026-03-24 04:49Windows fatal exception: stack overflow.
- Issue#184BigStationW2026-03-22 05:39Windows fatal exception: stack overflow.
- Issue#143pollockjj2026-03-21 17:41support FLUX.2-dev
- Issue comment#136pollockjj2026-03-21 17:40I have a 5090 and four V100s. Can they run together using this plugin?
- Issue comment#176seamouse1232026-03-21 10:16qwen disappered and took me 2 hours to see!!
- Issue#121pollockjj2026-03-21 08:39Implement an Inference Engine for distributed parallelism on multi GPU
- Issue#121pollockjj2026-03-21 08:39Implement an Inference Engine for distributed parallelism on multi GPU
- Issue#181pollockjj2026-03-21 08:07New build causing Windows Stack Overflow on all workflows
- Issue comment#181pollockjj2026-03-21 08:07New build causing Windows Stack Overflow on all workflows
- Issue#178pollockjj2026-03-21 03:20Loading CLIP with CLIPLoaderMultiGPU crashes ComfyUI v0.17.1 - v0.17.2
- Pull request#183pollockjj2026-03-21 03:20
- Issue#181tecknight2026-03-18 00:56New build causing Windows Stack Overflow on all workflows
- Issue#177pollockjj2026-03-17 22:58cudaErrorIllegalAddress during cross-device DLPack transfer (RTX 3090 Ti + RTX 5070)
- Issue comment#179Duemellon2026-03-17 22:56Device Switching Node - Feel free to comment or integrate
- Issue#138pollockjj2026-03-16 22:17Fresh install of Comfy and MultiGPU reports CheckpointLoaderAdvancedMultiGPU NOT FOUND
- Issue#142pollockjj2026-03-16 22:17ModuleNotFoundError: No module named 'accelerate'
- Issue#166pollockjj2026-03-16 22:16CLIPLoaderMultiGPU to cuda:1 give completly different results compared to cuda:0 (Wan)
- Issue#158pollockjj2026-03-16 22:16[DEPRECATION WARNING] Detected import of deprecated legacy
- Issue#177tsondo2026-03-14 22:36cudaErrorIllegalAddress during cross-device DLPack transfer (RTX 3090 Ti + RTX 5070)
- Issue comment#171pollockjj2026-03-11 08:13CUDA error: invalid argument when using gguf distorch with cpu donor
- Issue comment#144pollockjj2026-03-11 02:43no other GPU anymore possible to choose after ComfyUI upgrade
- Issue comment#162pollockjj2026-03-10 22:41ValueError: too many values to unpack (expected 4)
Totals cover only the window loaded into ClickHouse and count events, not GitHub's lifetime totals — 387 stars here means stars gained during the window, not the repo's star count.