mARTin-B78/DGX-Spark-ComfyUI
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Change CUDA module loading to lazy loading

Browse Source
This commit is contained in:
luix93 2026-03-28 16:51:22 +01:00 committed by GitHub
parent 92fbd1708d
commit 0b07cb4a7e
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 1 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
README.md
View File

@ -118,7 +118,7 @@ The default `COMFYUI_FLAGS` are tuned for the Grace-Blackwell unified memory arc
| `PYTORCH_NO_CUDA_MEMORY_CACHING=1` | Disables PyTorch CUDA caching allocator — lets the unified memory manager handle allocations |
| `CUDA_MANAGED_FORCE_DEVICE_ALLOC=1` | Forces managed memory to prefer device allocation |
| `TORCH_COMPILE_DISABLE=1` / `TORCHDYNAMO_DISABLE=1` | Disables `torch.compile` — Triton does not support `sm_121a` yet |
| `CUDA_MODULE_LOADING=CUDA` | Loads CUDA modules eagerly for faster model loading |
| `CUDA_MODULE_LOADING=LAZY` | Loads CUDA modules lazily for faster model loading and more memory efficiency |
| `OMP_NUM_THREADS=20` | Tuned for the Grace CPU core count |
---