Torch Compile Speed Settings
TorchCompileSpeedSettings
torch.compile settings for maximum speed optimization. Speed Mode Features: - Uses inductor backend with max-autotune-no-cudagraphs - Enables dynamic compilation for better cache reuse - Disables CUDA graphs for flexibility - Enables all Triton autotune optimizations - First run: comprehensive autotune (slower) - Second run: cached execution (extremely fast) Author: eddy
Pack: ComfyUI-TorchCompileSpeed
custom_nodes.ComfyUI-TorchCompileSpeed
Inputs (12)
| Name | Type | Required |
|---|---|---|
| backend | COMBO | required |
| fullgraph | BOOLEAN | required |
| mode | COMBO | required |
| dynamic | BOOLEAN | required |
| dynamo_cache_size_limit | INT | required |
| compile_transformer_blocks_only | BOOLEAN | required |
| reuse_if_similar | BOOLEAN | required |
| experimental_ptx | BOOLEAN | required |
| ptx_fast_math | BOOLEAN | required |
| warmup_runs | INT | required |
| ptx_cache_dir | STRING | optional |
| dynamo_recompile_limit | INT | optional |
Outputs (1)
| Name | Type |
|---|---|
| torch_compile_args | WANCOMPILEARGS |