ComfyUI Node
PyramidFlow Torch Compile Settings
torch.compile settings, when connected to the model loader, torch.compile of the selected layers is attempted. Requires Triton and torch 2.5.0 is recommended
PyramidFlow Torch Compile Settings
- torch_compile_args
◄backendinductor►
◄fullgraphfalse►
◄modedefault►
◄compile_whole_modelfalse►
◄single_blockstrue►
◄double_blockstrue►
◄embedderstrue►
◄compile_resttrue►
◄dynamo_cache_size_limit64►
CategoryMochiWrapper
Inputs (9)
| Name | Type | Default | Description |
|---|---|---|---|
| backend | COMBO | inductor | 2 options: inductor, cudagraphs |
| fullgraph | BOOLEAN | false | Enable full graph mode |
| mode | COMBO | default | 4 options: default, max-autotune, max-autotune-no-cudagraphs, reduce-overhead |
| compile_whole_model | BOOLEAN | false | Compile the whole model, overrides other block settings |
| single_blocks | BOOLEAN | true | Compile single_blocks |
| double_blocks | BOOLEAN | true | Compile transformer blocks |
| embedders | BOOLEAN | true | Compile embedders |
| compile_rest | BOOLEAN | true | Compile the rest of the model (proj and norm out) |
| dynamo_cache_size_limit | INT | 640–1024 | torch._dynamo.config.cache_size_limit |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| torch_compile_args | PYRAMIDFLOW_COMPILEARGS | — |