This documentation was AI-generated. If you find any errors or have suggestions for improvement, please feel free to contribute! Edit on GitHubThe TorchCompileModel node applies PyTorch compilation to a model to optimize its performance. It creates a copy of the input model and wraps it with PyTorch’s compilation functionality using the specified backend. This can improve the model’s execution speed during inference.
Inputs
| Parameter | Data Type | Required | Range | Description |
|---|---|---|---|---|
model | MODEL | Yes | - | The model to be compiled and optimized |
backend | STRING | Yes | ”inductor" "cudagraphs” | The PyTorch compilation backend to use for optimization |
Outputs
| Output Name | Data Type | Description |
|---|---|---|
model | MODEL | The compiled model with PyTorch compilation applied |