# DirectML

SD.Next includes support for PyTorch-DirectML.

> [!IMPORTANT]
> DirectML support is marked End-of-Life (EOL) and will be removed in a future release.
> `torch-directml` has not received updates in over one year and is currently superseded by `ROCm` or `ZLUDA`.  

## How to

Add `--use-directml` to the command-line arguments.

For details, go to [Installation](Installation#initial-installation).

## Performance

Performance is significantly lower than ZLUDA and ROCm.

If your card is relatively new and you prefer Windows, use ZLUDA.

If you are comfortable with Linux, use ROCm.

## FAQ

### DirectML does not collect garbage memory

PyTorch-DirectML does not access graphics memory by indexing. Because PyTorch-DirectML's tensor implementation extends [OpaqueTensorImpl](https://github.com/pytorch/pytorch/blob/main/aten/src/ATen/OpaqueTensorImpl.h), we cannot access the actual storage of a tensor.

### An error occurs with no error message

If you met `RuntimeError` with no error message (or empty), please report us via GitHub issue or Discord. (please check whether there's a duplicated issue)
If you get `RuntimeError` with no error message, report it on GitHub or Discord after checking for duplicates.

### It does not work properly with FP16

If it works with FP32, please report us via GitHub issue or Discord. (please check whether there's a duplicated issue)
If it works with FP32 but not FP16, report it on GitHub or Discord after checking for duplicates.

### The terminal is suddenly frozen during generation

Please report us via GitHub issue or Discord. (please check whether there's a duplicated issue)
Please report it on GitHub or Discord after checking for duplicates.

## Olive (experimental support)

Refer to [ONNX Runtime](ONNX-Runtime)