# Training on AMD GPUs

TRL trainers run on AMD Instinct GPUs without code changes. PyTorch's ROCm build exposes them as `cuda` devices, so the commands in the rest of the documentation work as they are.

## Installation

Install PyTorch's ROCm build, then TRL in the same environment:

```bash
pip install torch torchvision --index-url https://download.pytorch.org/whl/rocm7.2
pip install trl
```

## Attention kernels

The FlashAttention kernels used in the examples of this documentation (`kernels-community/flash-attn2`, `flash-attn3` and `flash-attn4`) have no ROCm build. On AMD GPUs, use [`kernels-community/aiter-flash-attn`](https://huggingface.co/kernels-community/aiter-flash-attn), downloaded from the Hub by the [`kernels`](https://github.com/huggingface/kernels) library (`pip install trl[kernels]`):

```python
from trl import SFTConfig

training_args = SFTConfig(model_init_kwargs={"attn_implementation": "kernels-community/aiter-flash-attn@v2"})
```

The kernel has no attention-logit soft-capping, and Transformers skips `attn_logit_softcapping` without a warning, so models that use it, such as Gemma 2, train with uncapped attention.

## Generation with vLLM

GRPO and RLOO with `use_vllm=True` need vLLM's ROCm build, see the [vLLM installation guide](https://docs.vllm.ai/en/latest/getting_started/installation/gpu.html).