miles/docs/getting-started/installation.md

title, description

title	description
Installation	Install Miles on NVIDIA or AMD GPUs. Docker is the recommended path.

There are three ways to install Miles. Docker is recommended because Miles pins patched versions of SGLang, Megatron-LM, and a few CUDA kernels.

Method 1: Docker (recommended)

```bash
docker pull radixark/miles:latest

docker run --rm \
  --gpus all --ipc=host --shm-size=32g \
  --ulimit memlock=-1 --ulimit stack=67108864 \
  --network=host \
  -it radixark/miles:latest /bin/bash
```

```bash
docker pull rlsys/miles:MI350-355-latest    # or MI300-latest

docker run --rm \
  --device /dev/dri --device /dev/kfd \
  --group-add video --ipc=host --shm-size=32g \
  --cap-add SYS_PTRACE --security-opt seccomp=unconfined \
  --privileged \
  -it rlsys/miles:MI350-355-latest /bin/bash
```

The image ships with:

• PyTorch (matching the container's CUDA / ROCm version)
• Megatron-LM, SGLang, FlashAttention-3, DeepGEMM, Apex
• Ray, uv, and Miles installed editable at /root/miles

See Hardware requirements for per-GPU status.

Method 2: From source

Clone and install in an existing environment.

git clone https://github.com/radixark/miles.git
cd miles
pip install -r requirements.txt
pip install -e . --no-deps

Patched dependencies. Miles pins patched versions of SGLang and Megatron-LM. Installing them yourself at the wrong commit is the most common source of bug reports — use Docker if you can.

Method 3: Update an existing container

If you already run a Miles image and want the latest code:

cd /root/miles
git pull --rebase
pip install -e . --no-deps
ray stop && ray start --head --port=6379

Verify

Confirm Miles imports and the GPUs are visible:

python -c "import miles; print('Miles import OK')"
nvidia-smi

If either command fails, see Debugging.

Hardware requirements

Hardware	Status
NVIDIA GB300 / GB200 / B300 / B200	Production
NVIDIA H200 / H100	Production (CI guarded)
NVIDIA A100	Supported — FP8 features disabled
AMD MI300X, MI325, MI350X, MI355X	Supported via ROCm

For multi-node training you also need a high-bandwidth interconnect — InfiniBand, RoCEv2, or Slingshot — and 200+ GB/s per node. Single-node jobs run fine over NVLink only.

Next steps

• Quick Start — run your first training job.
• Core concepts — the mental model behind Miles.
• Training backends — Megatron vs FSDP.