# TODO

## ROCm / vLLM on Strix Halo (gfx1151)

The Framework Desktop runs **Ubuntu 24.04 LTS (noble)**, which aligns
with AMD's ROCm 7.x packaging. The deploy installs `rocminfo` and
`librocm-smi-dev` host-side; heavier ROCm bits (full HIP toolchain,
device-mapped libraries) still run inside containers that ship their
own ROCm stack. The host stays slim by design.

### Open questions

- **Does `rocm/vllm:latest` actually run on Strix Halo's iGPU?** vLLM's
  AMD support officially targets datacenter cards (MI300X / gfx942).
  gfx1151 (RDNA 3.5 consumer) is a different ISA. If the stock image
  doesn't initialize the device, try `rocm/vllm-dev:nightly` or build
  from source against ROCm 7.x with `-DAMDGPU_TARGETS=gfx1151`.

### If you ever want full host-side ROCm

For native ROCm work on the host (compiling HIP kernels, full toolchain):
1. Bump `ROCM_VERSION` and `AMDGPU_INSTALL_DEB` in
   `pyinfra/framework/deploy.py` to the latest release.
2. Add a step that runs `amdgpu-install -y --usecase=rocm --no-dkms`
   (currently avoided to stay slim — ~25 GB toolchain).
3. `./run.sh`.

For container-only workflows (current default), no action is needed —
container images update independently of the host.

## Pick a coding model (StrixHaloSetup Phase 6)

Open question — research current Strix Halo benchmarks before
committing. Candidates: Qwen3-Coder, DeepSeek-Coder-V3.x, GLM-4.6,
Devstral, Kimi-K2. Track Kimi Linear separately via the weekly routine
referenced in `StrixHaloSetup.md`.