Skip to content

[Feature Request] Windows AMD GPU support via DirectML / torch-directml #531

Description

@ragsteban

Environment:

  • OS: Windows 11 Pro (AtlasOS)
  • GPU: AMD RX 6600 (RDNA 2, 8GB VRAM)
  • Voicebox: v0.4.5

Issue:
GPU is not detected — app falls back to CPU only.
The CUDA backend download only supports NVIDIA GPUs.

Expected:
The README mentions "AMD ROCm, Intel XPU, and DirectML are also supported
where available through PyTorch." However, the bundled backend does not
include torch-directml, so AMD GPUs on Windows get no acceleration.

ROCm is Linux-only, so Windows AMD users have no GPU path currently.

Request:
Add a downloadable DirectML backend (torch-directml) for Windows AMD/Intel
GPU users, similar to how the CUDA backend is offered as an optional download.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions