Has anyone here managed to get Dia-TTS-Server running on ROCm with AMD GPUs?
Docs only mention CUDA + CPU fallback, but I’m on an AMD card (RX 6800). Right now it only runs on CPU, which is painfully slow.
Would love to know:
- if there’s a clean way to point it at ROCm PyTorch,
- any config/env tweaks needed,
- or if someone already has a working Docker/compose setup for AMD.
Basically just trying to avoid burning hours hacking together my own build if someone’s already solved it. Any pointers would be appreciated.