Running Chatterbox TTS on AMD ROCm: Workarounds for CUDA-Only Dependencies
JackStrawWitchita · reddit · 2026-08-23
The author shares their experience getting Chatterbox TTS running on an AMD 7900 XTX with ROCm 7.2, identifying and resolving two dependency roadblocks.
Issues and Fixes
- faster-whisper crashes: It relies on ctranslate2, which lacks ROCm support (CUDA only). Workaround: Enable "Bypass Whisper Checking" in the UI to skip the validation step, sacrificing auto-retry but retaining generation capability.
- torchaudio.save/load failures: Newer torchaudio versions route I/O through torchcodec, whose pre-built wheels are CUDA-exclusive. Fix: Replace torchaudio calls with custom wrapper functions using the soundfile library (already a dependency). The author provides the drop-in code snippets.
Takeaway
While core PyTorch works on ROCm, projects often break on CUDA-only dependencies like ctranslate2 and torchcodec.
More from Infra
- NAS boot failure复盘: Store metadata on separate drives — TheZachMueller · 2026-08-23
- Nvidia Rubin NVL72 Rack Could Cost ~$8M — zephyr_z9 · 2026-08-23
- Vertex AI Users Complain: No Native Hard Spending Cap, Must DIY via Pub/Sub — MeetingWeird9418 · 2026-08-23
- Open source MCP server implements HTTP 402 micropayment gateway — EstablishmentTough18 · 2026-08-23
- DeepSeek V4 Flash on M2 Ultra: lossless repack to 141GiB, 25.8 t/s beats M3 Ultra — Agusx1211 · 2026-08-23
- Running Ollama LFM2.5-2.6B on Intel iGPU under Linux: a how-to guide — Coolsh0e · 2026-08-23