v0.54.2 v0.3.1
ysharma3501/NovaSRJul 13, 2026 → Jul 15, 20262 days between releases
What's New
Newer release adds public warmup API and backend compatibility; older release is unrelated project v0.54.2
New Features
- Added public model.warmup(prefill_len=100) API
- Maintained private _warmup() as compatibility alias
- Updated demo Space to install PyPI package
Improvements
- Improved backend interface compatibility between Torch and GGML
- Optimized Torch CUDA graph capture
- Treated GGML warmup as safe no-op
Bug Fixes
- Resolved backend interface mismatch reported in issue #354
Upgrade Notes
Install via pip install faster-qwen3-tts==0.3.1; internal _warmup() remains available as a compatibility alias.
The Two Releases
Release v0.54.2
Jul 13, 2026· 3 months ago
This release introduces support for secure TLS connections to the Hubble Relay, allows local account users to change their passwords, and disables temperature in AI requests. It also fixes a slow bulk alert status change issue.
Full release notesv0.3.1
Jul 15, 2026· 3 months ago
This release exposes a public backend warmup API to resolve interface mismatches between Torch and GGML backends, ensuring compatibility across different implementations.
Full release notes