Expand description
TTS service — concurrent slot-based serving of TtsModelExecutor.
Multiple TTS requests can be processed in parallel, each on its own executor slot. Slots share nothing (each has its own model weights + KV cache). Future: Phase 2 will share weights across slots to reduce memory.
The struct is named TtsService to free the TtsEngine identifier
for the trait of the same name in ferrum-interfaces.
Structs§
- TtsService
- Concurrent TTS service. Implements
TtsEngine(the modality trait) andInferenceEngine(lifecycle).