Text-Cue-Guided TS-ASR weights
model.pt is the checkpoint for Text-Cue-Guided TS-ASR, a target-speaker speech
recognizer that transcribes only the speaker who says a given text cue. It holds
every weight the model needs: the cue encoder, the Conformer acoustic encoder,
the audio projector, and Qwen2.5-0.5B with its LoRA adapters.
- SHA-256:
870aef70a8a8c2b292b3eb959ea4a0907edc751848729070aea04465b449fcf5 - Inference code and demo:
YuCeong-May/TS-ASR
Licenses
- Model weights: FunASR Model Open Source License 1.1, see
MODEL_LICENSE. - Code: MIT, see
LICENSE. - The embedded Qwen2.5-0.5B weights are under Apache-2.0.