asr-models-qnn
k2-fsa/sherpa-onnxasr-models-qnnNov 10, 2025by csukuangfj
AI Summary
This release details the constraints of QNN (Qualcomm NPU) regarding dynamic input shapes, explaining that models are limited to fixed durations (e.g., 10 seconds) and are padded or truncated accordingly.
Key Highlights
- Explanation of QNN dynamic input shape limitation (fixed duration support).
- Models are padded or truncated to the max duration indicated in the filename.
- Documentation link for QNN model usage.
New Features
- QNN usage documentation and constraints explanation
Full Release Notes
Since QNN does not support dynamic input shapes, we limit the maximum duration the model can handle. For example, if the limit is 10 seconds, any input shorter than 10 seconds will be padded to 10 seconds, and inputs longer than 10 seconds will be truncated to that length. The model name indicates the maximum duration the model can handle. See https://k2-fsa.github.io/sherpa/onnx/qnn/index.html for how to use them. If you want to reduce the initialization time for the first run, please use files from https://github.com/k2-fsa/sherpa-onnx/releases/tag/asr-models-qnn-binary