Release v1.7.3

• Support TDNN models of the yesno recipe from icefall

Release v1.7.2

• Add non-streaming Python websocket server

Release v1.7.0

• Fix building with GPU on Linux
• Refactor code to make it easier support other kinds of streaming models. The breaks the Python API. sherpa_onnx.OnlineRecognizer() should be changed to sherpa_onnx.OnlineRecognizer.from_transducer().

Release v1.6.2

• Bugfixes

Release v1.6.0

• Support Whisper models for non-streaming speech recognition

Release v1.5.5

Fix building wheels on Windows (#229)

Release v1.5.4

Add go-api-examples.

Tested on the following platforms

• Linux (x86_64, aarch64, arm)
• Windows (x86_64, x86)
• macOS (x86_64, aarch64)

Release v1.5.3

• Support Android x86
• Publish precompiled Android libraries
• Support model_type for loading streaming models

Release v1.5.2

• Reduce initialization time for non-streaming models
• Add non-streaming speech recognition examples for MFC

Release v1.5.1

• Add MFC examples