Code
How openai/whisper is laid out, and where to start for common changes.
There is no import map for this report yet.
Where to start reading
- Repository root
- Repository root
- Likely public API
- Likely public API
- Likely start-up file
- Suggested starting pointAI
- Suggested starting pointAI
- Suggested starting pointAI
Find files for a task
AIAdd support for a new language's text normalization
- Import and register the normalizer in the normalizers package.
- Update transcribe.py to select the normalizer based on detected or specified language.
- Add test cases for the new language's normalization rules.
Implement a custom decoding strategy
- Add a new decoding class inheriting from the base decoder pattern in decoding.py.
- Register the strategy as a callable option in the model's decode method.
- Update transcribe.py to accept the new strategy as a parameter.
Add a new inference optimization
- Implement kernel or optimization logic, optionally using Triton.
- Integrate the optimization into the model's forward pass or decoding loop.
- Benchmark and document timing impact using timing utilities.
Change the public API
Trace runtime behavior
Add or update tests
Change build or automation
RepoPilot lists up to 500 files, so a file missing here may still exist.
- Files marked “likely” are guessed from their names.