Open Issues Need Help
View All on GitHubAI Summary: This issue requires refactoring the `neural_audio_tokenizer.py` script to centralize version number management and improve the behavior of the command-line verbose flag. The version number should be defined in a single constant and used throughout the script, with micro-versions in comments or less critical areas replaced by 'x'. A new logging system needs to be implemented and integrated to provide proper output for the verbose mode.
Tim's Ears - Music and sound reasoning data for LLMs to hear audio via tokens and data files generated from processing any types of files
AI Summary: This issue addresses a bug in the neural audio tokenizer where codebooks are not being correctly serialized, saved, and loaded. The goal is to enable the reuse of initialized codebooks across program invocations, with an option to force re-initialization via a command-line flag. This fix is crucial before tagging the v0.1.7 release.
Tim's Ears - Music and sound reasoning data for LLMs to hear audio via tokens and data files generated from processing any types of files
AI Summary: This issue requests the implementation of inline comments found within the `neural_audio_tokenizer.py` file. The author is seeking to incorporate specific considerations and potential changes suggested in those comments, as well as any other necessary or beneficial modifications identified by the developer.
Tim's Ears - Music and sound reasoning data for LLMs to hear audio via tokens and data files generated from processing any types of files
AI Summary: This issue requires a review and refactoring of the `neural_audio_tokenizer.py` script to improve code consistency and logic. Specific updates are needed for optional Encodec sections to enhance token diversity, moving beyond simple jittering, and general cleanup/modernization of command-line arguments is also requested.
Tim's Ears - Music and sound reasoning data for LLMs to hear audio via tokens and data files generated from processing any types of files