Change8

b8102

📦 llama-cppView on GitHub →
2 features🐛 3 fixes🔧 2 symbols

Summary

This release introduces support for the LFM2.5-Audio-1.5B tokenizer and includes several internal code improvements and fixes related to attention layers and model conversion.

Migration Steps

  1. To convert the LFM2.5-Audio-1.5B tokenizer, use the command: python3 convert_hf_to_gguf.py /path/to/LFM2.5-Audio-1.5B/audio_detokenizer

✨ New Features

  • Added support for the tokenizer from LFM2.5-Audio-1.5B, which acts as an embedding model with different input and output embedding sizes.
  • Added support for LFM2 SWA model.

🐛 Bug Fixes

  • Reworked check for attention layers.
  • Set vocab to none.
  • Moved helper function definitions to cpp file.

Affected Symbols