Change8

b8196

📦 llama-cppView on GitHub →
🐛 1 fixes🔧 1 symbols

Summary

This release fixes a bug in tensor shape formatting to correctly handle large vocabulary sizes, such as those found in Gemma 3, and provides updated pre-compiled binaries for multiple platforms.

🐛 Bug Fixes

  • Fixed an issue where tensor shapes were incorrectly formatted using only 5 digits, which failed for models with vocabulary sizes exceeding 99,999 (e.g., Gemma 3's 262,208 vocab size).

Affected Symbols