• MQA implementation
• Ops refactorings and optimizations
• Bugfixes
• Model exporting script (util/convert_weights.py)

Important Note: With the MQA implementation, older 2B model artifacts need to be updated. Please re-download weights from Kaggle and ensure you have the latest version (-mqa or version 3).

What's Changed

• Clean up docs for developers by @austinvhuang in https://github.com/google/gemma.cpp/pull/102
• MQA Implementation for 2B models by @ufownl in https://github.com/google/gemma.cpp/pull/114
• Enhancing Utility Functions in ops.h by @enum-class in https://github.com/google/gemma.cpp/pull/105
• Added a missing space in app.h by @villesundell in https://github.com/google/gemma.cpp/pull/115
• Fix compilation error when HWY_COMPILER_GCC_ACTUAL < 1300 by @ufownl in https://github.com/google/gemma.cpp/pull/120
• .bazelversion: Bazel 7.1.1 by @LINKIWI in https://github.com/google/gemma.cpp/pull/122
• Add standalone tool to compress weights. by @szabadka in https://github.com/google/gemma.cpp/pull/125
• 1.07x speedup: merge MQA parallel sections as suggested by @veluca93 by @copybara-service in https://github.com/google/gemma.cpp/pull/126