Olivier Chafik
Olivier Chafik
Refactoring of chat / tool-call logic (follow up to https://github.com/ggml-org/llama.cpp/pull/11016) along the lines of @ggerganov's suggestions ([ref](https://github.com/ggml-org/llama.cpp/pull/11016#issuecomment-2660837676)): - Move all `common_chat_*` from `common.*` to `common/chat.*` - Moved minja headers to...
Allows specifying a JSON schema by file (currently only flag is `-j` / `--json-schema` which takes the full schema itself as argument)
**TL;DR: fixes tool calling of Qwen 2.5 Coder 0.5B/1.5B/3B/7B/... at any temperature** - Added support for regex grammar triggers, and respect when they should be matching at the start only...