3.0.0-beta.15 (2024-04-04)

Bug Fixes

• create a context with no parameters (#188) (6267778)
• improve chat wrappers tokenization (#182) (35e6f50)
• use the new llama.cpp CUDA flag (#182) (35e6f50)
• adapt to breaking llama.cpp changes (#183) (6b012a6)

Features

• automatically adapt to current free VRAM state (#182) (35e6f50)
• inspect gguf command (#182) (35e6f50)
• inspect measure command (#182) (35e6f50)
• readGgufFileInfo function (#182) (35e6f50)
• GGUF file metadata info on LlamaModel (#182) (35e6f50)
• JinjaTemplateChatWrapper (#182) (35e6f50)
• use the tokenizer.chat_template header from the gguf file when available - use it to find a better specialized chat wrapper or use JinjaTemplateChatWrapper with it as a fallback (#182) (35e6f50)
• simplify generation CLI commands: chat, complete, infill (#182) (35e6f50)
• Windows on Arm prebuilt binary (#181) (f3b7f81)

Shipped with llama.cpp release b2608

To use the latest llama.cpp release available, run npx --no node-llama-cpp download --release latest. (learn more)