oxibonsai-core
GGUF Q1_0_g128 parser, tensor types, and configuration for OxiBonsai.
Provides the foundational data types and I/O layer: GGUF file loading (v1/v2/v3), Q1_0_g128 block tensor deserialization, Qwen3 model configuration, streaming GGUF parser, GGUF writer, model card generation, and all shared error types.
Part of the OxiBonsai project.
Features
- GGUF v1/v2/v3 reader with forward-compatibility layer
GgufStreamParser— state-machine streaming parser for network-loaded modelsGgufWriter— produce valid GGUF byte streams with metadata and tensorsQwen3Config— model configuration for Bonsai-8B, 4B, and 1.7B variantsBlockQ1_0G128/OneBitTensor— Q1_0_g128 block tensor typesModelCard— structured model card (author, license, tags) embedded in GGUF- K-quant formats:
BlockQ2K,BlockQ4K - Memory budget estimation and tensor metadata lazy loading
mmapfeature for zero-copy model file access
Feature Flags
| Flag | Description | Default |
|---|---|---|
mmap |
Memory-mapped file access via memmap2 |
enabled |
wasm |
WASM-safe builds (no memmap2) |
disabled |
Usage
[]
= "0.1.0"
License
Apache-2.0 — COOLJAPAN OU