Module gguf 

GGUF Export (spec §7.2)

Pure Rust writer for GGUF format (llama.cpp compatible). WASM compatible - no C/C++ dependencies.

Format Structure

┌─────────────────────────────────────────┐
│ Magic: "GGUF" (4 bytes)                 │
│ Version: u32 (currently 3)              │
│ Tensor count: u64                       │
│ Metadata KV count: u64                  │
├─────────────────────────────────────────┤
│ Metadata KV pairs                       │
├─────────────────────────────────────────┤
│ Tensor info array                       │
├─────────────────────────────────────────┤
│ Tensor data (aligned)                   │
└─────────────────────────────────────────┘

Reference: [GGUF2023] Gerganov, G. (2023). GGUF Format.

Structs

GgufHeader

GGUF file header

GgufTensor

A tensor to be exported to GGUF format

GgufTensorInfo

GGUF tensor info

Enums

GgmlType

GGUF tensor types (from ggml)

GgufValue

GGUF metadata value

GgufValueType

GGUF value types (from ggml)

Constants

GGUF_DEFAULT_ALIGNMENT

Default alignment for tensor data

GGUF_MAGIC

GGUF magic number: “GGUF” in little-endian

GGUF_VERSION

GGUF format version (v3 is current)

Functions

export_tensors_to_gguf

Export tensors to GGUF format

padding_for_alignment

Calculate padding bytes needed for alignment

write_metadata_kv

Write a metadata key-value pair