Modules§
- gguf
- GGUF reader (little-endian). Loads dense tensors and Q4_0 / Q8_0 GGML blocks expanded to
f32payloads in IR. Other quant types still error with a descriptive message. See GGUF spec. - onnx
- pytorch
- PyTorch checkpoints are pickled (often inside a ZIP). Practical Rust support without an embedded Python interpreter is limited, so we accept:
- safetensors