rust-ml-huggingface-porting

What problem does it solve?

Port HuggingFace models to Candle with correct tensor naming to prevent weight-loading and matmul errors during integration.

Core Features & Use Cases

• Pattern catalog: Catalogs common tensor name patterns across HuggingFace families to identify naming mismatches.
• Prefix-detection: Detects presence or absence of model prefix (model.) and adjusts paths accordingly.
• Contiguity fixes: Guides how to enforce tensor contiguity before matmul to avoid runtime errors.
• Use Case: Debugs weight loading in safetensors when converting models to Candle.

Quick Start

Use this skill to inspect a HuggingFace safetensors file and apply correct naming patterns to ensure Candle loads weights without error.