HuggingFace Trainer: Custom LLaMA Training
A complete training script for LLaMA (250M–8B) with a custom tokenizer, label masking, gradient checkpointing, and multi-GPU support via Accelerate.
Huggingface
Custom Tokenizer for Audio LLMs
Building a HuggingFace-compatible tokenizer for neural audio codec tokens — vocabulary design, special tokens, and uploading to the Hub.