This repo is a tiny language model mimicking LLMs. I would like to create LLM-like model from scratch. To some degree LLM could also be abbreviated to little language model. :D
- Collect Data
- Clean & Augment Data
- Make Tokenizer
- Make Model Based on Transformer
- Pretrain Model
- LLM Alignment
- Transfer to downstream tasks.
- SFT Training
- Distribute Data Parallel
- Distill model
- RAG
- Agent