微笑fla

It's a wonderful work and as a beginner I appreciate for being able to learn a lot from it. However, I have no access to the DIHARD datasets, so I...

Instruction-FT chatGLM with LORA and Chinese self-instruct data

1

We have tried fine-tuning chatGLM , an awesome dialogue model aligned with human preference, with LORA. I think it indeed learns from new datas. https://github.com/27182812/ChatGLM-chinese-insturct

fix bug auto loading llamatokenizer

2

# What does this PR do? Huggingface decapoda-research/llama-7b-hf config decide the name of tokenizer LLaMATokenizer, while in transformers it is LlamaTokenizer. Unify the name as LLaMATokenizer, so that we can...