Update README.md

update tokenizer

use torch.nn.functional.gelu instead

Update README.md

fix forward

fix: missing metadata

fuse qkv

Update modeling_gpt2_mq.py

init

initial commit