alxndrTL / mamba.py

A simple and efficient Mamba implementation in pure PyTorch and MLX.
MIT License
960 stars 86 forks source link

Integration to `transformers` #19

Closed ArthurZucker closed 6 months ago

ArthurZucker commented 6 months ago

Hey! I ported mamba to transformers and think your approach to replace the naive scan would be great there! Would you like to open a PR? 🤗 (to https://github.com/huggingface/transformers)

alxndrTL commented 6 months ago

Hello, thank you for reaching me out, sure I will start working on it !

alxndrTL commented 6 months ago

Hello @ArthurZucker, I just posted the PR to transformers :)

ArthurZucker commented 5 months ago

Thanks! will check it out!