New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Implent two types of rope & Add support to tinyllama-1.1B-chat-v0.2 #408

Closed

magician-blue wants to merge 1 commit into karpathy:master from magician-blue:master

magician-blue commented Sep 26, 2023

I have implented 2 type of ROPE:

falcon style roate the first and second half of the dims
llama style roate the even and odd dims

Support TinyLlama-1.1B-chat-v0.2

orginal model https://huggingface.co/PY007/TinyLlama-1.1B-Chat-v0.2 and github https://github.com/jzhang38/TinyLlama
converted model and tokenizer can be downloaded here
TinyLlama-1.1B is the smallest billion parameter llama architecture model. (however, its rope architecture is falcon's)

I have hardcode the str_lookup part to recognize the '\n' which is <0x0A> in the tokenizer.model.


          Implent two types of rope.

b77aa76

Add support to tinyllama-1.1B-chat-v0.2

magician-blue closed this

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet