Fabian Degen comments

Repositories
Issues
Comments

Results 4 comments of


                                            Fabian Degen

[Bug Report] Qwen model implementation is too inaccurate

Hi! After all the einsum removals that were done over the past two weeks, I took a look at this again, and it seems like the logits match very well!...

[Bug Report] The output from HookedTransformer is not identical compared to Huggingface model for Lllama 3

Hi! After all the einsum removals that were done over the past two weeks, I took a look at this again, and it seems like the logits match very well!...

[Bug Report] Gemma-2-2b-it output logit doesn't match with huggingface

Hi! After all the einsum removals that were done over the past two weeks, I took a look at this again, and it seems like the logits match very well!...

[Bug Report] Gemma-2-2b-it output logit doesn't match with huggingface

Hey! How are you loading Llama in? When you load Llama with from_pretrained_no_processing the residual stream values should be very close.