search

OpenMOSS / Language-Model-SAEs

For OpenMOSS Mechanistic Interpretability Team's Sparse Autoencoder (SAE) research.
21 stars 3 forks source link

[Proposal] Improve lm_config.json #22

Open Hzfinfdu opened 1 week ago

Hzfinfdu commented 1 week ago

What about doing all stuff with relative path? e.g. saving model path in lm_config.json.

I will revisit this later

dest1n1s commented 1 week ago

Currently LanguageModelConfig accept model_name and model_from_pretrained_path. Do you mean merging these two into one field?

Hzfinfdu commented 1 week ago

Sry for the confusion. I meant to save model_from_pretrained_path as relative path instead of absolute one.