search

neulab / RIPPLe

Code for the paper "Weight Poisoning Attacks on Pre-trained Models" (ACL 2020)
BSD 3-Clause "New" or "Revised" License
138 stars 18 forks source link

Question about model path to extract the replacement embedding #2

Closed ziqi-zhang closed 4 years ago

ziqi-zhang commented 4 years ago

Hi Paul,

Thanks for sharing the code and adding so many detailed comments.

I tried to run the code for a basic baseline but encountered a problem: how do I get the model to extract the replacement embedding? Here in the example_manifesto.yaml is the path to load the model but I don't find how to save the model into logs/sst_clean_ref_2.

I'm pretty new to NLP so can you please tell me how to train or download the model?

Thanks, Ziqi

pmichel31415 commented 4 years ago

Hi Ziqi, glad you liked our paper.

this file corresponds to a fine-tuned version of the pre-trained model. This is the model from which the replacement embeddings are extracted. You can obtain this model by running run_glue.py with the appropriate parameters, then copying the resulting file to logs/sst_clean_ref_2 (or updating your manifesto accordingly).

Hope that helps! Feel free to reopen the issue if you still have any problem