Skip to content

paritea/LlmClassifier

Repository files navigation

LlmClassifier

CSE-584 midterms

Directory structure and further details :

  • The code for making the xi's along with the code calling the LLM pipelines via hugging-face is in the CodeForDatasetCreation folder.
  • The completions for each LLM are kept in the datasets_separated folder.
  • The classifier code is in the Bert Classifier.ipynb file. Unfortunately, the lime generated images aren't visible in the jupyter notebook itself. Check out the MLMidtermReport pdf to see the images!
  • Finally, the report details the entire process of building the classifier!

Steps for reproducing the results :

  1. Run the classifier file in kaggle to not deal with any additional package installations as pytorch, transformers and other libraries used are pre-installed there.
  2. To run the file, you will have to upload or use (if you choose to do so locally) the separated datasets instead of the combined one.

Hardware requirements :

I used the gpu freely available on kaggle for this. It will take a long time if run just on a cpu.

About

LLM Classifier

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published