LLM-Research

Begin by ensuring you have the appropriate packages installed (run pip install -r requirements.txt).

Modify the files as follows:

LIME

• generate_lime_expl.py
  • adjust prompts (PE, EP)
  • change model to desired
  • ensure sentences.pickle exists (if not, first run generate_model_expl.py. Optionally can comment out the model generation code to be left with sentence.pickle)
  • adjust batch size as well as start and end (sentence index)
• process.py
  • change prompt to same as generate_lime_expl.py
  • change response filename to the one in generate_lime_expl.py
  • ensure accuracy is commented out
  • set PE to the appropriate value based on prompt
  • Rename print statements/variables based on the current model

Model Explanations:

• generate_model_expl.py
  • adjust prompts (PE, EP)
  • change model to desired
  • change response and label filenames (line 82 & 84)
• process.py
  • change prompt to same as generate_model_expl.py
  • change response filename to the one in generate_model_expl.py
  • ensure accuracy is not commented
  • set PE to the appropriate value based on prompt
  • Rename print statements/variables based on the current model