Step 1 - Set Environment

Create a Virtual Environment

python -m venv .venv

Activate

Then dpeending on your os and environment, activate the virtual environment.
For example, from shell terminal:

source .venv/bin/activate.csh

Install Libraries

pip install -r requirements.txt

Run scripts

phi

python phi.py {args}

use the following args:

--model             Choose between model variants: "phi-2" for the smaller, lightweight model or "phi-3.5" for the slightly larger, more competitive model. Default is "phi-3.5".
--prompt            Specify the type of prompting method: "zero_shots" for no examples or "few_shots" for examples in the prompt. Default is "few_shots".
--prompt-version    Choose the prompt version: "basic" for simple prompts or "comprehensive" for more detailed ones. Default is "comprehensive".
--output-dir        Specify the directory where the results will be saved. Default is "results/".
--run-all           Use all choices for model, prompt, and prompt-version```

Note that you can choose several model, prompts and prompt versions, for example python phi.py --model phi-2 phi-3.5 --prompt zero_shots few_shots To run all the configurations possible, use the --run-all flag: python phi.py --run-all

results are automatically stored in the results folder

TweetEval

To evaluate TweetEval and visualize the plots, run:

python tweet_eval.py

TODO before submitting

• remove the cache variable