CLTL Team @ ArgMining 2022 Shared Task
See description of the task here.
We are a team of awesome CLTL PhDs:
Installation
Install packages with pip install -r requirements.txt
. This file may or may not be complete/up-to-date..
Wandb integration
See here on how to set it up. For convenience, I'm using python-dotenv for creating the environment variable for the wandb project name
- Create a
.env
file in project root. - Create following variable e.g.
WANDB_PROJECT=argmining2022-mtl
Usage
GPT3 models
First, please create an account with OpenAI. Get your API key and set it as an environmental variable. For running the best approach using GPT3 (for example on the test set), run:
python prompting.py --n_shot 4 --prompt_style 5
To evaluate a file with predictions, you can do:
from baseline import evaluate_from_file
evaluate_from_file("predictions/dump_prompt_5_results.json")
MTL training
See the arguments in train_mtl.py
.