parse_requests.py

import os
from datasets import load_dataset, Dataset

skip_statuses = ['COMPLETED', 'FAILED', 'RUNNING']
TOKEN = os.environ.get("DEBUG")
requests_dataset = load_dataset("AIEnergyScore/requests_debug", split="test")

def normalize_task(task):
    # Makes assumption about how the task names are being written, and called.
    return '_'.join(task.split()).lower()


requests_dset = requests_dataset.to_pandas()

for model, task in requests_dset[['model','task']].loc[requests_dset['status'] == 'PENDING'].to_dict(orient= 'split', index=False)['data']:
    print("%s,%s" % (model, normalize_task(task)))

# Custom errors we will rerun.
for model, task in requests_dset[['model','task']].loc[~requests_dset['status'].isin(skip_statuses)].to_dict(orient= 'split', index=False)['data']:
    print("%s,%s" % (model, normalize_task(task)))