failed_run.py

import argparse
import os

from datasets import load_dataset, Dataset
from huggingface_hub import HfApi


TOKEN = os.environ.get("DEBUG")
api = HfApi(token=TOKEN)

parser = argparse.ArgumentParser()
parser.add_argument(
    "--run_dir",
    default=None,
    type=str,
    required=True,
    help="Path to the run directory.",
)
parser.add_argument(
    "--model_name",
    default=None,
    type=str,
    required=True,
    help="Model to benchmark.",
)
args = parser.parse_args()

# Updating request
dataset = load_dataset("EnergyStarAI/requests_debug", split="test", token=TOKEN).to_pandas()

# Set benchmark to failed
dataset.loc[dataset["model"].isin(args.model_name), ['status']] = "FAILED"

# Read error message
with open(f"{args.run_dir}/error.log", 'r') as file:
    error_message = file.read()

# Add a new column for the error message if necessary
if "error_message" not in dataset.columns:
    dataset["error_message"] = ""
dataset.loc[dataset["model"].isin(args.model_name), ['error_message']] = error_message

updated_dataset = Dataset.from_pandas(dataset)
updated_dataset.push_to_hub("EnergyStarAI/requests_debug", split="test", token=TOKEN)

print("Status set to FAILED")