Skip to content

Instantly share code, notes, and snippets.

@harrytormey
Created March 13, 2024 20:52
Show Gist options
  • Star 0 You must be signed in to star a gist
  • Fork 0 You must be signed in to fork a gist
  • Save harrytormey/4d3aa3906061b6b9ec149a43f98b2c3d to your computer and use it in GitHub Desktop.
Save harrytormey/4d3aa3906061b6b9ec149a43f98b2c3d to your computer and use it in GitHub Desktop.
How to iterate over data in parquet file from hugging face, example uses SWEBench data.
import pandas as pd
df = pd.read_parquet("./test-00000-of-00001-dc7762b94638c186.parquet")
# Filter the DataFrame for entries from the Django repo
django_entries = df[df["repo"].str.contains("Django", case=False)]
if not django_entries.empty:
for index, row in django_entries.iterrows():
print(row["problem_statement"], "\n")
created_at_datetime = pd.to_datetime(row["created_at"])
created_at_readable = created_at_datetime.strftime("%Y-%m-%d %H:%M:%S")
print(f"Created At: {created_at_readable}")
# Assess the size of the patch
patch_lines = (
row["patch"].count("\n") + 1
) # +1 to count the last line if it doesn't end with a newline
print(f"Patch Size: {patch_lines} lines\n")
else:
print("No entries from the Django repository were found.")
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment