push_results.py

importglob
importsys

importpandasaspd
fromhuggingface_hubimporthf_hub_download, upload_file
fromhuggingface_hub.utilsimportEntryNotFoundError


sys.path.append(".")
fromutilsimportBASE_PATH, FINAL_CSV_FILE, GITHUB_SHA, REPO_ID, collate_csv# noqa: E402


defhas_previous_benchmark() ->str:
csv_path=None
try:
csv_path=hf_hub_download(repo_id=REPO_ID, repo_type="dataset", filename=FINAL_CSV_FILE)
exceptEntryNotFoundError:
csv_path=None
returncsv_path


deffilter_float(value):
ifisinstance(value, str):
returnfloat(value.split()[0])
returnvalue


defpush_to_hf_dataset():
all_csvs=sorted(glob.glob(f"{BASE_PATH}/*.csv"))
collate_csv(all_csvs, FINAL_CSV_FILE)

# If there's an existing benchmark file, we should report the changes.
csv_path=has_previous_benchmark()
ifcsv_pathisnotNone:
current_results=pd.read_csv(FINAL_CSV_FILE)
previous_results=pd.read_csv(csv_path)

numeric_columns=current_results.select_dtypes(include=["float64", "int64"]).columns
numeric_columns= [
cforcinnumeric_columnsifcnotin ["batch_size", "num_inference_steps", "actual_gpu_memory (gbs)"]
 ]

forcolumninnumeric_columns:
previous_results[column] =previous_results[column].map(lambdax: filter_float(x))

# Calculate the percentage change
current_results[column] =current_results[column].astype(float)
previous_results[column] =previous_results[column].astype(float)
percent_change= ((current_results[column] -previous_results[column]) /previous_results[column]) *100

# Format the values with '+' or '-' sign and append to original values
current_results[column] =current_results[column].map(str) +percent_change.map(
lambdax: f" ({'+'ifx>0else''}{x:.2f}%)"
 )
# There might be newly added rows. So, filter out the NaNs.
current_results[column] =current_results[column].map(lambdax: x.replace(" (nan%)", ""))

# Overwrite the current result file.
current_results.to_csv(FINAL_CSV_FILE, index=False)

commit_message=f"upload from sha: {GITHUB_SHA}"ifGITHUB_SHAisnotNoneelse"upload benchmark results"
upload_file(
repo_id=REPO_ID,
path_in_repo=FINAL_CSV_FILE,
path_or_fileobj=FINAL_CSV_FILE,
repo_type="dataset",
commit_message=commit_message,
 )


if__name__=="__main__":
push_to_hf_dataset()