Spaces:
Running
Running
File size: 2,684 Bytes
ee31436 56bf4e8 ee31436 18716bc 56bf4e8 ad27ecb ee31436 18716bc ad27ecb ee31436 ad27ecb 56bf4e8 e157bd5 ee31436 ad27ecb ee31436 3abeb28 56bf4e8 ee31436 56bf4e8 ee31436 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 |
import json
import os
import pandas as pd
import src.display.formatting as formatting
import src.display.utils as utils
import src.leaderboard.read_evals as read_evals
def get_leaderboard_df(results_path: str, requests_path: str, cols: list, benchmark_cols: list) -> pd.DataFrame:
print(results_path, requests_path)
raw_data = read_evals.get_raw_eval_results(results_path, requests_path)
#print("raw_data:",raw_data)
all_data_json = [v.to_dict() for v in raw_data]
#print(all_data_json)
df = pd.DataFrame.from_records(all_data_json)
print("all results:",df)
# exit()
try:
df = df.sort_values(by=[utils.AutoEvalColumn.hallucination_rate.name], ascending=True)
df = df[cols].round(decimals=2)
# filter out if any of the benchmarks have not been produced
df = df[formatting.has_no_nan_values(df, benchmark_cols)]
except:
pass
return df
def get_evaluation_queue_df(save_path: str, cols: list) -> list[pd.DataFrame]:
entries = [entry for entry in os.listdir(save_path) if not entry.startswith(".")]
all_evals = []
for entry in entries:
if ".json" in entry:
file_path = os.path.join(save_path, entry)
with open(file_path) as fp:
data = json.load(fp)
data[utils.EvalQueueColumn.model.name] = data["model"] #formatting.make_clickable_model(data["model"])
data[utils.EvalQueueColumn.revision.name] = data.get("revision", "main")
all_evals.append(data)
elif ".md" not in entry:
# this is a folder
sub_entries = [e for e in os.listdir(f"{save_path}/{entry}") if not e.startswith(".")]
for sub_entry in sub_entries:
file_path = os.path.join(save_path, entry, sub_entry)
with open(file_path) as fp:
data = json.load(fp)
data[utils.EvalQueueColumn.model.name] = formatting.make_clickable_model(data["model"])
data[utils.EvalQueueColumn.revision.name] = data.get("revision", "main")
all_evals.append(data)
pending_list = [e for e in all_evals if e["status"] in ["PENDING", "RERUN"]]
running_list = [e for e in all_evals if e["status"] == "RUNNING"]
finished_list = [e for e in all_evals if e["status"].startswith("FINISHED") or e["status"] == "PENDING_NEW_EVAL"]
df_pending = pd.DataFrame.from_records(pending_list, columns=cols)
df_running = pd.DataFrame.from_records(running_list, columns=cols)
df_finished = pd.DataFrame.from_records(finished_list, columns=cols)
return df_finished[cols], df_running[cols], df_pending[cols]
|