brunneis's picture
Add Solbench score
5a908b8 unverified
raw
history blame
5.35 kB
#!/usr/bin/env python
# -*- coding: utf-8 -*-
# flake8: noqa E501
import json
import os
import pandas as pd
from src.display.formatting import has_no_nan_values, make_clickable_model
from src.display.utils import AutoEvalColumn, EvalQueueColumn
from src.leaderboard.read_evals import get_raw_eval_results
from src.utils import get_model_name_from_filepath
def get_leaderboard_df(results_path: str, requests_path: str, cols: list, benchmark_cols: list) -> pd.DataFrame:
"""Creates a dataframe from all the individual experiment results"""
raw_data = get_raw_eval_results(results_path, requests_path)
all_data_json = [v.to_dict() for v in raw_data]
df = pd.DataFrame.from_records(all_data_json)
# df = df.sort_values(by=[AutoEvalColumn.average.name], ascending=False)
df = df.sort_values(by=[AutoEvalColumn.solbench.name], ascending=False)
df = df[cols].round(decimals=2)
# filter out if any of the benchmarks have not been produced
df = df[has_no_nan_values(df, benchmark_cols)]
return df
# def get_evaluation_requests_df(save_path: str, cols: list) -> list[pd.DataFrame]:
# """Creates the different dataframes for the evaluation requestss requestes"""
# 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)
# try:
# with open(file_path, encoding='utf-8') as fp:
# data = json.load(fp)
# except UnicodeDecodeError as e:
# print(f"Unicode decoding error in {file_path}: {e}")
# continue
# # data[EvalQueueColumn.model.name] = make_clickable_model(data["model"])
# model_name = get_model_name_from_filepath(file_path)
# data[EvalQueueColumn.model.name] = make_clickable_model(model_name)
# data[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)
# try:
# with open(file_path, encoding='utf-8') as fp:
# data = json.load(fp)
# except json.JSONDecodeError:
# print(f"Error reading {file_path}")
# continue
# except UnicodeDecodeError as e:
# print(f"Unicode decoding error in {file_path}: {e}")
# continue
# # data[EvalQueueColumn.model.name] = make_clickable_model(data["model"])
# model_name = get_model_name_from_filepath(file_path)
# data[EvalQueueColumn.model.name] = make_clickable_model(model_name)
# data[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]
def get_evaluation_requests_df(save_path: str, cols: list) -> list[pd.DataFrame]:
"""Creates the different dataframes for the evaluation requestss requested."""
all_evals = []
def process_file(file_path):
try:
with open(file_path, 'r', encoding='utf-8') as fp:
data = json.load(fp)
except (json.JSONDecodeError, UnicodeDecodeError) as e:
print(f"Error reading or decoding {file_path}: {e}")
return None
model_name = get_model_name_from_filepath(file_path)
# data[EvalQueueColumn.model.name] = make_clickable_model(data["model"])
data[EvalQueueColumn.model.name] = make_clickable_model(model_name)
data[EvalQueueColumn.revision.name] = data.get("revision", "main")
return data
for root, _, files in os.walk(save_path):
for file in files:
if file.endswith('.json'):
file_path = os.path.join(root, file)
data = process_file(file_path)
if data:
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]