src/json2df.py

import os
import json
import pandas as pd
from models_info import model_info

directory = 'data/raw-eval-outputs'
data = []

def model_hyperlink(link, model_name):
    return f'<a target="_blank" href="{link}" style="color: var(--link-text-color); text-decoration: underline;text-decoration-style: dotted;">{model_name}</a>'

def make_clickable_names(df):
    df["Model"] = df.apply(
        lambda row: model_hyperlink(row["Link"], row["Model"]), axis=1
    )
    return df

# Iterate over all the files in the directory
for filename in os.listdir(directory):
    if filename.endswith(".json"):
        filepath = os.path.join(directory, filename)
        with open(filepath, 'r') as f:
            json_data = json.load(f)
            model_name = filename.replace("_results.json", "")
            
            # Extract the accuracy values
            results = json_data['results']
            row = {'Model': model_name}
            for key, value in results.items():
                row[key] = round(value['acc,none'] * 100, 2)
            
            # Add the tuning type and link to the row
            row['T'] = model_info[model_name]['tuning']
            row['Link'] = model_info[model_name]['link']
            
            data.append(row)


df = pd.DataFrame(data)
df = make_clickable_names(df)
df.drop(columns=["Link"], inplace=True)

df['medmcqa_diff'] = (df['medmcqa_g2b'] - df['medmcqa_orig_filtered']).round(2)
df['medqa_diff'] = (df['medqa_4options_g2b'] - df['medqa_4options_orig_filtered']).round(2)

# Reorder columns
cols = [
    "T", 
    "Model", 
    "b4bqa", 
    "b4b", 
    "medmcqa_g2b", 
    "medmcqa_orig_filtered", 
    "medmcqa_diff", 
    "medqa_4options_g2b", 
    "medqa_4options_orig_filtered", 
    "medqa_diff"
] + [col for col in df.columns if col not in [
    "T", "Model", "b4bqa", "b4b", "medmcqa_g2b", "medmcqa_orig_filtered", "medmcqa_diff", "medqa_4options_g2b", "medqa_4options_orig_filtered", "medqa_diff"
]]
df = df[cols]


output_csv = 'data/csv/models_data.csv'
df.to_csv(output_csv, index=False)

print(f"DataFrame saved to {output_csv}")