scripts/readme.py

import os
import json
import pandas as pd
from tabulate import tabulate
import typer

def create_readme_for_model(model_dir: str, project_url: str):
    # Path to the metrics and meta files
    metrics_file = os.path.join(model_dir, 'metrics.json')
    meta_file = os.path.join(model_dir, 'model-best', 'meta.json')

    # DataFrame for the model's overall performance metrics
    overall_df = pd.DataFrame(columns=['Metric', 'Value'])

    # DataFrame for the model's per-label performance metrics
    per_label_df = pd.DataFrame(columns=['Label', 'Precision', 'Recall', 'F-Score'])

    # Read and add metrics data
    if os.path.exists(metrics_file):
        with open(metrics_file, 'r') as file:
            metrics = json.load(file)
            overall_df = overall_df.append({'Metric': 'Precision', 'Value': round(metrics['spans_sc_p'] * 100, 1)}, ignore_index=True)
            overall_df = overall_df.append({'Metric': 'Recall', 'Value': round(metrics['spans_sc_r'] * 100, 1)}, ignore_index=True)
            overall_df = overall_df.append({'Metric': 'F-Score', 'Value': round(metrics['spans_sc_f'] * 100, 1)}, ignore_index=True)

            # Extract and add per-type metrics
            for label, scores in metrics.get('spans_sc_per_type', {}).items():
                per_label_df = per_label_df.append({
                    'Label': label,
                    'Precision': round(scores['p'] * 100, 1),
                    'Recall': round(scores['r'] * 100, 1),
                    'F-Score': round(scores['f'] * 100, 1)
                }, ignore_index=True)

    # Sort the per_label_df by Label
    per_label_df.sort_values(by='Label', inplace=True)

    # Convert the DataFrames to Markdown tables
    overall_markdown = tabulate(overall_df, headers='keys', tablefmt='pipe', showindex=False)
    per_label_markdown = tabulate(per_label_df, headers='keys', tablefmt='pipe', showindex=False)

    # Read meta.json file
    meta_info = ""
    if os.path.exists(meta_file):
        with open(meta_file, 'r') as file:
            meta_data = json.load(file)
            for key, value in meta_data.items():
                meta_info += f"- **{key}**: {value}\n"

    # README content
    readme_content = f"""
# Placing the Holocaust spaCy Model - {os.path.basename(model_dir).capitalize()}

This is a spaCy model trained as part of the placingholocaust spaCy project. Training and evaluation code, along with the dataset, can be found at the following URL: [Placingholocaust SpaCy Project]({project_url})

## Model Performance
{overall_markdown}

## Performance per Label
{per_label_markdown}

## Meta Information
{meta_info}
"""

    # Write the README content to a file
    readme_file = os.path.join(model_dir, 'README.md')
    with open(readme_file, 'w') as file:
        file.write(readme_content)

    print(f"README created in {model_dir}")

def create_all_readmes(project_url: str):
    # Directories for each model type
    model_dirs = ['training/sm', 'training/md', 'training/lg', 'training/trf']

    for dir in model_dirs:
        create_readme_for_model(dir, project_url)

if __name__ == "__main__":
    project_url = "https://huggingface.co/datasets/placingholocaust/spacy-project"
    typer.run(lambda: create_all_readmes(project_url))