Papers with Code Model Index

Warnings

• gotcha The package name on PyPI is `model-index` (with a hyphen). An alias package `modelindex` (without a hyphen) exists but primarily serves as a dependency for `model-index`. Always use `pip install model-index` to ensure you get the correct, maintained library.
  Fix: Ensure installation is `pip install model-index` and not `pip install modelindex`.
• gotcha The `model-index` library relies heavily on the correct structure and syntax of `model-index.yml` and referenced metadata files (YAML, JSON, Markdown). Incorrect formatting, broken relative paths, or non-adherence to the Papers with Code schema can lead to parsing errors or validation failures.
  Fix: Carefully review the official `model-index` documentation and examples for required YAML schema and file linking conventions. Utilize the `mi` command-line tool (if available) for validation.
• gotcha While a Python API exists for programmatic interaction (e.g., loading the index), much of the official documentation and usage examples emphasize defining metadata via `model-index.yml` files and processing them with the provided command-line interface (CLI) tool (`mi`). Explicit Python API quickstart examples for building an index programmatically from scratch are less prevalent.
  Fix: For basic usage, consider starting with the `model-index.yml` file approach and CLI tools. If programmatic generation or complex manipulation is needed, consult the library's source code or advanced documentation for API specifics.

Install

• pip install model-index Install latest version

Imports

• ModelIndex

  from modelindex.models import ModelIndex

  Based on common Python package structure for data models. Explicit `models` submodule is standard for classes representing data schemas.

Quickstart

This quickstart demonstrates how to define a `model-index.yml` file and then programmatically load and inspect its contents using the `model-index` Python library. It highlights how the library facilitates reading structured model metadata into Python objects.

import os
from pathlib import Path
from modelindex.models import ModelIndex

# 1. Create a dummy model-index.yml file for demonstration
model_index_content = '''
Models:
  - Name: MyAwesomeModelV1
    Metadata:
      Task: Image Classification
      Dataset: CIFAR-10
      Metrics:
        Accuracy: 0.92
      Paper: https://arxiv.org/abs/2301.00001
      Code: https://github.com/myuser/myawesomemodel
      Weights: https://example.com/weights.pth
      License: MIT
  - Name: MyAwesomeModelV2
    Metadata:
      Task: Image Classification
      Dataset: CIFAR-10
      Metrics:
        Accuracy: 0.94
      Paper: https://arxiv.org/abs/2302.00002
      Code: https://github.com/myuser/myawesomemodel
      Weights: https://example.com/weights_v2.pth
      License: Apache-2.0
'''

index_file_path = "model-index.yml"
Path(index_file_path).write_text(model_index_content)

print(f"Created dummy {index_file_path}")

try:
    # 2. Load the model index from the YAML file
    # Note: The exact loading method might vary; .load_from_file() is a common pattern.
    # The library's core is parsing these files.
    model_index = ModelIndex.load_from_file(index_file_path)

    print(f"\nSuccessfully loaded {len(model_index.models)} model(s).")

    # 3. Access data from the loaded index
    for i, model in enumerate(model_index.models):
        print(f"\n--- Model {i+1} ---")
        print(f"Name: {model.name}")
        print(f"Task: {model.metadata['Task']}")
        print(f"Accuracy: {model.metadata['Metrics']['Accuracy']}")
        print(f"Paper: {model.metadata['Paper']}")

except Exception as e:
    print(f"An error occurred during loading or processing: {e}")
finally:
    # Clean up the dummy file
    if Path(index_file_path).exists():
        os.remove(index_file_path)
        print(f"\nCleaned up {index_file_path}")