Skip to content

resource_listing.md

Latest commit

 

History

History
101 lines (69 loc) · 3.43 KB

resource_listing.md

File metadata and controls

101 lines (69 loc) · 3.43 KB

Listing Supported Models and Datasets

Protify provides several ways to view and explore the supported models and datasets. This documentation explains how to use these features.

Using the README Toggle Sections

The main README.md file contains expandable toggle sections for both models and datasets:

  • Currently Supported Models: Click the toggle to expand and see a complete table of models with their descriptions, sizes, and types.
  • Currently Supported Datasets: Click the toggle to expand and see a complete table of datasets with their descriptions, types, and tasks.

Command-Line Listing

Protify provides command-line utilities for listing models and datasets with detailed information:

Listing Models

To list all supported models with their descriptions:

# List all supported models
python -m src.protify.base_models.get_base_models --list

# To download standard models
python -m src.protify.base_models.get_base_models --download

Listing Datasets

To list all supported datasets with their descriptions:

# List all datasets
python -m src.protify.data.dataset_utils --list

# Get information about a specific dataset
python -m src.protify.data.dataset_utils --info EC

Combined Listing

For a combined view of both models and datasets:

# List both models and datasets
python -m src.protify.resource_info --all

# List only standard models and datasets
python -m src.protify.resource_info --all --standard-only

# List only models
python -m src.protify.resource_info --models

# List only datasets
python -m src.protify.resource_info --datasets

Programmatic Access

You can also access model and dataset information programmatically:

# For models
from src.protify.resource_info import model_descriptions
from src.protify.base_models.get_base_models import currently_supported_models, standard_models

# Get information about a specific model
model_info = model_descriptions.get('ESM2-150', {})
print(f"Model: ESM2-150")
print(f"Description: {model_info.get('description', 'N/A')}")
print(f"Size: {model_info.get('size', 'N/A')}")
print(f"Type: {model_info.get('type', 'N/A')}")

# For datasets
from src.protify.resource_info import dataset_descriptions
from src.protify.data.supported_datasets import supported_datasets

# Get information about a specific dataset
dataset_info = dataset_descriptions.get('EC', {})
print(f"Dataset: EC")
print(f"Description: {dataset_info.get('description', 'N/A')}")
print(f"Type: {dataset_info.get('type', 'N/A')}")
print(f"Task: {dataset_info.get('task', 'N/A')}")

Model Group Types

Models in Protify are generally grouped into the following categories:

  1. Protein Language Models: Pre-trained models that have learned protein properties from large-scale sequence data (e.g., ESM2, ProtBert)
  2. Baseline Controls: Models with random weights for comparison (e.g., Random, Random-Transformer)

Dataset Group Types

Datasets are categorized by their task types:

  1. Multi-label Classification: Datasets where each protein can have multiple labels (e.g., EC, GO-CC)
  2. Classification: Binary or multi-class classification tasks (e.g., DeepLoc-2, DeepLoc-10)
  3. Regression: Prediction of continuous values (e.g., enzyme-kcat, optimal-temperature)
  4. Protein-Protein Interaction: Tasks focused on protein interactions (e.g., human-ppi, gold-ppi)
  5. Token-wise Classification/Regression: Residue-level prediction tasks (e.g., SecondaryStructure-3)