Welcome to DeepMTP’s documentation!

DeepMTP is a Python package implementing a flexible two-branch neural network architecture that makes it compatible with the majority of multi-target prediction problems.

Documentation

Release:

0.0.24.dev15+gf823f3ee0

Date:

Jul 31, 2026

Contents:

• Installation
  • Requirements
  • Install from PyPI
  • Optional integrations
  • Install from source
• Migrating from DeepMTP 0.0.22 to 0.0.23
  • Before upgrading
  • Python and dependencies
  • Imports
  • Configuration
  • Data preparation
  • Prediction and training lifecycle
  • Checkpoints
  • Optional integrations and offline runs
  • Verification checklist
• Loading a dataset
  • Loading a built-in dataset
  • Loading a custom dataset
• Configuration options
  • General configuration
  • Gradient accumulation and clipping
  • Instance and target branch hyperparameters
• Architecture visualization
  • Python API
  • Parameter statistics
  • Command line
  • SVG export
• Reproducible checkpoints
  • Foundation adapter format
  • Dense preprocessing replay
  • Tabular preprocessing
  • Loading and compatibility
• Weights & Biases tracking
  • Run organization
  • Artifacts and lineage
  • Prediction Tables and charts
  • Restoring a model Artifact
  • Privacy-conscious defaults
• Multiclass classification
  • Label contract
  • Model configuration
  • Metric averaging
  • Prediction output
  • Compatibility and current limits
• Mixed tabular inputs
  • Input format
  • Configuration
  • Preprocessing and checkpoints
  • Precomputed embeddings
• Sparse high-dimensional inputs
  • Supported inputs
  • Configuration
  • Memory and preprocessing
  • Benchmarking
• Token-sequence inputs
  • Input format
  • Configuration
  • Structured batches
  • Compatibility and checkpoints
• Foundation-model sequence encoders
  • Supported model contract
  • Tokenizing raw values
  • Length policy
  • Configuring the branch
  • Pooling
  • Training modes
  • LoRA configuration
  • Compact adapter checkpoints
  • Composite inputs
  • Reproducibility and checkpoints
  • Security and offline use
• Molecular graph inputs
  • Installation
  • Input contract
  • Data preparation
  • Branch configuration
  • Batch contract
  • Composite graph components
  • Scope
• Composite branch inputs
  • Configuration
  • Component fusion
  • Optional components
  • Modality dropout
  • Supported components
  • Batch and encoder contract
  • Compatibility and current limits
• Hyperparameter Optimization
  • Hyperband
  • Combining Hyperband with DeepMTP
• Demo notebooks
• Project policies
• API reference
  • DeepMTP package
• Credits

Indices and tables

• Index

• Module Index

• Search Page