skorch-dev/skorch

Features

• Training Lifecycle Management - Manages the end-to-end training lifecycle using early stopping, checkpointing, and learning rate scheduling.
• Scikit-learn Wrappers - Provides a scikit-learn compatible interface for PyTorch neural networks, enabling integration into standard machine learning pipelines.
• Deep Learning Training Toolsets - Provides a training management toolkit with integrated checkpointing, early stopping, and learning rate scheduling.
• Neural Network Training Toolkits - Offers a comprehensive toolkit for managing the PyTorch training lifecycle, including checkpointing and learning rate scheduling.
• Hyperparameter Optimizers - Automates the search for optimal neural network hyperparameters using grid search and cross-validation interfaces.
• Training Callbacks - Implements modular training callbacks for automated checkpointing, early stopping, and learning rate adjustments.
• Workflow Pipelines - Embeds neural networks into standard data workflows to automate feature scaling and transformation steps.
• Estimator Wrappers - Adapts deep learning modules into a standardized estimator interface for compatibility with conventional ML pipelines.
• Dynamic Input Dimension Inference - Automatically configures neural network input layer shapes by inspecting the first batch of training data.
• Experiment Logging - Integrates recording of training statistics and model state checkpoints to external tracking tools.
• Hyperparameter Optimization Tools - Implements a system for tuning PyTorch model hyperparameters via standard grid search and cross-validation.
• Learning Rate Schedulers - Dynamically adjusts training step sizes during optimization to improve numerical stability and accelerate convergence.
• Automatic Accelerator Selection - Provides dynamic logic to select the most efficient hardware backend, such as CPU or GPU, at runtime.
• Model Checkpointing - Saves periodic snapshots of model weights and optimizer states to enable recovery from training interruptions.
• Hyperparameter Tuning - Automates the search for optimal network configurations using scikit-learn's grid search and cross-validation tools.
• Model Performance Evaluators - Calculates predictive accuracy and statistical metrics using standardized scoring functions against benchmarks.
• Layer-Specific Training - Provides mechanisms to freeze specific layers or weights during training to facilitate efficient fine-tuning.
• Parameter Freezing Utilities - Provides mechanisms to freeze or unfreeze specific model parameters to facilitate efficient transfer learning.
• Tabular-to-Tensor Pipelines - Provides utilities for converting tabular dataframes and numerical arrays into standardized tensor representations.
• Training Progress Monitors - Displays real-time execution updates and training metrics via progress bars in command line or notebook environments.
• Transfer Learning - Supports adapting pre-trained weights to specialized tasks by freezing layers and fine-tuning model components.
• Transfer Learning Toolkits - Ships utilities for freezing specific PyTorch model layers and fine-tuning pre-trained weights for specialized tasks.
• Tabular Data Processing - Facilitates deep learning on structured datasets by converting standard dataframes and arrays into tensors.
• Model Fine-Tuning - Adapts pre-trained neural network weights to specialized datasets for high performance with minimal training.
• Tabular-to-Tensor Converters - Converts heterogeneous tabular data structures into the numerical tensor formats required for deep learning model inputs.
• Deep Learning Ecosystems - Scikit-learn compatible wrapper for PyTorch.
• Deep Learning Frameworks - Scikit-learn compatible neural network library for PyTorch.
• General Machine Learning - Scikit-learn compatible wrapper for PyTorch.
• Machine Learning - Scikit-learn compatible neural network library.
• Frameworks de Machine Learning - Scikit-learn compatible wrapper for PyTorch neural networks.

Historial de estrellas