Skip to main content

Introduction to PhyTorch

PhyTorch is a unified Python toolkit for fitting plant physiology models, from simple curve fitting to complex biochemical processes. With one consistent API, fit any model using the same simple pattern: fit(model, data, options).

What is PhyTorch?​

PhyTorch provides a unified framework for parameter estimation across all domains of plant physiology. Whether you're fitting a simple linear regression or a complex photosynthesis model, PhyTorch uses the same intuitive interface:

from phytorch import fit

result = fit(model, data)
result.plot()
result.write()

The same pattern works for:

  • Generic curve fitting (9 models: linear, sigmoidal, Michaelis-Menten, temperature responses, distributions)
  • Hydraulic models (vulnerability curves, pressure-volume relationships)
  • Photosynthesis models (biochemical C3 photosynthesis with environmental responses)
  • Stomatal conductance models (empirical and semi-empirical gs models)
  • Canopy architecture models (leaf angle distribution with canonical type classification)

Philosophy​

PhyTorch is built on three core principles:

  1. Unified API: One consistent interface for all models eliminates the need to learn different fitting procedures
  2. Simple to Complex: Start with basic curve fitting, scale up to multi-parameter physiological models seamlessly
  3. Automatic Intelligence: Built-in parameter initialization, plotting, and validation work automatically for all models

Overview​

Building on PhoTorch, PhyTorch extends robust photosynthesis modeling to a comprehensive plant physiology toolkit. By leveraging PyTorch's automatic differentiation and optimization algorithms, PhyTorch delivers efficient, reliable parameter estimation across the full spectrum of plant physiological processes.

Key Features​

  • Unified API: One function (fit) works for all models - from linear regression to complex photosynthesis
  • Comprehensive Model Library:
    • 9 generic curve-fitting models
    • 2 hydraulic models (vulnerability and P-V curves)
    • Photosynthesis (FvCB with temperature and light responses)
    • Stomatal conductance (Medlyn, Ball-Berry, and more)
    • Canopy architecture (leaf angle distribution with classification)
  • Automatic Visualization: Built-in plotting adapts to model type (1D curves, 3D surfaces for photosynthesis)
  • Smart Initialization: Auto-generates parameter starting values from your data
  • GPU-Accelerated: Leverages PyTorch for fast optimization and batch processing
  • Modular Design: Easily extend with custom models following the same base class pattern

Who Should Use PhyTorch?​

PhyTorch is designed for researchers in plant physiology, ecology, and agriculture who need to:

  1. Fit Physiological Models Simply: One unified interface for all model types eliminates API complexity
  2. Scale from Simple to Complex: Start with basic curve fitting, graduate to multi-parameter models without learning new syntax
  3. Process Data Efficiently: GPU acceleration and batch processing for high-throughput analysis
  4. Visualize Results Automatically: Built-in adaptive plotting for immediate quality assessment
  5. Ensure Reproducibility: Consistent API across models promotes standardized analysis workflows

Getting Started​

pip install phytorch-lib

Check out the Installation Guide for detailed instructions and the Getting Started Tutorial to fit your first model.