How to run a model with PCSE

To get familiar with WOFOST models and how to run the models, we recommend to first check out the PCSE documentation and explore the notebooks 01 Getting Started with PCSE.ipynb and 02 Running with custom input data.ipynb .

In a nutshell, we can run a model, for example, leaf_dynamics using diffWOFOST as:

from diffwofost.physical_models.utils import EngineTestHelper
from diffwofost.physical_models.config import Configuration
from diffwofost.physical_models.crop.leaf_dynamics import WOFOST_Leaf_Dynamics

# create config
leaf_dynamics_config = Configuration(
    CROP=WOFOST_Leaf_Dynamics,
    OUTPUT_VARS=["LAI", "TWLV"],
)

# create the engine
engine = EngineTestHelper(
    leaf_dynamics_config,  # this where the differentiable model is specified
)

# create the model
model = engine.setup(
    crop_parameters_provider,  # this provides the crop parameters
    weather_data_provider,
    agromanagement_provider,
    external_states,  # any external states if needed
)

# run the simulation with a fixed time step of one day
model.run_till_terminate()

# get the output
results = model.get_output()

See the notebooks in the examples section for more details and examples on how to run the models.

How to set computing device and data type

By default, the model will run on torch default device and dtype i.e. torch.get_default_device() and torch.get_default_dtype(). See the examples section for more details.

How to run a crop model (physics-based) with torch.nn.Module

The crop models (physics-based) in diffWOFOST are implemented as a SimulationObject, and they can be run using Engine, as explained above. To run a model using torch.nn.Module, you can simply create a wrapper class that inherits from torch.nn.Module and calls the Engine.setup() in the forward method. See the notebooks in the examples section.

How to run a crop model (ml-based)

The crop models (ml-based) in diffWOFOST are also implemented as a SimulationObject, and they can be run using Engine, as explained above. To run a model, you can simply specify the class and any ml-based model that the class uses in the Configuration when creating the Engine as:

from diffwofost.physical_models.utils import EngineTestHelper
from diffwofost.physical_models.config import Configuration
from diffwofost.ml_models.crop.partitioning import DVS_Partitioning_NN, PartitioningMLP


# create config
partition_config = Configuration(
    CROP=DVS_Partitioning_NN,  # this is ml-based class and a SimulationObject
    CROP_NN_MODEL=PartitioningMLP(hidden_size=32),  # as an example
    OUTPUT_VARS=["LAI", "TWLV"],
)

# create the engine
engine = EngineTestHelper(
    partition_config,  # this where the ml-based model is specified
)

# create the model
model = engine.setup(
    crop_parameters_provider,  # this provides the crop parameters
    weather_data_provider,
    agromanagement_provider,
    external_states,  # any external states if needed
)

# run the simulation with a fixed time step of one day
model.run_till_terminate()

# get the output
results = model.get_output()

How to replace a model with a machine learning model (Hybrid modeling)

The crop (ml-based) models in diffWOFOST are also implemented as a SimulationObject. The physics-based computations inside the class can be replaced with a machine learning model. See the source codes in ml_models/crop/ for examples of how to replace a physics-based model with a machine learning model, such as the partitioning model in WOFOST72. Also, there are notebooks in the examples section.

Note

The ml-based models in diffWOFOST, for example PartitioningMLP and PartitioningNN from ml_models/crop/partitioning.py are meant to be used as an example of how to replace a physics-based model with a machine learning model. They are not trained and are not evaluated against the original wofost model results. The purpose is to show how to run a ml-based model in diffWOFOST and how to integrate it with the physics-based models.