Uploading files and datasets

1. Problem 1/10. Model 10/2D_heat_conduction.py

@@ -0,0 +1,300 @@
+# Import packages
+import numpy as np
+import torch # device assignment for importance model
+
+from sympy import Symbol, Eq, Function, Number, Abs # defining symbols
+
+import modulus # load classes and functions from modulus directory
+from modulus.hydra import to_absolute_path, to_yaml, instantiate_arch  # configure settings and create architecture
+from modulus.hydra.config import ModulusConfig # load config files
+#from modulus.csv_utils.csv_rw import csv_to_dict  # load true solution for validation
+from modulus.continuous.solvers.solver import Solver # continuous time solver
+from modulus.continuous.domain.domain import Domain  # domain (custom classes + architecture)
+from modulus.geometry.csg.csg_2d import Rectangle  # CSG geometry
+from modulus.continuous.constraints.constraint import (  # constraints
+    PointwiseBoundaryConstraint,  # BC
+    PointwiseInteriorConstraint,  # interior/collocation points
+)
+from modulus.constraint import Constraint # don't know
+from modulus.continuous.validator.validator import PointwiseValidator  # adding validation dataset
+from modulus.continuous.inferencer.inferencer import PointwiseInferencer # infer variables
+from modulus.tensorboard_utils.plotter import ValidatorPlotter, InferencerPlotter # tensorboard
+from modulus.key import Key  # add keys to nodes
+from modulus.node import Node # add node to geometry
+
+from modulus.pdes import PDES # PDEs
+from modulus.graph import Graph # for defining importance graph model
+from modulus.architecture.fully_connected import FullyConnectedArch
+from modulus.architecture.fourier_net import FourierNetArch
+from modulus.architecture.siren import SirenArch
+from modulus.architecture.modified_fourier_net import ModifiedFourierNetArch
+from modulus.architecture.dgm import DGMArch
+
+
+class HeatConductionEquation2D(PDES):
+    """
+    Heat Conduction 2D
+    
+    Parameters
+    ==========
+    k : float, string
+        Conductivity doesn't matter in steady state problem with no heat source
+    """
+
+    name = "HeatConductionEquation2D"
+    def __init__(self):
+        # coordinates
+        x = Symbol("x")
+
+        # time
+        y = Symbol("y")
+
+        # make input variables
+        input_variables = {"x": x, "y": y}
+
+        # Temperature output
+        T = Function("T")(*input_variables)
+
+        # conductivity coefficient
+        # if type(c) is str:
+        #     c = Function(c)(*input_variables) # if c is function of independent variables
+        # elif type(c) in [float, int]:
+        #     c = Number(c)
+
+        # set equations
+        self.equations = {}
+        self.equations["heat_equation"] = T.diff(x, 2) + T.diff(y, 2)  # diff(variable,order of derivative)
+
+
+@modulus.main(config_path="conf", config_name="config")
+def run(cfg: ModulusConfig) -> None:
+    print(to_yaml(cfg))
+
+    input_keys = [Key("x"), Key("y")]
+    output_keys=[Key("T")]
+    print('Detecting architecture')
+    # make list of nodes to unroll graph on
+    heat_eq = HeatConductionEquation2D()
+    if cfg.custom.arch == "FullyConnectedArch":
+        print('Architecture: Fully Connected Arch')
+        heat_net = FullyConnectedArch(
+            input_keys=input_keys,
+            output_keys=output_keys,
+            adaptive_activations=cfg.custom.adaptive_activations,
+        )
+    elif cfg.custom.arch == "FourierNetArch":
+        print('Architecture: Fourier Network')
+        heat_net = FourierNetArch(
+            input_keys=input_keys,
+            output_keys=output_keys,
+            adaptive_activations=cfg.custom.adaptive_activations,
+            frequencies=("axis", [i / 2 for i in range(35)]),
+            frequencies_params=("axis", [i / 2 for i in range(35)]),
+        )
+    elif cfg.custom.arch == "SirenArch":
+        print('Architecture: SIREN Arch')
+        heat_net = SirenArch(
+            input_keys=input_keys,
+            output_keys=output_keys,
+        )
+    elif cfg.custom.arch == "ModifiedFourierNetArch":
+        heat_net = ModifiedFourierNetArch(
+            input_keys=input_keys,
+            output_keys=output_keys,
+            adaptive_activations=cfg.custom.adaptive_activations,
+        )
+    else:
+        sys.exit(
+            "Network not configured for this script. Please include the network in the script"
+        )
+    # heat_net = instantiate_arch(
+    #     input_keys=[Key("x"), Key("y")],
+    #     output_keys=[Key("T")],
+    #     cfg=cfg.arch.fully_connected,
+    # )
+
+    nodes = heat_eq.make_nodes() + [heat_net.make_node(name="heat_network", jit=cfg.jit)]
+    print('Nodes Initialised')
+
+    # make importance model
+    device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
+
+    if torch.cuda.is_available():
+        print('GPU Name ', torch.cuda.get_device_name(0))
+
+    importance_model_graph = Graph(
+        nodes,
+        invar=[Key("x"), Key("y")],
+        req_names=[
+            Key("T", derivatives=[Key("x")]),
+            Key("T", derivatives=[Key("y")]),
+        ],
+    ).to(device)
+
+    def importance_measure(invar):
+        outvar = importance_model_graph(
+            Constraint._set_device(invar, requires_grad=True)
+        )
+        importance = (
+            outvar["T__x"] ** 2
+            + outvar["T__y"] ** 2
+        ) ** 0.5 + 10
+        return importance.cpu().detach().numpy()
+
+
+    # make geometry
+    lower_bound = (-0.5, 0)
+    upper_bound = (0.5, 1)
+
+    x, y = Symbol("x"), Symbol("y")
+    rec = Rectangle(lower_bound, upper_bound) # (x_1,y_1), (x_2,y_2)
+
+    print('Geometry created')
+
+    # make domain
+    heat_domain = Domain()
+    print('Domain created')
+
+    # Adding constraints
+    # Bottom wall
+    bottom_wall = PointwiseBoundaryConstraint(
+        nodes=nodes,
+        geometry=rec,
+        outvar={"T": 0.0}, # outputs
+        batch_size=cfg.batch_size.Wall,
+        lambda_weighting={"T": 1.0 - 2 * Abs(x)},  # weight edges to be zero
+        criteria=Eq(y, 0.0), # coordinates for sampling
+        importance_measure=importance_measure,
+        quasirandom=cfg.custom.quasirandom,
+    )
+    heat_domain.add_constraint(bottom_wall, "bottom_wall")
+
+    print('Bottom wall BC created')
+
+    # Top wall
+    top_wall = PointwiseBoundaryConstraint(
+        nodes=nodes,
+        geometry=rec,
+        outvar={"T": 0.0}, # outputs
+        batch_size=cfg.batch_size.Wall,
+        lambda_weighting={"T": 1.0 - 2 * Abs(x)},  # weight edges to be zero
+        criteria=Eq(y, 1.0), # coordinates for sampling
+        importance_measure=importance_measure,
+        quasirandom=cfg.custom.quasirandom,
+    )
+    heat_domain.add_constraint(top_wall, "top_wall")
+
+    print('Top wall BC created')
+
+    # Left wall
+    left_wall = PointwiseBoundaryConstraint(
+        nodes=nodes,
+        geometry=rec,
+        outvar={"T": 1.0}, # outputs
+        batch_size=cfg.batch_size.Wall,
+        lambda_weighting={"T": 1.0 - 2 * Abs(y-0.5)},  # weight edges to be zero
+        criteria=Eq(x, -0.5), # coordinates for sampling
+        importance_measure=importance_measure,
+        quasirandom=cfg.custom.quasirandom,
+    )
+    heat_domain.add_constraint(left_wall, "left_wall")
+
+    print('Left wall BC created')
+
+    # Right wall
+    right_wall = PointwiseBoundaryConstraint(
+        nodes=nodes,
+        geometry=rec,
+        outvar={"T": 1.0}, # outputs
+        batch_size=cfg.batch_size.Wall,
+        lambda_weighting={"T": 1.0 - 2 * Abs(y-0.5)},  # weight edges to be zero
+        criteria=Eq(x, 0.5), # coordinates for sampling
+        importance_measure=importance_measure,
+        quasirandom=cfg.custom.quasirandom,
+    )
+    heat_domain.add_constraint(right_wall, "right_wall")
+
+    print('Right wall BC created')
+
+    # PDE constraint
+    x_bound = (-0.5,0.5)
+    y_bound = (0, 1)
+    # interior
+    interior = PointwiseInteriorConstraint(
+        nodes=nodes,
+        geometry=rec,
+        outvar={"heat_equation" : 0},
+        batch_size=cfg.batch_size.Interior,
+        bounds={x: x_bound, y: y_bound},
+        # lambda weights for now
+        lambda_weighting={"heat_equation": rec.sdf},
+        importance_measure=importance_measure,
+        quasirandom=cfg.custom.quasirandom,
+    )
+    heat_domain.add_constraint(interior, "interior")
+
+    print('Interior points created')
+
+    # Getting path (nvidia modulus only accepts absolute path to remove any ambiguity)
+    validation_path = to_absolute_path("final_data.npz")
+
+    # Adding VALIDATION data
+    data = np.load(validation_path) # Load data from file
+    keys = list(data.keys()) # all keys in the dictionary
+    print('Validation dataset keys ',keys)
+    nodes3D = data[keys[0]]
+    temperature = data[keys[1]]
+    boundary_nodal_coordinates3D = data[keys[2]]
+    boundary_solution = data[keys[3]]
+    # face3 = data[keys[4]]
+    # face4 = data[keys[5]]
+    # face5 = data[keys[6]]
+    # face6 = data[keys[7]]
+
+
+    # cutting the useless third dimension where there is nothing to predict
+    node = nodes3D[:,[0,2]] # remember there is nodes variable also, they should not clash
+    #temperature = temperature3D[:,[0,2]]
+    boundary_nodal_coordinates = boundary_nodal_coordinates3D[:,[0,2]]
+    #boundary_solution = boundary_solution3D[:,[0,2]]
+
+
+    # This is the required format for validation data
+    openfoam_invar_numpy = {}
+    openfoam_invar_numpy["x"] = node[:, 0][:, None]
+    openfoam_invar_numpy["y"] = node[:, 1][:, None]
+
+    openfoam_outvar_numpy = {}
+    openfoam_outvar_numpy["T"] = temperature
+
+    openfoam_validator = PointwiseValidator(
+        openfoam_invar_numpy,
+        openfoam_outvar_numpy,
+        nodes,
+        batch_size=1024,
+        plotter=ValidatorPlotter(),
+    )
+    heat_domain.add_validator(openfoam_validator)
+
+    print('Validation Data loaded')
+
+    # add INFERENCER data
+    grid_inference = PointwiseInferencer(
+        openfoam_invar_numpy,
+        ["T", "T__x", "T__y"],
+        nodes,
+        batch_size=1024,
+        plotter=InferencerPlotter(),
+    )
+    heat_domain.add_inferencer(grid_inference, "inf_data")
+
+    print('Inference outputs loaded')
+
+        # make solver
+    slv = Solver(cfg, heat_domain)
+
+    # start solver
+    slv.solve()
+
+if __name__ == "__main__":
+    run()

1. Problem 1/10. Model 10/conf/.ipynb_checkpoints/config-checkpoint.yaml

@@ -0,0 +1,36 @@
+defaults :
+  - modulus_default
+  - scheduler: tf_exponential_lr
+  - optimizer: adam
+  - loss: sum
+  - _self_
+
+
+jit: false
+summary_histograms: true
+save_filetypes : "vtk,npz"
+
+scheduler:
+  decay_rate: 0.95
+  decay_steps: 4000
+
+optimizer:
+  lr: 1e-3   
+  betas: [0.9, 0.999]
+
+training:
+  rec_validation_freq: 100
+  rec_inference_freq: 100
+  rec_monitor_freq: 100
+  rec_constraint_freq: 100
+  max_steps : 20000
+
+custom:
+  arch: "ModifiedFourierNetArch"
+  exact_continuity: False
+  quasirandom: True
+  adaptive_activations: True
+
+batch_size:
+  Wall: 1000
+  Interior: 4000

1. Problem 1/10. Model 10/conf/config.yaml

@@ -0,0 +1,36 @@
+defaults :
+  - modulus_default
+  - scheduler: tf_exponential_lr
+  - optimizer: adam
+  - loss: sum
+  - _self_
+
+
+jit: false
+summary_histograms: true
+save_filetypes : "vtk,npz"
+
+scheduler:
+  decay_rate: 0.95
+  decay_steps: 4000
+
+optimizer:
+  lr: 1e-3   
+  betas: [0.9, 0.999]
+
+training:
+  rec_validation_freq: 100
+  rec_inference_freq: 100
+  rec_monitor_freq: 100
+  rec_constraint_freq: 100
+  max_steps : 20000
+
+custom:
+  arch: "ModifiedFourierNetArch"
+  exact_continuity: False
+  quasirandom: True
+  adaptive_activations: True
+
+batch_size:
+  Wall: 1000
+  Interior: 4000