Update solvers (#434)
* Enable DDP training with batch_size=None and add validity check for split sizes * Refactoring SolverInterfaces (#435) * Solver update + weighting * Updating PINN for 0.2 * Modify GAROM + tests * Adding more versatile loggers * Disable compilation when running on Windows * Fix tests --------- Co-authored-by: giovanni <giovanni.canali98@yahoo.it> Co-authored-by: FilippoOlivo <filippo@filippoolivo.com>
This commit is contained in:
Dario Coscia
committed by
Nicola Demo
Nicola Demo
parent
780c4921eb
commit
9cae9a438f
@@ -1,143 +1,133 @@
|
||||
import torch
|
||||
import pytest
|
||||
from pina.problem import AbstractProblem, SpatialProblem
|
||||
from pina import Condition, LabelTensor
|
||||
from pina.condition import InputOutputPointsCondition
|
||||
from pina.problem import AbstractProblem
|
||||
from pina.solvers import SupervisedSolver
|
||||
from pina.model import FeedForward
|
||||
from pina.equation import Equation
|
||||
from pina.equation.equation_factory import FixedValue
|
||||
from pina.operators import laplacian
|
||||
from pina.domain import CartesianDomain
|
||||
from pina.trainer import Trainer
|
||||
|
||||
# in_ = LabelTensor(torch.tensor([[0., 1.]]), ['u_0', 'u_1'])
|
||||
# out_ = LabelTensor(torch.tensor([[0.]]), ['u'])
|
||||
from torch._dynamo.eval_frame import OptimizedModule
|
||||
|
||||
|
||||
# class NeuralOperatorProblem(AbstractProblem):
|
||||
# input_variables = ['u_0', 'u_1']
|
||||
# output_variables = ['u']
|
||||
|
||||
# conditions = {
|
||||
# 'data': Condition(input_points=in_, output_points=out_),
|
||||
# }
|
||||
class LabelTensorProblem(AbstractProblem):
|
||||
input_variables = ['u_0', 'u_1']
|
||||
output_variables = ['u']
|
||||
conditions = {
|
||||
'data': Condition(
|
||||
input_points=LabelTensor(torch.randn(20, 2), ['u_0', 'u_1']),
|
||||
output_points=LabelTensor(torch.randn(20, 1), ['u'])),
|
||||
}
|
||||
|
||||
|
||||
# class myFeature(torch.nn.Module):
|
||||
# """
|
||||
# Feature: sin(x)
|
||||
# """
|
||||
|
||||
# def __init__(self):
|
||||
# super(myFeature, self).__init__()
|
||||
|
||||
# def forward(self, x):
|
||||
# t = (torch.sin(x.extract(['u_0']) * torch.pi) *
|
||||
# torch.sin(x.extract(['u_1']) * torch.pi))
|
||||
# return LabelTensor(t, ['sin(x)sin(y)'])
|
||||
class TensorProblem(AbstractProblem):
|
||||
input_variables = ['u_0', 'u_1']
|
||||
output_variables = ['u']
|
||||
conditions = {
|
||||
'data': Condition(
|
||||
input_points=torch.randn(20, 2),
|
||||
output_points=torch.randn(20, 1))
|
||||
}
|
||||
|
||||
|
||||
# problem = NeuralOperatorProblem()
|
||||
# extra_feats = [myFeature()]
|
||||
# model = FeedForward(len(problem.input_variables), len(problem.output_variables))
|
||||
# model_extra_feats = FeedForward(
|
||||
# len(problem.input_variables) + 1, len(problem.output_variables))
|
||||
model = FeedForward(2, 1)
|
||||
|
||||
|
||||
# def test_constructor():
|
||||
# SupervisedSolver(problem=problem, model=model)
|
||||
def test_constructor():
|
||||
SupervisedSolver(problem=TensorProblem(), model=model)
|
||||
SupervisedSolver(problem=LabelTensorProblem(), model=model)
|
||||
assert SupervisedSolver.accepted_conditions_types == (
|
||||
InputOutputPointsCondition
|
||||
)
|
||||
|
||||
|
||||
# test_constructor()
|
||||
@pytest.mark.parametrize("batch_size", [None, 1, 5, 20])
|
||||
@pytest.mark.parametrize("use_lt", [True, False])
|
||||
@pytest.mark.parametrize("compile", [True, False])
|
||||
def test_solver_train(use_lt, batch_size, compile):
|
||||
problem = LabelTensorProblem() if use_lt else TensorProblem()
|
||||
solver = SupervisedSolver(problem=problem, model=model, use_lt=use_lt)
|
||||
trainer = Trainer(solver=solver,
|
||||
max_epochs=2,
|
||||
accelerator='cpu',
|
||||
batch_size=batch_size,
|
||||
train_size=1.,
|
||||
test_size=0.,
|
||||
val_size=0.,
|
||||
compile=compile)
|
||||
|
||||
trainer.train()
|
||||
if trainer.compile:
|
||||
assert (isinstance(solver.model, OptimizedModule))
|
||||
|
||||
|
||||
# def laplace_equation(input_, output_):
|
||||
# force_term = (torch.sin(input_.extract(['x']) * torch.pi) *
|
||||
# torch.sin(input_.extract(['y']) * torch.pi))
|
||||
# delta_u = laplacian(output_.extract(['u']), input_)
|
||||
# return delta_u - force_term
|
||||
@pytest.mark.parametrize("use_lt", [True, False])
|
||||
@pytest.mark.parametrize("compile", [True, False])
|
||||
def test_solver_validation(use_lt, compile):
|
||||
problem = LabelTensorProblem() if use_lt else TensorProblem()
|
||||
solver = SupervisedSolver(problem=problem, model=model, use_lt=use_lt)
|
||||
trainer = Trainer(solver=solver,
|
||||
max_epochs=2,
|
||||
accelerator='cpu',
|
||||
batch_size=None,
|
||||
train_size=0.9,
|
||||
val_size=0.1,
|
||||
test_size=0.,
|
||||
compile=compile)
|
||||
trainer.train()
|
||||
if trainer.compile:
|
||||
assert (isinstance(solver.model, OptimizedModule))
|
||||
|
||||
|
||||
# my_laplace = Equation(laplace_equation)
|
||||
@pytest.mark.parametrize("use_lt", [True, False])
|
||||
@pytest.mark.parametrize("compile", [True, False])
|
||||
def test_solver_test(use_lt, compile):
|
||||
problem = LabelTensorProblem() if use_lt else TensorProblem()
|
||||
solver = SupervisedSolver(problem=problem, model=model, use_lt=use_lt)
|
||||
trainer = Trainer(solver=solver,
|
||||
max_epochs=2,
|
||||
accelerator='cpu',
|
||||
batch_size=None,
|
||||
train_size=0.8,
|
||||
val_size=0.1,
|
||||
test_size=0.1,
|
||||
compile=compile)
|
||||
trainer.test()
|
||||
if trainer.compile:
|
||||
assert (isinstance(solver.model, OptimizedModule))
|
||||
|
||||
|
||||
# class Poisson(SpatialProblem):
|
||||
# output_variables = ['u']
|
||||
# spatial_domain = CartesianDomain({'x': [0, 1], 'y': [0, 1]})
|
||||
def test_train_load_restore():
|
||||
dir = "tests/test_solvers/tmp/"
|
||||
problem = LabelTensorProblem()
|
||||
solver = SupervisedSolver(problem=problem, model=model)
|
||||
trainer = Trainer(solver=solver,
|
||||
max_epochs=5,
|
||||
accelerator='cpu',
|
||||
batch_size=None,
|
||||
train_size=0.9,
|
||||
test_size=0.1,
|
||||
val_size=0.,
|
||||
default_root_dir=dir)
|
||||
trainer.train()
|
||||
|
||||
# conditions = {
|
||||
# 'gamma1':
|
||||
# Condition(domain=CartesianDomain({
|
||||
# 'x': [0, 1],
|
||||
# 'y': 1
|
||||
# }),
|
||||
# equation=FixedValue(0.0)),
|
||||
# 'gamma2':
|
||||
# Condition(domain=CartesianDomain({
|
||||
# 'x': [0, 1],
|
||||
# 'y': 0
|
||||
# }),
|
||||
# equation=FixedValue(0.0)),
|
||||
# 'gamma3':
|
||||
# Condition(domain=CartesianDomain({
|
||||
# 'x': 1,
|
||||
# 'y': [0, 1]
|
||||
# }),
|
||||
# equation=FixedValue(0.0)),
|
||||
# 'gamma4':
|
||||
# Condition(domain=CartesianDomain({
|
||||
# 'x': 0,
|
||||
# 'y': [0, 1]
|
||||
# }),
|
||||
# equation=FixedValue(0.0)),
|
||||
# 'D':
|
||||
# Condition(domain=CartesianDomain({
|
||||
# 'x': [0, 1],
|
||||
# 'y': [0, 1]
|
||||
# }),
|
||||
# equation=my_laplace),
|
||||
# 'data':
|
||||
# Condition(input_points=in_, output_points=out_)
|
||||
# }
|
||||
# restore
|
||||
new_trainer = Trainer(solver=solver, max_epochs=5, accelerator='cpu')
|
||||
new_trainer.train(
|
||||
ckpt_path=f'{dir}/lightning_logs/version_0/checkpoints/' +
|
||||
'epoch=4-step=5.ckpt')
|
||||
|
||||
# def poisson_sol(self, pts):
|
||||
# return -(torch.sin(pts.extract(['x']) * torch.pi) *
|
||||
# torch.sin(pts.extract(['y']) * torch.pi)) / (2 * torch.pi ** 2)
|
||||
# loading
|
||||
new_solver = SupervisedSolver.load_from_checkpoint(
|
||||
f'{dir}/lightning_logs/version_0/checkpoints/epoch=4-step=5.ckpt',
|
||||
problem=problem, model=model)
|
||||
|
||||
# truth_solution = poisson_sol
|
||||
test_pts = LabelTensor(torch.rand(20, 2), problem.input_variables)
|
||||
assert new_solver.forward(test_pts).shape == (20, 1)
|
||||
assert new_solver.forward(test_pts).shape == solver.forward(test_pts).shape
|
||||
torch.testing.assert_close(
|
||||
new_solver.forward(test_pts),
|
||||
solver.forward(test_pts))
|
||||
|
||||
|
||||
# def test_wrong_constructor():
|
||||
# poisson_problem = Poisson()
|
||||
# with pytest.raises(ValueError):
|
||||
# SupervisedSolver(problem=poisson_problem, model=model)
|
||||
|
||||
|
||||
# def test_train_cpu():
|
||||
# solver = SupervisedSolver(problem=problem, model=model)
|
||||
# trainer = Trainer(solver=solver,
|
||||
# max_epochs=200,
|
||||
# accelerator='gpu',
|
||||
# batch_size=5,
|
||||
# train_size=1,
|
||||
# test_size=0.,
|
||||
# val_size=0.)
|
||||
# trainer.train()
|
||||
# test_train_cpu()
|
||||
|
||||
|
||||
# def test_extra_features_constructor():
|
||||
# SupervisedSolver(problem=problem,
|
||||
# model=model_extra_feats,
|
||||
# extra_features=extra_feats)
|
||||
|
||||
|
||||
# def test_extra_features_train_cpu():
|
||||
# solver = SupervisedSolver(problem=problem,
|
||||
# model=model_extra_feats,
|
||||
# extra_features=extra_feats)
|
||||
# trainer = Trainer(solver=solver,
|
||||
# max_epochs=200,
|
||||
# accelerator='gpu',
|
||||
# batch_size=5)
|
||||
# trainer.train()
|
||||
# rm directories
|
||||
import shutil
|
||||
shutil.rmtree('tests/test_solvers/tmp')
|
||||
|
||||
Reference in New Issue
Block a user