def test_cart_frac_conversion():
import numpy as np
from chemtorch.structure.lattice import standardize_lattice
from chemtorch.structure.coordinates import frac2cart
from chemtorch.structure.coordinates import cart2frac
l0 = []
l1 = np.random.rand(1, 1) * 5
l2 = np.random.rand(2, 2) * 5
l3 = np.random.rand(3, 3) * 5
l0 = standardize_lattice(l0)
l1 = standardize_lattice(l1)
l2 = standardize_lattice(l2)
l3 = standardize_lattice(l3)
L = [l0, l1, l2, l3]
# 3 dimensional coordinates
R = np.random.rand(10, 3) * 5
for l in L:
R0 = R.copy()
R1 = frac2cart(cart2frac(R0, l), l)
R0 = R1.copy()
R1 = frac2cart(cart2frac(R0, l), l)
print(len(l), "dimension conversional")
print("First row of the coordinate is", R1[0])
print("maximum error is", np.max(R1 - R0))
# 4 dimensional coordinates
l4 = np.random.rand(4, 4) * 5
l4 = standardize_lattice(l4)
L.append(l4)
R = np.random.rand(10, 4) * 5
for l in L:
R0 = R.copy()
R1 = frac2cart(cart2frac(R0, l), l)
R0 = R1.copy()
R1 = frac2cart(cart2frac(R0, l), l)
print(len(l), "dimension conversional")
print("First row of the coordinate is", R1[0])
print("maximum error is", np.max(R1 - R0))
mlp = [num_feat, 50, 50, 50, num_engy]
weights, biases = get_weights(mlp, xavier=True), get_biases(mlp)
optimizer = torch.optim.Adam(biases + weights, lr=1e-4)
featFile = "tests\\data\\feat" if platform == 'win32' else "tests/data/feat"
engyFile = "tests\\data\\engy" if platform == 'win32' else "tests/data/engy"
filename = "tests\data\MOVEMENT.train" if platform == 'win32' else "tests/data/MOVEMENT.train"
mmt = read_PWMat_movement(filename,
get_forces=True,
get_velocities=True,
get_Ei=True,
get_Epot=True)
for iteration, (n_atoms, lattice, atom_types, Rfrac, F, V, Ei,
Epot) in enumerate(mmt):
lattice = standardize_lattice(lattice)
Rcart = frac2cart(Rfrac, lattice)
idxNb, Rij, maxNb = get_nb(Rcart, lattice, dcut=6.2)
dij, dijk, Rhat = get_distances(Rij)
g, g_dldl, g_dpdl = get_d_features(dij, dijk, Rhat, M2, M3, Router=6.2)
x = torch.from_numpy(g.astype(np.float32))
y = torch.from_numpy(Ei[:, None].astype(np.float32))
print(iteration)
pd.DataFrame(g.astype(np.float32)).to_csv(featFile,
header=False,
index=False,
mode="a")
pd.DataFrame(Ei[:, None].astype(np.float32)).to_csv(engyFile,