def test_empty_mpo(self):
mpiprint(
0, '\n' + '=' * 50 +
'\nTesting Initial Environment Using Identity MPO\n' + '-' * 50)
# Create an MPS
from cyclomps.tools.mps_tools import create_mps_list, make_mps_list_right
from cyclomps.tools.env_tools import calc_env, env_load_ten
d = 3
N = 10
mbd = 100
nState = 4
mpsList = create_mps_list([d] * N, mbd, nState)
mpsList = make_mps_list_right(mpsList)
# Load an mpo
mpoList = []
mpo = [None] * N
mpoList.append(mpo)
# Calculate the environment
envList = calc_env(mpsList, mpoList)
# We can assert that the environment is now an identity at each site
for site in range(N - 1, 0, -1):
ident_check = env_load_ten(envList, 0, site)
ident_check = einsum('ijk->ik', ident_check)
(nx, ny) = ident_check.shape
I = eye(nx)
diff = summ(abss(I - ident_check))
self.assertTrue(diff < 1e-8)
# We can check the norm to see if it agrees
mpiprint(0, 'Passed\n' + '=' * 50)
def test_one_site(self):
mpiprint(
0,
'\n' + '=' * 20 + '\nTesting One Site Optimization\n' + '=' * 20)
# Create an MPS
from cyclomps.tools.mps_tools import create_mps_list, make_mps_list_right
from cyclomps.tools.env_tools import calc_env, env_load_ten
from cyclomps.tools.diag_tools import eig1
d = 3
N = 10
mbd = 100
nState = 4
mpsList = create_mps_list([d] * N, mbd, nState)
mpsList = make_mps_list_right(mpsList)
# Load an mpo
mpoList = []
mpo = [None] * N
mpoList.append(mpo)
# Calculate the environment
envList = calc_env(mpsList, mpoList)
# Retrieve Relevant tensors for the local optimization
(mps, ) = retrieve_tensors(0, mpsList=mpsList)
(envl, envr) = retrieve_tensors(0, envList=envList)
(mpo, ) = retrieve_tensors(0, mpoList=mpoList)
# Send to diagonalization routine
E, mps0, ovlp = eig1(mps, mpo, envl, envr, alg='exact')
mpiprint(0, 'Exact Worked')
#E,mps0,ovlp = eig1(mps,mpo,envl,envr,alg='arnoldi')
#mpiprint(0,'Arnoldi Worked')
E, mps0, ovlp = eig1(mps, mpo, envl, envr, alg='davidson')
mpiprint(0, 'Davidson Worked')
mpiprint(0, 'Passed\n' + '=' * 20)
def test_tasep_mpo(self):
mpiprint(
0, '\n' + '=' * 50 +
'\nTesting Initial Environment Using TASEP MPO\n' + '-' * 50)
# Create an MPS
from cyclomps.tools.mps_tools import create_mps_list, make_mps_list_right, load_mps_list, calc_mps_norm
d = 2
N = 4
mbd = 10
nState = 4
mpsList = create_mps_list([d] * N, mbd, nState)
# Load the MPO
from cyclomps.mpo.tasep import return_mpo
mpo = return_mpo(N, (0.5, 0.5, 1.))
# Create a function to compute energy
def calc_energy(mps, mpo, load_mps=True):
if load_mps:
mps = load_mps_list(mps)
E = zeros(nState, dtype=mps[0][0].dtype)
for state in range(nState):
psi = einsum('apb,bqc,crd,dse->pqrs', mps[state][0],
mps[state][1], mps[state][2], mps[state][3])
op = einsum('iPpj,jQqk,kRrl,lSsm->PpQqRrSs', mpo[0][0],
mpo[0][1], mpo[0][2], mpo[0][3])
psi_conj = einsum('apb,bqc,crd,dse->pqrs', conj(mps[state][0]),
conj(mps[state][1]), conj(mps[state][2]),
conj(mps[state][3]))
E[state] = einsum('pqrs,PpQqRrSs,PQRS', psi, op, psi_conj)
norm = einsum('pqrs,pqrs->', psi, psi_conj)
E[state] /= norm
return E
# Compute energy
E1 = calc_energy(mpsList, mpo)
mpiprint(0, 'Initial Energy =\n{}'.format(E1))
# Right Canonicalize
mpsList = make_mps_list_right(mpsList)
# Compute energy again
E2 = calc_energy(mpsList, mpo)
mpiprint(0, 'Energy after canonicalization =\n{}'.format(E2))
# Calculate the environment
from cyclomps.tools.env_tools import calc_env, env_load_ten, load_env_list
envList = calc_env(mpsList, mpo)
# Calculate the energy using the environment
env = load_env_list(envList)
mps = load_mps_list(mpsList)
E3 = zeros(nState, dtype=mps[0][0].dtype)
for state in range(nState):
E3[state] = einsum('Ala,apb,lPpr,APB,Brb->', env[0][0], mps[0][0],
mpo[0][0], conj(mps[0][0]), env[0][1])
E3[state] /= calc_mps_norm(mpsList, state=state)
mpiprint(0, 'Energy from environment =\n{}'.format(E3))
# Check Results
self.assertTrue(summ(abss(E1[0] - E2[0])) < 1e-10)
self.assertTrue(summ(abss(E1[0] - E3[0])) < 1e-10)
mpiprint(0, 'Passed\n' + '=' * 50)
def test_retrieve(self):
mpiprint(
0, '\n' + '=' * 20 + '\nTesting Tensor Retrieve Function\n' +
'=' * 20)
# Create an MPS
from cyclomps.tools.mps_tools import create_mps_list, make_mps_list_right
from cyclomps.tools.env_tools import calc_env, env_load_ten
d = 3
N = 10
mbd = 100
nState = 4
mpsList = create_mps_list([d] * N, mbd, nState)
mpsList = make_mps_list_right(mpsList)
# Load an mpo
mpoList = []
mpo = [None] * N
mpoList.append(mpo)
# Calculate the environment
envList = calc_env(mpsList, mpoList)
# Retrieve Relevant tensors for the local optimization
(mps0, ) = retrieve_tensors(0, mpsList=mpsList)
(mps1, ) = retrieve_tensors(1, mpsList=mpsList)
(envL, envR) = retrieve_tensors(0, envList=envList)
(env2L, env2R) = retrieve_tensors(0, envList=envList, twoSite=True)
(mpo0, ) = retrieve_tensors(0, mpoList=mpoList)
(mpo1, ) = retrieve_tensors(1, mpoList=mpoList)
# Check to make sure all dimensions align
(n1, n2, n3) = mps0[0].shape
(n4, n5, n6) = mps1[0].shape
(n7, n8, n9) = envL[0].shape
(n10, n11, n12) = envR[0].shape
(n13, n14, n15) = env2L[0].shape
(n16, n17, n18) = env2R[0].shape
self.assertEqual(n1, n7)
self.assertEqual(n1, n13)
self.assertEqual(n3, n4)
self.assertEqual(n3, n12)
self.assertEqual(n3, n10)
self.assertEqual(n6, n16)
self.assertEqual(n6, n18)
# We can check the norm to see if it agrees
mpiprint(0, 'Passed\n' + '=' * 20)