def extract_data_sub(subfolder, transform_y = lambda y: np.log10(y)):
base = osio.cwd()
osio.cd(subfolder)
X = np.load('xnkl.array.npy')
y = float(open('jeff2_h.txt', 'r').readline())
y = transform_y(y)
osio.cd(base)
return X, y
from momo import osio, endl, flush
fs = [ 'nN', 'aA', 'cC' ]
mol = 'PEN'
for f in fs:
osio.cd(f)
osio >> '../gdma_formchk.sh {mol:s}_{f:s}'.format(mol=mol, f=f)
osio.cd(-1)
from momo import osio, endl, flush
fs = ['nN', 'aA', 'cC']
mol = 'PEN'
for f in fs:
osio.cd(f)
osio >> '../gdma_formchk.sh {mol:s}_{f:s}'.format(mol=mol, f=f)
osio.cd(-1)
return
# LOAD STRUCTURES
structures = [
soap.tools.setup_structure_ase(c.config_file, c.atoms)
for c in soap.tools.ase_load_all('in.configs')
]
struct_dict = {}
for struct in structures:
print struct.label
struct_dict[struct.label] = struct
# CHANGE TO WORKING DIRECTORY
osio.cd('out.files.hexcomb')
# SELECT STRUCTURE
struct = struct_dict[select_struct_label]
# CREATE NETWORK
hexcomb = SimSpaceNetwork(options)
hexcomb.createHexComb(c=1.4, N=N_network)
hexcomb.createNeighbours(cutoff=1.41)
hexcomb.writeXyz()
root_idx = 2*((2*N_network+1)*2*N_network + N_network)
"""
hexcomb.printInfo()
target_map = {}
tree = XmlTree('extract.pairs.xml')
pair_tags = tree.DiveForAll('pair')
pair_count = 0
for pair in pair_tags:
id1 = pair['id1'](int)
id2 = pair['id2'](int)
jeff2_h = pair['channel'].GetUnique('jeff2_h')(float)
if target_map.has_key(id1):
pass
else:
target_map[id1] = {}
target_map[id1][id2] = jeff2_h
# EXPAND STRUCTURES
osio.cd('frame0')
pair_folders = get_dirs('./', 'pair_')
n_pairs = len(pair_folders)
pair_folders = sorted(
pair_folders,
key=lambda f: int(f.split('_')[1]) * n_pairs + int(f.split('_')[2]))
xnkl_list = []
target_list = []
label_list = []
pair_id = 0
for folder in pair_folders:
pair_id += 1
# READ XYZ => CONFIG
osio.cd(folder)
osio << "Pair-ID %5d ('%s')" % (pair_id, folder) << endl
osio << "Run ..." << endl
out = osio >> osio.catch >> "ctp_parallel -l -v"
osio >> osio.devnull >> "ctp_run -l -v"
osio.DisconnectFromFile()
ls = osio >> osio.catch >> "ls"
osio << "Result of 'ls':" << endl
for item in ls.split():
osio << osio.item << item << endl
# DIRECTORY HIKING
osio << osio.mg << "DIRECTORY HIKING" << endl
try:
osio >> osio.assert_zero >> 'mkdir testdir'
except RuntimeError:
osio >> 'rm -r testdir'
osio >> 'mkdir testdir'
osio.cd('/home')
osio.cd(0)
osio.cd('testdir')
osio.cd(-2)
osio.cd(-1)
osio.cd('../')
cwd = osio.pwd()
osio << "Current directory:" << cwd << endl
# EXIT PROTOCOLS
osio << osio.mg << "EXIT PROTOCOLS" << endl
osio.okquit()
def extract_data(folder, N_train, N_test, regex):
def extract_data_sub(subfolder, transform_y = lambda y: np.log10(y)):
base = osio.cwd()
osio.cd(subfolder)
X = np.load('xnkl.array.npy')
y = float(open('jeff2_h.txt', 'r').readline())
y = transform_y(y)
osio.cd(base)
return X, y
def sort_subfolders(folders):
return sorted(folders, key = lambda f: int(f.split('_')[1])*len(folders)+int(f.split('_')[2]))
base = osio.cwd()
osio << "Base =" << base << endl
osio.cd(folder)
# GET SUBFOLDERS
subfolders = sort_subfolders(get_dirs('./', regex))
# LIMIT RANGE
N = N_train + N_test
assert N < len(subfolders)
subfolders = subfolders[:N]
# FIND DESCRIPTOR DIMENSIONS
dim_x = extract_data_sub(subfolders[0])[0].shape[0]
osio << "Descriptor dimension =" << dim_x << endl
# EXTRACT FEATURES & TARGETS
X_list = np.zeros((N, dim_x))
y_list = np.zeros((N))
idx = -1
for subfolder in subfolders:
idx += 1
osio << osio.back << subfolder << flush
X, y = extract_data_sub(subfolder)
X_list[idx] = X
y_list[idx] = y
osio << endl
# SPLIT ONTO TRAINING & TEST SET
X_train = np.zeros((N_train, dim_x))
y_train = np.zeros((N_train))
X_test = np.zeros((N_test, dim_x))
y_test = np.zeros((N_test))
X_train[0] = X_list[0]
y_train[0] = y_list[0]
count_train = 1
count_test = 0
for idx in range(1,N):
if float(count_train)/(count_train+count_test) < float(N_train)/(N):
X_train[count_train] = X_list[idx]
y_train[count_train] = y_list[idx]
count_train += 1
else:
X_test[count_test] = X_list[idx]
y_test[count_test] = y_list[idx]
count_test += 1
assert count_train == N_train
assert count_test == N_test
osio.cd(base)
return X_train, y_train, X_test, y_test
osio << "Run ..." << endl
out = osio >> osio.catch >> "ctp_parallel -l -v"
osio >> osio.devnull >> "ctp_run -l -v"
osio.DisconnectFromFile()
ls = osio >> osio.catch >> "ls"
osio << "Result of 'ls':" << endl
for item in ls.split():
osio << osio.item << item << endl
# DIRECTORY HIKING
osio << osio.mg << "DIRECTORY HIKING" << endl
try:
osio >> osio.assert_zero >> "mkdir testdir"
except RuntimeError:
osio >> "rm -r testdir"
osio >> "mkdir testdir"
osio.cd("/home")
osio.cd(0)
osio.cd("testdir")
osio.cd(-2)
osio.cd(-1)
osio.cd("../")
cwd = osio.pwd()
osio << "Current directory:" << cwd << endl
# EXIT PROTOCOLS
osio << osio.mg << "EXIT PROTOCOLS" << endl
osio.okquit()
ofs.write(
"""\
Title "{tag:s}"
File {fchk:s}
Multipoles
Limit 2
Limit 1 H
Radius H 0.35
Punch {tag:s}.mps
Start
Finish
""".format(
fchk=fchkfile, tag=tag
)
)
ofs.close()
GDMA = "/people/thnfs/homes/poelking/GDMA/gdma-2.2.06/bin/gdma"
dirs = ["nN", "aA", "cC"]
mol = "PEN"
for dir in dirs:
osio.cd(dir)
osio << "Directory: '%s'" % dir << endl
write_gdma_infile("{mol:s}_{dir:s}.fchk".format(mol=mol, dir=dir))
osio >> "{gdma:s} < gdma.in > gdma.out &".format(gdma=GDMA)
osio.cd(-1)
assert fchkfile[-5:] == '.fchk'
tag = fchkfile[:-5]
osio << "File 'gdma.in' in" << osio.cwd() << ":" << fchkfile << endl
ofs = open('gdma.in', 'w')
ofs.write('''\
Title "{tag:s}"
File {fchk:s}
Multipoles
Limit 2
Limit 1 H
Radius H 0.35
Punch {tag:s}.mps
Start
Finish
'''.format(fchk=fchkfile, tag=tag))
ofs.close()
GDMA = '/people/thnfs/homes/poelking/GDMA/gdma-2.2.06/bin/gdma'
dirs = ['nN', 'aA', 'cC']
mol = 'PEN'
for dir in dirs:
osio.cd(dir)
osio << "Directory: '%s'" % dir << endl
write_gdma_infile('{mol:s}_{dir:s}.fchk'.format(mol=mol, dir=dir))
osio >> '{gdma:s} < gdma.in > gdma.out &'.format(gdma=GDMA)
osio.cd(-1)
# LOAD STRUCTURES
structures = [
soap.tools.setup_structure_ase(c.config_file, c.atoms)
for c in soap.tools.ase_load_all('in.configs')
]
struct_dict = {}
for struct in structures:
print struct.label
struct_dict[struct.label] = struct
# SELECT STRUCTURE
struct = struct_dict[select_struct_label]
# CREATE TREE
osio.cd('out.files')
tree = SimSpaceTree(options)
tree.seed(struct)
N_generations = 2
N_spawn = [3,3]
N_generations = 4
N_spawn = [2,2,2,2]
N_generations = 9
N_spawn = [ 2 for n in range(N_generations) ]
authorize_spawn_fct = lambda joint: joint.node.energy <= 0.0
for n in range(N_generations):
gen_id = n+1
import espressopp as sxx
from momo import osio, endl, flush
import mpi4py.MPI as MPI
import os
osio << "Running on %d cores ..." % MPI.COMM_WORLD.size << endl
# COLLECT CONFIG FILES FROM FOLDER
config_file_dir = './configs'
osio.cd(config_file_dir)
config_files = sorted(os.listdir('./'))
config_list = sxx.tools.convert.aseread.AseConfigList(config_files, log=osio)
osio.cd(-1)
# SETUP SYSTEMS FOR CONFIGURATIONS
systems = []
for config in config_list:
osio << config.atoms << endl
system = sxx.tools.convert.aseread.setup_sxx_system(config.atoms, config.config_file)
systems.append(system)
# SET OPTIONS
options = sxx.soap.Options()
options.configureCenters(1., [1,77,119])
options.configureRealBasis(12,9,10.)
options.configureReciprocalBasis(10,7,8.)
b1 = sxx.Real3D(2.,0,0)
b2 = sxx.Real3D(0,2.,0)
b3 = sxx.Real3D(0,0,2.)
options.configureReciprocalLattice(b1, b2, b3)