def read(self,filepath):
print 'reading Procar',
lines = nstrip(readfile(filepath))
self.nk = int(lines[1].split(':')[1].split()[0])
self.nbands = int(lines[1].split(':')[2].split()[0])
self.nions = int(lines[1].split(':')[3])
#initialize
self.kvecs = zeros((3,self.nk),dtype = float)
self.kweights = zeros((self.nk),dtype = float)
self.ener = zeros((self.nk, self.nbands, self.nspin),dtype = float)
self.occ = zeros((self.nk, self.nbands,self.nspin),dtype = float)
self.weights = zeros((self.nk,self.nbands,self.nions,len(self.orbs),self.nspin),dtype = float)
#read weights
ipos = 2
for ispin in range(self.nspin):
ipos += 1; # print 'ipos for spin %i: %i' % (ispin,ipos)
for ik in range(self.nk):
self.kvecs[:,ik] = lines[ipos].split()[3:6]
self.kweights[ik] = float(lines[ipos].split()[8])
ipos += 2
for ib in range(self.nbands):
self.ener[ik,ib,ispin] = float(lines[ipos].split()[4])
self.occ[ik,ib,ispin] = float(lines[ipos].split()[7])
ipos += 3
for i in range(self.nions):
if mod(ipos,1e4)==0: print '-', #progress bar
tempw = lines[ipos].split()[1:self.norbs+1]
self.weights[ik,ib,i,:,ispin] = tempw
ipos += 1
ipos += 2
ipos += 1
print
def readSym(dir):
lines = readfile('{}/sym.out'.format(dir))
nops = int(lines[0].split(':')[1])
IBZvolCut = float(lines[1].split(':')[1])
# IBZvol = float(lines[2].split(':')[1])
IBZvol = None
return nops, IBZvolCut, IBZvol
def rd_ibzkpt(self,dir): #only extracts the kpoint vectors
self.vecs = zeros((3,self.nk),dtype = float64)
lines = nstrip(readfile(dir+'IBZKPT'))
nk2 = int(lines[1].split()[0])
if nk2 != self.nk: print nk2,self.nk, 'Warning! Nkpts in IBZKPTS differs from this program''s'
for ik in range(self.nk):
self.vecs[:,ik] = lines[ik+3].split()[:3] #don't read weights'
def readSym(dir):
lines = readfile('{}/sym.out'.format(dir))
nops = int(lines[0].split(':')[1])
IBZvolCut = float(lines[1].split(':')[1])
# IBZvol = float(lines[2].split(':')[1])
IBZvol = None
return nops, IBZvolCut, IBZvol
def get_eigen(self,dir):
lines = nstrip(readfile(dir+'eigenv'))
self.nk = int(lines[0].split()[0])
self.nbands = int(lines[0].split()[1])
self.ener = zeros((self.nk,self.nbands),dtype = float64)
for ik in range(self.nk):
for ib in range(self.nbands):
self.ener[ik,ib] = lines[ik+1].split()[ib+1]
def get_tetweights(self,dir):
lines = nstrip(readfile(dir+'tetwgt'))
self.ntet = int(lines[0].split()[0])
self.nbands = int(lines[0].split()[1])
# nspin = int(lines[0][2]) #haven't added spin yet
self.fwgt = zeros((self.ntet,self.nbands),dtype = float64)
for it in range(self.ntet):
for ib in range(self.nbands):
self.fwgt[it,ib] = lines[it+1].split()[ib+1] #First line and first column are not weights
def rd_ibzkpt(self,dir): #only extracts the tets ids
self.idk = zeros((4,self.ntet),dtype = int)
self.vwgt = zeros(self.ntet,dtype = float64)
lines = nstrip(readfile(dir+'IBZKPT'))
nk = int(lines[1].split()[0])
ntet2 = int(lines[4+nk].split()[0])
self.volscale = float(lines[4+nk].split()[1])
print 'VASP volume scale', self.volscale
if ntet2 != self.ntet: print 'Warning! NTET in IBZKPTS differs from this program''s'
for it in range(self.ntet):
self.vwgt[it] = lines[5+nk+it].split()[0]
self.idk[:,it] = lines[5+nk+it].split()[1:]
self.idk = self.idk -1 #(converting from Fortran counting to python)
def get_tetvecs(self,dir):
lines = nstrip(readfile(dir+'tetvecs'))
self.vecs = zeros((3,4,self.ntet),dtype = float64)
self.vol = zeros(self.ntet,dtype = float64)
print shape(self.vecs)
for it in range(self.ntet):
for ic in range(4):
print it, ic,lines[5*it+ic+1]
self.vecs[:,ic,it] = lines[5*it+ic+1].split()
va = self.vecs[:,1,it]-self.vecs[:,0,it]
vb = self.vecs[:,2,it]-self.vecs[:,0,it]
vc = self.vecs[:,3,it]-self.vecs[:,0,it]
self.vol[it] = abs(dot(va,cross(vb,vc))/2.0)
def incar(self,dir):
print 'reading ISPIN and LORBIT from INCAR'
lines = nstrip(readfile(dir+'INCAR'))
for line in lines:
if ('ISPIN' in line) or ('ispin' in line): self.nspin = int(line.split('=')[1])
if ('LORBIT' in line) or ('lorbit' in line):
lorbit = int(line.split('=')[1])
if lorbit in [0,10]: #no lm decomposition
self.orbs = ['s','p','d']
elif lorbit in [1,2,11,12]:
self.orbs = ['s','py','pz','px','dxy','dyz','dz2','dxz','dx2']
else:
sys.exit('Unrecognized LORBIT in INCAR')
self.norbs = len(self.orbs)
def read_eigenval(dir):
'''Read in k vectors and eigenvalues from vasp file'''
os.chdir(dir)
eigs = readfile('EIGENVAL')
nb = int(eigs[5].split()[2])
nk = int(eigs[5].split()[1])
print nb, nk
ks = zeros((nk,3))
eners = zeros((nk,nb))
for ik in range(nk):
istart = 7 + ik*(nb+2)
ks[ik,:] = [float(eigs[istart].split()[0]), float(eigs[istart].split()[1]), float(eigs[istart].split()[2])]
for ib in range(nb):
eners[ik,ib] = float(eigs[istart+ib+1].split()[1])
return [nb,nk,ks,eners]
def read_eigenval(dir):
'''Read in k vectors and eigenvalues from vasp file'''
os.chdir(dir)
eigs = readfile('EIGENVAL')
nb = int(eigs[5].split()[2])
nk = int(eigs[5].split()[1])
print nb, nk
ks = zeros((nk,3))
eners = zeros((nk,nb))
for ik in range(nk):
istart = 7 + ik*(nb+2)
ks[ik,:] = [float(eigs[istart].split()[0]), float(eigs[istart].split()[1]), float(eigs[istart].split()[2])]
for ib in range(nb):
eners[ik,ib] = float(eigs[istart+ib+1].split()[1])
return [nb,nk,ks,eners]
def incar(self, dir):
print 'reading ISPIN and LORBIT from INCAR'
lines = nstrip(readfile(dir + 'INCAR'))
for line in lines:
if ('ISPIN' in line) or ('ispin' in line):
self.nspin = int(line.split('=')[1])
if ('LORBIT' in line) or ('lorbit' in line):
lorbit = int(line.split('=')[1])
if lorbit in [0, 10]: #no lm decomposition
self.orbs = ['s', 'p', 'd']
elif lorbit in [1, 2, 11, 12]:
self.orbs = [
's', 'py', 'pz', 'px', 'dxy', 'dyz', 'dz2', 'dxz',
'dx2'
]
else:
sys.exit('Unrecognized LORBIT in INCAR')
self.norbs = len(self.orbs)
def read(self, filepath):
print 'reading Procar',
lines = nstrip(readfile(filepath))
self.nk = int(lines[1].split(':')[1].split()[0])
self.nbands = int(lines[1].split(':')[2].split()[0])
self.nions = int(lines[1].split(':')[3])
#initialize
self.kvecs = zeros((3, self.nk), dtype=float)
self.kweights = zeros((self.nk), dtype=float)
self.ener = zeros((self.nk, self.nbands, self.nspin), dtype=float)
self.occ = zeros((self.nk, self.nbands, self.nspin), dtype=float)
self.weights = zeros(
(self.nk, self.nbands, self.nions, len(self.orbs), self.nspin),
dtype=float)
#read weights
ipos = 2
for ispin in range(self.nspin):
ipos += 1
# print 'ipos for spin %i: %i' % (ispin,ipos)
for ik in range(self.nk):
self.kvecs[:, ik] = lines[ipos].split()[3:6]
self.kweights[ik] = float(lines[ipos].split()[8])
ipos += 2
for ib in range(self.nbands):
self.ener[ik, ib, ispin] = float(lines[ipos].split()[4])
self.occ[ik, ib, ispin] = float(lines[ipos].split()[7])
ipos += 3
for i in range(self.nions):
if mod(ipos, 1e4) == 0: print '-', #progress bar
tempw = lines[ipos].split()[1:self.norbs + 1]
self.weights[ik, ib, i, :, ispin] = tempw
ipos += 1
ipos += 2
ipos += 1
print
]
N = len(dirs)
#x = zeros[len(dirs)]
#y = zeros[len(dirs)]
labels = ['No correction', 'Blochl correction']
if len(labels) != len(dirs):
sys.exit(
'Number of graphing labels is different from the number of datasets!\n Stopping\n'
)
fig = figure()
ax1 = fig.add_subplot(111)
for idir, dir in enumerate(dirs):
print idir
lines = readfile(dir + file)
x = [lines[i].strip().split()[0] for i in range(len(lines))]
y = [lines[i].strip().split()[1] for i in range(len(lines))]
ax1.semilogy(x,
y,
color=cm.jet(1. * (idir + 1) / N),
linestyle='None',
marker='o',
label=labels[idir])
#ylim((1e-12,1e0))
legend(loc='upper right', prop={'size': 12})
title('Al:Al Tetrahedral methods\n Cubic mesh')
xlabel('n')
ylabel('error (eV)')
os.chdir(dirs[0])
os.chdir('../')
writefile([str(j)+'\n' for j in dets],'dets%s' % structi)
writefile([str(j)+'\n' for j in err],'err%s' % structi)
writefile([str(j)+'\n' for j in ef_err],'ef_err%s' % structi)
# log err vs n for all structs
os.chdir(tempdir)
fig = figure()
ax1 = fig.add_subplot(111)
# ax1.set_color_cycle(['r','b','g','c', 'm', 'y', 'k'])
xlabel('n in cubic grid')
ylabel('Error (eV)')
title('Structure noise '+title_detail+':\nTheoretical values: max n on struct 1; 1e-8 mark')
xlim((0,55))
#ylim((1e-12,1e0))
for i,structi in enumerate(structselect):
nlist = readfile('ns%s' % structi)
errlist = readfile('err%s' % structi)
ax1.semilogy(nlist, errlist,label=structi,linestyle='None',color=cm.jet(1.*(i+1)/len(structselect)), marker = 'o') # marker = 'o',
plt.legend(loc='upper right',prop={'size':14});
show()
fig.savefig('log_err_vs_n_structs')
# for structi in structselect:
# for arr in []
#
# #en_per_atom vs ns
# titleadd = ''+ title_detail
from analysisToolsVasp import readfile
file = 'structs.cubmesh/errlist_vs_n'
dirs = [
'/fslhome/bch/cluster_expansion/alal/equivk_f-16.tetra.noBlochl/', '/fslhome/bch/cluster_expansion/alal/equivk_f1-6.tetra/'
]
N = len(dirs)
#x = zeros[len(dirs)]
#y = zeros[len(dirs)]
labels = ['No correction','Blochl correction']
if len(labels) != len(dirs): sys.exit('Number of graphing labels is different from the number of datasets!\n Stopping\n')
fig = figure()
ax1 = fig.add_subplot(111)
for idir,dir in enumerate(dirs):
print idir
lines = readfile(dir+file)
x = [lines[i].strip().split()[0] for i in range(len(lines))]
y = [lines[i].strip().split()[1] for i in range(len(lines))]
ax1.semilogy(x, y,color=cm.jet(1.*(idir+1)/N),linestyle = 'None', marker = 'o', label=labels[idir])
#ylim((1e-12,1e0))
legend(loc='upper right',prop={'size':12});
title('Al:Al Tetrahedral methods\n Cubic mesh')
xlabel('n')
ylabel('error (eV)')
os.chdir(dirs[0]);os.chdir('../')
print dirs
fig.savefig('err_vs_n_multi')
print 'done'
if os.path.isdir(d) and d == filter2
])
for struct in structs:
if collateMeshMat: meshPlots.write('(* {} *)\n'.format(struct))
os.chdir(struct)
iplot += 1
calcs = sorted(
[d for d in os.listdir(os.getcwd()) if os.path.isdir(d)])
if len(calcs) > maxCalcs: maxCalcs = len(calcs)
if collateMeshMat:
for ic, calc in enumerate(calcs):
os.chdir(calc)
if ic == 0: os.system('cp bounds ../')
if os.path.exists('cell_IBZmesh.m'):
meshPlots.write('\t(* {} *)\n'.format(calc))
lines = readfile('cell_IBZmesh.m')
lines.append('\n\n')
meshPlots.writelines(lines)
os.chdir('../')
os.chdir(path)
if collateMeshMat:
meshPlots.close()
#external run paths are of the form extmethodpath/atom_convergence/11_atom
if not extpaths is None:
for ipath, extpath in enumerate(extpaths):
os.chdir(extpath)
method = extpath.split('/')[0]
if collateMeshMat:
meshPlots = open('IBZmeshPlots', 'w')
atomdirs = sorted([
latticeType, packingFraction, mink_reduce, lattvec_u,arenormal,\
unique_anorms,points_in_ppiped #,three_perp,
from numpy import array, arccos, dot, cross, pi, floor, sum, sqrt, exp, log, asarray
from numpy import transpose, rint, inner, multiply, size, argmin, argmax, nonzero, float64, identity
from numpy import ceil, real, unravel_index
from scipy.optimize import minimize
from copy import copy, deepcopy
fprec = float64
from numpy import zeros #use arrays, not "matrix" class
#from numpy.matlib import zeros, matrix #creates np.matrix rather than array, but limited to 2-D!!!! uses *, but array uses matrixmultiply
from numpy.linalg import norm, det, inv, eig
from numpy.random import randint, random
from itertools import combinations
sys.path.append(
'/bluehome2/bch/pythonscripts/cluster_expansion/analysis_scripts/')
from analysisToolsVasp import writeEnergiesOszicar, writedirnames, nstrip, writeNk, writeNkIBZ, \
writeElConverge, writeElSteps, writeCPUtime, enerparts, getdata, readfile, writefile, \
getms, writefermi, removezeros,getEf
'''reads chgcar from dir1 and multiplies by 2 (for superlattice of volume 2 along z axis),
and writes it twice into CHGCAR for dir2 '''
dir1 = '/fslhome/bch/cluster_expansion/alal/cubic_al/mp_c1,3/c1_8x8x8/BANDS/'
dir2 = '/fslhome/bch/cluster_expansion/alal/cubic_al/mp_c1,3/c3_8x8x4/BANDS/'
lines1 = readfile(dir1 + 'CHGCAR')
lines2 = readfile(dir2 + 'CHGCAR')
print 'Done'
def analyze(paths): #as used with the parameter search, paths will have only one entry. But keep consistent with interactive vaspoutCombineRunsExtData
extpath = None
useSym = False
coloring = 'method'
# coloring = 'indiv'
doLegend = True
doLabel = True
smoothFactor = 2.0
filter = '_' #string must be in dir name to be included
filter2 = None #'Cu_1' #for single structures. set to None if using filter1 only
summaryPath = paths[0]
#count the number of plots:
iplot = 0
maxCalcs = 0
maxNk = 0
methods = []
for ipath,path in enumerate(paths):
method = path.split('_')[-1].split('/')[0]
methods.append(method)
os.chdir(path)
if filter2 == None:
structs = sorted([d for d in os.listdir(os.getcwd()) if os.path.isdir(d) and filter in d])
else:
structs = sorted([d for d in os.listdir(os.getcwd()) if os.path.isdir(d) and d==filter2])
for struct in structs:
os.chdir(struct)
iplot += 1
calcs = sorted([d for d in os.listdir(os.getcwd()) if os.path.isdir(d)])
if len(calcs)>maxCalcs: maxCalcs = len(calcs)
os.chdir(path)
#external data is of the form extpath/atom_method/struct.csv. The csv has energies vs nK
if not extpath is None:
os.chdir(extpath)
atoms_methods = sorted([d for d in os.listdir(extpath) if os.path.isdir(d) and filter in d])# os.chdir(extpath)
for atom_method in atoms_methods:
atom = atom_method.split('_')[0]
os.chdir(atom_method)
os.system('rm -r .*lock*')
for structfile in os.listdir(os.getcwd()):
if atom not in structfile:
os.system('mv {} {}_{}'.format(structfile,atom,structfile)) #so that file has atom name at beginning
if filter2 == None:
structfiles = sorted([d for d in os.listdir(os.getcwd()) if os.path.getsize(d)>0])
else:
structfiles = sorted([d for d in os.listdir(os.getcwd()) if '_'.join(d.split('_')[:2])==filter2 and os.path.getsize(d)>0])
for structfile in structfiles:
iplot += 1
#count number of points in this structfile
lines = readfile(structfile)
if len(lines)>maxCalcs: maxCalcs = len(lines)
os.chdir(extpath)
nplots = iplot
if nplots < len(paths): sys.exit('Stop. Structures do not match filter')
data = zeros(nplots,dtype = [('ID', 'S25'),('color', 'S15'),('method', 'S15'),\
('nDone','int32'),('nAtoms','int32'),('nops','int8'),\
('IBZvolcut','float'),('IBZvol','float'),\
('eners', '{}float'.format(maxCalcs)), ('errs', '{}float'.format(maxCalcs)),\
('nKs', '{}int16'.format(maxCalcs)),('ns', '{}int8'.format(maxCalcs))])
# style.use('bmh')
# for i, item in enumerate(rcParams['axes.prop_cycle']):
# colorsList.append(item['color'])
style.use('fivethirtyeight')
# for i, item in enumerate(rcParams['axes.prop_cycle'][:-2]):
# colorsList.append(item['color'])
colorsList = [u'#30a2da', u'#fc4f30', u'#e5ae38', u'#6d904f', u'#8b8b8b',
u'#348ABD', u'#A60628', u'#7A68A6', u'#467821', u'#D55E00',
u'#CC79A7', u'#56B4E9', u'#009E73', u'#F0E442', u'#0072B2']
colorsList = colorsList + ['b','m','y','c','k']
rcParams.update({'figure.autolayout': True})
rcParams['axes.facecolor'] = 'white'
rcParams['axes.linewidth'] = 1.0
rcParams['axes.edgecolor'] = 'black' # axisbg=axescolor
rcParams['savefig.facecolor'] = 'white' # axisbg=axescolor
rcParams['lines.markersize'] = 4.5
#read all the data
iplot = -1
for ipath, path in enumerate(paths): #my data
tag = path.split('/')[-1][-7:]
os.chdir(path)
if filter2 == None:
structs = sorted([d for d in os.listdir(os.getcwd()) if os.path.isdir(d) and filter in d])
else:
structs = sorted([d for d in os.listdir(os.getcwd()) if os.path.isdir(d) and d==filter2])
nStructs = len(structs)
# print structs,path
for istruct,struct in enumerate(structs):
# print 'test', istruct, struct
# print 'struct',struct
os.chdir(struct)
if coloring == 'indiv':
# if iplot < nplots -1:
color = rgb2hex(cm.jet(1.*(iplot+1)/float(nplots)))
# else:
# color = 'k'
elif coloring == 'method':
# color = colorsList[ipath]
color = None
calcs = sorted([d for d in os.listdir(os.getcwd()) if os.path.isdir(d) and os.path.exists('{}/OUTCAR'.format(d))])
energies = []
nKs = []
ns = [] #the base n of the run run
nDone = 0
if useSym:
try:
nops,IBZvolcut,IBZvol = readSym(calcs[0])
except:
sys.exit('Stopping. readSym failed. Set useSym to False')
for calc in calcs:
if electronicConvergeFinish(calc):
ener = getEnergy(calc) #in energy/atom
if not areEqual(ener,0,1e-5):
nDone +=1
energies.append(ener)
if 'vc' in path:
nK = getNkIBZ(calc,'KPOINTS')
else:
nK = getNkIBZ(calc,'IBZKPT')
if nK > maxNk: maxNk = nK
nKs.append(nK)
ns.append(int(calc.split('_')[-1]))
#sort by increasing number of kpoints
if len(energies)>0:
iplot += 1
nKs = array(nKs)
energies = array(energies)
ns = array(ns)
order = argsort(nKs)
# print 'struct',struct
# print 'energies',energies
energies = energies[order]
ns = ns[order]
nKs = sort(nKs)
eref = energies[-1]#the last energy of each struct is that of the most kpoints
errs = abs(energies-eref)*1000 + 1e-4 #now in meV
data[iplot]['ID'] = '{} {}'.format(struct,tag)
nAtoms = getNatoms('{}/POSCAR'.format(calc))
data[iplot]['nAtoms'] = nAtoms
if useSym:
data[iplot]['nops'] = nops
data[iplot]['IBZvolcut'] = IBZvolcut
data[iplot]['nDone'] = nDone
data[iplot]['eners'][:nDone] = energies
data[iplot]['errs'][:nDone] = errs
data[iplot]['nKs'][:nDone] = nKs
data[iplot]['ns'][:nDone] = ns
data[iplot]['color'] = color
method = path.split('_')[-1].split('/')[0]
data[iplot]['method'] = method
os.chdir(path)
# os.chdir(extpath)
if not extpath is None:
os.chdir(extpath)
# print; print atoms_methods
for atom_method in atoms_methods:
os.chdir(atom_method)
if coloring == 'method':
color = None
if 'MP' in atom_method:
# color = colorsList[len(paths)]
method = 'MP'
elif 'Mueller' in atom_method:
# color = colorsList[len(paths)+1]
method = 'Mueller'
if method not in methods:
methods.append(method)
if filter2 == None:
structfiles = sorted([d for d in os.listdir(os.getcwd()) if os.path.getsize(d)>0])
else:
structfiles = sorted([d for d in os.listdir(os.getcwd()) if '_'.join(d.split('_')[:2])==filter2 and os.path.getsize(d)>0])
for structfile in structfiles:
if useSym:
nops,IBZvolcut,nAtoms = copyData(structfile,data)
if coloring == 'indiv':
if iplot < nplots -1:
color = cm.jet(1.*(iplot+1)/float(nplots))
else:
color = 'k'
iplot += 1
energies = []
nKs = []
lines = readfile(structfile)
for line in lines:
nK = int(line.split('\t')[0])
if nK > maxNk: maxNk = nK
nKs.append(nK)
energies.append(-float(line.split('\t')[1].split('\r')[0]))
nKs = array(nKs)
energies = array(energies)
nDone = len(energies)
order = argsort(nKs)
energies = energies[order]
eref = energies[-1]#the last energy of each struct is that of the most kpoints
nKs = sort(nKs)
errs = abs(energies-eref)*1000 + 1e-4 #now in meV
struct = '_'.join(structfile.split('_')[:2])
data[iplot]['ID'] = atom_method + struct
data[iplot]['nAtoms'] = nAtoms
if useSym:
data[iplot]['nops'] = nops
data[iplot]['IBZvolcut'] = IBZvolcut
data[iplot]['nDone'] = len(energies)
data[iplot]['eners'][:nDone] = energies
data[iplot]['errs'][:nDone] = errs
data[iplot]['nKs'][:nDone] = nKs
data[iplot]['color'] = color
data[iplot]['method'] = method
os.chdir(extpath)
nplots = iplot+1
lines = [' ID , nKIBZ , ener , err, nAtoms, nops,IBZcut\n']
for iplot in range(nplots):
n = data[iplot]['nDone']
for icalc in range(n):#data[iplot]['eners'][:n].tolist()
lines.append('{}_n{},{},{:15.12f},{:15.12f},{},{},{}\n'.format(data[iplot]['ID'],\
data[iplot]['ns'][icalc], data[iplot]['nKs'][icalc],\
data[iplot]['eners'][icalc],data[iplot]['errs'][icalc],\
data[iplot]['nAtoms'],data[iplot]['nops'],data[iplot]['IBZvolcut']))
writefile(lines,'{}/summary.csv'.format(summaryPath))
#plots
if maxNk > 1:
if filter[0] == '_':filter = '' #labels can't begin with _
# plotTypes = ['linear','loglog', 'loglinear'];ylabels = ['Vasp error energy/atom (eV)','Error (meV)','Error (meV)']
# print 'plot only loglog'
plotTypes = ['loglog']; ylabels = ['Error (meV)']
# plotTypes = []
xtext = 'N k-points'
for it,plotType in enumerate(plotTypes):
fig = figure()
ax1 = fig.add_subplot(111)
xlabel(xtext)
ylabel(ylabels[it])
# title('Convergence vs mesh method')
#ylim((1e-12,1e0))
oldmethod = ''
methods2 = []
for iplot in range(nplots):
labelStr = None
n = data[iplot]['nDone']
if coloring == 'method':
method = data[iplot]['method']
data[iplot]['color'] = colorsList[methods.index(method)]
if method != oldmethod and method not in methods2:
if doLabel: labelStr = '{} {}'.format(filter,data[iplot]['method'])
plotData(fig,summaryPath,data[iplot],n,plotType,filter,doLegend,labelStr)
oldmethod = method;labelStr = None
methods2.append(method)
else:
plotData(fig,summaryPath,data[iplot],n,plotType,filter,doLegend,labelStr)
elif coloring == 'indiv':
if doLabel: labelStr = '{} {}'.format(filter,data[iplot]['ID'])
plotData(data[iplot],n,plotType,filter,doLegend,labelStr)
#Method averaging
if coloring == 'method':
# print 'Averaging, plotting method errors'
nbins = int(10*ceil(log10(maxNk)))# 10 bins per decade
nKbins = array([(10.0**(1/10.0))**i for i in range(nbins)])
fig = figure()
ax1 = fig.add_subplot(111)
xlabel('N k-points (smoothed by factor {})'.format(int(smoothFactor)))
ylabel('Error (meV)')
methodCostsLogs = []
for im,method in enumerate(methods):
methnKmax = 0
binCounts = zeros(nbins,dtype = int32)
binErrs = zeros(nbins,dtype = float)
costLogs = zeros(nbins,dtype = float) # "Costs" relative to excellent Si Monkhorst Pack, which has err = 10^3/nK^3 + 10^-3 meV.
for iplot in range(nplots):
if data[iplot]['method'] == method:
for icalc in range(data[iplot]['nDone']-1):
nK = data[iplot]['nKs'][icalc]
if nK>methnKmax: methnKmax = nK
if nK>1:
for ibin in range(nbins):
if abs(log10(nK/nKbins[ibin])) <= log10(smoothFactor)\
and nKbins[ibin]<= maxNk:
binErrs[ibin] += data[iplot]['errs'][icalc]
costLogs[ibin] += log10(data[iplot]['errs'][icalc]/(10**3/(nK**3.0)+0.001))
binCounts[ibin] += 1
mask = where(binCounts>0)
binErrs2 = binErrs[mask[0]]
binCounts2 = binCounts[mask[0]]
nKbins2 = nKbins[mask[0]]
costLogs2 = costLogs[mask[0]]
nbins2 = len(nKbins2)
avgErrs = [binErrs2[ibin]/binCounts2[ibin] for ibin in range(nbins2)]
avgcostLogs = [costLogs2[ibin]/binCounts2[ibin] for ibin in range(nbins2)]
avgcostLins = [10**avgcostLogs[ibin] for ibin in range(nbins2)]
methodCostsLogs.append(mean(avgcostLogs))
loglog(nKbins2,avgErrs,label = method,\
color = colorsList[im], marker = None)
loglog(nKbins2,avgcostLins,label = None,\
color = colorsList[im], marker = None,linestyle=':')
# print 'Method',method, 'nKmax',methnKmax, 'avgLogCost', mean(avgcostLogs)
legend(loc='lower left',prop={'size':12});
fig.savefig('{}/methodErrs'.format(summaryPath))
close('all')
if maxNk > 1:
return [methodCostsLogs[0],mean(data['nDone'])] #there is only one method when running this routine
else:
return [100,0]
def weightfunc(FE,FEmax,FEmin):
maxratio = 1.0
return str(1.0 + (maxratio-1)*(FEmax-FE)/(FEmax-FEmin))
'''Reads formation energies in structures.in. If weighted, it changes the weights.
If not, it adds a line for each weight'''
#dir = '/fslhome/bch/cluster_expansion/graphene/csv1-8W/'
dir = '/fslhome/bch/cluster_expansion/graphene/test1/'
infile = dir+'structures.in.441.FEsorted'
#infile = dir+'structures.in'
outfile = dir+'structures.in'
FEs = []
indices = []
lines = readfile(infile)
if 'noweights' in lines[1]:
weights_in = False
lines[1] = 'weights\n'
else:
weights_in = True
#print 'weights_in', weights_in
#Find maximum and min formation energies
for i, line in enumerate(lines):
if 'FE' in line:
energy = float(line.split()[line.split().index('FE')+2].split(',')[0]) #FE = 1.2578, etc. remove any ","
FEs.append(energy)
if '#Energy' in line:
indices.append(i+2) #where weights should appear
FEmax = amax(FEs)
FEmin = amin(FEs)
def analyze(
paths
): #as used with the parameter search, paths will have only one entry. But keep consistent with interactive vaspoutCombineRunsExtData
extpath = None
useSym = False
coloring = 'method'
# coloring = 'indiv'
doLegend = True
doLabel = True
smoothFactor = 2.0
filter = '_' #string must be in dir name to be included
filter2 = None #'Cu_1' #for single structures. set to None if using filter1 only
summaryPath = paths[0]
#count the number of plots:
iplot = 0
maxCalcs = 0
maxNk = 0
methods = []
for ipath, path in enumerate(paths):
method = path.split('_')[-1].split('/')[0]
methods.append(method)
os.chdir(path)
if filter2 == None:
structs = sorted([
d for d in os.listdir(os.getcwd())
if os.path.isdir(d) and filter in d
])
else:
structs = sorted([
d for d in os.listdir(os.getcwd())
if os.path.isdir(d) and d == filter2
])
for struct in structs:
os.chdir(struct)
iplot += 1
calcs = sorted(
[d for d in os.listdir(os.getcwd()) if os.path.isdir(d)])
if len(calcs) > maxCalcs: maxCalcs = len(calcs)
os.chdir(path)
#external data is of the form extpath/atom_method/struct.csv. The csv has energies vs nK
if not extpath is None:
os.chdir(extpath)
atoms_methods = sorted([
d for d in os.listdir(extpath) if os.path.isdir(d) and filter in d
]) # os.chdir(extpath)
for atom_method in atoms_methods:
atom = atom_method.split('_')[0]
os.chdir(atom_method)
os.system('rm -r .*lock*')
for structfile in os.listdir(os.getcwd()):
if atom not in structfile:
os.system('mv {} {}_{}'.format(
structfile, atom,
structfile)) #so that file has atom name at beginning
if filter2 == None:
structfiles = sorted([
d for d in os.listdir(os.getcwd())
if os.path.getsize(d) > 0
])
else:
structfiles = sorted([
d for d in os.listdir(os.getcwd())
if '_'.join(d.split('_')[:2]) == filter2
and os.path.getsize(d) > 0
])
for structfile in structfiles:
iplot += 1
#count number of points in this structfile
lines = readfile(structfile)
if len(lines) > maxCalcs: maxCalcs = len(lines)
os.chdir(extpath)
nplots = iplot
if nplots < len(paths): sys.exit('Stop. Structures do not match filter')
data = zeros(nplots,dtype = [('ID', 'S25'),('color', 'S15'),('method', 'S15'),\
('nDone','int32'),('nAtoms','int32'),('nops','int8'),\
('IBZvolcut','float'),('IBZvol','float'),\
('eners', '{}float'.format(maxCalcs)), ('errs', '{}float'.format(maxCalcs)),\
('nKs', '{}int16'.format(maxCalcs)),('ns', '{}int8'.format(maxCalcs))])
# style.use('bmh')
# for i, item in enumerate(rcParams['axes.prop_cycle']):
# colorsList.append(item['color'])
style.use('fivethirtyeight')
# for i, item in enumerate(rcParams['axes.prop_cycle'][:-2]):
# colorsList.append(item['color'])
colorsList = [
u'#30a2da', u'#fc4f30', u'#e5ae38', u'#6d904f', u'#8b8b8b', u'#348ABD',
u'#A60628', u'#7A68A6', u'#467821', u'#D55E00', u'#CC79A7', u'#56B4E9',
u'#009E73', u'#F0E442', u'#0072B2'
]
colorsList = colorsList + ['b', 'm', 'y', 'c', 'k']
rcParams.update({'figure.autolayout': True})
rcParams['axes.facecolor'] = 'white'
rcParams['axes.linewidth'] = 1.0
rcParams['axes.edgecolor'] = 'black' # axisbg=axescolor
rcParams['savefig.facecolor'] = 'white' # axisbg=axescolor
rcParams['lines.markersize'] = 4.5
#read all the data
iplot = -1
for ipath, path in enumerate(paths): #my data
tag = path.split('/')[-1][-7:]
os.chdir(path)
if filter2 == None:
structs = sorted([
d for d in os.listdir(os.getcwd())
if os.path.isdir(d) and filter in d
])
else:
structs = sorted([
d for d in os.listdir(os.getcwd())
if os.path.isdir(d) and d == filter2
])
nStructs = len(structs)
# print structs,path
for istruct, struct in enumerate(structs):
# print 'test', istruct, struct
# print 'struct',struct
os.chdir(struct)
if coloring == 'indiv':
# if iplot < nplots -1:
color = rgb2hex(cm.jet(1. * (iplot + 1) / float(nplots)))
# else:
# color = 'k'
elif coloring == 'method':
# color = colorsList[ipath]
color = None
calcs = sorted([
d for d in os.listdir(os.getcwd())
if os.path.isdir(d) and os.path.exists('{}/OUTCAR'.format(d))
])
energies = []
nKs = []
ns = [] #the base n of the run run
nDone = 0
if useSym:
try:
nops, IBZvolcut, IBZvol = readSym(calcs[0])
except:
sys.exit('Stopping. readSym failed. Set useSym to False')
for calc in calcs:
if electronicConvergeFinish(calc):
ener = getEnergy(calc) #in energy/atom
if not areEqual(ener, 0, 1e-5):
nDone += 1
energies.append(ener)
if 'vc' in path:
nK = getNkIBZ(calc, 'KPOINTS')
else:
nK = getNkIBZ(calc, 'IBZKPT')
if nK > maxNk: maxNk = nK
nKs.append(nK)
ns.append(int(calc.split('_')[-1]))
#sort by increasing number of kpoints
if len(energies) > 0:
iplot += 1
nKs = array(nKs)
energies = array(energies)
ns = array(ns)
order = argsort(nKs)
# print 'struct',struct
# print 'energies',energies
energies = energies[order]
ns = ns[order]
nKs = sort(nKs)
eref = energies[
-1] #the last energy of each struct is that of the most kpoints
errs = abs(energies - eref) * 1000 + 1e-4 #now in meV
data[iplot]['ID'] = '{} {}'.format(struct, tag)
nAtoms = getNatoms('{}/POSCAR'.format(calc))
data[iplot]['nAtoms'] = nAtoms
if useSym:
data[iplot]['nops'] = nops
data[iplot]['IBZvolcut'] = IBZvolcut
data[iplot]['nDone'] = nDone
data[iplot]['eners'][:nDone] = energies
data[iplot]['errs'][:nDone] = errs
data[iplot]['nKs'][:nDone] = nKs
data[iplot]['ns'][:nDone] = ns
data[iplot]['color'] = color
method = path.split('_')[-1].split('/')[0]
data[iplot]['method'] = method
os.chdir(path)
# os.chdir(extpath)
if not extpath is None:
os.chdir(extpath)
# print; print atoms_methods
for atom_method in atoms_methods:
os.chdir(atom_method)
if coloring == 'method':
color = None
if 'MP' in atom_method:
# color = colorsList[len(paths)]
method = 'MP'
elif 'Mueller' in atom_method:
# color = colorsList[len(paths)+1]
method = 'Mueller'
if method not in methods:
methods.append(method)
if filter2 == None:
structfiles = sorted([
d for d in os.listdir(os.getcwd())
if os.path.getsize(d) > 0
])
else:
structfiles = sorted([
d for d in os.listdir(os.getcwd())
if '_'.join(d.split('_')[:2]) == filter2
and os.path.getsize(d) > 0
])
for structfile in structfiles:
if useSym:
nops, IBZvolcut, nAtoms = copyData(structfile, data)
if coloring == 'indiv':
if iplot < nplots - 1:
color = cm.jet(1. * (iplot + 1) / float(nplots))
else:
color = 'k'
iplot += 1
energies = []
nKs = []
lines = readfile(structfile)
for line in lines:
nK = int(line.split('\t')[0])
if nK > maxNk: maxNk = nK
nKs.append(nK)
energies.append(-float(line.split('\t')[1].split('\r')[0]))
nKs = array(nKs)
energies = array(energies)
nDone = len(energies)
order = argsort(nKs)
energies = energies[order]
eref = energies[
-1] #the last energy of each struct is that of the most kpoints
nKs = sort(nKs)
errs = abs(energies - eref) * 1000 + 1e-4 #now in meV
struct = '_'.join(structfile.split('_')[:2])
data[iplot]['ID'] = atom_method + struct
data[iplot]['nAtoms'] = nAtoms
if useSym:
data[iplot]['nops'] = nops
data[iplot]['IBZvolcut'] = IBZvolcut
data[iplot]['nDone'] = len(energies)
data[iplot]['eners'][:nDone] = energies
data[iplot]['errs'][:nDone] = errs
data[iplot]['nKs'][:nDone] = nKs
data[iplot]['color'] = color
data[iplot]['method'] = method
os.chdir(extpath)
nplots = iplot + 1
lines = [' ID , nKIBZ , ener , err, nAtoms, nops,IBZcut\n']
for iplot in range(nplots):
n = data[iplot]['nDone']
for icalc in range(n): #data[iplot]['eners'][:n].tolist()
lines.append('{}_n{},{},{:15.12f},{:15.12f},{},{},{}\n'.format(data[iplot]['ID'],\
data[iplot]['ns'][icalc], data[iplot]['nKs'][icalc],\
data[iplot]['eners'][icalc],data[iplot]['errs'][icalc],\
data[iplot]['nAtoms'],data[iplot]['nops'],data[iplot]['IBZvolcut']))
writefile(lines, '{}/summary.csv'.format(summaryPath))
#plots
if maxNk > 1:
if filter[0] == '_': filter = '' #labels can't begin with _
# plotTypes = ['linear','loglog', 'loglinear'];ylabels = ['Vasp error energy/atom (eV)','Error (meV)','Error (meV)']
# print 'plot only loglog'
plotTypes = ['loglog']
ylabels = ['Error (meV)']
# plotTypes = []
xtext = 'N k-points'
for it, plotType in enumerate(plotTypes):
fig = figure()
ax1 = fig.add_subplot(111)
xlabel(xtext)
ylabel(ylabels[it])
# title('Convergence vs mesh method')
#ylim((1e-12,1e0))
oldmethod = ''
methods2 = []
for iplot in range(nplots):
labelStr = None
n = data[iplot]['nDone']
if coloring == 'method':
method = data[iplot]['method']
data[iplot]['color'] = colorsList[methods.index(method)]
if method != oldmethod and method not in methods2:
if doLabel:
labelStr = '{} {}'.format(filter,
data[iplot]['method'])
plotData(fig, summaryPath, data[iplot], n, plotType,
filter, doLegend, labelStr)
oldmethod = method
labelStr = None
methods2.append(method)
else:
plotData(fig, summaryPath, data[iplot], n, plotType,
filter, doLegend, labelStr)
elif coloring == 'indiv':
if doLabel:
labelStr = '{} {}'.format(filter, data[iplot]['ID'])
plotData(data[iplot], n, plotType, filter, doLegend,
labelStr)
#Method averaging
if coloring == 'method':
# print 'Averaging, plotting method errors'
nbins = int(10 * ceil(log10(maxNk))) # 10 bins per decade
nKbins = array([(10.0**(1 / 10.0))**i for i in range(nbins)])
fig = figure()
ax1 = fig.add_subplot(111)
xlabel('N k-points (smoothed by factor {})'.format(
int(smoothFactor)))
ylabel('Error (meV)')
methodCostsLogs = []
for im, method in enumerate(methods):
methnKmax = 0
binCounts = zeros(nbins, dtype=int32)
binErrs = zeros(nbins, dtype=float)
costLogs = zeros(
nbins, dtype=float
) # "Costs" relative to excellent Si Monkhorst Pack, which has err = 10^3/nK^3 + 10^-3 meV.
for iplot in range(nplots):
if data[iplot]['method'] == method:
for icalc in range(data[iplot]['nDone'] - 1):
nK = data[iplot]['nKs'][icalc]
if nK > methnKmax: methnKmax = nK
if nK > 1:
for ibin in range(nbins):
if abs(log10(nK/nKbins[ibin])) <= log10(smoothFactor)\
and nKbins[ibin]<= maxNk:
binErrs[ibin] += data[iplot]['errs'][
icalc]
costLogs[ibin] += log10(
data[iplot]['errs'][icalc] /
(10**3 / (nK**3.0) + 0.001))
binCounts[ibin] += 1
mask = where(binCounts > 0)
binErrs2 = binErrs[mask[0]]
binCounts2 = binCounts[mask[0]]
nKbins2 = nKbins[mask[0]]
costLogs2 = costLogs[mask[0]]
nbins2 = len(nKbins2)
avgErrs = [
binErrs2[ibin] / binCounts2[ibin] for ibin in range(nbins2)
]
avgcostLogs = [
costLogs2[ibin] / binCounts2[ibin]
for ibin in range(nbins2)
]
avgcostLins = [10**avgcostLogs[ibin] for ibin in range(nbins2)]
methodCostsLogs.append(mean(avgcostLogs))
loglog(nKbins2,avgErrs,label = method,\
color = colorsList[im], marker = None)
loglog(nKbins2,avgcostLins,label = None,\
color = colorsList[im], marker = None,linestyle=':')
# print 'Method',method, 'nKmax',methnKmax, 'avgLogCost', mean(avgcostLogs)
legend(loc='lower left', prop={'size': 12})
fig.savefig('{}/methodErrs'.format(summaryPath))
close('all')
if maxNk > 1:
return [methodCostsLogs[0], mean(data['nDone'])
] #there is only one method when running this routine
else:
return [100, 0]
structs = sorted([d for d in os.listdir(os.getcwd()) if os.path.isdir(d) and filter in d])
else:
structs = sorted([d for d in os.listdir(os.getcwd()) if os.path.isdir(d) and d==filter2])
for struct in structs:
if collateMeshMat:meshPlots.write('(* {} *)\n'.format(struct))
os.chdir(struct)
iplot += 1
calcs = sorted([d for d in os.listdir(os.getcwd()) if os.path.isdir(d)])
if len(calcs) > maxCalcs: maxCalcs = len(calcs)
if collateMeshMat:
for ic, calc in enumerate(calcs):
os.chdir(calc)
if ic == 0: os.system('cp bounds ../')
if os.path.exists('cell_IBZmesh.m'):
meshPlots.write('\t(* {} *)\n'.format(calc))
lines = readfile('cell_IBZmesh.m')
lines.append( '\n\n')
meshPlots.writelines(lines)
os.chdir('../')
os.chdir(path)
if collateMeshMat:
meshPlots.close()
#external run paths are of the form extmethodpath/atom_convergence/11_atom
if not extpaths is None:
for ipath,extpath in enumerate(extpaths):
os.chdir(extpath)
method = extpath.split('/')[0]
if collateMeshMat:
meshPlots = open('IBZmeshPlots','w')
atomdirs = sorted([d for d in os.listdir(extpath) if os.path.isdir(d) and filter in d])# os.chdir(extpath)
'''Reads formation energies in structures.in or .holdout
Ignores weights, and writes for energy a scaled formation energy that spreads differences in FE for low energy,
and compresses differences in FE for high FE'''
#dir = '/fslhome/bch/cluster_expansion/graphene/csv1-8W/'
dir = '/fslhome/bch/cluster_expansion/graphene/test1/'
#infile = dir+'structures.in.441.FEsorted'
infile = dir + 'structures.holdout'
#infile = dir+'structures.in'
#outfile = dir+'structures.in'
outfile = dir + 'structures.holdout'
FEs = []
indices = []
lines = readfile(infile)
#Find maximum and min formation energies
for i, line in enumerate(lines):
if 'FE' in line:
energy = float(
line.split()[line.split().index('FE') +
2].split(',')[0]) #FE = 1.2578, etc. remove any ","
FEs.append(energy)
if '#Energy' in line:
indices.append(i) #where label 'Energy' appeared
FEmax = amax(FEs)
FEmin = amin(FEs)
indices = array(indices)
#Write new Fes
latticeType, packingFraction, mink_reduce, lattvec_u,arenormal,\
unique_anorms,points_in_ppiped #,three_perp,
from numpy import array, arccos, dot, cross, pi, floor, sum, sqrt, exp, log, asarray
from numpy import transpose,rint,inner,multiply,size,argmin,argmax,nonzero,float64, identity
from numpy import ceil,real,unravel_index
from scipy.optimize import minimize
from copy import copy,deepcopy
fprec=float64
from numpy import zeros #use arrays, not "matrix" class
#from numpy.matlib import zeros, matrix #creates np.matrix rather than array, but limited to 2-D!!!! uses *, but array uses matrixmultiply
from numpy.linalg import norm, det, inv, eig
from numpy.random import randint,random
from itertools import combinations
sys.path.append('/bluehome2/bch/pythonscripts/cluster_expansion/analysis_scripts/')
from analysisToolsVasp import writeEnergiesOszicar, writedirnames, nstrip, writeNk, writeNkIBZ, \
writeElConverge, writeElSteps, writeCPUtime, enerparts, getdata, readfile, writefile, \
getms, writefermi, removezeros,getEf
'''reads chgcar from dir1 and multiplies by 2 (for superlattice of volume 2 along z axis),
and writes it twice into CHGCAR for dir2 '''
dir1 = '/fslhome/bch/cluster_expansion/alal/cubic_al/mp_c1,3/c1_8x8x8/BANDS/'
dir2 = '/fslhome/bch/cluster_expansion/alal/cubic_al/mp_c1,3/c3_8x8x4/BANDS/'
lines1 = readfile(dir1 + 'CHGCAR')
lines2 = readfile(dir2 + 'CHGCAR')
print 'Done'