efbest = float(efermis[ibest])
print 'ebest', ebest
print 'e-fermi best', efbest
err = abs(en_per_atom-ebest)
ef_err = abs(efermis - efbest)
ns = [ms[j]*dets[j] for j in range(len(dets))]
#plots
[[ns,ms,en_per_atom,efermis,NkIBZ,NkfullBZ,dets],zerolist] = removezeros([ns,ms,en_per_atom,efermis,NkIBZ,NkfullBZ,dets])#for structures that were not finished and have zero entries
# ebest = en_per_atom[argmax(ns)]
# efbest = efermis[argmax(ns)]
err = abs(en_per_atom-ebest)+1e-8 #do these again with only the finished runs
ef_err = abs(efermis - efbest)+1e-8
#write important arrays to file
writefile([str(j)+'\n' for j in ns],'ns%s' % structi)
writefile([str(j)+'\n' for j in ms],'ms%s' % structi)
writefile([str(j)+'\n' for j in NkIBZ],'NkIBZ%s' % structi)
writefile([str(j)+'\n' for j in NkfullBZ],'NkfullBZ%s' % structi)
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')
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)
indices = array(indices)
#Write weights
for i in range(len(FEs)):
if weights_in:
lines[indices[i]-2] = '#Energy and weight'+'\n'
lines[indices[i]] = str(weightfunc(FEs[i],FEmax,FEmin))+'\n'
else: #have to add lines
lines.insert(indices[i],str(weightfunc(FEs[i],FEmax,FEmin))+'\n')
indices += 1 #increment all indices by one to handle new lines
writefile(lines,outfile)
print 'Number of structures weighted', len(FEs)
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]
# plotErrs()
data[iplot]['ID'] = '{} {}'.format(struct,meshMethod)
nAtoms = getNatoms('{}/POSCAR'.format(calc))
data[iplot]['nAtoms'] = nAtoms
data[iplot]['nDone'] = nDone
data[iplot]['eners'][:nDone] = energies
data[iplot]['errs'][:nDone] = errs
data[iplot]['nKs'][:nDone] = nKs
os.chdir(maindir)
lines = [' ID , nKIBZ , ener , err, nAtoms \n']
for iplot in range(nplots):
n = data[iplot]['nDone']
for icalc in range(n):#data[iplot]['eners'][:n].tolist()
lines.append('{},{},{:15.12f},{:15.12f},{}\n'.format(data[iplot]['ID'], data[iplot]['nKs'][icalc],\
data[iplot]['eners'][icalc],data[iplot]['errs'][icalc],data[iplot]['nAtoms']))
writefile(lines,'{}/summary.csv'.format(summaryPath))
fig = figure()
ax1 = fig.add_subplot(111)
# ax1.set_color_cycle(['r','b','g','c', 'm', 'y', 'k'])
xlabel('N kpoints')
ylabel('Vasp energy/atom (eV)')
title('Convergence vs mesh method')
#ylim((1e-12,1e0))
for iplot in range(nplots):
if iplot < nplots -1:
plotcolor = cm.jet(1.*(iplot+1)/float(nplots))
else:
plotcolor = 'k'
n = data[iplot]['nDone']
# print 'iplot',data[iplot]['eners'][:n], data[iplot]['nKs'][:n]
plot(data[iplot]['nKs'][:n],data[iplot]['eners'][:n],\
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]
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
for i in range(len(FEs)):
lines[indices[i]] = '#Energy (scaled formation energy)' + '\n'
lines[indices[i] + 1] = str(scaledFE(FEs[i], FEmax, FEmin)) + '\n'
writefile(lines, outfile)
print 'Number of structures with formation energies scaled', len(FEs)
print 'Done'
os.chdir('../')
os.chdir(extpath)
if collateMeshMat:
meshPlots.close()
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('{},{},{:15.12f},{:15.12f},{},{},{}\n'.format(data[iplot]['ID'],\
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 filter[0] == '_': filter = '' #labels can't begin with _
# plotTypes = ['linear','loglog'] #loglinear
# print 'plot only loglog'
plotTypes = ['loglog'] #loglinear
# plotTypes = []
ylabels = ['Vasp energy/atom (eV)', 'Error (meV)', 'Error (meV)']
xtext = 'N k-points'
for it, plotType in enumerate(plotTypes):
fig = figure()
ax1 = fig.add_subplot(111)
xlabel(xtext)
ylabel(ylabels[it])
# ebest = en_per_atom[argmax(ns)]
# efbest = efermis[argmax(ns)]
names2 = deepcopy(names); names = delete(names2,zerolist,axis=0)
parts2 = deepcopy(parts); parts2 = delete(parts,zerolist,axis=0)
parts3 = array([parts2[i]/dets[i] for i in range(len(dets))])
ibestnew = nonzero(en_per_atom == ebest)[0][0]
print 'New index for ebest', ibestnew
# for plots
err = abs(en_per_atom-ebest)+1e-9 #do these again with only the finished runs
ef_err = abs(efermis - efbest)+1e-9
#Write to file to simplify plotting for comparisons with other runs
errlist = [str(ns[i])+' '+str(err[i])+'\n' for i in range(len(ns))]
print errlist
writefile(errlist,'errlist_vs_n')
#en_per_atom vs ns
titleadd = ''+ title_detail
# plotxy(ns,en_per_atom,'en_per_atom', titleadd + 'Vasp energy vs n (defines grid)','n','eV')
# print ns
# print err
fig = figure()
semilogy(ns,err,'ro')
title(titleadd + ' Error vs n (defines grid)')
xlabel('n')
ylabel('error')
fig.savefig('vary_n_log_err')