validprogressbar = tqdm(valid_data, total=valid_data.steps)
for x, y in validprogressbar:
lval, accval = clf.validate(x, y)
validlvals.append(lval)
validaccvals.append(accval)
metrics = [('loss', numpy.mean(validlvals)),
('acc', numpy.mean(validaccvals))]
desc = monitor(metrics, 4)
validprogressbar.set_description(desc=desc, refresh=True)
validlvals = []
validaccvals = []
clf.close('/Users/administrateur/ArnaudModules/neuralyzer/Logs/Ref/model.ckpt')
del archi, clf
archi1 = Classifier(brickname='dependency')
archi2 = Classifier(brickname='reference')
clf = SiameseCLF(
archi2,
archi1,
height=h,
width=w,
colors=c,
learning_rate=0.001,
ref_path=
'/Users/administrateur/ArnaudModules/neuralyzer/Logs/Ref/model.ckpt',
trainlvals.append(lval)
trainaccvals.append(accval)
metrics = [('loss', numpy.mean(trainlvals)), ('acc', numpy.mean(trainaccvals))]
desc = monitor(metrics, 4)
trainprogressbar.set_description(desc=desc, refresh=True)
trainlvals = []
trainaccvals = []
print('VALIDATE')
validprogressbar = tqdm(valid_data, total=valid_data.steps)
for x, y in validprogressbar:
lval, accval = clf.validate(x, y)
validlvals.append(lval)
validaccvals.append(accval)
metrics = [('loss', numpy.mean(validlvals)), ('acc', numpy.mean(validaccvals))]
desc = monitor(metrics, 4)
validprogressbar.set_description(desc=desc, refresh=True)
validlvals = []
validaccvals = []
dirname = os.path.join(logfolder, 'dependency' + str(n))
if not os.path.isdir(dirname):
os.makedirs(dirname)
clf.close(os.path.join(dirname, 'model.ckpt'))
trainlvals.append(lval)
trainaccvals.append(accval)
metrics = [('loss', numpy.mean(trainlvals)), ('acc', numpy.mean(trainaccvals))]
desc = monitor(metrics, 4)
trainprogressbar.set_description(desc=desc, refresh=True)
trainlvals = []
trainaccvals = []
print('VALIDATE')
validprogressbar = tqdm(valid_data, total=valid_data.steps)
for x, y in validprogressbar:
lval, accval = clf.validate(x, y)
validlvals.append(lval)
validaccvals.append(accval)
metrics = [('loss', numpy.mean(validlvals)), ('acc', numpy.mean(validaccvals))]
desc = monitor(metrics, 4)
validprogressbar.set_description(desc=desc, refresh=True)
accplot.append(numpy.mean(validaccvals))
validlvals = []
validaccvals = []
clf.close(os.path.join(outfolder, 'model.ckpt'))
with open(os.path.join(outfolder, 'accuracyplot.p'), 'wb') as f:
pickle.dump(accplot, f)
validprogressbar = tqdm(valid_data, total=valid_data.steps)
accvals = []
for x, y in validprogressbar:
_, accval = clf.validate(x, y)
accvals.append(accval)
metrics = [('acc', numpy.mean(accvals))]
desc = monitor(metrics, 4)
validprogressbar.set_description(desc=desc, refresh=True)
ref_accuracy = numpy.mean(accvals)
clf.close()
del clf, archidep, archiref
# then evaluate dependency alone
archidep = Classifier(brickname='dependency' + str(n),
dropouts=[0.5, 0.5],
fcdropouts=[0.5])
archiref = Classifier(brickname='reference',
dropouts=[0.5, 0.5],
fcdropouts=[0.5])
clf = CLF(archidep,
height=h,
width=w,
colors=c,
learning_rate=0.0001,