valmetric = TopKMisclassification(k=5)
# dummy optimizer for benchmarking
# training implementation coming soon
opt_gdm = GradientDescentMomentum(0.0, 0.0)
opt_biases = GradientDescentMomentum(0.0, 0.0)
opt = MultiOptimizer({'default': opt_gdm, 'Bias': opt_biases})
# setup cost function as CrossEntropy
cost = Multicost(costs=[GeneralizedCost(costfunc=CrossEntropyMulti()),
GeneralizedCost(costfunc=CrossEntropyMulti()),
GeneralizedCost(costfunc=CrossEntropyMulti())],
weights=[1, 0., 0.]) # We only want to consider the CE of the main path
assert os.path.exists(args.model_file), 'script requires the trained weights file'
model.load_params(args.model_file)
model.initialize(test, cost)
print 'running speed benchmark...'
model.benchmark(test, cost, opt)
print '\nCalculating performance on validation set...'
test.reset()
mets = model.eval(test, metric=valmetric)
print 'Validation set metrics:'
print 'LogLoss: %.2f, Accuracy: %.1f %% (Top-1), %.1f %% (Top-5)' % (mets[0],
(1.0-mets[1])*100,
(1.0-mets[2])*100)
save_path='serialize_test.pkl')
lr_sched = PolySchedule(total_epochs=10, power=0.5)
opt_gdm = GradientDescentMomentum(0.01, 0.9, wdecay=0.0002, schedule=lr_sched)
opt_biases = GradientDescentMomentum(0.02, 0.9, schedule=lr_sched)
opt = MultiOptimizer({'default': opt_gdm, 'Bias': opt_biases})
if not args.resume:
# fit the model for 3 epochs
model.fit(train,
optimizer=opt,
num_epochs=3,
cost=cost,
callbacks=callbacks)
train.reset()
# get 1 image
for im, l in train:
break
train.exit_batch_provider()
save_obj((im.get(), l.get()), 'im1.pkl')
im_save = im.get().copy()
if args.resume:
(im2, l2) = load_obj('im1.pkl')
im.set(im2)
l.set(l2)
# run fprop and bprop on this minibatch save the results
out_fprop = model.fprop(im)
out_fprop_save = [x.get() for x in out_fprop]
model = Model(layers=pdict, dataset=train)
# configure callbacks
callbacks = Callbacks(model, progress_bar=True, output_file='temp1.h5',
serialize=1, history=3, save_path='serialize_test.pkl')
lr_sched = PolySchedule(total_epochs=10, power=0.5)
opt_gdm = GradientDescentMomentum(0.01, 0.9, wdecay=0.0002, schedule=lr_sched)
opt_biases = GradientDescentMomentum(0.02, 0.9, schedule=lr_sched)
opt = MultiOptimizer({'default': opt_gdm, 'Bias': opt_biases})
if not args.resume:
# fit the model for 3 epochs
model.fit(train, optimizer=opt, num_epochs=3, cost=cost, callbacks=callbacks)
train.reset()
# get 1 image
for im, l in train:
break
train.exit_batch_provider()
with open('im1.pkl', 'w') as fid:
pickle.dump((im.get(), l.get()), fid)
im_save = im.get().copy()
if args.resume:
with open('im1.pkl', 'r') as fid:
(im2, l2) = pickle.load(fid)
im.set(im2)
l.set(l2)
# run fprop and bprop on this minibatch save the results
out_fprop = model.fprop(im)
# dummy optimizer for benchmarking
# training implementation coming soon
opt_gdm = GradientDescentMomentum(0.0, 0.0)
opt_biases = GradientDescentMomentum(0.0, 0.0)
opt = MultiOptimizer({'default': opt_gdm, 'Bias': opt_biases})
# setup cost function as CrossEntropy
cost = Multicost(
costs=[
GeneralizedCost(costfunc=CrossEntropyMulti()),
GeneralizedCost(costfunc=CrossEntropyMulti()),
GeneralizedCost(costfunc=CrossEntropyMulti())
],
weights=[1, 0., 0.]) # We only want to consider the CE of the main path
assert os.path.exists(
args.model_file), 'script requires the trained weights file'
model.load_params(args.model_file)
model.initialize(test, cost)
print 'running speed benchmark...'
model.benchmark(test, cost, opt)
print '\nCalculating performance on validation set...'
test.reset()
mets = model.eval(test, metric=valmetric)
print 'Validation set metrics:'
print 'LogLoss: %.2f, Accuracy: %.1f %% (Top-1), %.1f %% (Top-5)' % (
mets[0], (1.0 - mets[1]) * 100, (1.0 - mets[2]) * 100)
