def train_epoch(_epoch, dataloader, model, loss_function):
# switch to train mode -> enable regularization layers, such as Dropout
model.train()
running_loss = 0.0
for i_batch, sample_batched in enumerate(dataloader, 1):
# get the inputs (batch)
inputs, labels, lengths, indices, get_aspect = sample_batched
# in this point i have to concatenate the get_aspect embedding to the input.
# sort batch (for handling inputs of variable length)
lengths, (inputs,
labels), get_aspect = sort_cnn_batch(lengths,
(inputs, labels),
get_aspect)
# convert to CUDA Variables
if torch.cuda.is_available():
inputs = Variable(inputs.cuda())
labels = Variable(labels.cuda())
else:
inputs = Variable(inputs)
labels = Variable(labels)
# 1 - zero the gradients
optimizer.zero_grad()
# 2 - forward pass: compute predicted y by passing x to the model
outputs = model(inputs, get_aspect)
# 3 - compute loss
_, labels = labels.squeeze().max(dim=1)
loss = loss_function(outputs, labels)
loss.backward()
# 5 - update weights
optimizer.step()
running_loss += loss.data[0]
# print statistics
progress(loss=loss.data[0],
epoch=_epoch,
batch=i_batch,
batch_size=BATCH_SIZE,
dataset_size=len(train_set))
return loss.data[0]
best_match = {}
for a in possible_matches:
if best_match == {}:
best_match = a
elif a["popularity"] > best_match["popularity"]:
best_match = a
found_match = best_match
# print "best match: " + best_match["name"]
else:
covered_artists_file.write("%s\n" % artist)
# print "none: " + artist
if found_match != {}:
write_obj = {
'name': found_match["name"],
'id': found_match["id"],
'genres': found_match["genres"],
'popularity': found_match["popularity"]
}
artist_info_file.write("%s\n" % write_obj)
covered_artists_file.write("%s\n" % artist)
found_artists_amount += 1
counter += 1
progress = ((counter + 0.0) / artist_amount) * 100
helpers.progress(progress)
covered_artists_amount += counter
helpers.endProgress(" Artists covered: " + str(covered_artists_amount))
_ewkLoose = np.full((nl_max), False, dtype=bool)
_ewkTight = np.full((nl_max), False, dtype=bool)
#output_tree.Branch('_ewkLoose', _ewkLoose, '_ewkLoose[_nL]/O')
#output_tree.Branch('_ewkTight', _ewkTight, '_ewkTight[_nL]/O')
if args.isTest:
eventRange = xrange(250)
else:
eventRange = xrange(Chain.GetEntries())
from objectSelection import isEwkLoose, isEwkTight
#output_file.cd('blackJackAndHookers')
#Fill new tree
for entry in eventRange:
progress(entry, len(eventRange))
Chain.GetEntry(entry)
if args.skim == 'ewkino':
if not isGoodEventEwkino(Chain): continue
elif args.skim == 'tauAN':
if not isGoodEventAN17_094(Chain): continue
elif args.skim == 'FR':
if not isGoodEventFakeRate(Chain): continue
elif args.skim == 'jana':
if not isGoodEventJana(Chain): continue
for l in xrange(Chain._nL):
_ewkLoose[l] = isEwkLoose(Chain, l)
_ewkTight[l] = isEwkTight(Chain, l)
output_tree.Fill()
def log_optimization_init(i, T, res=0.1, logger=None):
if logger is None:
logger = logging.getLogger('root')
v = progress(i,T, res=res)
if v is not None:
logger.info('%0.1f%% Models initialized', v*100)