def main(width=32, vector_length=8):
train_data, check_data, nr_tag = ancora_pos_tags()
model = FeedForward(
(layerize(flatten_sequences), BatchNorm(Embed(width, vector_length)),
ExtractWindow(nW=2), BatchNorm(ReLu(width)), BatchNorm(ReLu(width)),
Softmax(nr_tag)))
train_X, train_y = zip(*train_data)
dev_X, dev_y = zip(*check_data)
dev_y = model.ops.flatten(dev_y)
with model.begin_training(train_X, train_y) as (trainer, optimizer):
trainer.batch_size = 8
trainer.nb_epoch = 10
trainer.dropout = 0.2
trainer.dropout_decay = 0.
trainer.each_epoch.append(lambda: print(model.evaluate(dev_X, dev_y)))
for X, y in trainer.iterate(train_X, train_y):
y = model.ops.flatten(y)
yh, backprop = model.begin_update(X, drop=trainer.dropout)
d_loss, loss = categorical_crossentropy(yh, y)
optimizer.set_loss(loss)
backprop(d_loss, optimizer)
with model.use_params(optimizer.averages):
print(model.evaluate(dev_X, dev_y))
def main(width=100,
depth=4,
vector_length=64,
min_batch_size=1,
max_batch_size=32,
learn_rate=0.001,
momentum=0.9,
dropout=0.5,
dropout_decay=1e-4,
nb_epoch=20,
L2=1e-6):
cfg = dict(locals())
print(cfg)
if cupy is not None:
print("Using GPU")
Model.ops = CupyOps()
train_data, check_data, nr_tag = ancora_pos_tags()
extracter = FeatureExtracter('es', attrs=[LOWER, SHAPE, PREFIX, SUFFIX])
Model.lsuv = True
with Model.define_operators({
'**': clone,
'>>': chain,
'+': add,
'|': concatenate
}):
lower_case = HashEmbed(width, 100, column=0)
shape = HashEmbed(width // 2, 200, column=1)
prefix = HashEmbed(width // 2, 100, column=2)
suffix = HashEmbed(width // 2, 100, column=3)
model = (with_flatten(
(lower_case | shape | prefix | suffix) >> Maxout(width, pieces=3)
>> Residual(ExtractWindow(nW=1) >> Maxout(width, pieces=3))**depth
>> Softmax(nr_tag),
pad=depth))
train_X, train_y = preprocess(model.ops, extracter, train_data, nr_tag)
dev_X, dev_y = preprocess(model.ops, extracter, check_data, nr_tag)
n_train = float(sum(len(x) for x in train_X))
global epoch_train_acc
with model.begin_training(train_X[:5000], train_y[:5000],
**cfg) as (trainer, optimizer):
trainer.each_epoch.append(track_progress(**locals()))
trainer.batch_size = min_batch_size
batch_size = float(min_batch_size)
for X, y in trainer.iterate(train_X, train_y):
yh, backprop = model.begin_update(X, drop=trainer.dropout)
gradient = [yh[i] - y[i] for i in range(len(yh))]
backprop(gradient, optimizer)
trainer.batch_size = min(int(batch_size), max_batch_size)
batch_size *= 1.001
with model.use_params(trainer.optimizer.averages):
print(model.evaluate(dev_X, model.ops.flatten(dev_y)))
with open('/tmp/model.pickle', 'wb') as file_:
pickle.dump(model, file_)
def main(width=128,
depth=1,
vector_length=128,
min_batch_size=16,
max_batch_size=16,
learn_rate=0.001,
momentum=0.9,
dropout=0.5,
dropout_decay=1e-4,
nb_epoch=20,
L2=1e-6):
using_gpu = prefer_gpu()
if using_gpu:
torch.set_default_tensor_type('torch.cuda.FloatTensor')
cfg = dict(locals())
print(cfg)
train_data, check_data, nr_tag = ancora_pos_tags()
train_data = list(train_data)
check_data = list(check_data)
extracter = FeatureExtracter('es', attrs=[LOWER, SHAPE, PREFIX, SUFFIX])
with Model.define_operators({
'**': clone,
'>>': chain,
'+': add,
'|': concatenate
}):
lower_case = HashEmbed(width, 100, column=0)
shape = HashEmbed(width // 2, 200, column=1)
prefix = HashEmbed(width // 2, 100, column=2)
suffix = HashEmbed(width // 2, 100, column=3)
model = (with_flatten(
(lower_case | shape | prefix | suffix) >> Maxout(width, pieces=3))
>> PyTorchBiLSTM(width, width, depth) >> with_flatten(
Softmax(nr_tag)))
train_X, train_y = preprocess(model.ops, extracter, train_data, nr_tag)
dev_X, dev_y = preprocess(model.ops, extracter, check_data, nr_tag)
n_train = float(sum(len(x) for x in train_X))
global epoch_train_acc
with model.begin_training(train_X[:10], train_y[:10],
**cfg) as (trainer, optimizer):
trainer.each_epoch.append(track_progress(**locals()))
trainer.batch_size = min_batch_size
batch_size = float(min_batch_size)
for X, y in trainer.iterate(train_X, train_y):
yh, backprop = model.begin_update(X, drop=trainer.dropout)
gradient = [yh[i] - y[i] for i in range(len(yh))]
backprop(gradient, optimizer)
trainer.batch_size = min(int(batch_size), max_batch_size)
batch_size *= 1.001
print(model.evaluate(dev_X, model.ops.flatten(dev_y)))
with open('/tmp/model.pickle', 'wb') as file_:
pickle.dump(model, file_)
def main(
width=100,
depth=4,
vector_length=64,
min_batch_size=1,
max_batch_size=32,
learn_rate=0.001,
momentum=0.9,
dropout=0.5,
dropout_decay=1e-4,
nb_epoch=20,
L2=1e-6,
):
cfg = dict(locals())
print(cfg)
prefer_gpu()
train_data, check_data, nr_tag = ancora_pos_tags()
extracter = FeatureExtracter("es", attrs=[LOWER, SHAPE, PREFIX, SUFFIX])
Model.lsuv = True
with Model.define_operators({"**": clone, ">>": chain, "+": add, "|": concatenate}):
lower_case = HashEmbed(width, 100, column=0)
shape = HashEmbed(width // 2, 200, column=1)
prefix = HashEmbed(width // 2, 100, column=2)
suffix = HashEmbed(width // 2, 100, column=3)
model = with_flatten(
(lower_case | shape | prefix | suffix)
>> Maxout(width, pieces=3)
>> Residual(ExtractWindow(nW=1) >> Maxout(width, pieces=3)) ** depth
>> Softmax(nr_tag),
pad=depth,
)
train_X, train_y = preprocess(model.ops, extracter, train_data, nr_tag)
dev_X, dev_y = preprocess(model.ops, extracter, check_data, nr_tag)
n_train = float(sum(len(x) for x in train_X))
global epoch_train_acc
with model.begin_training(train_X[:5000], train_y[:5000], **cfg) as (
trainer,
optimizer,
):
trainer.each_epoch.append(track_progress(**locals()))
trainer.batch_size = min_batch_size
batch_size = float(min_batch_size)
for X, y in trainer.iterate(train_X, train_y):
yh, backprop = model.begin_update(X, drop=trainer.dropout)
gradient = [yh[i] - y[i] for i in range(len(yh))]
backprop(gradient, optimizer)
trainer.batch_size = min(int(batch_size), max_batch_size)
batch_size *= 1.001
with model.use_params(trainer.optimizer.averages):
print(model.evaluate(dev_X, model.ops.flatten(dev_y)))
with open("/tmp/model.pickle", "wb") as file_:
pickle.dump(model, file_)
def main(width=64,
depth=2,
vector_length=64,
min_batch_size=1,
max_batch_size=32,
dropout=0.9,
dropout_decay=1e-3,
nb_epoch=20,
L2=1e-6):
cfg = dict(locals())
print(cfg)
if cupy is not None:
print("Using GPU")
Model.ops = CupyOps()
train_data, check_data, nr_tag = ancora_pos_tags()
extracter = FeatureExtracter('es', attrs=[LOWER, SHAPE, PREFIX, SUFFIX])
with Model.define_operators({
'**': clone,
'>>': chain,
'+': add,
'|': concatenate
}):
lower_case = Embed(width, vector_length, 5000, column=0)
prefix = Embed(width, vector_length, 5000, column=2)
suffix = Embed(width, vector_length, 5000, column=3)
model = (layerize(flatten_sequences) >> (lower_case + prefix + suffix)
>> Residual(ExtractWindow(nW=1) >> Maxout(width))**depth >>
Softmax(nr_tag))
train_X, train_y = preprocess(model.ops, extracter, train_data, nr_tag)
dev_X, dev_y = preprocess(model.ops, extracter, check_data, nr_tag)
n_train = float(sum(len(x) for x in train_X))
global epoch_train_acc
with model.begin_training(train_X, train_y, **cfg) as (trainer, optimizer):
trainer.each_epoch.append(track_progress(**locals()))
trainer.batch_size = min_batch_size
batch_size = float(min_batch_size)
for X, y in trainer.iterate(train_X, train_y):
y = model.ops.flatten(y)
yh, backprop = model.begin_update(X, drop=trainer.dropout)
loss = ((yh - y)**2).sum() / y.shape[0]
if loss > 0.:
optimizer.set_loss(loss)
backprop(yh - y, optimizer)
trainer.batch_size = min(int(batch_size), max_batch_size)
batch_size *= 1.001
epoch_train_acc += (yh.argmax(axis=1) == y.argmax(axis=1)).sum()
if epoch_train_acc / n_train >= 0.999:
break
with model.use_params(trainer.optimizer.averages):
print(model.evaluate(dev_X, model.ops.flatten(dev_y)))
def main(width=32, nr_vector=1000):
train_data, check_data, nr_tag = ancora_pos_tags(encode_words=True)
model = with_flatten(
chain(HashEmbed(width, 1000), ReLu(width, width), ReLu(width, width),
Softmax(nr_tag, width)))
train_X, train_y = zip(*train_data)
dev_X, dev_y = zip(*check_data)
train_y = [to_categorical(y, nb_classes=nr_tag) for y in train_y]
dev_y = [to_categorical(y, nb_classes=nr_tag) for y in dev_y]
with model.begin_training(train_X, train_y) as (trainer, optimizer):
trainer.each_epoch.append(lambda: print(model.evaluate(dev_X, dev_y)))
for X, y in trainer.iterate(train_X, train_y):
yh, backprop = model.begin_update(X, drop=trainer.dropout)
backprop([yh[i] - y[i] for i in range(len(yh))], optimizer)
with model.use_params(optimizer.averages):
print(model.evaluate(dev_X, dev_y))
def main(width=32, nr_vector=1000):
train_data, check_data, nr_tag = ancora_pos_tags(encode_words=True)
model = with_flatten(
chain(
HashEmbed(width, nr_vector),
ReLu(width, width),
ReLu(width, width),
Softmax(nr_tag, width)))
train_X, train_y = zip(*train_data)
dev_X, dev_y = zip(*check_data)
train_y = [to_categorical(y, nb_classes=nr_tag) for y in train_y]
dev_y = [to_categorical(y, nb_classes=nr_tag) for y in dev_y]
with model.begin_training(train_X, train_y) as (trainer, optimizer):
trainer.each_epoch.append(
lambda: print(model.evaluate(dev_X, dev_y)))
for X, y in trainer.iterate(train_X, train_y):
yh, backprop = model.begin_update(X, drop=trainer.dropout)
backprop([yh[i]-y[i] for i in range(len(yh))], optimizer)
with model.use_params(optimizer.averages):
print(model.evaluate(dev_X, dev_y))
def main(width=128,
depth=4,
vector_length=64,
max_batch_size=32,
dropout=0.9,
drop_decay=1e-4,
nb_epoch=20,
L2=1e-5):
cfg = dict(locals())
Model.ops = CupyOps()
train_data, check_data, nr_tag = ancora_pos_tags()
with Model.define_operators({'**': clone, '>>': chain}):
model = (layerize(flatten_sequences) >> Embed(width, vector_length) >>
(ExtractWindow(nW=1) >> Maxout(width, pieces=3))**depth >>
Softmax(nr_tag))
train_X, train_y = preprocess(model.ops, train_data, nr_tag)
dev_X, dev_y = preprocess(model.ops, check_data, nr_tag)
n_train = float(sum(len(x) for x in train_X))
global epoch_train_acc
with model.begin_training(train_X, train_y, **cfg) as (trainer, optimizer):
trainer.each_epoch.append(track_progress(**locals()))
trainer.batch_size = 1
batch_size = 1.
for X, y in trainer.iterate(train_X, train_y):
y = model.ops.flatten(y)
yh, backprop = model.begin_update(X, drop=trainer.dropout)
backprop(yh - y, optimizer)
trainer.batch_size = min(int(batch_size), max_batch_size)
batch_size *= 1.001
epoch_train_acc += (yh.argmax(axis=1) == y.argmax(axis=1)).sum()
with model.use_params(trainer.optimizer.averages):
print(model.evaluate(dev_X, model.ops.flatten(dev_y)))
def ancora():
train_data, check_data, nr_class = datasets.ancora_pos_tags()
train_X, train_y = zip(*train_data)
dev_X, dev_y = zip(*check_data)
return (train_X[:100], train_y[:100]), (dev_X, dev_y)
def main(
width=128,
depth=1,
vector_length=128,
min_batch_size=16,
max_batch_size=16,
learn_rate=0.001,
momentum=0.9,
dropout=0.5,
dropout_decay=1e-4,
nb_epoch=20,
L2=1e-6,
):
using_gpu = prefer_gpu()
if using_gpu:
torch.set_default_tensor_type("torch.cuda.FloatTensor")
cfg = dict(locals())
print(cfg)
train_data, check_data, nr_tag = ancora_pos_tags()
train_data = list(train_data)
check_data = list(check_data)
extracter = FeatureExtracter("es", attrs=[LOWER, SHAPE, PREFIX, SUFFIX])
with Model.define_operators({"**": clone, ">>": chain, "+": add, "|": concatenate}):
lower_case = HashEmbed(width, 100, column=0)
shape = HashEmbed(width // 2, 200, column=1)
prefix = HashEmbed(width // 2, 100, column=2)
suffix = HashEmbed(width // 2, 100, column=3)
model = (
with_flatten(
(lower_case | shape | prefix | suffix) >> Maxout(width, pieces=3)
)
>> PyTorchBiLSTM(width, width, depth)
>> with_flatten(Softmax(nr_tag))
)
train_X, train_y = preprocess(model.ops, extracter, train_data, nr_tag)
dev_X, dev_y = preprocess(model.ops, extracter, check_data, nr_tag)
n_train = float(sum(len(x) for x in train_X))
global epoch_train_acc
with model.begin_training(train_X[:10], train_y[:10], **cfg) as (
trainer,
optimizer,
):
trainer.each_epoch.append(track_progress(**locals()))
trainer.batch_size = min_batch_size
batch_size = float(min_batch_size)
for X, y in trainer.iterate(train_X, train_y):
yh, backprop = model.begin_update(X, drop=trainer.dropout)
gradient = [yh[i] - y[i] for i in range(len(yh))]
backprop(gradient, optimizer)
trainer.batch_size = min(int(batch_size), max_batch_size)
batch_size *= 1.001
print(model.evaluate(dev_X, model.ops.flatten(dev_y)))
with open("/tmp/model.pickle", "wb") as file_:
pickle.dump(model, file_)
def main(
width=100,
depth=4,
vector_length=64,
min_batch_size=1,
max_batch_size=32,
learn_rate=0.001,
momentum=0.9,
dropout=0.5,
dropout_decay=1e-4,
nb_epoch=20,
L2=1e-6,
):
cfg = dict(locals())
print(cfg)
prefer_gpu()
train_data, check_data, nr_tag = ancora_pos_tags()
extracter = FeatureExtracter("es", attrs=[LOWER, SHAPE, PREFIX, SUFFIX])
Model.lsuv = True
with Model.define_operators({
"**": clone,
">>": chain,
"+": add,
"|": concatenate
}):
lower_case = HashEmbed(width, 100, column=0)
shape = HashEmbed(width // 2, 200, column=1)
prefix = HashEmbed(width // 2, 100, column=2)
suffix = HashEmbed(width // 2, 100, column=3)
model = (with_flatten(
(lower_case | shape | prefix | suffix) >> Maxout(width, pieces=3),
pad=depth) >> with_pad_and_mask(
MultiHeadedAttention(nM=width, nH=4)) >> with_flatten(
Softmax(nr_tag)))
train_X, train_y = preprocess(model.ops, extracter, train_data, nr_tag)
dev_X, dev_y = preprocess(model.ops, extracter, check_data, nr_tag)
n_train = float(sum(len(x) for x in train_X))
global epoch_train_acc
with model.begin_training(train_X[:5000], train_y[:5000], **cfg) as (
trainer,
optimizer,
):
trainer.each_epoch.append(track_progress(**locals()))
trainer.batch_size = min_batch_size
batch_size = float(min_batch_size)
for X, y in trainer.iterate(train_X, train_y):
yh, backprop = model.begin_update(X, drop=trainer.dropout)
gradient = [yh[i] - y[i] for i in range(len(yh))]
backprop(gradient, optimizer)
trainer.batch_size = min(int(batch_size), max_batch_size)
batch_size *= 1.001
with model.use_params(trainer.optimizer.averages):
print(model.evaluate(dev_X, model.ops.flatten(dev_y)))
with open("/tmp/model.pickle", "wb") as file_:
pickle.dump(model, file_)
