for start_idx in range(0, len(inputs) - batchsize + 1, batchsize):
if shuffle:
excerpt = indices[start_idx:start_idx + batchsize]
else:
excerpt = slice(start_idx, start_idx + batchsize)
yield inputs[excerpt], targets[excerpt]
# Create the Network
print "Embeddings shape", embeddings.shape
model = Sequential()
model.add(
FixedEmbedding(output_dim=embeddings.shape[1],
input_dim=embeddings.shape[0],
input_length=MAX_LENGTH,
weights=[embeddings]))
# Hidden + Softmax Layer
model.add(
LSTM(
output_dim=MAX_LENGTH,
init='glorot_uniform',
activation='tanh',
batch_input_shape=(None, MAX_LENGTH, embeddings.shape[1]),
return_sequences=True,
))
model.add(Dropout(0.5))
model.add(
LSTM(
output_dim=MAX_LENGTH,
init='glorot_uniform',
if shuffle:
excerpt = indices[start_idx:start_idx + batchsize]
else:
excerpt = slice(start_idx, start_idx + batchsize)
yield inputs[excerpt], targets[excerpt]
# Create the Network
print "Embeddings shape", embeddings.shape
model = Sequential()
# Embeddings layers, lookups the word indices and maps them to their dense vectors. FixedEmbeddings are _not_ updated during training
# If you switch it to an Embedding-Layer, they will be updated (training time increases significant)
model.add(
FixedEmbedding(output_dim=embeddings.shape[1],
input_dim=embeddings.shape[0],
input_length=n_in,
weights=[embeddings]))
# Hidden + Softmax Layer
model.add(
LSTM(output_dim=n_hidden,
init='glorot_uniform',
activation='tanh',
batch_input_shape=(None, n_in, embeddings.shape[1]),
return_sequences=True))
model.add(Dropout(0.5))
model.add(
LSTM(output_dim=n_hidden,
init='glorot_uniform',
activation='tanh',
batch_input_shape=(None, n_in, n_hidden)))
model.add(Dense(output_dim=n_out, activation='softmax'))
number_of_epochs = 10
minibatch_size = 35
embedding_size = embeddings.shape[1]
dim_case = 6
x = T.imatrix('x') # the data, one word+context per row
y = T.ivector('y') # the labels are presented as 1D vector of [int] labels
print "Embeddings shape", embeddings.shape
words = Sequential()
words.add(
FixedEmbedding(output_dim=embeddings.shape[1],
input_dim=embeddings.shape[0],
input_length=n_in,
weights=[embeddings])) #input_length=n_in,
words.add(Flatten())
casing = Sequential()
#casing.add(Embedding(output_dim=dim_case, input_dim=len(caseLookup), input_length=n_in))
casing.add(
FixedEmbedding(output_dim=caseMatrix.shape[1],
input_dim=caseMatrix.shape[0],
input_length=n_in,
weights=[caseMatrix]))
casing.add(Flatten())
model = Sequential()
model.add(Merge([words, casing], mode='concat'))