def create_split_streams(frame_files,label_files,left_context,right_context):
streams = []
for frame_file,label_file in izip(frame_files,label_files):
frame_stream = data_io.stream_file(frame_file)
frame_stream = data_io.context(frame_stream,
left=left_context,right=right_context)
label_stream = data_io.stream_file(label_file)
stream = data_io.zip_streams(frame_stream,label_stream)
streams.append(stream)
return streams
def create_split_streams(frame_files, label_files, left_context,
right_context):
streams = []
for frame_file, label_file in izip(frame_files, label_files):
frame_stream = data_io.stream_file(frame_file)
frame_stream = data_io.context(frame_stream,
left=left_context,
right=right_context)
label_stream = data_io.stream_file(label_file)
stream = data_io.zip_streams(frame_stream, label_stream)
streams.append(stream)
return streams
def create_split_streams(frame_files,left_context,right_context):
streams = []
for frame_file in frame_files:
stream = data_io.stream_file(frame_file)
stream = data_io.context(stream,
left=left_context,right=right_context)
stream = data_io.zip_streams(stream)
streams.append(stream)
return streams
def stream():
stream = data_io.random_select_stream(*[
data_io.stream_file('data/train.%02d.pklgz' % i)
for i in xrange(1, 20)
])
stream = data_io.buffered_sort(stream, key=lambda x: x[1].shape[0], buffer_items=128)
batched_stream = reader.batch_and_pad(stream, batch_size=16, mean=mean, std=std)
batched_stream = data_io.buffered_random(batched_stream, buffer_items=4)
return batched_stream
def validate():
stream = data_io.stream_file('data/train.%02d.pklgz' % 0)
stream = data_io.buffered_sort(stream, key=lambda x: x[1].shape[0], buffer_items=128)
batched_stream = reader.batch_and_pad(stream, batch_size=32, mean=mean, std=std)
total_cost = 0
total_frames = 0
for data, lengths in batched_stream:
batch_avg_cost = test(data,lengths)
batch_frames = np.sum(lengths)
total_cost += batch_avg_cost * batch_frames
total_frames += batch_frames
return total_cost / total_frames
P_learn = Parameters()
updates = updates.adam(parameters,gradients,learning_rate=0.00025,P=P_learn)
updates = normalise_weights(updates)
print "Compiling..."
train = theano.function(
inputs=[X,l],
outputs=batch_cost,
updates=updates,
)
test = theano.function(inputs=[X,l],outputs=batch_cost)
print "Calculating mean variance..."
rand_stream = data_io.random_select_stream(*[
data_io.stream_file('data/train.%02d.pklgz' % i)
for i in xrange(1, 20)
])
mean, std, count = reader.get_normalise(rand_stream)
print "Dataset count:", count
def stream():
stream = data_io.random_select_stream(*[
data_io.stream_file('data/train.%02d.pklgz' % i)
for i in xrange(1, 20)
])
stream = data_io.buffered_sort(stream, key=lambda x: x[1].shape[0], buffer_items=128)
batched_stream = reader.batch_and_pad(stream, batch_size=16, mean=mean, std=std)
batched_stream = data_io.buffered_random(batched_stream, buffer_items=4)
return batched_stream
