def merge_generators(filename, batchsize, seqlen):
X = iu.train_generator(filename + ".seq", batchsize, seqlen, "seq", "repeat")
# merge the marks here, and the split them up in training.
C = iu.train_generator(filename + ".chromtracks", batchsize, seqlen, "chrom", "repeat")
y = iu.train_generator(filename + ".labels", batchsize, seqlen, "labels", "repeat")
while True:
print X.next,
yield [X.next(), C.next()], y.next()
def merge_generators(path, batchsize, seqlen, mode):
dat_seq = train_generator(path['seq'], batchsize, seqlen, 'seq', mode)
dat_chromatin = []
for chromatin_track in path['chromatin_tracks']:
dat_chromatin.append(
train_generator(chromatin_track, batchsize, seqlen, 'chrom', mode))
y = train_generator(path['labels'], batchsize, seqlen, 'labels', mode)
while True:
combined_chrom_data = []
for chromatin_track_generators in dat_chromatin:
x = next(chromatin_track_generators)
combined_chrom_data.append(pd.DataFrame(x))
chromatin_features = pd.concat(combined_chrom_data, axis=1).values
sequence_features = next(dat_seq)
labels = next(y)
yield [sequence_features, chromatin_features], labels
def data_generator(path, batchsize, seqlen, bin_size):
dat_seq = train_generator(path['seq'], batchsize, seqlen, 'seq', 'repeat')
dat_chromatin = []
for chromatin_track in path['chromatin_tracks']:
dat_chromatin.append(
train_generator(chromatin_track, batchsize, seqlen, 'chrom',
'repeat'))
y = train_generator(path['labels'], batchsize, seqlen, 'labels', 'repeat')
while True:
combined_chrom_data = []
for chromatin_track_generators in dat_chromatin:
curr_chromatin_mark = next(chromatin_track_generators)
mark_resolution = curr_chromatin_mark.shape
assert (mark_resolution == (batchsize, seqlen/bin_size)),\
"Please check binning, specified bin size=50"
combined_chrom_data.append(pd.DataFrame(curr_chromatin_mark))
chromatin_features = pd.concat(combined_chrom_data, axis=1).values
print(chromatin_features.shape)
sequence_features = next(dat_seq)
labels = next(y)
yield [sequence_features, chromatin_features], labels
def data_generator(path, batchsize, seqlen):
X = train_generator(path['seq'], batchsize, seqlen, 'seq', 'repeat')
y = train_generator(path['labels'], batchsize, seqlen, 'labels', 'repeat')
while True:
yield next(X), next(y)
def val_generator(filename, batchsize, seqlen):
X = iu.train_generator(filename + ".seq", batchsize, seqlen, "seq", "repeat")
y = iu.train_generator(filename + ".labels", batchsize, seqlen, "labels", "repeat")
while True:
yield X.next(), y.next()