def main():
h5file = '/root/data/pcallier/amazon/temp.hd5'
amzn_path = '/root/data/pcallier/amazon/reviews_Health_and_Personal_Care.json.gz'
#azbw = AmazonBatchWriter(amzn_path, h5file)
#azbw.run()
from neon.backends.nervanagpu import NervanaGPU
ng = NervanaGPU(0, device_id=1)
NervanaObject.be = ng
ng.bsz = 128
train_set = DiskDataIterator(lambda: batcher(load_data('/root/data/amazon/test_amazon.json.gz')), 3000, 128, nvocab=67)
# random examples from each
for bidx, (X_batch, y_batch) in enumerate(train_set):
print "Batch {}:".format(bidx)
#print X_batch.get().T.sum(axis=1)
reviewnum = input("Pick review index to fetch and decode: ")
review = from_one_hot(X_batch.get().T[reviewnum].reshape(67, -1))
print ''.join(review)[::-1]
def main():
h5file = '/root/data/pcallier/amazon/temp.hd5'
amzn_path = '/root/data/pcallier/amazon/reviews_Health_and_Personal_Care.json.gz'
#azbw = AmazonBatchWriter(amzn_path, h5file)
#azbw.run()
from neon.backends.nervanagpu import NervanaGPU
ng = NervanaGPU(0, device_id=1)
NervanaObject.be = ng
ng.bsz = 128
train_set = DiskDataIterator(
lambda: batcher(load_data('/root/data/amazon/test_amazon.json.gz')),
3000,
128,
nvocab=67)
# random examples from each
for bidx, (X_batch, y_batch) in enumerate(train_set):
print "Batch {}:".format(bidx)
#print X_batch.get().T.sum(axis=1)
reviewnum = input("Pick review index to fetch and decode: ")
review = from_one_hot(X_batch.get().T[reviewnum].reshape(67, -1))
print ''.join(review)[::-1]
np.set_printoptions(threshold=np.nan)
print "Batch properties:"
print "Shape (data): {}".format(data.shape)
print "Shape (label): {}".format(label.shape)
print "Type: {}".format(type(data))
print
print "First record of first batch:"
print "Type (1 level in): {}".format(type(data[0]))
print "Type of record (2 levels in): {}".format(type(data[0,0]))
print data[0,0]
print "Sentiment label: {}".format(label[0,0])
print "Data in numpy format:"
oh = data_utils.to_one_hot(data[0,0])
print np.array_str(np.argmax(oh,axis=0))
print "Translated back into characters:\n"
print ''.join(data_utils.from_one_hot(oh))
# demo balanced batching
amz_balanced_batcher = batch_data(amz_train,balance_labels=True)
balanced_batch = amz_balanced_batcher.next()
print 'Balanced batch:'
balanced_label_counts = {}
for idx in range(balanced_batch[1].shape[0]):
label = balanced_batch[1][idx,0]
balanced_label_counts[label] = balanced_label_counts.get(label, 0) + 1
print balanced_label_counts
# Demo iterator utility classes
# iterate multiple times over same data
if args.iterator_demo:
# Demo dataIterator class
transformer_fun=None)
am_test_batch = batch_data.batch_data(amte,
normalizer_fun=None,transformer_fun=None)
# Spit out some sample data
next_batch = am_train_batch.next()
data, label = next_batch
np.set_printoptions(threshold=np.nan)
print "Batch properties:"
print "Length: {}".format(len(data))
print "Type: {}".format(type(data))
print
print "First record of first batch:"
print "Type (1 level in): {}".format(type(data[0]))
print "Type of record (2 levels in): {}".format(type(data[0,0]))
print data[0,0]
print "Sentiment label: {}".format(label[0])
print "In numpy format:"
oh = data_utils.to_one_hot(data[0,0])
print np.array_str(np.argmax(oh,axis=0))
print "Translated back into characters:\n"
print data_utils.from_one_hot(oh)
# dimension checks
second_batch_data, second_batch_label = second_batch = am_train_batch.next()
second_batch = list(second_batch)
print len(second_batch)
print "Data object type: ", type(second_batch_data)
print second_batch_data.shape
