def main():
args = parse_args()
gen_data(SEQ_LEN, args.start_range, args.end_range)
batch_size = 100
buckets = []
num_hidden = 300
num_embed = 512
num_lstm_layer = 2
num_epoch = 1
learning_rate = 0.1
momentum = 0.9
if args.cpu:
contexts = [mx.context.cpu(i) for i in range(1)]
else:
contexts = [mx.context.gpu(i) for i in range(1)]
vocab = default_build_vocab(os.path.join(DATA_DIR, TRAIN_FILE))
def sym_gen(seq_len):
return bi_lstm_unroll(seq_len,
len(vocab),
num_hidden=num_hidden,
num_embed=num_embed,
num_label=len(vocab))
init_c = [('l%d_init_c' % l, (batch_size, num_hidden))
for l in range(num_lstm_layer)]
init_h = [('l%d_init_h' % l, (batch_size, num_hidden))
for l in range(num_lstm_layer)]
init_states = init_c + init_h
data_train = BucketSentenceIter(os.path.join(DATA_DIR, TRAIN_FILE), vocab,
buckets, batch_size, init_states)
data_val = BucketSentenceIter(os.path.join(DATA_DIR, VALID_FILE), vocab,
buckets, batch_size, init_states)
if len(buckets) == 1:
symbol = sym_gen(buckets[0])
else:
symbol = sym_gen
model = mx.model.FeedForward(ctx=contexts,
symbol=symbol,
num_epoch=num_epoch,
learning_rate=learning_rate,
momentum=momentum,
wd=0.00001,
initializer=mx.init.Xavier(factor_type="in",
magnitude=2.34))
model.fit(
X=data_train,
eval_data=data_val,
eval_metric=mx.metric.np(Perplexity),
batch_end_callback=mx.callback.Speedometer(batch_size, 50),
)
model.save("sort")
def main():
args = parse_args()
gen_data(SEQ_LEN, args.start_range, args.end_range)
batch_size = 100
buckets = []
num_hidden = 300
num_embed = 512
num_lstm_layer = 2
num_epoch = 1
learning_rate = 0.1
momentum = 0.9
if args.cpu:
contexts = [mx.context.cpu(i) for i in range(1)]
else:
contexts = [mx.context.gpu(i) for i in range(1)]
vocab = default_build_vocab(os.path.join(DATA_DIR, TRAIN_FILE))
def sym_gen(seq_len):
return bi_lstm_unroll(seq_len, len(vocab),
num_hidden=num_hidden, num_embed=num_embed,
num_label=len(vocab))
init_c = [('l%d_init_c'%l, (batch_size, num_hidden)) for l in range(num_lstm_layer)]
init_h = [('l%d_init_h'%l, (batch_size, num_hidden)) for l in range(num_lstm_layer)]
init_states = init_c + init_h
data_train = BucketSentenceIter(os.path.join(DATA_DIR, TRAIN_FILE), vocab,
buckets, batch_size, init_states)
data_val = BucketSentenceIter(os.path.join(DATA_DIR, VALID_FILE), vocab,
buckets, batch_size, init_states)
if len(buckets) == 1:
symbol = sym_gen(buckets[0])
else:
symbol = sym_gen
model = mx.model.FeedForward(ctx=contexts,
symbol=symbol,
num_epoch=num_epoch,
learning_rate=learning_rate,
momentum=momentum,
wd=0.00001,
initializer=mx.init.Xavier(factor_type="in", magnitude=2.34))
model.fit(X=data_train, eval_data=data_val,
eval_metric = mx.metric.np(Perplexity),
batch_end_callback=mx.callback.Speedometer(batch_size, 50),)
model.save("sort")
def main():
tks = sys.argv[1:]
assert len(
tks
) >= 5, "Please provide 5 numbers for sorting as sequence length is 5"
batch_size = 1
buckets = []
num_hidden = 300
num_embed = 512
num_lstm_layer = 2
num_epoch = 1
learning_rate = 0.1
momentum = 0.9
contexts = [mx.context.cpu(i) for i in range(1)]
vocab = default_build_vocab(os.path.join(DATA_DIR, TRAIN_FILE))
rvocab = {}
for k, v in vocab.items():
rvocab[v] = k
_, arg_params, __ = mx.model.load_checkpoint("sort", 1)
for tk in tks:
assert (
tk in vocab
), "{} not in range of numbers that the model trained for.".format(
tk)
model = BiLSTMInferenceModel(SEQ_LEN,
len(vocab),
num_hidden=num_hidden,
num_embed=num_embed,
num_label=len(vocab),
arg_params=arg_params,
ctx=contexts,
dropout=0.0)
data = np.zeros((1, len(tks)))
for k in range(len(tks)):
data[0][k] = vocab[tks[k]]
data = mx.nd.array(data)
prob = model.forward(data)
for k in range(len(tks)):
print(rvocab[np.argmax(prob, axis=1)[k]])
def main():
tks = sys.argv[1:]
assert len(tks) >= 5, "Please provide 5 numbers for sorting as sequence length is 5"
batch_size = 1
buckets = []
num_hidden = 300
num_embed = 512
num_lstm_layer = 2
num_epoch = 1
learning_rate = 0.1
momentum = 0.9
contexts = [mx.context.cpu(i) for i in range(1)]
vocab = default_build_vocab(os.path.join(DATA_DIR, TRAIN_FILE))
rvocab = {}
for k, v in vocab.items():
rvocab[v] = k
_, arg_params, __ = mx.model.load_checkpoint("sort", 1)
for tk in tks:
assert (tk in vocab), "{} not in range of numbers that the model trained for.".format(tk)
model = BiLSTMInferenceModel(SEQ_LEN, len(vocab),
num_hidden=num_hidden, num_embed=num_embed,
num_label=len(vocab), arg_params=arg_params, ctx=contexts, dropout=0.0)
data = np.zeros((1, len(tks)))
for k in range(len(tks)):
data[0][k] = vocab[tks[k]]
data = mx.nd.array(data)
prob = model.forward(data)
for k in range(len(tks)):
print(rvocab[np.argmax(prob, axis = 1)[k]])
if __name__ == '__main__':
batch_size = 1
buckets = []
num_hidden = 300
num_embed = 512
num_lstm_layer = 2
num_epoch = 1
learning_rate = 0.1
momentum = 0.9
contexts = [mx.context.gpu(i) for i in range(1)]
vocab = default_build_vocab("./data/sort.train.txt")
rvocab = {}
for k, v in vocab.items():
rvocab[v] = k
_, arg_params, __ = mx.model.load_checkpoint("sort", 1)
model = BiLSTMInferenceModel(5,
len(vocab),
num_hidden=num_hidden,
num_embed=num_embed,
num_label=len(vocab),
arg_params=arg_params,
ctx=contexts,
dropout=0.0)
arr[:] = tmp
if __name__ == '__main__':
batch_size = 1
buckets = []
num_hidden = 300
num_embed = 512
num_lstm_layer = 2
num_epoch = 1
learning_rate = 0.1
momentum = 0.9
contexts = [mx.context.gpu(i) for i in range(1)]
vocab = default_build_vocab("./data/sort.train.txt")
rvocab = {}
for k, v in vocab.items():
rvocab[v] = k
_, arg_params, __ = mx.model.load_checkpoint("sort", 1)
model = BiLSTMInferenceModel(5, len(vocab),
num_hidden=num_hidden, num_embed=num_embed,
num_label=len(vocab), arg_params=arg_params, ctx=contexts, dropout=0.0)
tks = sys.argv[1:]
data = np.zeros((1, len(tks)))
for k in range(len(tks)):
data[0][k] = vocab[tks[k]]
