def build_graph(self, batch):
self.reset()
# Build Computation Graph according to length
bsize, length = batch.data.shape
if self.num_neg_sample == -1:
negSampleIdx = range(self.dict_size)
else:
negSampleIdx = np.array([
np.random.randint(low=0, high=self.dict_size)
for i in range(self.num_neg_sample)
])
self.neg_sample.value = np.int32(negSampleIdx)
self.h0.value = np.zeros([bsize, self.hidden_dim])
self.c0.value = np.zeros([bsize, self.hidden_dim])
h = self.h0
c = self.c0
outputs = []
for idx in range(length - 1):
in_i = self.input()
in_i.value = batch.data[:, idx] # Get value from batch
x = Embed(in_i, self.embed)
h, c = self.lstm_cell(x, h, c)
outputs.append(h)
self.output(
HingeLossOutput(Collect(outputs), self.sample_embed,
self.neg_sample))
self.expect(batch.data[:, 1:])
def build(self, batch):
# clear nodes above watermark
del self.nodes[self.watermark:]
# Remove retained inputs
del self.inputs[2:]
data = batch.data
B = data.shape[0]
T = data.shape[1]
hidden_dim = self.hidden_dim
self.h0.value = np.zeros((B, hidden_dim))
self.c0.value = np.zeros((B, hidden_dim))
collect = []
h = self.h0
c = self.c0
for t in range(T - 1):
x = self.input()
x.value = data[:, t]
wordvec = Embed(x, self.C2V)
cell = LSTMCell(self.wf, self.bf, self.wi, self.bi, self.wc,
self.bc, self.wo, self.bo, wordvec, h, c)
collect.append(SoftMax(Dot(cell.hout, self.V)))
h = cell.hout
c = cell.cout
self.output(Collect(collect))
self.expect(data[:, 1:T].T)
def build_graph(batch):
data = batch.data[0]
fwd_graph.reset()
bcwd_graph.reset()
bsize, length = data.shape
fwd_graph.h0.value = np.zeros([bsize, hidden_dim])
fwd_graph.c0.value = np.zeros([bsize, hidden_dim])
bcwd_graph.h0.value = np.zeros([bsize, hidden_dim])
bcwd_graph.c0.value = np.zeros([bsize, hidden_dim])
fwd_h = fwd_graph.h0
fwd_c = fwd_graph.c0
bcwd_h = bcwd_graph.h0
bcwd_c = bcwd_graph.c0
fwd_outputs = []
bcwd_outputs = []
for idx in range(length - 1):
# Build Fowward Graph
fwd_in_i = fwd_graph.input()
fwd_in_i.value = data[:, idx] # Get value from batch
fwd_x = Embed(fwd_in_i, fwd_graph.embed)
fwd_h, fwd_c = fwd_graph.lstm_cell(fwd_x, fwd_h, fwd_c)
fwd_out_i = SoftMax(Dot(fwd_h, fwd_graph.v2c))
fwd_outputs.append(fwd_out_i)
# Build Backward Graph
bcwd_in_i = bcwd_graph.input()
bcwd_in_i.value = data[:, length - 1 - idx] # Get value from batch
bcwd_x = Embed(bcwd_in_i, bcwd_graph.embed)
bcwd_h, bcwd_c = bcwd_graph.lstm_cell(bcwd_x, bcwd_h, bcwd_c)
bcwd_out_i = SoftMax(Dot(bcwd_h, bcwd_graph.v2c))
bcwd_outputs.append(bcwd_out_i)
fwd_graph.output(Collect(fwd_outputs))
fwd_graph.expect(data[:, 1:])
bcwd_graph.output(Collect(bcwd_outputs))
bcwd_graph.expect(np.flip(data, axis=1)[:, 1:])
def build_graph(self, batch):
enc_data = batch.data
self.reset()
bsize, enc_length = enc_data.shape
outputs = []
# Build Fwd Encode Graph
self.feh0.value = np.zeros([bsize, self.half_dim])
self.fec0.value = np.zeros([bsize, self.half_dim])
fh = self.feh0
fc = self.fec0
for idx in range(enc_length):
in_i = self.input()
in_i.value = enc_data[:, idx] # Get value from batch
x = Embed(in_i, self.feembed)
fh, fc = self.fenc_lstm_cell(x, fh, fc)
# Build Bwd Encode Graph
self.beh0.value = np.zeros([bsize, self.half_dim])
self.bec0.value = np.zeros([bsize, self.half_dim])
bh = self.beh0
bc = self.bec0
for idx in range(enc_length):
in_i = self.input()
in_i.value = enc_data[:,
enc_length - 1 - idx] # Get value from batch
x = Embed(in_i, self.beembed)
bh, bc = self.benc_lstm_cell(x, bh, bc)
# Build Decode Graph
h = Concat(fh, bh)
c = Concat(fc, bc)
self.encoded_h = h
self.encoded_c = c
# Build Decode Graph
decode_in = self.input()
decode_in.value = np.zeros([bsize])
decode_embed = Embed(decode_in, self.dembed)
x = decode_embed
for idx in range(self.predict_len):
h, c = self.dec_lstm_cell(x, h, c)
out_i = ArgMax(SoftMax(Dot(h, self.dv2c)))
outputs.append(out_i)
x = Embed(out_i, self.dembed)
self.output(Collect(outputs))
self.expect(np.zeros([bsize, self.predict_len]))
def build_graph(self, batch):
enc_data = batch.data[0]
dec_data = batch.data[1]
self.reset()
bsize, enc_length = enc_data.shape
dec_length = dec_data.shape[1]
outputs = []
# Build Fwd Encode Graph
self.feh0.value = np.zeros([bsize, self.half_dim])
self.fec0.value = np.zeros([bsize, self.half_dim])
fh = self.feh0
fc = self.fec0
for idx in range(enc_length):
in_i = self.input()
in_i.value = enc_data[:, idx] # Get value from batch
x = Embed(in_i, self.feembed)
fh, fc = self.fenc_lstm_cell(x, fh, fc)
# Build Bwd Encode Graph
self.beh0.value = np.zeros([bsize, self.half_dim])
self.bec0.value = np.zeros([bsize, self.half_dim])
bh = self.beh0
bc = self.bec0
for idx in range(enc_length):
in_i = self.input()
in_i.value = enc_data[:,
enc_length - 1 - idx] # Get value from batch
x = Embed(in_i, self.beembed)
bh, bc = self.benc_lstm_cell(x, bh, bc)
# Build Decode Graph
h = Concat(fh, bh)
c = Concat(fc, bc)
self.encoded_h = h
self.encoded_c = c
for idx in range(dec_length - 1):
in_i = self.input()
in_i.value = dec_data[:, idx]
x = Embed(in_i, self.dembed)
h, c = self.dec_lstm_cell(x, h, c)
out_i = SoftMax(Dot(h, self.dv2c))
outputs.append(out_i)
self.output(Collect(outputs))
self.expect(dec_data[:, 1:])
def build_graph(self, batch):
data = batch.data[1]
self.reset()
# Build Computation Graph according to length
bsize, length = data.shape
h, c = self.bow_encode(batch.data[0])
outputs = []
for idx in range(length - 1):
in_i = self.input()
in_i.value = data[:, idx] # Get value from batch
x = Embed(in_i, self.embed)
h, c = self.lstm_cell(x, h, c)
out_i = SoftMax(Dot(h, self.v2c))
outputs.append(out_i)
self.output(Collect(outputs))
self.expect(data[:, 1:])
def build_graph(self, batch):
self.reset()
# Build Computation Graph according to length
bsize, length = batch.data.shape
self.h0.value = np.zeros([bsize, self.hidden_dim])
self.c0.value = np.zeros([bsize, self.hidden_dim])
h = self.h0
c = self.c0
outputs = []
for idx in range(length - 1):
in_i = self.input()
in_i.value = batch.data[:, idx] # Get value from batch
x = Embed(in_i, self.embed)
h, c = self.lstm_cell(x, h, c)
out_i = SoftMax(Dot(h, self.v2c))
outputs.append(out_i)
self.output(Collect(outputs))
self.expect(batch.data[:, 1:])
def build_graph(batch):
data = batch.data[1]
decode_graph.reset()
# Build Computation Graph according to length
bsize, length = data.shape
'''
Change the function here to switch between encoders
'''
h, c = lstm_encode(batch.data[0])
outputs = []
for idx in range(length - 1):
in_i = decode_graph.input()
in_i.value = data[:, idx] # Get value from batch
x = Embed(in_i, decode_graph.embed)
h, c = decode_graph.lstm_cell(x, h, c)
out_i = SoftMax(Dot(h, decode_graph.v2c))
outputs.append(out_i)
decode_graph.output(Collect(outputs))
decode_graph.expect(data[:, 1:])
def build_graph(batch):
data = batch.data[0]
graph.reset()
# Build Computation Graph according to length
bsize, length = data.shape
graph.h0.value = np.zeros([bsize, hidden_dim])
graph.c0.value = np.zeros([bsize, hidden_dim])
h = graph.h0
c = graph.c0
outputs = []
for idx in range(length - 1):
in_i = graph.input()
in_i.value = data[:, idx] # Get value from batch
x = Embed(in_i, graph.embed)
h, c = graph.lstm_cell(x, h, c)
out_i = SoftMax(Dot(h, graph.v2c))
outputs.append(out_i)
graph.output(Collect(outputs))
graph.expect(data[:, 1:])
def build_graph(self, batch):
enc_data = batch.data
self.reset()
bsize, enc_length = enc_data.shape
outputs = []
# Build Encode Graph
self.h0.value = np.zeros([bsize, self.hidden_dim])
self.c0.value = np.zeros([bsize, self.hidden_dim])
h = self.h0
c = self.c0
for idx in range(enc_length):
in_i = self.input()
in_i.value = enc_data[:, idx] # Get value from batch
x = Embed(in_i, self.eembed)
h, c = self.enc_lstm_cell(x, h, c)
# out_i = SoftMax(Dot(h, graph.ev2c))
# outputs.append(out_i)
self.encoded_h = h
self.encoded_c = c
# Build Decode Graph
decode_in = self.input()
decode_in.value = np.zeros([bsize])
decode_embed = Embed(decode_in, self.dembed)
x = decode_embed
for idx in range(self.predict_len):
h, c = self.dec_lstm_cell(x, h, c)
out_i = ArgMax(SoftMax(Dot(h, self.dv2c)))
outputs.append(out_i)
x = Embed(out_i, self.dembed)
self.output(Collect(outputs))
self.expect(np.zeros([bsize, self.predict_len]))
def build_graph(self, batch):
enc_data = batch.data[0]
dec_data = batch.data[1]
self.reset()
bsize, enc_length = enc_data.shape
dec_length = dec_data.shape[1]
outputs = []
# Build Encode Graph
self.h0.value = np.zeros([bsize, self.hidden_dim])
self.c0.value = np.zeros([bsize, self.hidden_dim])
h = self.h0
c = self.c0
for idx in range(enc_length):
in_i = self.input()
in_i.value = enc_data[:, idx] # Get value from batch
x = Embed(in_i, self.eembed)
h, c = self.enc_lstm_cell(x, h, c)
# out_i = SoftMax(Dot(h, graph.ev2c))
# outputs.append(out_i)
self.encoded_h = h
self.encoded_c = c
# Build Decode Graph
for idx in range(dec_length - 1):
in_i = self.input()
in_i.value = dec_data[:, idx]
x = Embed(in_i, self.dembed)
h, c = self.dec_lstm_cell(x, h, c)
out_i = SoftMax(Dot(h, self.dv2c))
outputs.append(out_i)
self.output(Collect(outputs))
self.expect(dec_data[:, 1:])
def build_graph(self, batch):
enc_data = batch.data
dec_data = batch.expect
self.reset()
bsize = 1
enc_length = enc_data.shape[1]
dec_length = dec_data.shape[0]
outputs = []
hcollect = []
ccollect = []
self.feh0.value = np.zeros([bsize, self.half_dim])
self.fec0.value = np.zeros([bsize, self.half_dim])
self.beh0.value = np.zeros([bsize, self.half_dim])
self.bec0.value = np.zeros([bsize, self.half_dim])
for line_idx in range(self.num_line):
# Build Fwd Encode Graph
fh = self.feh0
fc = self.fec0
for idx in range(enc_length):
in_i = self.input()
in_i.value = enc_data[line_idx,
idx].reshape(1) # Get value from batch
x = Embed(in_i, self.feembed)
fh, fc = self.fenc_lstm_cell(x, fh, fc)
# Build Bwd Encode Graph
bh = self.beh0
bc = self.bec0
for idx in range(enc_length):
in_i = self.input()
in_i.value = enc_data[line_idx, enc_length - 1 - idx].reshape(
1) # Get value from batch
x = Embed(in_i, self.beembed)
bh, bc = self.benc_lstm_cell(x, bh, bc)
h = Concat(fh, bh)
c = Concat(fc, bc)
hcollect.append(h)
ccollect.append(c)
# Build Decode Graph
h = Average(Collect(hcollect))
c = Average(Collect(ccollect))
self.encoded_h = h
self.encoded_c = c
for idx in range(dec_length - 1):
in_i = self.input()
in_i.value = dec_data[idx].reshape(1)
x = Embed(in_i, self.dembed)
h, c = self.dec_lstm_cell(x, h, c)
out_i = SoftMax(Dot(h, self.dv2c))
outputs.append(out_i)
self.output(Collect(outputs))
self.expect(dec_data[1:])
def build_graph(self, batch):
enc_data = batch.data
self.reset()
bsize, enc_length = enc_data.shape
outputs = []
fwd_encode_result = [None] * enc_length
bwd_encode_result = [None] * enc_length
# Build Fwd Encode Graph
self.feh0.value = np.zeros([bsize, self.half_dim])
self.fec0.value = np.zeros([bsize, self.half_dim])
fh = self.feh0
fc = self.fec0
for idx in range(enc_length):
in_i = self.input()
in_i.value = enc_data[:, idx] # Get value from batch
x = Embed(in_i, self.feembed)
fh, fc = self.fenc_lstm_cell(x, fh, fc)
fwd_encode_result[idx] = fh
# Build Bwd Encode Graph
self.beh0.value = np.zeros([bsize, self.half_dim])
self.bec0.value = np.zeros([bsize, self.half_dim])
bh = self.beh0
bc = self.bec0
for idx in range(enc_length):
in_i = self.input()
in_i.value = enc_data[:,
enc_length - 1 - idx] # Get value from batch
x = Embed(in_i, self.beembed)
bh, bc = self.benc_lstm_cell(x, bh, bc)
bwd_encode_result[enc_length - 1 - idx] = bh
# Build Decode Graph
h = Concat(fh, bh)
# c = Concat(fc, bc)
self.encoded_h = h
# self.encoded_c = c
encode_result = []
for idx in range(enc_length):
encode_result.append(
Concat(fwd_encode_result[idx], bwd_encode_result[idx]))
encode_state = Collect(encode_result)
init = self.input()
init.value = np.zeros([bsize])
x = Embed(init, self.dembed)
for idx in range(self.predict_len):
c = Attention(encode_state, h)
h, c = self.dec_lstm_cell(x, h, c)
out_i = ArgMax(SoftMax(Dot(h, self.dv2c)))
outputs.append(out_i)
x = Embed(out_i, self.dembed)
self.output(Collect(outputs))
self.expect(np.zeros([bsize, self.predict_len]))