##data类型问题

import paddle.fluid as fluid
import paddle.fluid.layers as layers

emb_dim = 256
vocab_size = 10000
data = fluid.layers.data(name='x', shape=[-1, 100, 1],
dtype='int32')

emb = fluid.layers.embedding(input=data, size=[vocab_size, emb_dim], is_sparse=True)

batch_size = 20
max_len = 100
dropout_prob = 0.2
seq_len = 100
hidden_size = 150
num_layers = 1
init_h = layers.fill_constant( [num_layers, batch_size, hidden_size], 'float32', 0.0 )
init_c = layers.fill_constant( [num_layers, batch_size, hidden_size], 'float32', 0.0 )

rnn_out, last_h, last_c = layers.lstm(emb, init_h, init_c, max_len, hidden_size, num_layers, dropout_prob=dropout_prob)
rnn_out.shape # (-1, 100, 150)
last_h.shape # (1, 20, 150)
layt_c.shape # (1, 20, 150)