def seq2seq_f(encoder_inputs, decoder_inputs, do_decode):
return rl_seq2seq.embedding_attention_seq2seq(
encoder_inputs,
decoder_inputs,
cell,
num_encoder_symbols=source_vocab_size,
num_decoder_symbols=target_vocab_size,
embedding_size=size,
output_projection=output_projection,
feed_previous=do_decode,
dtype=dtype)
def seq2seq_f(encoder_inputs, decoder_inputs, do_decode):
return rl_seq2seq.embedding_attention_seq2seq(
encoder_inputs,
decoder_inputs,
cell,
num_encoder_symbols=self.source_vocab_size,
num_decoder_symbols=self.target_vocab_size,
embedding_size=self.emb_dim,
output_projection=output_projection,
feed_previous=do_decode, # 是否把上一轮的预测作为这一轮的输入 || 是否在测试
mc_search=self.mc_search, # TODO(Zhu) 文件位置：seq2seq._argmax_or_mcsearch 什么意思？
dtype=self.dtype)
def seq2seq_f(encoder_inputs, decoder_inputs, do_decode):
return rl_seq2seq.embedding_attention_seq2seq(
encoder_inputs,
decoder_inputs,
cell,
num_encoder_symbols=source_vocab_size,
num_decoder_symbols=target_vocab_size,
embedding_size=emb_dim,
output_projection=output_projection,
feed_previous=do_decode,
mc_search=self.mc_search,
dtype=tf.float32)
