crystal0913

我是用带韵律标签去生成lab，然后用这些lab去训练时长模型和声学模型的，不知道您是不是指这个意思？

Just modify the function decode() in decode.py and remove the `batcher` in `__init__` Example: ```py def decode(self, article): example = Example(article, [], self._vocab) b = [example for _ in range(self.batch_size)]...

In this paper [https://arxiv.org/pdf/1609.08144.pdf](https://arxiv.org/pdf/1609.08144.pdf), based on beam search algorithm, they includes a coverage penalty to favor translations that fully cover the source sentence according to the attention module.The scoring function...

Does the dimension match of trained weight W and feature matrix after appending the new node's feature vector ? the dimension of new feature matrix X is (n+1, n+1) while...

链接:https://pan.baidu.com/s/1gZoucMYIIXpd0E4H6pgTKA&shfl=sharepset 密码:bdl1

> @crystal0913 Hi，有个小问题，你的代码在predict的时候每一个隐藏层都用到了batch_normalization，但是我看merlin TTS在做inference的时候，dnn_generation里面应该是没有做batch_norm的。请问你是在训练的时候加了这个步骤，然后保存了每一层的mean和var吗 看了一下[merlin的模型定义](https://github.com/CSTR-Edinburgh/merlin/blob/9df27e9309edccb6e35511ead2e832f2fe98cc15/src/tensorflow_lib/model.py), 是有做batch_norm的。 "请问你是在训练的时候加了这个步骤，然后保存了每一层的mean和var吗" , 是的。

@shakingWaves 这个就是调用merlin提供的generate_wav， gen_test.py内容如下： import sys import os import configuration from utils.generate import generate_wav import logging import logging.config from logplot.logging_plotting import LoggerPlotter, cfg=configuration.cfg logging.setLoggerClass(LoggerPlotter) logger = logging.getLogger("main") config_file = sys.argv[1] config_file...