LeeRel1991

Results 4 comments of LeeRel1991

Hi @kitstar , would you like to implemet the Pack, Shape,Conv2DBackpropInput operators in your tensorflow parser? cause it is necessary for transpose convolution. the exact net I'm doing is resfcn256...

from [#28](https://github.com/huggingface/pytorch-openai-transformer-lm/issues/28) * n_ctx is the maximum number of token in an input sequence. * n_special is the number of special tokens used to format the input properly. For example...

我最开始训练时也总出现类似的问题,predict的关键点有偏移,进过分析后感觉是归一化的问题,后来尝试了各种归一化方法,有一定的改善,但并未完全解决。另外mask也有一些问题。 训练代码会逐步提供,但由于近期比较忙,所以可能得一段时间,抱歉,不过可以参考其他人的实现:https://github.com/jnulzl/PRNet。

prnet的训练好的模型在[作者官方版本](https://github.com/YadiraF/PRNet)里可以下载,具体的下载链接为百度盘或谷歌盘(https://github.com/YadiraF/PRNet#usage)