binshen

> Hello, thanks for your reply. I executed your code, but the computer's memory would gradually increase to full, and finally the program will be terminated. Is it the problem...

> 你就把他的code转成Tensorflow 2就行了。RTX30 系列的显卡需要更高版本的CUDA，而更高版本的CUDA不支持Tensorflow 1.8。 想问下你有转好的tensorflow 2的代码吗。我这能力有限，报错太多修改不好了 [email protected]

不好意思，刚刚看到了您把reshape的代码没编写到netSC里，抱歉

老师您在代码中通过netCE信道编码网络输出128*16*8*8的输出，但是在后续的reshape后大小变为 128 * （G_s + G_n）* 128，没有太理解您将输出的特征通道数16reshape成了8，然后增大了H*W值得意义，然后利用hard_mask对这个reshape后的8个通道进行了掩码。

> As stated, how can we use CompressAI to compress [y] instead of [y-mean]. Furthermore, how can we understand the advantages of compressing [y-mean]? Thank you very much for your...

想请教一下，在channel.py -> Channel.forward 中 output_real = batch_conv1d(signal_real, cof_real) - batch_conv1d(signal_imag, cof_imag) # (NxP)x(L+SMK-1) output_imag = batch_conv1d(signal_real, cof_imag) + batch_conv1d(signal_imag, cof_real) # (NxP)x(L+SMK-1) 这两句的作用是什么，没有太理解