Fix generate.py to write summary_out properly

[ghleokim]

Hello,

I am trying to generate audio with this model.

Giving wav_out_path properly, the output seems to be generated as intended. However, when I run a tensorboard server with logdir/ files, the summary file does not show up properly.

I looked up the issue tab on this repo. As mentioned in several other issues like #391, #338, #262, and #232, I couldn't see generated output. My case was the same.

By changing the shape inside summary_out(...) from [-1,1] to [1,-1] solved this problem.

[ghleokim]

1. Tensorboard output with original code

2. Tensorboard output with modified code

3. Also double-checked if the shown audio is identical to the generated output file.