Hi, I tried reproducing your results but the files generated by test.py in the tagging model contain ndarrays with nan values.  I guess, in the tagging network when you mention keep_prob, you actually meant rate.
Hi, I tried reproducing your results but the files generated by test.py in the tagging model contain ndarrays with nan values.

I guess, in the tagging network when you mention keep_prob, you actually meant rate.