I reported this issue for tensorflow.

tensorflow/tensorflow#100689

I think for inaSpeechSegmenter version 0.8.0 and above, if the user is installing it on windows, the library should use the cpu mode directly as they said that GPU mode isn't supported on Windows,

Reference

https://www.tensorflow.org/install/pip#windows-native