Convert kaldi feature extraction and nnet3 models into Tensorflow Lite models. Currently aimed at converting kaldi's x-vector models and diarization pipelines to tensorflow models.
The new version of tensorflow also adds tf.random.normal to list
of supported ops in TF Lite. This will now allow dithering to be
enabled in the TFLite models. So tests that were previously skipped
when dithering was enabled are now run.
tf.random.normal
to list of supported ops in TF Lite. This will now allow dithering to be enabled in the TFLite models. So tests that were previously skipped when dithering was enabled are now run.