Install on macOS¶
Running DELTA training on a macOS is mostly the same as running on Linux, except some minor differences.
You need to set up a working Python 3.6.x environment, either by using conda or manually build from source.
You can follow the instructions in
manual_setup.md to set up python and the required packages, e.g. Tensorflow.
tensorflow-gpu requires nvidia GPU, which might not be supported the latest macOS versions. You may want to use the
tensorflow package (no -gpu postfix) instead. Some models that uses cuDNN implementations will not work without a CUDA GPU however.
Notes for Kaldi¶
Building and running Kaldi on a macOS requires
gawk and other utilities which need to be installed via
https://brew.sh for details.
/usr/bin/ruby -e "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/master/install)" brew install wget gawk grep
mmseg package for Python2 is needed:
pip2 install mmseg
DELTA install section to install 3rd-party dependencies.