Visit the real daft-exprt github page for more information.

To make the above work in Colab, you will probably need access to Colab's terminal, which will require Colab pro [ This is probably removable, but I have no clue on how to do it ]

Steps:

• Create a cell that does the following, and run it.

!wget -O mini.sh https://repo.anaconda.com/miniconda/Miniconda3-py38_4.8.2-Linux-x86_64.sh
!chmod +x mini.sh
!bash ./mini.sh -b -f -p /usr/local

• Run the following steps in the terminal

eval "$(conda shell.bash hook)" 
conda create -n daft_exprt python=3.8 -y 
conda activate daft_exprt
conda config --add channels conda-forge
conda install montreal-forced-aligner
pip install -r pip_requirements.txt
pip install pyworld
pip install -e ../.[pytorch] --find-links https://download.pytorch.org/whl/torch_stable.html
make MFA_pretrained
conda install ipykernel

That's it! When running scripts, please use:

/usr/local/envs/daft_exprt/bin/python

instead of python, and the jupyter notebook itself does not "really" work yet.

It's generally recommended to use the GPU only on the actual model training.

Main Changes:

• Changed REAPER to pyworld, so that GLIBC 2.29 errors are avoided
• Updated MFA install to most recent version, which changes some parameters
• Found a solid method of getting conda with environments 'working' [ Can we just support it? ]