Implementation Steps - 2021 Requirements (packages needed and versions)

[ya3ya6]

(Please add this implementation guide to the readme file :)

Implementation Guide:

1. Install nvidia cuda toolkit or disable cuda before importing tensorflow in demo, by:

os.environ["CUDA_VISIBLE_DEVICES"] = "-1"

2. use python 3.6.0 (python 3.8 doesn't work with tensorflow v1)
3. use these set of requirements (versions):

numpy<=1.21
tensorflow==1.15
scipy
tqdm
librosa
ipython
contextvars==2.1
prompt-toolkit==2.0.1

4. also download the pretrained data mentioned in the readme file, and extract it in the root directory.

---------------- Optional Step ----------- Optionally If you want to stick to python 3.8 and tensorflow v2, you should import tensorflow v1 instead:

// change import tf to:
import tensorflow.compat.v1 as tf
tf.disable_v2_behavior()

python 3.8 and tensorflow v2 requirements is:

numpy==1.21
tensorflow
scipy
tqdm
librosa

and also you should change tf.variable_scope in model_graph file to tf.compat.v1.variable_scope