Installing Dependencies

Procedure

1. Use PuTTY to log in to the server as the root user.
2. Install Python.

   yum install python3-devel python3-pip

3. Check whether Python is successfully installed.

   python3 --version

   If the correct version is displayed, the installation is successful.

4. Create a Keras virtual environment.

   pip3 install virtualenv
   mkdir -p /path/to/keras
   python3 -m venv /path/to/keras
   source /path/to/keras/bin/activate

5. Go to the Keras environment.

   cd /path/to/keras

6. Download the TensorFlow package.

   wget https://github.com/lhelontra/tensorflow-on-arm/releases/download/v2.4.0/tensorflow-2.4.0-cp37-none-linux_aarch64.whl

7. Install TensorFlow.

   pip3 install tensorflow-2.4.0-cp37-none-linux_aarch64.whl --trusted-host mirrors.tools.huawei.com -i http://mirrors.tools.huawei.com/pypi/simple

   TensorFlow may fail to be installed due to h5py.

8. Download the h5py package and decompress it.

   wget http://mirrors.tools.huawei.com/pypi/packages/5f/97/a58afbcf40e8abecededd9512978b4e4915374e5b80049af082f49cebe9a/h5py-2.10.0.tar.gz
   tar xvf h5py-2.10.0.tar.gz

9. Install h5py.

   cd h5py-2.10.0
   python3 setup.py install

   Information similar to the following is displayed:

   

10. Install TensorFlow again.

    cd .. pip3 install tensorflow-2.4.0-cp37-none-linux_aarch64.whl --trusted-host mirrors.tools.huawei.com -i http://mirrors.tools.huawei.com/pypi/simple

    Information similar to the following is displayed: