Usage Recommendations (Teradata Package for Python on IBM PowerPC) - Teradata Package for Python

Recommendations

• Use conda-forge channel provided by conda package manager for scientific packages.

  conda install -c conda-forge numpy pandas scikit-learn lightgbm matplotlib

• Use pip only for pure Python packages.

  pip install teradataml --no-cache-dir

• Keep dependencies consistent.

  conda list  # Check what's installed via conda
  pip list    # Check what's installed via pip

Considerations

• Don’t install scientific packages via pip on PowerPC.

  pip install numpy pandas scikit-learn  # This will fail with compiler errors

• Don't try to compile from source.

  pip install --no-binary=:all: numpy   # This will fail on PowerPC

• Don't mix conda and pip for the same packages.

  conda install numpy
  pip install numpy  # This can cause conflicts

Environment Support