tensor-regression 0.0.4

A small wrapper around pytest_regressions for Tensors

tensor_regression

A small wrapper to simplify using pytest_regressions with Tensors.

This adds the following to pytest_regressions:

• Simple Tensor statistics (min, max, mean, std, shape, dtype, device, hash, etc.) are generated and saved in a .yaml file.
  • The simple statistics are used as a pre-check before comparing the full tensors.
  • These yaml files can be saved with git without having to worry about accidentally saving huge files.
• Full tensors are moved to CPU and saved in a .npy file (same as ndarrays_regression), and these .npy files are gitignored.
• Adds a --gen-missing argument (default True) which will generate any missing regression files without raising error, as opposed to pytest-regression's --regen-all which regenerates all regression files.