fixes #550: move torch tensors to cpu before to numpy array

[dkgaraujo]

As mentioned in keras-team/keras#18437, one remaining change from #750 is to first move the torch tensor to the CPU before passing it to a numpy array.

This is relevant for the Metal GPU case. While calling .cpu() is harmless on a torch tensor already in the CPU, in my observation directly moving the tensor from Metal to numpy returns an error, hence the need to move it first.

[google-cla]

Thanks for your pull request! It looks like this may be your first contribution to a Google open source project. Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA).

View this failed invocation of the CLA check for more information.

For the most up to date status, view the checks section at the bottom of the pull request.

[dkgaraujo]

Another thing I would mention is that the instructions for macOS users could be updated to reflect that access to GPU requires setting another environment variable as follows:

import os
os.environ[“KERAS_BACKEND”] = “torch”
os.environ[“PYTORCH_ENABLE_MPS_FALLBACK”] = “1”
import keras_core as keras

[codecov]

Codecov Report

Patch coverage: 100.00% and no project coverage change.

Comparison is base (f350416) 76.00% compared to head (ae75181) 76.00%.
Report is 2 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main     keras-team/keras-core#853   +/-   ##
=======================================
  Coverage   76.00%   76.00%           
=======================================
  Files         328      328           
  Lines       31103    31103           
  Branches     6052     6052           
=======================================
  Hits        23639    23639           
  Misses       5866     5866           
  Partials     1598     1598           

Flag	Coverage Δ
keras_core	75.91% <100.00%> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

Files Changed	Coverage Δ
..._core/trainers/data_adapters/torch_data_adapter.py	86.95% <100.00%> (ø)

☔ View full report in Codecov by Sentry.

📢 Have feedback on the report? Share it here.

[dkgaraujo]

I signed the CLA - I'm not sure if I should take another step to officially pass that check.

[fchollet]

LGTM, thanks! You can ignore the CLA issue.