Workaround for h5py memory leak

[soumickmj]

No description provided.

[facebook-github-bot]

Hi @soumickmj!

Thank you for your pull request and welcome to our community.

Action Required

In order to merge any pull request (code, docs, etc.), we require contributors to sign our Contributor License Agreement, and we don't seem to have one on file for you.

Process

In order for us to review and merge your suggested changes, please sign at https://code.facebook.com/cla. If you are contributing on behalf of someone else (eg your employer), the individual CLA may not be sufficient and your employer may need to sign the corporate CLA.

Once the CLA is signed, our tooling will perform checks and validations. Afterwards, the pull request will be tagged with CLA signed. The tagging process may take up to 1 hour after signing. Please give it that time before contacting us about it.

If you have received this in error or have any questions, please contact us at cla@fb.com. Thanks!

[facebook-github-bot]

Thank you for signing our Contributor License Agreement. We can now accept your code for this (and any) Meta Open Source project. Thanks!

[mmuckley]

Hello @soumickmj, the copy() commands threw an error for me because h5py datasets don't have the copy() command. You need to convert to a numpy array first.

I tried modifying the PR to convert to a numpy array prior to copy() commands and it didn't for me with the memory issue - I still observed exploding memory. This also included a case where I applied a deepcopy() to the dict.

[mmuckley]

Throws an error.

[mmuckley]

Throws an error.

[mmuckley]

Throws an error.

[soumickmj]

Hi @mmuckley
That's strange. h5py dataset does not have a copy method, but when "hf["kspace"][dataslice]" is called it becomes numpy.ndarray which is then having a copy method.
It is working fine for me with this code:

Converting to numpy explicitly and then copy: I tried earlier and that did not resolve the issue for me as well.
But directly calling copy is working for me.

I wonder what's the difference then.
What's the versinon of Python and h5py are you using?

[mmuckley]

This is the full error:

  File "/private/home/mmuckley/projects/fastmri/code/fastMRI/fastmri/data/mri_data.py", line 350, in __getitem__
    mask = np.asarray(hf["mask"].copy()) if "mask" in hf else None
AttributeError: 'Dataset' object has no attribute 'copy'

So the issue might be that you're calling it on the test set and returning None.

We actually have a unit test for __getitem__ here. Could you make sure that your code is passing all lints and unit tests before I test again? You should be able to get unit tests on your fork by renaming your branch to main.

[mmuckley]

Closing due to inactivity.