BUG: `Cannot convert a MPS Tensor to float64 dtype` on Apple M1 Max

[NickleDave]

BUG: `Cannot convert a MPS Tensor to float64 dtype` on Apple M1 Max

[NickleDave]

Copying my reply from the forum:
https://forum.vocalpy.org/t/vak-tweetynet-with-an-apple-m1-max/78/11?u=nicholdav

I think the problem isn't the way we save the arrays for the dataset. I confirmed this shouldn't be an issue but I'll spare you the details (basically, we transform inputs to float32 when we load them).

(I first thought it was because we were loading spectrograms as float64, but in fact we apply a transform to make them float32)

My best guess is that the error happens when lightning tries to take a computed metric value that is returned and put it in a tensor, in order to log that value. Because the returned value is float64, we get this error.

From the traceback you provided, we see where we're calling self.log:

 File "/opt/anaconda3/envs/TweetyNet/lib/python3.11/site-packages/vak/models/windowed_frame_classification_model.py", line 208, in validation_step
    self.log(f'val_{metric_name}', metric_callable(y_pred_labels, y_labels), batch_size=1)

Here's line 208 in version 1.0.0a1 that you're using:
https://github.com/vocalpy/vak/blob/3dcce70030ae9b1fd6d040e055def0d656a7512e/src/vak/models/windowed_frame_classification_model.py#L208

You can see that we're computing an edit distance metric using string labels. So it can't be the tensor inputs to the model, and it has to be the returned value. I put in a breakpoint() before that line and ran an eval file to confirm that, yes, the segment_edit_distance returns a numpy float with dtype float64. I'm guessing that's what causes the crash.

I think I have a fix here:
https://github.com/vocalpy/vak/tree/make-distance-metrics-return-tensors

I will raise a separate issue stating we need all metrics to return tensors, apply the fix I have in progress in that branch, and then release a new version

[NickleDave]

@all-contributors please add @VenetianRed for bug

[allcontributors]

@NickleDave

I've put up a pull request to add @VenetianRed! 🎉