Cameron Arshadi comments

Results 4 comments of


                                            Cameron Arshadi

ValueError: conflicting sizes for dimension in xr.open_dataset("reference://"...) VS. no error in xr.open_dataset(direct_file_path) for h5

I am also seeing this when trying to open a multiscale OME-Zarr dataset, since the scale arrays all have different dimensions. ``` File "C:\Users\cameron.arshadi\repos\aind-morphology-utils\src\aind_morphology_utils\movie_maker.py", line 592, in main() File "C:\Users\cameron.arshadi\repos\aind-morphology-utils\src\aind_morphology_utils\movie_maker.py",...

write_multiscale() writes to local path on Linux instead of cloud bucket when using dask>=2023.1.0

I agree that specifying `chunks` in `storage_options` is not necessary, since that option is ignored by the FSStore constructor and the zarr chunks are based on the dask array chunks.

How to train on only one GPU?

Also tried reducing the batch size to 32,16,etc but the program got stuck before even starting to iterate

[BUG] Cell detection ignores borders

Thanks @adamltyson , padding the image with zeros works ![image](https://user-images.githubusercontent.com/44130022/190221135-90955bb4-464b-4f44-9430-da3b7f15c8b3.png)