Janusz Lisiecki

@danbider you can put any value you like (you still need to provide a value in the file) as a label and just don't use it in the pipeline.

Hi @980202006, Do you want to read only a window from the npy file? If so please check the following parameters of [the `numpy reader`](https://docs.nvidia.com/deeplearning/dali/user-guide/docs/operations/nvidia.dali.fn.readers.numpy.html): ``` - rel_roi_end (float or...

The easiest solution would be to use [the `external source` operator](https://docs.nvidia.com/deeplearning/dali/user-guide/docs/examples/general/data_loading/parallel_external_source.html) (especially parallel mode).

@szkarpinski - sure, no rush.

@mayujie, If you cannot share your code I would disable randomness, or set the same seed for the CPU and GPU pipelines and compare the results to see if the...

Hi @danbider, It looks you run out of memory. Maybe one of your models consumes more memory. It is hard to tell. What you can do is to ask Pytorch...

Hi @alvitawa, > For me it appears to be the case that the gpu's memory usage accumulates across runs. I.e. the fist run is fine but the second run requires...

Hi @ricpruss, As the documentation states: ``` Pads all samples with the fill_value in the specified axes to match the biggest extent in the batch for those axes or to...

Hi @timcogan, The decoding itself is just part of the limitations. Other DALI operators also don't support 16-bit images, so we plan to extend decoders and operators at the same...

Hi @timcogan, For medical images, you may consider [cuCIM](https://docs.rapids.ai/api/cucim/stable/), as it provides the ability to provide slinging window reading pattern.