Janusz Lisiecki

Hi @zmasih, DALI maintains a memory pool to speed memory allocation for GPU and CPU pinned memory. However, the pool is not meant to carry information between pipeline instances but...

Hi @fversaci, We have a couple of ideas on how to approach this but this requires a noticeable effort and we cannot commit to any particular date at the moment....

Will keep you posted.

IBM conda package is for POWER architecture. Using conda it is still possible to install pip whl and this works in DALI case (even if it is not the most...

Right, I forgot about binary compatibility problems. Libs itself are not a big deal - DALI hashes the lib names and do other magic to prevent symbol clash for the...

Hi @otaj, I'm sorry, we don't have any updates we can share now. The DALI conda release is a part of a bigger strategy which details we can't share here.

Hi @buchholzmd, Thank you for reaching out. Indeed, DALI is a part of the Open Cognitive Environment (Open-CE), but the effort is not run by NVIDIA. Regarding the original ask,...

Hi @kristinat8, Thank you for reaching out. Can you try: ``` combined = fn.cat(decoded_imgs, depths.gpu(), axis=2) ```

Hi @kristinat8, I'm not sure if I understand your ask. - if you want to specify the `depths` layout - please use `layout` argument of [the `external source` operator](https://docs.nvidia.com/deeplearning/dali/user-guide/docs/operations/nvidia.dali.fn.external_source.html#nvidia.dali.fn.external_source) -...

Hi @kristinat8, Can you provide a full error log as it should print the name of the image that caused this error? It seems that one of the images in...