Rick Lan

@NadineSf What is your annotations? Are you trying to do OCR?

Useful if no new images are introduced and existing ones are frequently updated. However, in reality, new images are added more often and existing ones (eg latest, latest-gpu) rarely change.

If scripting and ENV are used to drive the image building process, it is not possible to use automated build nor docker's internal build infrastructure, because (1) scripting is not...

These containers might be better optimized for Nvidia hardware than the generic ones I built.

More about LTS: https://github.com/pytorch/pytorch.github.io/issues/828

@andresmafladelgado Would you share your command-line and cfg files? Did you modify the code base?

LTS is deprecated. https://pytorch.org/blog/pytorch-enterprise-support-update/

Looking at [releases](https://docs.nvidia.com/deeplearning/frameworks/pytorch-release-notes/rel-24-01.html#rel-24-01) from Nvidia NGC, it seems that each release use the latest version of each component (e.g. PyTorch, CUDA and TensorRT). This is different from the release matrix...