Karan Sikka

If anyone wants to try an easy fix that's probably incorrect in a lot of cases, but will likely cut incremental compile time in half: https://github.com/TypeStrong/tsify/compare/master...ksikka:perf-fix I don't have access...

I reinstalled (yes removed drivers on uninstall), no fix for me.

+1 for a callback that gets called after some piece of the DOM is updated by rivets. It would allow our parent views to listen for subview changes and potentially...

+1

Hi @Wulin-Tan and @YitingChang, I am starting to investigate support for multi-GPU on lightning pose. Could you provide some more context about your goals with multi-gpu - i.e. are you...

Hi @YitingChang and @Wulin-Tan support for multi-GPU was just added for supervised training. This works for both heatmap and heatmap_mchrnn as long as `losses_to_use: []`. To use it, set `num_gpus`...

Without multi-gpu on unsupervised, you might have success with gradient accumulation (see `accumulate_grad_batches` in the [configuration file docs](https://lightning-pose.readthedocs.io/en/latest/source/user_guide/config_file.html). Say you want an effective batch size of 8 but can only...

Hi @YitingChang and @Wulin-Tan support for multi-GPU was just added for unsupervised training and is now ready for you to try out! We've tested it and been pleased with the...

The lightning framework we use is auto-detecting SLURM and trying to do something smart. We haven't tested with SLURM before. The first thing to try is to unset some SLURM...

Thanks, I would try changing --ntasks-per-node to 4. Feel free to message on discord where we can go back and forth more quickly.