Cliff Woolley

You can easily query which algorithm cuDNN's heuristic has selected for your problem size and memory availability. You can also just pick an algorithm and force cuDNN to use it....

See https://github.com/lspvic/jupyter_tensorboard/pull/63

@asford - I've merged this PR into https://github.com/lspvic/jupyter_tensorboard/pull/63 because it's needed also for the Scalars card now in TensorBoard 2.4. Thank you!

If you want to use TB 2.3 or later, can you please try with https://github.com/lspvic/jupyter_tensorboard/pull/63 ?

https://github.com/lspvic/jupyter_tensorboard/pull/63 should fix the projector plugin, among other things.

If you want to use TB 2.3 or later, can you please try with https://github.com/lspvic/jupyter_tensorboard/pull/63 ?

Hi all, FYI the original issue here is fixed up in our 21.02 version of NVIDIA containers due to be released very soon -- the fixes needed were in jupyter-tensorboard...

For TensorBoard 2.2 or later as seen here, you'd probably be needing https://github.com/lspvic/jupyter_tensorboard/pull/63 .

/cc @wchargin

I've updated this PR from last year to also include TB 2.3 and TB 2.4 now. Plus the newer, simplified implementation should allow better future-proofness as well: Instead of pulling...