Samuel Pfrommer
Samuel Pfrommer
Same issue. [EDIT]: changing to wired internet fixed this. It seems wandb is a little finicky over wireless, atleast for me.
Ditto! FYI for others pulling nightly will get the feature: https://github.com/Lightning-AI/lightning/pull/17163
On further reflection, this is probably the idea way: * Exponential binning in the backend around the median. No outlier rejection. Serialize the bin bounds as well for the frontend....
Unfortunately, "smooth" histograms ended up being too low-performance. So I think that any kind of client-side outlier rejection is out of the question. I think now the only plausible way...
Probably best implementation would be to have the `log` list take: `io_vals, io_grads, io_grad_norms, param_vals, param_grads, param_grad_norms` The `attach` parameter should be [`vals`, `grads`, `grad_norms`]
This would be a good feature and should be possible. However, displaying profiler statistics as a color coding would clash with the highlight colors to show which node is being...
Really appreciate the detailed feedback! I agree the colors would be better for usability, and having different analysis modes is the better UI. My suggestion was mostly motivated by practical...