Jinze Xue

 A dirty work around: At this line https://github.com/wookayin/gpustat/blob/master/gpustat/core.py#L217 change `term.bold_black` to `term.bold_gray`

Any updates? I’m also observing this issue > switching to another app and switching back doesn't seem to preserve the scroll position.

Hi, thanks for the report! Could you provide a minimal example to reproduce this?

Hi, the error came from the openmm-ml wrapper. A temp fixed version work ONLY for GPU could be found at: https://github.com/yueyericardo/openmm-ml/commit/1d1d3f24f40becdcd8a36431c8d0900d98eb1304#diff-911692ca194bf903c77d038662969ad3277dcf2fa8b3b3048d95a5aa3af59de1 It is using cuaev `use_cuda_extension` for aev calculation, but...

After using share mem for ShfZ & ShfA ``` RUN Total AEV Forward Backward Force Optimizer Others Epoch time GPU 0 cu Energy train 21.4 ms 12.7 ms 15.6 ms...

original benchmark result: ``` ----------------------------------------------------------------------------------------------------------------------------------------- RUN PDB Size forward backward Others Total Total(200) Speedup GPU ----------------------------------------------------------------------------------------------------------------------------------------- 01 py aev fd small.pdb 264 1.8 ms 0.0 ms 0.0 ms 1.8 ms...

According to the notice, following updates are moved privately.

ping @samdow @zou3519

Great thanks to @zou3519 for the quick debugging!! I'm working on NVIDIA [Modulus](https://developer.nvidia.com/modulus) project, we are using functorch because it provides a lot of perfs for the Jacobian and Hessian...

Hi @zou3519 Our source code is free to download from the website, but it is not developed on GitHub. And our code base is also might too large to put...