Machine learning

Current implementation of Go binding can not specify options.

GPUOptions struct is in internal package. And go generate doesn't work for protobuf directory. So we can't specify GPUOptions for NewSession.

Followup to pytorch/pytorch#74955 (comment).

It turns out that that cmake version was just bad and we can now unpin cmake once again.

cc @seemethere @malfet @pytorch/pytorch-dev-infra

Currently many more Python projects like dask and optuna are using Python type hints. With the Python package of xgboost gaining more and more features, we should also adopt mypy as a safe guard against some type errors and for better code documentation.

What happened + What you expected to happen

The autoscaler pushes some logs to the Ray driver. The logs are prefixed with (scheduler) which is misleading.
The prefix should be (autoscaler).

Versions / Dependencies

Ray master.

Reproduction script

Submit a Ray task or actor which triggers upscaling to trigger the logs.

Issue Severity

Low: It annoys or frustrates me.

Problem

See #3856 . Developer would like the ability to configure whether the developer menu or viewer menu is displayed while they are developing on cloud IDEs like Gitpod or Github Codespaces

Solution

Create a config option

showDeveloperMenu: true | false | auto

where

• true: always shows the developer menu locally and while deployed
• false: always sho

🐛 Bug

Trainer profilers are typehinted with the deprecated BaseProfiler instead of Profiler. This means that you cannot use class_path initialization of profilers with LightningCLI.

Error message:

  - "pytorch_lightning.profiler.PyTorchProfiler" is not a subclass of <class 'pytorch_lightning.profiler.base.BaseProfiler'>
  - Expected a <class 'str'> but got "{'class_path':