Anton Schwaighofer

https://www.microsoft.com/en-us/research/people/antonsc/

Microsoft Research Cambridge, UK Principal ML Engineer @ Microsoft Health Futures

Results 50 comments of


                                            Anton Schwaighofer

Ensure that all AML runs from one PR build end up in the same experiment

Similar to #271

Support for more complex operations on YAML config files, working off YACS

Idea for inheritance: ``` TEMPLATE: ../foo/bar.yml SYSTEM: NUM_GPUS: 2 TRAIN: SCALES: (1, 2) ``` this should allow multiple levels of inheritance

Support for more complex operations on YAML config files, working off YACS

templates could also be added for specific parts of the tree

Support for more complex operations on YAML config files, working off YACS

Allow type checking on fields that are None by default. Could use `params` library, or do introspection to retrieve the type annotation

Bump pypa/gh-action-pypi-publish from 1.5.0 to 1.5.1

superseded by #589

ENH: Upgrade hi-ml-cpath to Python 3.9

> Why not straight to version 3.10? @mebristo , because some AzureML packages are not yet supported by 3.10

pylint is set up for use in VSCode, but it is not run as a tool in the workflows

torch will raise lots of problems, see https://stackoverflow.com/questions/50319943/pytorch-error-message-torch-has-no-member for a solution

hi-ml-multimodal should be included in coverage reports

@fepegar FYI

Bump numpy from 1.21.6 to 1.22.0 in /hi-ml-cpath

@fepegar , good point about jumping to 3.9, will get going with that.

Bump numpy from 1.21.6 to 1.22.0 in /hi-ml-cpath

Superseded by #589

1
2
3
4
5
›