CI/CD Pipeline setup #19
Comments
|
Yan, I'm taking the liberty of copying your initial email here as it is the best place for discussion Hi Ben, This item was assigned to you and me. If you don’t mind, I can do some research on possible solutions and put together a proposal. But we need to first define some requirements. Here are what I can think of: When PR merge request is received: Please feel free to add/change. Depending on the solution options we have, we may need to adjust these. |
|
We have been using 'GitLab CI/CD for GitHub' as it comes with full GPU test agent support. Shall we just go with that? Let's protect master so that it can only be updated via PR. Github provides 'auto-squash' option on PRs so that they are always a single commit on master. This is good because it keeps master history clean. I'd go further and suggest rewriting commit message for the PR so that it summarises the feature. Re checks, we should be able to configure CI so that testing happens automatically on PR submission (i.e. before it has been passed/merged). All unit and integration tests should be run on a CI basis, ideally. We can have a nightly build. We may wish to have longer running training tests that execute on it. It is a bit tricky given the unstable nature of some networks, however. Notification of the team; I guess you mean automatic emails; don't we get them from the issues, updates to the issues? If not, I guess that there is a hook of some kind. |
|
Thanks @atbenmurray, just to confirm that now the master can only be updated via PR, squash merging is enabled for PR. |
* Adding script to run unit tests and example test cases (#29) Adding script to run unit tests and example test cases * initial unit tests for dice loss (#27) * initial unit tests for 2d/3d unet * unit tests update - triggering unit tests via github workflow - renamed testconvolutions.py to test_convolutions.py - test unet test cases as variables for readability * initial unit tests for 2d/3d unet (#26) * initial unit tests for 2d/3d unet * unit tests update - triggering unit tests via github workflow - renamed testconvolutions.py to test_convolutions.py - test unet test cases as variables for readability * 14 code examples of monai input data pipeline (#24) * fixes cardiac example * update example cardiac segmentation * Create .gitlab-ci.yml (#30) an initial step towards #19 * tests intensity normalizer - revised to support both `[key]` and `key` as an input for apply_keys - added `NumpyImageTestCase2D` and `TorchImageTestCase2D` * style updates and new test cases: - adding copyright notice - validate user input before setting class member - one line space after copyright - testing multiple keys input data Co-authored-by: Eric Kerfoot <[email protected]> Co-authored-by: Isaac Yang <[email protected]>
* [DLMED] implement intensity normalization transform design according to our latest discussion: 1. input data is dict format with keys for fields. 2. only based on PyTorch and data shape is channel_last. * 9 part a adding test intensity normalisation transform (#33) * Adding script to run unit tests and example test cases (#29) Adding script to run unit tests and example test cases * initial unit tests for dice loss (#27) * initial unit tests for 2d/3d unet * unit tests update - triggering unit tests via github workflow - renamed testconvolutions.py to test_convolutions.py - test unet test cases as variables for readability * initial unit tests for 2d/3d unet (#26) * initial unit tests for 2d/3d unet * unit tests update - triggering unit tests via github workflow - renamed testconvolutions.py to test_convolutions.py - test unet test cases as variables for readability * 14 code examples of monai input data pipeline (#24) * fixes cardiac example * update example cardiac segmentation * Create .gitlab-ci.yml (#30) an initial step towards #19 * tests intensity normalizer - revised to support both `[key]` and `key` as an input for apply_keys - added `NumpyImageTestCase2D` and `TorchImageTestCase2D` * style updates and new test cases: - adding copyright notice - validate user input before setting class member - one line space after copyright - testing multiple keys input data Co-authored-by: Eric Kerfoot <[email protected]> Co-authored-by: Isaac Yang <[email protected]> * [DLMED] simplify intensity normalization transform for MVP Co-authored-by: Wenqi Li <[email protected]> Co-authored-by: Eric Kerfoot <[email protected]> Co-authored-by: Isaac Yang <[email protected]>
No description provided.
The text was updated successfully, but these errors were encountered: