- Implements helper functions to load BioViL-T text and image models.
- Updates the text and image inference engines to consume Biovil-T
models.
- Updates the grounding notebook with the new model and showing improved
results.
---------
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Fernando Pérez-García <fepegar@gmail.com>
Co-authored-by: Fernando Pérez-García <fperezgarcia@microsoft.com>
- Splits the content in `image/model/model.py` into separate files with
respect to their functionality, such as encoders and custom types.
- Creates a new encoder class that can handle multiple image inputs such
as current scan and previous scan. This way models can embed
multi-series data into a single vector representation.
<!--
## Guidelines
Please follow the guidelines for pull requests (PRs) in
[CONTRIBUTING](/CONTRIBUTING.md). Checklist:
- Ensure that your PR is small, and implements one change
- Give your PR title one of the prefixes listed in
[CONTRIBUTING](../docs/source/coding_guidelines.md#pull-request-titles)
to indicate what type of change that is
- Link the correct GitHub issue for tracking
- Add unit tests for all functions that you introduced or modified
- Run automatic code formatting / linting on all files ("Format
Document" Shift-Alt-F in VSCode)
- Ensure that documentation renders correctly in Sphinx by running `make
html` in the `docs` folder
## Change the default merge message
When completing your PR, you will be asked for a title and an optional
extended description. By default, the extended description will be a
concatenation of the individual
commit messages. Please DELETE/REPLACE that with a human readable
extended description for non-trivial PRs.
-->
---------
Co-authored-by: Stephanie Hyland <sthyland@microsoft.com>
Co-authored-by: Stephanie Hyland <corcra@users.noreply.github.com>
Co-authored-by: Fernando Pérez-García <fepegar@gmail.com>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Add custom forward that checkpoints activations to trade compute for
memory
---------
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
- Regression metrics are now read and compared semantically rather than
as text files. Hence, formatting of the JSON files no longer matters.
- Fix for bug: if a scalar metric is read from AML before a list metric,
the dataframe conversion fails. Hence we can now mix scalar and list
metrics without problems.
LoadMaskSubROId loads a sub section from a slide based a foreground png
mask with categorical labels for each tissue section.
---------
Co-authored-by: Anton Schwaighofer <antonsc@microsoft.com>
This PR contains changes necessary to enable running [MONAI's Auto3dSeg
tool](https://github.com/Project-MONAI/tutorials/tree/main/auto3dseg)
using hi-ml as a built-in feature for submitting to Azure:
- Allowing scripts of the form `python -m <module name>` to be run as an
AzureML job.
- Updating argument parsing to allow passing command line args
containing quotes.
- Replace logging with `print` statements to show job URL in other
packages.
---------
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Fixes#823.
There was an issue with the Azure ML Python SDK v2 disliking symlinks in
the source dir, but that seems to have been resolved now so this PR
removes the copying operation and replaces the targets with symlinks.
Fixes#805.
Results with the new image:
This image is very large, though :`(3950, 3300)`. Feel free to propose a
smaller one from the internet. Perhaps from Open-i.
<!--
## Guidelines
Please follow the guidelines for pull requests (PRs) in
[CONTRIBUTING](/CONTRIBUTING.md). Checklist:
- Ensure that your PR is small, and implements one change
- Give your PR title one of the prefixes listed in
[CONTRIBUTING](../docs/source/coding_guidelines.md#pull-request-titles)
to indicate what type of change that is
- Link the correct GitHub issue for tracking
- Add unit tests for all functions that you introduced or modified
- Run automatic code formatting / linting on all files ("Format
Document" Shift-Alt-F in VSCode)
- Ensure that documentation renders correctly in Sphinx by running `make
html` in the `docs` folder
## Change the default merge message
When completing your PR, you will be asked for a title and an optional
extended description. By default, the extended description will be a
concatenation of the individual
commit messages. Please DELETE/REPLACE that with a human readable
extended description for non-trivial PRs.
-->
---------
Co-authored-by: Stephanie Hyland <corcra@users.noreply.github.com>
Resolves#789.
<!--
## Guidelines
Please follow the guidelines for pull requests (PRs) in
[CONTRIBUTING](/CONTRIBUTING.md). Checklist:
- Ensure that your PR is small, and implements one change
- Give your PR title one of the prefixes listed in
[CONTRIBUTING](../docs/source/coding_guidelines.md#pull-request-titles)
to indicate what type of change that is
- Link the correct GitHub issue for tracking
- Add unit tests for all functions that you introduced or modified
- Run automatic code formatting / linting on all files ("Format
Document" Shift-Alt-F in VSCode)
- Ensure that documentation renders correctly in Sphinx by running `make
html` in the `docs` folder
## Change the default merge message
When completing your PR, you will be asked for a title and an optional
extended description. By default, the extended description will be a
concatenation of the individual
commit messages. Please DELETE/REPLACE that with a human readable
extended description for non-trivial PRs.
-->
---------
Co-authored-by: Shruthi42 <13177030+Shruthi42@users.noreply.github.com>
When guessing the number of output dimensions in the image encoder, the
model is in training mode and the batch norm stats get updated. This
shouldn't really happen and makes the checks in
https://github.com/microsoft/hi-ml/pull/807/files#r1112200449 fail. This
PR adds a line to ensure the encode is in `eval` mode before calculating
the number of output dims.
Fixes the issue described in
- https://github.com/microsoft/hi-ml/pull/730#issuecomment-1419302573
<!--
## Guidelines
Please follow the guidelines for pull requests (PRs) in
[CONTRIBUTING](/CONTRIBUTING.md). Checklist:
- Ensure that your PR is small, and implements one change
- Give your PR title one of the prefixes listed in
[CONTRIBUTING](../docs/source/coding_guidelines.md#pull-request-titles)
to indicate what type of change that is
- Link the correct GitHub issue for tracking
- Add unit tests for all functions that you introduced or modified
- Run automatic code formatting / linting on all files ("Format
Document" Shift-Alt-F in VSCode)
- Ensure that documentation renders correctly in Sphinx by running `make
html` in the `docs` folder
## Change the default merge message
When completing your PR, you will be asked for a title and an optional
extended description. By default, the extended description will be a
concatenation of the individual
commit messages. Please DELETE/REPLACE that with a human readable
extended description for non-trivial PRs.
-->
In this PR:
- `stratify_plots_by` parameter is introduced in BaseMIL to specify
metadata column on which the output PR and ROC curves can be stratified
- Functionality is added to plot the output curves stratified by
metadata
- Corresponding tests are added
---------
Co-authored-by: Harshita Sharma <t-hsharma@microsoft.com>
Closes#704. Includes the following changes:
* Allow users to access data using their own identity. This means that
we no longer need to store Account Keys / SAS tokens in the datastores
and avoids the information governance concerns there. Used by default,
if not available / not valid will revert to trying datastore identity
auth.
* Adds functionality for accessing input and output datasets in v2 jobs
via environment args rather than command line arguments.
---------
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Add resubmit_crossval_child flag to be able to submit a single fold run.
This is useful in case one of the crossvalidation folds fails and needs
to be resubmitted manually.
ROC curve is added as a plot option. Tests are updated for ROC and PR
curves
---------
Co-authored-by: Harshita Sharma <t-hsharma@microsoft.com>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Peter Hessey
Ozan Oktay
Kenza Bouzid
Anton Schwaighofer
dependabot[bot]
pre-commit-ci[bot]
Fernando Pérez-García
Peter Hessey
Harshita Sharma
