* Add hubconf file
* Refactor to minimise hubconf dependencies
* Pin hubconf dependencies
* Revert "Pin hubconf dependencies"
This reverts commit bc904a963e as it
didn't seem to work.
* Add support for newer versions of torch
* Add only the model folder to path
* Remove unnecessary try-except block
* Avoid duplicate definition of Hugging Face strings
* Import from a more appropriate module
* Add test to compare package and PyTorch Hub models
* Add version number to package __init__
* Remove branch name from PyTorch Hub repo string
* Check only fields from package model
* Remove unnecessary zip wrap
Co-authored-by: Shruthi42 <13177030+Shruthi42@users.noreply.github.com>
Co-authored-by: Shruthi42 <13177030+Shruthi42@users.noreply.github.com>
Change LABEL_COLUMN and N_CLASSES to be instance attributes instead of class attributes to offer more flexibility to change labels for a same data cohort and make split labels global in the repo.
Closes#480
* Add `run_inference_only` flag to trigger trainer.test only
* Add `src_checkpoint` to specify the azure ml run id from which to retrieve the weights checkpoint for inference or resume training
* Rename `ckpt_run_id` -> `ssl_ckpt_run_id` in BasEMIL container to differentiate between different checkpoint run_ids
* ENH: Update tiling script
* slowly updating the create tiles dataset script
* removed logging
* tiling script is updated
* submodule commit test
* [pre-commit.ci] auto fixes from pre-commit.com hooks
for more information, see https://pre-commit.ci
* remove subfolder
* solve flake8
* updated PANDA 5X path
* removed dublicated code
* [pre-commit.ci] auto fixes from pre-commit.com hooks
for more information, see https://pre-commit.ci
* flake8
* cleaned up test
* [pre-commit.ci] auto fixes from pre-commit.com hooks
for more information, see https://pre-commit.ci
* added cucim work around
* added cucim work around
* move test to gpu because of cucim
* test
* test
* make test gpu only
* update readme for gpu test
* [pre-commit.ci] auto fixes from pre-commit.com hooks
for more information, see https://pre-commit.ci
* update readme for gpu test
* added original function because of cucim nightmare
* another cucim work around
* remove relative histo folder
* added test
* more tests
* fix mypy
* make panda 5X default
* revert changes in prad dataset
* after Daniel's review
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>
Put all packages that are required only by the cpath folder into a requirements file, to later consume in package setup. Adding a script to create a full locked Conda environment.
Removed the automatic merging of requirements files of hi-ml because that now conflicts with the full environment.
* Update PR titles
Update PR titles for PR to be more precise
* Add more categories in PR titles
Add Test, fix, mnt and perf, rename style to sty.
* Remove Style
Remove Style as Sty is short alternative for style
Refactor MLRunner to contain all training code.
- Remove model_train function so that we can reuse the trainer for validation
- Break long functions into smaller testable functions
- Kill DDP processes after validation and load model checkpoints once for extra validation epoch and test
- Fix unit tests accordingly
* Prepare codebase for image model release
* Fix interpolation mode
* Fix circular imports
* Address reviewers' comments
* Improve documentation for get_similarity_map_from_raw_data
* Address comments on image module
* Add URL to image model weights
* Fix function name
* Support passing the weights path as Path
* Select device dynamically
* update hf paths
* Fix mypy error
* Fix alignment mode error when using NN interpolation
* Fix link to model weights
* Load weights on CPU and choose device in notebook
* Use same tag for text and image models
* Add missing return type
* Remove unused attribute
Co-authored-by: Ozan Oktay <ozan.oktay@microsoft.com>
Co-authored-by: Shruthi42 <13177030+Shruthi42@users.noreply.github.com>
* add hf revision and unify calls
* add missing header template
* Minor changes
* Fix text imports
* add a commit tag instead of hash since our CI cannot differentiate str from passcode
Co-authored-by: Fernando Pérez-García <fperezgarcia@microsoft.com>
Co-authored-by: Fernando Pérez-García <fepegar@gmail.com>
* Parameterize branch of argument to PIP
* Remove "notebooks" target from Makefile
* Pass branch name to demo notebook
Adapted from https://stackoverflow.com/a/58034787/3956024
* Echo branch name
* Run notebooks in new job
* Add steps to check out repo and install Python
* Introduce change not present in "main" to test CI
* Fix getting branch name
From https://stackoverflow.com/a/71158878/3956024
* Isolate parameterized cell
* Revert "Introduce change not present in "main" to test CI"
This reverts commit bee1504dfd.
Melissa Bristow
vale-salvatelli
Harshita Sharma
Kenza Bouzid
Fernando Pérez-García
Kenza Bouzid
maxilse
Anton Schwaighofer
SangamSwadik
Ozan Oktay
pre-commit-ci[bot]