Add training and evaluation code

code/RETFound/.gitignore

@@ -0,0 +1,134 @@
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[cod]
+*$py.class
+
+# C extensions
+*.so
+
+# Distribution / packaging
+.Python
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+pip-wheel-metadata/
+share/python-wheels/
+*.egg-info/
+.installed.cfg
+*.egg
+MANIFEST
+
+# PyInstaller
+#  Usually these files are written by a python script from a template
+#  before PyInstaller builds the exe, so as to inject date/other infos into it.
+*.manifest
+*.spec
+
+# Installer logs
+pip-log.txt
+pip-delete-this-directory.txt
+
+# Unit test / coverage reports
+htmlcov/
+.tox/
+.nox/
+.coverage
+.coverage.*
+.cache
+nosetests.xml
+coverage.xml
+*.cover
+*.py,cover
+.hypothesis/
+.pytest_cache/
+
+# Translations
+*.mo
+*.pot
+
+# Django stuff:
+*.log
+local_settings.py
+db.sqlite3
+db.sqlite3-journal
+
+# Flask stuff:
+instance/
+.webassets-cache
+
+# Scrapy stuff:
+.scrapy
+
+# Sphinx documentation
+docs/_build/
+
+# PyBuilder
+target/
+
+# Jupyter Notebook
+.ipynb_checkpoints
+
+# IPython
+profile_default/
+ipython_config.py
+
+# pyenv
+.python-version
+
+# pipenv
+#   According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control.
+#   However, in case of collaboration, if having platform-specific dependencies or dependencies
+#   having no cross-platform support, pipenv may install dependencies that don't work, or not
+#   install all needed dependencies.
+#Pipfile.lock
+
+# PEP 582; used by e.g. github.com/David-OConnor/pyflow
+__pypackages__/
+
+# Celery stuff
+celerybeat-schedule
+celerybeat.pid
+
+# SageMath parsed files
+*.sage.py
+
+# Environments
+.env
+.venv
+env/
+venv/
+ENV/
+env.bak/
+venv.bak/
+
+# Spyder project settings
+.spyderproject
+.spyproject
+
+# Rope project settings
+.ropeproject
+
+# VS Code project settings
+.vscode
+
+# mkdocs documentation
+/site
+
+# mypy
+.mypy_cache/
+.dmypy.json
+dmypy.json
+
+# Pyre type checker
+.pyre/
+
+IDRiD_data

code/RETFound/BENCHMARK.md

@@ -0,0 +1,45 @@
+## RETFound - Benckmark data split and model checkpoints
+
+
+We provide data split and model checkpoints to facilitate model comparison. Please remember to adjust the data_path, task, and nb_classes for model fine-tuning and evaluation.
+
+### Data split
+| Dataset | Download Link 1 |   Download Link 2 |
+| ------------- | ------------------ |------------------ |
+| APTOS2019 | [Google Drive](https://drive.google.com/file/d/162YPf4OhMVxj9TrQH0GnJv0n7z7gJWpj/view?usp=sharing)      |  [Baidu](https://pan.baidu.com/s/1uR8uUAnkO19lVT3beZuoMg) code:a2wg   |
+| MESSIDOR2 | [Google Drive](https://drive.google.com/file/d/1vOLBUK9xdzNV8eVkRjVdNrRwhPfaOmda/view?usp=sharing)        | [Baidu](https://pan.baidu.com/s/1uR8uUAnkO19lVT3beZuoMg) code:a2wg  |
+| IDRID | [Google Drive](https://drive.google.com/file/d/1c6zexA705z-ANEBNXJOBsk6uCvRnzmr3/view?usp=sharing)        |    [Baidu](https://pan.baidu.com/s/1uR8uUAnkO19lVT3beZuoMg) code:a2wg           |
+| PAPILA | [Google Drive](https://drive.google.com/file/d/1JltYs7WRWEU0yyki1CQw5-10HEbqCMBE/view?usp=sharing)      |  [Baidu](https://pan.baidu.com/s/1uR8uUAnkO19lVT3beZuoMg) code:a2wg   |
+| Glaucoma_fundus | [Google Drive](https://drive.google.com/file/d/18vSazOYDsUGdZ64gGkTg3E6jiNtcrUrI/view?usp=sharing)        | [Baidu](https://pan.baidu.com/s/1uR8uUAnkO19lVT3beZuoMg) code:a2wg |
+| JSIEC | [Google Drive](https://drive.google.com/file/d/1q0GFQb-dYwzIx8AwlaFZenUJItix4s8z/view?usp=sharing)        |       [Baidu](https://pan.baidu.com/s/1uR8uUAnkO19lVT3beZuoMg) code:a2wg       |
+| Retina | [Google Drive](https://drive.google.com/file/d/1vdmjMRDoUm9yk83HMArLiPcLDk_dm92Q/view?usp=sharing)      | [Baidu](https://pan.baidu.com/s/1uR8uUAnkO19lVT3beZuoMg) code:a2wg   |
+| OCTID | [Google Drive](https://drive.google.com/file/d/1I7nAvbkJG4UF29J3HcyIW53rVEFcKRgm/view?usp=sharing)        | [Baidu](https://pan.baidu.com/s/1uR8uUAnkO19lVT3beZuoMg) code:a2wg  |
+
+
+
+### Model checkpoints
+| Dataset | Download Link 1 |   Download Link 2 |
+| ------------- | ------------------ |------------------ |
+| APTOS2019 | [Google Drive](https://drive.google.com/drive/folders/16kL5V-1U7ACc-68PSHjAq6vyXRJvUoq3?usp=sharing)      |   [Baidu](https://pan.baidu.com/s/1Tr5Z7DMI8OTz7wpCnDj41g) code:ai05   |
+| MESSIDOR2 | [Google Drive](https://drive.google.com/drive/folders/1OTBRAHNbaytpwzwMHw9SWrltJouEEuxF?usp=sharing)        | [Baidu](https://pan.baidu.com/s/1Tr5Z7DMI8OTz7wpCnDj41g) code:ai05  |
+| IDRID | [Google Drive](https://drive.google.com/drive/folders/18Ml-B7nhejK4rnNG8upjqIARSlMP5kUc?usp=sharing)        |   [Baidu](https://pan.baidu.com/s/1Tr5Z7DMI8OTz7wpCnDj41g) code:ai05           |
+| PAPILA | [Google Drive](https://drive.google.com/drive/folders/1cHOX6C4NQVi9B6n-7Bxxg7b4-wdI4c73?usp=sharing)      |   [Baidu](https://pan.baidu.com/s/1Tr5Z7DMI8OTz7wpCnDj41g) code:ai05   |
+| Glaucoma_fundus | [Google Drive](https://drive.google.com/drive/folders/10JbanmVxjyX6mghXbxGnGVX1p9nwqsja?usp=sharing)        | [Baidu](https://pan.baidu.com/s/1Tr5Z7DMI8OTz7wpCnDj41g) code:ai05  |
+| JSIEC | [Google Drive](https://drive.google.com/drive/folders/1eosdBXsONUy49cwDO80AOzDHkHiPNJvv?usp=sharing)        |    [Baidu](https://pan.baidu.com/s/1Tr5Z7DMI8OTz7wpCnDj41g) code:ai05          |
+| Retina | [Google Drive](https://drive.google.com/drive/folders/1n7mXxN-ZUKauOrAlBAiF2E_36F6f0wZD?usp=sharing)      |  [Baidu](https://pan.baidu.com/s/1Tr5Z7DMI8OTz7wpCnDj41g) code:ai05   |
+| OCTID | [Google Drive](https://drive.google.com/drive/folders/14SQdLuIxfkiqz_zmpvNkd9Ka4NTW3Fml?usp=sharing)        | [Baidu](https://pan.baidu.com/s/1Tr5Z7DMI8OTz7wpCnDj41g) code:ai05  |
+
+
+
+### Official websites
+
+- APTOS2019: https://www.kaggle.com/competitions/aptos2019-blindness-detection/data
+- MESSIDOR2: https://www.adcis.net/en/third-party/messidor2/
+- IDRID: https://ieee-dataport.org/open-access/indian-diabetic-retinopathy-image-dataset-idrid
+- PAPILA: https://figshare.com/articles/dataset/PAPILA/14798004/1
+- Glaucoma_fundus: https://dataverse.harvard.edu/dataset.xhtml?persistentId=doi:10.7910/DVN/1YRRAC
+- JSIEC: https://zenodo.org/records/3477553
+- Retina: https://www.kaggle.com/datasets/jr2ngb/cataractdataset
+- OCTID: https://borealisdata.ca/dataverse/OCTID
+
+

code/RETFound/LICENSE

@@ -0,0 +1,400 @@
+
+Attribution-NonCommercial 4.0 International
+
+=======================================================================
+
+Creative Commons Corporation ("Creative Commons") is not a law firm and
+does not provide legal services or legal advice. Distribution of
+Creative Commons public licenses does not create a lawyer-client or
+other relationship. Creative Commons makes its licenses and related
+information available on an "as-is" basis. Creative Commons gives no
+warranties regarding its licenses, any material licensed under their
+terms and conditions, or any related information. Creative Commons
+disclaims all liability for damages resulting from their use to the
+fullest extent possible.
+
+Using Creative Commons Public Licenses
+
+Creative Commons public licenses provide a standard set of terms and
+conditions that creators and other rights holders may use to share
+original works of authorship and other material subject to copyright
+and certain other rights specified in the public license below. The
+following considerations are for informational purposes only, are not
+exhaustive, and do not form part of our licenses.
+
+     Considerations for licensors: Our public licenses are
+     intended for use by those authorized to give the public
+     permission to use material in ways otherwise restricted by
+     copyright and certain other rights. Our licenses are
+     irrevocable. Licensors should read and understand the terms
+     and conditions of the license they choose before applying it.
+     Licensors should also secure all rights necessary before
+     applying our licenses so that the public can reuse the
+     material as expected. Licensors should clearly mark any
+     material not subject to the license. This includes other CC-
+     licensed material, or material used under an exception or
+     limitation to copyright. More considerations for licensors:
+   wiki.creativecommons.org/Considerations_for_licensors
+
+     Considerations for the public: By using one of our public
+     licenses, a licensor grants the public permission to use the
+     licensed material under specified terms and conditions. If
+     the licensor's permission is not necessary for any reason--for
+     example, because of any applicable exception or limitation to
+     copyright--then that use is not regulated by the license. Our
+     licenses grant only permissions under copyright and certain
+     other rights that a licensor has authority to grant. Use of
+     the licensed material may still be restricted for other
+     reasons, including because others have copyright or other
+     rights in the material. A licensor may make special requests,
+     such as asking that all changes be marked or described.
+     Although not required by our licenses, you are encouraged to
+     respect those requests where reasonable. More_considerations
+     for the public: 
+   wiki.creativecommons.org/Considerations_for_licensees
+
+=======================================================================
+
+Creative Commons Attribution-NonCommercial 4.0 International Public
+License
+
+By exercising the Licensed Rights (defined below), You accept and agree
+to be bound by the terms and conditions of this Creative Commons
+Attribution-NonCommercial 4.0 International Public License ("Public
+License"). To the extent this Public License may be interpreted as a
+contract, You are granted the Licensed Rights in consideration of Your
+acceptance of these terms and conditions, and the Licensor grants You
+such rights in consideration of benefits the Licensor receives from
+making the Licensed Material available under these terms and
+conditions.
+
+Section 1 -- Definitions.
+
+  a. Adapted Material means material subject to Copyright and Similar
+     Rights that is derived from or based upon the Licensed Material
+     and in which the Licensed Material is translated, altered,
+     arranged, transformed, or otherwise modified in a manner requiring
+     permission under the Copyright and Similar Rights held by the
+     Licensor. For purposes of this Public License, where the Licensed
+     Material is a musical work, performance, or sound recording,
+     Adapted Material is always produced where the Licensed Material is
+     synched in timed relation with a moving image.
+
+  b. Adapter's License means the license You apply to Your Copyright
+     and Similar Rights in Your contributions to Adapted Material in
+     accordance with the terms and conditions of this Public License.
+
+  c. Copyright and Similar Rights means copyright and/or similar rights
+     closely related to copyright including, without limitation,
+     performance, broadcast, sound recording, and Sui Generis Database
+     Rights, without regard to how the rights are labeled or
+     categorized. For purposes of this Public License, the rights
+     specified in Section 2(b)(1)-(2) are not Copyright and Similar
+     Rights.
+  d. Effective Technological Measures means those measures that, in the
+     absence of proper authority, may not be circumvented under laws
+     fulfilling obligations under Article 11 of the WIPO Copyright
+     Treaty adopted on December 20, 1996, and/or similar international
+     agreements.
+
+  e. Exceptions and Limitations means fair use, fair dealing, and/or
+     any other exception or limitation to Copyright and Similar Rights
+     that applies to Your use of the Licensed Material.
+
+  f. Licensed Material means the artistic or literary work, database,
+     or other material to which the Licensor applied this Public
+     License.
+
+  g. Licensed Rights means the rights granted to You subject to the
+     terms and conditions of this Public License, which are limited to
+     all Copyright and Similar Rights that apply to Your use of the
+     Licensed Material and that the Licensor has authority to license.
+
+  h. Licensor means the individual(s) or entity(ies) granting rights
+     under this Public License.
+
+  i. NonCommercial means not primarily intended for or directed towards
+     commercial advantage or monetary compensation. For purposes of
+     this Public License, the exchange of the Licensed Material for
+     other material subject to Copyright and Similar Rights by digital
+     file-sharing or similar means is NonCommercial provided there is
+     no payment of monetary compensation in connection with the
+     exchange.
+
+  j. Share means to provide material to the public by any means or
+     process that requires permission under the Licensed Rights, such
+     as reproduction, public display, public performance, distribution,
+     dissemination, communication, or importation, and to make material
+     available to the public including in ways that members of the
+     public may access the material from a place and at a time
+     individually chosen by them.
+
+  k. Sui Generis Database Rights means rights other than copyright
+     resulting from Directive 96/9/EC of the European Parliament and of
+     the Council of 11 March 1996 on the legal protection of databases,
+     as amended and/or succeeded, as well as other essentially
+     equivalent rights anywhere in the world.
+
+  l. You means the individual or entity exercising the Licensed Rights
+     under this Public License. Your has a corresponding meaning.
+
+Section 2 -- Scope.
+
+  a. License grant.
+
+       1. Subject to the terms and conditions of this Public License,
+          the Licensor hereby grants You a worldwide, royalty-free,
+          non-sublicensable, non-exclusive, irrevocable license to
+          exercise the Licensed Rights in the Licensed Material to:
+
+            a. reproduce and Share the Licensed Material, in whole or
+               in part, for NonCommercial purposes only; and
+
+            b. produce, reproduce, and Share Adapted Material for
+               NonCommercial purposes only.
+
+       2. Exceptions and Limitations. For the avoidance of doubt, where
+          Exceptions and Limitations apply to Your use, this Public
+          License does not apply, and You do not need to comply with
+          its terms and conditions.
+
+       3. Term. The term of this Public License is specified in Section
+          6(a).
+
+       4. Media and formats; technical modifications allowed. The
+          Licensor authorizes You to exercise the Licensed Rights in
+          all media and formats whether now known or hereafter created,
+          and to make technical modifications necessary to do so. The
+          Licensor waives and/or agrees not to assert any right or
+          authority to forbid You from making technical modifications
+          necessary to exercise the Licensed Rights, including
+          technical modifications necessary to circumvent Effective
+          Technological Measures. For purposes of this Public License,
+          simply making modifications authorized by this Section 2(a)
+          (4) never produces Adapted Material.
+
+       5. Downstream recipients.
+
+            a. Offer from the Licensor -- Licensed Material. Every
+               recipient of the Licensed Material automatically
+               receives an offer from the Licensor to exercise the
+               Licensed Rights under the terms and conditions of this
+               Public License.
+
+            b. No downstream restrictions. You may not offer or impose
+               any additional or different terms or conditions on, or
+               apply any Effective Technological Measures to, the
+               Licensed Material if doing so restricts exercise of the
+               Licensed Rights by any recipient of the Licensed
+               Material.
+
+       6. No endorsement. Nothing in this Public License constitutes or
+          may be construed as permission to assert or imply that You
+          are, or that Your use of the Licensed Material is, connected
+          with, or sponsored, endorsed, or granted official status by,
+          the Licensor or others designated to receive attribution as
+          provided in Section 3(a)(1)(A)(i).
+
+  b. Other rights.
+
+       1. Moral rights, such as the right of integrity, are not
+          licensed under this Public License, nor are publicity,
+          privacy, and/or other similar personality rights; however, to
+          the extent possible, the Licensor waives and/or agrees not to
+          assert any such rights held by the Licensor to the limited
+          extent necessary to allow You to exercise the Licensed
+          Rights, but not otherwise.
+
+       2. Patent and trademark rights are not licensed under this
+          Public License.
+
+       3. To the extent possible, the Licensor waives any right to
+          collect royalties from You for the exercise of the Licensed
+          Rights, whether directly or through a collecting society
+          under any voluntary or waivable statutory or compulsory
+          licensing scheme. In all other cases the Licensor expressly
+          reserves any right to collect such royalties, including when
+          the Licensed Material is used other than for NonCommercial
+          purposes.
+
+Section 3 -- License Conditions.
+
+Your exercise of the Licensed Rights is expressly made subject to the
+following conditions.
+
+  a. Attribution.
+
+       1. If You Share the Licensed Material (including in modified
+          form), You must:
+
+            a. retain the following if it is supplied by the Licensor
+               with the Licensed Material:
+
+                 i. identification of the creator(s) of the Licensed
+                    Material and any others designated to receive
+                    attribution, in any reasonable manner requested by
+                    the Licensor (including by pseudonym if
+                    designated);
+
+                ii. a copyright notice;
+
+               iii. a notice that refers to this Public License;
+
+                iv. a notice that refers to the disclaimer of
+                    warranties;
+
+                 v. a URI or hyperlink to the Licensed Material to the
+                    extent reasonably practicable;
+
+            b. indicate if You modified the Licensed Material and
+               retain an indication of any previous modifications; and
+
+            c. indicate the Licensed Material is licensed under this
+               Public License, and include the text of, or the URI or
+               hyperlink to, this Public License.
+
+       2. You may satisfy the conditions in Section 3(a)(1) in any
+          reasonable manner based on the medium, means, and context in
+          which You Share the Licensed Material. For example, it may be
+          reasonable to satisfy the conditions by providing a URI or
+          hyperlink to a resource that includes the required
+          information.
+
+       3. If requested by the Licensor, You must remove any of the
+          information required by Section 3(a)(1)(A) to the extent
+          reasonably practicable.
+
+       4. If You Share Adapted Material You produce, the Adapter's
+          License You apply must not prevent recipients of the Adapted
+          Material from complying with this Public License.
+
+Section 4 -- Sui Generis Database Rights.
+
+Where the Licensed Rights include Sui Generis Database Rights that
+apply to Your use of the Licensed Material:
+
+  a. for the avoidance of doubt, Section 2(a)(1) grants You the right
+     to extract, reuse, reproduce, and Share all or a substantial
+     portion of the contents of the database for NonCommercial purposes
+     only;
+
+  b. if You include all or a substantial portion of the database
+     contents in a database in which You have Sui Generis Database
+     Rights, then the database in which You have Sui Generis Database
+     Rights (but not its individual contents) is Adapted Material; and
+
+  c. You must comply with the conditions in Section 3(a) if You Share
+     all or a substantial portion of the contents of the database.
+
+For the avoidance of doubt, this Section 4 supplements and does not
+replace Your obligations under this Public License where the Licensed
+Rights include other Copyright and Similar Rights.
+
+Section 5 -- Disclaimer of Warranties and Limitation of Liability.
+
+  a. UNLESS OTHERWISE SEPARATELY UNDERTAKEN BY THE LICENSOR, TO THE
+     EXTENT POSSIBLE, THE LICENSOR OFFERS THE LICENSED MATERIAL AS-IS
+     AND AS-AVAILABLE, AND MAKES NO REPRESENTATIONS OR WARRANTIES OF
+     ANY KIND CONCERNING THE LICENSED MATERIAL, WHETHER EXPRESS,
+     IMPLIED, STATUTORY, OR OTHER. THIS INCLUDES, WITHOUT LIMITATION,
+     WARRANTIES OF TITLE, MERCHANTABILITY, FITNESS FOR A PARTICULAR
+     PURPOSE, NON-INFRINGEMENT, ABSENCE OF LATENT OR OTHER DEFECTS,
+     ACCURACY, OR THE PRESENCE OR ABSENCE OF ERRORS, WHETHER OR NOT
+     KNOWN OR DISCOVERABLE. WHERE DISCLAIMERS OF WARRANTIES ARE NOT
+     ALLOWED IN FULL OR IN PART, THIS DISCLAIMER MAY NOT APPLY TO YOU.
+
+  b. TO THE EXTENT POSSIBLE, IN NO EVENT WILL THE LICENSOR BE LIABLE
+     TO YOU ON ANY LEGAL THEORY (INCLUDING, WITHOUT LIMITATION,
+     NEGLIGENCE) OR OTHERWISE FOR ANY DIRECT, SPECIAL, INDIRECT,
+     INCIDENTAL, CONSEQUENTIAL, PUNITIVE, EXEMPLARY, OR OTHER LOSSES,
+     COSTS, EXPENSES, OR DAMAGES ARISING OUT OF THIS PUBLIC LICENSE OR
+     USE OF THE LICENSED MATERIAL, EVEN IF THE LICENSOR HAS BEEN
+     ADVISED OF THE POSSIBILITY OF SUCH LOSSES, COSTS, EXPENSES, OR
+     DAMAGES. WHERE A LIMITATION OF LIABILITY IS NOT ALLOWED IN FULL OR
+     IN PART, THIS LIMITATION MAY NOT APPLY TO YOU.
+
+  c. The disclaimer of warranties and limitation of liability provided
+     above shall be interpreted in a manner that, to the extent
+     possible, most closely approximates an absolute disclaimer and
+     waiver of all liability.
+
+Section 6 -- Term and Termination.
+
+  a. This Public License applies for the term of the Copyright and
+     Similar Rights licensed here. However, if You fail to comply with
+     this Public License, then Your rights under this Public License
+     terminate automatically.
+
+  b. Where Your right to use the Licensed Material has terminated under
+     Section 6(a), it reinstates:
+
+       1. automatically as of the date the violation is cured, provided
+          it is cured within 30 days of Your discovery of the
+          violation; or
+
+       2. upon express reinstatement by the Licensor.
+
+     For the avoidance of doubt, this Section 6(b) does not affect any
+     right the Licensor may have to seek remedies for Your violations
+     of this Public License.
+
+  c. For the avoidance of doubt, the Licensor may also offer the
+     Licensed Material under separate terms or conditions or stop
+     distributing the Licensed Material at any time; however, doing so
+     will not terminate this Public License.
+
+  d. Sections 1, 5, 6, 7, and 8 survive termination of this Public
+     License.
+
+Section 7 -- Other Terms and Conditions.
+
+  a. The Licensor shall not be bound by any additional or different
+     terms or conditions communicated by You unless expressly agreed.
+
+  b. Any arrangements, understandings, or agreements regarding the
+     Licensed Material not stated herein are separate from and
+     independent of the terms and conditions of this Public License.
+
+Section 8 -- Interpretation.
+
+  a. For the avoidance of doubt, this Public License does not, and
+     shall not be interpreted to, reduce, limit, restrict, or impose
+     conditions on any use of the Licensed Material that could lawfully
+     be made without permission under this Public License.
+
+  b. To the extent possible, if any provision of this Public License is
+     deemed unenforceable, it shall be automatically reformed to the
+     minimum extent necessary to make it enforceable. If the provision
+     cannot be reformed, it shall be severed from this Public License
+     without affecting the enforceability of the remaining terms and
+     conditions.
+
+  c. No term or condition of this Public License will be waived and no
+     failure to comply consented to unless expressly agreed to by the
+     Licensor.
+
+  d. Nothing in this Public License constitutes or may be interpreted
+     as a limitation upon, or waiver of, any privileges and immunities
+     that apply to the Licensor or You, including from the legal
+     processes of any jurisdiction or authority.
+
+=======================================================================
+
+Creative Commons is not a party to its public
+licenses. Notwithstanding, Creative Commons may elect to apply one of
+its public licenses to material it publishes and in those instances
+will be considered the “Licensor.” The text of the Creative Commons
+public licenses is dedicated to the public domain under the CC0 Public
+Domain Dedication. Except for the limited purpose of indicating that
+material is shared under a Creative Commons public license or as
+otherwise permitted by the Creative Commons policies published at
+creativecommons.org/policies, Creative Commons does not authorize the
+use of the trademark "Creative Commons" or any other trademark or logo
+of Creative Commons without its prior written consent including,
+without limitation, in connection with any unauthorized modifications
+to any of its public licenses or any other arrangements,
+understandings, or agreements concerning use of licensed material. For
+the avoidance of doubt, this paragraph does not form part of the
+public licenses.
+
+Creative Commons may be contacted at creativecommons.org.

code/RETFound/README.md

@@ -0,0 +1,272 @@
+## RETFound - A foundation model for retinal images
+
+
+Official repo including a series of foundation models and applications for retinal images.<br>
+`[RETFound-MAE]`:[RETFound: a foundation model for generalizable disease detection from retinal images](https://www.nature.com/articles/s41586-023-06555-x).<br>
+`[RETFound-DINOv2]`:[Revealing the Impact of Pre-training Data on Medical Foundation Models](https://www.researchsquare.com/article/rs-6080254/v1).<br>
+`[DINOv2]`:[General-purpose vision foundation models DINOv2 by Meta](https://github.com/facebookresearch/dinov2).<br>
+`[DINOv3]`:[General-purpose vision foundation models DINOv3 by Meta](https://github.com/facebookresearch/dinov3).<br>
+
+
+Please contact 	**ykzhoua@gmail.com** or **yukun.zhou.19@ucl.ac.uk** if you have questions.
+
+
+### 📝Key features
+
+- RETFound is pre-trained on 1.6 million retinal images with self-supervised learning
+- RETFound has been validated in multiple disease detection tasks
+- RETFound can be efficiently adapted to customised tasks
+
+
+### 🎉News
+
+- 🐉2025/09: **Preprint benchmarking DINOv3, DINOv2, and RETFound is [available](https://arxiv.org/abs/2509.03421)!**
+- 🐉2025/09: **We included state-of-the-art DINOv3 into fine-tuning pipeline for retinal applications!**
+- 🐉2025/02: **We organised the model weights on HuggingFace, no more manual downloads needed!**
+- 🐉2025/02: **Multiple [pre-trained weights](https://huggingface.co/YukunZhou), including MAE-based and DINOV2-based, are added!**
+- 🐉2025/02: **We update the version of packages, such as CUDA12+ and PyTorch 2.3+!**
+- 🐉2024/01: [Feature vector notebook](https://github.com/rmaphoh/RETFound_MAE/blob/main/latent_feature.ipynb) are now online!
+- 🐉2024/01: [Data split and model checkpoints](BENCHMARK.md) for public datasets are now online!
+- 🎄2023/12: [Colab notebook](https://colab.research.google.com/drive/1_X19zdMegmAlqPAEY0Ao659fzzzlx2IZ?usp=sharing) is now online - free GPU & simple operation!
+
+
+### 🔧Install environment
+
+1. Create environment with conda:
+
+```
+conda create -n retfound python=3.11.0 -y
+conda activate retfound
+```
+
+2. Install dependencies
+
+```
+pip install torch==2.5.1 torchvision==0.20.1 --index-url https://download.pytorch.org/whl/cu121
+git clone https://github.com/rmaphoh/RETFound/
+cd RETFound
+pip install -r requirements.txt
+pip install ipykernel
+python -m ipykernel install --user --name retfound --display-name "Python (retfound)"
+```
+
+
+### 🌱Fine-tuning with RETFound weights
+
+1. Get access to the pre-trained models on HuggingFace (register an account and fill in the form) and go to step 2:
+<table><tbody>
+<!-- START TABLE -->
+<!-- TABLE HEADER -->
+<th valign="bottom"></th>
+<th valign="bottom">ViT-Large</th>
+<th valign="bottom">Source</th>
+<!-- TABLE BODY -->
+<tr><td align="left">RETFound_mae_natureCFP</td>
+<td align="center"><a href="https://huggingface.co/YukunZhou/RETFound_mae_natureCFP">access</a></td>
+<td align="center"><a href="https://www.nature.com/articles/s41586-023-06555-x">Nature RETFound paper</a></td>
+</tr>
+<!-- TABLE BODY -->
+<tr><td align="left">RETFound_mae_natureOCT</td>
+<td align="center"><a href="https://huggingface.co/YukunZhou/RETFound_mae_natureOCT">access</a></td>
+<td align="center"><a href="https://www.nature.com/articles/s41586-023-06555-x">Nature RETFound paper</a></td>
+</tr>
+<!-- TABLE BODY -->
+<tr><td align="left">RETFound_mae_meh</td>
+<td align="center"><a href="https://huggingface.co/YukunZhou/RETFound_mae_meh">access</a></td>
+<td align="center"><a href="https://www.researchsquare.com/article/rs-6080254/v1">FM data paper</a></td>
+</tr>
+<!-- TABLE BODY -->
+<tr><td align="left">RETFound_mae_shanghai</td>
+<td align="center"><a href="https://huggingface.co/YukunZhou/RETFound_mae_shanghai">access</a></td>
+<td align="center"><a href="https://www.researchsquare.com/article/rs-6080254/v1">FM data paper</a></td>
+</tr>
+<!-- TABLE BODY -->
+<tr><td align="left">RETFound_dinov2_meh</td>
+<td align="center"><a href="https://huggingface.co/YukunZhou/RETFound_dinov2_meh">access</a></td>
+<td align="center"><a href="https://www.researchsquare.com/article/rs-6080254/v1">FM data paper</a></td>
+</tr>
+<!-- TABLE BODY -->
+<tr><td align="left">RETFound_dinov2_shanghai</td>
+<td align="center"><a href="https://huggingface.co/YukunZhou/RETFound_dinov2_shanghai">access</a></td>
+<td align="center"><a href="https://www.researchsquare.com/article/rs-6080254/v1">FM data paper</a></td>
+</tr>
+</tbody></table>
+
+2. Login in your HuggingFace account, where HuggingFace token can be [created and copied](https://huggingface.co/settings/tokens).
+```
+huggingface-cli login --token YOUR_HUGGINGFACE_TOKEN
+```
+
+**Optional**: if your machine and server cannot access HuggingFace due to internet wall, run the command below (Do not run it if you can access):
+```
+export HF_ENDPOINT=https://hf-mirror.com
+```
+
+3. If you would like to fine-tune [DINOv2](https://github.com/facebookresearch/dinov2) and [DINOv3](https://github.com/facebookresearch/dinov3), please visit their GitHub repositories to download the model weights and put them in the RETFound folder.
+
+4. Organise your data into this directory structure (Public datasets used in this study can be [downloaded here](BENCHMARK.md))
+
+```
+├── data folder
+    ├──train
+        ├──class_a
+        ├──class_b
+        ├──class_c
+    ├──val
+        ├──class_a
+        ├──class_b
+        ├──class_c
+    ├──test
+        ├──class_a
+        ├──class_b
+        ├──class_c
+``` 
+
+
+
+5. Start fine-tuning by running `sh train.sh`.
+
+
+In `train.sh`, the model can be selected by changing the hyperparameters `MODEL`, `MODEL_ARCH`, `FINETUNE`:
+
+**RETFound**:
+
+| MODEL           | MODEL_ARCH               | FINETUNE                 | SIZE                     |
+|-----------------|--------------------------|--------------------------|--------------------------|
+| RETFound_mae    | retfound_mae             | RETFound_mae_natureCFP   | ~300M                    |
+| RETFound_mae    | retfound_mae             | RETFound_mae_natureOCT   | ~300M                    |
+| RETFound_mae    | retfound_mae             | RETFound_mae_meh         | ~300M                    |
+| RETFound_mae    | retfound_mae             | RETFound_mae_shanghai    | ~300M                    |
+| RETFound_dinov2 | retfound_dinov2          | RETFound_dinov2_meh      | ~300M                    |
+| RETFound_dinov2 | retfound_dinov2          | RETFound_dinov2_shanghai | ~300M                    |
+
+
+**DINOv3**:
+
+| MODEL           | MODEL_ARCH               | FINETUNE                         | SIZE                     |
+|-----------------|--------------------------|----------------------------------|--------------------------|
+| Dinov3          | dinov3_vits16            | dinov3_vits16_pretrain.pth       | ~21M                     |
+| Dinov3          | dinov3_vits16plus        | dinov3_vits16plus_pretrain.pth   | ~29M                     |
+| Dinov3          | dinov3_vitb16            | dinov3_vitb16_pretrain.pth       | ~86M                     |
+| Dinov3          | dinov3_vitl16            | dinov3_vitl16_pretrain.pth       | ~300M                    |
+| Dinov3          | dinov3_vith16plus        | dinov3_vith16plus_pretrain.pth   | ~840M                    |
+| Dinov3          | dinov3_vit7b16           | dinov3_vit7b16_pretrain.pth      | ~6.7B                    |
+
+
+**DINOv2**:
+
+| MODEL           | MODEL_ARCH               | FINETUNE                     | SIZE                     |
+|-----------------|--------------------------|------------------------------|--------------------------|
+| Dinov2          | dinov2_vits14            | dinov2_vits14_pretrain.pth   | ~21M                     |
+| Dinov2          | dinov2_vitb14            | dinov2_vitb14_pretrain.pth   | ~86M                     |
+| Dinov2          | dinov2_vitl14            | dinov2_vitl14_pretrain.pth   | ~300M                    |
+| Dinov2          | dinov2_vitg14            | dinov2_vitg14_pretrain.pth   | ~1.1B                    |
+
+
+Change the DATA_PATH to your dataset directory.
+
+```
+# ==== Model settings ====
+# adaptation {finetune,lp}
+ADAPTATION="finetune"
+MODEL="RETFound_dinov2"
+MODEL_ARCH="retfound_dinov2"
+FINETUNE="RETFound_dinov2_meh"
+
+# ==== Data settings ====
+# change the dataset name and corresponding class number
+DATASET="MESSIDOR2"
+NUM_CLASS=5
+
+# =======================
+DATA_PATH="PATH TO THE DATASET"
+TASK="${MODEL_ARCH}_${DATASET}_${ADAPTATION}"
+
+torchrun --nproc_per_node=1 --master_port=48766 main_finetune.py \
+  --model "${MODEL}" \
+  --model_arch "${MODEL_ARCH}" \
+  --finetune "${FINETUNE}" \
+  --savemodel \
+  --global_pool \
+  --batch_size 24 \
+  --world_size 1 \
+  --epochs 50 \
+  --nb_classes "${NUM_CLASS}" \
+  --data_path "${DATA_PATH}" \
+  --input_size 224 \
+  --task "${TASK}" \
+  --adaptation "${ADAPTATION}" 
+
+```
+
+
+
+6. For evaluation only (download data and model checkpoints [here](BENCHMARK.md); change the DATA_PATH below)
+
+
+```
+# ==== Model/settings (match training) ====
+ADAPTATION="finetune"
+MODEL="RETFound_dinov2"
+MODEL_ARCH="retfound_dinov2"
+FINETUNE="RETFound_dinov2_meh"
+
+# ==== Data/settings (match training) ====
+DATASET="MESSIDOR2"
+NUM_CLASS=5
+
+# =======================
+DATA_PATH="PATH TO THE DATASET"
+TASK="${MODEL_ARCH}_${DATASET}_${ADAPTATION}"
+
+# Path to the trained checkpoint (adjust if you saved elsewhere)
+CKPT="./output_dir/${TASK}/checkpoint-best.pth"
+
+# ==== Evaluation only ====
+torchrun --nproc_per_node=1 --master_port=48766 main_finetune.py \
+  --model "${MODEL}" \
+  --model_arch "${MODEL_ARCH}" \
+  --savemodel \
+  --global_pool \
+  --batch_size 128 \
+  --world_size 1 \
+  --nb_classes "${NUM_CLASS}" \
+  --data_path "${DATA_PATH}" \
+  --input_size 224 \
+  --task "${TASK}" \
+  --adaptation "${ADAPTATION}" \
+  --eval \
+  --resume "${CKPT}"
+
+```
+
+
+### 📃Citation
+
+If you find this repository useful, please consider citing this paper:
+
+
+```
+@article{zhou2023foundation,
+  title={A foundation model for generalizable disease detection from retinal images},
+  author={Zhou, Yukun and Chia, Mark A and Wagner, Siegfried K and Ayhan, Murat S and Williamson, Dominic J and Struyven, Robbert R and Liu, Timing and Xu, Moucheng and Lozano, Mateo G and Woodward-Court, Peter and others},
+  journal={Nature},
+  volume={622},
+  number={7981},
+  pages={156--163},
+  year={2023},
+  publisher={Nature Publishing Group UK London}
+}
+```
+
+```
+@misc{zhou2025generalistversusspecialistvision,
+      title={Generalist versus Specialist Vision Foundation Models for Ocular Disease and Oculomics}, 
+      author={Yukun Zhou and Paul Nderitu and Jocelyn Hui Lin Goh and Justin Engelmann and Siegfried K. Wagner and Anran Ran and Hongyang Jiang and Lie Ju and Ke Zou and Sahana Srinivasan and Hyunmin Kim and Takahiro Ninomiya and Zheyuan Wang and Gabriel Dawei Yang and Eden Ruffell and Dominic Williamson and Rui Santos and Gabor Mark Somfai and Carol Y. Cheung and Tien Yin Wong and Daniel C. Alexander and Yih Chung Tham and Pearse A. Keane},
+      year={2025},
+      eprint={2509.03421},
+      archivePrefix={arXiv},
+      primaryClass={eess.IV},
+      url={https://arxiv.org/abs/2509.03421}, 
+}
+```

code/RETFound/RETFound_mae_natureCFP/.gitattributes

@@ -0,0 +1,35 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tar filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.zst filter=lfs diff=lfs merge=lfs -text
+*tfevents* filter=lfs diff=lfs merge=lfs -text

code/RETFound/RETFound_mae_natureCFP/README.md

@@ -0,0 +1,83 @@
+---
+tags:
+- pytorch
+extra_gated_fields:
+  First Name: text
+  Last Name: text
+  Affiliation: text
+  Job title:
+    type: select
+    options:
+    - Student
+    - Research Graduate
+    - AI researcher
+    - AI developer/engineer
+    - Other
+  geo: ip_location
+extra_gated_button_content: Submit
+---
+# Model Card for RETFound_MAE_MEH
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+This modelcard aims to provide a pre-trained vision foundation model [RETFound](https://github.com/rmaphoh/RETFound_MAE), pre-trained with Masked Autoencoder.
+
+This is the official weight for [RETFound Nature paper](https://www.nature.com/articles/s41586-023-06555-x)
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+- **Developed by:** Yukun Zhou
+- **Model type:** Pre-trained model
+- **License:** Creative Commons Attribution-NonCommercial 4.0 International Public License (CC BY-NC 4.0)
+
+### Model Sources
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [RETFound](https://github.com/rmaphoh/RETFound_MAE)
+- **Paper:** [Nature paper](https://www.nature.com/articles/s41586-023-06555-x)
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+This repo contains the model weight. After granted the access, please fill the token in the [code](https://github.com/rmaphoh/RETFound_MAE).
+
+The code will automatically download the model and run the training. 
+
+
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+
+- **Hardware Type:** 4 * NVIDIA A100 80GB
+- **Hours used:** 14 days
+- **Cloud Provider:** UCL CS Cluster & Shanghai Jiaotong University Cluster
+
+
+## Citation
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+```
+@article{zhou2023foundation,
+  title={A foundation model for generalizable disease detection from retinal images},
+  author={Zhou, Yukun and Chia, Mark A and Wagner, Siegfried K and Ayhan, Murat S and Williamson, Dominic J and Struyven, Robbert R and Liu, Timing and Xu, Moucheng and Lozano, Mateo G and Woodward-Court, Peter and others},
+  journal={Nature},
+  volume={622},
+  number={7981},
+  pages={156--163},
+  year={2023},
+  publisher={Nature Publishing Group UK London}
+}
+```
+
+## Model Card Contact
+
+**ykzhoua@gmail.com** or **yukun.zhou.19@ucl.ac.uk**

code/RETFound/RETFound_mae_natureCFP/config.json

@@ -0,0 +1,15 @@
+{
+  "architectures": ["ViT-Large"],
+  "image modality": "CFP",
+  "model_type": "vit",  
+  "hidden_size": 1024,
+  "num_attention_heads": 16,
+  "num_hidden_layers": 24,
+  "image_size": 224,
+  "patch_size": 16,
+  "pretrained": true,
+  "dataset": "MEH MIDAS",
+  "fine_tuned": false,
+  "author": "Yukun Zhou",
+  "license": "CC BY-NC 4.0"
+}

code/RETFound/RETFound_mae_natureOCT/.gitattributes

@@ -0,0 +1,35 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tar filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.zst filter=lfs diff=lfs merge=lfs -text
+*tfevents* filter=lfs diff=lfs merge=lfs -text

code/RETFound/RETFound_mae_natureOCT/README.md

@@ -0,0 +1,83 @@
+---
+tags:
+- pytorch
+extra_gated_fields:
+  First Name: text
+  Last Name: text
+  Affiliation: text
+  Job title:
+    type: select
+    options:
+    - Student
+    - Research Graduate
+    - AI researcher
+    - AI developer/engineer
+    - Other
+  geo: ip_location
+extra_gated_button_content: Submit
+---
+# Model Card for RETFound_MAE_MEH
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+This modelcard aims to provide a pre-trained vision foundation model [RETFound](https://github.com/rmaphoh/RETFound_MAE), pre-trained with Masked Autoencoder.
+
+This is the official weight for [RETFound Nature paper](https://www.nature.com/articles/s41586-023-06555-x)
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+- **Developed by:** Yukun Zhou
+- **Model type:** Pre-trained model
+- **License:** Creative Commons Attribution-NonCommercial 4.0 International Public License (CC BY-NC 4.0)
+
+### Model Sources
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [RETFound](https://github.com/rmaphoh/RETFound_MAE)
+- **Paper:** [Nature paper](https://www.nature.com/articles/s41586-023-06555-x)
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+This repo contains the model weight. After granted the access, please fill the token in the [code](https://github.com/rmaphoh/RETFound_MAE).
+
+The code will automatically download the model and run the training. 
+
+
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+
+- **Hardware Type:** 4 * NVIDIA A100 80GB
+- **Hours used:** 14 days
+- **Cloud Provider:** UCL CS Cluster & Shanghai Jiaotong University Cluster
+
+
+## Citation
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+```
+@article{zhou2023foundation,
+  title={A foundation model for generalizable disease detection from retinal images},
+  author={Zhou, Yukun and Chia, Mark A and Wagner, Siegfried K and Ayhan, Murat S and Williamson, Dominic J and Struyven, Robbert R and Liu, Timing and Xu, Moucheng and Lozano, Mateo G and Woodward-Court, Peter and others},
+  journal={Nature},
+  volume={622},
+  number={7981},
+  pages={156--163},
+  year={2023},
+  publisher={Nature Publishing Group UK London}
+}
+```
+
+## Model Card Contact
+
+**ykzhoua@gmail.com** or **yukun.zhou.19@ucl.ac.uk**

code/RETFound/RETFound_mae_natureOCT/config.json

@@ -0,0 +1,15 @@
+{
+  "architectures": ["ViT-Large"],
+  "image modality": "OCT",
+  "model_type": "vit",  
+  "hidden_size": 1024,
+  "num_attention_heads": 16,
+  "num_hidden_layers": 24,
+  "image_size": 224,
+  "patch_size": 16,
+  "pretrained": true,
+  "dataset": "MEH MIDAS",
+  "fine_tuned": false,
+  "author": "Yukun Zhou",
+  "license": "CC BY-NC 4.0"
+}

code/RETFound/engine_finetune.py

@@ -0,0 +1,149 @@
+import os
+import csv
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+import numpy as np
+import matplotlib.pyplot as plt
+from typing import Iterable, Optional
+from timm.data import Mixup
+from timm.utils import accuracy
+from sklearn.metrics import (
+    accuracy_score, roc_auc_score, f1_score, average_precision_score,
+    hamming_loss, jaccard_score, recall_score, precision_score, cohen_kappa_score
+)
+from pycm import ConfusionMatrix
+import util.misc as misc
+import util.lr_sched as lr_sched
+
+def train_one_epoch(
+    model: torch.nn.Module,
+    criterion: torch.nn.Module,
+    data_loader: Iterable,
+    optimizer: torch.optim.Optimizer,
+    device: torch.device,
+    epoch: int,
+    loss_scaler,
+    max_norm: float = 0,
+    mixup_fn: Optional[Mixup] = None,
+    log_writer=None,
+    args=None
+):
+    """Train the model for one epoch."""
+    model.train(True)
+    metric_logger = misc.MetricLogger(delimiter="  ")
+    metric_logger.add_meter('lr', misc.SmoothedValue(window_size=1, fmt='{value:.6f}'))
+    print_freq, accum_iter = 20, args.accum_iter
+    optimizer.zero_grad()
+    
+    if log_writer:
+        print(f'log_dir: {log_writer.log_dir}')
+    
+    for data_iter_step, (samples, targets) in enumerate(metric_logger.log_every(data_loader, print_freq, f'Epoch: [{epoch}]')):
+        if data_iter_step % accum_iter == 0:
+            lr_sched.adjust_learning_rate(optimizer, data_iter_step / len(data_loader) + epoch, args)
+        
+        samples, targets = samples.to(device, non_blocking=True), targets.to(device, non_blocking=True)
+        if mixup_fn:
+            samples, targets = mixup_fn(samples, targets)
+        
+        with torch.cuda.amp.autocast():
+            outputs = model(samples)
+            loss = criterion(outputs, targets)
+        loss_value = loss.item()
+        loss /= accum_iter
+        
+        loss_scaler(loss, optimizer, clip_grad=max_norm, parameters=model.parameters(), create_graph=False,
+                    update_grad=(data_iter_step + 1) % accum_iter == 0)
+        if (data_iter_step + 1) % accum_iter == 0:
+            optimizer.zero_grad()
+        
+        torch.cuda.synchronize()
+        metric_logger.update(loss=loss_value)
+        min_lr = 10.
+        max_lr = 0.
+        for group in optimizer.param_groups:
+            min_lr = min(min_lr, group["lr"])
+            max_lr = max(max_lr, group["lr"])
+
+        metric_logger.update(lr=max_lr)
+
+        loss_value_reduce = misc.all_reduce_mean(loss_value)
+        if log_writer is not None and (data_iter_step + 1) % accum_iter == 0:
+            """ We use epoch_1000x as the x-axis in tensorboard.
+            This calibrates different curves when batch size changes.
+            """
+            epoch_1000x = int((data_iter_step / len(data_loader) + epoch) * 1000)
+            log_writer.add_scalar('loss/train', loss_value_reduce, epoch_1000x)
+            log_writer.add_scalar('lr', max_lr, epoch_1000x)
+    
+    metric_logger.synchronize_between_processes()
+    print("Averaged stats:", metric_logger)
+    return {k: meter.global_avg for k, meter in metric_logger.meters.items()}
+
+@torch.no_grad()
+def evaluate(data_loader, model, device, args, epoch, mode, num_class, log_writer):
+    """Evaluate the model."""
+    criterion = nn.CrossEntropyLoss()
+    metric_logger = misc.MetricLogger(delimiter="  ")
+    os.makedirs(os.path.join(args.output_dir, args.task), exist_ok=True)
+    
+    model.eval()
+    true_onehot, pred_onehot, true_labels, pred_labels, pred_softmax = [], [], [], [], []
+    
+    for batch in metric_logger.log_every(data_loader, 10, f'{mode}:'):
+        images, target = batch[0].to(device, non_blocking=True), batch[-1].to(device, non_blocking=True)
+        target_onehot = F.one_hot(target.to(torch.int64), num_classes=num_class)
+        
+        with torch.cuda.amp.autocast():
+            output = model(images)
+            loss = criterion(output, target)
+        output_ = nn.Softmax(dim=1)(output)
+        output_label = output_.argmax(dim=1)
+        output_onehot = F.one_hot(output_label.to(torch.int64), num_classes=num_class)
+        
+        metric_logger.update(loss=loss.item())
+        true_onehot.extend(target_onehot.cpu().numpy())
+        pred_onehot.extend(output_onehot.detach().cpu().numpy())
+        true_labels.extend(target.cpu().numpy())
+        pred_labels.extend(output_label.detach().cpu().numpy())
+        pred_softmax.extend(output_.detach().cpu().numpy())
+    
+    accuracy = accuracy_score(true_labels, pred_labels)
+    hamming = hamming_loss(true_onehot, pred_onehot)
+    jaccard = jaccard_score(true_onehot, pred_onehot, average='macro')
+    average_precision = average_precision_score(true_onehot, pred_softmax, average='macro')
+    kappa = cohen_kappa_score(true_labels, pred_labels)
+    f1 = f1_score(true_onehot, pred_onehot, zero_division=0, average='macro')
+    roc_auc = roc_auc_score(true_onehot, pred_softmax, multi_class='ovr', average='macro')
+    precision = precision_score(true_onehot, pred_onehot, zero_division=0, average='macro')
+    recall = recall_score(true_onehot, pred_onehot, zero_division=0, average='macro')
+    
+    score = (f1 + roc_auc + kappa) / 3
+    if log_writer:
+        for metric_name, value in zip(['accuracy', 'f1', 'roc_auc', 'hamming', 'jaccard', 'precision', 'recall', 'average_precision', 'kappa', 'score'],
+                                       [accuracy, f1, roc_auc, hamming, jaccard, precision, recall, average_precision, kappa, score]):
+            log_writer.add_scalar(f'perf/{metric_name}', value, epoch)
+    
+    print(f'val loss: {metric_logger.meters["loss"].global_avg}')
+    print(f'Accuracy: {accuracy:.4f}, F1 Score: {f1:.4f}, ROC AUC: {roc_auc:.4f}, Hamming Loss: {hamming:.4f},\n'
+          f' Jaccard Score: {jaccard:.4f}, Precision: {precision:.4f}, Recall: {recall:.4f},\n'
+          f' Average Precision: {average_precision:.4f}, Kappa: {kappa:.4f}, Score: {score:.4f}')
+    
+    metric_logger.synchronize_between_processes()
+    
+    results_path = os.path.join(args.output_dir, args.task, f'metrics_{mode}.csv')
+    file_exists = os.path.isfile(results_path)
+    with open(results_path, 'a', newline='', encoding='utf8') as cfa:
+        wf = csv.writer(cfa)
+        if not file_exists:
+            wf.writerow(['val_loss', 'accuracy', 'f1', 'roc_auc', 'hamming', 'jaccard', 'precision', 'recall', 'average_precision', 'kappa'])
+        wf.writerow([metric_logger.meters["loss"].global_avg, accuracy, f1, roc_auc, hamming, jaccard, precision, recall, average_precision, kappa])
+    
+    if mode == 'test':
+        cm = ConfusionMatrix(actual_vector=true_labels, predict_vector=pred_labels)
+        cm.plot(cmap=plt.cm.Blues, number_label=True, normalized=True, plot_lib="matplotlib")
+        plt.savefig(os.path.join(args.output_dir, args.task, 'confusion_matrix_test.jpg'), dpi=600, bbox_inches='tight')
+        np.savez(os.path.join(args.output_dir, args.task, 'test_pred.npz'), y_true=np.array(true_labels), y_prob=np.array(pred_softmax))
+    
+    return {k: meter.global_avg for k, meter in metric_logger.meters.items()}, score

code/RETFound/examples/RETFound_MESSIDOR2_demo.ipynb

@@ -0,0 +1,223 @@
+{
+ "cells": [
+  {
+   "cell_type": "markdown",
+   "id": "76b39fb1",
+   "metadata": {
+    "jp-MarkdownHeadingCollapsed": true
+   },
+   "source": [
+    "## Jupyter notebook example - Classification task\n",
+    "### Example using [MESSIDOR2](https://www.adcis.net/en/third-party/messidor2/) dataset\n",
+    "**Application**: Using RETFound for five-category diabetic retinopathy classification\n",
+    "\n",
+    "**Author**: Yukun Zhou\n",
+    "\n",
+    "**Date**: 30 Nov 2025\n",
+    "\n",
+    "**Performance**:\n",
+    "\n",
+    "<table align=\"left\">\n",
+    "<tr>\n",
+    "  <th>Accuracy</th>\n",
+    "  <th>Recall</th>\n",
+    "  <th>F1 Score</th>\n",
+    "  <th>ROC AUC</th>\n",
+    "  <th>PR AUC</th>\n",
+    "</tr>\n",
+    "<tr>\n",
+    "  <td>0.7091</td>\n",
+    "  <td>0.5616</td>\n",
+    "  <td>0.6078</td>\n",
+    "  <td>0.9037</td>\n",
+    "  <td>0.6863</td>\n",
+    "</tr>\n",
+    "</table>\n"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "7ec435a7",
+   "metadata": {},
+   "source": [
+    "## 1. Install environment\n",
+    "1. Follow [RETFound README](https://github.com/rmaphoh/RETFound) to install environment\n",
+    "2. Restart this Jupyter Notebook\n",
+    "3. Select Kernel retfound"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "7cbf5e93-6ca0-4401-88e6-64e39968e7cd",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import sys, torch\n",
+    "from pathlib import Path\n",
+    "import os\n",
+    "\n",
+    "PROJECT_ROOT = Path.cwd().resolve()\n",
+    "\n",
+    "if PROJECT_ROOT.name == 'examples': PROJECT_ROOT = PROJECT_ROOT.parent\n",
+    "os.chdir(PROJECT_ROOT)\n",
+    "\n",
+    "print('Project root:', PROJECT_ROOT)\n",
+    "print(\"sys.executable:\", sys.executable)\n",
+    "print(\"torch version:\", torch.__version__)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "ed67953f",
+   "metadata": {},
+   "source": [
+    "## 2. Prepare MESSIDOR2 dataset\n",
+    "1. Download from the [shared data pool](https://github.com/rmaphoh/RETFound/blob/main/BENCHMARK.md).\n",
+    "2. Put the data folder under the project directory, e.g. \"RETFound/MESSIDOR2\"\n"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "357be2fa-a914-4d1f-8759-76b2b1c3f20f",
+   "metadata": {},
+   "source": [
+    "## 3. Hyperparameter and path settings\n",
+    "1. Can choose finetune or lp (linear probe)\n",
+    "2. Model selection [info](https://github.com/rmaphoh/RETFound#:~:text=In%20train.sh%2C%20the%20model%20can%20be%20selected%20by%20changing%20the%20hyperparameters%20MODEL%2C%20MODEL_ARCH%2C%20FINETUNE%3A)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "5f675843",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from pathlib import Path\n",
+    "ADAPTATION='finetune'\n",
+    "MODEL='RETFound_dinov2'\n",
+    "MODEL_ARCH='retfound_dinov2'\n",
+    "FINETUNE='RETFound_dinov2_meh'\n",
+    "DATASET='MESSIDOR2'\n",
+    "NUM_CLASS=5\n",
+    "DATA_PATH=PROJECT_ROOT/DATASET\n",
+    "BATCH_SIZE=24\n",
+    "EPOCHS=50\n",
+    "INPUT_SIZE=224\n",
+    "WORLD_SIZE=1\n",
+    "TASK=f\"{MODEL_ARCH}_{DATASET}_{ADAPTATION}\"\n",
+    "OUTPUT_DIR=PROJECT_ROOT/'output_dir'/TASK\n",
+    "print('DATA_PATH:',DATA_PATH)\n",
+    "print('TASK:',TASK)\n",
+    "print('OUTPUT_DIR:',OUTPUT_DIR)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "6ac04845",
+   "metadata": {},
+   "source": [
+    "## 4. Fine-tuning and testing RETFound on MESSIDOR2"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "d23ff751",
+   "metadata": {
+    "scrolled": true
+   },
+   "outputs": [],
+   "source": [
+    "import sys\n",
+    "\n",
+    "!{sys.executable} main_finetune.py \\\n",
+    "  --model {MODEL} \\\n",
+    "  --model_arch {MODEL_ARCH} \\\n",
+    "  --finetune {FINETUNE} \\\n",
+    "  --savemodel \\\n",
+    "  --global_pool \\\n",
+    "  --batch_size {BATCH_SIZE} \\\n",
+    "  --epochs {EPOCHS} \\\n",
+    "  --nb_classes {NUM_CLASS} \\\n",
+    "  --data_path {DATA_PATH} \\\n",
+    "  --input_size {INPUT_SIZE} \\\n",
+    "  --task {TASK} \\\n",
+    "  --adaptation {ADAPTATION}"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "84ce93ac",
+   "metadata": {},
+   "source": [
+    "## 5. Evaluation-only"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "0af0f8a7",
+   "metadata": {
+    "scrolled": true
+   },
+   "outputs": [],
+   "source": [
+    "import sys\n",
+    "\n",
+    "CKPT = OUTPUT_DIR / \"checkpoint-best.pth\"\n",
+    "\n",
+    "!{sys.executable} main_finetune.py \\\n",
+    "  --model {MODEL} \\\n",
+    "  --model_arch {MODEL_ARCH} \\\n",
+    "  --finetune {FINETUNE} \\\n",
+    "  --savemodel \\\n",
+    "  --global_pool \\\n",
+    "  --batch_size 128 \\\n",
+    "  --nb_classes {NUM_CLASS} \\\n",
+    "  --data_path {DATA_PATH} \\\n",
+    "  --input_size {INPUT_SIZE} \\\n",
+    "  --task {TASK} \\\n",
+    "  --adaptation {ADAPTATION} \\\n",
+    "  --eval \\\n",
+    "  --resume {CKPT}\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "02d2dce7-31c2-48e2-87ce-9223b74cf94e",
+   "metadata": {},
+   "outputs": [],
+   "source": []
+  }
+ ],
+ "metadata": {
+  "environment": {
+   "kernel": "retfound",
+   "name": "workbench-notebooks.m128",
+   "type": "gcloud",
+   "uri": "us-docker.pkg.dev/deeplearning-platform-release/gcr.io/workbench-notebooks:m128"
+  },
+  "kernelspec": {
+   "display_name": "retfound_jupyter (Local)",
+   "language": "python",
+   "name": "retfound"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.11.0"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}

code/RETFound/latent_feature.ipynb

@@ -0,0 +1,196 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "0ae19951",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import os\n",
+    "import torch\n",
+    "import torch.nn as nn\n",
+    "import numpy as np\n",
+    "import pandas as pd\n",
+    "from PIL import Image\n",
+    "import models_vit as models\n",
+    "from huggingface_hub import hf_hub_download\n",
+    "np.set_printoptions(threshold=np.inf)\n",
+    "np.random.seed(1)\n",
+    "torch.manual_seed(1)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "90c3d964",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "def prepare_model(chkpt_dir, arch='RETFound_mae'):\n",
+    "    \n",
+    "    # load model\n",
+    "    checkpoint = torch.load(chkpt_dir, map_location='cpu', weights_only=False)\n",
+    "    \n",
+    "    # build model\n",
+    "    if arch=='RETFound_mae':\n",
+    "        model = models.__dict__[arch](\n",
+    "            img_size=224,\n",
+    "            num_classes=5,\n",
+    "            drop_path_rate=0,\n",
+    "            global_pool=True,\n",
+    "        )\n",
+    "        msg = model.load_state_dict(checkpoint['model'], strict=False)\n",
+    "    else:\n",
+    "        model = models.__dict__[arch](\n",
+    "            num_classes=5,\n",
+    "            drop_path_rate=0,\n",
+    "            args=None,\n",
+    "        )\n",
+    "        msg = model.load_state_dict(checkpoint['teacher'], strict=False)\n",
+    "    return model\n",
+    "\n",
+    "def run_one_image(img, model, arch):\n",
+    "    \n",
+    "    x = torch.tensor(img)\n",
+    "    x = x.unsqueeze(dim=0)\n",
+    "    x = torch.einsum('nhwc->nchw', x)\n",
+    "    \n",
+    "    x = x.to(device, non_blocking=True)\n",
+    "    latent = model.forward_features(x.float())\n",
+    "    \n",
+    "    if arch=='dinov2_large':\n",
+    "        latent = latent[:, 1:, :].mean(dim=1,keepdim=True)\n",
+    "        latent = nn.LayerNorm(latent.shape[-1], eps=1e-6).to(device)(latent)\n",
+    "    \n",
+    "    latent = torch.squeeze(latent)\n",
+    "\n",
+    "    return latent\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "9a250363",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "def get_feature(data_path,\n",
+    "                chkpt_dir,\n",
+    "                device,\n",
+    "                arch='RETFound_mae'):\n",
+    "     #loading model\n",
+    "    model_ = prepare_model(chkpt_dir, arch)\n",
+    "    model_.to(device)\n",
+    "\n",
+    "    img_list = os.listdir(data_path)\n",
+    "    \n",
+    "    name_list = []\n",
+    "    feature_list = []\n",
+    "    model_.eval()\n",
+    "    \n",
+    "    finished_num = 0\n",
+    "    for i in img_list:\n",
+    "        finished_num+=1\n",
+    "        if (finished_num%1000 == 0):\n",
+    "            print(str(finished_num)+\"finished\")\n",
+    "        \n",
+    "        img = Image.open(os.path.join(data_path, i))\n",
+    "        img = img.resize((224, 224))\n",
+    "        img = np.array(img) / 255.\n",
+    "        img[...,0] = (img[...,0] - img[...,0].mean())/img[...,0].std()\n",
+    "        img[...,1] = (img[...,1] - img[...,1].mean())/img[...,1].std()\n",
+    "        img[...,2] = (img[...,2] - img[...,2].mean())/img[...,2].std()\n",
+    "        assert img.shape == (224, 224, 3)\n",
+    "        \n",
+    "        latent_feature = run_one_image(img, model_,arch)\n",
+    "        \n",
+    "        name_list.append(i)\n",
+    "        feature_list.append(latent_feature.detach().cpu().numpy())\n",
+    "        \n",
+    "    return [name_list,feature_list]\n",
+    "\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "54acfcd7",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "chkpt_dir = hf_hub_download(repo_id=\"YukunZhou/RETFound_dinov2_meh\", filename=\"RETFound_dinov2_meh.pth\")\n",
+    "data_path = 'DATA_PATH'\n",
+    "device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')\n",
+    "arch='dinov2_large'"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "0296f74e",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "[name_list,feature]=get_feature(data_path,\n",
+    "                chkpt_dir,\n",
+    "                device,\n",
+    "                arch=arch)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 23,
+   "id": "925d3994",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "#save the feature\n",
+    "df_feature = pd.DataFrame(feature)\n",
+    "df_imgname = pd.DataFrame(name_list)\n",
+    "df_visualization = pd.concat([df_imgname,df_feature], axis=1)\n",
+    "column_name_list = []\n",
+    "\n",
+    "for i in range(1024):\n",
+    "    column_name_list.append(\"feature_{}\".format(i))\n",
+    "df_visualization.columns = [\"name\"] + column_name_list\n",
+    "df_visualization.to_csv(\"Feature.csv\",index=False)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "7f0d13a7-2b46-40eb-ab48-5f90a6aeecb5",
+   "metadata": {},
+   "outputs": [],
+   "source": []
+  }
+ ],
+ "metadata": {
+  "environment": {
+   "kernel": "test",
+   "name": "common-cu121.m123",
+   "type": "gcloud",
+   "uri": "us-docker.pkg.dev/deeplearning-platform-release/gcr.io/base-cu121:m123"
+  },
+  "kernelspec": {
+   "display_name": "Python_test (Local)",
+   "language": "python",
+   "name": "test"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.11.0"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}

code/RETFound/main_finetune.py

@@ -0,0 +1,451 @@
+#!/usr/bin/env python3
+
+# =========================
+import argparse
+import datetime
+import json
+import os
+import time
+from pathlib import Path
+import warnings
+import faulthandler
+
+# =========================
+import numpy as np
+import torch
+import torch.backends.cudnn as cudnn
+from torch.utils.tensorboard import SummaryWriter
+from timm.models.layers import trunc_normal_
+from timm.data.mixup import Mixup
+from huggingface_hub import hf_hub_download, login  # login imported as in original
+
+# =========================
+import models_vit as models
+import util.lr_decay as lrd
+import util.misc as misc
+from util.datasets import build_dataset
+from util.pos_embed import interpolate_pos_embed
+from util.misc import NativeScalerWithGradNormCount as NativeScaler
+from engine_finetune import train_one_epoch, evaluate
+
+# =========================
+faulthandler.enable()
+warnings.simplefilter(action="ignore", category=FutureWarning)
+
+
+def get_args_parser():
+    parser = argparse.ArgumentParser(
+        "MAE fine-tuning / linear probing for image classification", add_help=False
+    )
+
+    # ---- Core training
+    parser.add_argument("--batch_size", default=128, type=int,
+                        help="Batch size per GPU (effective batch size = batch_size * accum_iter * #gpus)")
+    parser.add_argument("--epochs", default=50, type=int)
+    parser.add_argument("--accum_iter", default=1, type=int,
+                        help="Gradient accumulation steps")
+
+    # ---- Model parameters
+    parser.add_argument("--model", default="vit_large_patch16", type=str, metavar="MODEL",
+                        help="Model entry in models_vit.py")
+    parser.add_argument("--model_arch", default="dinov3_vits16", type=str, metavar="MODEL_ARCH",
+                        help="Backbone architecture key (e.g., dinov2_vitl14, convnext_base, etc.)")
+    parser.add_argument("--input_size", default=256, type=int, help="Image size")
+    parser.add_argument("--drop_path", type=float, default=0.2, metavar="PCT", help="Drop path rate")
+    parser.add_argument("--global_pool", action="store_true"); parser.set_defaults(global_pool=True)
+    parser.add_argument("--cls_token", action="store_false", dest="global_pool",
+                        help="Use class token instead of global pool for classification")
+
+    # ---- Optimizer parameters
+    parser.add_argument("--clip_grad", type=float, default=None, metavar="NORM", help="Clip grad norm")
+    parser.add_argument("--weight_decay", type=float, default=0.05, help="Weight decay")
+    parser.add_argument("--lr", type=float, default=None, metavar="LR", help="Absolute LR (overrides blr)")
+    parser.add_argument("--blr", type=float, default=5e-3, metavar="LR",
+                        help="Base LR: lr = blr * total_batch_size / 256")
+    parser.add_argument("--layer_decay", type=float, default=0.65, help="Layer-wise LR decay (ViT)")
+    parser.add_argument("--min_lr", type=float, default=1e-6, metavar="LR", help="Lower LR bound")
+    parser.add_argument("--warmup_epochs", type=int, default=10, metavar="N", help="Warmup epochs")
+
+    # ---- Augmentation
+    parser.add_argument("--color_jitter", type=float, default=None, metavar="PCT")
+    parser.add_argument("--aa", type=str, default="rand-m9-mstd0.5-inc1", metavar="NAME")
+    parser.add_argument("--smoothing", type=float, default=0.1)
+
+    # ---- Random erase
+    parser.add_argument("--reprob", type=float, default=0.25, metavar="PCT")
+    parser.add_argument("--remode", type=str, default="pixel")
+    parser.add_argument("--recount", type=int, default=1)
+    parser.add_argument("--resplit", action="store_true", default=False)
+
+    # ---- Mixup/Cutmix
+    parser.add_argument("--mixup", type=float, default=0.0)
+    parser.add_argument("--cutmix", type=float, default=0.0)
+    parser.add_argument("--cutmix_minmax", type=float, nargs="+", default=None)
+    parser.add_argument("--mixup_prob", type=float, default=1.0)
+    parser.add_argument("--mixup_switch_prob", type=float, default=0.5)
+    parser.add_argument("--mixup_mode", type=str, default="batch")
+
+    # ---- Finetuning & adaptation
+    parser.add_argument("--finetune", default="", type=str, help="Checkpoint id/path (see model rules below)")
+    parser.add_argument("--task", default="", type=str, help="Task name for logging/output grouping")
+    parser.add_argument("--adaptation", default="finetune", choices=["finetune", "lp"],
+                        help="Adaptation strategy: finetune=full fine-tune, lp=linear probe (train head only)")
+
+    # ---- Dataset & paths
+    parser.add_argument("--data_path", default="./data/", type=str)
+    parser.add_argument("--nb_classes", default=8, type=int)
+    parser.add_argument("--output_dir", default="./output_dir")
+    parser.add_argument("--log_dir", default="./output_logs")
+
+    # >>> NEW: training data efficiency <<<
+    parser.add_argument(
+        "--dataratio", type=str, default="1.0",
+        help=('Training data ratio(s) for subsampling in build_dataset. '
+              'Use a single float in (0,1] (e.g., 0.25) or a comma-separated list '
+              '(e.g., "1.0,0.5,0.25") if your build_dataset supports sweeps.')
+    )
+    parser.add_argument(
+        "--stratified", action="store_true",
+        help="If set, subsample training data in a class-stratified manner (requires support in build_dataset)."
+    )
+
+    # ---- Runtime
+    parser.add_argument("--device", default="cuda")
+    parser.add_argument("--seed", default=0, type=int)
+    parser.add_argument("--resume", default="", help="Resume full state (optimizer, scaler, etc.)")
+    parser.add_argument("--start_epoch", default=0, type=int, metavar="N")
+    parser.add_argument("--eval", action="store_true", help="Evaluation only")
+    parser.add_argument("--dist_eval", action="store_true", default=False,
+                        help="Distributed evaluation (faster monitoring during training)")
+    parser.add_argument("--num_workers", default=10, type=int)
+    parser.add_argument("--pin_mem", action="store_true"); parser.set_defaults(pin_mem=True)
+
+    # ---- Distributed
+    parser.add_argument("--world_size", default=1, type=int)
+    parser.add_argument("--local_rank", default=-1, type=int)
+    parser.add_argument("--dist_on_itp", action="store_true")
+    parser.add_argument("--dist_url", default="env://")
+
+    # ---- Misc
+    parser.add_argument("--savemodel", action="store_true", default=True, help="Save best model")
+    parser.add_argument("--norm", default="IMAGENET", type=str)
+    parser.add_argument("--enhance", action="store_true", default=False)
+    parser.add_argument("--datasets_seed", default=2026, type=int)
+
+    return parser
+
+
+# =========================
+# Main
+# =========================
+def main(args, criterion):
+    # ---- Optionally load args from resume (when training)
+    if args.resume and not args.eval:
+        resume_path = args.resume
+        checkpoint = torch.load(args.resume, map_location="cpu")
+        print(f"Load checkpoint (args) from: {args.resume}")
+        args = checkpoint["args"]
+        args.resume = resume_path
+
+    # ---- Distributed setup
+    misc.init_distributed_mode(args)
+
+    print(f"job dir: {os.path.dirname(os.path.realpath(__file__))}")
+    print(f"{args}".replace(", ", ",\n"))
+
+    device = torch.device(args.device)
+
+    # ---- Reproducibility
+    seed = args.seed + misc.get_rank()
+    torch.manual_seed(seed)
+    np.random.seed(seed)
+    cudnn.benchmark = True
+
+    # ---- Build model
+    if args.model == "RETFound_mae":
+        model = models.__dict__[args.model](
+            img_size=args.input_size,
+            num_classes=args.nb_classes,
+            drop_path_rate=args.drop_path,
+            global_pool=args.global_pool,
+        )
+    else:
+        model = models.__dict__[args.model](
+            num_classes=args.nb_classes,
+            drop_path_rate=args.drop_path,
+            args=args,
+        )
+
+    # ---- Load pre-trained weights (if requested and not eval-only)
+    if args.finetune and not args.eval:
+        print(f"Preparing to load pre-trained weights: {args.finetune}")
+
+        if args.model in ["Dinov3", "Dinov2"]:
+            checkpoint_path = args.finetune  # local path
+        elif args.model in ["RETFound_dinov2", "RETFound_mae"]:
+            if os.path.exists(args.finetune):
+                checkpoint_path = args.finetune  # local path (patched)
+            else:
+                print(f"Downloading pre-trained weights from Hugging Face Hub: {args.finetune}")
+                checkpoint_path = hf_hub_download(
+                    repo_id=f"YukunZhou/{args.finetune}",
+                    filename=f"{args.finetune}.pth",
+                )
+        else:
+            raise ValueError(
+                f"Unsupported model '{args.model}'. "
+                f"Expected one of: Dinov3, Dinov2, RETFound_dinov2, RETFound_mae"
+            )
+
+        checkpoint = torch.load(checkpoint_path, map_location="cpu")
+        print(f"Loaded pre-trained checkpoint from: {checkpoint_path}")
+
+        if args.model in ["Dinov3", "Dinov2"]:
+            checkpoint_model = checkpoint
+        elif args.model == "RETFound_dinov2":
+            checkpoint_model = checkpoint["teacher"]
+        else:  # RETFound_mae
+            checkpoint_model = checkpoint["model"]
+
+        # -- Key hygiene
+        checkpoint_model = {k.replace("backbone.", ""): v for k, v in checkpoint_model.items()}
+        checkpoint_model = {k.replace("mlp.w12.", "mlp.fc1."): v for k, v in checkpoint_model.items()}
+        checkpoint_model = {k.replace("mlp.w3.", "mlp.fc2."): v for k, v in checkpoint_model.items()}
+
+        # -- Remove classifier if shape mismatched
+        state_dict = model.state_dict()
+        for k in ["head.weight", "head.bias"]:
+            if k in checkpoint_model and checkpoint_model[k].shape != state_dict[k].shape:
+                print(f"Removing key {k} from pretrained checkpoint")
+                del checkpoint_model[k]
+
+        # -- Interpolate pos embed (ViT)
+        interpolate_pos_embed(model, checkpoint_model)
+
+        # -- Load backbone weights (non-strict)
+        _ = model.load_state_dict(checkpoint_model, strict=False)
+
+        # -- Re-init head
+        if hasattr(model, "head") and hasattr(model.head, "weight"):
+            trunc_normal_(model.head.weight, std=2e-5)
+
+    # ---- Datasets & samplers
+    dataset_train = build_dataset(is_train="train", args=args)
+    dataset_val   = build_dataset(is_train="val",   args=args)
+    dataset_test  = build_dataset(is_train="test",  args=args)
+
+    num_tasks   = misc.get_world_size()
+    global_rank = misc.get_rank()
+
+    if not args.eval:
+        sampler_train = torch.utils.data.DistributedSampler(
+            dataset_train, num_replicas=num_tasks, rank=global_rank, shuffle=True
+        )
+        print(f"Sampler_train = {sampler_train}")
+        if args.dist_eval:
+            if len(dataset_val) % num_tasks != 0:
+                print("Warning: dist eval with dataset not divisible by #procs; results may differ slightly.")
+            sampler_val = torch.utils.data.DistributedSampler(
+                dataset_val, num_replicas=num_tasks, rank=global_rank, shuffle=True
+            )
+        else:
+            sampler_val = torch.utils.data.SequentialSampler(dataset_val)
+
+    if args.dist_eval:
+        if len(dataset_test) % num_tasks != 0:
+            print("Warning: dist eval test set not divisible by #procs; results may differ slightly.")
+        sampler_test = torch.utils.data.DistributedSampler(
+            dataset_test, num_replicas=num_tasks, rank=global_rank, shuffle=True
+        )
+    else:
+        sampler_test = torch.utils.data.SequentialSampler(dataset_test)
+
+    # ---- Logging
+    if global_rank == 0 and args.log_dir is not None and not args.eval:
+        os.makedirs(args.log_dir, exist_ok=True)
+        log_writer = SummaryWriter(log_dir=os.path.join(args.log_dir, args.task))
+    else:
+        log_writer = None
+
+    # ---- DataLoaders
+    if not args.eval:
+        data_loader_train = torch.utils.data.DataLoader(
+            dataset_train, sampler=sampler_train,
+            batch_size=args.batch_size, num_workers=args.num_workers,
+            pin_memory=args.pin_mem, drop_last=True,
+        )
+        print(f"len of train_set: {len(data_loader_train) * args.batch_size}")
+
+        data_loader_val = torch.utils.data.DataLoader(
+            dataset_val, sampler=sampler_val,
+            batch_size=args.batch_size, num_workers=args.num_workers,
+            pin_memory=args.pin_mem, drop_last=False,
+        )
+
+    data_loader_test = torch.utils.data.DataLoader(
+        dataset_test, sampler=sampler_test,
+        batch_size=args.batch_size, num_workers=args.num_workers,
+        pin_memory=args.pin_mem, drop_last=False,
+    )
+
+    # ---- Mixup/CutMix
+    mixup_fn = None
+    mixup_active = (args.mixup > 0) or (args.cutmix > 0.) or (args.cutmix_minmax is not None)
+    if mixup_active:
+        print("Mixup is activated!")
+        mixup_fn = Mixup(
+            mixup_alpha=args.mixup, cutmix_alpha=args.cutmix, cutmix_minmax=args.cutmix_minmax,
+            prob=args.mixup_prob, switch_prob=args.mixup_switch_prob, mode=args.mixup_mode,
+            label_smoothing=args.smoothing, num_classes=args.nb_classes
+        )
+
+    # ---- Eval-only: resume weights
+    if args.resume and args.eval:
+        checkpoint = torch.load(args.resume, map_location="cpu")
+        print(f"Load checkpoint for eval from: {args.resume}")
+        model.load_state_dict(checkpoint["model"])
+
+    model.to(device)
+    model_without_ddp = model
+
+    # ---- Adaptation toggle
+    if args.adaptation == "lp":
+        for name, param in model.named_parameters():
+            param.requires_grad = ("head" in name)
+        print("[Adaptation] Linear probe: training classifier head only.")
+    else:
+        print("[Adaptation] Full fine-tuning: training all parameters.")
+
+    # ---- Count trainable params
+    n_parameters = sum(p.numel() for p in model.parameters() if p.requires_grad)
+    print(f"number of trainable params (M): {n_parameters / 1.e6:.2f}")
+
+    # ---- LR scaling by effective batch size
+    eff_batch_size = args.batch_size * args.accum_iter * misc.get_world_size()
+    if args.lr is None:
+        args.lr = args.blr * eff_batch_size / 256
+    print(f"base lr: {args.lr * 256 / eff_batch_size:.2e}")
+    print(f"actual lr: {args.lr:.2e}")
+    print(f"accumulate grad iterations: {args.accum_iter}")
+    print(f"effective batch size: {eff_batch_size}")
+
+    # ---- DDP (if available)
+    if args.distributed and torch.cuda.device_count() > 1:
+        ddp_kwargs = {}
+        if args.adaptation == "lp":
+            ddp_kwargs["find_unused_parameters"] = True
+        model = torch.nn.parallel.DistributedDataParallel(
+            model, device_ids=[args.gpu], **ddp_kwargs
+        )
+        model_without_ddp = model.module
+    else:
+        model_without_ddp = model  # single-GPU
+
+    # ---- Optimizer param groups (after freezing)
+    no_weight_decay = (model_without_ddp.no_weight_decay()
+                       if hasattr(model_without_ddp, "no_weight_decay") else [])
+
+
+    param_groups = lrd.param_groups_lrd(
+        model_without_ddp,
+        weight_decay=args.weight_decay,
+        no_weight_decay_list=no_weight_decay,
+        layer_decay=args.layer_decay,
+    )
+    for g in param_groups:
+        g["params"] = [p for p in g["params"] if p.requires_grad]
+
+    optimizer = torch.optim.AdamW(param_groups, lr=args.lr)
+    loss_scaler = NativeScaler()
+    print(f"criterion = {criterion}")
+
+    # ---- Load previous full state (optimizer, scaler, etc.)
+    misc.load_model(args=args, model_without_ddp=model_without_ddp,
+                    optimizer=optimizer, loss_scaler=loss_scaler)
+
+    # =========================
+    # Eval-only Short Circuit
+    # =========================
+    if args.eval:
+        if "checkpoint" in locals() and isinstance(checkpoint, dict) and ("epoch" in checkpoint):
+            print(f"Test with the best model at epoch = {checkpoint['epoch']}")
+        test_stats, auc_roc = evaluate(
+            data_loader_test, model, device, args, epoch=0, mode="test",
+            num_class=args.nb_classes, log_writer=log_writer
+        )
+        return
+
+    # =========================
+    # Train Loop
+    # =========================
+    print(f"Start training for {args.epochs} epochs")
+    start_time = time.time()
+    max_score = 0.0
+    best_epoch = 0
+
+    for epoch in range(args.start_epoch, args.epochs):
+        if args.distributed:
+            data_loader_train.sampler.set_epoch(epoch)
+
+        train_stats = train_one_epoch(
+            model, criterion, data_loader_train,
+            optimizer, device, epoch, loss_scaler,
+            args.clip_grad, mixup_fn,
+            log_writer=log_writer, args=args
+        )
+
+        val_stats, val_score = evaluate(
+            data_loader_val, model, device, args, epoch, mode="val",
+            num_class=args.nb_classes, log_writer=log_writer
+        )
+
+        if max_score < val_score:
+            max_score = val_score
+            best_epoch = epoch
+            if args.output_dir and args.savemodel:
+                misc.save_model(
+                    args=args, model=model, model_without_ddp=model_without_ddp,
+                    optimizer=optimizer, loss_scaler=loss_scaler, epoch=epoch, mode="best"
+                )
+        print(f"Best epoch = {best_epoch}, Best score = {max_score:.4f}")
+
+        if log_writer is not None:
+            log_writer.add_scalar("loss/val", val_stats["loss"], epoch)
+            log_writer.flush()
+
+        log_stats = {**{f"train_{k}": v for k, v in train_stats.items()},
+                     "epoch": epoch,
+                     "n_parameters": n_parameters}
+
+        if args.output_dir and misc.is_main_process():
+            with open(os.path.join(args.output_dir, args.task, "log.txt"), "a", encoding="utf-8") as f:
+                f.write(json.dumps(log_stats) + "\n")
+
+    # =========================
+    # Final Test (Best Ckpt)
+    # =========================
+    ckpt_path = os.path.join(args.output_dir, args.task, "checkpoint-best.pth")
+    checkpoint = torch.load(ckpt_path, map_location="cpu")
+    model_without_ddp.load_state_dict(checkpoint["model"], strict=False)
+    model.to(device)
+    print(f"Test with the best model, epoch = {checkpoint.get('epoch', -1)}:")
+    _test_stats, _auc_roc = evaluate(
+        data_loader_test, model, device, args, -1, mode="test",
+        num_class=args.nb_classes, log_writer=None
+    )
+
+    total_time = time.time() - start_time
+    total_time_str = str(datetime.timedelta(seconds=int(total_time)))
+    print(f"Training time {total_time_str}")
+
+
+if __name__ == "__main__":
+    args = get_args_parser()
+    args = args.parse_args()
+
+    criterion = torch.nn.CrossEntropyLoss()
+
+    if args.output_dir:
+        Path(args.output_dir).mkdir(parents=True, exist_ok=True)
+
+    main(args, criterion)

code/RETFound/models_vit.py

@@ -0,0 +1,105 @@
+
+from functools import partial
+
+import timm.models.vision_transformer
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+from torch import Tensor
+from timm.models.layers import trunc_normal_
+
+class VisionTransformer(timm.models.vision_transformer.VisionTransformer):
+    """ Vision Transformer with support for global average pooling
+    """
+    def __init__(self, global_pool=False, **kwargs):
+        super(VisionTransformer, self).__init__(**kwargs)
+
+        self.global_pool = global_pool
+        if self.global_pool:
+            norm_layer = kwargs['norm_layer']
+            embed_dim = kwargs['embed_dim']
+            self.fc_norm = norm_layer(embed_dim)
+
+            del self.norm  # remove the original norm
+
+    def forward_features(self, x):
+        B = x.shape[0]
+        x = self.patch_embed(x)
+
+        cls_tokens = self.cls_token.expand(B, -1, -1)  # stole cls_tokens impl from Phil Wang, thanks
+        x = torch.cat((cls_tokens, x), dim=1)
+        x = x + self.pos_embed
+        x = self.pos_drop(x)
+
+        for blk in self.blocks:
+            x = blk(x)
+
+        if self.global_pool:
+            x = x[:, 1:, :].mean(dim=1,keepdim=True)  # global pool without cls token
+            outcome = self.fc_norm(x)
+        else:
+            x = self.norm(x)
+            outcome = x[:, 0]
+
+        return outcome
+
+
+def RETFound_mae(**kwargs):
+    model = VisionTransformer(
+        patch_size=16, embed_dim=1024, depth=24, num_heads=16, mlp_ratio=4, qkv_bias=True,
+        norm_layer=partial(nn.LayerNorm, eps=1e-6), **kwargs)
+    return model
+
+
+
+def Dinov2(args, **kwargs):
+    
+    if args.model_arch == 'dinov2_vits14':
+        arch = 'vit_small_patch14_dinov2.lvd142m'
+    elif args.model_arch == 'dinov2_vitb14':
+        arch = 'vit_base_patch14_dinov2.lvd142m'
+    elif args.model_arch == 'dinov2_vitl14':
+        arch = 'vit_large_patch14_dinov2.lvd142m'
+    elif args.model_arch == 'dinov2_vitg14':
+        arch = 'vit_giant_patch14_dinov2.lvd142m'
+    else:
+        raise ValueError(f"Unknown model_arch '{args.model_arch}'. "
+                         f"Expected one of: dinov2_vits14, dinov2_vitb14, dinov2_vitl14, dinov2_vitg14")
+        
+    model = timm.create_model(
+        arch,
+        pretrained=True,
+        img_size=224,
+        **kwargs
+    )
+    return model
+
+
+
+def RETFound_dinov2(args, **kwargs):
+    model = timm.create_model(
+        'vit_large_patch14_dinov2.lvd142m',
+        pretrained=True,
+        img_size=224,
+        **kwargs
+    )
+    return model
+
+
+def Dinov3(args, **kwargs):
+    # Load ViT-L/16 backbone (hub model has `head = Identity` by default)
+    model = torch.hub.load(
+        repo_or_dir="facebookresearch/dinov3",
+        model=args.model_arch,
+        pretrained=False,   # main() will load your checkpoint
+        trust_repo=True,
+    )
+
+    # Figure out feature dimension for the probe
+    feat_dim = getattr(model, "embed_dim", None) or getattr(model, "num_features", None)
+    model.head = nn.Linear(feat_dim, args.nb_classes)
+    trunc_normal_(model.head.weight, std=2e-5)
+    if model.head.bias is not None:
+        nn.init.zeros_(model.head.bias)
+
+    return model

code/RETFound/output_logs/adam_005pct/events.out.tfevents.1782887956.qs-55163-1487918-ai-1040476-default0-0.476175.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7f96fd06ad4d36e7cf9a52f1e0118185d2c17480154cc55826f1e87b92cc5760
+size 88

code/RETFound/output_logs/adam_010pct/events.out.tfevents.1782887911.qs-55163-1487918-ai-1040476-default0-0.470020.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c0e785911f32393fbac2a5d827a89347f5578ac8b256007671e2d6fc01e8a32a
+size 88

code/RETFound/output_logs/adam_025pct/events.out.tfevents.1782887757.qs-55163-1487918-ai-1040476-default0-0.451324.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b4c2a7b4ef9db0365f153d4e681a1abcdc0e7e49585d9e5ef26fdd0d640c624c
+size 36644

code/RETFound/output_logs/adam_050pct/events.out.tfevents.1782887757.qs-55163-1487918-ai-1040476-default0-0.451322.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0959643206d2b20a91bda9f5aaa4191b99e98260519c5daa69117d265ffdabce
+size 45778

code/RETFound/output_logs/adam_100pct/events.out.tfevents.1782887757.qs-55163-1487918-ai-1040476-default0-0.451323.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b5985aaaba9b06f78ab1b533af7bbde844778864f1c369d2a62d94bb74f90847
+size 64044

code/RETFound/output_logs/airogs_005pct/events.out.tfevents.1782888170.qs-55163-1487918-ai-1040476-default0-0.501553.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cb8d443582ed9dfbab23bfcf5de3eaa94ef74d387aafa9136b00443a000a5e79
+size 35620

code/RETFound/output_logs/airogs_010pct/events.out.tfevents.1782888170.qs-55163-1487918-ai-1040476-default0-0.501550.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:59d058195317c681bfbc653f87b88227e8ee2ec6987bcef8e5e297bd87765899
+size 57436

code/RETFound/output_logs/airogs_025pct/events.out.tfevents.1782888068.qs-55163-1487918-ai-1040476-default0-0.491038.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1d19073047a8ab892ceeccb0ba03c1a50e5022f35d7a8715ba1d99b44508ed1d
+size 36034

code/RETFound/output_logs/airogs_050pct/events.out.tfevents.1782887997.qs-55163-1487918-ai-1040476-default0-0.481305.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1b9eab235939451532c93b74746dfbacd2c04855c9450249d6c4c3a753a287e6
+size 113028

code/RETFound/output_logs/airogs_100pct/events.out.tfevents.1782887993.qs-55163-1487918-ai-1040476-default0-0.478479.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:46b1d4d79b57dd32a5b328fd5938f2604c2414a24dfcc1a1c0695e1889d09aad
+size 137623

code/RETFound/output_logs/papila_005pct/events.out.tfevents.1782888500.qs-55163-1487918-ai-1040476-default0-0.596649.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a4b50a8b3deb96ceec271daef062980b28975f7233bb83aac82a0e1c154f759d
+size 88

code/RETFound/output_logs/papila_010pct/events.out.tfevents.1782888470.qs-55163-1487918-ai-1040476-default0-0.590487.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:76491fe64534b2eef2bf950300f318b4118f758be852f8d40e6ab0a235453962
+size 88

code/RETFound/output_logs/papila_025pct/events.out.tfevents.1782888304.qs-55163-1487918-ai-1040476-default0-0.547656.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d6c721a36f47af3e3da2a4f47143daefa457136603fea81395eca49a83972ebb
+size 36644

code/RETFound/output_logs/papila_050pct/events.out.tfevents.1782888259.qs-55163-1487918-ai-1040476-default0-0.535042.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:269ae475ea16bd86d06b9e48aa4e87abe167bcdbcb540d449277cba120b77b90
+size 45778

code/RETFound/output_logs/papila_100pct/events.out.tfevents.1782888171.qs-55163-1487918-ai-1040476-default0-0.501618.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:47261cac7b4ca480461833e6fff0a961172ce4d4090516beb47ef517713213cd
+size 68612

code/RETFound/output_logs/retfound/events.out.tfevents.1781502758.qs-55163-1487918-ai-1040476-default0-0.964106.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d54ad1698ac243112b323b01c2a0a28f1bf5d66afe93a6a643c1e9b23527228c
+size 64044

code/RETFound/output_logs/retfound/events.out.tfevents.1781502758.qs-55163-1487918-ai-1040476-default0-0.964108.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:de0b929f80f169a372a6f749a5e5e9b17a839e74da9ab52e8e3af1aea0d8a1d0
+size 441940

code/RETFound/output_logs/retfound/events.out.tfevents.1781502758.qs-55163-1487918-ai-1040476-default0-0.964236.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e8ac8b04c07d7df4f3084d6247aae9f76bf104b35dc638200e09a75b71999d19
+size 164520

code/RETFound/output_logs/retfound/events.out.tfevents.1781502916.qs-55163-1487918-ai-1040476-default0-0.981217.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d78f8f6f47a4c545c04a72a69f2ece568522ebe82262e08cfd127012e6ccb8e7
+size 68612

code/RETFound/output_logs/retfound/events.out.tfevents.1781503026.qs-55163-1487918-ai-1040476-default0-0.990775.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:33c3bd9b8fda689167bc22583cc4bf65718d6fd7ff684b010c53078e65ef15a4
+size 68612

code/RETFound/output_logs/retfound/events.out.tfevents.1781503167.qs-55163-1487918-ai-1040476-default0-0.1001549.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2b5aa04bfcfe824f73aa38e3a1e0047787671470b7dddc876dc1f624ab860378
+size 264686

code/RETFound/output_logs/retfound/events.out.tfevents.1781503317.qs-55163-1487918-ai-1040476-default0-0.1009431.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7503f4ec51360aa54bfc1c9cef39e6f4c5b9915d642344c1674a7a68dc278cfa
+size 196488

code/RETFound/output_logs/retfound/events.out.tfevents.1782888167.qs-55163-1487918-ai-1040476-default0-0.499798.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1831a225218409b22f8bfc12c04e1f7dc2578d428ac0de5668dc147461481f48
+size 45778

code/RETFound/output_logs/retfound/events.out.tfevents.1782888168.qs-55163-1487918-ai-1040476-default0-0.499928.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:24f8970fd8263768a36d832360e7f3c1360556a2b99b81c8f3cc2635fbd03344
+size 36644

code/RETFound/output_logs/retfound/events.out.tfevents.1782888169.qs-55163-1487918-ai-1040476-default0-0.499800.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d5a8f968920f47c5c33fd25a0ef918909d5c3656bb8f2b1233bc5238f99e09a8
+size 64044

code/RETFound/output_logs/retfound/events.out.tfevents.1782888299.qs-55163-1487918-ai-1040476-default0-0.545237.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c72746b35f476fc0672e04c7b03133887a85113790d429e5e06ae668b5f56296
+size 51306

code/RETFound/output_logs/retfound/events.out.tfevents.1782888359.qs-55163-1487918-ai-1040476-default0-0.561100.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9b0d289429f92dc01ef3b1649a83e9ff38c1f4955fb586277620f050c9bd412e
+size 51306

code/RETFound/output_logs/retfound/events.out.tfevents.1782888412.qs-55163-1487918-ai-1040476-default0-0.568878.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8df0846151a1bc95ef63417a2f6a1c473c3e8141ae043f36e3591fcf8af6a3d6
+size 441940

code/RETFound/output_logs/retfound/events.out.tfevents.1782888443.qs-55163-1487918-ai-1040476-default0-0.582303.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0f2d814e3bbb0b17670e32d8fe41fe891adb1ec1a450efe90637389cfa8d1657
+size 229236

code/RETFound/output_logs/retfound/events.out.tfevents.1782888625.qs-55163-1487918-ai-1040476-default0-0.614730.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5f8cbde8d4b219ce08fa53d1e265e4fab9920b278417929718ede09de8d37afc
+size 122884

code/RETFound/output_logs/retfound/events.out.tfevents.1782888664.qs-55163-1487918-ai-1040476-default0-0.621091.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:72d20c42662bfec27932a0d9b0f5bc6e09c1787f37a1cae3075660b28205582c
+size 57436

code/RETFound/output_logs/retfound/events.out.tfevents.1782888736.qs-55163-1487918-ai-1040476-default0-0.633055.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ef5ac96bba4b45897919ac78a3e44269af289f42de4d039da7a16e66e314c185
+size 35620

code/RETFound/output_logs/retfound/events.out.tfevents.1782888782.qs-55163-1487918-ai-1040476-default0-0.643431.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9fb255cbf2aa90f23a0eefb748df1d6a911e9fa341a79c90d4710b9bf969ef18
+size 68612

code/RETFound/output_logs/retfound/events.out.tfevents.1782888827.qs-55163-1487918-ai-1040476-default0-0.652209.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ea10c9f7ca9102ba5d73f322721785c0f6bbed2a996e458db737b4b036f87729
+size 45778

code/RETFound/output_logs/retfound/events.out.tfevents.1782888888.qs-55163-1487918-ai-1040476-default0-0.659616.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:71875430fccc73d65ca7e6f86cbb329161f4b0823888bfcf177b30fb3ba3c3ea
+size 36644