Automated AWS Model Creator 1.1.1 [TEMPORARY CLOSED]

Caution

Temporary application closure due to the removal of public access to the developers' private ECR image version 4.3-py310-cu121. The application may still be usable in uploading of dataset and/or downloading results and models from existing training jobs but will not be able to train due to missing access to the image 4.3-py310-cu121 for YOLOv5 Custom Dataset Training. Contact @kerrlabajo for re-allowing public access to the image.

This is a secondary release of the now Automated AWS Model Creator Windows Form Application in .NET 4.8. As of this version, this is production-ready for Mitsumi AEC team of the Cebu Mitsumi Inc. and potentially for other users that intends to train an AI Vision System using YOLOv5 in AWS.

Note

This application is only applicable to Windows 10/11.

Warning

Users are assumed to have prepared a pre-processed and annotated dataset. The model category is supervised learning, object detection in images. The required annotation of the dataset should also be supported for YOLOv5, such as Roboflow for dataset annotation. See the dataset creation section in this link. Otherwise, users may use public and free-to-use datasets from the Roboflow Universe to download a copy, then upload it using this application to train the model for it.

Warning

Users are assumed to be a registered Amazon Web Services (AWS) IAM User with having setup the necessary prerequisites before using the application.

Note

The required image tag version in the user's private ECR to make this application working in any AWS account would be the 2.7-py310-cu121 or more. Otherwise, general users outside of the Mitsumi AEC team would be fetching from the developers' private ECR image version 4.3-py310-cu121.

Installation

1. Download and install AutomatedAWSModelCreator_Installer.msi if you intend to install as administrator.
2. Otherwise, download AutomatedAWSModelCreator_v1.1.1.13.zip if you intend to install as non-administrator. Proceed by executing the setup.exe in the archive for the installation.
3. Run the AutomatedAWSModelCreator that will be shown in your Start Menu.

What's Changed

• CreateTrainingJobRequest by @Ra-Jay in #40
• Retrieve Available Datasets from S3 Bucket by @Ra-Jay in #39
• Fetch Downloads by @Ra-Jay in #41
• Feat/live tail by @JhonrayAcojedo in #42
• Docker Enhancement by @kerrlabajo in #43
• Fetching Dataset Fix by @Ra-Jay in #44
• Implement get UserName to use for user-based custom uploads and training jobs by @kerrlabajo in #45
• Feat/minor fixes by @JhonrayAcojedo in #46
• Download Feature by @Ra-Jay in #47
• Inline Documentation by @Ra-Jay in #48
• Features push by @JhonrayAcojedo in #49
• Fix: No items message stacking on continuous fetching by @Ra-Jay in #50
• Fix: output list sorting by @Ra-Jay in #51
• Customized minor fixes by @kerrlabajo in #52
• Feat/minor fixes by @JhonrayAcojedo in #53
• Major features improvement from version 1.1.0.0 to 1.1.1.x by @kerrlabajo in #54
• Add an installation source via this GitHub repository raw/main/publish/ by @kerrlabajo in #55
• Revert back to disk installation by @kerrlabajo in #56
• Temp branch by @JhonrayAcojedo in #57
• Handle exceptions in failure reason of the training script by @kerrlabajo in #58
• Generalize instructions from any potential error that occurs in subprocess by @kerrlabajo in #59
• Feat/minor fixes by @JhonrayAcojedo in #60
• Minor revisions in docker to add proper failure reason in training job by @kerrlabajo in #61
• Allow generic dataset's yaml config to allow different type of datasets regardless of structure by @kerrlabajo in #62
• Major update to renaming the entire solution and projects by @kerrlabajo in #63
• Remove unused tests and general update by @kerrlabajo in #64

Full Changelog: 1.1.0...1.1.1

LSC-Trainer 1.1.0

This is an initial release of the LSC-Trainer Windows Form Application in .NET 4.8. As of this version, this is production-ready for Mitsumi AEC team of the Cebu Mitsumi Inc. This application is only applicable to Windows 10/11. Users are assumed to be a registered Amazon Web Services (AWS) IAM User with having setup the necessary prerequisites before using the application. The required image tag version to make this application working in any AWS account would be the 2-6-py310-cu121.

Installation

1. Download and install LSC-Trainer-Installer.msi.
2. Run the LSC-Trainer that will be shown in your Start Menu.

Uninstallation

1. Look for Control Panel > Uninstall a Program.
2. Uninstall LSC-Trainer.

Technical Documentation

What's Changed

• DotNet environment setup by @Ra-Jay in #3
• update sagemakerclient initialization by @Ra-Jay in #4
• Add YAML reader and its dependencies by @JhonrayAcojedo in #5
• Connection checking by @Ra-Jay in #6
• Completed the training function by @JhonrayAcojedo in #7
• Implement training logger by @JhonrayAcojedo in #8
• AWS SageMaker Training Job Creation Implementation by @kerrlabajo in #1
• AWS S3 Implementations by @kerrlabajo in #9
• Dockerize Ubuntu with CUDA and YOLOv5 environment by @kerrlabajo in #10
• AWS SageMaker Input Data Configuration by @kerrlabajo in #12
• Progress tracking implementation by @Ra-Jay in #11
• Additional s3 implementation by @kerrlabajo in #13
• AWS SageMaker Export Job Request Implementation by @kerrlabajo in #14
• Partial Application Testing using Free Account by @kerrlabajo in #15
• Fix overlooked development commits by @kerrlabajo in #16
• Follow up minor fixes from dev by @kerrlabajo in #17
• Build and Push New Training Image that has Scripts Caller by @kerrlabajo in #21
• Training Job Logging Feature by @kerrlabajo in #18
• UI QoL Changes by @kerrlabajo in #20
• Login Credentials and Role Handling Feature by @kerrlabajo in #19
• Feat/training logging by @JhonrayAcojedo in #22
• Feat/training script update by @kerrlabajo in #23
• Multi-GPU Distributed Data Parallel Training by @kerrlabajo in #24
• Save User Credentials by @kerrlabajo in #26
• Save User Credentials by @kerrlabajo in #28
• Refactor Code by @kerrlabajo in #25
• Create dotnet.yml by @JhonrayAcojedo in #29
• Partial Multi-Instance GPU DDP Training Setup by @kerrlabajo in #32
• Multi-Node Single/Multi-GPU DDP Training by @kerrlabajo in #34
• Feat/dynamic dropdown by @JhonrayAcojedo in #33
• Features Implementation by @kerrlabajo in #35
• Unit tests by @Ra-Jay in #30
• Feat/create winforms build tests by @JhonrayAcojedo in #36
• Add TestCredentialsSaved by @JhonrayAcojedo in #37
• Implement partial build and unit testing by @kerrlabajo in #38

New Contributors

• @Ra-Jay made their first contribution in #3
• @JhonrayAcojedo made their first contribution in #5
• @kerrlabajo made their first contribution in #1

Full Changelog: https://github.com/kerrlabajo/LSC-Trainer/commits/1.1.0