Skip to content

Latest commit

 

History

History
executable file
·
419 lines (301 loc) · 27.3 KB

README.md

File metadata and controls

executable file
·
419 lines (301 loc) · 27.3 KB
 
Official Site VISIT      Apply for Trial TRY IT OUT      SLACK Community WELCOME
 
 

📫 Feedback on usage issues: [email protected] / Professional consulting for server equipment: [email protected]

 
 

English | 简体中文

Citations

If you wish to refer to YMIR in your work, please use the following BibTeX entry.

@inproceedings{huang2021ymir,
      title={YMIR: A Rapid Data-centric Development Platform for Vision Applications},
      author={Phoenix X. Huang and Wenze Hu and William Brendel and Manmohan Chandraker and Li-Jia Li and Xiaoyu Wang},
      booktitle={Proceedings of the Data-Centric AI Workshop at NeurIPS},
      year={2021},
}

What's new

Version 2.0.0 updated on 11/08/2022

YMIR platform

  • A new model performance diagnosis module.
  • A new function for visual evaluation of model inference results.
  • Adding a public algorithm library with a variety of built-in high-precision algorithms.
  • One-click deployment function, supporting the deployment of algorithms to prerequisite certified devices.
  • New operating instruction.
  • Refactory code structure.

Docker

View more ymir-executor-fork

Within the public dockerimage

  • Update yolov5 training image: youdaoyzbx/ymir-executor:ymir2.0.0-yolov5-cu111-tmi
  • Update mmdetection training image: youdaoyzbx/ymir-executor:ymir2.0.0-mmdet-cu111-tmi
  • Update yolov5 image with rv1126 chip deployment support: youdaoyzbx/ymir-executor:ymir2.0.0-yolov5-cu111-tmid

More code updates ymir-dev.

Deployment Prerequisite (optional)

YMIR supports deploying the trained model and public algorithm model directly to the certified device, for more hardware specs, please check the details.

 
 

Introduction

Catalog

1. Introduction to AI SUITE-YMIR

As a streamlined model development product, YMIR(You Mine In Recursion) focuses on the dataset versioning and model iteration in the AI SUITE open-source series.

 
 

AI commercialization is currently reaching a stage of maturity in terms of computing hardwares, algorithms, etc. The adoption of AI often encounter challenges such as a lack of skilled developers, high development costs and long iteration cycles.

As a platform, YMIR provides an end-to-end AI development system. This platform reduces costs for companies using artificial intelligence and accelerates the adoption of artificial intelligence. YMIR provides ML developers with one-stop services for data processing, model training, and other steps required in the AI development cycle.

The YMIR platform provides effective model development capabilities with a data-centric approach. The platform integrates concepts such as active learning methods, data and model version control, and workspace. Parallel rapid iteration of data sets and projects for multiple specific tasks are realized by YMIR. The platform uses an open API design, so third-party tools can also be integrated into the system.

1.1. Main functions

A typical model development process can usually be summarized in a few steps: defining the task, preparing the data, training the model, evaluating the model, and deploying the model.

  • Define the target: Before starting an AI development project, it is important to be clear about what is to be analyzed. This will help developers correctly convert the problem into several typical ML modeling tasks, such as image classification, object detection, etc. Different tasks have different data requirements.

  • Prepare data: Data preparation is the foundation of a successful AI project. The most important task in this step is to ensure the quality of data and its annotations. Collect all the required data at the beginning is the optimal situation for many projects. Therefore, the project developer may find that some data is missing in subsequent stages. Additional data could be necessary upon the project needs.

  • Train model: This operation is commonly referred to "modeling". This step refers to the exploration and analysis of prepared data to discover the internal patterns and any links between the input and the expected prediction target. The result of this step is usually one or more machine learning models. These models can be applied to new data to obtain predictions. Developers train their own models using mainstream model training frameworks, such as pytorch, tensorflow, darknet, etc.

  • Evaluate model: The entire development process has not yet been completed after training the model. Models need to be evaluated and checked before being put into production. Normally, get a production-quality model all at once is not so easy. You need to adjust parameters, and iterate the model continuously. Some common metrics can help you evaluate models quantitatively and pick a satisfactory model. Some common metrics can help you to evaluate models quantitatively.

  • Deploy model: Models are developed and trained based on previously available data (possibly test data). After a satisfactory model is obtained, it will be applied to real data to make predictions at scale.

YMIR platform mainly meets the needs of users to produce models at scale, provides users with a good and easy-to-use display interface, and facilitates the management and viewing of data and models. The platform contains main functional modules such as project management, tag management, model deployment, system configuration, dockerimage management, etc. It supports the realization of the following main functions.

Function Module Primary Function Secondary Function Function Description
Project Management Project Management Project Editing Supports adding, deleting, and editing projects and project information
Project Management Iteration Management Iteration Preparation Supports setting up the dataset and model information needed for iteration
Project Management Iteration Management Iteration Steps Support to populate the data from the previous round to the next step corresponding to the task parameters
Project Management Iteration Management Iteration Charts Support to display the datasets and models generated during the iterative process in the interface as a graphical comparison
Project Management Dataset Management Import datasets Support users to import prepared datasets by copying public datasets, url addresses, paths, and local imports
Project Management Data Set Management View Data Sets Supports visualization of image data and annotations, and viewing of historical information
Project Management Data Set Management Edit Data Set Support editing and deleting data sets
Project Management Dataset Management Dataset Versions Support creating new dataset versions on the source dataset, with the version number incremented by time
Project Management Data Set Management Data Preprocessing Support image data fusion, filtering, sampling operations
Project Management Data Set Management Data Mining Supports finding the most beneficial data for model optimization in a large number of data sets
Project Management Data Set Management Data Annotation Support for adding annotations to image data
Project Management Data Set Management Data Inference Supports adding annotations to a data set by specifying a model
Project Management Model Management Model Import Support local import of model files to the platform
Project Management Model Management Training Models Support to select datasets, labels, and adjust training parameters to train models according to requirements, and view the corresponding model results after completion
Project Management Model Management Model Validation Support uploading a single image to check the performance of the model in real images through visualization to verify the accuracy of the model
Tag management Tag management Add tags Support adding primary names and aliases of training tags
Model Deployment Algorithm Management Public Algorithm Support algorithm customization, view public algorithms and try them out, support adding to my algorithms
Model Deployment Algorithm Management Public Algorithm Support publishing my algorithms to public algorithms
Model Deployment Algorithm Management My Algorithms Support for viewing and editing my published algorithms and added algorithms
Model Deployment Algorithm Management Deploy Algorithms Support deploying my algorithms to devices and viewing deployment history
Model Deployment Device Management View Devices Support viewing device information and deployment history
Model Deployment Device Management Edit Device Support adding, deploying, and deleting devices
Model Deployment Device Management Support Devices Support viewing and purchasing of supported devices
System Configuration Mirror Management My Mirrors Support for adding custom mirrors to the system (available to administrators only)
System Configuration Mirror Management Public Mirror Support for viewing public mirrors uploaded by others and copying them to your own system
System Configuration Permissions Configuration Permissions Management Support for configuring user permissions (available only to administrators)

1.2. Apply for trial

We provide an online trial version for your convenience. If you need, please fill out the Apply for YMIR Trial , and we will send the trial information to your email address.

2. Installation

How do users choose to install GUI or CMD?

  1. The GUI verision with the supports of model training and model iteration is more suitable for ordinary users.

  2. If you need to modify the default configuration of the system, it is recommended to install CMD;

  3. If you have already deployed the existing version of ymir, please refer to the Upgrade Instructions.

This chapter contains the installation instructions for YMIR-GUI. If you need to use CMD, please refer to the Ymir-CMD user guide.

2.1. Environment dependencies

1.NVIDIA drivers shall be properly installed before installing YMIR. For detailed instructions, see https://www.nvidia.cn/geforce/drivers/.

  1. Docker and Docker Compose installation:
## check the maximum CUDA version supported by the host
nvidia-smi
## for Host support cuda 11+, check nvidia-docker
sudo docker run --rm --gpus all nvidia/cuda:11.0.3-base-ubuntu20.04 nvidia-smi
## for Host support cuda 10+, check nvidia-docker
sudo docker run --rm --gpus all nvidia/cuda:10.2-base-ubuntu18.04 nvidia-smi
## those commands should result in a console output shown below:
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 510.60.02    Driver Version: 510.60.02    CUDA Version: 11.6     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  Tesla P4            Off  | 00000000:0B:00.0 Off |                    0 |
| N/A   62C    P0    55W /  75W |   4351MiB /  7680MiB |     94%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+
+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
|    0   N/A  N/A      8132      C                                    4349MiB |
+-----------------------------------------------------------------------------+
  1. Hardware Suggestions
  • NVIDIA GeForce RTX 2080 Ti or higher is recommended.

  • The maximum CUDA version supported by the host >= 11.2

2.2. Installation of YMIR-GUI

The user must ensure that all the conditions in Cuda environment dependencies have been met, otherwise the installation may fail.

The YMIR-GUI project package is on DockerHub and the steps to install and deploy YMIR are as follows:

  1. Clone the deployment project YMIR to the local server:
git clone https://github.com/IndustryEssentials/ymir.git
  1. If there is no available GPU and you need to install CPU mode, please change it to CPU boot mode by modifying the .env file to change the SERVER_RUNTIME parameter to runc:

# nvidia for gpu, runc for cpu.

SERVER_RUNTIME=runc

  1. If you do not need to use the label free labeling platform, you can directly execute the start command with the default configuration: bash ymir.sh start.It is recommended not to use the sudo command, otherwise it may cause insufficient privileges.
  • When the service starts, it asks the user if they want to send usage reports to the YMIR development team, the default is yes if you do not enter it.
  • The default port number for YMIR's Model Deployment module is 18801. If there is a conflict that needs to be modified, you need to go to the YMIR directory and modify the .env file to configure the ModelDeployment port and MySQL access password:
DEPLOY_MODULE_HOST_PORT=18801
DEPLOY_MODULE_URL=${DEPLOY_MODULE_HOST_PORT}
DEPLOY_MODULE_MYSQL_ROOT_PASSWORD=deploy_db_passwd

Execute the start command after the modification: bash ymir.sh start.

  1. After the service successfully started, YMIR will be available at http://localhost:12001/. If you need to stop the service, run the command: bash ymir.sh stop

  2. The default initial user is super administrator, you can check account and password through the .env file under the project path and modify it before deployment. It is recommended to change the password through the user management interface after the service deployment is completed.

 
  1. Enter the .env file configuration, set the outgoing mailbox information, and only after the configuration is complete can you send email notifications.
# Email Notification
EMAILS_ENABLED=True
FRONTEND_ENTRYPOINT=<YMIR FRONTEND URL>
SMTP_TLS=
SMTP_PORT=
SMTP_HOST=
SMTP_USER=
SMTP_PASSWORD=
EMAILS_FROM_EMAIL= <SENDER EMAIL ADDRESS>
EMAILS_FROM_NAME=ymir-project
EMAIL_RESET_TOKEN_EXPIRE_HOURS=1
EMAIL_TEMPLATES_DIR=/app/email-templates/build

3. Use YMIR-GUI: typical model production process

 
 

As shown in the figure, YMIR divides the model development process into multiple steps. Details about how to run each step are listed in the subsequent sections.

Data and labels are necessary for the training of deep learning, and the training requires a large amount of data with labels. However, what exists in reality is a large amount of unlabeled data, which is too costly in terms of labor and time if all of them are manually labeled.

Therefore, YMIR platform, through active learning, first attains an initial model by local import or a small amount of labeled data, and uses this initial model to mine the most beneficial data for model capability improvement from a large amount of data. After the mining is completed, only this part of the data is labeled and the original training dataset is expanded efficiently.

The updated dataset is used to train the model again to improve the model capability. The YMIR platform provides a more efficient approach than labeling the entire data and then training it, reducing the cost of labeling low-quality data. Through the cycle of mining, labeling, and training, high quality data is expanded and the model capability is improved.

This section uses a complete model iteration process as an example to illustrate how to use the YMIR platform. Please check Operating Instructions.

4. For advanced users: YMIR-CMD (command line) user's guide

This chapter contains the instructions for the YMIR-CMD. If you need to use the GUI, please refer to Ymir-GUI Installation.

4.1 Installation

Mode I. Pip Installation

# Requires >= Python3.8.10
$ pip install ymir-cmd
$ mir --vesion

Mode II. Installation from the source

$ git clone --recursive https://github.com/IndustryEssentials/ymir.git
$ cd ymir/ymir/command
$ python setup.py clean --all install
$ mir --version

4.2 Typical model production process

process-en

The above figure shows a typical process of model training: 1) the user prepares external data, 2) imports it into the system, 3) appropriately filters the data, and 4) begins training to obtain a model (possibly with low accuracy). 5) selects images in a dataset to be mined that are suitable for further training based on this model, 6) annotates these images, 7) merges the annotated results with the original training set, and 8) uses the merged results to run the training process again to obtain a better model. This section implement the process shown above using the command line. For details, please check the CMD usage instructions.

5. Get the code

5.1. Code contribution

Any code in the YMIR repo should follow the coding standards and will be checked in the CI tests.

  • Functional code needs to be unit tested.

  • Use flake8 or black to format the code before committing. Both of these follow the PEP8 and Google Python Style style guides.

  • mypy - Python must go through static type checking.

Also check out MSFT Encoding Style for more advice.

5.2. About training, inference, mining docker images and model package structure

Check this document for training, inference and mining details.

Check this document for model package structure details.

6. Design concept

We use the concept of code version control in Git to manage our data and models, use the concept of branches to create new projects so that different tasks on the same set of images can run in parallel. The additions, retrievals, updates, and deletions of datasets and basic operations are created by commits to branches. Logically, each commit stores an updated version of the dataset or new model, as well as the metadata of the operation that led to this change. Finally, only the data changes are merged into the main branch. This branch conceptually aggregates all the data annotated by many projects on the platform. Please see Life of a dataset for specific design concepts.

7. MISC

7.1. FAQ

Why did the upload of the local dataset fail?

Regardless of whether the dataset has a label file, the images folder and annotations folder must be created. The images are placed in the images folder and the format is limited to jpg, jpeg, and png. The annotation files are placed in the annotations folder and the format is the pascal (when there is no annotation file, the folder is empty). Please put the images and annotations in the same folder and compress them into a ".zip" compressed package (not a .rar compressed format).

How should I obtain training and mining configuration files?

The default profile template needs to be extracted in the mirror.

The training image youdaoyzbx/ymir-executor:ymir2.0.0-yolov5-cu111-tmi has a configuration file template located at: /img-man/training-template.yaml

Mining and inference mirrors The configuration file templates for youdaoyzbx/ymir-executor:ymir2.0.0-yolov5-cu111-tmi are located at: /img-man/mining-template.yaml (mining) and /img-man/infer-template. yaml (infer).

How can the trained model be used outside the system?

After successful training, the system will output the ID of the model. The user can find the corresponding file according to this id at --model-location. In fact, it is a tar file that can be extracted directly using the tar command to get the "mxnet" model file in parameters and JSON format.

How to solve the deployment, debugging and operation problems encountered in windows system?

It has not been fully tested on Windows server, so we cannot provide service support for the time being.

How to import models I've already trained?

See this document.

All Contributors