tensormsa_old

KO | EN

gitlite — search

#react #typescript #golang #machine-learning #rust #java #nodejs #windows #cli #awesome-list #pytorch #security #openai #open-source

(★ 9)

TensorFlow micro service architecture

tensormsa_old 최신버젼 다운로드

최종 버전 다운로드 (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# TensorMSA : Tensorflow Micro Service Architecture

[![HOYA ver0.1](https://github.com/TensorMSA/TensorMSA/blob/2d99ed93c2d6fa2b61b6f3a852ec6fb8f87e7b5c/video.PNG?raw=true)](https://youtu.be/sxx9l5gWbk0 "HOYA ver0.1 - Click to Watch!")

1.TensorMSA 
 - Tensor Micro Service Architecture is a project started to make TensorFlow more accessable from Java legacy systems
 with out modifying too much source codes. 
 - We know there are AI platforms like Azure ML,Nvidia digits, AWS ML and etc, but our goal is little bit diffrent form theirs.
 - More focus on enterprise support (continuous training, neural network management, history management, user managemetn and etc) 
 - More focus on neural network (UI/UX based network configuration, work flow based management and etc) 
 
2. Function 
 - REST APIs corresponding to Tensorflow 
 - JAVA API component interface with python REST APIS
 - Easy to use UI component provide NN configuration, train remotly, save & load NN models, handling train data sets
 - Train NN models via Spark cluster supported 
 - Android mobile SDK are also part of the plan (gather data and predict) 
 
3. Schedule 
 - start project : 2016.8
 - start dev : 2016.9
 - pilot version : 2016.12
 - version 0.1 target date : 2017.4

4. Stack 
 - FE : React(ES6), SVG, D3, Pure CSS
 - BE : Django F/W, Tensorflow, PostgreSQL, Spark

5. Methodology 
 - Agile (CI, TDD, Pair programming and Cloud)

# Overview
Like described bellow, purpose of this project is provide deep learning management system via rest service so that non 
python legacy systems can use deep learning easily

 <img src="https://raw.githubusercontent.com/seungwookim/TensorMSA/master/ProjectDesc3.png" width="750"/>

# Docker(Cluster Mode)*[(Docker Hub)](https://hub.docker.com/r/tmddno1/tensormsa/)**[(usage)](http://wp.me/p7xrpI-dr)*
 - Docker Packages 
 ```python
 data-master : docker pull tmddno1/tfmsa_name_node:v2
 data-slave : docker pull tmddno1/tfmsa_data_node:v2
 tfmsa-was : docker pull tmddno1/tensormsa:v4
 CI tools : docker pull tmddno1/jenkins:v1 
 ```

- Start master/slave node container *[(set up)](http://wp.me/p7xrpI-eH)* 
 ```python
 docker run --net=host -d tmddno1/tfmsa_name_node:v2
 docker run --net=host -d tmddno1/tfmsa_name_node:v2
 ```
 - Start tfmsa-was container (dev purpose) 
 
 ```python
 docker run -it --env="DISPLAY" --env="QT_X11_NO_MITSHM=1" --volume="/tmp/.X11-unix:/tmp/.X11-unix:rw" -p 8989:8989 -p 8888:8888 -p 5432:5432 tmddno1/tensormsa:v4
 ```
 - postgres tool (dev purpose) 
 
 ```python
 pgadmin3 &
 ``` 
 - Start tfmsa-was container (dev purpose) 
 
 ```python
 docker run -it -p 8989:8989 -d --name tmddno1/tensormsa:v1
 ``` 
 - Start CI Tool 
 
 ```python
 docker run -it -p 8080:8080 -d --name tmddno1/jenkins:v1
 ``` 
 - Check servers 
 ```python
 TensorMSA : http://locahost:8989 
 Jenkins : http://locahost:8080 
 Hadoop :http://localhost:50070 
 Yarn : http://localhost:8088 
 Hbase : http://localhost:9095
 ``` 
# Docker - Settings 
 - Server information 
 ```python
 path : /home/dev/TensorMSA/TensorMSA/settings.py
 
 # custom setting need for tensormsa
 SPARK_HOST = '8b817bad1154:7077'
 SPARK_CORE = '1'
 SPARK_MEMORY = '1G'
 SPARK_WORKER_CORE = '2'
 SPARK_WORKER_MEMORY = '4G'
 HBASE_HOST = '52.78.179.14'
 HBASE_HOST_PORT = 9090
 FILE_ROOT = '/tensormsa'
 HDFS_HOST = '587ed1df9441:9000'
 HDFS_ROOT = '/tensormsa'
 HDFS_DF_ROOT = '/tensormsa/dataframe'
 HDFS_IMG_ROOT = '/tensormsa/image'
 HDFS_CONF_ROOT = '/tensormsa/config'
 HDFS_MODEL_ROOT = '/tensormsa/model'
 TRAIN_SERVER1 = '52.78.20.251:8989'
 ```

# Docker Trouble Shooting 
  *[(Base Size Trouble Shooting)](http://wp.me/p7xrpI-ep)* : if you suffer "not enough space" related error with docker

# Install*[(link)](http://hugrypiggykim.com/2016/09/03/python-tensorflow-django-%ea%b0%9c%eb%b0%9c%ed%99%98%ea%b2%bd-%ea%b5%ac%ec%b6%95-%ec%a2%85%ed%95%a9/)*

1.Install Anaconda 
 - download Anaconda : https://www.continuum.io/downloads
 - install (make sure anaconda works as default interpreter) 
 
 ```python
 bash /home/user/Downloads/Anaconda2-4.1.1-Linux-x86_64.sh
 ```
 ```python
 vi ~/.bashrc
 export PATH="$HOME/anaconda2/bin;$PATH"
 ```

2.Install Tensorflow 
 - install Tensorflow using conda 
 
 ```python
 conda create -n tensorflow python=2.7
 source activate tensorflow
 conda install -c conda-forge tensorflow
 ```

3.Install Django 
 - install Django, Django Rest Framework and Postgresql plugin
 
 ```python
 [Django]
 conda install -c anaconda django=1.9.5
 [Django Rest Frame Work]
 conda install -c ioos djangorestframework=3.3.3
 [postgress plugin]
 conda install -c anaconda psycopg2=2.6.1
 [pygments]
 conda install -c anaconda pygments=2.1.3
 ```

4.Install Postgresql

- install
 ```python
 yum install postgresql-server
 ```
 
 - check account and set pass
 ```python
 cat /etc/passwd | grep postgres
 sudo passwd postgres
 ```
 
 - check PGDATA
 ```python
 cat /var/lib/pgsql/.bash_profile
 env | grep PGDATA
 ```
 
 - init and run
 ```python
 sudo -i -u postgres
 initdb
 pg_ctl start
 ps -ef | grep postgress
 ```
 
 - connect and create database
 ```python
 # psql
 postgres=# create database tensormsa ;
 postgres=# select * from pg_database ;
 ``` 
 
 - create user for TesorMsA
 ```python
 postgres=#CREATE USER tfmsauser WITH PASSWORD '1234';
 postgres=#ALTER ROLE tfmsauser SET client_encoding TO 'utf8'; 
 postgres=#ALTER ROLE tfmsauser SET default_transaction_isolation TO 'read committed'; 
 postgres=#ALTER ROLE testuser SET imezone TO 'UTC';
 postgres=#GRANT ALL PRIVILEGES ON DATABASE tensormsa TO tfmsauser;
 ```

5.get TensorMSA form git 
 ```python
 git clone https://github.com/TensorMSA/TensorMSA.git
 ```

5.migrate database 
 - get to project folder where you can see 'manage.py'

```python
       python manage.py makemigrations 
       python manage.py migrate
   ```

6.run server 
 - run server with bellow command

```python
       ip addr | grep "inet "
       python manage.py runserver localhost:8989
   ```

# REST API / JAVA API Documents 
 - we are still on research process 
 - will be prepared on 2017

# Contributions *[(Desigin Link)](https://docs.google.com/presentation/d/105lw-nC9a37qJvKXsyBh045pGaBa7lqbCUI4V2mfjKc/pub?start=false&loop=false&delayms=3000)*
 
 <img src="https://github.com/TensorMSA/TensorMSA/blob/master/HOYA%20F_W%20Design%20Document.jpg?raw=true" width="750"/>

1. Data Base 
 - Train history data 
 - Work Flow data 
 - Data Preprocess with spark & UI
 - GPU Cluster server info
 - SSO & Authority (manager, servers, mobile users)
 - Neural Network UI/UX config data
 - odit columns 
 - schedule job info
 - plugin info
 - convert vchar field to json (case use json)
 - store file type data on postgresql
 - store raw text data
 - store dictionary (for RNN)
 - store video, audio 
 - Code based Custom Neural Net info store

2. View      
 - Intro Page : notice pops up and etc 
 - Top Menu : server management, user management, workflow, batch jobs, neural nets, plugins, etc 
 - NeuralNet Menu : steps we have now, but will be related on workflow nodes 
 - WorkFlow Menu : define extract data(ETL), preprocess, neuralnet, etc 
 - Batch jobs : time or event based Workflow waker 
 - Server Management : manage Hadoop, Hbase, Spark, Database, etc server ip & port

3. Neural Net         
 - basic : linear regression, logistic regression, clustering      
 - more nets : rnn, residual, lrcn, auto encoder

tensormsa_old

tensormsa_old 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

tensormsa_old

tensormsa_old 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

Download README