Deep Learning Studio (DLS) is a software toolkit and platform that simplifies designing, training, testing and deployment of deep learning AI models. The main features of Deep Learning Studio are the following:
This component allows for the design and training of the model without any programming. This also allows expert engineers to save time that otherwise will be spent in coding. Further editor has many innovative features like built in help and design time validation of a neural network. It is geared for making deep learning developers more efficient. DLS provides full flexibility to design arbitrary models including CNN, RNN, and LSTM. Automatic type encoding (for example, one hot encoding) is done by the software automatically based on selected data type.
DLS allows you to organize and manage all your datasets using an easy to use dashboard.
AutoML feature of Deep Learning Studio can design a starting neural network for your dataset. This can be a good starting point for more complex neural network designs.
DLS can transparently use all available GPUs on the machine it is installed on.
DLS has an optimized engine to read and feed data from your disks so that a GPU does not get under-utilized.
All models and their training results are automatically saved so you don't lose any work. DLS's built in versioning system lets you go back to previous models if required.
DLS allows deployment of built models with a single click as REST API and a simple form-based web application.
Built models can be exported as human readable PyThon code for PyTorch, Keras, MxNet, TensorFlow, CNTK or ONNX.
For coding-based projects, DLS provides pre-configured Conda development environments for almost all deep learning frameworks like TensorFlow and Keras. These environments are accessible through terminal or Jupyter Notebook.
DLS is developed by Deep Cognition and is available for both cloud and local machine. It supports (Linux and Windows) OS.
DLS is free for all developers and can be used for commercial use. You will need a free account at https://deepcognition.ai/ for using this software.
Create persistent directories where your datasets and work will be stored $ mkdir -p $HOME/dls/database $HOME/dls/keras $HOME/dls/data
Sample nvidia-docker run:
$ nvidia-docker run \ -d \ --rm \ --name deep_learning_studio \ -p 8880:80 -p 8881:80 -p 8888:8880 -p 8886:8888 -p 8889:3000 \ -v $HOME/dls/data:/data \ -v $HOME/dls/database:/home/app/database \ -v $HOME/dls/keras:/root/.keras \ -e DLS_EULA_AGREED=y \ nvcr.io/partners/deep-learning-studio:latest
-d: run container in detached mode
--name deep_learning_studio: assign name
deep_learning_studio to the container
-p 8880:80 -p 8881:80 -p 8888:8880 -p 8886:8888 -p 8889:3000 : expose required ports for the application
-v $HOME/dls/data:/data: this will be the main directory where your uploaded datasets and project will be stored.
-v $HOME/dls/database:/home/app/database: directory where DLS will store settings and configuration.
-v $HOME/dls/keras:/root/.keras: settings for Keras will be stored here.
-e DLS_EULA_AGREED=y: by selecting this option you accept DLS EULA
Once container is launched you can open http://127.0.0.1:8880/ link from your browser to access DLS.
Getting Started Videos: http://deepcognition.ai/resources/videos/