IBM today announced the launch of its new Deep Learning as a Service (DLaaS) program for AI developers.
With Deep Learning as a Service (DLaaS), developers will be able to train neural networks using popular frameworks such as TensorFlow, PyTorch and Caffe without buying and maintaining expensive hardware.
The service allows data scientists to train their models using only the resources they need, paying only for the duration of GPU usage.
Each cloud-based data processing unit is built for ease of use and is ready for programming deep learning networks without the need for users to manage the infrastructure. According to a white paper published by IBM researchers:
Users can select a set of supported deep learning frameworks, a neural network model, training data, and cost constraints. The service then takes care of the rest, providing them with an interactive, iterative AI training experience.
To use the services, users simply need to prepare their data, upload it, and start training. They can then download the training results to their app.
It seems pretty straightforward and could potentially save weeks of programming, according to TNW.
IBM is reportedly working to address the difficulty of training neural networks, or at least reduce development time. According to a blog post from the company:
This deep learning as a service is an experimental training environment, meaning users don’t have to worry about scheduling and issue management. The entire training lifecycle is managed automatically, and results can be viewed in real time and reviewed later. Each training session is automatically started, monitored, and stopped upon completion, saving users time and money by only paying for the resources they use.
The new Deep Learning as a Service (DLaaS) service is powered by the excellent Watson platform. This means it has been tested on one of the most advanced AI systems on the planet.
For more information, you can check out IBM's blog.
