Deepstributed is an online app. It makes
unimaginably easy to cope with three ML processes that usually make life harder.
At first we help you to connect your GPUs to our app. Yes, you can manage your experiments from Deepstributed
even on your own computing resources. And then, if you want to run more, you can reach for the community GPUs
or GPU-as-a-Service resources.
Then we let you specify which resources and preconfigured runtime frameworks you want to use
for a certain experiment. PyTorch, Tensorflow, Caffe? Yes, we have them all. And many more.
Finally, you upload your code, data and run the experiment on the cheapest resources possible
and no configuration effort. Nice, huh?
Join our beta tests
Have you ever had to wait for your resources to finish other tasks? Optimize the usage of your GPUs by queuing, prioritizing and automating deployment of your ML/DL experiments. With Deepstributed you just drop your code in a queue and wait for notification that your computations are finished.
We know how frustrating and time-consuming the continuesly and repeating preparation and configuration of your infrastructure for ML/DL experiments might be. Leave it to us. Deploy your Deepstributed experiments with a single click so you could focus on what you are the best in – perfecting your DL models and polishing your cutting-edge AI solutions.
Your local compute resources are not sufficient to handle you ML/DL experiments backlog? With Deepstributed you can extend your pool of resources on the fly with almost no administrative effort. You just point the resources to be used. And it does not matter if these are your own local/remote resources nor if they come from your favorite GPU-as-a-Service infrastructure provider.
Have you ever paid for hired resources when they were idle? Forget about it. Just add resources from one or more GPU-as-a-Service providers to your computing pool. Deepstributed will not only choose the cheapest resources currently available but also shut down for you machines once they finish computations.
Is there a time when your GPUs are idle? Make them earning real money for you. Just set the price and availability periods. Deepstributed will make them available for all our users according to your personally defined sharing schedule. GPUs shared pool will be also available for your disposal.
If you work as a research group or a small company you can utilize our Team workspaces functionality, where you can share your computing resources, machine learning experiments and experiment data in a secure and isolated environment. What is more, our flexible workspace management system allows you to share resources and data between various workspaces to help you organize your work if you are working on a larger number of projects.
Deepstributed has been tested on another testbed infrastructure. This time we have used to Fed4FIRE+ testbed for evaluation of our distributed task deployment mechanisms in various scenarios, also involving heavy network and computing resources loads generated by dedicated testbed tools. All of the tests have been successful and our Platform have proven to be reliable and efficient even in such circumstances. The results of this experimentation show that Deepstributed can enter the next stage of development and should be ready to enter the market in a not so far future.
Deepstributed platfrom was successfully tested on FIWARE Lab infrastructure (www.fiware.org). We can use FIWARE machines also as a host for platform and as a shared resources. Integration with FIWARE Community allows their clients building research environment based on Deepstributed.
We’ve reached the first great milestone in development of our community. Our Deepstributed community have reached 100 developers and the numbers are still increasing. Such development indicates that we are on the right track and motivates us to provide you with new features and solutions in our Platform. We would like to thank all of you for supporting us in our journey in provisioning of smart infrastructure supporting development of cutting age AI solutions.
Knowledge, passion and determination drives us to action. We are the ones who believe in what we do. We know problems and challenges of research environment, because we are a part of it. Meet us, than you won’t need to trust or believe us, you will know that we are your solution.
Deepstributed provides fully-functional ready-to-use environments prepared for selected deep learning libraries. All popular frameworks are supported, including TensorFlow, PyTorch, MXNet and Caffe. All you need to do is select your favorite frameworks and libraries and point the location of your code and training data.