Problem Setting GPU instances for deep learning is a super expensive and daunting task. As a deep learning researcher, I used to spend a lot of time setting up the environment, fixing CUDA issues, and downloading data before I start using GPUs for model training. Some of the other issues faced are - Availability of modern GPU instances at affordable prices [Reserved Instances]. Ability to pause and resume the instance from wherever you left after a few hours or days. Flexibility to work on a single GPU for prototyping models and switch to powerful GPUs to train models faster. Ability to pause the instance programmatically. Sometimes you need approvals to use a particular kind of GPU. I hate it happening on the last day of a Kaggle competition. Most of these features are only available from major cloud providers, but they are very expensive and need special knowledge to implement. The Solution - How JarvisCloud accelerates your DL journey? Simple UI Launch optimized GPU instance in less than 30 seconds. Access instance either using JupyterLab (with HTTPS ?), SSH into the instance, or connect using VSCode. Prototype on a smaller GPU and train your model on a more powerful and multi GPU setup. Pause the instance from your code and sleep well. Pay for what you use - minute billing. Choose from a different range of GPUs including A100. Please check our platform, your feedback goes a long way, so please comment and let us know what you think!?. Our early adopters have loved the platform and hope you love it too.
