Paperspace, a Brooklyn-based startup has launched an AI PaaS offering called Gradient. Based on the serverless delivery model, Gradient removes the friction involved in launching and configuring GPU-backed VMs to train machine learning and deep learning models.
Data scientists and developers spend a lot of time configuring the right environment needed for creating machine learning experiments and models. Firstly, they need to launch a Linux virtual machine powered by a GPU. This step is followed by installing required NVIDIA tools such as graphic drivers, CUDA runtime, and cuDNN libraries. Once the VM is configured for the GPU, they spend a significant amount of time installing libraries and frameworks such as Keras, TensorFlow, Caffe, Torch, Microsoft CNTK and Apache MXNet. With multiple dependencies and fragmented toolkits, the installation and configuration process often turns cumbersome.
Paperspace is removing the barrier to ML experimentation and model generation through Gradient. Users will never have to spin up VMs or install any software to get started. They can gain access to a fully-configured environment that’s ready to run machine learning jobs.
Read the entire article at Forbes