Easy to Use
PyMarlin was designed to make PyTorch training as easy as possible while still getting the benefits of GPU and distributed node acceleration.
Focus on What Matters
Focus on your scenario code and data preprocessing pipeline, we'll take care of the training loop and optimizations for you.
Scale out to hundreds of GPUs with AzureML
Run the same code in all environments and just use a simple configuration change to scale from a single CPU on your dev machines to hundreds of GPU's in AzureML or other cloud services.