Scaletorch.ai

At Scaletorch we feel that lack of scalable computing power is what holds back deep learning engineers and AI researchers from breakthrough work in applied or pure AI. The core of Scaletorch is DAP (Disaggregated Asynchronous Processing Engine), an engine that relies on asynchronous and disaggregated execution of Pytorch training workloads. This results in training running 4x-30x faster on the same number of GPUs. DAP requires no changes to your Pytorch script. Just launch a Pytorch training job with DAP enabled and see you training time and costs drop like a rock in water.