Storage solution provider Quobyte has announced the release of a new TensorFlow Filesystem Plug-in that will allow developers to integrate TensorFlow apps directly to Quobyte without having to go through the operating system kernel. This integration will increase GPU utilization and speed up the model training and inference stages of machine learning workflows, Quobyte explained.
According to Quobyte, the plugin can be used on Linux Systems and with Google Cloud Platform (GCP). Organizations can use the plugin to train models locally using sample data sets, and then train at scale on GCP.
The TensorFlow Filesystem Plug-in is useful at all stages of machine learning, not just the model and inference stages. Its broad set of access protocols and clients makes it well-suited for the data ingestion and preparation stages as well, according to the company.
“With our new plug-in offering AI/ML training performance is no longer limited by available storage performance, specifically throughput performance,” said Bjorn Kolbeck, CEO of Quobyte. “With the higher accuracy of results, scalability to handle bigger data sets and flexibility to run on-prem, cloud, edge, we believe we are providing the most positive customer experience.”