Go back to the [[AI Glossary]]
#TensorFlow #GoogleCloud
A programmable linear algebra accelerator with on-chip high bandwidth memory that is optimized for machine learning workloads. Multiple TPU chips are deployed on a TPU device.