Testing conducted by Lambda in March 2022 using a production Tensorbook, 16-inch MacBook Pro system with Apple M1 Max with a 32-core GPU and 64GB RAM, Google Colab instance running a K80 GPU, and Google Colab+ instance running on a P100 GPU. Tensorbook running Ubuntu 20.04.4, TensorFlow 2.7.0, Nvidia driver r510, CUDA 11.1.
MacBook Pro running MacOS 12.2, TensorFlow 2.7.0. Google Colab instances running Ubuntu 18.04.5, TensorFlow 2.8.0, Nvidia driver r460, CUDA 11.2. Performance tests are conducted using specific computer systems and reflect the approximate performance of Tensorbook.
Throughput benchmark of ResNet50v2 on a 224x224 input with any batch size, allowing for XLA and mixed precision datatypes.
Throughput benchmark of distilbert-base-uncased on a 128 length input with any batch size, allowing for XLA and mixed precision datatypes.
