- TensorFlow 1.0. The roll out of TensorFlow 1.0 took center stage in the keynote address. This release includes new higher-level APIs, however, the big news is the performance and scalability. Google touted a 58X speedup scaling to 64 GPUs (8 node cluster with 4 K80’s each) with Inception v3.
- XLA – TensorFlow Compiled. XLA is a just-in-time (JIT) compiler or ahead-of-time (AOT) compiler for TensorFlow designed to improve speed, memory usage, and portability on server and mobile platforms. While still considered experimental, it’s worth a look.
- TensorBoard. There was considerable excitement around TensorBoard, the TensorFlow network visualizer, including the plans for Tensorflow debugger integration.
- Skin Cancer Image Classification. Skin cancer is the most common type of cancer in the US. Learn how TensorFlow is used to classify benign versus malignant lesions with accuracy comparable to a dermatologist.
While the highlights provided do not attempt to cover all topics addressed at Dev Summit, hopefully, they are sufficient to peak your interest to investigate the event replay further.