A Spectrogram is a picture of sound. A common approach for audio classification tasks is to use spectrograms as input and simply treat the audio as an image. After several tries I finally got an optimized way to integrate the spectrogram generation pipeline into the tensorflow computational graph.

The CIFAR-10 dataset consists of 60000 32x32 colour images in 10 classes, with 6000 images per class. In this blog post I’ll cover how to export a trained CIFAR-10 tensorflow model to Tensorflow Serving.

Tensorflow Serving is the most optimized way to serve a Tensorflow machine learning model in production. The official tutorials only cover the python client, so lets implement a Go/Golang one.

Microservices become the new standard for scalable and maintanable arquitectures. Monolithic aplications are splited into smaller pieces to keep things simple and decoupled. Tracing comes in hand when you need to aggregate the events during the request lifecycle between a single or multiple services.

I’m a person that loves to learn new things, and assembly is one of the most recent topics I decide to dig into.