MLOps

This blog aims to provide a basic understanding of GPUs for ML/MLOps Engineers, without going into extensive technical details. The focus will be on developing a broad knowledge of GPUs to aid in hardware decision-making for managing ML/DL workloads and pipelines. This blog will not include comparisons of different GPU architectures for ML workloads. Before we begin, let’s review some key terminologies that will be used throughout this blog: Latency: The time delay between the initiation of a task and the detection of its effects....

Kaldi is a very popular Automatic Speech Recognition (ASR) toolkit. In my experience, setting up Kaldi on a system can take huge amount of time. Therefore, it is best to use containerization approach. Following is the Dockerfile that I wrote. We will be using a cuda enabled Ubuntu 18.04 image from Nvidia NGC: FROM nvidia/cuda:11.1.1-cudnn8-devel-ubuntu18.04 LABEL Author="umerjamil16@github.io" Email="umer1694@gmail.com" LABEL Description="Kaldi ASR Image" Vendor="x" Version="1.0" RUN sh -c 'echo "APT { Get { AllowUnauthenticated \"1\"; }; };" > /etc/apt/apt....