May 2023 Summaries

Lambda Demos` is a service that simplifies the process of hosting machine learning demos, allowing users to host their Gradio-built demos with just a few clicks using their existing repository URL. The platform provides access to NVIDIA GPUs for training and smaller instance types for demo hosting, catering to both machine learning practitioners and end-users who want to share their work. To use Lambda Demos, users must follow specific requirements, including building the app using Gradio, having a git repository with a README.md file containing certain properties. The service aims to streamline the process of hosting demos and provide an easy way for users to share their models with others, while also ensuring that licensing and permissions are respected.

You can host Stable Diffusion and other models with a few clicks using the new Lambda GPU Cloud Demos feature. To do this, you simply need to copy and paste the URL of the Gradio app repository into the demo source URL field in the Lambda Cloud dashboard. After setting up the demo, it will be deployed on a cloud instance in 10-15 minutes, depending on the instance type chosen. Once deployed, you can launch the demo by clicking on its name, allowing users to generate images with the model without waiting for a queue. To avoid being charged for the instance after use, it's recommended to terminate the instance when not in use. The Lambda GPU Cloud Demos feature makes it easy and convenient to host your own ML models, providing an alternative to waiting for queues or hosting models on-premises.

Lambda Cloud has deployed a fleet of NVIDIA H100 Tensor Core GPUs, offering users a powerful, cost-effective way to accelerate high-performance computing and deep learning workloads. The deployment includes 1x NVIDIA H100 PCIe GPU instances for $2.40/GPU/hr, with transparent pricing and pay-by-the-second billing. Lambda Cloud also offers several flexible deployment options, including private cloud clusters, scalable server configurations, and colocation services, allowing teams to access NVIDIA H100 GPUs depending on their unique needs.