Blog

Our latest feature is built to help you make the most out of your on-prem hardware: utilize idle GPUs, adjust GPU usage for every ML job, and forget about managing priority queues. It’s live and ready for you to give it a spin (no pun intended).

Introducing an out-of-the-box solution that gives all Valohai users automatic, immutable, and secure audit logs that ensure traceability for navigating compliance requirements, debugging issues, and improving accountability within teams.

AMD's MI300X GPU can outperform Nvidia's H100 in LLM inference benchmarks, offering larger memory and higher bandwidth. Read our benchmark in full, get the details, and discover how this impacts AI hardware performance and model capabilities.

We’ve built the Model Hub to help you streamline and automate model lifecycle management. Leverage Valohai for lineage tracking, performance comparison, workflow automation, access control, regulatory compliance, and more.

Don’t miss out on Valohai’s upcoming updates on AI governance and the AI EU Act, examples of machine learning pipelines in production, new features, and GPU benchmarks. Subscribe to our newsletter.

Valohai’s new experimental feature selects compute instances based on where the data has been cached already, helping you reduce data transfer overhead and increase model iteration speed.

Our new partnership enables you to seamlessly access OVHcloud’s scalable and secure environments from the Valohai MLOps platform without changing your preferred ML workflows.

Valohai’s latest feature helps you avoid unnecessary costs by reusing the results of matching pipeline steps from previous executions. This feature is already available to all Valohai users!

We’ve built significant enhancements into our platform to further empower data science teams in accelerating time-to-market and optimizing operational costs. These enhancements tackle model iteration speed, efficient resource utilization, and dataset management.

Our new feature monitors CPU, GPU, and memory usage and alerts you when your machines operate below 50% capacity. This allows you to optimize resource usage and reduce costs.

Valohai’s Model Registry is a centralized hub for managing model lifecycle from development to production. Think of it as a single source of truth for model versions and lineage.

We designed our new Kubernetes support so that Data Science teams can effortlessly manage and scale their workflows on top of Kubernetes and enhance their overall machine-learning operations.

We're excited to announce that Valohai now supports Slurm, an open-source workload manager used in HPC environments. Valohai users can now scale their ML workflows with Slurm-based clusters with unprecedented ease and efficiency.

LLMs and other generative models make ripples everywhere from established enterprises to innovative startups, and beyond. But what did successful adoption look like in 2023? And what can we expect in 2024?

We’ve built a template for fine-tuning Mistral 7B on Valohai. Mistral is an excellent combination of size and performance, and by fine-tuning it using a technique called LoRA, we can be very cost-efficient.