Lambda Labs, which is a company that provides on-demand access to GPUs in the cloud for AI, is democratizing who can train large AI models with the launch of Lambda 1-Click Clusters.
Normally, companies that offer access to GPU clusters require minimum reservations that ensure only the largest enterprises can afford them. But smaller companies rarely have a need to be paying for a year’s worth of GPUs; typically they only need a couple of weeks’ time to run their experiments, Lambda Labs explained.
1-Click Clusters have a reservation minimum of just two weeks, and offer access to multi-node clusters of 16 to 512 NVIDIA H100 Tensor Core GPUs with NVIDIA Quantum-2 InfiniBand 400 Gb/s. These clusters also come with pre-installed versions of popular AI frameworks, including PyTorch, TensorFlow, NVIDIA CUDA, NVIDIA cuDNN, and NVIDIA drivers.
“ML teams will no longer be forced to choose between flexibility and performance. No more choosing between wasting weeks of idle GPU time on a long-term contract or hoping an on-demand instance with many H100 GPUs will magically free up when it is needed most,” the company wrote in a blog post.
1-Click Clusters are available today and can be requested within a customer’s Lambda On-Demand Cloud account. The pricing is $4.49 per GPU per hour, and there are no egress or ingress fees. Storage can be paid for at $0.20 per GB per month.
