Spell Blog
MLOps
A brief history of learning rate schedulers and adaptive optimizers
April 16th, 2021
Learning rate schedulers and optimizers: a brief history
MLOps
Training larger-than-memory PyTorch models using gradient checkpointing
April 6th, 2021
Gradient checkpointing is a key technique for training large models on GPU
Company
Spell Joins the AI Infrastructure Alliance as a Core Member
March 16th, 2021
Spell partners with the AIIA, joining the inaugural industry-wide effort to build an AI Canonical Stack
MLOps
More effective Tensorflow 2: Best practices and design patterns
February 19th, 2021
We’ll walk through solutions to common problems you might run into when using TF on non-trivial projects and discuss best practices for optimal productivity.
Product Updates
ML observability with Spell and Arize
February 9th, 2021
Announcing our new partnership with Arize, bringing sophisticated model observability tools to Spell model servers.
MLOps
Speeding up model training with PyTorch JIT
February 3rd, 2021
PyTorch JIT can be used to enable 2x-3x speedups on custom module code.
Ready to Get Started?
Create an account in minutes or connect with our team to learn how Spell can accelerate your business.