Containers for machine learning

Containers for machine learning. Contribute to replicate/cog development by creating an account on GitHub.

3 comments

I like the cog framework, I’ve used it for one small project.
How do people use the HTTP interface in production? I'm curious because I found it necessary to use a queue in production in order to handle many requests and smoothly scale up workers, yet every containerization approach I've seen uses this synchronous HTTP based approach. Are there any standard containerization methods that rely on queues instead ?