456 points by ml_hacker 6 months ago flag hide 8 comments
tarunpabs 6 months ago next
Love seeing GraphQL and TensorFlow used together. I've been using this combo for some side projects and it's been great. I think it could be a game-changer for ML performance.
gkamradt 6 months ago next
Totally agreed! I've used this combination in a production environment, and it significantly sped up our ML processes. It also made it easier to fetch and manage data.
eddyc 6 months ago next
I've worked on a similar project, and I can confirm that TensorFlow and GraphQL can significantly improve performance. However, be careful with batching and resolving – it can be tricky sometimes.
hmason 6 months ago prev next
I haven't tried using GraphQL with TensorFlow for ML, but I'm interested in learning more. Can you share any resources or articles on how to get started?
tarunpabs 6 months ago next
@hmason, starting with this guide is a good idea: <https://graphql.org/learn/best-practices/#machine-learning-and-graphql>. Then, I'd suggest looking at some TensorFlow and GraphQL wrappers, like this one: <https://github.com/hasura/graphql-engine/blob/master/community/learn-hasura/graphql-apis-for-machine-learning/graphql-tensorflow.md>.
guest 6 months ago prev next
Has anyone had any issues with API compatibility and the rate of new releases in GraphQL and TensorFlow? I'm concerned it could lead to compatibility issues over time.
vigoda 6 months ago next
@guest, while there can be compatibility issues, most major frameworks provide ways to manage these problems. For instance, Hasura's GraphQL engine has a compatibility matrix, and TensorFlow supports backward compatibility for most releases.
jlw 6 months ago next
Great to know! I think the community will grow, and I suspect compatibility issues will be better addressed as more developers get on board.