馃殌 Feature
Deployment, the thing we as ML devs always keep pushing forward can't be pushed forward no more.
Let's first discuss ideas and work on minimal examples before jumping head first on this one. I myself don't really know how deployment works, so I'll be starting with very simple apps and I will be sharing my journey here.
This is going to be a hard one, let's not do this alone, collective brain power is what we need right now, so I'll tag everyone 馃槄
@oke-aditya @paras-jain @ai-fast-track
lgvaz
All 4 comments
As we previously discussed, it's a good a idea to start with streamlit, so I'll do that
lgvaz
on 19 Jun 2020
- Add docker support.
- Publish to PyPi experimental branch.
- Add support for Amazon Container Registry and Google Container Registry.
- Try with Kubernetes as well.
- Run with KubeFlow and AirFlow.
oke-aditya
on 19 Jun 2020
Start small and build up on that. The way I see it:
1- Use streamlit
2- Deploy locally with a toy example (MantisFasterRCNN + Toy Dataset)
3- Use docker locally
4- Deploy docker container on the cloud:
5- Rinse and repeat
ai-fast-track
on 19 Jun 2020
97 should be the first big step.
oke-aditya
on 21 Jun 2020
Related issues
shimsan
路
3Comments
lgvaz
路
5Comments
lgvaz
路
3Comments
lgvaz
路
4Comments
lgvaz
路
6Comments
Most helpful comment
Start small and build up on that. The way I see it:
1- Use streamlit
2- Deploy locally with a toy example (MantisFasterRCNN + Toy Dataset)
3- Use docker locally
4- Deploy docker container on the cloud:
5- Rinse and repeat