Aaron Pham

> I have not actually but can you tell me more about this compat layer? Oh just that running this on an older version of bentoml.

Can you try out the latest change on the `main` branch?

Sure, 0477d103dd454c9c5d2b0edea06ac1b3ed5c8c61

https://github.com/moby/buildkit/issues/879 😮‍💨

The team has investigated this issue and the plan is to have to fix with 1.0.8 - https://github.com/bentoml/BentoML/pull/3047 ensure that generated dockerfile fits with kaniko - https://github.com/bentoml/yatai-deployment/pull/39 will revert back...

This should be fixed in the latest release. cc @benjamintanweihao

Thanks for the mock. I will take a stab at it.

From the quick look at your mock @chlee1016, I notice that you are creating a session lazily. I believe this is the root cause of tensorflow taking a lot of...

I assume this is for docker container?

you can build it on top of the Docker GPU image we provided and work from there 😄