We're working on a sagemaker training flyte connector but am noticing that we can't seem to chain traditional tasks with the connector eg., have a task that creates a training dataset and then pass the s3 output to the connector. Is this by design? Looking through the docs, I'm seeing that the configurations for all the connectors on the examples are static — including the sagemaker endpoints connector, which defines separate workflows for training and creating the deployment. Any way to connect these into a single workflow?

Wdym you should be able to chain

We want a connector task to be able to accept input from the output of another task. One problem I have experienced is that if I wraps a connector task inside a Python task, it gets the

error=Cannot find agent for task category: sagemaker-training-job_v0.

error. Not sure if @brash-london-45337 is experiencing the same problem?

I think I might have figured it out — seems like I was trying to pass the task output into a config before passing it to the connector. Building up the full config in a task, returning that and then passing it to the connector seems to be working now 😅

Ya that’s what it should be - btw promises is what’s changing 😇

we are very excited to hear what's coming!