Siddharth Murching

@gogasca awesome, that'd be great! I'll reach out to you over open-source Slack (let me know if there's a better way) to set up time to discuss :)

@jaklan sorry for the delay!! To clarify, does this problem occur specifically with docker-based projects (e.g. when `mlflow run`-ing a project whose MLproject file contains a `docker_env` field)?

Got it, thanks for the clarification - it's surprising to me as AFAICT we basically execute a `COPY` dockerfile command to copy files into the container, so .dockerignore should be...

Apologies for the delay, thanks all for chiming in! @ambiSanthu the approach of adding dependencies based on the model's flavor makes sense. I wonder if instead of adding a `--no-conda`...

One alternative could be to copy over tags known to be absolute/workspace agnostic, e.g. tags like `mlflow.source.git.commit` should just be copied over since they should be the same regardless of...

> Since the MLflowClient constructor doesn't allow you to pass in credentials, you can't have two (or more) clients active in the same program. This is needed to make the...

This would be great btw - currently any package that depends on `mlflow_export_import` needs to do so as described in [this SO post](https://stackoverflow.com/a/54894359/12111347), which in turn prevents that package from...

Thanks @MinuraPunchihewa ! We're actively thinking about how to simplify model management across environments, we should have an update in a few weeks - certainly it may make sense to...

@Zethson thanks for filing this, just to confirm - it looks like the "docker not found" exception is raised from within your docker container (i.e. within `multistep_example:latest`). If you run...

@Zethson I see two potential solutions to the problem: 1. Introduce a --no-docker option, which will allow for running entry points without trying to create a new docker container. You...