I'm seeing an error that gets thrown if a flow is solved with conda and the corresponding local environment is removed. Steps to reproduce with the following flow `helloconda.py`:

from metaflow import FlowSpec, step, conda_base

libraries = {
    "numpy": "1.24.3",
}
@conda_base(libraries=libraries, python="3.10.9")
class HelloConda(FlowSpec):

    @step
    def start(self):
        import numpy
        print(numpy.__version__)

        self.next(self.end)
    @step
    def end(self):
        print("HelloConda is all done.")


if __name__ == "__main__":
    HelloConda()

• Run with

python helloconda.py --environment=conda run

successfully • Run

conda env list

and delete the recently created conda environment with

conda env remove -n <env_name>

• The error

Step: start, Error: CondaValueError: cannot mix specifications with conda package filenames

will now be thrown This is while using S3 as a datastore. Any thoughts?

are you using the conda implementation from the extensions or from metaflow?

From metaflow

what is the version of conda/mamba that you are using?

@conda

is supposed to account for this scenario - it could be regression with the return codes in the

conda

binary that our implementation should account for.

This is happening with version 2.9.2