mwoss / mlflow-stock-market-example Goto Github PK

Hi,
very nice project! I found a typo in the MLproject file (entrypoint: main) that lets the mlflow run command fail:
Typo char ["]
bucket_name" {type:string, default: stock-market-models}
Should be char [:]
bucket_name: {type:string, default: stock-market-models}

In main.py the chaching does not work for me, because in the _already_ran function there are some string, int comparisons.

I found the following:
run_value = full_run.data.params.get(param_key) if run_value != param_value:
run_value stores the value as a string, but param_value stores it as an int -> comparison does not work as expected.
I fixed it for myself with following:
if type(run_value) != type(param_value): param_value = str(param_value)

if run_info.status != RunStatus.FINISHED:
run_info.status stores 'FINISHED' as string, but RunStatus.FINISHED is stored as enum = int -> comparison does not work as expected.
I fixed it with the following:
if run_info.status != RunStatus.to_string(RunStatus.FINISHED):

After these changes the workflow worked for me and it reused the already executed steps correctly.
I ran it with python 3.7.3 I don't know if that caused the problems.