[Feature Request] Integration to MLOPS System
Is your feature request related to a problem? Please describe. Sometimes we may need to record the run in Promptflow for later evaluation using other MLops systems (e.g. MLFlow/ClearML). But I found it was quite difficult to integrate. E.g. In ClearML, a global experiment object need to be maintained during the experiment. Therefore, adoption to ClearML need to pass the experiment object to each node
Describe the solution you'd like Integration supports to other MLops system. Or new MLops for promptflow?
Describe alternatives you've considered I finally integrated promptflow to ClearML, but many tricks need to be applied.
Additional context Add any other context or screenshots about the feature request here.
Hi @sparkdemotivator, "It seems like your main concern currently is the need to individually pass the experiment object to each node using flow input and output linking, which you think is a cumbersome process. It appears you might prefer a global variable setting, allowing downstream nodes to access this object without manually passing it across nodes. Is that correct?
Hi @sparkdemotivator, "It seems like your main concern currently is the need to individually pass the experiment object to each node using flow input and output linking, which you think is a cumbersome process. It appears you might prefer a global variable setting, allowing downstream nodes to access this object without manually passing it across nodes. Is that correct?
Yes. It was the pain point of integrating to my MLops system (ClearML). If you have better ideas to integrate with other opensource MLops system without using much effort in tackling those cumbersome process, it is also welcome to suggest.