External reviews
External reviews are not included in the AWS star rating for the product.
MLFLow as a great addition to our ML focused data pipelines
What do you like best about the product?
MLFlow has been instrumental in providing a decoupled interface between training and prediction part of our ML pipelines where we can store model metadata as part of training cycle and then in prediction we are able to leverage and pick model with highest ROC or just latest by chronological sorting and overall just do good job in tracking our experiments.
What do you dislike about the product?
Sometimes slow to render in databricks environment as UI but that could be related to our databricks setup.
What problems is the product solving and how is that benefiting you?
MLFlow is very helpful in ability to do concurrent run of training and prediction as prediction doesn't have to wait for training to finish, we can just pick last successful experiment from MLFlow and leverage it for predictions.
- Leave a Comment |
- Mark review as helpful
MLflow is a very useful open source tool
What do you like best about the product?
MLflow tracking has been a major advantage for keeping up the record of the results of the experiments we carry out on the data using different parameters. Tracking the results and parameters is very iseful for achieving the most optimized solution.
What do you dislike about the product?
One small counter point is that it is not an easy tool and requires all in depth knowledge for making the best use of it.
What problems is the product solving and how is that benefiting you?
I am currently utilizing the MLflow MLflow tracking utility.
Recommendations to others considering the product:
Highly recommend this tool to the users.
Great tool, helped me alot.
What do you like best about the product?
Centralization and remote servers and API
What do you dislike about the product?
Nothing specific. I like everything about mlflow.
What problems is the product solving and how is that benefiting you?
experiment tracking and ML metadata
MLflow makes ML life cycle management quite streamlined with easy implementation.
What do you like best about the product?
I like how it forces the developer to follow a certain code style which can basically help maintain the codebase much easily over time and have a proper documentation over it.
What do you dislike about the product?
I think there could be improvements within the documentation over how to use MLflow within existing codebases.
What problems is the product solving and how is that benefiting you?
My manager wanted to have a visual UI to track and monitor ml projects and metrics and also be able to import a model quickly and try it out, mlflow makes it really easy to do that.
Recommendations to others considering the product:
Read through the documentation
Mlflow is currently the most useful tool for tracking performance of machine learning models
What do you like best about the product?
It's very useful when it comes to tracking performance of machine learning models. Acts like a dashboard that would otherwise have to be built from scratch.
What do you dislike about the product?
There aren't any major downsides but the model training part according to me is still better run locally for comfortable experiments
What problems is the product solving and how is that benefiting you?
There are some generic models that are used and needs to be kept a track of the performance of the model with respect to recent data. And if the model performance is declining we retrain the model using mlflow
Great for model refresh
What do you like best about the product?
Great for model refreshes and comparison
What do you dislike about the product?
Ui and artifact loads can be improved...
What problems is the product solving and how is that benefiting you?
Model storage, reuse.
Recommendations to others considering the product:
Great for storing your artifact.
Easy and fun to use
What do you like best about the product?
I learned to datamine with Python on Databricks and I use it daily. It is a nice software, user friendly and easy to connect to multiple sources
What do you dislike about the product?
The errors can be a little more explanatory than what it is currently.
What problems is the product solving and how is that benefiting you?
Helping the client make business decisions using the purchase and engagement data on the Azure Cloud
Great experience. It helped us create/deploy custom models on sagemaker.
What do you like best about the product?
The features. There are hell lot of features present on MLFlow
What do you dislike about the product?
Haven't explored everything. But yeah, maybe better documentation.
What problems is the product solving and how is that benefiting you?
Solving platform issues of aws sagemaker. We deployed a custom model which used to power ranking algorithm on the listing page.
Best Open-Source Platform to calibrate the models with keeping tracking of the experiments and store
What do you like best about the product?
One please with all nessarery fetaure.
1. tracking
2. storage of models and files related documentation like log, config file.
3. validation of the model with meteric feature and plot crossponding to it and genrate the report from the experiments.
4. Data callibation with data 🧱, SQL and other cloud providers.
1. tracking
2. storage of models and files related documentation like log, config file.
3. validation of the model with meteric feature and plot crossponding to it and genrate the report from the experiments.
4. Data callibation with data 🧱, SQL and other cloud providers.
What do you dislike about the product?
Anything which dislike is nothing tell yet, But If we can build something like feature where we can do more advanced anlaytics crossponding to parameters and meterics and generate the different plot from that tabuler data, like we have d-tale (github: https://github.com/man-group/dtale), beacuse I was running some simulation run with differnet experiments and then write the report crossponding with the experiments and with differnet signnificant plot, demonstrates the report write up.
What problems is the product solving and how is that benefiting you?
I was running some simulation run with differnet experiments and then write the report crossponding with the experiments and with differnet signnificant plot, demonstrates the report write up.
Good for code collaboration
What do you like best about the product?
Multiple people can write code in the same file at the same time. We use Databricks in Machine Learning.
What do you dislike about the product?
The cluster gets shut down after sometimes, leading to loss of data on the RAM
What problems is the product solving and how is that benefiting you?
Code collaboration, Machine Learning
showing 361 - 370