How to Deploy ML model to AWS Sagemaker with mlflow and Docker

How to Deploy ML model to AWS Sagemaker with mlflow and Docker - Step by step

Рет қаралды 13,464

Күн бұрын

Пікірлер: 37

@DataScienceGarage 3 жыл бұрын

Thank you for watching this video. Check the Github repo for this video. It will be updated and modified upon your request. Let me know if any! github.com/vb100/deploy-ml-mlflow-aws/

@MirrorNeuron 2 жыл бұрын

Thanks for the brilliant tutorial. Can you also please add a bonus tutorial on how to automate the prediction steps. So that whenever there is a new data it will trigger the prediction.

@TheDataScienceChannel 2 жыл бұрын

Using both mlflow and sagemaker separately in the past. The two work best when use with the other

@johnliang3786 Жыл бұрын

great and detailed tutorial for beginner like me, thanks

@DataScienceGarage Жыл бұрын

Happy that ii helped, thanks for watching!

@MirrorNeuron 2 жыл бұрын

For others while running the command "build-and-push-container" make sure your docker engine is running on your local machine.

@nibinjoseph2136 4 ай бұрын

Its getting failed for me every single time.

@krzysztofformella Жыл бұрын

Great video, really informative! Thank you. Unfortunetely outdated because of new mlflow's methods to perform deployment.

@DataScienceGarage Жыл бұрын

Thank you for such feedback! Appreciate! :)

@scar2080 2 жыл бұрын

Brilliant and clean delivery.. you are awesome mate.. Can you make another video on using mlflow + sagemaker + ( Gensim topic modelling )/FBPhrophet ?

@DataScienceGarage 2 жыл бұрын

Hi! Thanks for your feedback, really appreciate! It is a very interesting topic you're suggesting. Incorporate Gensim topic modelling into pipeline would be cool. Thanks for that, I will invest my time into this, and if will prepare something, I'll come back with tutorial here - sometime in short future. :)

@scar2080 2 жыл бұрын

@@DataScienceGarage thanks.

@pravinkumar54 3 жыл бұрын

thank you so much. much needed

@DataScienceGarage 3 жыл бұрын

Thank for such feedback, appreciate it!

@nibinjoseph2136 6 ай бұрын

Excellent!!!!

@DataScienceGarage 6 ай бұрын

Thanks for feedback! :)

@abdjanshvamdjsj 2 жыл бұрын

Great video. Just one question, when you did mlflow sagemake build-and-push, it didn't package the model in the image? The image only had the necessary environment installed. That's why in the deploy script you had to provide both the URI to model and the image separately.

@deepakmk663 2 жыл бұрын

This was the go to video for mlops using mlflow and aws. I have a quick question, is there a way to deploy pertained DL model without training job? If so please let me know

@DataScienceGarage 2 жыл бұрын

Yes, it is possible. Pre-trained model is one of the artifacts generated after training. You should put your pre-trained model, create conda.yaml file, requirements.txt file. And then push these artifacts into AWS ECR.

@deepakmk663 2 жыл бұрын

@@DataScienceGarage Thank you for your response. I tried reading model using pytorch and log model in mlflow. This is generating yaml file with default libraries.

@vishalwaghmare3130 Жыл бұрын

Very helpful video :-)

@DataScienceGarage Жыл бұрын

Thanks a lot! Glad it was useful :)

@MirrorNeuron 2 жыл бұрын

Hi there, I have few more libraries in the train.py and also have pip installed them. But it doesn't show up in the requirements.txt file. Why is that so? kindly let me know.

@DataScienceGarage 2 жыл бұрын

If I understood correctly, could you add these libraries manually in requerements.txt? You can add as many libraries as you want, depend on your actual project.

@MirrorNeuron 2 жыл бұрын

@@DataScienceGarage thank you. I just did. also can you point to a docker based deployment where the training ML model is done within sagemaker

@philtoa334 2 жыл бұрын

Very Nice thanks.

@DataScienceGarage 2 жыл бұрын

Thanks for watching!

@mdh5213 2 жыл бұрын

Hey, can you help me with something? after deploying my model when I am about to test my model it says "CancelledError: Session has been closed."

@victorgabrielsouzabarbosa5488 2 жыл бұрын

AMAZING!!!!!!

@DataScienceGarage 2 жыл бұрын

Thanks for watching! Really appreciate your feedback!

@scar2080 2 жыл бұрын

Hey Mate! Quick question? Can you tell me how to choose your compute instance for training in this situation? Lets say i want to use 64 GB Ram and 16 CPU Just a scenario Hehehehe.. how can we do it?? i see you did training offline right? can we also do it online?? Please throw me some light here

@sugammehta0301 2 жыл бұрын

How do you integrate flask api with this?

@DataScienceGarage 2 жыл бұрын

There is no Flask in this tutorial. The idea is to create a Docker image which contain model artifacts with the help of mlfflow and deploy it to AWS SageMaker through Amazon ECR.

@sugammehta0301 2 жыл бұрын

@@DataScienceGarageok great😁but would really love to have a tutorial on integration with flask soon...thanks

@DataScienceGarage 2 жыл бұрын

@@sugammehta0301 there could be couple of options. One, to use AWS Beamstalk, another - use Kubernetes to initialize Flask API. I will create this kind of tutorial one day :)

@sugammehta0301 2 жыл бұрын

@@DataScienceGarage great thank you :))

@tanb13 2 жыл бұрын

I am getting the error 'Failed to lookup host: 354565582869' while running the command mlflow sagemaker build-and-push-container. Can you please advise how to fix it?