LightFM on AWS sagemaker | Is it possible to train model

Question

I am using LightFM for recommender system https://github.com/lyst/lightfm

Now I want to move my model to AWS Sagemaker where this is not part of build in algorithm, now I want to train my model using this algorithm and also want to leverage Sagemaker capability for huge data, I am following this link to run my custom model https://docs.aws.amazon.com/sagemaker/latest/dg/your-algorithms.html

Now, it seems I need to install docker image of this algorithm and too much other things too, is it any simple way to train my model without pre-build algorithm.

Answer 1

You'll have to put the algorithm into a docker container and bring it to SageMaker for training. You may want to check out SageMaker sample notebooks to get some examples of preparing the docker images. For example, https://github.com/awslabs/amazon-sagemaker-examples/tree/master/advanced_functionality/r_bring_your_own