A mini ML system which can train some prediction models for recommendation on a given user+item click interaction data set and automatically deploy models to be served by a simple python web API end point
-
~/recsysdemo$ chmod a+x rundocker.sh && ./rundocker.sh
-
Open http://localhost:9001/tree/notebooks in your browser to examine the notebooks
-
Open http://localhost:5000/docs in your browser to examine the prediction api.
-
If you system has CUDA + GPU configured properly you can try below command to startup
-
~/recsysdemo$ chmod a+x rundocker.sh && ./rundocker_gpu.sh
curl --location --request POST 'http://localhost:5000/predict' \
--header 'Content-Type: application/json' \
--data-raw '{
"req_id":"1242765",
"id":732810,
"user_id":28349,
"store_id":366,
"device":"app_ios",
"platform":"iOS App",
"channel":"Direct",
"created_at":"2021-02-03 23:47:27",
"num_of_items_req":5
}'