Updated the Dockerfile and the script to build the image and push it to DockerHub.
In off_sample_wrapper.py added logging and forwarding ds_id and batch_id in off-sample service.
In resources.py, predict() and cpu_ram_consumption() are now run in separate processes. The last function collects metrics once per second: CPU, memory consumption, RSS by predict process. All of this data is logged for each batch in CloudWatch.
off_sample_wrapper.py
added logging and forwardingds_id
andbatch_id
in off-sample service.resources.py
,predict()
andcpu_ram_consumption()
are now run in separate processes. The last function collects metrics once per second: CPU, memory consumption, RSS by predict process. All of this data is logged for each batch in CloudWatch.