Use Insightedge 1.0.0
dist dir - temporarily saved jar file there which allows to run a TextClassification example in Zeppelin. It also contains configuration file for spark, you might need to set this properties for Zeppelin interpreter but I'm not sure.
zeppelin dir - contains zeppelin files: notebooks and etc.
runTextClassification.sh - runs text classification example as spark job.
- Clone and build forked BigDL project form insightedge branch
- Download pre-trained model.
- Download images
- Specify path ro model, images and BigDL jar in rumImageClassification.sh
- run rumImageClassification.sh. Compare with image categories here
If you wish to run the demo on an EC2 instance, you can use the private AMI ami-07a02f28095fd6f6e(Ireland) or ami-0b3b158ac47ab8833 (Ohio). The AMI is a private one that is available in Ireland and Ohio regions (can be easily copied to differnet regions). It is under our SEs AWS account.
Instructions:
- Have SSH Key and a Security Group with ports 22, 8080 and 9443 opened. If you are going to use the instance multiple times, an Elastic IP might be useful.
- Start new EC2 Instance / Spot instance(Spotinst) with the relevant AMI. It was tested with t2.2xlarge instance.
- After the instance has started, SSH into it and run
supervisorctl start bigdl
. This will start the text classification job and once it finishes (10 mins~), it starts the BigDL WebUI and the long running Job that "listens" for inputs from the BigDL WebUI.
You can monitor the progress via Spark Master UI: http://<PUBLIC_IP>:8080/ The BigDL WebUI will be available at: https://<PUBLIC_IP>:9443/
If the long running job has stopped immediatly, reboot the machine and try again.