Automation of DL/ML model for better accuracy
We always face problem in getting better accuracy . We want our ml/dl model to have at least 90% accuracy. We keep on doing some tweaks in our model to get better accuracy and it takes lot of time to do that.
I’ve created an automation with the help of docker and Jenkins which would train the model automatically as soon as you upload your code in GitHub and if the model acquire accuracy less than 90% , it’ll automatically perform some changes in the code and train the model again until we get an accuracy more than 90%.
This is DeepLearning Dockerfile
This is my code.py
After it’s done we build our docker file by writing docker command..
docker build -t ddl .
Now as soon as the developer upload the code in GitHub Jenkins role come in play. Here we copying the files in our folder DeepLearning and this is our job1
Now JOB2 , it’ll build the image we created from Dockerfile and run it and it’ll automatically checks if it is a Machine Learning code or Deep Learning code and job1 is set as an upstream , job2 runs only if job1 is successful.
Now Job3 will check if the model accuracy is less than 95% , if yes it’ll train the model again with some tweak in the code i.e increasing the no. of epochs and hidden layer we can also do feature scaling and feature elimination.
After tweaking in code we get..
Accuracy 97% that’s great right ?
