MLOPS TASK-3 (MACHINE LEARNING AND OPERATIONS)

Problems with their ideas are below:-

Job1 : Pull the Github repo automatically when some I push repo to Github.

Job2 : By looking at the code or program file, Jenkins should automatically start the respective machine learning software installed interpreter install image container to deploy code and start training( eg. If code uses CNN, then Jenkins should start the container that has already installed all the softwares required for the cnn processing).

Idea:-job2 is completed by using the bash script which will read line by line from the file and if it found “Conv2D(filters)” it will place into cnn folder and start the cnn container which have all the required lib already, else it will palce the file into ann folder rest do the same as done with cnn.

Job3 : Train your model and predict accuracy or metrics.

Idea:- This job will fetch the file name with extesnion .py and start execution with it’s respective model, and store the accuracy in cnn/ann_accuracy.txt

Job4 : if metrics accuracy is less than 80% , then tweak the machine learning model architecture.

Idea:- This job will check the ann/cnn_accuracy and is the accuracy of model is found to be < 80% then it’ll start the genric code of model-tweak of respetive model(for cnn and ann are different)

Say if the model with accuracy < 80 is of cnn that code first check the number of times it has already running (by default counter is 1) if the counter is 1 it’ll check the output shape of last pooling layer it the output shape of the last pooling layer is < 7 then it’ll tweak only by droping the learning rate and increasing the number of epochs.

But it the output shape is grater than 7 that it’ll fine tune the model and will add conv and pooling layer unitl the o/p shape of pooling layer become less then 7, after that it’ll add the 3 dense layer and the activation function and loss function is that was used by the model before tweaking.

And other things like save model at every epoch if the container stops in b/w it’ll resume thetraining right where it left.

Job5:Retrain the model or notify that the best model is being created

idea :- this job is always triggered by job4 and check if the acc of model is less < 80 then it will trigger job4 by curl command

job6: If container where app is running. fails due to any reason then this job should automatically start the container again from where the last trained model left

idea:-this job usses poll scm and check if container is running or not if not running then it start the training.

Other ScreenShots are as follows:-