What is usefull on developing?

I finished Production Specialization and I have question about instruments form this course for ML in development? How do you see flow of development at all?

From usefull instrument in this course I cought:

  1. ConfusionMatrix for recognizing bad input data
  2. Automatical data labeling (but it is enough rare on practice, as I understand)
  3. Anomalies detection and validation of input data
  4. Features selection (specially by importance), transformation to use raw data directly to input.
  5. Autotuner
  6. Fairness detection
  7. Shapely would be usefull too in debugging, I think

Maybe do you have anything more to add for development mode?

Hello @someone555777
Congratulations on completing this Specialization :confetti_ball:
You can check out this short course:


to learn about more advanced MLOPs tools

https://www.deeplearning.ai/short-courses/evaluating-debugging-generative-ai/

Wishing you the best!
Thank you

I’ve just briefly checked and see that the main thought of course is using of wandb that will automatically will try to analyse key things. But we used near the same features with tensorflow + google cloud in ml for prod secialisation. So, if I understand correct, we learned much more things in it like about features engeneering for example.

So, I would like to have a discussion under the instruments, that we learned in mlops specialization, if possible. It will be much more extended.

It would be nice to have a pattern of initial file, that contains all important for development piplines that contains all usefull features from Deployment mode from this course and maybe another, for example that I noted before:

And where I just can add my data with model and debug. I have access to a few courses that are connected with piplines and can’t find lab, that fully contain all usefull features for debugging. Everything is enough scattered.