/learns

masyahook learns

This repo contains my personal learnings around various topics in Machine Learning, CI/CD, and other tech stuff. I will be updating this repo as I learn new things.

Things I learnt

CI/CD for Machine Learning

As part of CI/CD for Machine Learning course by Weights and Biases, we implemented a few modules and workflows. The corresponding Weights and Biases project can be found here. Please checkout hw/wandb-run-base-comparison branch, and explore the corresponding PR.

  • src/train_and_eval.py - a dummy pipeline that trains and evaluates a machine learning model. Everything is logged in Weights and Biases (you can find the project here). To run the pipeline, you can use the following command:

    python -m src.train_and_eval --model_type <model_type> --tags <tags>

  • src/compare_runs.py - a script that compares the runs of a project in Weights and Biases. You can pass wandb_run_id to compare corresponding run with baseline run. To run the script, you can use the following command:

    python -m src.compare_runs <wandb_run_id>

  • .github/workflows/chatops.yaml - a GitHub Actions workflow that triggers a chatbot to notify the user about the status of the pipeline. To jobs can be triggered:

    • add-label - adds a label to the PR using the comment /label <label>.
    • compare-runs - compares the runs of a project in Weights and Biases using the comment /wandb <wandb_run_id>.