caropen / course-ds-base Goto Github PK

From course-ds-base created by mnrozhkov: iterative#7

Branch: step-3-reusable-code

Tasks:

• move reusable code to .py modules

From course-ds-base created by mnrozhkov: iterative#13

Branch: step-6-data-and-model-version-control
(continue from iterative#11)

Tasks:

• add remote storage
• store file.txt to remote storage
• delete file.txt in workspace
• restore file.txt from DVC cache with dvc checkout file.txt
• clean cache with dvc gc
• try to restore file.txt from DVC cache with dvc checkout file.txt (failure is expected)
• restore file.txt from DVC Remote Storage dvc pull (success is expected)
• update 'file.txt' (version 2) and push it to remote storage
• switch btw Git branches and 'file.txt' versions

From course-ds-base created by mnrozhkov: iterative#14

Branch: step-6-data-and-model-version-control

Tasks:

• show example of dvc list and dvc get commands

From course-ds-base created by mnrozhkov: iterative#5

Branch: step-1-organize-ml-project

Tasks:

• Build a prototype with Jupyter Notebook

From course-ds-base created by mnrozhkov: iterative#9

Branch: step-5-automate-ml-pipeline

Tasks:

• install DVC (in a virtual environment)
• automate ML pipeline with DVC
• setup dependencies and outputs
• setup parameters

Requirements:

• for each stage, explicitly define dependencies (code and data)

From course-ds-base created by mnrozhkov: iterative#17

Branch: step-9-studio

Tasks:

• connect a GitHub repo
• walk through comparing metrics, params, plots, and running new experiments

From course-ds-base created by mnrozhkov: iterative#6

Branch: step-2-create-config-file

Tasks:

• create params.yaml file
• move configs into params.yaml

From course-ds-base created by mnrozhkov: iterative#2

Branch: step-1-organize-ml-project
Tasks:

• checkout to the branch: step-1-organize-ml-project
• create the project structure
• commit & push to remote

Check yourself:

• branch: step-1-organize-ml-project_SOLUTION

From course-ds-base created by mnrozhkov: iterative#16

Branch: step-7-metrics-and-experiments

Tasks:

• save 'reports/classes.csv` file
• get confusion matrix plot withdvc plots show reports/classes.csv --template confusion --template confusion
• dvc plots show
• dvc plots diff

From course-ds-base created by mnrozhkov: iterative#11

Branch: step-6-data-and-model-version-control

Tasks:

• create a file.txt
• add file.txt to version control with DVC
• create a datadir and add it to version control with DVC

From course-ds-base created by mnrozhkov: iterative#19

Branch: step-9-experimenting-workflow

Tasks:
Experiment 1

• run a new experiment with updated parameter:
  dvc exp run -S train.cv=2
• show metrics with CLI:
  dvc exp show
• apply to the current branch / commit & push to remote /
  dvc exp apply <exp>
git add . & git commit -m "Experiment 1: train.cv=2"

Experiment 2

• run a new experiment
  dvc exp run -S train.estimator_name=svm
• create a branch / commit / push to remote / check updates in Studio
  dvc exp branch <exp> experiment-2
git add . & git commit -m "Experiment 2: SVM"
git push origin experiment-2
• merge to step-9-experimenting / check updates in Studio

From course-ds-base created by mnrozhkov: iterative#1

Branch: main

Task:

• Fork & clone repo
• create a branch: git_basics
• checkout, create a file, commit & push to remote
• Checkout diffs for main VS. git_basics
• Create a PR and merge to main (on remote)

From course-ds-base created by mnrozhkov: iterative#4

Branch: step-1-organize-ml-project
Tasks:

• Show README, typing & docstring
• Use GitHub issues for task tracking

From course-ds-base created by mnrozhkov: iterative#15

Branch: step-7-metrics-and-experiments

Tasks:

• save 'reports/metrics.json` file
• specify metrics in dvc.yaml
• dvc metrics show
• dvc metrics diff

From course-ds-base created by mnrozhkov: iterative#3

Branch: step-1-organize-ml-project
Tasks:

• create venv
• install python dependencies
• activate the environment and run Jupyter Notebook

From course-ds-base created by mnrozhkov: iterative#12

Branch: step-6-data-and-model-version-control
(continue from iterative#11)

Tasks:

• add remote storage
• push data and models controlled by DVC to remote storage with dvc push
• delete files in workspace
• restore siles from DVC Remote Storage dvc pull

From course-ds-base created by mnrozhkov: iterative#10

Branch: step-5-automate-ml-pipeline_SOLUTION

Tasks:

• run dvc repro (assume we do this for 1st time)
• run dvc repro (no changes, DVC skips all stages)
• run dvc repro -f (force running ML pipeline)
• run dvc repro train (run only train stage from dvc.yaml)
• update train stage configs and run dvc repro
• update src/stage/train.py code and run dvc repro
• remove reports/metrics.json and run dvc repro

From course-ds-base created by mnrozhkov: iterative#8

Branch: step-4-build-ml-pipeline

Tasks:

• create src/stages directory

• create .py modules for each pipeline stage:

  • data_load.py
  • data_split.py
  • featurize.py
  • train.py
  • evaluate.py

• run each stage

Requirements:

• use params.yaml to manage stages configuration