dmqclib 0.6.2

This package aims to offer helper functions that simplify model building and evaluation

dmqclib

dmqclib is a Python library that provides a configuration-driven workflow for machine learning, simplifying dataset preparation, model training, and data classification. It is a core component of the AIQC project that aims to enhance anomaly detection in CTD (Conductivity, Temperature, Depth) data.

Installation

The package is available on PyPI and conda-forge.

Using pip:

pip install dmqclib

Using conda:

conda install -c conda-forge dmqclib

Core Concepts

The library is designed around a three-stage workflow:

1. Dataset Preparation: Prepare feature datasets from raw data and generate training, validation, and test data sets.
2. Training & Evaluation: Train machine learning models and evaluate their performance using cross-validation.
3. Classification: Apply a trained model to classify new, unseen data.

Each stage is controlled by a YAML configuration file, allowing you to define and reproduce your entire workflow with ease.

Usage

The general workflow for any task in dmqclib follows these steps:

1. Generate a Configuration Template: Create a starter YAML file for the task (e.g., prepare, train, classify).
2. Customize the Configuration: Edit the YAML file to specify paths, dataset names, and other parameters.
3. Run the Task: Load the configuration and execute the main function for the task.

1. Dataset Preparation

This workflow processes your input data and creates training, validation, and test sets.

Step 1: Generate a configuration template.

import dmqclib as dm

# This creates 'prepare_config.yaml' with predefined sections
dm.write_config_template(file_name="/path/to/prepare_config.yaml", stage="prepare")

Step 2: Customize prepare_config.yaml. You must edit the file to set the correct input/output paths and define your dataset. See the Configuration section for details.

Step 3: Run the preparation process.

import dmqclib as dm

config = dm.read_config("/path/to/prepare_config.yaml")
dm.create_training_dataset(config)

This generates the following output folders:

• summary: Statistics of input data used for normalization.
• select: Profiles with bad observation flags (positive samples) and good profiles (negative samples).
• locate: Observation records for both positive and negative profiles.
• extract: Features extracted from the observation records.
• training: The final training, validation, and test datasets.

2. Model Training and Evaluation

This workflow uses the prepared dataset to train a model and evaluate its performance.

Step 1: Generate a training configuration template.

import dmqclib as dm

dm.write_config_template(file_name="/path/to/training_config.yaml", stage="train")

Step 2: Customize training_config.yaml. Edit the file to point to your prepared dataset and define training parameters.

Step 3: Train and evaluate the model.

import dmqclib as dm

config = dm.read_config("/path/to/training_config.yaml")
dm.train_and_evaluate(config)

This generates the following output folders:

• validate: Results from the cross-validation process.
• build: The final trained models and their evaluation results on the test dataset.

3. Data Classification

This workflow applies a trained model to classify all observations in a dataset.

Step 1: Generate a classification configuration template.

import dmqclib as dm

dm.write_config_template(file_name="/path/to/classification_config.yaml", stage="classify")

Step 2: Customize classification_config.yaml. Edit the file to point to the input data and the trained model.

Step 3: Run classification.

import dmqclib as dm

config = dm.read_config("/path/to/classification_config.yaml")
dm.classify_dataset(config)

This workflow processes a dataset using a trained model and generates:

• classify: The final classification results and a summary report.

Configuration

Configuration is managed via YAML files. The write_config_template function provides a starting point that you must customize for each module.

1. Dataset Preparation (stage="prepare")

The preparation config requires you to modify two key sections:

• path_info_sets: Defines the location of input and output data.

  path_info_sets:
    - name: data_set_1
      common:
        base_path: /path/to/data # EDIT: Root output directory
      input:
        base_path: /path/to/input # EDIT: Directory with input files
        step_folder_name: ""
      split:
        step_folder_name: training
  

• data_sets: Defines a specific dataset to be processed.

  data_sets:
    - name: dataset_0001  # EDIT: Your data set name
      dataset_folder_name: dataset_0001  # EDIT: Your output folder
      input_file_name: nrt_cora_bo_4.parquet # EDIT: Your input filename
  

2. Training and Evaluation (stage="train")

The training config links the prepared data to the model training process.

• path_info_sets: Defines where to find the prepared dataset and where to save model artifacts.

  path_info_sets:
    - name: data_set_1
      common:
        base_path: /path/to/data # EDIT: Root output directory
      input:
        step_folder_name: training
  

• training_sets: Links to a dataset prepared in the previous workflow.

  training_sets:
    - name: training_0001  # EDIT: Your training name
      dataset_folder_name: dataset_0001  # EDIT: Your output folder
  

3. Classification (stage="classify")

The classification config uses a trained model to classify new data.

• path_info_sets: Defines paths for raw data, models, and classification results.

  path_info_sets:
    - name: data_set_1
      common:
        base_path: /path/to/data # EDIT: Root output directory
      input:
        base_path: /path/to/input # EDIT: Directory with input files
        step_folder_name: ""
      model:
        base_path: /path/to/model  # EDIT: Directory with model files
        step_folder_name: model
      concat:
        step_folder_name: classification # EDIT: Directory with classification results
  

• classification_sets: Defines a specific dataset to be classified.

  classification_sets:
    - name: classification_0001  # EDIT: Your classification name
      dataset_folder_name: dataset_0001  # EDIT: Your output folder
      input_file_name: nrt_cora_bo_4.parquet   # EDIT: Your input filename
  

Contributing & Development

We welcome contributions! Please use the following guidelines for development.

Environment Setup

We recommend using uv + mamba for managing the development environment.

1. Install python and uv (e.g., via conda or mamba):

   # Using mamba (recommended)
   mamba create -n uvenv -c conda-forge python=3.12 uv pip
   mamba activate uvenv
   

2. Navigate to the project root and create the virtual environment:

   cd /path/to/dmqclib
   uv sync
   

3. Activate the virtual environment:

   source .venv/bin/activate
   

4. Install the library in editable mode:

   uv pip install -e .
   

Running Tests

Unit tests are run with pytest.

run pytest -v

Code Style (Linting & Formatting)

We use Ruff for linting and formatting.

Linting:

# Lint the library source code
ruff check src

# Lint the test code
ruff check tests

Formatting:

# Format the library source code
ruff format src

# Format the test code
ruff format tests

Documentation (for Maintainers)

Project documentation is hosted on Read the Docs.

Building Docs Locally

1. Update Docstrings (Requires Google Gemini API Key):

   # Update docstrings for source files
   python ./docs/scripts/update_docstrings.py src docs/scripts/prompt_main.txt
   
   # Update docstrings for test files
   python ./docs/scripts/update_docstrings.py tests docs/scripts/prompt_unittest.txt
   

2. Review Docstrings: Manually review all modified files. Remove generated headers/footers and correct any sections marked with "Issues:".

3. Update API Documents: From the project root, run:

   uv run sphinx-apidoc -f --remove-old --module-first -o docs/source/api src/dmqclib
   

4. Build HTML: From the project root, run:

   cd docs
   uv run make html
   cd ..
   

   You can view the generated site by opening docs/build/html/index.html in a browser.

Deployment (for Maintainers)

PyPI

The package is published to PyPI automatically via a GitHub Action whenever a new release is created on GitHub.

conda-forge (Manual)

Bump version

Updating the package on conda-forge involves creating a pull request to the conda-forge/dmqclib-feedstock repository.

1. Fork and clone the dmqclib-feedstock repository.
2. Sync with upstream (e.g., add conda-forge/dmqclib-feedstock as a remote named upstream and git rebase upstream/main).
3. Update the forked repo:

   git checkout main                      # Go to your local main branch
   git fetch upstream                     # Get latest changes from original repo
   git rebase upstream/main               # Make your local main perfectly linear with original
   git push origin main --force           # Update your GitHub fork's main (optional but good practice)
   

4. Create a new branch (e.g., git checkout -b update_vX.Y.Z).
5. Generate a strict recipe (e.g., grayskull pypi dmqclib --strict-conda-forge).
6. Review recipes/meta.yaml and ensure it meets conda-forge standards.
7. Rerender the feedstock (e.g., conda smithy rerender -c auto).
8. Commit, push, and open a pull request to the staged-recipes repository.
9. Merge it after passing CI.

Initial upload

Submitting the package on conda-forge involves creating a pull request to the conda-forge/staged-recipes repository.

1. Fork and clone the staged-recipes repository.
2. Create a new branch (e.g., git checkout -b dmqclib-recipe).
3. Generate a strict recipe: grayskull pypi dmqclib --strict-conda-forge.
4. Review recipes/dmqclib/meta.yaml and ensure it meets conda-forge standards.
5. Commit, push, and open a pull request to the staged-recipes repository.

Anaconda.org (Manual)

Publishing to the <username> channel on Anaconda.org is a manual process.

1. Install build tools:

   mamba install -c conda-forge conda-build anaconda-client grayskull
   

2. Generate Recipe: From the project root, run grayskull pypi dmqclib. This creates dmqclib/meta.yaml.

3. Build Package: conda build dmqclib

4. Upload Package:

   anaconda login
   anaconda upload /path/to/your/conda-bld/noarch/dmqclib-*.conda
   

5. Cleanup: Copy dmqclib/meta.yaml to conda/meta.yaml for version control and remove the temporary dmqclib directory.