Quickstart¶

1. Using the library¶

To use the library, the simplest way is to install it from the packages available:

on conda-forge for Linux, macOS, and Windows:
```
conda install -c conda-forge plaid
```
on PyPI for Linux:
```
pip install pyplaid
```
on Spack for Linux, macOS, and Windows:
```
spack install py-plaid
```

Installation notes

Conda-forge packages for Linux, macOS, and Windows, as well as the Linux PyPI package, include a bundled pyCGNS dependency. Non-Linux PyPI installations require a separate pyCGNS installation and are untested.
A Spack package recipe is available for Linux, macOS, and Windows, but has only been tested on Linux.
On Apple Silicon, users can force an osx-64 conda environment with CONDA_SUBDIR=osx-64 to install the existing macOS-64 builds under Rosetta.

2. Core concepts¶

The main classes of plaid are Sample, ProblemDefinition and Infos. There is no public top-level dataset container class. Datasets are accessed through backend-specific collections and Converter objects, so users typically materialize individual Sample objects only when needed instead of loading a whole dataset as one in-memory PLAID object.

The main use case is efficient, backend-agnostic access to heterogeneous PLAID samples in ML pipelines. The reading API is therefore centered on backend datasets and converters:

from plaid.storage import init_from_disk

datasetdict, converterdict = init_from_disk(local_folder)

dataset = datasetdict[split]
converter = converterdict[split]

for i in range(len(dataset)):
    plaid_sample = converter.to_plaid(dataset, i)

Sample features can then be retrieved as follows:

from plaid.storage import load_problem_definitions_from_disk

pb_defs = load_problem_definitions_from_disk(local_folder)
pb_def = pb_defs["my_pb_def"]

plaid_sample = ... # see above to instantiate a plaid sample

for t in plaid_sample.get_all_time_values():
    for path in pb_def.input_features:
        feature = plaid_sample.get_feature_by_path(path=path, time=t)
        ...
    for path in pb_def.output_features:
        feature = plaid_sample.get_feature_by_path(path=path, time=t)
        ...

These instructions are valid regardless of the storage backend or the heterogeneity of the data.

The package also ships two command-line tools:

plaid-check /path/to/plaid_dataset
plaid-viewer --datasets-root /path/to/datasets

These concepts are described in more detail in Concepts, including efficient dataset generation and writing.

3. Going further¶

Explore Examples & Tutorials for practical use cases and advanced techniques.

The API Reference provides detailed information on all available classes and methods.