PyYel

PyYel is a personnal library that aims at helping the deployement of strong data science tools, from data handling to deep learning.

Quick start

1. Install the library.

your_path> pip install PyYel

2. Import the library into you code.

import pyl

3. Import the relevant features.

from pyl.models.LLM import LLMDecodingPhi, LLMEncodingBARTLargeMNLI
from pyl.models.CNN import CNNClassificationResNet

Content

Data

A collection of features to manipulate the data. Can be used to implement pipelines, preprocessing, data augmentation...

• Augmentations: a compilation of classes featuring methods to augment a datapoint of various type.

  • ImageAugmentation : features a handfull of functions that can augment any type of data, as well as its labels.
  • TODO

• Reduction: acompilation of classes featuring methods to reduce datapoint of various type.

  • TODO/TO-REWORK

• Utils: a collection of powerful tools that permit an easy manipulation of the datapoints.

  • TODO/TO-REWORK

Models

The neural networks implementations. These are grouped by types and tasks.

• CNN (Convolutional Neural Networks)

Note: Traditionnal computer vision networks. Features a model builder to design custom small-sized networks.

• FCN (Fully Connected Networks)

• LLM (Large Language Models)

Note: NLP transformers.

• LVM (Large Vision Models)

Note: Computer vision transformers.

Utils

A collection of higher-level tools, that simplifies the manipulation of the library

TODO/TO-REWORK

Notes

TODO