yaac 0.1.4

YAAC - Python package for loading and using trained AI models

yaac

YAAC - Python package for loading and train AI models.

Installation

Install from PyPI:

pip install yaac

Requirements

HuggingFace Token (for DINOv3 ConvNeXt-Tiny backbone)

If you're loading models that use the DINOv3 ConvNeXt-Tiny backbone (convnext_tiny_dinov3), you'll need a HuggingFace token because the model repository is gated.

1. Request access: Visit https://huggingface.co/facebook/dinov3-convnext-tiny-pretrain-lvd1689m and request access to the repository
2. Generate a token: Create a token at https://huggingface.co/settings/tokens
3. Set environment variable: Export the token as an environment variable:

   export HUGGINGFACE_TOKEN=your_token_here
   

The token is only required when loading models with the ConvNeXt-Tiny backbone. Models using ResNet18 or other backbones don't require a token.

Quick Start

Load a trained model and run inference:

from yaac.common.model_loader import load_model_from_checkpoint
import torch

# Load your trained model
model, config = load_model_from_checkpoint("path/to/checkpoint", device="cuda")

# Run inference
image = torch.randn(1, 3, 224, 224)  # Your image tensor
with torch.no_grad():
    predictions = model(image)
    processed = model.postprocess(predictions)

print(f"Predictions: {processed}")

What is yaac?

yaac is a Python package that provides:

• Model Loading: Load trained image classification models from exported checkpoints (safetensors + config.json)
• Model Interface: Standardized TrainableModel interface for consistent model usage
• SIC Models: Support for Simple Image Classifier (SIC) models with configurable backbones and heads

Models are trained using YAAC's infrastructure and exported in a format compatible with this package.

Documentation

• Contributing Guide - For developers who want to contribute
• Publishing Guide - For maintainers publishing new versions

License

Apache 2.0 - See LICENSE for details.