open-gigaai/giga-models

A Comprehensive Repository for Multi-modal, Generative, and Perceptual Models

| Quick Start | Contributing | License | Citation |

🔥 Latest News

• [2025.12.08] 😊 We provide a refactored version of the Cosmos-Predict2.5 and Cosmos-Transfer2.5 training and inference code, which makes the code more concise and readable and easier to use.
• [2025.11.27] 🎉 We released GigaBrain-0, a novel VLA foundation model empowered by world model-generated data.
• [2025.11.26] 🎉 We released GigaWorld-0, a unified world model framework designed explicitly as a data engine to advance embodied AI.
• [2025.10.29] 😊 We provided a clean PyTorch reproduction of Pi0 and Pi0.5. It supports end-to-end training and inference with better performance and lightweight deployment.
• [2025.10.29] 🎉 We released GigaModels.

✨ Introduction

GigaModels is an open-source project offering an intuitive, high-performance infrastructure for a wide range of models. This comprehensive toolkit empowers users throughout the entire workflow, from training and inference to deployment and model compression.

We are dedicated to continuously integrating the latest advancements in open-source technology. Exciting updates and innovative features are always on the horizon—stay tuned!

⚡ Installation

GigaModels can be installed directly from source for the latest updates:

conda create -n giga_models python=3.11.10
conda activate giga_models
git clone https://github.com/open-gigaai/giga-models.git
cd giga-models
pip3 install -e .

🚀 Quick Start

GigaModels is designed to be very simple to use. You can easily load and utilize the model using load_pipeline or XXPipeline. Here is an example of how to use:

# Load the Grounding DINO model with load_pipeline
from PIL import Image
from giga_models import load_pipeline

image = Image.open(image_path)
pipe = load_pipeline('detection/grounding_dino/swint_ogc')
pred_boxes, pred_labels, pred_scores = pipe(image, det_labels)

# Load the Depth Anything model with DepthAnythingPipeline
from giga_models import DepthAnythingPipeline

pipe = DepthAnythingPipeline('depth-anything/Depth-Anything-V2-Large-hf').to('cuda')
depth_image = pipe(image)

More details on using GigaModels can be found in the projects folder.

🧨 Tasks & Pipelines

	Task	Pipeline	Inference	Training
VLA	VLA	GigaBrain-0 Pi0 Pi0.5	GigaBrain-0 Pi0 Pi0.5	GigaBrain-0 Pi0 Pi0.5
Diffusion	Diffusion	GigaWorld-0 Cosmos-Predict2.5 Cosmos-Transfer2.5	GigaWorld-0 Cosmos-Predict2.5 Cosmos-Transfer2.5	GigaWorld-0 Cosmos-Predict2.5 Cosmos-Transfer2.5
Vision	Depth Estimation	Depth Anything DPT Video Depth Anything	Depth Anything DPT Video Depth Anything
	Detection	Grounding DINO	Grounding DINO
	Edge Detection	Canny HED Lineart MLSD PidiNet	Canny HED Lineart MLSD PidiNet
	Frame Interpolation	Film	Film
	Image Restoration	PromptIR	PromptIR
	Keypoints	OpenPose RTMPose	OpenPose RTMPose
	Optical Flow	UniMatch	UniMatch
	Segmentation	Grounded SAM 2 Segment Anything UperNet	Grounded SAM 2 Segment Anything UperNet
	Shot Boundary Detection	TransNetV2	TransNetV2

🤝 Contributing

We welcome contributions! Please see our Contributing Guide for details.

📄 License

This project is licensed under the Apache License 2.0 - see the LICENSE file for details.

📖 Citation

@misc{gigaai2025gigamodels,
    author = {GigaAI},
    title = {GigaModels: A Comprehensive Repository for Multi-modal, Generative, and Perceptual Models},
    year = {2025},
    publisher = {GitHub},
    journal = {GitHub repository},
    howpublished = {\url{https://github.com/open-gigaai/giga-models}}
}