Skip to content

README.md

Latest commit

 

History

History
585 lines (443 loc) · 24 KB

README.md

File metadata and controls

585 lines (443 loc) · 24 KB
Raw
Geti Library

A low-code transfer learning framework for training, evaluating, optimizing, and deploying computer vision models

Key Features • Supported Tasks & Models • Installation • Quick Start • Docs • License

PyPI

python pytorch openvino numpy

Codecov OpenSSF Scorecard Pre-Merge Test License Downloads

Geti Library - getitune

The Geti™ library (getitune) is a low-code transfer learning framework for Computer Vision. Its API and CLI let you train, evaluate, optimize, and deploy models quickly, even without deep expertise in deep learning. It supports diverse combinations of model architectures, learning methods, and task types built on PyTorch and the OpenVINO™ toolkit.

Each supported task ships with curated "recipes": YAML files that bundle the model, data pipeline, and training configuration into a single one-stop entry point. Recipes are validated on standard datasets so you get a strong baseline out of the box.

Key Features

  • Multi-task support: classification, object detection, instance segmentation, semantic segmentation, and keypoint detection, see the full model list below.
  • Tiling for large images across detection and segmentation tasks.
  • Multiple backends: train with PyTorch Lightning, Ultralytics YOLO, export and run inference with ONNX and OpenVINO™.
  • Hardware acceleration: Intel GPU (XPU) and NVIDIA CUDA support.
  • Datumaro data frontend with automatic format detection (COCO, YOLO, VOC, native).
  • Distributed training across multiple GPUs.
  • Mixed-precision training to reduce memory and increase batch size.
  • Class-incremental learning to extend an existing model with new classes.
  • Deployment to OpenVINO™ IR and ONNX formats, with inference via OpenVINO™ ModelAPI.

Supported Tasks & Models

All recipes live under src/getitune/recipe/<task>/. Pass any of these YAMLs directly to the API as model=.... Recipes whose name ends in _tile enable the tiling pipeline for large images.

Task Recipe directory Example recipes
Classification (multi-class / multi-label / hierarchical) multi_class_cls, multi_label_cls, h_label_cls dino_v2, vit_tiny, efficientnet_b0, efficientnet_b3, efficientnet_v2, mobilenet_v3_large
Object detection detection atss_mobilenetv2, ssd_mobilenetv2, yolox_{tiny,s,l,x}, rtdetr_50, dfine_x, deim_dfine_{l,m,x}, deimv2_{s,m,l}, rfdetr_{nano,small,medium,large}, yolo26_{n,s,m}
Instance segmentation instance_segmentation maskrcnn_{r50,swint,efficientnetb2b}, rtmdet_inst_tiny, rfdetr_seg_{nano,small,medium,large,xlarge,2xlarge}, yolo26_{n,s,m}_seg
Semantic segmentation semantic_segmentation dino_v2, litehrnet_{s,18,x}, segnext_{t,s,b} (with _tile variants)
Keypoint detection keypoint_detection rtmpose_tiny

Each task directory also ships an openvino_model.yaml recipe for running and optimizing pre-exported OpenVINO IR models via OVEngine.

Licensing Information: Ultralytics YOLO models are distributed under the AGPL-3.0 license, an OSI approved license ideal for open-source research, academic, and personal projects. For commercial use, enhanced support, and tailored licensing terms, please explore flexible Ultralytics licensing options at https://www.ultralytics.com/license.

Installation

Quick Install

# With uv (recommended)
# CPU-only by default
uv pip install "getitune"

# Or with pip
pip install "getitune"

# For hardware-specific PyTorch wheels, see "Advanced Installation: Specify Hardware Backend" below.

⚠️ Note: PyPI version doesn't support Ultralytics YOLO models. To use Ultralytics YOLO models, you must install from source.

Advanced Installation: Specify Hardware Backend

getitune ships three mutually exclusive extras that select the right PyTorch wheel for your hardware:

Extra PyTorch wheel Use when Setup Guide
[cpu] torch==2.10.0+cpu (Linux/Windows) or default torch==2.10.0 (macOS) No GPU, or running on Apple silicon. —
[xpu] torch==2.10.0+xpu + triton-xpu Intel discrete or integrated GPUs. Intel GPU drivers
[cuda] torch==2.10.0+cu128 NVIDIA GPUs with CUDA 12.8 drivers. NVIDIA CUDA Toolkit 
# Intel GPU (XPU)
uv pip install "getitune[xpu]" --extra-index-url https://download.pytorch.org/whl/xpu

# NVIDIA GPU (CUDA 12.8)
uv pip install "getitune[cuda]" --extra-index-url https://download.pytorch.org/whl/cu128

# CPU-only (no extra index needed)
uv pip install "getitune[cpu]"

macOS: PyTorch's +cpu wheel is only published for Linux and Windows. The [cpu] extra resolves this automatically and installs the default torch==2.10.0 wheel on macOS. Ultralytics YOLO models: The PyPI version doesn't include Ultralytics YOLO support. To use YOLO26 models, you must install from source.

Advanced Installation: Install from Source with Ultralytics YOLO Support 
git clone https://github.com/open-edge-platform/geti.git
cd geti/library

# Recommended: use uv to honor the lockfile
uv sync                      # CPU-only
uv sync --extra xpu          # Intel GPU (XPU) — setup: https://github.com/intel/compute-runtime/releases
uv sync --extra cuda         # NVIDIA GPU (CUDA 12.8) — setup: https://developer.nvidia.com/cuda-12-8-0-download-archive

# Or with pip in a virtual environment
python -m venv .venv && source .venv/bin/activate

# CPU-only
pip install -e ".[cpu]"

# Intel GPU (XPU)
pip install -e ".[xpu]" \
  --extra-index-url https://download.pytorch.org/whl/xpu

# NVIDIA GPU (CUDA 12.8)
pip install -e ".[cuda]" \
  --extra-index-url https://download.pytorch.org/whl/cu128

Ultralytics YOLO models: Add --extra ultralytics for uv sync or [ultralytics] for pip install:

uv sync --extra xpu --extra ultralytics  # Intel GPU + YOLO

# or with pip
pip install -e ".[xpu,ultralytics]" --extra-index-url https://download.pytorch.org/whl/xpu  #Intel GPU + YOLO

Quick Start

Discovering Recipes and Models

To explore available models and recipes:

from getitune.utils import list_models

# List all available models names
all_models = list_models()

# List all available recipes / configuration files (full YAML paths)
all_recipes = list_models(return_recipes=True)

# Filter by task
detection_models = list_models(task="DETECTION")

# Filter by pattern
efficient_models = list_models(pattern="*efficient*")

Then pass any model name to create_engine(model="...", data="...").

Note on Model Resolution:

  • If you pass a recipe YAML path (with .yaml or .yml suffix) that doesn't exist on disk, a FileNotFoundError is raised.
  • If you pass a model name that matches recipes under multiple tasks, a ValueError is raised listing the matches. Pass task= to disambiguate.
  • Use list_models(task="...", return_recipes=True) to get full recipe paths instead of just model names.

Training

Getitune supports an API-based training approach:

from getitune.engine import create_engine

# Initialize and train using a recipe and dataset
engine = create_engine(
    model="src/getitune/recipe/classification/multi_class_cls/efficientnet_b0.yaml", # any supported model name from the model catalog (see below example), recipe.yaml path or instantiated model class directly
    data="tests/assets/classification_cifar10", # path to dataset root (any supported format, e.g., COCO, VOC, YOLO, Datumaro) or DataModule instance
    work_dir="./my_workspace",  # Defaults to "./getitune-workspace"
    device="auto",              # "auto", "cpu", "gpu", "0", "xpu", etc.
)
engine.train(max_epochs=50)
engine.test()

Tip: Pass model= as a recipe YAML path or a model name (e.g., "efficientnet_b0"), or a weights path (.xml, .onnx). If a model name matches recipes under multiple tasks, pass task= to disambiguate (e.g., task="DETECTION"). If you want to use an Ultralytics YOLO model, you can pass a YAML file in Ultralytics format as data=.

Export

Export a trained model to OpenVINO IR or ONNX format:

from getitune.engine import create_engine
from getitune.types import ExportFormat, ExportPrecision

engine = create_engine(
    model="efficientnet_b0",
    data="/path/to/dataset",
    work_dir="./my_workspace",
)
engine.train(max_epochs=50)

# Export to FP32 OpenVINO IR (default)
ov_ir_path = engine.export()

# Export to FP32 ONNX
onnx_path = engine.export(export_format=ExportFormat.ONNX)

# Export to FP16 ONNX. Same for OpenVINO IR.
onnx_path = engine.export(export_format=ExportFormat.ONNX, precision=ExportPrecision.FP16)

Validation and Inference

Getitune provides inference via PyTorch and OpenVINO backends (utilizing ModelAPI):

from getitune.engine import create_engine

# PyTorch inference with model name
engine = create_engine(
    model="efficientnet_b0",
    data="/path/to/dataset",
)
test_metrics = engine.test() # test on test subset
predictions = engine.predict() # predict on test subset
# OpenVINO inference (from exported model)
ov_engine = create_engine(
    model="/path/to/exported_model.xml",
    data="/path/to/dataset",
)
ov_engine.test() # test on test subset
ov_engine.predict() # predict on test subset
# ONNX inference (from exported model)
ov_engine = create_engine(
    model="/path/to/exported_model.onnx",
    data="/path/to/dataset",
)
ov_engine.test() # test on test subset
ov_engine.predict() # predict on test subset

Optimization

Apply post-training quantization to reduce model size and accelerate inference:

from getitune.engine import create_engine

# Load an exported OpenVINO model and optimize it
ov_engine = create_engine(
    model="/path/to/exported_model.xml",
    data="/path/to/dataset",
)
ov_engine.optimize()  # post-training quantization via NNCF
test_metrics = ov_engine.test() # test on test subset with optimized model
predictions = ov_engine.predict() # predict on test subset with optimized model

Note: The recommended calibration set size for optimization is around 300 images. Calibration images are automatically taken from the training subset of your dataset.

After .optimize() the model in the Engine is replaced with an INT8 quantized version. To re-validate or run inference with the original FP32/FP16 model, pass the model XML path directly to .test() / .predict() or create the Engine again from the original .xml.

Dataset Support

When you pass a path to data=, getitune uses Datumaro to auto-detect the dataset format. Supported formats:

Format Detection method
COCO annotations/ directory with COCO JSON files
YOLO data.yaml file (Ultralytics layout)
Pascal VOC JPEGImages/, Annotations/, ImageSets/ directories
Datumaro (native) metadata.json + data.parquet at root

Zip archives are also accepted, Datumaro extracts them on import.

# Works the same regardless of format, just point to the dataset root
engine = create_engine(
    data="/path/to/dataset_root",
    model="src/getitune/recipe/detection/yolox_s.yaml",
)
engine.train()

Note: If you are working with Ultralytics YOLO models, you can pass a YOLO Ultralytics data.yaml file directly to the data= argument of create_engine.

Advanced Usage

Direct Backend Engine Usage

The backends engines can also be instantiated directly:

# -- Lightning Backend --
from getitune.backend.lightning.engine import LightningEngine
from getitune.models import EfficientNet

engine = LightningEngine(
    model=EfficientNet(label_info=10, model_name="efficientnet_b0"),
    data="/path/to/dataset",
    work_dir="./my_workspace",
    device="auto",
)
engine.train(max_epochs=50)
engine.test()
engine.export()


# -- Ultralytics Backend --
from getitune.backend.ultralytics.engine import UltralyticsEngine
from getitune.backend.ultralytics.models import UltralyticsDetectionModel

engine = UltralyticsEngine(
    model=UltralyticsDetectionModel(model_name="yolo26s"),
    data="/path/to/yolo_dataset/data.yaml",
    work_dir="./yolo_workspace",
    device="auto",
)
engine.train(epochs=50)
engine.test()
engine.export()

> ⚠️ **Note**: Ultralytics YOLO models and the `UltralyticsEngine` backend require [installing from source](#advanced-installation-install-from-source) with the `[ultralytics]` extra.
> The PyPI package does **not** include Ultralytics support.


# -- OpenVINO Backend (inference) --
from getitune.backend.openvino.engine import OVEngine

engine = OVEngine(
    model="/path/to/exported_model.xml",
    data="/path/to/dataset",
)
engine.test()
engine.optimize()
Common Engine Parameters

Customize engine creation with the following parameters:

engine = create_engine(
    model="efficientnet_b0",
    data="/path/to/data",
    work_dir="./my_workspace",         # Working directory for checkpoints/logs; defaults to "./getitune-workspace"
    device="gpu",                       # "auto", "cpu", "gpu", "0", "1", "xpu", etc.
    checkpoint="/path/to/weights.pt",  # Optional pretrained checkpoint for warm-start training
    task="MULTI_CLASS_CLS",             # Required only if model name matches multiple tasks
)
engine.train(max_epochs=50)
Override training hyperparameters

You can override engine-level training parameters directly:

engine.train(
    max_epochs=50,
    seed=42,
    deterministic=True,
    precision="16-mixed",          # mixed-precision training
    gradient_clip_val=1.0,
    check_val_every_n_epoch=5,
)

For model-level hyperparameters like learning rate and optimizer, you can pass them when instantiating a model class directly:

from torch.optim import AdamW
from getitune.models import EfficientNet

model = EfficientNet(
    label_info=datamodule.label_info, # or simply num_classes, e.g., int value
    model_name="efficientnet_b0",
    optimizer=lambda params: AdamW(params, lr=0.001, weight_decay=0.01),
)

engine = create_engine(data=datamodule, model=model)
engine.train(max_epochs=100)

Alternatively, you can set these in a custom recipe YAML (copy and modify an existing one):

# my_recipe.yaml — custom learning rate and optimizer
task: MULTI_CLASS_CLS
model:
  class_path: getitune.backend.lightning.models.classification.multiclass_models.efficientnet.EfficientNetMulticlassCls
  init_args:
    label_info: 1000
    model_name: efficientnet_b0

    optimizer:
      class_path: torch.optim.AdamW
      init_args:
        lr: 0.001
        weight_decay: 0.01

    scheduler:
      class_path: getitune.backend.lightning.schedulers.LinearWarmupSchedulerCallable
      init_args:
        num_warmup_steps: 5
        main_scheduler_callable:
          class_path: lightning.pytorch.cli.ReduceLROnPlateau
          init_args:
            mode: max
            factor: 0.5
            patience: 3
            monitor: val/accuracy

data: src/getitune/recipe/_base_/data/classification.yaml

overrides:
  max_epochs: 100
Override augmentations and datamodule

For augmentations, override the data config. Augmentations run on CPU (augmentations_cpu) and GPU (augmentations_gpu) separately:

# my_data.yaml — stronger augmentations
task: MULTI_CLASS_CLS
input_size: [224, 224]
train_subset:
  subset_name: train
  batch_size: 32
  num_workers: 8
  augmentations_cpu:
    - class_path: torchvision.transforms.v2.RandomResizedCrop
      init_args:
        size: [224, 224]
        scale: [0.08, 1.0]
  augmentations_gpu:
    - class_path: kornia.augmentation.RandomHorizontalFlip
      init_args:
        p: 0.5
    - class_path: kornia.augmentation.ColorJiggle
      init_args:
        brightness: 0.4
        contrast: 0.4
        saturation: 0.4
        hue: 0.1
        p: 0.8
    - class_path: kornia.augmentation.RandomGaussianBlur
      init_args:
        kernel_size: [3, 3]
        sigma: [0.1, 2.0]
        p: 0.3
    - class_path: kornia.augmentation.Normalize
      init_args:
        mean: [0.485, 0.456, 0.406]
        std: [0.229, 0.224, 0.225]

Then reference your custom data config from your recipe with data: my_data.yaml.

Augmentations can also be overridden directly via the API by creating a DataModule instance and passing it to the engine:

from getitune.data.module import DataModule
from getitune.config.data import SubsetConfig
from getitune.models import EfficientNet
import kornia.augmentation as K
from torchvision.transforms import v2

datamodule = DataModule(
    task="MULTI_CLASS_CLS",
    data_root="tests/assets/classification_cifar10",
    train_subset=SubsetConfig(
        augmentations_cpu=[
            v2.Resize((256, 256)),
        ],
        augmentations_gpu=[
            K.RandomErasing(p=0.5),
            K.ColorJitter(brightness=0.2, contrast=0.2, saturation=0.2, hue=0.1),
        ],
        batch_size=32,
        num_workers=8,
    ),
    val_subset=SubsetConfig(
        subset_name="val",
        augmentations_cpu=[
            v2.Resize((256, 256)),
        ],
        batch_size=32,
        num_workers=8,
    ),
    test_subset=SubsetConfig(
        subset_name="test",
        augmentations_cpu=[
            v2.Resize((256, 256)),
        ],
        batch_size=32,
        num_workers=8,
    ),
)

model = EfficientNet(label_info=datamodule.label_info, model_name="efficientnet_b0")
engine = create_engine(data=datamodule, model=model)
engine.train()

Note: GPU augmentations (augmentations_gpu) are supported only for the Lightning backend and will be ignored for Ultralytics. For YOLO models, all augmentations should be placed on CPU via torchvision.transforms.v2.

Available model classes:

  • Detection: ATSS, SSD, YOLOX, RTDETR, DFine, DEIMDFine, DEIMV2, RFDETR, UltralyticsDetectionModel
  • Instance segmentation: MaskRCNN, MaskRCNNTV, RTMDetInst, RFDETRSeg, UltralyticsInstSegModel
  • Semantic segmentation: DinoV2Seg, LiteHRNet, SegNext
  • Classification: EfficientNet, MobileNetV3, VisionTransformer, TimmModel, TVModel
  • Keypoint: RTMPose

License

The core Geti™ Library (getitune) is licensed under Apache License Version 2.0. By contributing to the project, you agree to the license and copyright terms therein and release your contribution under these terms.

Ultralytics YOLO models are distributed under the AGPL-3.0 license, an OSI approved license ideal for open-source research, academic, and personal projects. For commercial use, enhanced support, and tailored licensing terms, please explore flexible Ultralytics licensing options at https://www.ultralytics.com/license.