Skip to content

Reference for ultralytics/engine/model.py

Note

This file is available at https://github.com/ultralytics/ultralytics/blob/main/ultralytics/engine/model.py. If you spot a problem please help fix it by contributing a Pull Request 🛠️. Thank you 🙏!


ultralytics.engine.model.Model

Model(
    model: Union[str, Path] = "yolo11n.pt",
    task: str = None,
    verbose: bool = False,
)

Bases: Module

A base class for implementing YOLO models, unifying APIs across different model types.

This class provides a common interface for various operations related to YOLO models, such as training, validation, prediction, exporting, and benchmarking. It handles different types of models, including those loaded from local files, Ultralytics HUB, or Triton Server.

Attributes:

Name Type Description
callbacks Dict

A dictionary of callback functions for various events during model operations.

predictor BasePredictor

The predictor object used for making predictions.

model Module

The underlying PyTorch model.

trainer BaseTrainer

The trainer object used for training the model.

ckpt Dict

The checkpoint data if the model is loaded from a *.pt file.

cfg str

The configuration of the model if loaded from a *.yaml file.

ckpt_path str

The path to the checkpoint file.

overrides Dict

A dictionary of overrides for model configuration.

metrics Dict

The latest training/validation metrics.

session HUBTrainingSession

The Ultralytics HUB session, if applicable.

task str

The type of task the model is intended for.

model_name str

The name of the model.

Methods:

Name Description
__call__

Alias for the predict method, enabling the model instance to be callable.

_new

Initializes a new model based on a configuration file.

_load

Loads a model from a checkpoint file.

_check_is_pytorch_model

Ensures that the model is a PyTorch model.

reset_weights

Resets the model's weights to their initial state.

load

Loads model weights from a specified file.

save

Saves the current state of the model to a file.

info

Logs or returns information about the model.

fuse

Fuses Conv2d and BatchNorm2d layers for optimized inference.

predict

Performs object detection predictions.

track

Performs object tracking.

val

Validates the model on a dataset.

benchmark

Benchmarks the model on various export formats.

export

Exports the model to different formats.

train

Trains the model on a dataset.

tune

Performs hyperparameter tuning.

_apply

Applies a function to the model's tensors.

add_callback

Adds a callback function for an event.

clear_callback

Clears all callbacks for an event.

reset_callbacks

Resets all callbacks to their default functions.

Examples:

>>> from ultralytics import YOLO
>>> model = YOLO("yolo11n.pt")
>>> results = model.predict("image.jpg")
>>> model.train(data="coco8.yaml", epochs=3)
>>> metrics = model.val()
>>> model.export(format="onnx")

This constructor sets up the model based on the provided model path or name. It handles various types of model sources, including local files, Ultralytics HUB models, and Triton Server models. The method initializes several important attributes of the model and prepares it for operations like training, prediction, or export.

Parameters:

Name Type Description Default
model Union[str, Path]

Path or name of the model to load or create. Can be a local file path, a model name from Ultralytics HUB, or a Triton Server model.

'yolo11n.pt'
task str | None

The task type associated with the YOLO model, specifying its application domain.

None
verbose bool

If True, enables verbose output during the model's initialization and subsequent operations.

False

Raises:

Type Description
FileNotFoundError

If the specified model file does not exist or is inaccessible.

ValueError

If the model file or configuration is invalid or unsupported.

ImportError

If required dependencies for specific model types (like HUB SDK) are not installed.

Examples:

>>> model = Model("yolo11n.pt")
>>> model = Model("path/to/model.yaml", task="detect")
>>> model = Model("hub_model", verbose=True)
Source code in ultralytics/engine/model.py
def __init__(
    self,
    model: Union[str, Path] = "yolo11n.pt",
    task: str = None,
    verbose: bool = False,
) -> None:
    """
    Initializes a new instance of the YOLO model class.

    This constructor sets up the model based on the provided model path or name. It handles various types of
    model sources, including local files, Ultralytics HUB models, and Triton Server models. The method
    initializes several important attributes of the model and prepares it for operations like training,
    prediction, or export.

    Args:
        model (Union[str, Path]): Path or name of the model to load or create. Can be a local file path, a
            model name from Ultralytics HUB, or a Triton Server model.
        task (str | None): The task type associated with the YOLO model, specifying its application domain.
        verbose (bool): If True, enables verbose output during the model's initialization and subsequent
            operations.

    Raises:
        FileNotFoundError: If the specified model file does not exist or is inaccessible.
        ValueError: If the model file or configuration is invalid or unsupported.
        ImportError: If required dependencies for specific model types (like HUB SDK) are not installed.

    Examples:
        >>> model = Model("yolo11n.pt")
        >>> model = Model("path/to/model.yaml", task="detect")
        >>> model = Model("hub_model", verbose=True)
    """
    super().__init__()
    self.callbacks = callbacks.get_default_callbacks()
    self.predictor = None  # reuse predictor
    self.model = None  # model object
    self.trainer = None  # trainer object
    self.ckpt = None  # if loaded from *.pt
    self.cfg = None  # if loaded from *.yaml
    self.ckpt_path = None
    self.overrides = {}  # overrides for trainer object
    self.metrics = None  # validation/training metrics
    self.session = None  # HUB session
    self.task = task  # task type
    model = str(model).strip()

    # Check if Ultralytics HUB model from https://hub.ultralytics.com
    if self.is_hub_model(model):
        # Fetch model from HUB
        checks.check_requirements("hub-sdk>=0.0.12")
        session = HUBTrainingSession.create_session(model)
        model = session.model_file
        if session.train_args:  # training sent from HUB
            self.session = session

    # Check if Triton Server model
    elif self.is_triton_model(model):
        self.model_name = self.model = model
        self.overrides["task"] = task or "detect"  # set `task=detect` if not explicitly set
        return

    # Load or create new YOLO model
    if Path(model).suffix in {".yaml", ".yml"}:
        self._new(model, task=task, verbose=verbose)
    else:
        self._load(model, task=task)

    # Delete super().training for accessing self.model.training
    del self.training

device property

device: device

Retrieves the device on which the model's parameters are allocated.

This property determines the device (CPU or GPU) where the model's parameters are currently stored. It is applicable only to models that are instances of nn.Module.

Returns:

Type Description
device

The device (CPU/GPU) of the model.

Raises:

Type Description
AttributeError

If the model is not a PyTorch nn.Module instance.

Examples:

>>> model = YOLO("yolo11n.pt")
>>> print(model.device)
device(type='cuda', index=0)  # if CUDA is available
>>> model = model.to("cpu")
>>> print(model.device)
device(type='cpu')

names property

names: Dict[int, str]

Retrieves the class names associated with the loaded model.

This property returns the class names if they are defined in the model. It checks the class names for validity using the 'check_class_names' function from the ultralytics.nn.autobackend module. If the predictor is not initialized, it sets it up before retrieving the names.

Returns:

Type Description
Dict[int, str]

A dict of class names associated with the model.

Raises:

Type Description
AttributeError

If the model or predictor does not have a 'names' attribute.

Examples:

>>> model = YOLO("yolo11n.pt")
>>> print(model.names)
{0: 'person', 1: 'bicycle', 2: 'car', ...}

task_map property

task_map: dict

Provides a mapping from model tasks to corresponding classes for different modes.

This property method returns a dictionary that maps each supported task (e.g., detect, segment, classify) to a nested dictionary. The nested dictionary contains mappings for different operational modes (model, trainer, validator, predictor) to their respective class implementations.

The mapping allows for dynamic loading of appropriate classes based on the model's task and the desired operational mode. This facilitates a flexible and extensible architecture for handling various tasks and modes within the Ultralytics framework.

Returns:

Type Description
Dict[str, Dict[str, Any]]

A dictionary where keys are task names (str) and values are

dict

nested dictionaries. Each nested dictionary has keys 'model', 'trainer', 'validator', and

dict

'predictor', mapping to their respective class implementations.

Examples:

>>> model = Model()
>>> task_map = model.task_map
>>> detect_class_map = task_map["detect"]
>>> segment_class_map = task_map["segment"]
Note

The actual implementation of this method may vary depending on the specific tasks and classes supported by the Ultralytics framework. The docstring provides a general description of the expected behavior and structure.

transforms property

transforms

Retrieves the transformations applied to the input data of the loaded model.

This property returns the transformations if they are defined in the model. The transforms typically include preprocessing steps like resizing, normalization, and data augmentation that are applied to input data before it is fed into the model.

Returns:

Type Description
object | None

The transform object of the model if available, otherwise None.

Examples:

>>> model = YOLO("yolo11n.pt")
>>> transforms = model.transforms
>>> if transforms:
...     print(f"Model transforms: {transforms}")
... else:
...     print("No transforms defined for this model.")

__call__

__call__(
    source: Union[
        str, Path, int, Image.Image, list, tuple, np.ndarray, torch.Tensor
    ] = None,
    stream: bool = False,
    **kwargs
) -> list

Alias for the predict method, enabling the model instance to be callable for predictions.

This method simplifies the process of making predictions by allowing the model instance to be called directly with the required arguments.

Parameters:

Name Type Description Default
source str | Path | int | Image | ndarray | Tensor | List | Tuple

The source of the image(s) to make predictions on. Can be a file path, URL, PIL image, numpy array, PyTorch tensor, or a list/tuple of these.

None
stream bool

If True, treat the input source as a continuous stream for predictions.

False
**kwargs Any

Additional keyword arguments to configure the prediction process.

{}

Returns:

Type Description
List[Results]

A list of prediction results, each encapsulated in a Results object.

Examples:

>>> model = YOLO("yolo11n.pt")
>>> results = model("https://ultralytics.com/images/bus.jpg")
>>> for r in results:
...     print(f"Detected {len(r)} objects in image")
Source code in ultralytics/engine/model.py
def __call__(
    self,
    source: Union[str, Path, int, Image.Image, list, tuple, np.ndarray, torch.Tensor] = None,
    stream: bool = False,
    **kwargs,
) -> list:
    """
    Alias for the predict method, enabling the model instance to be callable for predictions.

    This method simplifies the process of making predictions by allowing the model instance to be called
    directly with the required arguments.

    Args:
        source (str | Path | int | PIL.Image | np.ndarray | torch.Tensor | List | Tuple): The source of
            the image(s) to make predictions on. Can be a file path, URL, PIL image, numpy array, PyTorch
            tensor, or a list/tuple of these.
        stream (bool): If True, treat the input source as a continuous stream for predictions.
        **kwargs (Any): Additional keyword arguments to configure the prediction process.

    Returns:
        (List[ultralytics.engine.results.Results]): A list of prediction results, each encapsulated in a
            Results object.

    Examples:
        >>> model = YOLO("yolo11n.pt")
        >>> results = model("https://ultralytics.com/images/bus.jpg")
        >>> for r in results:
        ...     print(f"Detected {len(r)} objects in image")
    """
    return self.predict(source, stream, **kwargs)

__getattr__

__getattr__(name)

Enables accessing model attributes directly through the Model class.

This method provides a way to access attributes of the underlying model directly through the Model class instance. It first checks if the requested attribute is 'model', in which case it returns the model from the module dictionary. Otherwise, it delegates the attribute lookup to the underlying model.

Parameters:

Name Type Description Default
name str

The name of the attribute to retrieve.

required

Returns:

Type Description
Any

The requested attribute value.

Raises:

Type Description
AttributeError

If the requested attribute does not exist in the model.

Examples:

>>> model = YOLO("yolo11n.pt")
>>> print(model.stride)
>>> print(model.task)
Source code in ultralytics/engine/model.py
def __getattr__(self, name):
    """
    Enables accessing model attributes directly through the Model class.

    This method provides a way to access attributes of the underlying model directly through the Model class
    instance. It first checks if the requested attribute is 'model', in which case it returns the model from
    the module dictionary. Otherwise, it delegates the attribute lookup to the underlying model.

    Args:
        name (str): The name of the attribute to retrieve.

    Returns:
        (Any): The requested attribute value.

    Raises:
        AttributeError: If the requested attribute does not exist in the model.

    Examples:
        >>> model = YOLO("yolo11n.pt")
        >>> print(model.stride)
        >>> print(model.task)
    """
    if name == "model":
        return self._modules["model"]
    return getattr(self.model, name)

add_callback

add_callback(event: str, func) -> None

Adds a callback function for a specified event.

This method allows registering custom callback functions that are triggered on specific events during model operations such as training or inference. Callbacks provide a way to extend and customize the behavior of the model at various stages of its lifecycle.

Parameters:

Name Type Description Default
event str

The name of the event to attach the callback to. Must be a valid event name recognized by the Ultralytics framework.

required
func Callable

The callback function to be registered. This function will be called when the specified event occurs.

required

Raises:

Type Description
ValueError

If the event name is not recognized or is invalid.

Examples:

>>> def on_train_start(trainer):
...     print("Training is starting!")
>>> model = YOLO("yolo11n.pt")
>>> model.add_callback("on_train_start", on_train_start)
>>> model.train(data="coco8.yaml", epochs=1)
Source code in ultralytics/engine/model.py
def add_callback(self, event: str, func) -> None:
    """
    Adds a callback function for a specified event.

    This method allows registering custom callback functions that are triggered on specific events during
    model operations such as training or inference. Callbacks provide a way to extend and customize the
    behavior of the model at various stages of its lifecycle.

    Args:
        event (str): The name of the event to attach the callback to. Must be a valid event name recognized
            by the Ultralytics framework.
        func (Callable): The callback function to be registered. This function will be called when the
            specified event occurs.

    Raises:
        ValueError: If the event name is not recognized or is invalid.

    Examples:
        >>> def on_train_start(trainer):
        ...     print("Training is starting!")
        >>> model = YOLO("yolo11n.pt")
        >>> model.add_callback("on_train_start", on_train_start)
        >>> model.train(data="coco8.yaml", epochs=1)
    """
    self.callbacks[event].append(func)

benchmark

benchmark(**kwargs)

Benchmarks the model across various export formats to evaluate performance.

This method assesses the model's performance in different export formats, such as ONNX, TorchScript, etc. It uses the 'benchmark' function from the ultralytics.utils.benchmarks module. The benchmarking is configured using a combination of default configuration values, model-specific arguments, method-specific defaults, and any additional user-provided keyword arguments.

Parameters:

Name Type Description Default
**kwargs Any

Arbitrary keyword arguments to customize the benchmarking process. These are combined with default configurations, model-specific arguments, and method defaults. Common options include: - data (str): Path to the dataset for benchmarking. - imgsz (int | List[int]): Image size for benchmarking. - half (bool): Whether to use half-precision (FP16) mode. - int8 (bool): Whether to use int8 precision mode. - device (str): Device to run the benchmark on (e.g., 'cpu', 'cuda'). - verbose (bool): Whether to print detailed benchmark information.

{}

Returns:

Type Description
Dict

A dictionary containing the results of the benchmarking process, including metrics for different export formats.

Raises:

Type Description
AssertionError

If the model is not a PyTorch model.

Examples:

>>> model = YOLO("yolo11n.pt")
>>> results = model.benchmark(data="coco8.yaml", imgsz=640, half=True)
>>> print(results)
Source code in ultralytics/engine/model.py
def benchmark(
    self,
    **kwargs,
):
    """
    Benchmarks the model across various export formats to evaluate performance.

    This method assesses the model's performance in different export formats, such as ONNX, TorchScript, etc.
    It uses the 'benchmark' function from the ultralytics.utils.benchmarks module. The benchmarking is
    configured using a combination of default configuration values, model-specific arguments, method-specific
    defaults, and any additional user-provided keyword arguments.

    Args:
        **kwargs (Any): Arbitrary keyword arguments to customize the benchmarking process. These are combined with
            default configurations, model-specific arguments, and method defaults. Common options include:
            - data (str): Path to the dataset for benchmarking.
            - imgsz (int | List[int]): Image size for benchmarking.
            - half (bool): Whether to use half-precision (FP16) mode.
            - int8 (bool): Whether to use int8 precision mode.
            - device (str): Device to run the benchmark on (e.g., 'cpu', 'cuda').
            - verbose (bool): Whether to print detailed benchmark information.

    Returns:
        (Dict): A dictionary containing the results of the benchmarking process, including metrics for
            different export formats.

    Raises:
        AssertionError: If the model is not a PyTorch model.

    Examples:
        >>> model = YOLO("yolo11n.pt")
        >>> results = model.benchmark(data="coco8.yaml", imgsz=640, half=True)
        >>> print(results)
    """
    self._check_is_pytorch_model()
    from ultralytics.utils.benchmarks import benchmark

    custom = {"verbose": False}  # method defaults
    args = {**DEFAULT_CFG_DICT, **self.model.args, **custom, **kwargs, "mode": "benchmark"}
    return benchmark(
        model=self,
        data=kwargs.get("data"),  # if no 'data' argument passed set data=None for default datasets
        imgsz=args["imgsz"],
        half=args["half"],
        int8=args["int8"],
        device=args["device"],
        verbose=kwargs.get("verbose"),
    )

clear_callback

clear_callback(event: str) -> None

Clears all callback functions registered for a specified event.

This method removes all custom and default callback functions associated with the given event. It resets the callback list for the specified event to an empty list, effectively removing all registered callbacks for that event.

Parameters:

Name Type Description Default
event str

The name of the event for which to clear the callbacks. This should be a valid event name recognized by the Ultralytics callback system.

required

Examples:

>>> model = YOLO("yolo11n.pt")
>>> model.add_callback("on_train_start", lambda: print("Training started"))
>>> model.clear_callback("on_train_start")
>>> # All callbacks for 'on_train_start' are now removed
Notes
  • This method affects both custom callbacks added by the user and default callbacks provided by the Ultralytics framework.
  • After calling this method, no callbacks will be executed for the specified event until new ones are added.
  • Use with caution as it removes all callbacks, including essential ones that might be required for proper functioning of certain operations.
Source code in ultralytics/engine/model.py
def clear_callback(self, event: str) -> None:
    """
    Clears all callback functions registered for a specified event.

    This method removes all custom and default callback functions associated with the given event.
    It resets the callback list for the specified event to an empty list, effectively removing all
    registered callbacks for that event.

    Args:
        event (str): The name of the event for which to clear the callbacks. This should be a valid event name
            recognized by the Ultralytics callback system.

    Examples:
        >>> model = YOLO("yolo11n.pt")
        >>> model.add_callback("on_train_start", lambda: print("Training started"))
        >>> model.clear_callback("on_train_start")
        >>> # All callbacks for 'on_train_start' are now removed

    Notes:
        - This method affects both custom callbacks added by the user and default callbacks
          provided by the Ultralytics framework.
        - After calling this method, no callbacks will be executed for the specified event
          until new ones are added.
        - Use with caution as it removes all callbacks, including essential ones that might
          be required for proper functioning of certain operations.
    """
    self.callbacks[event] = []

embed

embed(
    source: Union[str, Path, int, list, tuple, np.ndarray, torch.Tensor] = None,
    stream: bool = False,
    **kwargs
) -> list

Generates image embeddings based on the provided source.

This method is a wrapper around the 'predict()' method, focusing on generating embeddings from an image source. It allows customization of the embedding process through various keyword arguments.

Parameters:

Name Type Description Default
source str | Path | int | List | Tuple | ndarray | Tensor

The source of the image for generating embeddings. Can be a file path, URL, PIL image, numpy array, etc.

None
stream bool

If True, predictions are streamed.

False
**kwargs Any

Additional keyword arguments for configuring the embedding process.

{}

Returns:

Type Description
List[Tensor]

A list containing the image embeddings.

Raises:

Type Description
AssertionError

If the model is not a PyTorch model.

Examples:

>>> model = YOLO("yolo11n.pt")
>>> image = "https://ultralytics.com/images/bus.jpg"
>>> embeddings = model.embed(image)
>>> print(embeddings[0].shape)
Source code in ultralytics/engine/model.py
def embed(
    self,
    source: Union[str, Path, int, list, tuple, np.ndarray, torch.Tensor] = None,
    stream: bool = False,
    **kwargs,
) -> list:
    """
    Generates image embeddings based on the provided source.

    This method is a wrapper around the 'predict()' method, focusing on generating embeddings from an image
    source. It allows customization of the embedding process through various keyword arguments.

    Args:
        source (str | Path | int | List | Tuple | np.ndarray | torch.Tensor): The source of the image for
            generating embeddings. Can be a file path, URL, PIL image, numpy array, etc.
        stream (bool): If True, predictions are streamed.
        **kwargs (Any): Additional keyword arguments for configuring the embedding process.

    Returns:
        (List[torch.Tensor]): A list containing the image embeddings.

    Raises:
        AssertionError: If the model is not a PyTorch model.

    Examples:
        >>> model = YOLO("yolo11n.pt")
        >>> image = "https://ultralytics.com/images/bus.jpg"
        >>> embeddings = model.embed(image)
        >>> print(embeddings[0].shape)
    """
    if not kwargs.get("embed"):
        kwargs["embed"] = [len(self.model.model) - 2]  # embed second-to-last layer if no indices passed
    return self.predict(source, stream, **kwargs)

eval

eval()

Sets the model to evaluation mode.

This method changes the model's mode to evaluation, which affects layers like dropout and batch normalization that behave differently during training and evaluation.

Returns:

Type Description
Model

The model instance with evaluation mode set.

Examples:

model = YOLO("yolo11n.pt") model.eval()

Source code in ultralytics/engine/model.py
def eval(self):
    """
    Sets the model to evaluation mode.

    This method changes the model's mode to evaluation, which affects layers like dropout and batch normalization
    that behave differently during training and evaluation.

    Returns:
        (Model): The model instance with evaluation mode set.

    Examples:
        >> model = YOLO("yolo11n.pt")
        >> model.eval()
    """
    self.model.eval()
    return self

export

export(**kwargs) -> str

Exports the model to a different format suitable for deployment.

This method facilitates the export of the model to various formats (e.g., ONNX, TorchScript) for deployment purposes. It uses the 'Exporter' class for the export process, combining model-specific overrides, method defaults, and any additional arguments provided.

Parameters:

Name Type Description Default
**kwargs Dict

Arbitrary keyword arguments to customize the export process. These are combined with the model's overrides and method defaults. Common arguments include: format (str): Export format (e.g., 'onnx', 'engine', 'coreml'). half (bool): Export model in half-precision. int8 (bool): Export model in int8 precision. device (str): Device to run the export on. workspace (int): Maximum memory workspace size for TensorRT engines. nms (bool): Add Non-Maximum Suppression (NMS) module to model. simplify (bool): Simplify ONNX model.

{}

Returns:

Type Description
str

The path to the exported model file.

Raises:

Type Description
AssertionError

If the model is not a PyTorch model.

ValueError

If an unsupported export format is specified.

RuntimeError

If the export process fails due to errors.

Examples:

>>> model = YOLO("yolo11n.pt")
>>> model.export(format="onnx", dynamic=True, simplify=True)
'path/to/exported/model.onnx'
Source code in ultralytics/engine/model.py
def export(
    self,
    **kwargs,
) -> str:
    """
    Exports the model to a different format suitable for deployment.

    This method facilitates the export of the model to various formats (e.g., ONNX, TorchScript) for deployment
    purposes. It uses the 'Exporter' class for the export process, combining model-specific overrides, method
    defaults, and any additional arguments provided.

    Args:
        **kwargs (Dict): Arbitrary keyword arguments to customize the export process. These are combined with
            the model's overrides and method defaults. Common arguments include:
            format (str): Export format (e.g., 'onnx', 'engine', 'coreml').
            half (bool): Export model in half-precision.
            int8 (bool): Export model in int8 precision.
            device (str): Device to run the export on.
            workspace (int): Maximum memory workspace size for TensorRT engines.
            nms (bool): Add Non-Maximum Suppression (NMS) module to model.
            simplify (bool): Simplify ONNX model.

    Returns:
        (str): The path to the exported model file.

    Raises:
        AssertionError: If the model is not a PyTorch model.
        ValueError: If an unsupported export format is specified.
        RuntimeError: If the export process fails due to errors.

    Examples:
        >>> model = YOLO("yolo11n.pt")
        >>> model.export(format="onnx", dynamic=True, simplify=True)
        'path/to/exported/model.onnx'
    """
    self._check_is_pytorch_model()
    from .exporter import Exporter

    custom = {
        "imgsz": self.model.args["imgsz"],
        "batch": 1,
        "data": None,
        "device": None,  # reset to avoid multi-GPU errors
        "verbose": False,
    }  # method defaults
    args = {**self.overrides, **custom, **kwargs, "mode": "export"}  # highest priority args on the right
    return Exporter(overrides=args, _callbacks=self.callbacks)(model=self.model)

fuse

fuse()

Fuses Conv2d and BatchNorm2d layers in the model for optimized inference.

This method iterates through the model's modules and fuses consecutive Conv2d and BatchNorm2d layers into a single layer. This fusion can significantly improve inference speed by reducing the number of operations and memory accesses required during forward passes.

The fusion process typically involves folding the BatchNorm2d parameters (mean, variance, weight, and bias) into the preceding Conv2d layer's weights and biases. This results in a single Conv2d layer that performs both convolution and normalization in one step.

Raises:

Type Description
TypeError

If the model is not a PyTorch nn.Module.

Examples:

>>> model = Model("yolo11n.pt")
>>> model.fuse()
>>> # Model is now fused and ready for optimized inference
Source code in ultralytics/engine/model.py
def fuse(self):
    """
    Fuses Conv2d and BatchNorm2d layers in the model for optimized inference.

    This method iterates through the model's modules and fuses consecutive Conv2d and BatchNorm2d layers
    into a single layer. This fusion can significantly improve inference speed by reducing the number of
    operations and memory accesses required during forward passes.

    The fusion process typically involves folding the BatchNorm2d parameters (mean, variance, weight, and
    bias) into the preceding Conv2d layer's weights and biases. This results in a single Conv2d layer that
    performs both convolution and normalization in one step.

    Raises:
        TypeError: If the model is not a PyTorch nn.Module.

    Examples:
        >>> model = Model("yolo11n.pt")
        >>> model.fuse()
        >>> # Model is now fused and ready for optimized inference
    """
    self._check_is_pytorch_model()
    self.model.fuse()

info

info(detailed: bool = False, verbose: bool = True)

Logs or returns model information.

This method provides an overview or detailed information about the model, depending on the arguments passed. It can control the verbosity of the output and return the information as a list.

Parameters:

Name Type Description Default
detailed bool

If True, shows detailed information about the model layers and parameters.

False
verbose bool

If True, prints the information. If False, returns the information as a list.

True

Returns:

Type Description
List[str]

A list of strings containing various types of information about the model, including model summary, layer details, and parameter counts. Empty if verbose is True.

Raises:

Type Description
TypeError

If the model is not a PyTorch model.

Examples:

>>> model = Model("yolo11n.pt")
>>> model.info()  # Prints model summary
>>> info_list = model.info(detailed=True, verbose=False)  # Returns detailed info as a list
Source code in ultralytics/engine/model.py
def info(self, detailed: bool = False, verbose: bool = True):
    """
    Logs or returns model information.

    This method provides an overview or detailed information about the model, depending on the arguments
    passed. It can control the verbosity of the output and return the information as a list.

    Args:
        detailed (bool): If True, shows detailed information about the model layers and parameters.
        verbose (bool): If True, prints the information. If False, returns the information as a list.

    Returns:
        (List[str]): A list of strings containing various types of information about the model, including
            model summary, layer details, and parameter counts. Empty if verbose is True.

    Raises:
        TypeError: If the model is not a PyTorch model.

    Examples:
        >>> model = Model("yolo11n.pt")
        >>> model.info()  # Prints model summary
        >>> info_list = model.info(detailed=True, verbose=False)  # Returns detailed info as a list
    """
    self._check_is_pytorch_model()
    return self.model.info(detailed=detailed, verbose=verbose)

is_hub_model staticmethod

is_hub_model(model: str) -> bool

Check if the provided model is an Ultralytics HUB model.

This static method determines whether the given model string represents a valid Ultralytics HUB model identifier.

Parameters:

Name Type Description Default
model str

The model string to check.

required

Returns:

Type Description
bool

True if the model is a valid Ultralytics HUB model, False otherwise.

Examples:

>>> Model.is_hub_model("https://hub.ultralytics.com/models/MODEL")
True
>>> Model.is_hub_model("yolo11n.pt")
False
Source code in ultralytics/engine/model.py
@staticmethod
def is_hub_model(model: str) -> bool:
    """
    Check if the provided model is an Ultralytics HUB model.

    This static method determines whether the given model string represents a valid Ultralytics HUB model
    identifier.

    Args:
        model (str): The model string to check.

    Returns:
        (bool): True if the model is a valid Ultralytics HUB model, False otherwise.

    Examples:
        >>> Model.is_hub_model("https://hub.ultralytics.com/models/MODEL")
        True
        >>> Model.is_hub_model("yolo11n.pt")
        False
    """
    return model.startswith(f"{HUB_WEB_ROOT}/models/")

is_triton_model staticmethod

is_triton_model(model: str) -> bool

Checks if the given model string is a Triton Server URL.

This static method determines whether the provided model string represents a valid Triton Server URL by parsing its components using urllib.parse.urlsplit().

Parameters:

Name Type Description Default
model str

The model string to be checked.

required

Returns:

Type Description
bool

True if the model string is a valid Triton Server URL, False otherwise.

Examples:

>>> Model.is_triton_model("http://localhost:8000/v2/models/yolov8n")
True
>>> Model.is_triton_model("yolo11n.pt")
False
Source code in ultralytics/engine/model.py
@staticmethod
def is_triton_model(model: str) -> bool:
    """
    Checks if the given model string is a Triton Server URL.

    This static method determines whether the provided model string represents a valid Triton Server URL by
    parsing its components using urllib.parse.urlsplit().

    Args:
        model (str): The model string to be checked.

    Returns:
        (bool): True if the model string is a valid Triton Server URL, False otherwise.

    Examples:
        >>> Model.is_triton_model("http://localhost:8000/v2/models/yolov8n")
        True
        >>> Model.is_triton_model("yolo11n.pt")
        False
    """
    from urllib.parse import urlsplit

    url = urlsplit(model)
    return url.netloc and url.path and url.scheme in {"http", "grpc"}

load

load(weights: Union[str, Path] = 'yolo11n.pt') -> Model

Loads parameters from the specified weights file into the model.

This method supports loading weights from a file or directly from a weights object. It matches parameters by name and shape and transfers them to the model.

Parameters:

Name Type Description Default
weights Union[str, Path]

Path to the weights file or a weights object.

'yolo11n.pt'

Returns:

Type Description
Model

The instance of the class with loaded weights.

Raises:

Type Description
AssertionError

If the model is not a PyTorch model.

Examples:

>>> model = Model()
>>> model.load("yolo11n.pt")
>>> model.load(Path("path/to/weights.pt"))
Source code in ultralytics/engine/model.py
def load(self, weights: Union[str, Path] = "yolo11n.pt") -> "Model":
    """
    Loads parameters from the specified weights file into the model.

    This method supports loading weights from a file or directly from a weights object. It matches parameters by
    name and shape and transfers them to the model.

    Args:
        weights (Union[str, Path]): Path to the weights file or a weights object.

    Returns:
        (Model): The instance of the class with loaded weights.

    Raises:
        AssertionError: If the model is not a PyTorch model.

    Examples:
        >>> model = Model()
        >>> model.load("yolo11n.pt")
        >>> model.load(Path("path/to/weights.pt"))
    """
    self._check_is_pytorch_model()
    if isinstance(weights, (str, Path)):
        self.overrides["pretrained"] = weights  # remember the weights for DDP training
        weights, self.ckpt = attempt_load_one_weight(weights)
    self.model.load(weights)
    return self

predict

predict(
    source: Union[
        str, Path, int, Image.Image, list, tuple, np.ndarray, torch.Tensor
    ] = None,
    stream: bool = False,
    predictor=None,
    **kwargs
) -> List[Results]

Performs predictions on the given image source using the YOLO model.

This method facilitates the prediction process, allowing various configurations through keyword arguments. It supports predictions with custom predictors or the default predictor method. The method handles different types of image sources and can operate in a streaming mode.

Parameters:

Name Type Description Default
source str | Path | int | Image | ndarray | Tensor | List | Tuple

The source of the image(s) to make predictions on. Accepts various types including file paths, URLs, PIL images, numpy arrays, and torch tensors.

None
stream bool

If True, treats the input source as a continuous stream for predictions.

False
predictor BasePredictor | None

An instance of a custom predictor class for making predictions. If None, the method uses a default predictor.

None
**kwargs Any

Additional keyword arguments for configuring the prediction process.

{}

Returns:

Type Description
List[Results]

A list of prediction results, each encapsulated in a Results object.

Examples:

>>> model = YOLO("yolo11n.pt")
>>> results = model.predict(source="path/to/image.jpg", conf=0.25)
>>> for r in results:
...     print(r.boxes.data)  # print detection bounding boxes
Notes
  • If 'source' is not provided, it defaults to the ASSETS constant with a warning.
  • The method sets up a new predictor if not already present and updates its arguments with each call.
  • For SAM-type models, 'prompts' can be passed as a keyword argument.
Source code in ultralytics/engine/model.py
def predict(
    self,
    source: Union[str, Path, int, Image.Image, list, tuple, np.ndarray, torch.Tensor] = None,
    stream: bool = False,
    predictor=None,
    **kwargs,
) -> List[Results]:
    """
    Performs predictions on the given image source using the YOLO model.

    This method facilitates the prediction process, allowing various configurations through keyword arguments.
    It supports predictions with custom predictors or the default predictor method. The method handles different
    types of image sources and can operate in a streaming mode.

    Args:
        source (str | Path | int | PIL.Image | np.ndarray | torch.Tensor | List | Tuple): The source
            of the image(s) to make predictions on. Accepts various types including file paths, URLs, PIL
            images, numpy arrays, and torch tensors.
        stream (bool): If True, treats the input source as a continuous stream for predictions.
        predictor (BasePredictor | None): An instance of a custom predictor class for making predictions.
            If None, the method uses a default predictor.
        **kwargs (Any): Additional keyword arguments for configuring the prediction process.

    Returns:
        (List[ultralytics.engine.results.Results]): A list of prediction results, each encapsulated in a
            Results object.

    Examples:
        >>> model = YOLO("yolo11n.pt")
        >>> results = model.predict(source="path/to/image.jpg", conf=0.25)
        >>> for r in results:
        ...     print(r.boxes.data)  # print detection bounding boxes

    Notes:
        - If 'source' is not provided, it defaults to the ASSETS constant with a warning.
        - The method sets up a new predictor if not already present and updates its arguments with each call.
        - For SAM-type models, 'prompts' can be passed as a keyword argument.
    """
    if source is None:
        source = ASSETS
        LOGGER.warning(f"WARNING ⚠️ 'source' is missing. Using 'source={source}'.")

    is_cli = (ARGV[0].endswith("yolo") or ARGV[0].endswith("ultralytics")) and any(
        x in ARGV for x in ("predict", "track", "mode=predict", "mode=track")
    )

    custom = {"conf": 0.25, "batch": 1, "save": is_cli, "mode": "predict"}  # method defaults
    args = {**self.overrides, **custom, **kwargs}  # highest priority args on the right
    prompts = args.pop("prompts", None)  # for SAM-type models

    if not self.predictor:
        self.predictor = (predictor or self._smart_load("predictor"))(overrides=args, _callbacks=self.callbacks)
        self.predictor.setup_model(model=self.model, verbose=is_cli)
    else:  # only update args if predictor is already setup
        self.predictor.args = get_cfg(self.predictor.args, args)
        if "project" in args or "name" in args:
            self.predictor.save_dir = get_save_dir(self.predictor.args)
    if prompts and hasattr(self.predictor, "set_prompts"):  # for SAM-type models
        self.predictor.set_prompts(prompts)
    return self.predictor.predict_cli(source=source) if is_cli else self.predictor(source=source, stream=stream)

reset_callbacks

reset_callbacks() -> None

Resets all callbacks to their default functions.

This method reinstates the default callback functions for all events, removing any custom callbacks that were previously added. It iterates through all default callback events and replaces the current callbacks with the default ones.

The default callbacks are defined in the 'callbacks.default_callbacks' dictionary, which contains predefined functions for various events in the model's lifecycle, such as on_train_start, on_epoch_end, etc.

This method is useful when you want to revert to the original set of callbacks after making custom modifications, ensuring consistent behavior across different runs or experiments.

Examples:

>>> model = YOLO("yolo11n.pt")
>>> model.add_callback("on_train_start", custom_function)
>>> model.reset_callbacks()
# All callbacks are now reset to their default functions
Source code in ultralytics/engine/model.py
def reset_callbacks(self) -> None:
    """
    Resets all callbacks to their default functions.

    This method reinstates the default callback functions for all events, removing any custom callbacks that were
    previously added. It iterates through all default callback events and replaces the current callbacks with the
    default ones.

    The default callbacks are defined in the 'callbacks.default_callbacks' dictionary, which contains predefined
    functions for various events in the model's lifecycle, such as on_train_start, on_epoch_end, etc.

    This method is useful when you want to revert to the original set of callbacks after making custom
    modifications, ensuring consistent behavior across different runs or experiments.

    Examples:
        >>> model = YOLO("yolo11n.pt")
        >>> model.add_callback("on_train_start", custom_function)
        >>> model.reset_callbacks()
        # All callbacks are now reset to their default functions
    """
    for event in callbacks.default_callbacks.keys():
        self.callbacks[event] = [callbacks.default_callbacks[event][0]]

reset_weights

reset_weights() -> Model

Resets the model's weights to their initial state.

This method iterates through all modules in the model and resets their parameters if they have a 'reset_parameters' method. It also ensures that all parameters have 'requires_grad' set to True, enabling them to be updated during training.

Returns:

Type Description
Model

The instance of the class with reset weights.

Raises:

Type Description
AssertionError

If the model is not a PyTorch model.

Examples:

>>> model = Model("yolo11n.pt")
>>> model.reset_weights()
Source code in ultralytics/engine/model.py
def reset_weights(self) -> "Model":
    """
    Resets the model's weights to their initial state.

    This method iterates through all modules in the model and resets their parameters if they have a
    'reset_parameters' method. It also ensures that all parameters have 'requires_grad' set to True,
    enabling them to be updated during training.

    Returns:
        (Model): The instance of the class with reset weights.

    Raises:
        AssertionError: If the model is not a PyTorch model.

    Examples:
        >>> model = Model("yolo11n.pt")
        >>> model.reset_weights()
    """
    self._check_is_pytorch_model()
    for m in self.model.modules():
        if hasattr(m, "reset_parameters"):
            m.reset_parameters()
    for p in self.model.parameters():
        p.requires_grad = True
    return self

save

save(filename: Union[str, Path] = 'saved_model.pt') -> None

Saves the current model state to a file.

This method exports the model's checkpoint (ckpt) to the specified filename. It includes metadata such as the date, Ultralytics version, license information, and a link to the documentation.

Parameters:

Name Type Description Default
filename Union[str, Path]

The name of the file to save the model to.

'saved_model.pt'

Raises:

Type Description
AssertionError

If the model is not a PyTorch model.

Examples:

>>> model = Model("yolo11n.pt")
>>> model.save("my_model.pt")
Source code in ultralytics/engine/model.py
def save(self, filename: Union[str, Path] = "saved_model.pt") -> None:
    """
    Saves the current model state to a file.

    This method exports the model's checkpoint (ckpt) to the specified filename. It includes metadata such as
    the date, Ultralytics version, license information, and a link to the documentation.

    Args:
        filename (Union[str, Path]): The name of the file to save the model to.

    Raises:
        AssertionError: If the model is not a PyTorch model.

    Examples:
        >>> model = Model("yolo11n.pt")
        >>> model.save("my_model.pt")
    """
    self._check_is_pytorch_model()
    from copy import deepcopy
    from datetime import datetime

    from ultralytics import __version__

    updates = {
        "model": deepcopy(self.model).half() if isinstance(self.model, nn.Module) else self.model,
        "date": datetime.now().isoformat(),
        "version": __version__,
        "license": "AGPL-3.0 License (https://ultralytics.com/license)",
        "docs": "https://docs.ultralytics.com",
    }
    torch.save({**self.ckpt, **updates}, filename)

track

track(
    source: Union[str, Path, int, list, tuple, np.ndarray, torch.Tensor] = None,
    stream: bool = False,
    persist: bool = False,
    **kwargs
) -> List[Results]

Conducts object tracking on the specified input source using the registered trackers.

This method performs object tracking using the model's predictors and optionally registered trackers. It handles various input sources such as file paths or video streams, and supports customization through keyword arguments. The method registers trackers if not already present and can persist them between calls.

Parameters:

Name Type Description Default
source Union[str, Path, int, List, Tuple, ndarray, Tensor]

Input source for object tracking. Can be a file path, URL, or video stream.

None
stream bool

If True, treats the input source as a continuous video stream. Defaults to False.

False
persist bool

If True, persists trackers between different calls to this method. Defaults to False.

False
**kwargs Any

Additional keyword arguments for configuring the tracking process.

{}

Returns:

Type Description
List[Results]

A list of tracking results, each a Results object.

Raises:

Type Description
AttributeError

If the predictor does not have registered trackers.

Examples:

>>> model = YOLO("yolo11n.pt")
>>> results = model.track(source="path/to/video.mp4", show=True)
>>> for r in results:
...     print(r.boxes.id)  # print tracking IDs
Notes
  • This method sets a default confidence threshold of 0.1 for ByteTrack-based tracking.
  • The tracking mode is explicitly set in the keyword arguments.
  • Batch size is set to 1 for tracking in videos.
Source code in ultralytics/engine/model.py
def track(
    self,
    source: Union[str, Path, int, list, tuple, np.ndarray, torch.Tensor] = None,
    stream: bool = False,
    persist: bool = False,
    **kwargs,
) -> List[Results]:
    """
    Conducts object tracking on the specified input source using the registered trackers.

    This method performs object tracking using the model's predictors and optionally registered trackers. It handles
    various input sources such as file paths or video streams, and supports customization through keyword arguments.
    The method registers trackers if not already present and can persist them between calls.

    Args:
        source (Union[str, Path, int, List, Tuple, np.ndarray, torch.Tensor], optional): Input source for object
            tracking. Can be a file path, URL, or video stream.
        stream (bool): If True, treats the input source as a continuous video stream. Defaults to False.
        persist (bool): If True, persists trackers between different calls to this method. Defaults to False.
        **kwargs (Any): Additional keyword arguments for configuring the tracking process.

    Returns:
        (List[ultralytics.engine.results.Results]): A list of tracking results, each a Results object.

    Raises:
        AttributeError: If the predictor does not have registered trackers.

    Examples:
        >>> model = YOLO("yolo11n.pt")
        >>> results = model.track(source="path/to/video.mp4", show=True)
        >>> for r in results:
        ...     print(r.boxes.id)  # print tracking IDs

    Notes:
        - This method sets a default confidence threshold of 0.1 for ByteTrack-based tracking.
        - The tracking mode is explicitly set in the keyword arguments.
        - Batch size is set to 1 for tracking in videos.
    """
    if not hasattr(self.predictor, "trackers"):
        from ultralytics.trackers import register_tracker

        register_tracker(self, persist)
    kwargs["conf"] = kwargs.get("conf") or 0.1  # ByteTrack-based method needs low confidence predictions as input
    kwargs["batch"] = kwargs.get("batch") or 1  # batch-size 1 for tracking in videos
    kwargs["mode"] = "track"
    return self.predict(source=source, stream=stream, **kwargs)

train

train(trainer=None, **kwargs)

Trains the model using the specified dataset and training configuration.

This method facilitates model training with a range of customizable settings. It supports training with a custom trainer or the default training approach. The method handles scenarios such as resuming training from a checkpoint, integrating with Ultralytics HUB, and updating model and configuration after training.

When using Ultralytics HUB, if the session has a loaded model, the method prioritizes HUB training arguments and warns if local arguments are provided. It checks for pip updates and combines default configurations, method-specific defaults, and user-provided arguments to configure the training process.

Parameters:

Name Type Description Default
trainer BaseTrainer | None

Custom trainer instance for model training. If None, uses default.

None
**kwargs Any

Arbitrary keyword arguments for training configuration. Common options include: data (str): Path to dataset configuration file. epochs (int): Number of training epochs. batch_size (int): Batch size for training. imgsz (int): Input image size. device (str): Device to run training on (e.g., 'cuda', 'cpu'). workers (int): Number of worker threads for data loading. optimizer (str): Optimizer to use for training. lr0 (float): Initial learning rate. patience (int): Epochs to wait for no observable improvement for early stopping of training.

{}

Returns:

Type Description
Dict | None

Training metrics if available and training is successful; otherwise, None.

Raises:

Type Description
AssertionError

If the model is not a PyTorch model.

PermissionError

If there is a permission issue with the HUB session.

ModuleNotFoundError

If the HUB SDK is not installed.

Examples:

>>> model = YOLO("yolo11n.pt")
>>> results = model.train(data="coco8.yaml", epochs=3)
Source code in ultralytics/engine/model.py
def train(
    self,
    trainer=None,
    **kwargs,
):
    """
    Trains the model using the specified dataset and training configuration.

    This method facilitates model training with a range of customizable settings. It supports training with a
    custom trainer or the default training approach. The method handles scenarios such as resuming training
    from a checkpoint, integrating with Ultralytics HUB, and updating model and configuration after training.

    When using Ultralytics HUB, if the session has a loaded model, the method prioritizes HUB training
    arguments and warns if local arguments are provided. It checks for pip updates and combines default
    configurations, method-specific defaults, and user-provided arguments to configure the training process.

    Args:
        trainer (BaseTrainer | None): Custom trainer instance for model training. If None, uses default.
        **kwargs (Any): Arbitrary keyword arguments for training configuration. Common options include:
            data (str): Path to dataset configuration file.
            epochs (int): Number of training epochs.
            batch_size (int): Batch size for training.
            imgsz (int): Input image size.
            device (str): Device to run training on (e.g., 'cuda', 'cpu').
            workers (int): Number of worker threads for data loading.
            optimizer (str): Optimizer to use for training.
            lr0 (float): Initial learning rate.
            patience (int): Epochs to wait for no observable improvement for early stopping of training.

    Returns:
        (Dict | None): Training metrics if available and training is successful; otherwise, None.

    Raises:
        AssertionError: If the model is not a PyTorch model.
        PermissionError: If there is a permission issue with the HUB session.
        ModuleNotFoundError: If the HUB SDK is not installed.

    Examples:
        >>> model = YOLO("yolo11n.pt")
        >>> results = model.train(data="coco8.yaml", epochs=3)
    """
    self._check_is_pytorch_model()
    if hasattr(self.session, "model") and self.session.model.id:  # Ultralytics HUB session with loaded model
        if any(kwargs):
            LOGGER.warning("WARNING ⚠️ using HUB training arguments, ignoring local training arguments.")
        kwargs = self.session.train_args  # overwrite kwargs

    checks.check_pip_update_available()

    overrides = yaml_load(checks.check_yaml(kwargs["cfg"])) if kwargs.get("cfg") else self.overrides
    custom = {
        # NOTE: handle the case when 'cfg' includes 'data'.
        "data": overrides.get("data") or DEFAULT_CFG_DICT["data"] or TASK2DATA[self.task],
        "model": self.overrides["model"],
        "task": self.task,
    }  # method defaults
    args = {**overrides, **custom, **kwargs, "mode": "train"}  # highest priority args on the right
    if args.get("resume"):
        args["resume"] = self.ckpt_path

    self.trainer = (trainer or self._smart_load("trainer"))(overrides=args, _callbacks=self.callbacks)
    if not args.get("resume"):  # manually set model only if not resuming
        self.trainer.model = self.trainer.get_model(weights=self.model if self.ckpt else None, cfg=self.model.yaml)
        self.model = self.trainer.model

    self.trainer.hub_session = self.session  # attach optional HUB session
    self.trainer.train()
    # Update model and cfg after training
    if RANK in {-1, 0}:
        ckpt = self.trainer.best if self.trainer.best.exists() else self.trainer.last
        self.model, _ = attempt_load_one_weight(ckpt)
        self.overrides = self.model.args
        self.metrics = getattr(self.trainer.validator, "metrics", None)  # TODO: no metrics returned by DDP
    return self.metrics

tune

tune(use_ray=False, iterations=10, *args, **kwargs)

Conducts hyperparameter tuning for the model, with an option to use Ray Tune.

This method supports two modes of hyperparameter tuning: using Ray Tune or a custom tuning method. When Ray Tune is enabled, it leverages the 'run_ray_tune' function from the ultralytics.utils.tuner module. Otherwise, it uses the internal 'Tuner' class for tuning. The method combines default, overridden, and custom arguments to configure the tuning process.

Parameters:

Name Type Description Default
use_ray bool

If True, uses Ray Tune for hyperparameter tuning. Defaults to False.

False
iterations int

The number of tuning iterations to perform. Defaults to 10.

10
*args List

Variable length argument list for additional arguments.

()
**kwargs Dict

Arbitrary keyword arguments. These are combined with the model's overrides and defaults.

{}

Returns:

Type Description
Dict

A dictionary containing the results of the hyperparameter search.

Raises:

Type Description
AssertionError

If the model is not a PyTorch model.

Examples:

>>> model = YOLO("yolo11n.pt")
>>> results = model.tune(use_ray=True, iterations=20)
>>> print(results)
Source code in ultralytics/engine/model.py
def tune(
    self,
    use_ray=False,
    iterations=10,
    *args,
    **kwargs,
):
    """
    Conducts hyperparameter tuning for the model, with an option to use Ray Tune.

    This method supports two modes of hyperparameter tuning: using Ray Tune or a custom tuning method.
    When Ray Tune is enabled, it leverages the 'run_ray_tune' function from the ultralytics.utils.tuner module.
    Otherwise, it uses the internal 'Tuner' class for tuning. The method combines default, overridden, and
    custom arguments to configure the tuning process.

    Args:
        use_ray (bool): If True, uses Ray Tune for hyperparameter tuning. Defaults to False.
        iterations (int): The number of tuning iterations to perform. Defaults to 10.
        *args (List): Variable length argument list for additional arguments.
        **kwargs (Dict): Arbitrary keyword arguments. These are combined with the model's overrides and defaults.

    Returns:
        (Dict): A dictionary containing the results of the hyperparameter search.

    Raises:
        AssertionError: If the model is not a PyTorch model.

    Examples:
        >>> model = YOLO("yolo11n.pt")
        >>> results = model.tune(use_ray=True, iterations=20)
        >>> print(results)
    """
    self._check_is_pytorch_model()
    if use_ray:
        from ultralytics.utils.tuner import run_ray_tune

        return run_ray_tune(self, max_samples=iterations, *args, **kwargs)
    else:
        from .tuner import Tuner

        custom = {}  # method defaults
        args = {**self.overrides, **custom, **kwargs, "mode": "train"}  # highest priority args on the right
        return Tuner(args=args, _callbacks=self.callbacks)(model=self, iterations=iterations)

val

val(validator=None, **kwargs)

Validates the model using a specified dataset and validation configuration.

This method facilitates the model validation process, allowing for customization through various settings. It supports validation with a custom validator or the default validation approach. The method combines default configurations, method-specific defaults, and user-provided arguments to configure the validation process.

Parameters:

Name Type Description Default
validator BaseValidator | None

An instance of a custom validator class for validating the model.

None
**kwargs Any

Arbitrary keyword arguments for customizing the validation process.

{}

Returns:

Type Description
DetMetrics

Validation metrics obtained from the validation process.

Raises:

Type Description
AssertionError

If the model is not a PyTorch model.

Examples:

>>> model = YOLO("yolo11n.pt")
>>> results = model.val(data="coco8.yaml", imgsz=640)
>>> print(results.box.map)  # Print mAP50-95
Source code in ultralytics/engine/model.py
def val(
    self,
    validator=None,
    **kwargs,
):
    """
    Validates the model using a specified dataset and validation configuration.

    This method facilitates the model validation process, allowing for customization through various settings. It
    supports validation with a custom validator or the default validation approach. The method combines default
    configurations, method-specific defaults, and user-provided arguments to configure the validation process.

    Args:
        validator (ultralytics.engine.validator.BaseValidator | None): An instance of a custom validator class for
            validating the model.
        **kwargs (Any): Arbitrary keyword arguments for customizing the validation process.

    Returns:
        (ultralytics.utils.metrics.DetMetrics): Validation metrics obtained from the validation process.

    Raises:
        AssertionError: If the model is not a PyTorch model.

    Examples:
        >>> model = YOLO("yolo11n.pt")
        >>> results = model.val(data="coco8.yaml", imgsz=640)
        >>> print(results.box.map)  # Print mAP50-95
    """
    custom = {"rect": True}  # method defaults
    args = {**self.overrides, **custom, **kwargs, "mode": "val"}  # highest priority args on the right

    validator = (validator or self._smart_load("validator"))(args=args, _callbacks=self.callbacks)
    validator(model=self.model)
    self.metrics = validator.metrics
    return validator.metrics



📅 Created 1 year ago ✏️ Updated 2 months ago