Allreduce gradients. Args: params (list[torch.Parameters]): List of parameters of a model coalesce (bool, optional): Whether allr

Clip bboxes to fit the image shape. Args: bboxes (ndarray): Shape (..., 4*k) img_shape (tuple[int]): (height, width) of the image

Convert a BGR image to grayscale image. Args: img (ndarray): The input image. keepdim (bool): If False (by default), then return

Build PyTorch DataLoader. In distributed training, each GPU/process has a dataloader. In non-distributed training, there is only one dataload

Build datasets.

Builder for drop out layers.

Build loss.

A decorator factory to check if prerequisites are satisfied. Args: prerequisites (str of list[str]): Prerequisites to be checked.

Contrast distortion.

3x3 convolution with padding.

Copy gradients from fp16 model to fp32 weight copy.

Copy updated params from fp32 weight copy to fp16 model.

Get current learning rates. Returns: list[float] | dict[str, list[float]]: Current learning rates of all param gr

Denormalize input grid from range [0, 1] to [-1, 1] Args: grid (Tensor): The grid to be denormalize, range [0, 1]. Returns: Te

Evaluate the dataset. Args: results (list): Testing results of the dataset. metric (str | list[str]): Metrics to be e

Extract features from images.

Flattens predictions in the batch (binary case) Remove labels equal to 'ignore_index'.

Flattens predictions in the batch.

Convert FLOPs number into a string. Note that Here we take a multiply-add counts as one FLOP. Args: flops (float): FLOPs number to b

Placeholder of forward function.

Forward function.

Placeholder of forward function.

Scatters points into voxels. Args: points (torch.Tensor): Points to be reduced into voxels. coors (torch.Tensor): Cor

Read data from a given ``filepath`` with 'rb' mode. Args: filepath (str or Path): Path to read data. Returns:

Randomly get a crop bounding box.

Get frame by index. Args: frame_id (int): Index of the expected frame, 0-based. Returns: ndarray or None: Re

Get hostname and username. Return empty string if exception raised, e.g. ``getpass.getuser()`` will lead to error in docker container

Calculate parameter number of a model. Args: model (nn.module): The model for parameter number calculation. Returns: float:

Check if params contain overflow.

Check whether it is a list of some type. A partial method of :func:`is_seq_of`.

Check if a method of base class is overridden in derived class. Args: method (str): the method name to check. base_class (type):

Cast elements of an iterable object into some type. Args: inputs (Iterable): The input object. dst_type (type): Destination type.

Concatenate all file paths. Args: filepath (str or Path): Path to be concatenated. Returns: str: The result

Load state_dict to a module. This method is modified from :meth:`torch.nn.Module.load_state_dict`. Default value for ``strict`` is set to ``F

Computes gradient of the Lovasz extension w.r.t sorted errors. See Alg. 1 in paper.

Binary Lovasz hinge loss. Args: logits (torch.Tensor): [P], logits at each prediction (between -infty and +infty). la

Multi-class Lovasz-Softmax loss. Args: probs (torch.Tensor): [P, C], class probabilities at each prediction (between 0 and 1)

Test model with multiple gpus. This method tests model with multiple gpus and collects the results under two different modes: gpu and cpu mod

nn.Module: normalization layer after the first convolution layer

Convert parameter number into a string. Args: num_params (float): Parameter number to be converted. units (str | None): Converted

Prepare results dict for pipeline.

Write data to a given ``filepath`` with 'wb' mode. Note: ``put`` should create a directory if the directory of ``filepath``

Write data to a given ``filepath`` with 'w' mode. Note: ``put_text`` should create a directory if the directory of ``

Randomly sample an img_scale when ``multiscale_mode=='range'``. Args: img_scales (list[tuple]): Images scale range for sampling.

Randomly sample an img_scale when ``ratio_range`` is specified. A ratio will be randomly sampled from the range specified by ``ratio_

Convert roi based relative point coordinates to image based absolute point coordinates. Args: rois (Tensor): RoIs or BBoxes, shape (N

Calculate the new size to be rescaled to. Args: old_size (tuple[int]): The old size (w, h) of image. scale (float | tuple[int]):

Resize pos_embed weights. Resize pos_embed using bicubic interpolate method. Args: pos_embed (torch.Tensor): pos_embed we

Save checkpoint to file. The checkpoint will have 3 fields: ``meta``, ``state_dict`` and ``optimizer``. By default ``meta`` will contain vers

Scan a directory to find the interested files. Args: dir_path (str | obj:`Path`): Path of the directory. suffix (str | tuple(str)

Scatter inputs to target gpus. The only difference from original :func:`scatter` is to add support for :type:`~mmcv.parallel.DataContainer`.

Scatters tensor across multiple GPUs.

Draw `result` over `img`. Args: img (str or Tensor): The image to be displayed. result (Tensor): The semantic segment

Time since the last checking. Either :func:`since_start` or :func:`since_last_check` is a checking operation. Returns (float

Test with single GPU. Args: model (nn.Module): Model to be tested. data_loader (utils.data.Dataloader): Pytorch data loader.

Start the timer.

Track the progress of tasks execution with a progress bar. Tasks are done with a simple for-loop. Args: func (callable): The functio

update the current loss scale value when overflow happens.

Wrap the FP32 model to FP16. If you are using PyTorch >= 1.6, torch.cuda.amp is used as the backend, otherwise, original mmcv implementation

Returns a diagonal matrix of size [n, n]. The diagonal are all "-inf". This is for avoiding calculating the overlapped element in the Criss-C