Set transformation parameters if required.

Set transformation parameters if required.

Set transformation parameters if required.

Set transformation parameters if required.

Set transformation parameters if required.

All gathers the provided tensors from all processes across machines. Args: tensors (list): tensors to perform all gather across all p

Run all_gather on arbitrary picklable data (not necessarily tensors). Args: data: any picklable object group: a torch proces

All reduce the provided tensors from all processes across machines. Args: tensors (list): tensors to perform all reduce across all pr

Create AugMix PyTorch transform :param config_str: String defining configuration of random augmentation. Consists of multiple sections separated

Create a AutoAugment transform :param config_str: String defining configuration of auto augmentation. Consists of multiple sections separated

Returns: (float): the average number of seconds between every start/reset and pause.

Convert BN parameters to Sub-BN parameters if model contains Sub-BNs. Args: key (OrderedDict): source dict of parameters. mdo

Calculates loss according to cfg. For pre-training, losses are defined and registered in `SSL_LOSSES`. Different losses can be combined b

Determine whether the loss is NaN (not a number). Args: loss (loss): loss to check whether is NaN.

Remove the temporary files. Args: file_to_remove (string or BytesIO): If given "string", file will be removed.

Construct an optimizer. Supported optimizers include: SGD: Herbert Robbins, and Sutton Monro. "A stochastic approximation method.

Compute the system memory (RAM) usage for the current device (GB). Returns: usage (float): used memory (GB). total (float): t

Post processing for epic dataset. Args: cfg (Config): the global config object. video_list (list): videos name list.

Post processing for part videos in epic dataset. Args: cfg (Config): the global config object. video_list (list): videos name

Calculate and log the final ensembled metrics for joint verb and noun training. ks (tuple): list of top-k values for topk_ac

Calculate and log the final ensembled metrics, for separate verb and noun training. ks (tuple): list of top-k values for top

Args: x (tensor): feature to perform convolution on. alpha (tensor): calibration weight for the base weights.

Args: x (tensor): feature to perform convolution on. alpha (tensor): calibration weight for the base weights.

Args: feat (tensor): feature to perform convolution on. reshape_required (bool): True if intput feat is the shape of

Feature gating as used in S3D-G

Returns: x (Tensor): classification predictions. logits (Tensor): global average pooled features.

Returns: x (dict): dictionary of classification predictions, with keys "verb_class" and "noun_class" indicating

Returns: x (Tensor): classification predictions. logits (Tensor): global average pooled features.

Returns: x (dict): dictionary of classification predictions, with keys "verb_class" and "noun_class" indicating

Returns: x (dict): dictionary with keys "move_x", indicating the category prediction on the x-axis. l

Returns: x (dict): dictionary with keys "move_y", indicating the category prediction on the y-axis. l

Returns: x (Tensor): joint prediction on both axes. logits (Tensor): global average pooled features.

Args: x (dict): { "video" (tensor): Features for sliding windows. } Returns: outp

Returns: x (Tensor): classification predictions. logits (Tensor): global average pooled features.

Returns: x (dict): dictionary of classification predictions, with keys "verb_class" and "noun_class" indicating

Set all the bn layers to eval mode. Args: model (model): model to set bn layers to eval mode.

Prints all registered class.

Retrieves the lr for the given epoch (as specified by the lr policy). Args: cur_epoch (float): current poch id. cfg (Config):

Returns: The rank of the current process within the local (per-machine) process group.

Returns: The size of the per-machine process group, i.e. the number of processes per machine.

Retrieve the learning rate of the current epoch with the option to perform warm up in the beginning of the training stage. Args:

Calculate the current average value of the deque.

Initialize variables needed for distributed training.

Initializes the default process group. Args: local_rank (int): the rank on the current local machine. local_world_size (int):