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

Benchmark the speed of data loading in PySlowFast. Args: cfg (CfgNode): configs. Details can be found in slowfast/config

Builds the backbone of the model.

Builds the loss function for the model.

Builds the video model. Args: cfg (configs): configs that contains the hyper-parameters to build the backbone. Details can be

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

Classifies the features into classes.

Collate a batch of data points. Args: batch (list): A list of tuples containing the image tensor, the label tensor,

Collate a batch of data points. Args: batch (list): A list of tuples containing the image tensor and label tensor.

Collate a batch of data points. Args: batch (list): A list of tuples containing the image tensor, the label tensor,

color transfer for [0,1] float32 inputs

Compute and update the batch norm stats to make it more precise. During training both bn stats and the weight are changing after every iterat

Construct a stochastic gradient descent or ADAM optimizer with momentum. Details can be found in: Herbert Robbins, and Sutton Monro. "A s

Convert the caffe2_layer_name to pytorch format by apply the list of regular expressions. Args: caffe2_layer_name

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

returns either a context list that includes one that will disable zero.Init or an empty context list

Run AVA evaluation given annotation/prediction files.

Returns the features from the backbone given the input data.

Calculate and log the final ensembled metrics. ks (tuple): list of top-k values for topk_accuracies. For example, ks = (1