get Exp object by file or name. If exp_file and exp_name are both provided, get Exp by exp_file. Args: exp_file (str): file path

starts with conv layer then has `num_blocks` `ResLayer`

Returns the original image and target at an index for mixup Note: not using self.__getitem__(), as any transformations passed in coul

pytorch-accurate time

Return a process group based on gloo backend, containing all the ranks The result is cached.

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

check if model is in parallel mode.

Helper function to synchronize (barrier) among all processes when using distributed training

Returns: list[int]: size of the tensor, on each rank Tensor: padded tensor that has the max size

Run gather on arbitrary picklable data (not necessarily tensors). Args: data: any picklable object dst (int): destination ra

Single class NMS implemented in Numpy.

Args: size (int): the total number of data of the underlying dataset to sample from shuffle (bool): whether to shuffl

Args: depth (int): depth of darknet used in model, usually use [21, 53] for this param. in_channels (int): number of

Find an available port of current machine / node.

Accumulate per image evaluation results and store the result in self.eval. Does not support changing parameter settings from those u

`after_iter` contains two parts of logic: * log information * reset setting of resize

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

Apply all reduce function for python dict object. NOTE: make sure that every py_dict has the same keys and values are in the same shape.

All reduce norm statistics in different devices.