Clips fp32 master gradients via ``torch.nn.utils.clip_grad_norm``. Args: max_norm (float or int): max norm of the gradie

r"""Opens all layers in model for training. Examples:: >>> from torchreid.utils import open_all_layers >>> open_all_l

Returns env info as a string. Code source: github.com/facebookresearch/maskrcnn-benchmark

Complete reductions if full pipeline is not selected or overlap is not allowed.

Complete reductions if full pipeline is not selected or overlap is not allowed.

Complete reductions if full pipeline is not selected or overlap is not allowed.

Returns number of parameters and FLOPs. .. note:: (1) this function only provides an estimate of the theoretical time complexity

Computes weight normalized weight value to assign value to module attribute with name `name`. Arguments: module (

Recursively traverse module and its children to replace all instances of ``torch.nn.modules.batchnorm._BatchNorm`` with :class:`apex.parallel

r"""Counts number of parameters in a model while ignoring ``self.classifier``. Args: model (nn.Module): network model. Examples::

Creates process groups to be used for syncbn of a give ``group_size`` and returns process group that current GPU participates in. ``grou

detach_hidden()

detach_hidden()

Multiproc file which will launch a set of processes locally for multi-gpu usage: python -m apex.parallel.multiproc main.py ...

compute distance matrix

Returns query and gallery of a test dataset, each containing tuples of (img_path(s), pid, camid). Args: name (str): datas

Filter trace markers to remove certain file names.

Note that 1 FMA = 2 flops.