r"""Gets files to load for this sample. Args: index (int): Index in [0, len(dataset)]. Returns: (tuple):

r"""Gets files to load for this sample. Args: index (int): Index in [0, len(dataset)]. Returns: keys (dict):

r"""Save network weights, optimizer parameters, scheduler parameters in the checkpoint. Args: cfg (obj): Global configuration.

Compute the trace of matrix functions Arguments: A: square matrix in trace(exp(-t*A)), where t is temperature m: number of L

r"""Operations to do before starting an epoch. Args: current_epoch (int): Current number of epoch.

r"""Operations to do before starting an iteration. Args: data (dict): Data used for the current iteration. current_it

r"""Model specific custom start of iteration process. We will do two things. First, put all the data to GPU. Second, we will resize the

r""" Record forward, backward, loss time for the discriminator update.

r""" Record forward, backward, loss, and model average time for the generator update.

r""" Record time before applying model average.

Slow reference implementation of `upfirdn2d()` using standard PyTorch ops.

r"""Get vgg16 layers

r"""Get vgg19 layers

r"""Dummy member function to be overloaded by the child class. In the child class, you can write down whatever you want to track.

r"""Write all loss values to tensorboard.

r"""Write values to tensorboard. By default, we will log the time used per iteration, time used per epoch, generator learning rate, and

r"""Write out lmdb containing (key, contents of filepath) to file. Args: filepaths (list): List of filepath strings. keys (list):

r"""Calculate height map given a voxel grid [Y, X, Z] as input. The height is defined as the Y index of the surface (non-air) block Args:

r"""Add filepath and key to list of filepaths and keys. Args: filepath (str): Filepath to add. key (str): LMDB key for this filep

r"""Use clustering to compute the features. Args: cfg (obj): Global configuration file. train_data_loader (obj): Dataloader for i

compute optical flow color map :param u: horizontal optical flow :param v: vertical optical flow :return:

r"""Compute losses on the generated flow maps. Args: flow (tensor or list of tensors): Generated flow maps. warpe

r"""Compute the kid score. Args: kid_path (str): Location for store feature activations. data_loader (obj): PyTorch dataloader ob

r"""Compute losses on the generated occlusion masks. Args: occ_mask (tensor or list of tensors): Generated occlusion masks.

r""" Args: data_x: numpy.ndarray([N, feature_dim], dtype=np.float32) data_y: numpy.ndarray([N, feature_dim], dtype=np.float32)

r"""Concat prev and now frames in first dimension, up to t frames.

r"""Draw edges by connecting the keypoints onto the label map. Args: pts (Px3 numpy array): Keypoint xy coordinates + confidence.

Get file path for our dataset structure.

r"""Returns the output of the residual branch. Args: x (tensor): Input tensor. cond_inputs (list of tensors) : Condit

Convert applicable model parameters to fp16

r""" Copy state_dictionary from source to target. Here source is the regular module and the target is the moving average module. Basic

r"""Create output subdir for this sequence. Args: output_dir (str): Root output dir. key (str): LMDB key which contai

r"""Create output subdir for this sequence. Args: output_dir (str): Root output dir. key (str): LMDB key which contai

r"""Weight initialization of GANcraft components.

r"""Define the list of keypoints that should be connected to form the edges. Args: basic_points_only (bool): Whether to use only the basi

r"""Detach tensors in the dict. Args: output (dict): Output dict. Returns: output (dict): Detached output dict.

r"""Every trainer should implement its own discriminator forward.

r"""Discriminate a sequence of images. Args: real_B (NxCxHxW tensor): Real image. fake_B (NxCxHxW tensor): Fake image

r"""Encoder images to get their content and style codes. Args: images (tensor): Input image tensor.

r"""Encoder images to get their content and style codes. Args: images (tensor): Input image tensor.

r"""Compute feature embeddings for an image image. TODO(Ting-Chun): To make this funciton dataset independent. Args: net_E (nn.Module

r"""Encode the reference image to get features for weight generation. Args: ref_image ((NxK)x3xHxW tensor): Reference images.

r"""Things to do after an epoch. Args: data (dict): Data used for the current iteration. current_epoch (int): Curren

r"""Things to do after an iteration. Args: data (dict): Data used for the current iteration. current_epoch (int): Cur

r"""Use only the valid keypoints by looking at the detection confidences. If the confidences for all keypoints in an edge are above threshold,

r"""Finetune the model for a few iterations on the inference data.

convert optical flow into color image :param flow_data: :return: color image

r"""Generates flows and masks for warping reference / previous images. Args: label (NxCxHxW tensor): Target label map.

r"""Write the value in the tensorboard. Args: step (int): Epoch or iteration number.

r"""Coarse-to-fine generator forward. Args: data (dict) : Dictionary of input data. random_style (bool): Always set t

r"""Fuse operation min mean for hinge loss computation of negative samples

r"""Fuse operation min mean for hinge loss computation of positive samples

r"""Converting voxel center array to voxel corner array. The size of the produced array grows by 1 on every dimension. Args: voxel (t

r"""Every trainer should implement its own generator forward.

r"""Get CPU affinity

r"""Get all filenames from list. Args: list_name (str): Path to filename list. Returns: all_filenames (dict): Folder name for

r"""Get temporally skipped frames from the input frames. Args: past_frames (list of tensors): Past real frames / generator outputs.

r"""Get confirm token Args: response: Check if the file exists. Returns:

r"""Adaptively generate weights for layer i in main branch convolutions. Args: x (NxCxHxW tensor): Input features. i

r"""Get height and width of crop. Args: augmentation (dict): Dict of applied augmentations. Returns: (dict): - cro

Generate a unique id based on date. Returns: str: Return uid string, e.g. '20171122171307111552'.

r"""Compute discriminator losses. Args: net_D_output (dict): Output of the discriminator.

r""" Compute edge maps for a given input instance map. Args: t (4D tensor): Input instance map. Returns: (4D tensor): Output

r"""Get pixel coordinates of the face bounding box. Args: data_cfg (obj): Data configuration. pose (NxCxHxW tensor): Pose label m

r"""Compute generator losses. Args: data_t (dict): Training data at the current time t. net_G_output (dict): Output o

r"""Get a grid ranging [-1, 1] of 2D/3D coordinates. Args: batchsize (int) : Batch size. size (tuple) : (height, width) or (depth

r"""Do stuff.

r"""Get coordinates of the hand bounding box. Args: data_cfg (obj): Data configuration. pose (NxCxHxW tensor): Pose label map.

r"""Get the data types which should be put in LMDB. Args: cfg: Configuration object.

r"""Adaptively generate weights for SPADE in layer i of generator. Args: x (NxCxHxW tensor): Input features. i (int):

r"""Return the optimizer object. Args: cfg (obj): Global config. net_G (obj): Generator network. net_D (obj): Discriminat

r"""Get if convs should be partial or not. Args: num_downs (int) : How many downsamples at current layer. Returns:

r"""Evaluating img2img network to obtain pseudo-ground truth images. Args: pseudo_gen (callable): Function converting mask to ima

r"""Get temporally skipped frames from the input frames. Args: all_frames (NxTxCxHxW tensor): All past frames. frame (Nx1xCxHxW t

r"""Return dataset objects for testing Args: cfg (obj): Global configuration file. Returns: (obj): Val data loader. It may n

r"""Get the visualization output of test function. Args: data (dict): Training data at the current iteration.

r"""Get the visualization output of test function. Args: data (dict): Training data at the current iteration.

r"""Get train parameters. Args: net (obj): Network object. param_names_start_with (list of strings): Params whose names