:param d_model: dimension of the model :param length: length of positions :return: length*d_model position matrix

args provides: normalize_text_clip num_layers

some evaluation about the lossy evalution

add evaluation path

add finetune

add finetune

add the weights for each loss

add the weights for each loss

add some freq related options.

add the weights for each loss. this is copied from motion clip. include lambda_rc, lambda_rcxyz, lambda_vel, lambda_velxyz

add args for motion clip model

add options for optimization setting

add the path for data loading, model training, result saving

check the load file path exist. make the evaluation save path.

check the load file path exist. make the finetune save path.

check the load file path exist. make the finetune save path.

check if the file is empty. args file_path: str, the path of the file. returns bool, True if the file is empty.

check if the file is empty. args file_path: str, the path of the file. returns bool, True if the file is empty.

check that the generated motion has the same number as the caption

check that the generated motion has the same number as the caption

the collate function for the smal motion dataset

the collate function for the smal motion dataset

the collate function for the smpl motion dataset

the collate function for the smpl motion dataset

compute the cosine similarity loss

compute diversity

compute diversity

compute the lambdas

compute the lossy velocity loss

compute the lossy velocity loss

compute multimodality from prediction

compute r precision

compute r precision

compute the velocity loss

convert seconds to days, hours, minutes, seconds.

convert seconds to days, hours, minutes, seconds.

create dataloader only for mm evaluation.

create dataloader from the caption path and generated motion path

create dataloader only for mm evaluation.

create the loss recorder

create the loss recorder

Generate a unique name by adding a random prefix to the base name.

Generate a unique name by adding a random prefix to the base name.

get the caption from the file name

get the caption from the file name

get the list of generated motion path.

get the match score of the generated motions

get the match score of the generated motions

get the text path from the maa path

get the text path from the maa path

get top k of the generated motions

get top k of the generated motions

Process motion data and update the data structures.

Process motion data and update the data structures.

initialize the semantic head

Check if the motion's length is within the valid range.

Check if the motion's length is within the valid range.

convert lengths to mask

load animal id from meta file.

load animal id from meta file.

load animal id from meta file.

load caption from the caption file path.

load caption from the caption file path.

load caption from the caption file path.

load generated motion from the generated motion file path.

load generated motion from the generated motion file path.

load generated motion from the generated motion file path.

load motion length from the caption file path.

load motion length from the caption file path.

load motion length from the caption file path.

load the pretrained SMPL autoencoder

load offset from animal id

load offset from animal id

load offset from animal id

Parse a single line and return extracted details.

Parse a single line and return extracted details.

the current motion is represented as 321 and the motion should be 36*6

the current motion is represented as 321 and the motion should be 36*6

the current motion is represented as 321 and the motion should be 36*6

recover the text from the tokens. this function is used for create caption of the visualzation.

recover the text from the tokens. this function is used for create caption of the visualzation.

recover the text from the tokens. this function is used for create caption of the visualzation.