nuwa pytorch
0.7.8
Implementación de NÜWA, red de atención de última generación para síntesis de texto a video, en Pytorch. También contiene una extensión para la generación de vídeo y audio, utilizando un enfoque de decodificador dual.
Yannic Kilcher
Lector profundo
Marzo de 2022: veo señales de vida con una versión difícil de mnist en movimiento
Abril de 2022: parece que un método basado en difusión ha tomado el nuevo trono para SOTA. Sin embargo, continuaré con NUWA, ampliándolo para usar códigos de múltiples cabezas + transformador causal jerárquico. Creo que esa dirección está sin explotar para mejorar esta línea de trabajo.
$ pip install nuwa-pytorchPrimero entrene al VAE
import torch
from nuwa_pytorch import VQGanVAE
vae = VQGanVAE (
dim = 512 ,
channels = 3 , # default is 3, but can be changed to any value for the training of the segmentation masks (sketches)
image_size = 256 , # image size
num_layers = 4 , # number of downsampling layers
num_resnet_blocks = 2 , # number of resnet blocks
vq_codebook_size = 8192 , # codebook size
vq_decay = 0.8 # codebook exponential decay
)
imgs = torch . randn ( 10 , 3 , 256 , 256 )
# alternate learning for autoencoder ...
loss = vae ( imgs , return_loss = True )
loss . backward ()
# and the discriminator ...
discr_loss = vae ( imgs , return_discr_loss = True )
discr_loss . backward ()
# do above for many steps
# return reconstructed images and make sure they look ok
recon_imgs = vae ( imgs )Luego, con tu VAE aprendido
import torch
from nuwa_pytorch import NUWA , VQGanVAE
# autoencoder
vae = VQGanVAE (
dim = 64 ,
num_layers = 4 ,
image_size = 256 ,
num_conv_blocks = 2 ,
vq_codebook_size = 8192
)
# NUWA transformer
nuwa = NUWA (
vae = vae ,
dim = 512 ,
text_num_tokens = 20000 , # number of text tokens
text_enc_depth = 12 , # text encoder depth
text_enc_heads = 8 , # number of attention heads for encoder
text_max_seq_len = 256 , # max sequence length of text conditioning tokens (keep at 256 as in paper, or shorter, if your text is not that long)
max_video_frames = 10 , # number of video frames
image_size = 256 , # size of each frame of video
dec_depth = 64 , # video decoder depth
dec_heads = 8 , # number of attention heads in decoder
dec_reversible = True , # reversible networks - from reformer, decoupling memory usage from depth
enc_reversible = True , # reversible encoders, if you need it
attn_dropout = 0.05 , # dropout for attention
ff_dropout = 0.05 , # dropout for feedforward
sparse_3dna_kernel_size = ( 5 , 3 , 3 ), # kernel size of the sparse 3dna attention. can be a single value for frame, height, width, or different values (to simulate axial attention, etc)
sparse_3dna_dilation = ( 1 , 2 , 4 ), # cycle dilation of 3d conv attention in decoder, for more range
shift_video_tokens = True # cheap relative positions for sparse 3dna transformer, by shifting along spatial dimensions by one
). cuda ()
# data
text = torch . randint ( 0 , 20000 , ( 1 , 256 )). cuda ()
video = torch . randn ( 1 , 10 , 3 , 256 , 256 ). cuda () # (batch, frames, channels, height, width)
loss = nuwa (
text = text ,
video = video ,
return_loss = True # set this to True, only for training, to return cross entropy loss
)
loss . backward ()
# do above with as much data as possible
# then you can generate a video from text
video = nuwa . generate ( text = text , num_frames = 5 ) # (1, 5, 3, 256, 256) En el artículo, también presentan una forma de condicionar la generación de vídeo en función de máscaras de segmentación. También puedes hacer esto fácilmente, siempre que entrenes a un VQGanVAE en los bocetos de antemano.
Luego, usará NUWASketch en lugar de NUWA , que puede aceptar el boceto VAE como referencia.
ex.
import torch
from nuwa_pytorch import NUWASketch , VQGanVAE
# autoencoder, one for main video, the other for the sketch
vae = VQGanVAE (
dim = 64 ,
num_layers = 4 ,
image_size = 256 ,
num_conv_blocks = 2 ,
vq_codebook_size = 8192
)
sketch_vae = VQGanVAE (
dim = 512 ,
channels = 5 , # say the sketch has 5 classes
num_layers = 4 ,
image_size = 256 ,
num_conv_blocks = 2 ,
vq_codebook_size = 8192
)
# NUWA transformer for conditioning with sketches
nuwa = NUWASketch (
vae = vae ,
sketch_vae = sketch_vae ,
dim = 512 , # model dimensions
sketch_enc_depth = 12 , # sketch encoder depth
sketch_enc_heads = 8 , # number of attention heads for sketch encoder
sketch_max_video_frames = 3 , # max number of frames for sketches
sketch_enc_use_sparse_3dna = True , # whether to use 3d-nearby attention (of full attention if False) for sketch encoding transformer
max_video_frames = 10 , # number of video frames
image_size = 256 , # size of each frame of video
dec_depth = 64 , # video decoder depth
dec_heads = 8 , # number of attention heads in decoder
dec_reversible = True , # reversible networks - from reformer, decoupling memory usage from depth
enc_reversible = True , # reversible encoders, if you need it
attn_dropout = 0.05 , # dropout for attention
ff_dropout = 0.05 , # dropout for feedforward
sparse_3dna_kernel_size = ( 5 , 3 , 3 ), # kernel size of the sparse 3dna attention. can be a single value for frame, height, width, or different values (to simulate axial attention, etc)
sparse_3dna_dilation = ( 1 , 2 , 4 ), # cycle dilation of 3d conv attention in decoder, for more range
cross_2dna_kernel_size = 5 , # 2d kernel size of spatial grouping of attention from video frames to sketches
cross_2dna_dilation = 1 , # 2d dilation of spatial attention from video frames to sketches
shift_video_tokens = True # cheap relative positions for sparse 3dna transformer, by shifting along spatial dimensions by one
). cuda ()
# data
sketch = torch . randn ( 2 , 2 , 5 , 256 , 256 ). cuda () # (batch, frames, segmentation classes, height, width)
sketch_mask = torch . ones ( 2 , 2 ). bool (). cuda () # (batch, frames) [Optional]
video = torch . randn ( 2 , 10 , 3 , 256 , 256 ). cuda () # (batch, frames, channels, height, width)
loss = nuwa (
sketch = sketch ,
sketch_mask = sketch_mask ,
video = video ,
return_loss = True # set this to True, only for training, to return cross entropy loss
)
loss . backward ()
# do above with as much data as possible
# then you can generate a video from sketch(es)
video = nuwa . generate ( sketch = sketch , num_frames = 5 ) # (1, 5, 3, 256, 256) Este repositorio también ofrecerá una variante de NUWA que puede producir tanto vídeo como audio. Por ahora, el audio deberá codificarse manualmente.
import torch
from nuwa_pytorch import NUWAVideoAudio , VQGanVAE
# autoencoder
vae = VQGanVAE (
dim = 64 ,
num_layers = 4 ,
image_size = 256 ,
num_conv_blocks = 2 ,
vq_codebook_size = 100
)
# NUWA transformer
nuwa = NUWAVideoAudio (
vae = vae ,
dim = 512 ,
num_audio_tokens = 2048 , # codebook size for audio tokens
num_audio_tokens_per_video_frame = 32 , # number of audio tokens per video frame
cross_modality_attn_every = 3 , # cross modality attention every N layers
text_num_tokens = 20000 , # number of text tokens
text_enc_depth = 1 , # text encoder depth
text_enc_heads = 8 , # number of attention heads for encoder
text_max_seq_len = 256 , # max sequence length of text conditioning tokens (keep at 256 as in paper, or shorter, if your text is not that long)
max_video_frames = 10 , # number of video frames
image_size = 256 , # size of each frame of video
dec_depth = 4 , # video decoder depth
dec_heads = 8 , # number of attention heads in decoder
enc_reversible = True , # reversible encoders, if you need it
dec_reversible = True , # quad-branched reversible network, for making depth of twin video / audio decoder independent of network depth. recommended to be turned on unless you have a ton of memory at your disposal
attn_dropout = 0.05 , # dropout for attention
ff_dropout = 0.05 , # dropout for feedforward
sparse_3dna_kernel_size = ( 5 , 3 , 3 ), # kernel size of the sparse 3dna attention. can be a single value for frame, height, width, or different values (to simulate axial attention, etc)
sparse_3dna_dilation = ( 1 , 2 , 4 ), # cycle dilation of 3d conv attention in decoder, for more range
shift_video_tokens = True # cheap relative positions for sparse 3dna transformer, by shifting along spatial dimensions by one
). cuda ()
# data
text = torch . randint ( 0 , 20000 , ( 1 , 256 )). cuda ()
audio = torch . randint ( 0 , 2048 , ( 1 , 32 * 10 )). cuda () # (batch, audio tokens per frame * max video frames)
video = torch . randn ( 1 , 10 , 3 , 256 , 256 ). cuda () # (batch, frames, channels, height, width)
loss = nuwa (
text = text ,
video = video ,
audio = audio ,
return_loss = True # set this to True, only for training, to return cross entropy loss
)
loss . backward ()
# do above with as much data as possible
# then you can generate a video from text
video , audio = nuwa . generate ( text = text , num_frames = 5 ) # (1, 5, 3, 256, 256), (1, 32 * 5 == 160) Esta biblioteca ofrecerá algunas utilidades para facilitar la formación. Para empezar, puede utilizar la clase VQGanVAETrainer para encargarse de entrenar el VQGanVAE . Simplemente ajuste el modelo y pase también la ruta de la carpeta de imágenes, así como los distintos hiperparámetros de entrenamiento.
import torch
from nuwa_pytorch import VQGanVAE , VQGanVAETrainer
vae = VQGanVAE (
dim = 64 ,
image_size = 256 ,
num_layers = 5 ,
vq_codebook_size = 1024 ,
vq_use_cosine_sim = True ,
vq_codebook_dim = 32 ,
vq_orthogonal_reg_weight = 10 ,
vq_orthogonal_reg_max_codes = 128 ,
). cuda ()
trainer = VQGanVAETrainer (
vae , # VAE defined above
folder = '/path/to/images' , # path to images
lr = 3e-4 , # learning rate
num_train_steps = 100000 , # number of training steps
batch_size = 8 , # batch size
grad_accum_every = 4 # gradient accumulation (effective batch size is (batch_size x grad_accum_every))
)
trainer . train ()
# results and model checkpoints will be saved periodically to ./results Para entrenar NUWA, primero debe organizar una carpeta de archivos .gif con los archivos .txt correspondientes que contengan su título. Debería organizarse como tal.
ex.
video-and-text-data
┣ cat.gif
┣ cat.txt
┣ dog.gif
┣ dog.txt
┣ turtle.gif
┗ turtle.txt
Luego cargarás tu VQGan-VAE previamente entrenado y entrenarás a NUWA con las clases GifVideoDataset y NUWATrainer .
import torch
from nuwa_pytorch import NUWA , VQGanVAE
from nuwa_pytorch . train_nuwa import GifVideoDataset , NUWATrainer
# dataset
ds = GifVideoDataset (
folder = './path/to/videos/' ,
channels = 1
)
# autoencoder
vae = VQGanVAE (
dim = 64 ,
image_size = 256 ,
num_layers = 5 ,
num_resnet_blocks = 2 ,
vq_codebook_size = 512 ,
attn_dropout = 0.1
)
vae . load_state_dict ( torch . load ( './path/to/trained/vae.pt' ))
# NUWA transformer
nuwa = NUWA (
vae = vae ,
dim = 512 ,
text_enc_depth = 6 ,
text_max_seq_len = 256 ,
max_video_frames = 10 ,
dec_depth = 12 ,
dec_reversible = True ,
enc_reversible = True ,
attn_dropout = 0.05 ,
ff_dropout = 0.05 ,
sparse_3dna_kernel_size = ( 5 , 3 , 3 ),
sparse_3dna_dilation = ( 1 , 2 , 4 ),
shift_video_tokens = True
). cuda ()
# data
trainer = NUWATrainer (
nuwa = nuwa , # NUWA transformer
dataset = dataset , # video dataset class
num_train_steps = 1000000 , # number of training steps
lr = 3e-4 , # learning rate
wd = 0.01 , # weight decay
batch_size = 8 , # batch size
grad_accum_every = 4 , # gradient accumulation
max_grad_norm = 0.5 , # gradient clipping
num_sampled_frames = 10 , # number of frames to sample
results_folder = './results' # folder to store checkpoints and samples
)
trainer . train () Esta biblioteca depende de esta biblioteca de cuantificación de vectores, que viene con una serie de mejoras (vqgan mejorado, regularización de libro de códigos ortogonal, etc.). Para utilizar cualquiera de estas mejoras, puede configurar los parámetros de palabras clave del cuantificador vectorial anteponiendo vq_ en la inicialización VQGanVAE .
ex. coseno sim propuesto en vqgan mejorado
from nuwa_pytorch import VQGanVAE
vae = VQGanVAE (
dim = 64 ,
image_size = 256 ,
num_layers = 4 ,
vq_use_cosine_sim = True
# VectorQuantize will be initialized with use_cosine_sim = True
# https://github.com/lucidrains/vector-quantize-pytorch#cosine-similarity
). cuda () @misc { wu2021nuwa ,
title = { N"UWA: Visual Synthesis Pre-training for Neural visUal World creAtion } ,
author = { Chenfei Wu and Jian Liang and Lei Ji and Fan Yang and Yuejian Fang and Daxin Jiang and Nan Duan } ,
year = { 2021 } ,
eprint = { 2111.12417 } ,
archivePrefix = { arXiv } ,
primaryClass = { cs.CV }
} @misc { esser2021taming ,
title = { Taming Transformers for High-Resolution Image Synthesis } ,
author = { Patrick Esser and Robin Rombach and Björn Ommer } ,
year = { 2021 } ,
eprint = { 2012.09841 } ,
archivePrefix = { arXiv } ,
primaryClass = { cs.CV }
} @misc { iashin2021taming ,
title = { Taming Visually Guided Sound Generation } ,
author = { Vladimir Iashin and Esa Rahtu } ,
year = { 2021 } ,
eprint = { 2110.08791 } ,
archivePrefix = { arXiv } ,
primaryClass = { cs.CV }
} @misc { ding2021cogview ,
title = { CogView: Mastering Text-to-Image Generation via Transformers } ,
author = { Ming Ding and Zhuoyi Yang and Wenyi Hong and Wendi Zheng and Chang Zhou and Da Yin and Junyang Lin and Xu Zou and Zhou Shao and Hongxia Yang and Jie Tang } ,
year = { 2021 } ,
eprint = { 2105.13290 } ,
archivePrefix = { arXiv } ,
primaryClass = { cs.CV }
} @misc { kitaev2020reformer ,
title = { Reformer: The Efficient Transformer } ,
author = { Nikita Kitaev and Łukasz Kaiser and Anselm Levskaya } ,
year = { 2020 } ,
eprint = { 2001.04451 } ,
archivePrefix = { arXiv } ,
primaryClass = { cs.LG }
} @misc { shazeer2020talkingheads ,
title = { Talking-Heads Attention } ,
author = { Noam Shazeer and Zhenzhong Lan and Youlong Cheng and Nan Ding and Le Hou } ,
year = { 2020 } ,
eprint = { 2003.02436 } ,
archivePrefix = { arXiv } ,
primaryClass = { cs.LG }
} @misc { shazeer2020glu ,
title = { GLU Variants Improve Transformer } ,
author = { Noam Shazeer } ,
year = { 2020 } ,
url = { https://arxiv.org/abs/2002.05202 }
} @misc { su2021roformer ,
title = { RoFormer: Enhanced Transformer with Rotary Position Embedding } ,
author = { Jianlin Su and Yu Lu and Shengfeng Pan and Bo Wen and Yunfeng Liu } ,
year = { 2021 } ,
eprint = { 2104.09864 } ,
archivePrefix = { arXiv } ,
primaryClass = { cs.CL }
} @inproceedings { ho2021classifierfree ,
title = { Classifier-Free Diffusion Guidance } ,
author = { Jonathan Ho and Tim Salimans } ,
booktitle = { NeurIPS 2021 Workshop on Deep Generative Models and Downstream Applications } ,
year = { 2021 } ,
url = { https://openreview.net/forum?id=qw8AKxfYbI }
} @misc { liu2021swin ,
title = { Swin Transformer V2: Scaling Up Capacity and Resolution } ,
author = { Ze Liu and Han Hu and Yutong Lin and Zhuliang Yao and Zhenda Xie and Yixuan Wei and Jia Ning and Yue Cao and Zheng Zhang and Li Dong and Furu Wei and Baining Guo } ,
year = { 2021 } ,
eprint = { 2111.09883 } ,
archivePrefix = { arXiv } ,
primaryClass = { cs.CV }
} @misc { crowson2022 ,
author = { Katherine Crowson } ,
url = { https://twitter.com/RiversHaveWings/status/1478093658716966912 }
}La atención es la forma más rara y pura de generosidad. - Simone Weil
