mmdit

MMDiT

Implementierung einer einzelnen Schicht des MMDiT, vorgeschlagen von Esser et al. in Stable Diffusion 3, in Pytorch

Neben einer reinen Reproduktion werde ich auch auf > 2 Modalitäten verallgemeinern, da ich mir ein MMDiT für Bilder, Audio und Text vorstellen kann.

Bietet auch eine improvisierte Variante der Selbstaufmerksamkeit, die die zu verwendenden Gewichte durch erlerntes Gating adaptiv auswählt. Diese Idee entstand aus adaptiven Faltungen, die von Kang et al. angewendet wurden. für GigaGAN.

Installieren

$ pip install mmdit

Verwendung

 import torch
from mmdit import MMDiTBlock

# define mm dit block

block = MMDiTBlock (
    dim_joint_attn = 512 ,
    dim_cond = 256 ,
    dim_text = 768 ,
    dim_image = 512 ,
    qk_rmsnorm = True
)

# mock inputs

time_cond = torch . randn ( 2 , 256 )

text_tokens = torch . randn ( 2 , 512 , 768 )
text_mask = torch . ones (( 2 , 512 )). bool ()

image_tokens = torch . randn ( 2 , 1024 , 512 )

# single block forward

text_tokens_next , image_tokens_next = block (
    time_cond = time_cond ,
    text_tokens = text_tokens ,
    text_mask = text_mask ,
    image_tokens = image_tokens
)

Als solche kann eine verallgemeinerte Version verwendet werden

 import torch
from mmdit . mmdit_generalized_pytorch import MMDiT

mmdit = MMDiT (
    depth = 2 , 
    dim_modalities = ( 768 , 512 , 384 ),
    dim_joint_attn = 512 ,
    dim_cond = 256 ,
    qk_rmsnorm = True
)

# mock inputs

time_cond = torch . randn ( 2 , 256 )

text_tokens = torch . randn ( 2 , 512 , 768 )
text_mask = torch . ones (( 2 , 512 )). bool ()

video_tokens = torch . randn ( 2 , 1024 , 512 )

audio_tokens = torch . randn ( 2 , 256 , 384 )

# forward

text_tokens , video_tokens , audio_tokens = mmdit (
    modality_tokens = ( text_tokens , video_tokens , audio_tokens ),
    modality_masks = ( text_mask , None , None ),
    time_cond = time_cond ,
)

Zitate

 @article { Esser2024ScalingRF ,
    title   = { Scaling Rectified Flow Transformers for High-Resolution Image Synthesis } ,
    author  = { Patrick Esser and Sumith Kulal and A. Blattmann and Rahim Entezari and Jonas Muller and Harry Saini and Yam Levi and Dominik Lorenz and Axel Sauer and Frederic Boesel and Dustin Podell and Tim Dockhorn and Zion English and Kyle Lacey and Alex Goodwin and Yannik Marek and Robin Rombach } ,
    journal = { ArXiv } ,
    year    = { 2024 } ,
    volume  = { abs/2403.03206 } ,
    url     = { https://api.semanticscholar.org/CorpusID:268247980 }
}

 @inproceedings { Darcet2023VisionTN ,
    title   = { Vision Transformers Need Registers } ,
    author  = { Timoth'ee Darcet and Maxime Oquab and Julien Mairal and Piotr Bojanowski } ,
    year    = { 2023 } ,
    url     = { https://api.semanticscholar.org/CorpusID:263134283 }
}

 @article { Zhu2024HyperConnections ,
    title   = { Hyper-Connections } ,
    author  = { Defa Zhu and Hongzhi Huang and Zihao Huang and Yutao Zeng and Yunyao Mao and Banggu Wu and Qiyang Min and Xun Zhou } ,
    journal = { ArXiv } ,
    year    = { 2024 } ,
    volume  = { abs/2409.19606 } ,
    url     = { https://api.semanticscholar.org/CorpusID:272987528 }
}