facebookresearch DiT issues

facebookresearch / DiT

Official PyTorch Implementation of "Scalable Diffusion Models with Transformers"

Other

6.37k stars 569 forks source link

issues

Newest

Newest Most commented Recently updated Oldest Least commented Least recently updated

torch run sample_ddp.py fails around 49k

#101 zhengqigao opened 1 week ago
0
How to fine tune a pre-trained checkpoint using custom datasets?

#100 zhengyu-su closed 2 months ago
0
[Question Again] Why DiT-XL/2 takes 119 GFlops to generate 256x256 images?

#99 zheweijushi opened 2 months ago
1
Capi

#98 taigablk opened 2 months ago
1
Does input expects the image to be between 0 to 1?

#97 artemi8 opened 3 months ago
0
high cfg scale，low image diverse

#96 wytcsuch opened 3 months ago
0
Visualing q-k attention of ViT layers

#95 unlugi opened 4 months ago
0
Question on Evaluation

#94 Ting-Justin-Jiang opened 4 months ago
0
Debug

#93 YecanLee closed 5 months ago
1
Research

#92 YecanLee closed 5 months ago
1
Fix reshape dimensions from (h, h) to (h, w) for correct image reshaping

#91 YesianRohn opened 5 months ago
2
Scout

#90 YecanLee closed 6 months ago
1
The model could not be fitted if not predict xstart

#89 JJLi0427 opened 6 months ago
4
Scout

#88 YecanLee closed 6 months ago
1
about fused_attention

#87 AndyCA111 opened 6 months ago
0
Performance for patch size = 1

#86 NrealWJX opened 6 months ago
0
training batch

#85 ZiangWu-77 opened 6 months ago
0
DiT results on CIFAR10

#84 yuanzhi-zhu opened 6 months ago
4
Do the pre-trained DiT chekpoints contain EMA weights?

#83 jmkim0309 opened 7 months ago
1
Clarification on Zero Initialization in FinalLayer of DiT Model

#82 denemmy opened 7 months ago
3
Adapt to Ascend NPU

#81 qyliuAI opened 7 months ago
0
Green image during inference.

#80 constan1 opened 7 months ago
0
Bugs Fixing and Supporting for Multi-nodes

#79 WangWenhao0716 opened 7 months ago
5
time embedding use cat[cos, sin]

#78 shy19960518 opened 7 months ago
0
possible bug for sampling script: y_null = torch.tensor([1000] * n)

#77 forever208 closed 5 months ago
1
Request to adapt to Ascend NPU

#76 qyliuAI opened 7 months ago
0
sample_ddp failed (CUDA error: device-side assert triggered)

#75 forever208 closed 7 months ago
5
How to condition on an image?

#74 amirshamaei opened 8 months ago
3
image generation label doesn't match validation label

#73 eezhang123 closed 8 months ago
0
How do you calculate flops?

#72 xinwangChen opened 8 months ago
4
RuntimeError: CUDA error: CUBLAS_STATUS_EXECUTION_FAILED when calling cublasLtMatmul

#71 wuzelei123 opened 8 months ago
0
Prompt-conditioning model instead of class-conditioning

#70 anarabiyev opened 8 months ago
6
Giving Prompt instead of classes

#69 anarabiyev opened 8 months ago
1
[Snorkell.ai] Please review the generated documentation

#68 sumansaurabh opened 8 months ago
1
[Question] Why DiT-XL/2 takes 119 GFlops to generate 256x256 images?

#67 void-main closed 8 months ago
3
How to fit it to inpainting?

#66 hdjsjyl opened 8 months ago
2
Add frame dimension

#65 FeSens opened 9 months ago
2
No VAE?

#64 yuedajiong opened 9 months ago
3
why report Segmentation fault?

#63 YuyangYin opened 9 months ago
2
typo error fix #24

#62 ghost closed 8 months ago
1
[DiT video]

#61 tomguluson92 opened 9 months ago
3
Trained weights on other ddpm models

#60 oscarwooberry opened 9 months ago
0
Is the model code properly embedding the input tokens?

#59 ey-cai closed 9 months ago
0
Can a pretrained DiT directly be used in classification task?

#58 Colezwhy closed 9 months ago
1
Request to publish cross-attention and in-context-conditioning code?

#57 zf223669 opened 11 months ago
4
Epochs question

#56 l-cr closed 8 months ago
3
Why loss contains both 'mse' and 'vb'

#55 TtuHamg closed 12 months ago
1
when training for cfg, Why only utilize half of the input

#54 jinge170 opened 12 months ago
1
when training for cfg, Why only utilize half of the input？

#53 jinge170 opened 1 year ago
0
Confused with nll in vlb loss implementation.

#52 SeminKim opened 1 year ago
0