Skip to content

Issues: tianweiy/DMD2

How about DMD2 on img2img task?
#17 opened May 31, 2024 by dawei03896
Open 2
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

29 Open 29 Closed
29 Open 29 Closed
Author
Filter by author
Loading
Label
Filter by label
Loading
Use alt + click/return to exclude labels
or ⇧ + click/return for logical OR
Projects
Filter by project
Loading
Milestones
Filter by milestone
Loading
Assignee
Filter by who’s assigned
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated Best match
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀

Issues list

Sharing My Understanding of Distribution Matching Loss—Open to Thoughts
#60 opened Jan 27, 2025 by omrastogi
Question: Why is denoising_dict needed when we already have real_train_dict?
#59 opened Jan 21, 2025 by pldlgb
1
No discriminator...
#58 opened Jan 18, 2025 by dreamyou070
1
Is there any SDXL training wandb logs for reference?
#57 opened Jan 8, 2025 by foreverpiano
NaN gradient after loss.backward()
#56 opened Jan 6, 2025 by zhuole1025
Impact of batch size on training performance
#55 opened Dec 18, 2024 by XingtongGe
1
Generator scheduler mismatch in training and inference
#53 opened Dec 5, 2024 by mamicro-li
don't support both options fsdp and gradient_checkpointing . why?
#52 opened Nov 15, 2024 by yanmj0601
2
any loss curves for the first phase of training
#51 opened Nov 6, 2024 by SummerTrains
5
The DMD loss increases during the first stage of training.
#49 opened Sep 13, 2024 by CS123n
2
About Loss
#48 opened Sep 5, 2024 by joelulu
Why use the intermediate estimated x_0|t as the distribution matching input， but not the final image of generator?
#45 opened Aug 30, 2024 by vv12kant
2
Support SD3？
#44 opened Aug 26, 2024 by mliand
1
DMD2 for Flux Dev
#43 opened Aug 24, 2024 by pandayummy
2
train a 4step SDXL got AssertionError: Invalid: `_post_backward_hook_state
#41 opened Aug 9, 2024 by joelulu
8
FSDP is too slow, why not use deepspeed?
#40 opened Aug 8, 2024 by sunhaha123
1
DMD2 and Kandinsky, or Playground 2.5?
#37 opened Jul 21, 2024 by Quick-Eyed-Sky
muti-step model for sdv1.5
#36 opened Jul 20, 2024 by kmpartner
5
LoRA training
#35 opened Jul 19, 2024 by SYZhang0805
4
how long is the extra regression loss pre-training in 1-step SDXL distillation trained?
#34 opened Jul 17, 2024 by grovessss
1
When i applyed the dmd loss to velocity prediction , the loss varies a lot。
#33 opened Jul 16, 2024 by icelighting
4
After training, there are two pytorch_model files. Which one should I use?
#31 opened Jul 15, 2024 by koking0
5
is there plan for releasing code for grad accumulation?
#23 opened Jun 26, 2024 by grovessss
4
Convert 4-step unet into a checkpoint file
#20 opened Jun 11, 2024 by coastbluet
6
sdv15 train with torchrun successful, but sdxl train with fsdp failed.
#18 opened May 31, 2024 by conquer-pan
ProTip! Type g p on any issue or pull request to go back to the pull request listing page.