Finetune stable video diffusion, but the output video frames are all the same as the first frame, why ? #6418

Activity

1. jiarenyf

   added

   bugSomething isn't workingSomething isn't working

   on Jan 1, 2024
2. sayakpaul commented on Jan 2, 2024

   

   sayakpaul

   on Jan 2, 2024

   Member

   More actions

   This is not a library-specific issue. Please use the Discord forum (https://hf.co/join/discord) for such issues.
3. sayakpaul

   closed this as completedon Jan 2, 2024
4. SystemErrorWang commented on Jan 12, 2024

   

   SystemErrorWang

   on Jan 12, 2024

   More actions

   Hello @jiarenyf , would you please tell me how did you finetune the stable video diffusion? what scheduler did you use, and did you get any problem when calling "add_noise" function during training? Or would it be convenient for you to share the environment or version of diffusers? Thank you a lot.
5. sayakpaul commented on Jan 12, 2024

   

   sayakpaul

   on Jan 12, 2024

   Member

   More actions

   https://github.com/pixeli99/SVD_Xtend
6. steverhoades

   added 2 commits that reference this issue on Jan 13, 2024

   Fixes huggingface#6418 Advanced Dreambooth LoRa Training

   This issue will close once commit b238c1e is merged into the 'main' branch.

   b238c1e

   Fixes huggingface#6418 Advanced Dreambooth LoRa Training

   This issue will close once commit 6ccc66c is merged into the 'main' branch.

   6ccc66c
7. sayakpaul

   added a commit that references this issue on Jan 16, 2024

   Fixes training resuming: Advanced Dreambooth LoRa Training (#6566)

   ...

   This issue will close once commit 181280b is merged into the 'main' branch.

   181280b
8. AmericanPresidentJimmyCarter

   added a commit that references this issue on Apr 26, 2024

   Fixes training resuming: Advanced Dreambooth LoRa Training (huggingfa…

   ...

   This issue will close once commit 60877b3 is merged into the 'main' branch.

   60877b3