Issues

• Step Loss Trend Issue in Training

  SalesforceAIResearch/DiffusionDPO#23 · Filter by author lucasxu777 opened on May 12, 2025
  2

• Datasets No Longer Available

  SalesforceAIResearch/DiffusionDPO#22 · Filter by author lucasxu777 opened on May 2, 2025
  2

• The gradient norm is too large

  SalesforceAIResearch/DiffusionDPO#21 · Filter by author hitachinsk opened on Mar 14, 2025

• About the derivation in Supplementary Material S2

  SalesforceAIResearch/DiffusionDPO#20 · Filter by author Ideal-Bushgorman opened on Feb 27, 2025

• Dependency Issues During Environment Installation

  SalesforceAIResearch/DiffusionDPO#19 · Filter by author djFatNerd opened on Feb 16, 2025
  1

• How long does it take to train SDXL?

  SalesforceAIResearch/DiffusionDPO#18 · Filter by author wxbs opened on Feb 12, 2025

• the training_loss_l and the training_loss_w both went to a big value

  SalesforceAIResearch/DiffusionDPO#17 · Filter by author wxbs opened on Dec 29, 2024

• Potential Issue on data loader in distributed setting.

  SalesforceAIResearch/DiffusionDPO#15 · Filter by author dingyuan-shi opened on Sep 16, 2024
  4

• Why local batch size should be 1? A bigger local batch size may cause a performance degradation?

  SalesforceAIResearch/DiffusionDPO#13 · Filter by author DwanZhang-AI opened on Jul 26, 2024
  2

• The learning rate in the paper and the code are inconsistent

  SalesforceAIResearch/DiffusionDPO#12 · Filter by author kjzju opened on Jun 8, 2024
  1

• I can't reproduce the results of the paper

  SalesforceAIResearch/DiffusionDPO#11 · Filter by author Mowenyii opened on Jun 2, 2024
  5

• The trick of the filtering Pick-a-pic dataset

  SalesforceAIResearch/DiffusionDPO#10 · Filter by author zengziru opened on May 21, 2024
  2

• Distributed training

  SalesforceAIResearch/DiffusionDPO#9 · Filter by author dain5832 opened on May 21, 2024

• Significant Issue with Hugging Face Accelerate Version

  SalesforceAIResearch/DiffusionDPO#8 · Filter by author YangYongJin opened on Apr 28, 2024
  4

• How do I get formula (9) using ELBO?

  SalesforceAIResearch/DiffusionDPO#6 · Filter by author Mowenyii opened on Apr 10, 2024

• Why add same noise to pairwise images?

  SalesforceAIResearch/DiffusionDPO#3 · Filter by author ChenDRAG opened on Mar 10, 2024
  1

• The loss becomes almost 0 after 1 epoch of training.

  SalesforceAIResearch/DiffusionDPO#2 · Filter by author ChenDRAG opened on Mar 10, 2024
  3

• question regarding loss

  SalesforceAIResearch/DiffusionDPO#1 · Filter by author orgw opened on Feb 13, 2024
  4