Issues

• Margin-Based Remasking Strategy for Improved Text-to-Image Generation

  Gen-Verse/MMaDA#67 · Filter by author zhongweixie opened on May 13, 2026

• Concetenating input images twice while training does not seem intuitive.

  Gen-Verse/MMaDA#66 · Filter by author kingston-aditya opened on Mar 11, 2026

• Release training code of magvitv2

  Gen-Verse/MMaDA#65 · Filter by author LuBen-001 opened on Jan 24, 2026

• Possible bug in the t2i function regarding the calculation of uncond_prefix

  Gen-Verse/MMaDA#64 · Filter by author fishmeoww opened on Jan 5, 2026

• How to solve the tokenization mismatch?

  Gen-Verse/MMaDA#63 · Filter by author mumu-code opened on Jan 4, 2026

• Multi-node (2 machines x 8 GPUs) training hangs at accelerator.prepare / DeepSpeed init (torch.distributed.DistStoreError timeout)

  Gen-Verse/MMaDA#62 · Filter by author Ffree1 opened on Dec 29, 2025

• Question about EOS token used in training

  Gen-Verse/MMaDA#61 · Filter by author Aran-Dong opened on Nov 25, 2025

• Question about Unified DLLM

  Gen-Verse/MMaDA#60 · Filter by author ZechengLi19 opened on Nov 19, 2025

• releasing UniGRPO RL training code？

  Gen-Verse/MMaDA#59 · Filter by author JoshonSmith opened on Nov 17, 2025
  1

• Question about the pseudocode

  Gen-Verse/MMaDA#58 · Filter by author Ftj072514 opened on Nov 4, 2025

• Could these models support video or multi-pictures inference or train?

  Gen-Verse/MMaDA#57 · Filter by author kawagebo12 opened on Oct 24, 2025
  1

• Training on customized CoT multimodal data

  Gen-Verse/MMaDA#56 · Filter by author o0t1ng0o opened on Oct 21, 2025
  1

• ImportError: cannot import name 'ConfigMixin' from 'models.modeling_utils'

  Gen-Verse/MMaDA#53 · Filter by author Bin-sam opened on Sep 25, 2025

• multi-turn

  Gen-Verse/MMaDA#52 · Filter by author coincidence2020 opened on Sep 10, 2025
  1

• When will MMaDA-8B-Max be released?

  Gen-Verse/MMaDA#51 · Filter by author JoeLeelyf opened on Sep 10, 2025
  1

• Question on Image Generation Task Evaluation

  Gen-Verse/MMaDA#48 · Filter by author bc-li opened on Sep 6, 2025

• Block lengths used during training

  Gen-Verse/MMaDA#46 · Filter by author ToomasRo opened on Aug 9, 2025
  1

• Regarding the lack of thinking content in the MMaDA-8B-MixCoT output in text2image

  Gen-Verse/MMaDA#45 · Filter by author kkkeepgoing opened on Aug 7, 2025
  2

• About Autoregressive

  Gen-Verse/MMaDA#44 · Filter by author ChinChyi opened on Jul 28, 2025
  1

• The multimodal_cot dataset and their captions cannot be downloaded as they are not available for download.

  Gen-Verse/MMaDA#43 · Filter by author kdywt opened on Jul 25, 2025

• lack of answer_lengths_lm in stage2

  Gen-Verse/MMaDA#42 · Filter by author oiabtt opened on Jul 18, 2025

• Collaboration in WeChat -- Code is expired?

  Gen-Verse/MMaDA#41 · Filter by author hamzaamjad opened on Jul 18, 2025
  1

• Reproduction help: Discrepancy in GSM8K reproduction – 40.63 % vs. 73.4 %

  Gen-Verse/MMaDA#40 · Filter by author zrporz opened on Jul 14, 2025
  7

• Evaluation Details on Interleaved Text-image Examples

  Gen-Verse/MMaDA#39 · Filter by author warriors-30 opened on Jul 14, 2025
  1

• 关于二维码过期了

  Gen-Verse/MMaDA#38 · Filter by author XiantaoHu opened on Jul 9, 2025
  2