Issues

• where could we get the 1/10 training dataset

  JIA-Lab-research/VisionZip#32 · Filter by author zijian0312 opened on Dec 5, 2025
  3

• Question about the token merge for QwenVL

  JIA-Lab-research/VisionZip#31 · Filter by author nzomi opened on Sep 4, 2025
  2

• error when using in llava-1.6-vicuna-7b

  JIA-Lab-research/VisionZip#30 · Filter by author Sugar929 opened on Jul 22, 2025
  4

• There was no speed improvement and gpu memory did not decrease.

  JIA-Lab-research/VisionZip#29 · Filter by author coder4nlp opened on Jul 2, 2025
  5

• On the Convergence of Attention Positions in SigLIP Across Different Images: A Contrast with CLIP

  JIA-Lab-research/VisionZip#28 · Filter by author HAOobserver opened on Jun 27, 2025
  3

• Visionzip with Qwen2.5-VL

  JIA-Lab-research/VisionZip#26 · Filter by author aurora-xin opened on Jun 12, 2025
  3

• Acceleration effect on Qwen2.5VL

  JIA-Lab-research/VisionZip#25 · Filter by author zry13 opened on Jun 9, 2025
  1

• qwen2.5-vl speed

  JIA-Lab-research/VisionZip#24 · Filter by author vmmm123 opened on Jun 6, 2025
  1

• Whether you could kindly share the training code for VisionZip？

  JIA-Lab-research/VisionZip#23 · Filter by author Blissy-32 opened on May 11, 2025
  1

• Questions regarding token calculation and FLOPs measurement

  JIA-Lab-research/VisionZip#22 · Filter by author zychen8 opened on Apr 23, 2025
  1

• how can I select image tokens to protect privacy

  JIA-Lab-research/VisionZip#21 · Filter by author qianshanwei opened on Apr 6, 2025
  1

• ValueError: not enough values to unpack (expected 2, got 1)

  JIA-Lab-research/VisionZip#20 · Filter by author TDHHTXF opened on Mar 22, 2025
  6

• The following error occurred when I used lmms_eval to evaluate the model

  JIA-Lab-research/VisionZip#19 · Filter by author lihuang258 opened on Mar 19, 2025
  4

• Is visionzip compatible with FlashAttention2?

  JIA-Lab-research/VisionZip#18 · Filter by author Beniko95J opened on Mar 19, 2025
  5

• The setting of Llava-Next

  JIA-Lab-research/VisionZip#17 · Filter by author TungChintao opened on Feb 27, 2025
  4

• Similarity of this paper and LLaVA-PruMerge

  JIA-Lab-research/VisionZip#15 · Filter by author JackYFL opened on Feb 14, 2025
  1

• Reproduce Video-Llava

  JIA-Lab-research/VisionZip#14 · Filter by author liyucheng09 opened on Jan 23, 2025
  3

• why use complex tome code and r params?

  JIA-Lab-research/VisionZip#13 · Filter by author testpppppp opened on Jan 10, 2025
  1

• use visionzip from scratch

  JIA-Lab-research/VisionZip#12 · Filter by author alanshaoTT opened on Dec 29, 2024
  4

• Support the new model

  JIA-Lab-research/VisionZip#9 · Filter by author Yangsenqiao opened on Dec 16, 2024
  2

• support llava_onevision

  JIA-Lab-research/VisionZip#7 · Filter by author defaultak01 opened on Dec 13, 2024
  4

• Can VisionZip be used on VLMs deployed on NPU?

  JIA-Lab-research/VisionZip#6 · Filter by author geniusxxx opened on Dec 12, 2024
  3

• Does it support llava-video?

  JIA-Lab-research/VisionZip#5 · Filter by author zjukop opened on Dec 11, 2024
  3

• Llama3.2-vision and Qwen2-VL Support

  JIA-Lab-research/VisionZip#4 · Filter by author boom-bang opened on Dec 9, 2024
  7