Issues

• Verify evals on Papers with Code

  Vision-CAIR/LongVU#52 · Filter by author NielsRogge opened on Aug 28, 2026

• mat1 and mat2 shapes cannot be multiplied (1152x2688 and 256x3584)

  Vision-CAIR/LongVU#51 · Filter by author meng8407 opened on Sep 23, 2025

• May I ask if there are any files that support testing on video benchmarks using lmms-eval?

  Vision-CAIR/LongVU#50 · Filter by author SplendidYuan opened on Aug 14, 2025

• test video between 7b and 3b model, the results are all bad, could you release a 70b model version

  Vision-CAIR/LongVU#49 · Filter by author csjiyw opened on Jun 26, 2025

• Question about Cross-modal Query

  Vision-CAIR/LongVU#48 · Filter by author cokeshao opened on Apr 16, 2025

• May I ask which parts were trained during the fine-tuning phase

  Vision-CAIR/LongVU#47 · Filter by author SplendidYuan opened on Mar 1, 2025
  1

• Question about inference configuration and reproduction of results

  Vision-CAIR/LongVU#46 · Filter by author liys0304 opened on Feb 25, 2025

• Any way to calculate the sum of input tokens

  Vision-CAIR/LongVU#45 · Filter by author Hotpottttt opened on Feb 23, 2025

• Is deepspeed possible for fine-tuning LongVU?

  Vision-CAIR/LongVU#44 · Filter by author tcm03 opened on Feb 3, 2025

• Out of memory when using 8 * A800-80G for video inference

  Vision-CAIR/LongVU#43 · Filter by author caichaoxiang opened on Jan 27, 2025
  3

• Checkpoint loading issue: ValueError

  Vision-CAIR/LongVU#41 · Filter by author andyhandsom6 opened on Jan 24, 2025
  1

• Re. evaluation script

  Vision-CAIR/LongVU#40 · Filter by author pritamqu opened on Jan 20, 2025

• Where to find the answers of egoschema?

  Vision-CAIR/LongVU#39 · Filter by author jyC23333 opened on Jan 14, 2025
  1

• Training Time

  Vision-CAIR/LongVU#38 · Filter by author Hui-design opened on Jan 13, 2025
  3

• Request for requirements.txt for LLaMA-3.2 Image Training

  Vision-CAIR/LongVU#36 · Filter by author tsw123tsw opened on Jan 8, 2025

• Minor bug in MVBench style evaluation when the dataset size isn't divisible by the number of GPUs

  Vision-CAIR/LongVU#35 · Filter by author geomlyd opened on Dec 18, 2024
  1

• Is the parameter "vision_sampler_0.pos_embed_i" learnable?

  Vision-CAIR/LongVU#34 · Filter by author jiinhui opened on Dec 14, 2024

• The weights after sft of video data cannot be inferred

  Vision-CAIR/LongVU#33 · Filter by author orzgugu opened on Dec 11, 2024
  3

• The accuracy of llama2_2_3B on mvBench is a bit low.

  Vision-CAIR/LongVU#32 · Filter by author DuZzzs opened on Dec 6, 2024
  2

• Fine-tuning from an pre-trained image checkpoint has mismatch issue

  Vision-CAIR/LongVU#31 · Filter by author Amshaker opened on Dec 5, 2024

• attention warning and incomplete text description

  Vision-CAIR/LongVU#28 · Filter by author Holmes-GU opened on Dec 1, 2024
  1

• Selective Feature reduction via cross modal query

  Vision-CAIR/LongVU#25 · Filter by author insafim opened on Nov 30, 2024
  1

• Question: Video vs Frames?

  Vision-CAIR/LongVU#24 · Filter by author cbasavaraj opened on Nov 26, 2024
  4

• Need understanding number of tokens

  Vision-CAIR/LongVU#22 · Filter by author miguelcarvtalka opened on Nov 18, 2024
  4

• Did you ever encounter issues with Trainer's _save_checkpoint and FSDP?

  Vision-CAIR/LongVU#21 · Filter by author geomlyd opened on Nov 15, 2024
  14