Repository Issue Activity (beta)

#2113 Verify evals on Papers with Code

#1712 Fine-tuning data format for video with non-uniform timestamps?

#2110 pip install qwen-vl-utils==0.0.14

#1896 坐标识别完全不准确

#2105 Has anyone successfully extracted raw visual tokens before Qwen 3 VL projection?

#1756 vllm批量推理时报错CUDA error: an illegal memory access was encountered

#1485 Qwen-VL 系列坐标体系的变化解惑 | The Ever-Changing Coordinate System Has Been Confusing Me

#2112 Add REFUTE scientific critique + calibration benchmark

#2111 Enhance README with reference links

#1581 Reproduce Qwen3-VL-30B-A3B-Instruct on MMMU

#1178 About Android World Testing Result

#2109 qwen-vl-utils/pyproject.toml [project.urls] still point to Qwen2-VL repo

#2108 Qwen3-VL单卡训练和多卡训练速度相当甚至更快

#2107 Unz2359

#2095 Output text can't exceed 375 words

#2093 why "loss" is so big when training Qwen3-30B-A3B

#2102 Qwen3-VL Grounding任务的bbox需要做什么处理吗？

#2104 离线推理报错Engine core initialization failed. See root cause above. Failed core proc(s): {}

#663 图像grounding微调

#2014 qwen3-vl-8b-instruct在mmstar benchmark上的复现问题

#2101 Question: Recommended output format for oriented bounding boxes in Qwen3-VL grounding

#2099 `video_min_pixels` / `video_max_pixels` are silently ignored during video fine-tuning

#1430 Batch vs. sequential inference give different hidden states (diff starts at first layer, ~1e-3–1e-4)

#1463 Flash Attention 影响批量推理结果

#1956 Qwen2.5-VL-7B-Instruct single/batched inference mismatch