QwenLM/Qwen2-VL

Issues

• qwen2-vl-72B in V100 gptq error CUDA out of memory

  #619 opened 5 days ago by diandianliu
  4

• 通过VLLM推理，可以支持多图理解吗？

  #612 opened 12 days ago by wangbing35
  1

• 训练的时候，batch 内的所有图片会拼到一起？

  #595 opened 18 days ago by lambda-lee
  1

• 请教下vision-rope和m-rope该如何理解

  #621 opened 5 days ago by Lier007
  0

• What are the algorithms behind the reasoning ability of QVQ and QWQ?

  #620 opened 5 days ago by ConflictAvoidance
  0

• visual grounding poor performance after fine-tuning

  #584 opened 25 days ago by pengzhansun
  2

• 关于使用AutoGPTQ量化的问题

  #594 opened 18 days ago by liangyi-qianwan
  2

• IndexError: index 0 is out of bounds for dimension 0 with size 0 for " if cache_position is None or (cache_position is not None and cache_position[0] == 0):"

  #596 opened 16 days ago by xjywhu
  7

• Is there a plan to open source the Naive Dynamic Resolution ViT base model?

  #618 opened 5 days ago by quinwu
  0

• Hidden states of image patches

  #617 opened 5 days ago by debraj135
  0

• 怎么让其做目标检测，输出图像中物体的坐标？

  #588 opened 20 days ago by MathamPollard
  2

• 使用量化模型Qwen2-VL-7B-Instruct-AWQ 安装autoawq以后 出现cannot import name 'shard_checkpoint' from 'transformers.modeling_utils' (F:\ai\qwen_vl_simple_api-main\py310\lib\site-packages\transformers\modeling_utils.py)错误

  #589 opened 19 days ago by Joneshan907
  1

• Qwen2vl vllm grounding任务效果不如transformers推理

  #601 opened 14 days ago by nansanhao
  3

• 论文里的例子无法复现

  #605 opened 13 days ago by xiexie123
  3

• Increase batch size results in a non-linear increase in computation time

  #616 opened 9 days ago by Wonder1905
  0

• how to get the past_key_values

  #598 opened 14 days ago by WheatJH
  1

• Quick Question on Evaluation Method

  #615 opened 10 days ago by Ivesfu
  0

• how to implement the phrase 1 pretraining by contrasive learning?

  #614 opened 10 days ago by indeedzcr
  0

• Interleaved images and text

  #587 opened 19 days ago by pbarker
  1

• How to specify the output as json format? 如何指定用json格式输出

  #610 opened 11 days ago by yyxssm
  1

• docker版本启动成功，提交prompt报错

  #613 opened 11 days ago by akfheaven
  0

• 增大model_max_length，为什么原输入的sequence length会增加好几倍？

  #611 opened 12 days ago by Hothan01
  0

• Catastrophic forgetting

  #600 opened 15 days ago by HaithemH
  1

• 如何给字典新增加一些special token, 之前qwen-vl 字典里面还有<|extra_*|> 可以用。请问qwen2-vl 如何增加或者使用特殊符号呢？

  #603 opened 13 days ago by qxhhou
  1

• Qwen2-VL is supported in PaddleMIX by Paddle Team!

  #608 opened 13 days ago by luyao-cv
  0

• 使用vllm部署的服务会因某些具体的图片输入而整个服务挂掉

  #607 opened 13 days ago by peixin-lin
  0

• Qwen2-VL-7B-Instruct: 7% of images cannot be described correctly

  #606 opened 13 days ago by nifengxiaofei
  0

• sft阶段max_length参数

  #604 opened 13 days ago by endlesstalking
  0

• 容器部署Qwen2-VL-2B-Instruct 打包镜像迁移到无互联网设备调用报错

  #602 opened 14 days ago by sujunze
  1

• qwen2-vl-2b-instruct版本微调自己的数据集之后，训练正常，推理报错：IndexError: index 1 is out of bounds for dimension 0 with size 1

  #599 opened 15 days ago by xgl0626
  1

• AttributeError: 'Qwen2VLForConditionalGeneration' object has no attribute 'quantize'. Did you mean: 'dequantize'?

  #597 opened 15 days ago by liangyi-qianwan
  0

• RuntimeError: Key and Value must have the same sequence length

  #593 opened 18 days ago by fuxuelinwudi
  1

• OCR识别任务中经常出现一串数字多0、少0的问题，是否有好的微调方案呢？

  #592 opened 18 days ago by Apricot1225
  0

• How to change patch size for fine tuning

  #591 opened 18 days ago by saribolagh
  0

• 请问用这个模型训练grounding任务之后，怎么计算评价指标

  #590 opened 19 days ago by yfllllll
  0

• Correct enviroment for custom Qwen2-VL GPTQ quantization

  #580 opened a month ago by bhavyajoshi-mahindra
  0

• 测试发现某些提示词加图片会出现输出重复问题

  #575 opened a month ago by Mylszd
  1

• llamafactory微调后无法加载模型

  #581 opened a month ago by Youmo1
  5

• 使用LLaMA-Factory训练，loss为0，但预测结果与真实结果差很多。

  #586 opened 22 days ago by jjl001
  1

• batch inference: the second output contains the context of the first image

  #583 opened 25 days ago by pianogGG
  0

• 无法复现Qwen2-VL-7B-Instruct在RealworldQA上的结果。

  #585 opened 24 days ago by shaonan1993
  0

• Is it normal for Qwen2-VL-72B-Instruct served using vllm to be 12x faster than huggingface pipeline?

  #582 opened a month ago by yxchng
  0

• 如何在垂直领域上继续预训练

  #579 opened a month ago by ybshaw
  0

• In-context learning example

  #578 opened a month ago by RongKaiWeskerMA
  0

• 多轮对话

  #577 opened a month ago by aaa1233211
  1

• Batch Inference

  #576 opened a month ago by Wangbiao2
  0

• missing requirement file

  #574 opened a month ago by habaohaba
  0

• Single text prediction changes.

  #573 opened a month ago by anshkumar
  0

• 视频清晰度和时长太长，使用7B-instruct推理时显存爆了，请问使用processor如何滑动的方式做识别呢？

  #572 opened a month ago by jxyxiangyu
  0

• flash_attn_2_cuda: 找不到指定的模块

  #571 opened a month ago by joseph16388
  0