OpenBMB/MiniCPM-V

Issues

• README vllm section improved.

  #701 opened 16 hours ago by singing-cat
  0

• ❓ [QUESTION] - Help Needed: Modify Refusal Removal Scripts to Work with MiniCPM-V-2_6

  #700 opened 2 days ago by NullEqualsZero
  0

• [BUG] why is the vocab_size of MiniCPM-V-2_6 not the same as Qwen2-7B?

  #699 opened 3 days ago by LSC527
  0

• [BUG] <title>TypeError: CPMTrainer.training_step() takes 3 positional arguments but 4 were given

  #698 opened 4 days ago by FrancisFan98
  1

• [BUG] 为什么如何调整参数，显存占用都是接近80G？

  #660 opened a month ago by DankoZhang
  3

• [llama] <.so版本不一致>

  #697 opened 5 days ago by Heiheicxx
  0

• 💡 [REQUEST] - <title>MiniCPM-V 2.6 is supported in PaddleMIX by Paddle Team!

  #695 opened 5 days ago by luyao-cv
  1

• [BUG] <title>IsADirectoryError: [Errno 21] Is a directory: '../new_LLM_2024_12_09/minicpm_original/Minicpm_v2.6_moscol/minicpm-v-2_6/assets'

  #693 opened 6 days ago by sjjjoaps
  0

• pydantic_core._pydantic_core.ValidationError: 1 validation error for DeepSpeedZeroConfig

  #694 opened 6 days ago by xueyuG
  0

• 💡 [REQUEST] - <title>讀取並理解文件PDF

  #687 opened 17 days ago by aybs2060
  2

• [BUG] Visual embedding vector size

  #690 opened 13 days ago by rokopi-byte
  1

• 模型训练后推理出现乱码

  #688 opened 14 days ago by MickeyFei
  1

• 关于VisionEncoder里面的VIT: SigLIP SoViT-400m/14

  #692 opened 12 days ago by henrycCoder
  0

• 💡 [REQUEST] - 使用OCR识别场景中，若图片有手写删除的痕迹，是否可以准确理解并屏蔽不需的信息？

  #678 opened 12 days ago by evanlin88
  2

• 提示词“使用OCR识别”，识别结果有：正文/标题的Markdown格式输出，是否可以通过提示词屏蔽？

  #677 opened 12 days ago by evanlin88
  2

• [ollama] - 请问这个数据有误吗？

  #691 opened 12 days ago by auroralzone
  0

• [BUG] <title>

  #689 opened 13 days ago by 66RomanReigns
  0

• 💡 [REQUEST] - <title>給模型一個全新的名字跟形象

  #682 opened 25 days ago by aybs2060
  2

• [BUG] <title>多图单轮对话全量SFT，loss 变为零

  #679 opened a month ago by Evan9978
  5

• minicpm-V2.6提取图片中的文本信息怎么输出置信度呢

  #686 opened 18 days ago by zxy1029
  0

• 推理时，能否直接输出视觉特征？

  #675 opened 20 days ago by kiter-zero
  1

• 请问sft训练数据中能混入无图数据吗

  #673 opened 20 days ago by 128Ghe980
  1

• [vllm] - <title> vllm 部署了API接口，示例里边有单张图片的推理，但是多张图片的推理，有参考示例吗？

  #674 opened a month ago by lh3707
  1

• [BUG] <title>Is there any restriction on images or datasets for fine-tuning vision encoder or llm of MiniCPM-v2.6?

  #683 opened 25 days ago by Wytheyayaya
  1

• 求教：为什么Resampler后，还可以做OCR识别，感觉已经在压缩的过程中，丢失了信息，做OCR识别任务会降低精度

  #684 opened 23 days ago by alphanlp
  1

• [BUG] <title>MiniCPM-V 2.6跑demo，本身机器显存8G，使用cuda提示显存不够

  #685 opened 20 days ago by nw-wf
  1

• [BUG] <title> lora脚本微调错误

  #651 opened 2 months ago by sanhyzx
  1

• 💡 [REQUEST] - <few shot learning（in-context learning）评估脚本或评估设置>

  #681 opened a month ago by ColorDavid
  0

• [BUG] TypeError: 'Image' object is not iterable

  #680 opened a month ago by tattain404
  2

• 你好，由于 minicpm2.6 的 pretrain ckpt 没有开源，所以这部分评测代码也没有提供，我们的 few-shot 推理方式可以参考 readme 部分的 icl 推理方式。

  #676 opened a month ago by ColorDavid
  1

• [BUG] <title> Multi turns conversation 's output is not complete!

  #663 opened a month ago by FantasticZihao
  5

• 💡 [REQUEST] - 希望能够实现函数调用的功能，并且提供说明文档，方便第三方写插件

  #669 opened a month ago by Johnson-yue
  3

• MiniCPM-V-2.6 promp template的明文格式？

  #661 opened a month ago by apepkuss
  3

• 💡 [REQUEST] - <title>预训练相关

  #672 opened a month ago by zhangzhixun1999
  1

• 微调中断后如何继续训练

  #670 opened a month ago by wanghaolan123
  1

• Can you provide pre-training files for MiniiCPM-V-2.0 version?

  #671 opened a month ago by Leke-G
  0

• Accelerator' object has no attribute 'deepspeed_engine_wrapped'

  #668 opened a month ago by alanMachineLeraning
  0

• RuntimeError: Input type (torch.cuda.HalfTensor) and weight type (torch.HalfTensor) should be the same

  #667 opened a month ago by alanMachineLeraning
  0

• [BUG] <title>token截断导致训练label被删除

  #662 opened a month ago by tong-1989
  2

• [BUG] <title> 使用git拉取的DeepSpeed仓库按照命令pip install e . 安装的deepspeed版本号为deepspeed-0.15.4+unknown，后面运行微调命令时，遇到报错，不知道是不是和这个unknown的出现有关系

  #666 opened a month ago by xueyuG
  0

• [BUG] <title>为什么对于低分辨率256*256的图像输入的时候，调整max_slice_number = 1/9，对于结果有巨大的差异？

  #665 opened a month ago by flawsss
  1

• [BUG] 视频微调lora出现警告：Token indices sequence length is longer than the specified maximum sequence length for this model (7902 > 2048). Running this sequence through the model will result in indexing errors

  #664 opened a month ago by hshc123
  0

• LLM部署MiniCPM-V-2.6并发调用推理报错

  #659 opened 2 months ago by cheng358
  1

• [BUG] <title> Is there a way to output logits?

  #653 opened 2 months ago by zhaowenZhou
  1

• [BUG] MiniCPM-V-2.6, with no image input, answers "As a large language model trained by OpenAI"

  #654 opened 2 months ago by emanuelevivoli
  1

• 请问为什么MiniCPM-V-2_6-int4的参数量只有4.76B？

  #655 opened 2 months ago by quanmou
  1

• [llamacpp] - <title> 为什么llamacpp执行量化模型还要指定一个 f16的 mmproj-model-f16.gguf ?

  #657 opened 2 months ago by friendmine
  1

• [BUG] WebDemo中使用Few Shot功能，如果不传图片无法正常使用

  #658 opened 2 months ago by gongjimin
  0

• [BUG] <title> finetune/dataset.py 有bug

  #652 opened 2 months ago by bingo-todd
  4

• [BUG] <title> 使用llama.cpp遇到 Missing required key: general.description

  #656 opened 2 months ago by friendmine
  1