OpenBMB MiniCPM-V issues

OpenBMB / MiniCPM-V

MiniCPM-V 2.6: A GPT-4V Level MLLM for Single Image, Multi Image and Video on Your Phone

Apache License 2.0

12.76k stars 894 forks source link

issues

Newest

Newest Most commented Recently updated Oldest Least commented Least recently updated

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 alphanlp opened 1 day ago
0
[BUG] <title>Is there any restriction on images or datasets for fine-tuning vision encoder or llm of MiniCPM-v2.6?

#683 Wytheyayaya opened 3 days ago
0
💡 [REQUEST] - <title>給模型一個全新的名字跟形象

#682 aybs2060 opened 3 days ago
0
💡 [REQUEST] - <few shot learning（in-context learning）评估脚本或评估设置>

#681 ColorDavid opened 5 days ago
0
[BUG] TypeError: 'Image' object is not iterable

#680 tattain404 closed 5 days ago
2
[BUG] <title>多图单轮对话全量SFT，loss 变为零

#679 Evan9978 opened 6 days ago
3
💡 [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 evanlin88 opened 1 week ago
0
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 evanlin88 opened 1 week ago
0
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 ColorDavid opened 1 week ago
1
推理时，能否直接输出视觉特征？

#675 kiter-zero opened 1 week ago
0
[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 lh3707 opened 1 week ago
0
请问sft训练数据中能混入无图数据吗

#673 128Ghe980 opened 1 week ago
1
💡 [REQUEST] - <title>预训练相关

#672 zhangzhixun1999 opened 1 week ago
1
Can you provide pre-training files for MiniiCPM-V-2.0 version?

#671 Leke-G opened 2 weeks ago
0
微调中断后如何继续训练

#670 wanghaolan123 opened 2 weeks ago
1
💡 [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 Johnson-yue opened 2 weeks ago
3
Accelerator' object has no attribute 'deepspeed_engine_wrapped'

#668 alanMachineLeraning opened 2 weeks ago
0
RuntimeError: Input type (torch.cuda.HalfTensor) and weight type (torch.HalfTensor) should be the same

#667 alanMachineLeraning opened 2 weeks ago
0
[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 xueyuG opened 2 weeks ago
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 flawsss opened 2 weeks ago
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 hshc123 closed 2 weeks ago
0
[BUG] <title> Multi turns conversation 's output is not complete!

#663 FantasticZihao opened 2 weeks ago
5
[BUG] <title>token截断导致训练label被删除

#662 tong-1989 opened 2 weeks ago
2
MiniCPM-V-2.6 promp template的明文格式？

#661 apepkuss opened 3 weeks ago
3
[BUG] 为什么如何调整参数，显存占用都是接近80G？

#660 DankoZhang opened 3 weeks ago
2
LLM部署MiniCPM-V-2.6并发调用推理报错

#659 cheng358 opened 3 weeks ago
1
[BUG] WebDemo中使用Few Shot功能，如果不传图片无法正常使用

#658 gongjimin opened 3 weeks ago
0
[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 friendmine opened 3 weeks ago
1
[BUG] <title> 使用llama.cpp遇到 Missing required key: general.description

#656 friendmine closed 3 weeks ago
1
请问为什么MiniCPM-V-2_6-int4的参数量只有4.76B？

#655 quanmou opened 4 weeks ago
1
[BUG] MiniCPM-V-2.6, with no image input, answers "As a large language model trained by OpenAI"

#654 emanuelevivoli opened 1 month ago
1
[BUG] <title> Is there a way to output logits?

#653 zhaowenZhou opened 1 month ago
1
[BUG] <title> finetune/dataset.py 有bug

#652 bingo-todd opened 1 month ago
4
[BUG] <title> lora脚本微调错误

#651 sanhyzx opened 1 month ago
1
[BUG] <title>llama.cpp CLIP cannot encode some images after building graph CLIP无法编码特定图片

#650 yzyhyt opened 1 month ago
0
swift显存控制不住的涨

#649 2013358072 opened 1 month ago
3
deepspeed 配置文件bug

#648 BajieZheng opened 1 month ago
0
[BUG] <title> MiniCPM-V 2.6 在mathvista-minitest中测试性能远不如report

#647 WentaoTan opened 1 month ago
0
[BUG] <title>【多图推理】请问如何使用transformer进行minicpm-v 2.6的多图推理？

#646 popoyaya opened 1 month ago
1
[BUG] <title>data fetch error

#645 ydaiwxl closed 1 month ago
2
[BUG] <title> 请问在用deepspeed zero3 训练的过程因为minicpm navit的逻辑会导致不同rank上的image feature size 不同这样会hang住，这个是怎么解决的？zero2是没有问题的

#644 royzhang12 opened 1 month ago
2
[BUG] <title> Inference error. Replacing the LLM part with Llama-3.1 70B quantized causing error ( RuntimeError: shape mismatch: value tensor of shape [1024] cannot be broadcast to indexing result of shape [1025] )

#643 CCRss opened 1 month ago
7
[Fix] Trainer interface error when eval minicpm-v-2.6

#642 moonmengmeng opened 1 month ago
0
Highlighting detected features in an Image with a boundary box

#641 SrikanthChellappa opened 1 month ago
2
[vllm] - 一张A10，vllm api方式启动，报显存不足

#640 thend-wk opened 1 month ago
3
[BUG] 微信群加不了了，抓紧在更新一个吧

#639 ychy00001 closed 1 month ago
1
lora微调过程中卡住不动

#638 ljhjxt closed 1 month ago
1
[BUG] <title> no gradient when only tune encoder part.

#637 YaxinLi0-0 closed 1 month ago
0
[vllm] - 请求优化现有的batch inference模块

#636 Hibari36 opened 1 month ago
1
[BUG] <title>NotImplementedError: Cannot copy out of meta tensor; no data! Please use torch.nn.Module.to_empty() instead of torch.nn.Module.to() when moving module from meta to a different device.

#635 wshiman opened 1 month ago
1