MiniGPT-4 icon indicating copy to clipboard operation
MiniGPT-4 copied to clipboard

Now you can run MiniGPT-4 locally!!!!

Open cjl09 opened this issue 1 year ago • 17 comments

我将mini-gpt4部署到本地了,现在不需要使用web上传图片了。可批量加载图片,欢迎来玩。 项目地址:[MiniGPT-4-local] (https://github.com/cjl09/MiniGPT-4-local) "I have deployed Mini-GPT4 locally, and now there's no need to use web for uploading images. It can batch load images, and you are welcome to play around with it. Project link: [MiniGPT-4-local] (https://github.com/cjl09/MiniGPT-4-local)

cjl09 avatar Jul 20 '23 15:07 cjl09

i got this error Initializing model usage: ipykernel_launcher.py [-h] --cfg_path CFG_PATH --img_path IMG_PATH [--gpu_id GPU_ID] [--num_beams NUM_BEAMS] [--temperature TEMPERATURE] [--options OPTIONS [OPTIONS ...]] ipykernel_launcher.py: error: the following arguments are required: --cfg_path, --img_path An exception has occurred, use %tb to see the full traceback.

SystemExit: 2

how to add the proper path ?

salmaelgayar avatar Aug 07 '23 11:08 salmaelgayar

你想想问一下 除了demo_local.py文件 对别的文件进行了修改吗

djm0539 avatar Aug 10 '23 10:08 djm0539

2 谢谢ypu的回复, 我的问题是无法运行项目 我在 google collab 上运行它,路径 img 路径和配置路径出错? 您能给我提供运行它的详细步骤,详细路径吗? 或者如何从命令提示符运行它?

[佩卡蒂特](url)

salmaelgayar avatar Aug 10 '23 11:08 salmaelgayar

你想想问一下 除了demo_local.py文件 对别的文件进行了修改吗

没有,只要改demo就行了。

cjl09 avatar Aug 18 '23 03:08 cjl09

2 谢谢ypu的回复, 我的问题是无法运行项目 我在 google collab 上运行它,路径 img 路径和配置路径出错? 您能给我提供运行它的详细步骤,详细路径吗? 或者如何从命令提示符运行它?

佩卡蒂特

cfg怎么给和官方文档里要求的一样,图片路径给到你所有图片所在的图片路径就行。比如:/path/my_image/

cjl09 avatar Aug 18 '23 03:08 cjl09

when i upload my images ,it has the error"RuntimeError: The size of tensor a (4) must match the size of tensor b (3) at non-singleton dimension 0",Do you have any requirements for image size in your code

wwcc1107 avatar Nov 15 '23 03:11 wwcc1107

@wwcc1107 I have encountered this problem and I have solved it, if you are using the example images given by minigpt4, these images are 4 dimensions instead of ordinary RGB, you just need to add this code in the demo_local.py file when the model inputs the image: img = img.convert ('RGB')

Haonote avatar Nov 15 '23 12:11 Haonote

@wwcc1107 I have encountered this problem and I have solved it, if you are using the example images given by minigpt4, these images are 4 dimensions instead of ordinary RGB, you just need to add this code in the demo_local.py file when the model inputs the image: img = img.convert ('RGB')

thank you. but have you meet the other error,because the conversation.py has also been modified,so i have the other error"AttributeError: 'LlamaForCausalLM' object has no attribute 'embed_tokens'"

wwcc1107 avatar Nov 15 '23 12:11 wwcc1107

@wwcc1107 I have encountered this problem and I have solved it, if you are using the example images given by minigpt4, these images are 4 dimensions instead of ordinary RGB, you just need to add this code in the demo_local.py file when the model inputs the image: img = img.convert ('RGB')

i used minigptv2!

wwcc1107 avatar Nov 15 '23 12:11 wwcc1107

请问这个可以实现多轮对话吗? Can this achieve multiple rounds of dialogue?

CLmsx avatar Nov 16 '23 00:11 CLmsx

请问这个可以实现多轮对话吗? Can this achieve multiple rounds of dialogue?

可以,你把图片改成单张,然后自己把history加上,我记得他是做了多轮对话的。

cjl09 avatar Nov 16 '23 05:11 cjl09

请问这个可以实现多轮对话吗? Can this achieve multiple rounds of dialogue?

可以,你把图片改成单张,然后自己把history加上,我记得他是做了多轮对话的。

请问怎样将history加上呢

wwcc1107 avatar Nov 16 '23 09:11 wwcc1107

请问这个可以实现多轮对话吗? Can this achieve multiple rounds of dialogue? 请问你实现这个批量图像也能多轮对话吗

wwcc1107 avatar Nov 23 '23 12:11 wwcc1107

请问这个可以实现多轮对话吗? Can this achieve multiple rounds of dialogue?

可以,你把图片改成单张,然后自己把history加上,我记得他是做了多轮对话的。

请问,怎样实现,我批量推理图片,同时也实现多轮对话呢

wwcc1107 avatar Nov 23 '23 12:11 wwcc1107

请问这个可以实现多轮对话吗?

可以,你把图片改成单张,然后自己把history加上,我记得他是做了多轮对话的。

Is the batch processing of multiple images entered at once?

Spring24ch avatar Jan 26 '24 05:01 Spring24ch

我将mini-gpt4部署到本地了,现在不需要使用web上传图片了。可批量加载图片,欢迎来玩。 项目地址:[MiniGPT-4-local] (https://github.com/cjl09/MiniGPT-4-local) “我已经在本地部署了 Mini-GPT4,现在不需要使用 Web 上传图片了。它可以批量加载图像,欢迎您使用它。项目链接:[MiniGPT-4-local](https://github.com/cjl09/MiniGPT-4-local)

您好,请问批量加载图片是一次处理多张图片吗?怎么实现的呢,对那部分进行了修改呢?

Spring24ch avatar Jan 26 '24 05:01 Spring24ch

@cjl09 您好,感谢您的分享,想请教下,我在用minigpt-4版本的时候,用您的代码,出现问题说img是Image类型,然后我手动用F.to_tensor转化成了tensor,出现新的报错RuntimeError: Sizes of tensors must match except in dimension 1. Expected size 1 but got size 3 for tensor number 1 in the list.请问这个是因为我没有用v2的版本原因吗

ZhouCX117 avatar May 01 '24 12:05 ZhouCX117