MiniGPT-4

MiniGPT-4 copied to clipboard

It seems that conda torch 1.12 will be overrided by pip torch 2.0 due to peft requires torch 1.13 So I changes the torch version to 1.13.1 and torchvision to...

sharpbai

Is this normal? I used 13B model and asked the same question about the same pic in the eample, but get a very short answer?

1

Hey guys, I made step by step Installation according to README file, and I can run the project with no error. But somehow the result is really bad, I used...

silent780

Colab Pro - invalid load key

2

When running the demo code, with 13B Vicunia, I encounter an error when trying to load the model checkpoint. Seee the following stdout: 2023-04-25 03:08:17.957341: W tensorflow/compiler/tf2tensorrt/utils/py_utils.cc:38] TF-TRT Warning: Could...

BenHamm

I packaged MiniGPT-4+Vicuna 7B/13B up for image captioning and questioning on replicate.com. Usable as an API and just one-click to the public!

2

https://replicate.com/nelsonjchen/minigpt-4_vicuna-7b https://replicate.com/nelsonjchen/minigpt-4_vicuna-13b https://github.com/nelsonjchen/MiniGPT-4 I repackaged and threw this up on replicate.com for some image captioning duties. I can't say whether it is better or worse than BLIP2 for my use...

nelsonjchen

Is it possible to run on cpu by using 4bit via converting to ggml and quantizing to 4bit?

1

Repos like [Gpt4all](https://github.com/nomic-ai/gpt4all), [llama.cpp](https://github.com/ggerganov/llama.cpp), and [alpaca.cpp](https://github.com/antimatter15/alpaca.cpp) runs on the cpu quite fast while using less resources.

Jake36921

有墙外大神可以帮忙搬运google drive上的文件到墙内云盘吗？

2

或者bt种子什么的，小弟先谢谢了。。 哎，这个翻墙真是麻烦啊。。。

ZanoZ

部署成功了，但是识别图片结果为什么看起来很离谱

22

 识别结果为什么这么离谱呢，是一个小白按着教程部署的，不知道哪里有问题

kimiller

i have two 12GB GPU, and how to use them in model?

4

fact in here, i have two 12gb GPU, and i can use 13B model in theory, but there are no any note about how to use two GPU to inference,...

WakingHours-GitHub

Low GPU utilization on 4090, Windows 10

5

Thanks for the great project, I really like it. Is it just me or is it normal to have only around 35-45% GPU utilization while it's genereting the reply? All...

Korner83

The procedure of downloading Llama

1

Hi, can u tell me the procedure of downloading the original weights of Llama-13b-hf? According to your instruction, I fill in the [form](https://docs.google.com/forms/d/e/1FAIpQLSfqNECQnMkycAp2jP4Z9TFX0cGR4uf7b_fBxjY_OjhJILlKGA/viewform?usp=send_form) in [their website](https://huggingface.co/docs/transformers/main/model_doc/llama), however, they has not...

XiongweiWu