minigpt-4 本地部署

minigpt-4 git主页。
笔者参考了深度学习笔记–本地部署Mini-GPT4，使用了http链接，

huggingface下载llama和vicuna权重的download.txt分别如下：

http://huggingface.co/decapoda-research/llama-7b-hf/resolve/main/.gitattributes
http://huggingface.co/decapoda-research/llama-7b-hf/resolve/main/LICENSE
http://huggingface.co/decapoda-research/llama-7b-hf/resolve/main/README.md
http://huggingface.co/decapoda-research/llama-7b-hf/resolve/main/config.json
http://huggingface.co/decapoda-research/llama-7b-hf/resolve/main/generation_config.json
http://huggingface.co/decapoda-research/llama-7b-hf/resolve/main/pytorch_model-00001-of-00033.bin
http://huggingface.co/decapoda-research/llama-7b-hf/resolve/main/pytorch_model-00002-of-00033.bin
http://huggingface.co/decapoda-research/llama-7b-hf/resolve/main/pytorch_model-00003-of-00033.bin
http://huggingface.co/decapoda-research/llama-7b-hf/resolve/main/pytorch_model-00004-of-00033.bin
http://huggingface.co/decapoda-research/llama-7b-hf/resolve/main/pytorch_model-00005-of-00033.bin
http://huggingface.co/decapoda-research/llama-7b-hf/resolve/main/pytorch_model-00006-of-00033.bin
http://huggingface.co/decapoda-research/llama-7b-hf/resolve/main/pytorch_model-00007-of-00033.bin
http://huggingface.co/decapoda-research/llama-7b-hf/resolve/main/pytorch_model-00008-of-00033.bin
http://huggingface.co/decapoda-research/llama-7b-hf/resolve/main/pytorch_model-00009-of-00033.bin
http://huggingface.co/decapoda-research/llama-7b-hf/resolve/main/pytorch_model-00010-of-00033.bin
http://huggingface.co/decapoda-research/llama-7b-hf/resolve/main/pytorch_model-00011-of-00033.bin
http://huggingface.co/decapoda-research/llama-7b-hf/resolve/main/pytorch_model-00012-of-00033.bin
http://huggingface.co/decapoda-research/llama-7b-hf/resolve/main/pytorch_model-00013-of-00033.bin
http://huggingface.co/decapoda-research/llama-7b-hf/resolve/main/pytorch_model-00014-of-00033.bin
http://huggingface.co/decapoda-research/llama-7b-hf/resolve/main/pytorch_model-00015-of-00033.bin
http://huggingface.co/decapoda-research/llama-7b-hf/resolve/main/pytorch_model-00016-of-00033.bin
http://huggingface.co/decapoda-research/llama-7b-hf/resolve/main/pytorch_model-00017-of-00033.bin
http://huggingface.co/decapoda-research/llama-7b-hf/resolve/main/pytorch_model-00018-of-00033.bin
http://huggingface.co/decapoda-research/llama-7b-hf/resolve/main/pytorch_model-00019-of-00033.bin
http://huggingface.co/decapoda-research/llama-7b-hf/resolve/main/pytorch_model-00020-of-00033.bin
http://huggingface.co/decapoda-research/llama-7b-hf/resolve/main/pytorch_model-00021-of-00033.bin
http://huggingface.co/decapoda-research/llama-7b-hf/resolve/main/pytorch_model-00022-of-00033.bin
http://huggingface.co/decapoda-research/llama-7b-hf/resolve/main/pytorch_model-00023-of-00033.bin
http://huggingface.co/decapoda-research/llama-7b-hf/resolve/main/pytorch_model-00024-of-00033.bin
http://huggingface.co/decapoda-research/llama-7b-hf/resolve/main/pytorch_model-00025-of-00033.bin
http://huggingface.co/decapoda-research/llama-7b-hf/resolve/main/pytorch_model-00026-of-00033.bin
http://huggingface.co/decapoda-research/llama-7b-hf/resolve/main/pytorch_model-00027-of-00033.bin
http://huggingface.co/decapoda-research/llama-7b-hf/resolve/main/pytorch_model-00028-of-00033.bin
http://huggingface.co/decapoda-research/llama-7b-hf/resolve/main/pytorch_model-00029-of-00033.bin
http://huggingface.co/decapoda-research/llama-7b-hf/resolve/main/pytorch_model-00030-of-00033.bin
http://huggingface.co/decapoda-research/llama-7b-hf/resolve/main/pytorch_model-00031-of-00033.bin
http://huggingface.co/decapoda-research/llama-7b-hf/resolve/main/pytorch_model-00032-of-00033.bin
http://huggingface.co/decapoda-research/llama-7b-hf/resolve/main/pytorch_model-00033-of-00033.bin
http://huggingface.co/decapoda-research/llama-7b-hf/resolve/main/pytorch_model.bin.index.json
http://huggingface.co/decapoda-research/llama-7b-hf/resolve/main/special_tokens_map.json
http://huggingface.co/decapoda-research/llama-7b-hf/resolve/main/tokenizer.model
http://huggingface.co/decapoda-research/llama-7b-hf/resolve/main/tokenizer_config.json
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42

http://huggingface.co/lmsys/vicuna-7b-delta-v1.1/resolve/main/.gitattributes
http://huggingface.co/lmsys/vicuna-7b-delta-v1.1/resolve/main/README.md
http://huggingface.co/lmsys/vicuna-7b-delta-v1.1/resolve/main/config.json
http://huggingface.co/lmsys/vicuna-7b-delta-v1.1/resolve/main/generation_config.json
http://huggingface.co/lmsys/vicuna-7b-delta-v1.1/resolve/main/pytorch_model-00001-of-00002.bin
http://huggingface.co/lmsys/vicuna-7b-delta-v1.1/resolve/main/pytorch_model-00002-of-00002.bin
http://huggingface.co/lmsys/vicuna-7b-delta-v1.1/resolve/main/pytorch_model.bin.index.json
http://huggingface.co/lmsys/vicuna-7b-delta-v1.1/resolve/main/special_tokens_map.json
http://huggingface.co/lmsys/vicuna-7b-delta-v1.1/resolve/main/tokenizer.model
http://huggingface.co/lmsys/vicuna-7b-delta-v1.1/resolve/main/tokenizer_config.json
1
2
3
4
5
6
7
8
9
10

下载权重的脚本如下，使用了wget ${file} --no-check-certificate 绕开https检查：

#! /bin/bash
while read file; do
	wget ${file} --no-check-certificate 
done < download.txt
1
2
3
4

笔者的环境下，安装FastChat 0.1.10会导致依赖冲突：

The conflict is caused by:
    fschat 0.1.10 depends on tokenizers>=0.12.1
    transformers 4.35.0.dev0 depends on tokenizers<0.15 and >=0.14

To fix this you could try to:
1. loosen the range of package versions you've specified
2. remove package versions to allow pip attempt to solve the dependency conflict
1
2
3
4
5
6
7

因此改为安装别的版本fastchat

pip install fschat
1

再补充安装accelerate

pip install accelerate
1

之后执行python demo时，缺什么依赖就直接pip install安装即可。作者没有列出requirements.txt。

python demo.py --cfg-path eval_configs/minigpt4_eval.yaml  --gpu-id 0
1

其它修改

执行python demo.py仍然会遇到其它错误：

如果在下载bert tokenizer和bert config时报HTTPS错误，可以到huggingface把文件下载下来，再修改Minigpt-4项目的源码models/blip2.py：

tokenizer = BertTokenizer.from_pretrained("/root/minigpt4/bert-base-uncased", local_files_only=True)
...
encoder_config = BertConfig.from_pretrained("/root/minigpt4/bert-base-uncased", local_files_only=True)       
1
2
3

让代码监听网络，而不只是127.0.0.1

demo.launch(share=True, enable_queue=True, server_name='0.0.0.0')
1