centos-LLM-生物信息-BioGPT安装

参考

GitHub - microsoft/BioGPT

https://github.com/microsoft/BioGPT

BioGPT:用于生物医学文本生成和挖掘的生成式预训练转换器 |生物信息学简报 |牛津学术 --- BioGPT: generative pre-trained transformer for biomedical text generation and mining | Briefings in Bioinformatics | Oxford Academic

https://academic.oup.com/bib/article/23/6/bbac409/6713511

No such file or directory: 'fairseq\version.txt · Issue #5518 · facebookresearch/fairseq

https://github.com/facebookresearch/fairseq/issues/5518

环境

centos 7,anaconda3,CUDA 11.6


目录

问题描述

BioGPT,是一个在大规模生物医学文献上预先训练的特定领域的生成 Transformer 语言模型。

本文介绍了它在本地服务器上的安装过程,下一篇文章将会写具体用法。

安装

要求:(来自github页面readme)

  1. Python version == 3.10
  2. PyTorch version == 1.12.0
  3. fairseq version == 0.12.0(要先装PyTorch)
  4. Moses
  5. fastBPE
  6. sacremoses
  7. sklearn

PyTorch

安装 CUDA Toolkit 11.6

手动下载 CUDA Toolkit 11.6

shell 复制代码
wget https://developer.download.nvidia.com/compute/cuda/11.6.0/local_installers/cuda_11.6.0_510.39.01_linux.run
sudo sh cuda_11.6.0_510.39.01_linux.run

安装的时候会有选项,这样选:

然后选择Install。

这个要选择哪个?选择Choose components to upgrade

shell 复制代码
Existing installation of CUDA Toolkit 11.6 found:                            
x Upgrade all                                                                  
x Choose components to upgrade                                                 
x No, abort installation   

设置环境变量:

shell 复制代码
echo 'export PATH=/usr/local/cuda-11.6/bin:$PATH' >> ~/.bashrc
echo 'export LD_LIBRARY_PATH=/usr/local/cuda-11.6/lib64:$LD_LIBRARY_PATH' >> ~/.bashrc
source ~/.bashrc

安装PyTorch

使用清华源。

shell 复制代码
conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/pytorch/
conda install pytorch==1.12.0 torchvision==0.13.0 torchaudio==0.12.0

检查PyTorch能否识别GPU:输出True就对了

shell 复制代码
python -c "import torch; print(torch.cuda.is_available())"

fairseq

shell 复制代码
pip --default-timeout=1000 install fairseq -i https://pypi.tuna.tsinghua.edu.cn/simple

有可能需要:

shell 复制代码
sudo yum install -y python3-devel gcc g++ make cmake
pip install cython -i https://pypi.tuna.tsinghua.edu.cn/simple

官方的做法是下面。但我试了,1是要科学上网,2是会报错:TimeoutError: The read operation timed out或者ERROR: Could not find a version that satisfies the requirement cython (from versions: none)。安装了cython也不行。

shell 复制代码
git clone https://github.com/pytorch/fairseq
cd fairseq
pip install --editable ./

# on MacOS:
# CFLAGS="-stdlib=libc++" pip install --editable ./

# to install the latest stable release (0.10.x)
# pip install fairseq

报错:FileNotFoundError: Errno 2 No such file or directory: 'fairseq/version.txt'

完整报错:

shell 复制代码
  Installing build dependencies ... done
  Getting requirements to build wheel ... error
  error: subprocess-exited-with-error
  
  × Getting requirements to build wheel did not run successfully.
  │ exit code: 1
  ╰─> [16 lines of output]
      Traceback (most recent call last):
        File "/home/xxx/anaconda3/envs/biogpt/lib/python3.10/site-packages/pip/_vendor/pyproject_hooks/_in_process/_in_process.py", line 389, in <module>
          main()
        File "/home/xxx/anaconda3/envs/biogpt/lib/python3.10/site-packages/pip/_vendor/pyproject_hooks/_in_process/_in_process.py", line 373, in main
          json_out["return_val"] = hook(**hook_input["kwargs"])
        File "/home/xxx/anaconda3/envs/biogpt/lib/python3.10/site-packages/pip/_vendor/pyproject_hooks/_in_process/_in_process.py", line 143, in get_requires_for_build_wheel
          return hook(config_settings)
        File "/tmp/pip-build-env-ubaynszi/overlay/lib/python3.10/site-packages/setuptools/build_meta.py", line 334, in get_requires_for_build_wheel
          return self._get_build_requires(config_settings, requirements=[])
        File "/tmp/pip-build-env-ubaynszi/overlay/lib/python3.10/site-packages/setuptools/build_meta.py", line 304, in _get_build_requires
          self.run_setup()
        File "/tmp/pip-build-env-ubaynszi/overlay/lib/python3.10/site-packages/setuptools/build_meta.py", line 320, in run_setup
          exec(code, locals())
        File "<string>", line 27, in <module>
        File "<string>", line 18, in write_version_py
      FileNotFoundError: [Errno 2] No such file or directory: 'fairseq/version.txt'
      [end of output]
  
  note: This error originates from a subprocess, and is likely not a problem with pip.
error: subprocess-exited-with-error

× Getting requirements to build wheel did not run successfully.
│ exit code: 1
╰─> See above for output.

note: This error originates from a subprocess, and is likely not a problem with pip.

原因:参考下面的链接,是pip版本太高了,我的pip是25.0.1,降到24就可以了。

No such file or directory: 'fairseq\version.txt · Issue #5518 · facebookresearch/fairseq

https://github.com/facebookresearch/fairseq/issues/5518

shell 复制代码
python -m pip install pip==24.0 

Moses

shell 复制代码
git clone https://github.com/moses-smt/mosesdecoder.git
export MOSES=${PWD}/mosesdecoder

fastBPE

shell 复制代码
git clone https://github.com/glample/fastBPE.git
export FASTBPE=${PWD}/fastBPE
cd fastBPE
g++ -std=c++11 -pthread -O3 fastBPE/main.cc -IfastBPE -o fast

sacremoses

shell 复制代码
pip install sacremoses

sklearn

shell 复制代码
pip install scikit-learn -i https://pypi.tuna.tsinghua.edu.cn/simple
相关推荐
深蓝AI1 小时前
AI Agent 入门实战:用 Function Calling 让大模型学会调用工具
aigc
Kyrie6782 小时前
SkillOpt:把 Agent 的技能文件当作可训练参数
人工智能
leeyi2 小时前
可观测性:Langfuse、Langsmith 集成
aigc·agent·ai编程
冬奇Lab2 小时前
Workflow 系列(07):工程化与版本管理——Workflow 的 CI/CD
人工智能·工作流引擎
两万五千个小时2 小时前
Claude Code 上下文管理(一):为什么 Agent 会"失忆"?
人工智能·架构·开源
两万五千个小时2 小时前
Claude Code 上下文管理(二):零 Token 消耗的压缩三板斧
人工智能·程序员·开源
冬奇Lab3 小时前
每日一个开源项目(第150篇):caveman - 为什么用很多 token,少 token 也行——给 AI Agent 装上穴居人嘴巴
人工智能·开源·资讯
贵慜_Derek3 小时前
MAI-04|干净数据在工程上意味着什么:MAI 预训练数据治理
人工智能·算法·llm
feelmylife593 小时前
Agent 记忆设计架构 — 分层记忆:什么时候该记住,什么时候该忘记
人工智能
阿黎梨梨3 小时前
揭秘大语言模型的底层逻辑:从文本分词到高维向量的计算之旅
javascript·人工智能