(预发布)[阿维笔记]分析优化CloudStudio高性能工作空间的GPU训练速度和效果

刘维克2025-06-06 15:56

0

1.摘要

本教程使用 腾讯云 CloudStudio. CloudStudio 提供每月免费5W分钟CPU环境和1W分钟GPU环境(16C32G & Tesla T4@16GVRAM)

  • 检查了nvidia-driver和conda虚拟环境.
  • 分析2025年哪个版本PyTorch更适合该环境(仅讨论计算机视觉任务).
  • 分析解释nvidia-driver, CUDA toolkits和CuDNN等兼容性问题.

2.部署CloudStudio高性能空间

2.1 硬件环境

  • 点击CloudStudio链接cloud.tencent.com/product/clo..., 注册 腾讯云 .
  • 进入CloudStudio主页面ide.cloud.tencent.com/, 点击高性能工作空间,点击新建,点击左下角免费基础型,点新建,等待 几分钟 .
  • 刷新 页面等待出现新的空间条目,点击这个 新的条目 ,开始使用高性能空间.
  • 如下是成功打开 高性能工作空间 的效果

2.2 软件环境

由于这个免费基础型环境自带 deepseek 等依赖(用不到)和conda环境,我们这里直接使用conda环境的(base)环境

  • 这里已经自动激活了(base)的conda默认虚拟环境,使用pip list检查已经安装的依赖.这里并没有需要的PyTorch,所以我们需要手动安装:

    点击展开软件环境检查log

    bash 复制代码 
    (base) root@VM-0-80-ubuntu:/workspace# which conda
/root/miniforge3/bin/conda
(base) root@VM-0-80-ubuntu:/workspace# pip list
Package                 Version
----------------------- -----------
archspec                0.2.3
asttokens               3.0.0
boltons                 24.0.0
Brotli                  1.1.0
certifi                 2024.12.14
cffi                    1.17.1
charset-normalizer      3.4.1
colorama                0.4.6
comm                    0.2.2
conda                   24.11.2
conda-libmamba-solver   24.11.1
conda-package-handling  2.4.0
conda_package_streaming 0.11.0
debugpy                 1.8.11
decorator               5.1.1
distro                  1.9.0
exceptiongroup          1.2.2
executing               2.1.0
frozendict              2.4.6
h2                      4.1.0
hpack                   4.0.0
hyperframe              6.0.1
idna                    3.10
importlib_metadata      8.5.0
ipykernel               6.29.5
ipython                 8.31.0
jedi                    0.19.2
jsonpatch               1.33
jsonpointer             3.0.0
jupyter_client          8.6.3
jupyter_core            5.7.2
libmambapy              2.0.5
matplotlib-inline       0.1.7
menuinst                2.2.0
nest_asyncio            1.6.0
packaging               24.2
parso                   0.8.4
pexpect                 4.9.0
pickleshare             0.7.5
pip                     24.3.1
platformdirs            4.3.6
pluggy                  1.5.0
prompt_toolkit          3.0.48
psutil                  6.1.1
ptyprocess              0.7.0
pure_eval               0.2.3
pycosat                 0.6.6
pycparser               2.22
Pygments                2.18.0
PySocks                 1.7.1
python-dateutil         2.9.0.post0
pyzmq                   26.2.0
requests                2.32.3
ruamel.yaml             0.18.8
ruamel.yaml.clib        0.2.8
setuptools              75.6.0
six                     1.17.0
stack_data              0.6.3
tornado                 6.4.2
tqdm                    4.67.1
traitlets               5.14.3
truststore              0.10.0
typing_extensions       4.12.2
urllib3                 2.3.0
wcwidth                 0.2.13
wheel                   0.45.1
zipp                    3.21.0
zstandard               0.23.0

  • 使用nvidia-smi等命令检查GPU和nvidia-driver等版本,确定应该装什么版本PyTorch.

    点击展开环境检查log

    log 复制代码 
    (base) root@VM-0-80-ubuntu:/workspace# nvidia-smi
Thu Jun  5 13:40:40 2025       
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 525.105.17   Driver Version: 525.105.17   CUDA Version: 12.0     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  Tesla T4            On   | 00000000:00:09.0 Off |                    0 |
| N/A   33C    P8    11W /  70W |      5MiB / 15360MiB |      0%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+

+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
|  No running processes found                                                 |
+-----------------------------------------------------------------------------+
(base) root@VM-0-80-ubuntu:/workspace# python --version
Python 3.10.11

    !TIP

    选择PyTorch2.4.x是因为从2.4.x开始(图像分类,分割,检测等)算子速度有所优化,而2.4.1是2.4.0的更新版

    选择cu118版本是因为nvidia-smi的结果NVIDIA-SMI 525.105.17 Driver Version: 525.105.17 CUDA Version: 12.0也就是目前的525版本nvidia-driver最高支持到CUDA toolkits 12.0的CUDA toolkit和CuDNN.而Pytorch给的release只有11.8, 12.1和12.4由于无法使用高于CUDA toolkit 12.0(不含),所以只能使用cu118.再者cu118也是cu11x中最新,最稳定(不很确定)的版本.另外,实测cu12.x在图像任务训练领域不会比cu117, cu118快.

    这里其实是可以使用高于cu12.0的依赖的,但是不建议新手使用.有兴趣可以查阅docs.nvidia.com/deploy/cuda.... 也就是说,在表面上nvidia-driver最高支持到12.0的情况下,直接安装并使用CUDA toolkits 12.1, 12.4e也是可以的.

    如果你需要其他版本PyTorch请到pytorch.org/get-started...

  • 安装GPU版PyTorch: pip install torch==2.4.1 torchvision==0.19.1 torchaudio==2.4.1 --index-url https://download.pytorch.org/whl/cu118

    点击展开`PyTorch GPU`安装log

    log 复制代码 
    (base) root@VM-0-80-ubuntu:/workspace# pip install torch==2.4.1 torchvision==0.19.1 torchaudio==2.4.1 --index-url https://download.pytorch.org/whl/cu118
Looking in indexes: https://download.pytorch.org/whl/cu118
Collecting torch==2.4.1
  Downloading https://download.pytorch.org/whl/cu118/torch-2.4.1%2Bcu118-cp310-cp310-linux_x86_64.whl (857.6 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 857.6/857.6 MB 8.6 MB/s eta 0:00:00
Collecting torchvision==0.19.1
  Downloading https://download.pytorch.org/whl/cu118/torchvision-0.19.1%2Bcu118-cp310-cp310-linux_x86_64.whl (6.3 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 6.3/6.3 MB 9.8 MB/s eta 0:00:00
Collecting torchaudio==2.4.1
  Downloading https://download.pytorch.org/whl/cu118/torchaudio-2.4.1%2Bcu118-cp310-cp310-linux_x86_64.whl (3.3 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 3.3/3.3 MB 6.7 MB/s eta 0:00:00
Collecting filelock (from torch==2.4.1)
  Downloading https://download.pytorch.org/whl/filelock-3.13.1-py3-none-any.whl.metadata (2.8 kB)
Requirement already satisfied: typing-extensions>=4.8.0 in /root/miniforge3/lib/python3.10/site-packages (from torch==2.4.1) (4.12.2)
Collecting sympy (from torch==2.4.1)
  Downloading https://download.pytorch.org/whl/sympy-1.13.3-py3-none-any.whl.metadata (12 kB)
Collecting networkx (from torch==2.4.1)
  Downloading https://download.pytorch.org/whl/networkx-3.3-py3-none-any.whl.metadata (5.1 kB)
Collecting jinja2 (from torch==2.4.1)
  Downloading https://download.pytorch.org/whl/Jinja2-3.1.4-py3-none-any.whl.metadata (2.6 kB)
Collecting fsspec (from torch==2.4.1)
  Downloading https://download.pytorch.org/whl/fsspec-2024.6.1-py3-none-any.whl.metadata (11 kB)
Collecting nvidia-cuda-nvrtc-cu11==11.8.89 (from torch==2.4.1)
  Downloading https://download.pytorch.org/whl/cu118/nvidia_cuda_nvrtc_cu11-11.8.89-py3-none-manylinux1_x86_64.whl (23.2 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 23.2/23.2 MB 14.5 MB/s eta 0:00:00
Collecting nvidia-cuda-runtime-cu11==11.8.89 (from torch==2.4.1)
  Downloading https://download.pytorch.org/whl/cu118/nvidia_cuda_runtime_cu11-11.8.89-py3-none-manylinux1_x86_64.whl (875 kB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 875.6/875.6 kB 16.0 MB/s eta 0:00:00
Collecting nvidia-cuda-cupti-cu11==11.8.87 (from torch==2.4.1)
  Downloading https://download.pytorch.org/whl/cu118/nvidia_cuda_cupti_cu11-11.8.87-py3-none-manylinux1_x86_64.whl (13.1 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 13.1/13.1 MB 21.3 MB/s eta 0:00:00
Collecting nvidia-cudnn-cu11==9.1.0.70 (from torch==2.4.1)
  Downloading https://download.pytorch.org/whl/cu118/nvidia_cudnn_cu11-9.1.0.70-py3-none-manylinux2014_x86_64.whl (663.9 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 663.9/663.9 MB 10.0 MB/s eta 0:00:00
Collecting nvidia-cublas-cu11==11.11.3.6 (from torch==2.4.1)
  Downloading https://download.pytorch.org/whl/cu118/nvidia_cublas_cu11-11.11.3.6-py3-none-manylinux1_x86_64.whl (417.9 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 417.9/417.9 MB 9.7 MB/s eta 0:00:00
Collecting nvidia-cufft-cu11==10.9.0.58 (from torch==2.4.1)
  Downloading https://download.pytorch.org/whl/cu118/nvidia_cufft_cu11-10.9.0.58-py3-none-manylinux1_x86_64.whl (168.4 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 168.4/168.4 MB 12.5 MB/s eta 0:00:00
Collecting nvidia-curand-cu11==10.3.0.86 (from torch==2.4.1)
  Downloading https://download.pytorch.org/whl/cu118/nvidia_curand_cu11-10.3.0.86-py3-none-manylinux1_x86_64.whl (58.1 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 58.1/58.1 MB 13.0 MB/s eta 0:00:00
Collecting nvidia-cusolver-cu11==11.4.1.48 (from torch==2.4.1)
  Downloading https://download.pytorch.org/whl/cu118/nvidia_cusolver_cu11-11.4.1.48-py3-none-manylinux1_x86_64.whl (128.2 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 128.2/128.2 MB 12.5 MB/s eta 0:00:00
Collecting nvidia-cusparse-cu11==11.7.5.86 (from torch==2.4.1)
  Downloading https://download.pytorch.org/whl/cu118/nvidia_cusparse_cu11-11.7.5.86-py3-none-manylinux1_x86_64.whl (204.1 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 204.1/204.1 MB 11.9 MB/s eta 0:00:00
Collecting nvidia-nccl-cu11==2.20.5 (from torch==2.4.1)
  Downloading https://download.pytorch.org/whl/cu118/nvidia_nccl_cu11-2.20.5-py3-none-manylinux2014_x86_64.whl (142.9 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 142.9/142.9 MB 12.5 MB/s eta 0:00:00
Collecting nvidia-nvtx-cu11==11.8.86 (from torch==2.4.1)
  Downloading https://download.pytorch.org/whl/cu118/nvidia_nvtx_cu11-11.8.86-py3-none-manylinux1_x86_64.whl (99 kB)
Collecting triton==3.0.0 (from torch==2.4.1)
  Downloading https://download.pytorch.org/whl/triton-3.0.0-1-cp310-cp310-manylinux2014_x86_64.manylinux_2_17_x86_64.whl (209.4 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 209.4/209.4 MB 12.0 MB/s eta 0:00:00
Collecting numpy (from torchvision==0.19.1)
  Downloading https://download.pytorch.org/whl/numpy-2.1.2-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (60 kB)
Collecting pillow!=8.3.*,>=5.3.0 (from torchvision==0.19.1)
  Downloading https://download.pytorch.org/whl/pillow-11.0.0-cp310-cp310-manylinux_2_28_x86_64.whl.metadata (9.1 kB)
Collecting MarkupSafe>=2.0 (from jinja2->torch==2.4.1)
  Downloading https://download.pytorch.org/whl/MarkupSafe-2.1.5-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (25 kB)
Collecting mpmath<1.4,>=1.1.0 (from sympy->torch==2.4.1)
  Downloading https://download.pytorch.org/whl/mpmath-1.3.0-py3-none-any.whl (536 kB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 536.2/536.2 kB 7.5 MB/s eta 0:00:00
Downloading https://download.pytorch.org/whl/pillow-11.0.0-cp310-cp310-manylinux_2_28_x86_64.whl (4.4 MB)
   ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 4.4/4.4 MB 9.4 MB/s eta 0:00:00
Downloading https://download.pytorch.org/whl/filelock-3.13.1-py3-none-any.whl (11 kB)
Downloading https://download.pytorch.org/whl/fsspec-2024.6.1-py3-none-any.whl (177 kB)
Downloading https://download.pytorch.org/whl/Jinja2-3.1.4-py3-none-any.whl (133 kB)
Downloading https://download.pytorch.org/whl/networkx-3.3-py3-none-any.whl (1.7 MB)
   ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 1.7/1.7 MB 6.6 MB/s eta 0:00:00
Downloading https://download.pytorch.org/whl/numpy-2.1.2-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (16.3 MB)
   ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 16.3/16.3 MB 13.0 MB/s eta 0:00:00
Downloading https://download.pytorch.org/whl/sympy-1.13.3-py3-none-any.whl (6.2 MB)
   ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 6.2/6.2 MB 32.9 MB/s eta 0:00:00
Installing collected packages: mpmath, sympy, pillow, nvidia-nvtx-cu11, nvidia-nccl-cu11, nvidia-cusparse-cu11, nvidia-curand-cu11, nvidia-cufft-cu11, nvidia-cuda-runtime-cu11, nvidia-cuda-nvrtc-cu11, nvidia-cuda-cupti-cu11, nvidia-cublas-cu11, numpy, networkx, MarkupSafe, fsspec, filelock, triton, nvidia-cusolver-cu11, nvidia-cudnn-cu11, jinja2, torch, torchvision, torchaudio
Successfully installed MarkupSafe-2.1.5 filelock-3.13.1 fsspec-2024.6.1 jinja2-3.1.4 mpmath-1.3.0 networkx-3.3 numpy-2.1.2 nvidia-cublas-cu11-11.11.3.6 nvidia-cuda-cupti-cu11-11.8.87 nvidia-cuda-nvrtc-cu11-11.8.89 nvidia-cuda-runtime-cu11-11.8.89 nvidia-cudnn-cu11-9.1.0.70 nvidia-cufft-cu11-10.9.0.58 nvidia-curand-cu11-10.3.0.86 nvidia-cusolver-cu11-11.4.1.48 nvidia-cusparse-cu11-11.7.5.86 nvidia-nccl-cu11-2.20.5 nvidia-nvtx-cu11-11.8.86 pillow-11.0.0 sympy-1.13.3 torch-2.4.1+cu118 torchaudio-2.4.1+cu118 torchvision-0.19.1+cu118 triton-3.0.0
WARNING: Running pip as the 'root' user can result in broken permissions and conflicting behaviour with the system package manager, possibly rendering your system unusable.It is recommended to use a virtual environment instead: https://pip.pypa.io/warnings/venv. Use the --root-user-action option if you know what you are doing and want to suppress this warning.
(base) root@VM-0-80-ubuntu:/workspace#

  • 检查PyTorch GPU是否安装成功python -c "import torch;print(torch.cuda.get_device_name(torch.cuda.current_device()))"

log 复制代码 
(base) root@VM-0-80-ubuntu:/workspace# python -c "import  torch;print(torch.cuda.get_device_name(torch.cuda.current_device()))"
Tesla T4
相关推荐
ReinaXue12 分钟前
大模型【进阶】（五）：低秩适配矩阵LORA的深度认识
人工智能·深度学习·神经网络·语言模型·自然语言处理·transformer
DisonTangor21 分钟前
商汤InternLM发布最先进的开源多模态推理模型——Intern-S1
人工智能·深度学习·开源·aigc
都叫我大帅哥1 小时前
深度学习的"Hello World"：多层感知机全解指南
python·深度学习
格林威1 小时前
Baumer工业相机堡盟工业相机如何通过YoloV8深度学习模型实现轮船检测识别（C#代码UI界面版）
人工智能·深度学习·数码相机·yolo·视觉检测
Gession-杰1 小时前
OpenCV图像梯度、边缘检测、轮廓绘制、凸包检测大合集
人工智能·opencv·计算机视觉
计算机sci论文精选2 小时前
CVPR 2024 3D传感框架实现无监督场景理解新纪元
人工智能·机器学习·计算机视觉·3d·cvpr·传感技术
钟屿2 小时前
Multiscale Structure Guided Diffusion for Image Deblurring 论文阅读
论文阅读·图像处理·人工智能·深度学习·计算机视觉
go54631584653 小时前
Python点阵字生成与优化：从基础实现到高级渲染技术
开发语言·人工智能·python·深度学习·分类·数据挖掘
Coovally AI模型快速验证3 小时前
避开算力坑！无人机桥梁检测场景下YOLO模型选型指南
人工智能·深度学习·yolo·计算机视觉·目标跟踪·无人机
水军总督4 小时前
OpenCV+Python
python·opencv·计算机视觉
热门推荐
01Qwen3-Coder 快速上手教程 | Qwen Code + Claude Code02全球最强模型Grok4，国内已可免费使用！（附教程）03vue数据变化但页面不变04KGG转MP3工具|非KGM文件|解密音频05干翻 Typora！MilkUp：完全免费的桌面端 Markdown 编辑器！06扣子开源本地部署教程 丨Coze智能体小白喂饭级指南07【2025.7.18】更新vscode后所有.vue文件template标签后报红的临时解决办法，Vue - Official 插件3.0.2导致08ChatGPT Agent 完全使用指南：2025年7月最新功能详解09《魔兽世界》提示lua警告的含义及解决方法10这次领先Cursor！体验了Trae 2.0 SOLO 模式，超酷！