TensorFlow Environment Setup

TensorFlow 2.x has specific requirements for Python versions:

Python 3.9-3.11 is recommendedThese versions offer the best compatibility and stability.

Check Current Python Version

# 检查 Python 版本
python --version
# 或
python3 --version

# 检查 pip 版本
pip --version
# 或
pip3 --version

Python Installation Options

Option 1: Official Python (Recommended for Beginners)

• Visitpython.orgDownload and Install
• Check "Add Python to PATH" during installation
• Includes the pip package manager

Option 2: Anaconda/Miniconda (Recommended for Data Science)

• Anaconda: Complete data science environment
• Miniconda: Lightweight version
• Built-in environment and package management

Option 3: System Package Manager

# Ubuntu/Debian
sudo apt update
sudo apt install python3 python3-pip

# macOS (使用 Homebrew)
brew install python3

# Windows (使用 Chocolatey)
choco install python3

Virtual Environment Setup (Highly Recommended)

Why Use a Virtual Environment

Benefits：

• Dependency isolation: Avoids package version conflicts between different projects
• Clean environment: Keeps the system Python environment clean
• Easy management: Easily delete and recreate environments
• Reproducibility: Makes it easy to reproduce environments on different machines

Creating a Virtual Environment with venv

# 创建虚拟环境
python -m venv tensorflow_env

# 激活虚拟环境
# Windows:
tensorflow_env\Scripts\activate
# macOS/Linux:
source tensorflow_env/bin/activate

# 确认激活（提示符会显示环境名）
which python  # 应该指向虚拟环境中的 python

# 升级 pip
python -m pip install --upgrade pip

Creating a Virtual Environment with Conda

# 创建新环境
conda create -n tensorflow_env python=3.10

# 激活环境
conda activate tensorflow_env

# 列出所有环境
conda env list

# 在环境中安装包
conda install pip

Virtual Environment Management

# 查看已安装的包
pip list

# 生成依赖文件
pip freeze > requirements.txt

# 从文件安装依赖
pip install -r requirements.txt

# 退出虚拟环境
deactivate  # venv
conda deactivate  # conda

# 删除虚拟环境
rm -rf tensorflow_env  # venv
conda env remove -n tensorflow_env  # conda

TensorFlow Installation

CPU Version Installation (Suitable for Learning and Lightweight Tasks)

# 确保在虚拟环境中
pip install tensorflow

# 验证安装
python -c "import tensorflow as tf; print('TensorFlow version:', tf.__version__)"

CPU Version Features：

• Advantages: Simple installation, no additional configuration required
• Disadvantages: Slower training speed, suitable for small-scale data
• Applicable Scenarios: Learning, prototyping, inference deployment

GPU Version Installation (Suitable for Large-Scale Training)

GPU acceleration can significantly improve training speed, especially for deep learning tasks.

3.2.1 NVIDIA GPU Requirements

Hardware Requirements：

• NVIDIA GPU (Compute Capability 3.5 or higher)
• 8GB+ VRAM (Recommended)

View GPU Information：

# Windows
nvidia-smi

# Linux
lspci | grep -i nvidia

CUDA and cuDNN Installation

TensorFlow 2.15+ and CUDA Version Correspondence：

Installation Steps：

1. Download and Install CUDA Toolkit

   • VisitNVIDIA CUDA Toolkit
   • Select the corresponding version to download and install
   • Add to system PATH

2. Download and Install cuDNN

   • VisitNVIDIA cuDNN
   • NVIDIA account registration required
   • Extract to the CUDA installation directory

3. Verify CUDA Installation

   nvcc --version
   nvidia-smi

Install TensorFlow GPU Version

# TensorFlow 2.10+ 统一包名
pip install tensorflow

# 验证 GPU 可用性
python -c "
import tensorflow as tf
print('TensorFlow version:', tf.__version__)
print('GPU available:', tf.config.list_physical_devices('GPU'))
print('Built with CUDA:', tf.test.is_built_with_cuda())
"

Installing via Conda

# CPU 版本
conda install tensorflow

# 或从 conda-forge
conda install -c conda-forge tensorflow

# GPU 版本（自动安装 CUDA）
conda install tensorflow-gpu

Installing a Specific Version

# 安装特定版本
pip install tensorflow==2.14.0

# 安装预发布版本
pip install tf-nightly

# 升级到最新版本
pip install --upgrade tensorflow

Development Tools Installation

upyter Notebook/Lab

Jupyter is the standard development environment for data science and machine learning:

# 安装 Jupyter Notebook
pip install jupyter notebook

# 或安装 JupyterLab（推荐）
pip install jupyterlab

# 启动 Jupyter
jupyter notebook
# 或
jupyter lab

Jupyter Extensions：

# 安装有用的扩展
pip install jupyter_contrib_nbextensions
jupyter contrib nbextension install --user

# 变量检查器
pip install nbextensions

IDE Selection

Visual Studio Code (Recommended)：

• Install Python extension
• Install Jupyter extension
• Intelligent code completion and debugging

PyCharm：

• Professional Python IDE
• Powerful debugging features
• Built-in Git support

Google Colab (Cloud Option)：

• Free GPU usage
• Pre-installed common libraries
• No local configuration required

Required Python Packages

# 数据处理
pip install numpy pandas matplotlib seaborn

# 科学计算
pip install scipy scikit-learn

# 图像处理
pip install pillow opencv-python

# 进度条和实用工具
pip install tqdm

# 创建完整的要求文件
cat > requirements.txt << EOF
tensorflow>=2.12.0
numpy>=1.21.0
pandas>=1.3.0
matplotlib>=3.5.0
seaborn>=0.11.0
scikit-learn>=1.0.0
jupyter>=1.0.0
tqdm>=4.60.0
pillow>=8.0.0
EOF

# 批量安装
pip install -r requirements.txt

Configuration Verification

Complete Installation Verification Script

Createverify_installation.pyfile:

Run verification:

python verify_installation.py

Performance Benchmark Test

Create a simple performance test:

Common Problems and Solutions

Installation Issues

Problem 1: pip installation timeout

pip install --user tensorflow

Problem 2: Permission error

# 创建新的虚拟环境
python -m venv fresh_env
source fresh_env/bin/activate  # Linux/Mac
# 或 fresh_env\Scripts\activate  # Windows
pip install tensorflow

Problem 3: Version conflict

# 创建新的虚拟环境
python -m venv fresh_env
source fresh_env/bin/activate  # Linux/Mac
# 或 fresh_env\Scripts\activate  # Windows
pip install tensorflow

GPU-Related Issues

Problem 1: CUDA version mismatch

• Check TensorFlow and CUDA version compatibility
• Reinstall the matching CUDA version

Problem 2: Insufficient GPU memory

# 设置 GPU 内存增长
gpus = tf.config.experimental.list_physical_devices('GPU')
if gpus:
    for gpu in gpus:
        tf.config.experimental.set_memory_growth(gpu, True)

Problem 3: GPU not found

# 检查 NVIDIA 驱动
nvidia-smi

# 检查 CUDA 安装
nvcc --version

# 重新安装 GPU 支持
pip uninstall tensorflow
pip install tensorflow

Jupyter-Related Issues

Problem 1: Virtual environment not visible in Jupyter

# 安装 ipykernel
pip install ipykernel

# 添加虚拟环境到 Jupyter
python -m ipykernel install --user --name tensorflow_env --display-name "Python (TensorFlow)"

Problem 2: Jupyter fails to start

# 重新安装 Jupyter
pip uninstall jupyter notebook
pip install jupyter notebook

# 或使用 conda
conda install jupyter

Development Environment Optimization

GPU Memory Management

# GPU 配置优化
import tensorflow as tf

# 方法 1：设置内存增长
gpus = tf.config.experimental.list_physical_devices('GPU')
if gpus:
    try:
        for gpu in gpus:
            tf.config.experimental.set_memory_growth(gpu, True)
    except RuntimeError as e:
        print(e)

# 方法 2：限制内存使用
if gpus:
    try:
        tf.config.experimental.set_memory_limit(gpus[0], 4096)  # 4GB
    except RuntimeError as e:
        print(e)

Performance Optimization Settings

# 启用混合精度训练（提升性能）
from tensorflow.keras.mixed_precision import experimental as mixed_precision

policy = mixed_precision.Policy('mixed_float16')
mixed_precision.set_policy(policy)

# 启用 XLA 编译优化
tf.config.optimizer.set_jit(True)

Development Tools Configuration

VS Code Configuration (.vscode/settings.json)：

{
    "python.defaultInterpreterPath": "./tensorflow_env/bin/python",
    "python.linting.enabled": true,
    "python.linting.pylintEnabled": true,
    "jupyter.defaultKernel": "tensorflow_env"
}

Environment Maintenance

Regular Updates

# 更新 TensorFlow
pip install --upgrade tensorflow

# 更新所有包
pip list --outdated
pip install --upgrade package_name

# 或批量更新
pip freeze | grep -v "^-e" | cut -d = -f 1 | xargs pip install -U

Environment Backup and Migration

# 导出环境
pip freeze > requirements.txt
conda env export > environment.yml

# 在新机器上恢复环境
pip install -r requirements.txt
conda env create -f environment.yml