Kenhuang Academy
  • English

      • PyTorch Environment Installation

      System Requirements

      Operating System Support

      • Windows: Windows 10/11 (64-bit)
      • macOS: macOS 10.15+
      • Linux: Mainstream distributions (Ubuntu 18.04+, CentOS 7+)

      Python Version

      • Python 3.8 - 3.11 (recommended 3.9 or 3.10)

      Hardware Requirements

      • CPU: Modern processor supporting AVX instruction set
      • Memory: At least 8GB RAM (recommended 16GB+)
      • GPU: NVIDIA GPU (optional, for CUDA acceleration)

      Installation Method Selection

      The simplest and most direct installation method, suitable for most users.

      2. Conda Installation

      Suitable for users who need to manage multiple Python environments.

      3. Source Compilation

      Suitable for advanced users who need custom configuration or the latest development version.

      Detailed Installation Steps

      Method 1: Using pip Installation

      1. Check Python Environment

      python --version
pip --version
      # Create virtual environment
python -m venv pytorch_env

# Activate virtual environment
# Windows
pytorch_env\Scripts\activate
# macOS/Linux
source pytorch_env/bin/activate

      3. Install PyTorch

      CPU Version (Suitable for learning and small-scale experiments):

      pip install torch torchvision torchaudio

      GPU Version (Requires NVIDIA GPU and CUDA):

      # CUDA 11.8 version
pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118

# CUDA 12.1 version
pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu121

      Method 2: Using Conda Installation

      1. Install Anaconda or Miniconda

      Download and install from: https://www.anaconda.com/

      2. Create Conda Environment

      conda create -n pytorch_env python=3.10
conda activate pytorch_env

      3. Install PyTorch

      # CPU version
conda install pytorch torchvision torchaudio cpuonly -c pytorch

# GPU version
conda install pytorch torchvision torchaudio pytorch-cuda=11.8 -c pytorch -c nvidia

      CUDA Environment Configuration

      1. Check GPU Support

      import torch
print(f"CUDA available: {torch.cuda.is_available()}")
print(f"CUDA version: {torch.version.cuda}")
print(f"GPU count: {torch.cuda.device_count()}")
if torch.cuda.is_available():
    print(f"GPU name: {torch.cuda.get_device_name(0)}")

      2. Install CUDA Toolkit (if needed)

      Visit NVIDIA website to download corresponding version of CUDA Toolkit: https://developer.nvidia.com/cuda-toolkit

      3. Verify CUDA Installation

      nvcc --version
nvidia-smi

      Verify Installation

      Create test script test_pytorch.py:

      import torch
import torchvision
import numpy as np

def test_pytorch_installation():
    print("=== PyTorch Installation Verification ===")
    
    # Basic info
    print(f"PyTorch version: {torch.__version__}")
    print(f"TorchVision version: {torchvision.__version__}")
    
    # CUDA support
    print(f"CUDA available: {torch.cuda.is_available()}")
    if torch.cuda.is_available():
        print(f"CUDA version: {torch.version.cuda}")
        print(f"cuDNN version: {torch.backends.cudnn.version()}")
        print(f"GPU count: {torch.cuda.device_count()}")
        for i in range(torch.cuda.device_count()):
            print(f"GPU {i}: {torch.cuda.get_device_name(i)}")
    
    # Basic tensor operation test
    print("\n=== Basic Functionality Test ===")
    
    # CPU tensor
    x = torch.randn(3, 4)
    y = torch.randn(4, 5)
    z = torch.mm(x, y)
    print(f"CPU matrix multiplication result shape: {z.shape}")
    
    # GPU tensor (if available)
    if torch.cuda.is_available():
        x_gpu = x.cuda()
        y_gpu = y.cuda()
        z_gpu = torch.mm(x_gpu, y_gpu)
        print(f"GPU matrix multiplication result shape: {z_gpu.shape}")
        print(f"GPU tensor device: {z_gpu.device}")
    
    # Automatic differentiation test
    x = torch.randn(2, 2, requires_grad=True)
    y = x.pow(2).sum()
    y.backward()
    print(f"Automatic differentiation gradient shape: {x.grad.shape}")
    
    # Neural network module test
    import torch.nn as nn
    model = nn.Linear(10, 1)
    input_tensor = torch.randn(5, 10)
    output = model(input_tensor)
    print(f"Neural network output shape: {output.shape}")
    
    print("\n✅ PyTorch installation verification successful!")

if __name__ == "__main__":
    test_pytorch_installation()

      Run test:

      python test_pytorch.py

      Common Problem Solutions

      1. Import Error

      # Error: ImportError: No module named 'torch'
# Solution: Ensure correct virtual environment, reinstall PyTorch

      2. CUDA Version Mismatch

      # Error: RuntimeError: CUDA runtime version mismatch
# Solution: Install PyTorch matching system CUDA version

      3. Memory Insufficient

      # Error: RuntimeError: CUDA out of memory
# Solution: Reduce batch size or use CPU training

      4. Windows Long Path Issue

      # Enable long path support on Windows
git config --system core.longpaths true

      1. IDE Selection

      • PyCharm: Powerful Python IDE
      • VS Code: Lightweight with rich plugins
      • Jupyter Notebook: Suitable for experiments and learning

      2. Essential Plugins/Extensions

      • Python syntax highlighting
      • Code autocompletion
      • Git integration
      • Debugger support

      3. Useful Python Packages

      pip install jupyter matplotlib seaborn pandas scikit-learn tqdm

      Performance Optimization Suggestions

      1. Use Appropriate Data Type

      # Using float32 instead of float64 saves memory
x = torch.randn(1000, 1000, dtype=torch.float32)

      2. Enable cuDNN Benchmark

      import torch.backends.cudnn as cudnn
cudnn.benchmark = True  # Suitable for fixed input sizes

      3. Set Thread Count

      torch.set_num_threads(4)  # Adjust based on CPU core count

      Update PyTorch

      pip Update

      pip install --upgrade torch torchvision torchaudio

      conda Update

      conda update pytorch torchvision torchaudio -c pytorch

      Summary

      Correctly installing and configuring PyTorch environment is the first step to success in deep learning projects. Suggestions:

      1. Prioritize using virtual environments to isolate project dependencies
      2. Choose CPU or GPU version based on hardware conditions
      3. Regularly update to latest stable version
      4. Keep development environment clean and consistent

      After installation, you can start exploring the powerful features of PyTorch!