RTX 3080 AI算力入门指南：从环境配置到首个深度学习模型

1次阅读

共计 3476 个字符，预计需要花费 9 分钟才能阅读完成。

RTX 3080 作为 NVIDIA 的消费级显卡，拥有以下关键 AI 计算特性：

10GB GDDR6X 显存：高带宽设计（760GB/s）适合大 batch size 训练
8704 个 CUDA 核心：提供通用并行计算能力
第 2 代 Tensor Core：支持 FP16/FP32 混合精度计算，理论 AI 性能提升 2 - 4 倍
PCIe 4.0 接口：数据传输速度是 PCIe 3.0 的两倍

从 NVIDIA 官网下载最新 Game Ready 驱动（建议版本 510+）

安装对应版本的 CUDA Toolkit（推荐 11.7）：

wget https://developer.download.nvidia.com/compute/cuda/11.7.0/local_installers/cuda_11.7.0_515.43.04_linux.run
sudo sh cuda_11.7.0_515.43.04_linux.run

添加环境变量到~/.bashrc：

export PATH=/usr/local/cuda-11.7/bin${PATH:+:${PATH}}
export LD_LIBRARY_PATH=/usr/local/cuda-11.7/lib64${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}

下载与 CUDA 版本匹配的 cuDNN（需注册 NVIDIA 开发者账号）

解压并复制文件：

tar -xzvf cudnn-linux-x86_64-8.5.0.96_cuda11-archive.tar.xz
sudo cp cudnn-*-archive/include/cudnn*.h /usr/local/cuda/include
sudo cp -P cudnn-*-archive/lib/libcudnn* /usr/local/cuda/lib64
sudo chmod a+r /usr/local/cuda/include/cudnn*.h /usr/local/cuda/lib64/libcudnn*

pip install torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cu117

import torch
import torchvision
from torch import nn
from torch.utils.data import DataLoader

# 1. 设备检测
device = 'cuda' if torch.cuda.is_available() else 'cpu'
print(f"Using {device} device")

# 2. 数据准备
transform = torchvision.transforms.Compose([torchvision.transforms.ToTensor(),
    torchvision.transforms.Normalize((0.5,), (0.5,))
])

train_data = torchvision.datasets.CIFAR10(
    root='data',
    train=True,
    download=True,
    transform=transform
)

train_loader = DataLoader(train_data, batch_size=128, shuffle=True)

# 3. 模型定义
class CNN(nn.Module):
    def __init__(self):
        super().__init__()
        self.conv1 = nn.Conv2d(3, 32, kernel_size=3, padding=1)
        self.conv2 = nn.Conv2d(32, 64, kernel_size=3, padding=1)
        self.pool = nn.MaxPool2d(2, 2)
        self.fc1 = nn.Linear(64 * 8 * 8, 512)
        self.fc2 = nn.Linear(512, 10)

    def forward(self, x):
        x = self.pool(torch.relu(self.conv1(x)))
        x = self.pool(torch.relu(self.conv2(x)))
        x = torch.flatten(x, 1)
        x = torch.relu(self.fc1(x))
        x = self.fc2(x)
        return x

model = CNN().to(device)

# 4. 训练配置
criterion = nn.CrossEntropyLoss()
optimizer = torch.optim.Adam(model.parameters(), lr=0.001)

# 5. 训练循环
for epoch in range(10):
    running_loss = 0.0
    for i, (inputs, labels) in enumerate(train_loader):
        inputs, labels = inputs.to(device), labels.to(device)

        optimizer.zero_grad()
        outputs = model(inputs)
        loss = criterion(outputs, labels)
        loss.backward()
        optimizer.step()

        running_loss += loss.item()
        if i % 100 == 99:
            print(f'Epoch {epoch+1}, Batch {i+1}: loss {running_loss/100:.3f}')
            running_loss = 0.0

from torch.cuda.amp import autocast, GradScaler

scaler = GradScaler()

for inputs, labels in train_loader:
    inputs, labels = inputs.to(device), labels.to(device)

    optimizer.zero_grad()

    with autocast():
        outputs = model(inputs)
        loss = criterion(outputs, labels)

    scaler.scale(loss).backward()
    scaler.step(optimizer)
    scaler.update()

10GB 显存典型配置：
224×224 图像：batch_size=128-256
512×512 图像：batch_size=32-64
1024×1024 图像：batch_size=8-16

train_loader = DataLoader(
    train_data,
    batch_size=128,
    shuffle=True,
    num_workers=4,  # 通常设置为 CPU 核心数的 1 / 2 到 3 /4
    pin_memory=True,  # 启用快速数据传输到 GPU
    persistent_workers=True
)

症状：CUDA out of memory
解决方案：
减小 batch size

使用梯度累积：

accumulation_steps = 4
for i, (inputs, labels) in enumerate(train_loader):
    loss = criterion(model(inputs), labels) / accumulation_steps
    loss.backward()

    if (i+1) % accumulation_steps == 0:
        optimizer.step()
        optimizer.zero_grad()