[PyTorch] Anaconda, PyTorch 설치 및 실행

[실습] Anaconda 실습 환경 설정

https://www.anaconda.com/download

Download Anaconda Distribution | Anaconda

Download 후 기본 설정 값으로 다운해주자.

 

가상 환경을 만들고 접속하자.

(base) C:\Users\032> conda create -n torch_book python=3.9.0

(base) C:\Users\032> conda env list
# conda environments:
#
# base                 * C:\Users\032\anaconda3
# torch_book             C:\Users\032\anaconda3\envs\torch_book

(base) C:\Users\032> activate torch_book

# (torch_book) C:\Users\032> conda deactivate

# (base) C:\Users\032> conda env remove -n torch_book
# Remove all packages in environment C:\Users\032\anaconda3\envs\torch_book:

 

PyTorch, Jupyter notebook을 설치하고 실행하자.

 

https://pytorch.org/

PyTorch

 

(torch_book) C:\Users\032> pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118
(torch_book) C:\Users\032> pip install jupyter notebook

# pythorch 설치 확인
(torch_book) C:\Users\032> python
Python 3.9.0 (default, Nov 15 2020, 08:30:55) [MSC v.1916 64 bit (AMD64)] :: Anaconda, Inc. on win32
Type "help", "copyright", "credits" or "license" for more information.
>>> import torch
>>> torch.cuda.is_available()
True
>>> exit()

(torch_book) C:\Users\032> jupyter notebook

 

 

Cuda 버젼 확인 >

https://docs.nvidia.com/cuda/cuda-toolkit-release-notes/index.html

https://ko.wikipedia.org/wiki/CUDA

 

그래픽 카드 드라이버의 버전은 528.02이고 설치 가능한 최대 CUDA 버전은 12.0입니다. 

(base) C:\Users\032>nvidia-smi
Fri Jan 17 10:01:31 2025
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 528.02       Driver Version: 528.02       CUDA Version: 12.0     |
|-------------------------------+----------------------+----------------------+
| GPU  Name            TCC/WDDM | Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  NVIDIA GeForce ... WDDM  | 00000000:01:00.0 Off |                  N/A |
| N/A   42C    P0    14W /  50W |      0MiB /  4096MiB |      0%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+

+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
|  No running processes found                                                 |
+-----------------------------------------------------------------------------+

---

[실습] PyTorch 기초 문법

pip install requests

import torchvision.transforms as transforms

mnist_transform = transforms.Compose([
    transforms.ToTensor(),
    transforms.Normalize((0.5,) , (1.0,)) # 평균이 0.5, 표준편차가 1.0
])

from torchvision.datasets import MNIST
import requests
download_root = r'C:\Users\032\SKT_AI'

train_dataset = MNIST(download_root, transform=mnist_transform, train=True, download=True)
valid_dataset = MNIST(download_root, transform=mnist_transform, train=False,download=True)
test_dataset = MNIST(download_root, transform=mnist_transform, train=False, download=True)

import torch.nn as nn
model = nn.Linear(in_features=1, out_features=1, bias=True)

class MLP(Module):
    def __init__(self, inputs):
        super(MLP, self).__init__()
        self.layer = Linear(inputs,1)
        self.activation = Sigmoid()
    def forward(self,X):
        X = self.layer(X)
        X = self.activation(X)
        return X
        

import torch.nn as nn
class MLP(nn.Module):
    def __init__(self):
        super(MLP,self).__init__()
        self.layer1 = nn.Sequential(
            nn.Conv2d(in_channels=3, out_channels=64, kernel_size=5),
            nn.ReLU(inplace=True),
            nn.MaxPool2d(2))
        self.layer2 = nn.Sequential(
            nn.Conv2d(in_channels=64, out_channels=30, kernel_size=5),
            nn.ReLU(inplace=True),
            nn.MaxPool2d(2))
        self.layer3 = nn.Sequential(
            nn.Linear(in_features=30*5*5, out_features=10, bias=True),
            nn.ReLU(inplace=True))

    def forward(self,x):
        x=self.layer1(x)
        x=self.layer2(x)
        x=x.view(x.shape[0], -1)
        x=self.layer3(x)
        return x

model = MLP()

print("Printing childrne\n-----------")
print(list(model.children()))
print("\n\n Printing Modules \n------")
print(list(model.modules()))

---

pip install torchmetrics

import torch
import torchmetrics

# 랜덤한 예측값과 타겟 생성
preds = torch.randint(5, (10,))   # 클래스 0~4
target = torch.randint(5, (10,)) # 클래스 0~4

# 정확도 계산 (멀티클래스)
acc = torchmetrics.functional.accuracy(preds, target, task="multiclass", num_classes=5)
print(f"Accuracy: {acc:.4f}")
# Accuracy: 0.4000

---

# 라이브러리 호출
import torch
import torch.nn as nn
import torch.optim as optim
from torchvision import datasets, transforms
from torchvision.datasets import MNIST
from torch.utils.data import DataLoader

# 하이퍼파라미터 세팅
batch_size = 64
epoch = 10
lr = 0.001

# 데이터 전처리 정의
transform = transforms.Compose([
    transforms.ToTensor(),
    transforms.Normalize((0.5,), (0.5,))
])

# 데이터 로딩
download_root = '../chap02/data/MNIST_DATASET'

train_dataset = MNIST(download_root, transform=transform, train=True, download=True)
valid_dataset = MNIST(download_root, transform=transform, train=False, download=True)
test_dataset = MNIST(download_root, transform=transform, train=False, download=True)

train_loader = DataLoader(dataset=train_dataset, batch_size=batch_size, shuffle=True)
valid_loader = DataLoader(dataset=valid_dataset, batch_size=batch_size, shuffle=True)
test_loader = DataLoader(dataset=test_dataset, batch_size=batch_size, shuffle=True)

# 모델 정의
class SimpleNN(nn.Module):
    def __init__(self):
        super(SimpleNN, self).__init__()
        # 레이어 정의
        self.fc1 = nn.Linear(28 * 28, 128)
        self.fc2 = nn.Linear(128, 64)
        self.fc3 = nn.Linear(64, 10)

    def forward(self, x):
        x = x.view(-1, 28 * 28)  # 2차원 -> 1차원 형태로 변환
        x = torch.relu(self.fc1(x))
        x = torch.relu(self.fc2(x))
        x = self.fc3(x)
        return x

model = SimpleNN()

# 옵티마이저와 손실 함수 정의
loss_fn = nn.CrossEntropyLoss()
optimizer = optim.Adam(model.parameters(), lr=lr)

# 모델 학습
for i in range(epoch):
    for images, labels in train_loader:
        optimizer.zero_grad()  # 그라디언트 초기화
        preds = model(images)  # 모델 예측
        loss = loss_fn(preds, labels)  # 손실 계산
        loss.backward()  # 역전파
        optimizer.step()  # 가중치 업데이트

    # 1 epoch 완료
    print(f'[Epoch]| [{i + 1}/{epoch}], [Loss]| {loss.item():.4f}')
    

# [Epoch]| [1/10], [Loss]| 0.3773
# [Epoch]| [2/10], [Loss]| 0.0620
# [Epoch]| [3/10], [Loss]| 0.0596
# [Epoch]| [4/10], [Loss]| 0.0694
# [Epoch]| [5/10], [Loss]| 0.0087
# [Epoch]| [6/10], [Loss]| 0.0036
# [Epoch]| [7/10], [Loss]| 0.0704
# [Epoch]| [8/10], [Loss]| 0.0185
# [Epoch]| [9/10], [Loss]| 0.0634
# [Epoch]| [10/10], [Loss]| 0.0047

# model valid
model.eval() # 모델 eval 시작 = 파라미터 업데이트 X
total = 0 
correct = 0

with torch.no_grad(): # 기울기 계산 비활성화
    for images, labels in test_loader:
        preds = model(images) # 모델 예측
        _, predicted = torch.max(preds.data, 1)
        total += labels.size(0) # 전체 샘플 수
        correct += (predicted == labels).sum().item()

accuracy = (correct / total) * 100

# 정확도 계산
print(f'[Accuracy of the model on test dataset]: {accuracy:.2f}')

# [Accuracy of the model on test dataset]: 97.45