pytorch 연습

pytorch-practice-001.ipynb

import torch
import torch.nn as nn

# 입력 데이터의 크기: (배치 크기, 채널, 높이, 너비)
input_size = (1, 4, 4, 4)

# Conv2d 레이어 정의
conv = nn.Conv2d(in_channels=4, out_channels=2, kernel_size=(3,3), stride=1, padding=1, groups=2)

# 입력 데이터 생성
input_data = torch.randn(input_size)

# 컨볼루션 연산 수행
output = conv(input_data)

# 출력 데이터의 크기 출력
print("Input size:", input_data.size())
print(input_data)

print("Output size:", output.size())
print(output)

import torch.nn as nn

class Model(nn.Module):
    def __init__(self):
        super(Model, self).__init__()
        self.fc1 = nn.Linear(100, 50)
        self.bn = nn.BatchNorm1d(num_features=50)
        self.relu = nn.ReLU()
        self.fc2 = nn.Linear(50, 10)

    def forward(self, x):
        x = self.fc1(x)
        x = self.bn(x)
        x = self.relu(x)
        x = self.fc2(x)
        return x

print(input_data)

import torch
import torch.nn as nn

# 입력 텐서의 크기가 10이고 출력 텐서의 크기가 20인 선형 변환을 수행하는 nn.Linear 모듈 생성
linear = nn.Linear(4, 10)

# 입력 텐서 생성 (크기가 10인 벡터)
input_tensor = torch.zeros(4)

print("Input Tensor Size: ", input_tensor.size())
print(input_tensor)

# 선형 변환 수행 (입력 텐서를 출력 텐서로 변환)
output_tensor = linear(input_tensor)

# print("Input Tensor Size: ", input_tensor.size())
# print(input_tensor)
print("Linear Weight", linear.weight)
print("Linear Bias", linear.bias)
print(sum(linear.bias))
print("Output Tensor Size: ", output_tensor.size())
print(output_tensor)


x = torch.randn(1, 2, 3, 4)
flatten = nn.Flatten()
x_flatten = flatten(x)

print(x)
print(x_flatten)

import torch
import numpy as np
import pandas as pd

import matplotlib.pyplot as plt
import seaborn as sns

import glob
import IPython.display as ipd

glob.glob('./**/*mnist*.*')
['./sample_data/mnist_test.csv', './sample_data/mnist_train_small.csv']

df = pd.read_csv('./sample_data/mnist_train_small.csv')
df_test = pd.read_csv('./sample_data/mnist_test.csv')
# print(df.shape, df_test.shape)

def imshowTen(imgs,labels, n):
    fig,axs = plt.subplots(1, n, figsize=(n,2))

    for i, ax in enumerate(axs.flat):
        # print(torch.from_numpy(imgs[i].reshape(-1, 28, 28, 1)))
        ax.imshow(imgs[i], cmap='gray')
        ax.set_title(f"{labels[i]}")
        ax.set_xticks([])
        ax.set_yticks([])


n = 17
# print(df.iloc[:10])
data = df.iloc[:n, 1:].to_numpy()
labels = df.iloc[:n, 0].to_numpy()

print(data.shape)
imgs = data.reshape(-1,28,28,1)
imshowTen(imgs,labels, n)