storage以一维数组存储数值本身，metadata存储tensor相关信息
import logging
logging.debug("This is a debug message")
logging.info("This is an info message")
logging.warning("This is a warning message")
logging.error("This is an error message")
logging.critical("This is a critical message")output：默认只输出warning以上级别的信息
WARNING:root:This is a warning message
ERROR:root:This is an error message
CRITICAL:root:This is a critical message修改打印级别等配置
import logging
logging.basicConfig(level=logging.DEBUG, # 日志级别
# filename='test.log', # 写入本地文件
# filemode='a', # 追加模式,"w"为覆盖写入
format='%(asctime)s - [line:%(lineno)d] - %(levelname)s: %(message)s' # 日志格式
)启动命令：
tensorboard --logdir my_log
使用：
from torch.utils.tensorboard import SummaryWriter
writer = SummaryWriter(log_dir='runs/mock_accuracy')
for i in range(100):
writer.add_scalar(tag="accuracy", #可以暂时理解为图像的名字
scalar_value=i * random.uniform(0.8, 1), # 纵坐标的值
global_step=i # 当前是第几次迭代，可以理解为横坐标的值
)# 训练模型
total_steps = 0
for epoch in range(num_epochs):
for i, (inputs, targets) in enumerate(dataloader):
model.train()
outputs = model(inputs)
loss = criterion(outputs, targets)
optimizer.zero_grad()
loss.backward()
optimizer.step()
total_steps += 1
# 每个step记录损失到TensorBoard
writer.add_scalar('Loss/train', loss, total_steps)
if total_steps % 10 == 0:
print(f'Epoch [{epoch+1}/{num_epochs}], Step [{total_steps}], Loss: {loss.item():.4f}')
# 关闭SummaryWriter
writer.close()from torchsummary import summary
summary(model,(),bs)netron.app
api-key、登录
89220449c735e0cf8207f23230cf7b70ec98c04d
pip install wandb
import os
os.environ["wandb_api_key"]="89220449c735e0cf8207f23230cf7b70ec98c04d"
import wandb
wandb.login()管理超参
# train.py
import wandb
import random
epochs = 10
lr = 0.01
run = wandb.init(
# Set the project where this run will be logged
project="my-awesome-project",
# Track hyperparameters and run metadata
config={
"learning_rate": lr,
"epochs": epochs,
},
)
offset = random.random() / 5
print(f"lr: {lr}")
for epoch in range(2, epochs):
acc = 1 - 2**-epoch - random.random() / epoch - offset
loss = 2**-epoch + random.random() / epoch + offset
print(f"epoch={epoch}, accuracy={acc}, loss={loss}")
wandb.log({"accuracy": acc, "loss": loss})for children in model.children(): # 遍历所有直接子模块
print(children)
print('-----------------')
# for item in model.named_modules(): # 递归遍历所有子模块
# print(item)
# print("====================================")import logging
class EarlyStop:
def __init__(self, patience=30):
self.best_fiteness = 0.0 # 指标
self.best_epoch = 0
self.patience = patience or float("inf") # 早停的阈值
def __call__(self, epoch, fitness):
if fitness > self.best_fiteness:
self.best_fiteness = fitness
self.best_epoch = epoch
delta = epoch - self.best_epoch
stop = delta >= self.patience
if stop:
logging.info(f"Early stopping at epoch {epoch}")
return stop
stopper = EarlyStop()
stop = False
for epoch in range(epochs):
# train
# val
stop = stopper(epoch, fitness)
if stop:
break在使用imshow()展示图片前，需要先将图片读取出来。
读取图片可以通过pillow库，也可以用matplotlib本身自带的image模块实现。
import os
import matplotlib.image as mpimg
from PIL import Image
import matplotlib.pyplot as plt
import numpy as np
import matplotlib as mpl
mpl.rcParams['font.sans-serif'] = ['SimHei'] # 中文字体支持
# 使用pillow库读取图片，关闭坐标轴
img = Image.open(r"./jupyter/matplotlib/images/1.jpg")
fig = plt.figure(figsize=(8, 4))
ax1 = fig.add_subplot(111)
ax1.imshow(img)
ax1.axis('off')
plt.show()import numpy as np
import matplotlib.pyplot as plt
from sklearn.manifold import TSNE
vecArr = np.random.rand(1000,512) #构造1000个512维向量, 换成实际需要聚类的向量即可
tsneData = TSNE().fit_transform(vecArr)
#开始进行可视化
f = plt.figure(figsize=(10,10))
ax = plt.subplot(aspect='equal')
sc = ax.scatter(tsneData[:,0], tsneData[:,1])
plt.xlim(-50,50)
plt.ylim(-50,50)
ax.axis('off')
ax.axis('tight')
plt.show()
