Lec8

正文

讲解用的代码

Copyimport random  
  
def run_training_run_txt_log(epochs, lr):  
    print(f"Training for {epochs} epochs with learning rate {lr}")

    offset = random.random() / 5  
   
    for epoch in range(2, epochs):  
        # simulating a training run  
        acc = 1 - 2 ** -epoch - random.random() / epoch - offset  
        loss = 2 ** -epoch + random.random() / epoch + offset  
        print(f"epoch={epoch}, acc={acc}, loss={loss}")  
  
# run a training run with a learning rate of 0.1  
run_training_run_txt_log(epochs=10, lr=0.01)

W&B

Copyimport wandb
import random

def run_training_run(epochs, lr):  
      print(f"Training for {epochs} epochs with learning rate {lr}")  
  
      wandb.init(  
            # Set the project where this run will be logged  
            project="example",  
            # 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):  
            # simulating a training run  
            acc = 1 - 2 ** -epoch - random.random() / epoch - offset  
            loss = 2 ** -epoch + random.random() / epoch + offset  
            print(f"epoch={epoch}, acc={acc}, loss={loss}")  
            wandb.log({"acc": acc, "loss": loss})  
  
      wandb.finish()  
  
run_training_run(epochs=10, lr=0.01)

wandb.finish()的作用和含义: 之前已经通过 wandb.log()、wandb.init(config=...)、artifact 上传等操作产生了一些数据，这些数据可能还没有全部同步到 WandB 服务器。finish() 会等待/触发这些已产生的数据完成同步，并把这个 run 标记为结束

Multiple experiments

Copydef run_multiple_training_runs(epochs, lrs):  
    for epoch in epochs:  
        for lr in lrs:  
            run_training_run(epoch, lr)  
  
# Try different values for the learning rate  
epochs = [100, 120, 140]  
lrs = [0.1, 0.01, 0.001, 0.0001]  
run_multiple_training_runs(epochs, lrs)

根据原始教程整理....