Use kaggle to run Li Hongyi's machine learning homework

With hw4 For example
register kaggle Account needs tizi To get the verification code , No need to log in later tizi
Use GPU You need to fill in your mobile phone number , Also needed tizi Verify the captcha , There are... Every week 37 Hours of gpu Use your time , You can also use cpu

1. First read the training data

According to the directory structure on the right

Modify the path to read the dataset

def parse_args():
	"""arguments"""
	config = {
    
	# Modify the path to read the dataset 
		"data_dir": "/kaggle/input/ml2022spring-hw4/Dataset",
		"save_path": "model.ckpt",
		"batch_size": 32,
		"n_workers": 8,
		"valid_steps": 2000,#2000
		"warmup_steps": 1000, #1000
		"save_steps": 10000, #10000
		"total_steps": 70000, #70000
	}

2. Read the training set address / Save the model address /test Data set prediction results

def parse_args():
	"""arguments"""
	config = {
    
		"data_dir": "/kaggle/input/ml2022spring-hw4/Dataset",
		"model_path": "/kaggle/working/model.ckpt",
		"output_path": "/kaggle/working/output.csv",
	}

	return config

3. Use pip

First turn the right internet Turn on

pip Remember to add before ！