236 lines
7.5 KiB
Python
236 lines
7.5 KiB
Python
import importlib.util
|
|
import os
|
|
import requests
|
|
import subprocess
|
|
import time
|
|
import traceback
|
|
|
|
EXP_URL = "http://10.2.1.195:880"
|
|
PROJECT_NAME = "challenge_svm"
|
|
LOW_ACCURACY_THRESHOLD = 0.6
|
|
|
|
|
|
def load_user_function(file_path, function_name):
|
|
"""
|
|
动态加载函数。
|
|
"""
|
|
spec = importlib.util.spec_from_file_location("user_code", file_path)
|
|
user_code = importlib.util.module_from_spec(spec)
|
|
spec.loader.exec_module(user_code)
|
|
return getattr(user_code, function_name)
|
|
|
|
|
|
def get_commit_hash_url():
|
|
"""
|
|
获取最新的提交哈希值
|
|
"""
|
|
try:
|
|
commit_hash = (
|
|
subprocess.check_output(["git", "rev-parse", "HEAD"])
|
|
.strip()
|
|
.decode("utf-8")
|
|
)
|
|
path = os.getcwd().replace("/workspace/", "")
|
|
url = f"https://git.intcplab.cn/{path}/commit/{commit_hash}"
|
|
return url
|
|
except subprocess.CalledProcessError as e:
|
|
print(f"Error: {e}")
|
|
return None
|
|
|
|
|
|
def evaluate_svm(
|
|
user_train_func,
|
|
user_test_func,
|
|
train_data,
|
|
train_labels,
|
|
test_data,
|
|
test_labels,
|
|
log_details,
|
|
):
|
|
"""
|
|
使用编写的训练 SVM 函数训练权重，并用预测函数对测试集进行分类，并计算准确率。
|
|
|
|
Args:
|
|
- user_train_func (function): 编写的训练 SVM 函数。
|
|
- user_test_func (function): 编写的预测函数。
|
|
- train_data (numpy array): 训练数据集。
|
|
- train_labels (numpy array): 训练标签集。
|
|
- test_data (numpy array): 测试数据集。
|
|
- test_labels (numpy array): 测试标签集。
|
|
- log_details (list): 日志记录列表。
|
|
|
|
Returns:
|
|
- float: 分类准确率。
|
|
"""
|
|
correct = 0
|
|
total = 67 # SVM 测试样本数
|
|
temp_log_details = []
|
|
|
|
if len(test_data) != total:
|
|
log_details.append("测试集样本数不匹配，请检查 load_data 函数！")
|
|
print("测试集样本数不匹配，请检查 load_data 函数！")
|
|
log_details.append("成绩为零！")
|
|
print("成绩为零！")
|
|
return 0.0
|
|
|
|
if len(set(test_labels)) != 3:
|
|
log_details.append(
|
|
"本次 SVM 挑战赛要求不合并死亡与安乐死（即使用 SVM 完成三分类问题），请检查 load_data 函数！"
|
|
)
|
|
print(
|
|
"本次 SVM 挑战赛要求不合并死亡与安乐死（即使用 SVM 完成三分类问题），请检查 load_data 函数！"
|
|
)
|
|
log_details.append("成绩为零！")
|
|
print("成绩为零！")
|
|
return 0.0
|
|
|
|
try:
|
|
# 训练模型
|
|
time_start = time.time()
|
|
wegihts = user_train_func(train_data, train_labels)
|
|
time_end = time.time()
|
|
log_details.append(f"训练时间: {(time_end - time_start)*1000:.10f}ms")
|
|
print(f"训练时间: {(time_end - time_start)*1000:.10f}ms")
|
|
|
|
for i in range(len(test_data)):
|
|
# 获取测试数据中的每个样本
|
|
result = user_test_func(test_data[i], wegihts)
|
|
if result == test_labels[i] and test_labels[i] != -1:
|
|
correct += 1
|
|
else:
|
|
print(f"第 {i+1} 个数据被分类到 {result}，但实际是 {test_labels[i]}")
|
|
temp_log_details.append(f"{i+1}")
|
|
|
|
accuracy = correct / total
|
|
if accuracy >= LOW_ACCURACY_THRESHOLD:
|
|
log_details.append(f"以下测试次序的数据分类错误:")
|
|
log_details.append(", ".join(temp_log_details))
|
|
else:
|
|
log_details.append("分类准确率较低，成绩中不记录错误日志")
|
|
log_details.append(f"分类准确率: {accuracy * 100:.2f}%")
|
|
return accuracy
|
|
except Exception as e:
|
|
log_details.append(f"评测期间发生错误: {e}")
|
|
print(f"评测期间发生错误: {e}")
|
|
log_details.append(traceback.format_exc())
|
|
print(traceback.format_exc())
|
|
log_details.append("成绩为零！")
|
|
print("成绩为零！")
|
|
return 0.0
|
|
|
|
|
|
def upload_score(name, id, score_data):
|
|
"""
|
|
上传成绩到服务器。
|
|
"""
|
|
url = f"{EXP_URL}/upload_score"
|
|
try:
|
|
print("上传成绩中...")
|
|
# print(score_data)
|
|
response = requests.post(
|
|
url,
|
|
json={
|
|
"id": id,
|
|
"name": name,
|
|
"score": score_data["score"],
|
|
"running_time": score_data["running_time"],
|
|
"time": time.strftime(
|
|
"%Y-%m-%d %H:%M:%S", time.gmtime(time.time() + 8 * 3600)
|
|
),
|
|
"details": score_data["details"],
|
|
"project": PROJECT_NAME,
|
|
"hash": get_commit_hash_url(),
|
|
},
|
|
)
|
|
if response.status_code == 200:
|
|
print(response.json()["message"])
|
|
else:
|
|
print("上传成绩失败!\n原因:", response.json()["message"])
|
|
except Exception as e:
|
|
print(f"上传成绩出错: {e}")
|
|
|
|
|
|
def main(name, id):
|
|
user_code_file = "function.py" # 代码文件
|
|
user_train_function_name = "train_svm" # 训练 svm 函数名
|
|
user_test_function_name = "predict_svm" # 预测函数名
|
|
user_load_dataset_name = "load_data" # 加载数据集函数名
|
|
|
|
# 加载自定义函数
|
|
log_details = []
|
|
try:
|
|
user_load_dataset = load_user_function(user_code_file, user_load_dataset_name)
|
|
user_train_func = load_user_function(user_code_file, user_train_function_name)
|
|
user_test_func = load_user_function(user_code_file, user_test_function_name)
|
|
except Exception as e:
|
|
log_details.append(f"加载自定义函数时出错: {e}")
|
|
print("\n".join(log_details))
|
|
return
|
|
|
|
load_start_time = time.time()
|
|
|
|
# 加载数据集
|
|
try:
|
|
train_data, train_labels = user_load_dataset("horse-colic.data")
|
|
test_data, test_labels = user_load_dataset("horse-colic.test")
|
|
except Exception as e:
|
|
print(f"加载数据集时出错: {e}")
|
|
print("请检查 load_data 函数是否正确！")
|
|
print("本次评测无效，记录不做提交！")
|
|
return
|
|
|
|
load_end_time = time.time()
|
|
|
|
# 记录评估时间
|
|
start_time = time.time()
|
|
|
|
# 评估分类准确率
|
|
accuracy = evaluate_svm(
|
|
user_train_func,
|
|
user_test_func,
|
|
train_data,
|
|
train_labels,
|
|
test_data,
|
|
test_labels,
|
|
log_details,
|
|
)
|
|
|
|
# 计算运行时间
|
|
running_time = time.time() - start_time
|
|
|
|
# 输出结果
|
|
print("")
|
|
log_details.append(f"得分: {accuracy * 100:.2f}")
|
|
print(f"得分: {accuracy * 100:.2f}")
|
|
log_details.append(
|
|
f"加载数据集时间: {(load_end_time - load_start_time)*1000:.10f}ms"
|
|
)
|
|
print(f"加载数据集时间: {(load_end_time - load_start_time)*1000:.10f}ms")
|
|
log_details.append(f"运行时间: {running_time * 1000:.10f}ms")
|
|
print(f"运行时间: {running_time * 1000:.10f}ms")
|
|
|
|
# 上传成绩
|
|
score_data = {
|
|
"score": accuracy * 100,
|
|
"running_time": f"{running_time * 1000:.10f}ms",
|
|
"details": "<br>".join(log_details),
|
|
}
|
|
|
|
upload_score(name, id, score_data)
|
|
|
|
|
|
if __name__ == "__main__":
|
|
try:
|
|
from function import name, id
|
|
|
|
print(f"正在评测 {name}@{id} 的代码...\n")
|
|
main(name, id)
|
|
except Exception as e:
|
|
log_details = [
|
|
f"评测出错: {e}",
|
|
traceback.format_exc(),
|
|
"请检查代码和返回值类型是否正确！",
|
|
"本次评测无效，记录不做提交！",
|
|
]
|
|
print("\n".join(log_details))
|