67 lines
2.2 KiB
Python
67 lines
2.2 KiB
Python
|
import psutil
|
|||
|
|
import time
|
|||
|
|
from datetime import datetime
|
|||
|
|
from pynvml import *
|
|||
|
|
|
|||
|
|
def get_memory_usage():
|
|||
|
|
mem = psutil.virtual_memory()
|
|||
|
|
return mem.percent, mem.used / (1024 ** 3), mem.total / (1024 ** 3)
|
|||
|
|
|
|||
|
|
def get_cpu_usage():
|
|||
|
|
return psutil.cpu_percent(interval=1) # 获取CPU利用率（百分比）
|
|||
|
|
|
|||
|
|
def get_gpu_usage():
|
|||
|
|
try:
|
|||
|
|
nvmlInit()
|
|||
|
|
device_count = nvmlDeviceGetCount()
|
|||
|
|
gpu_stats = []
|
|||
|
|
for i in range(device_count):
|
|||
|
|
handle = nvmlDeviceGetHandleByIndex(i)
|
|||
|
|
name = nvmlDeviceGetName(handle).decode("utf-8")
|
|||
|
|
mem_info = nvmlDeviceGetMemoryInfo(handle)
|
|||
|
|
util = nvmlDeviceGetUtilizationRates(handle)
|
|||
|
|
gpu_stats.append({
|
|||
|
|
"name": name,
|
|||
|
|
"gpu_util": util.gpu,
|
|||
|
|
"mem_util": util.memory,
|
|||
|
|
"mem_used": mem_info.used / (1024 ** 3),
|
|||
|
|
"mem_total": mem_info.total / (1024 ** 3)
|
|||
|
|
})
|
|||
|
|
nvmlShutdown()
|
|||
|
|
return gpu_stats
|
|||
|
|
except Exception:
|
|||
|
|
return [{"name": "N/A", "gpu_util": 0, "mem_util": 0, "mem_used": 0, "mem_total": 0}]
|
|||
|
|
|
|||
|
|
def main():
|
|||
|
|
log_file = "monitor_log.txt"
|
|||
|
|
with open(log_file, "a", encoding="utf-8") as f:
|
|||
|
|
while True:
|
|||
|
|
now = datetime.now().strftime("%Y-%m-%d %H:%M:%S")
|
|||
|
|
|
|||
|
|
# CPU
|
|||
|
|
cpu_percent = get_cpu_usage()
|
|||
|
|
|
|||
|
|
# 内存
|
|||
|
|
mem_percent, mem_used, mem_total = get_memory_usage()
|
|||
|
|
|
|||
|
|
# GPU
|
|||
|
|
gpu_stats = get_gpu_usage()
|
|||
|
|
|
|||
|
|
# 拼接日志内容
|
|||
|
|
log_line = (f"[{now}] CPU: {cpu_percent:.1f}% | "
|
|||
|
|
f"Memory: {mem_percent:.1f}% ({mem_used:.2f} GB / {mem_total:.2f} GB)")
|
|||
|
|
|
|||
|
|
for gpu in gpu_stats:
|
|||
|
|
log_line += (f" | GPU: {gpu['name']} {gpu['gpu_util']}% "
|
|||
|
|
f"(Mem {gpu['mem_used']:.2f} GB / {gpu['mem_total']:.2f} GB)")
|
|||
|
|
|
|||
|
|
# 输出
|
|||
|
|
print(log_line)
|
|||
|
|
f.write(log_line + "\n")
|
|||
|
|
f.flush()
|
|||
|
|
|
|||
|
|
time.sleep(9) # 这里+CPU统计的1秒，总体大约10秒一个周期
|
|||
|
|
|
|||
|
|
if __name__ == "__main__":
|
|||
|
|
main()
|