def get_infos():
"""Get all information about all your graphics cards.
Returns:
dict: The returned result is a dict with 3 keys: count, driver_version and devices:
count: Number of gpus found
driver_version: The version of the system’s graphics driver
devices: It's a list and every item is a namedtuple Device which has 10 fields, for exzample id, name and fan_speed etc.
It should be noted that the Process field is also a namedtuple which has 11 fields.
"""
infos = {}
Device = namedtuple(
"Device",
[
"id",
"name",
"free",
"used",
"total",
"temperature",
"fan_speed",
"power_usage",
"power_state",
"process",
],
)
Process = namedtuple(
"Process",
[
"pid",
"memory_percent",
"status",
"username",
"num_threads",
"cpu_num",
"cpu_percent",
"name",
"cmdline",
"used_gpu_mem",
"create_time",
],
)
driver_version = pynvml.nvmlSystemGetDriverVersion().decode()
device_count = pynvml.nvmlDeviceGetCount()
devices = []
for i in range(device_count):
handle = pynvml.nvmlDeviceGetHandleByIndex(i)
name = pynvml.nvmlDeviceGetName(handle).decode()
mem_info = pynvml.nvmlDeviceGetMemoryInfo(handle)
power_usage = pynvml.nvmlDeviceGetPowerUsage(
handle) # Power usage in milliwatts mW
processes = pynvml.nvmlDeviceGetComputeRunningProcesses(
handle) # Which processes are using the GPU
# process_info = [(item.pid, item.usedGpuMemory) for item in process_info]
process_info = []
for p in processes:
# append Process object to process_info
pid = p.pid
used_gpu_mem = p.usedGpuMemory
p = psutil.Process(pid=pid)
_ = p.cpu_percent()
time.sleep(0.05)
process_info.append(
Process(
pid=pid,
memory_percent=p.memory_percent(),
status=p.status(),
username=p.username(),
num_threads=p.num_threads(),
cpu_num=p.cpu_num(),
cpu_percent=p.cpu_percent(),
name=p.name(),
cmdline=" ".join(p.cmdline()),
used_gpu_mem=used_gpu_mem,
create_time=p.create_time(),
))
try:
fan_speed = pynvml.nvmlDeviceGetFanSpeed(handle)
except pynvml.NVMLError_NotSupported as e:
fan_speed = None
power_usage = pynvml.nvmlDeviceGetPowerUsage(handle)
power_state = pynvml.nvmlDeviceGetPowerState(handle)
temperature = pynvml.nvmlDeviceGetTemperature(
handle, pynvml.NVML_TEMPERATURE_GPU)
devices.append(
Device(
id=i,
name=name,
free=mem_info.free,
used=mem_info.used,
total=mem_info.total,
temperature=temperature,
fan_speed=fan_speed,
power_usage=power_usage,
power_state=power_state,
process=process_info,
))
infos["count"] = device_count
infos["driver_version"] = driver_version
infos["devices"] = devices
return infos
def pwr_state(self):
"""Get the Device current Power State"""
return pynvml.nvmlDeviceGetPowerState(self.dev)
print(' 共 %s 块 GPU，名称为：'%deviceCount)
for i in range(deviceCount):
handle = pynvml.nvmlDeviceGetHandleByIndex(i)
print(" GPU", i, ":", pynvml.nvmlDeviceGetName(handle))
print('--------------')
for i in range(deviceCount):
print('查看第 %s 块GPU的显存、温度、风扇、电源: '%i)
handle = pynvml.nvmlDeviceGetHandleByIndex(i)
info = pynvml.nvmlDeviceGetMemoryInfo(handle)
print("Memory Total: %0.2f G"%(info.total/1024/1024/1024)) # 总的显存大小
print("Memory Free: %0.2f G "%(info.free/1024/1024/1024)) # 剩余显存大小
print("Memory Used: %0.2f G "%(info.used/1024/1024/1024))
print("Memory Used percent: %0.2f %% "%(info.used/info.total*100))
print("Temperature is %d C"%(pynvml.nvmlDeviceGetTemperature(handle,0)))
print("Fan speed is ",pynvml.nvmlDeviceGetFanSpeed(handle))
print("Power ststus",pynvml.nvmlDeviceGetPowerState(handle))
print('--------------')
#最后要关闭管理工具
pynvml.nvmlShutdown()
#%% ----------------- os ----------------------
os.listdir(r'c:\windows')
os.getcwd() # 当前工作目录
os.chdir('C:\Users\Python_Folder') # 改变工作目录到dirname
os.curdir # 返回当前工作目录
os.__file__ # D:\envs\py27\lib\os.pyc
os.rename("python26","python21")
shutil.move("python21","python20")
os.path.dirname(os.__file__) # 获取路径名 D:\software\Anaconda3\envs\py27\lib
memory_max = 0
temprature_max = 0
GPU_load_max = 0
while(1):
info = psutil.virtual_memory()
memory_max = max(memory_max, info.percent)
print(u'内存占比：%4.1f/%4.1f' % (info.percent, memory_max), end=" ")
meminfo = pynvml.nvmlDeviceGetMemoryInfo(handle)
print("GPU memory: %dMB/%dMB" % (meminfo.used / pow(1024, 2), meminfo.total / pow(1024, 2)), end=" ")
temprature = pynvml.nvmlDeviceGetTemperature(handle, 0)
temprature_max = max(temprature_max, temprature)
print("temprature: %d℃/%d℃" % (temprature, temprature_max), end=" ")
PowerState = pynvml.nvmlDeviceGetPowerState(handle)
print("PowerState: %d" % (PowerState), end=" ")
PowerUsage = pynvml.nvmlDeviceGetPowerUsage(handle)
PowerManagementDefaultLimit = pynvml.nvmlDeviceGetPowerManagementDefaultLimit(handle)
print("PowerUsage: %d/%d" % (PowerUsage, PowerManagementDefaultLimit), end=" ")
GPU_load = PowerUsage / PowerManagementDefaultLimit * 100
GPU_load_max = max(GPU_load_max, GPU_load)
print("Power load: %4.1f/%4.1f%%" % (GPU_load, GPU_load_max))
time.sleep(1)
if 0:
print(u'cpu个数：',psutil.cpu_count())
print(u'内存使用：', psutil.Process(os.getpid()).memory_info().rss)
except np.NVMLError, err:
temp = 'NA'
try:
powMan = pn.nvmlDeviceGetPowerManagementMode(handle) #获取设备当前的电源管理模式
except pn.NVMLError, err:
powMan = 'NA'
try:
graphics_clock = pn.nvmlDeviceGetClockInfo(handle, pn.NVML_CLOCK_GRAPHICS) #检索设备的当前时钟速度
except pn.NVMLError, err:
graphics_clock = 'NA'
try:
mem_clock = pn.nvmlDeviceGetClockInfo(handle, pn.NVML_CLOCK_MEM)
except np.NVMLError, err:
mem_clock = 'NA'
try:
perf_stat = pn.nvmlDeviceGetPowerState(handle) #检索设备的当前性能状
except np.NVMLError, err:
perf_stat = 'NA'
tmp_dict['Gpu_Id'] = gpu_id
tmp_dict['Product_Name'] = product_name
tmp_dict['Mode'] = mode
tmp_dict['Current_Driver_Model'] = Current_driver_model
tmp_dict['Total_Memory'] = mem_total
tmp_dict['Used_Memory'] = mem_used
tmp_dict['Free_Memory'] = mem_free
tmp_dict['GPU_Util'] = gpu_util+'%'
tmp_dict['Memory_Util'] = mem_util+'%'
tmp_dict['Temperature'] = str(temp)+'C'
tmp_dict['Graphics_Clock'] = graphics_clock
tmp_dict['Memory_Clock'] = mem_clock
tmp_dict['Perf_State'] = perf_stat
def _get_device_power_state(gpu):
power_state = "P{}".format(pynvml.nvmlDeviceGetPowerState(gpu))
return {'power_state': power_state}