测试深度学习环境(GPU是否可调用)

迷南。 2022-12-26 08:17 278阅读 0赞

**安装好CUDA和cuDNN之后好不好用呢？当然要测试一下：**

> 代码思想部分来源于几大开源模型源代码，还没有仔细写，以后有时间再补充完整吧

# 一、输出本机的配置 #

import platform
    import sys
    import os
    import subprocess
    import glob
    
    IS_WINDOWS = sys.platform == 'win32'
    
    
    def _find_cuda_home():
        cuda_home = os.environ.get('CUDA_HOME') or os.environ.get('CUDA_PATH')
        if cuda_home is None:
            try:
                which = 'where' if IS_WINDOWS else 'which'
                nvcc = subprocess.check_output([which, 'nvcc']).decode().rstrip('\r\n')
                cuda_home = os.path.dirname(os.path.dirname(nvcc))
            except Exception:
                # Guess #3
                if IS_WINDOWS:
                    cuda_homes = glob.glob(
                        'C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v*.*')
                    if len(cuda_homes) == 0:
                        cuda_home = ''
                    else:
                        cuda_home = cuda_homes[0]
                else:
                    cuda_home = '/usr/local/cuda'
                if not os.path.exists(cuda_home):
                    cuda_home = None
        return cuda_home
    
    
    def _get_nvcc_info(cuda_home):
        nvcc = None
        if cuda_home is not None and os.path.isdir(cuda_home):
            try:
                nvcc = os.path.join(cuda_home, 'bin/nvcc')
                nvcc = subprocess.check_output("{} -V".format(nvcc), shell=True).decode()
                nvcc = nvcc.strip().split('\n')[-1]
            except subprocess.SubprocessError:
                nvcc = "Not Available"
        return nvcc
    
    
    def _get_gpu_info():
        try:
            gpu_info = subprocess.check_output(['nvidia-smi', '-L']).decode().strip()
            gpu_info = gpu_info.split('\n')
            for i in range(len(gpu_info)):  # i是GPU数量
                gpu_info[i] = ' '.join(gpu_info[i].split(' ')[:4])
        except:
            gpu_info = ' Can not get GPU information. Please make sure CUDA have been installed successfully.'
        return gpu_info
    
    
    env_info = {}
    env_info['platform'] = platform.platform()
    env_info['architecture'] = platform.architecture()
    env_info['uname'] = platform.uname()
    env_info['python_branch'] = platform.python_branch()
    env_info['python_implementation'] = platform.python_implementation()
    env_info['python_revision'] = platform.python_revision()
    env_info['python_version'] = platform.python_version()
    env_info['python_build'] = platform.python_build()
    env_info['python_compiler'] = platform.python_compiler()
    
    env_info['Python'] = sys.version.replace('\n', '')
    
    cuda_home = _find_cuda_home()
    env_info['cuda_home'] = cuda_home
    env_info['NVCC'] = _get_nvcc_info(cuda_home)
    env_info['CUDA_VISIBLE_DEVICES'] = os.environ.get('CUDA_VISIBLE_DEVICES')
    env_info['GPU'] = _get_gpu_info()
    
    try:
        gcc = subprocess.check_output(['gcc', '--version']).decode()
        gcc = gcc.strip().split('\n')[0]
        env_info['GCC'] = gcc
    except:
        pass
    
    for k, v in env_info.items():
        print('{}: {}'.format(k, v))

# 二、测试tensorflow-gpu #

import tensorflow as tf
    
    print(tf.test.is_gpu_available())
    
    hello=tf.constant(‘hello,world’)
    sess=tf.Session()
    print(sess.run(hello))
    
    a = tf.constant([1.0, 2.0, 3.0, 4.0, 5.0, 6.0], shape=[2, 3], name='a')
    b = tf.constant([1.0, 2.0, 3.0, 4.0, 5.0, 6.0], shape=[3, 2], name='b')
    c = tf.matmul(a, b)
    sess = tf.Session(config=tf.ConfigProto(log_device_placement=True))
    print(sess.run(c))

# 三、测试pytorch-gpu #

import torch
    
    print(torch.cuda.is_available())
    
    ngpu= 1
    device = torch.device("cuda:0" if (torch.cuda.is_available() and ngpu > 0) else "cpu")
    print(device)
    print(torch.cuda.get_device_name(0))
    print(torch.rand(3,3).cuda())

# 四、测试paddlepaddle-gpu #

import paddle
    
    env_info = {}
    compiled_with_cuda = paddle.is_compiled_with_cuda()
    env_info['Paddle compiled with cuda'] = compiled_with_cuda
    
    if compiled_with_cuda:
        v = paddle.get_cudnn_version()
        v = str(v // 1000) + '.' + str(v % 1000 // 100)
        env_info['cudnn'] = v
        if 'gpu' in paddle.get_device():
            gpu_nums = paddle.distributed.ParallelEnv().nranks
        else:
            gpu_nums = 0
        env_info['GPUs used'] = gpu_nums
    
    env_info['PaddlePaddle'] = paddle.__version__
    
    for k, v in env_info.items():
        print('{}: {}'.format(k, v))

import paddle.fluid
    
    paddle.fluid.install_check.run_check()
    # 如果出现“Your Paddle Fluid is installed succesfully!”，说明已成功安装