issue 连接
https://github.com/numpy/numpy/issues/3477
今天使用np.genfromtxt加载一个5G不到的csv文件,把我30G内存+10Gswap都干满了还是报了out of memory.
一开始以为是open方法慢呢,原来是np.genfromtxt
一定要注意!!!!
np.genfromtxt非常的慢,并且需要读取文件的10倍内存
有问题的代码实例
with open(bin_file, 'r') as f: #读取5G的文件居然需要10倍大小的内存!!!!! csv = np.genfromtxt(f, delimiter=",", dtype='float32') csv = csv.reshape((count, shape_x, shape_y, shape_c)) print(csv.shape) for i in range(count): dset[i,:,:,:] = csv[i, :,:,:]
替换方案
使用np.loadtxt
def write_h5(file): print("load file ", file) #csv = np.genfromtxt(f, delimiter=",", dtype='float32') csv = np.loadtxt(file, delimiter=',') print("load completed! ") csv = csv.reshape((total_count, shape_x, shape_y, shape_c)) h5f = h5py.File(file_name, 'w') print("create h5file ", file_name) h5f.create_dataset(name='data', data=csv) print("create h5file dataset") h5f.close()
————————————————
版权声明:本文为CSDN博主「安安爸Chris」的原创文章,遵循CC 4.0 BY-SA版权协议,转载请附上原文出处链接及本声明。
原文链接:https://blog.csdn.net/mimiduck/article/details/116954108