Question

我尝试使用h5py将numpy数组保存在HDF文件中，如下所示：

with h5py.File("mfcc_aligned.hdf", "w") as aligned_f:
    # do stuff to create two numpy arrays, training_X and training_Y
    print(len(training_X)) # this returns the number of elements I expect in the the numpy arr
    aligned_f.create_dataset("train_X", data=training_X)
    aligned_f.create_dataset("train_Y", data=training_Y)
    # if I add a line here to access the datasets I just created, I see that aligned_f does indeed have two keys train_X and train_Y with the shapes I expect

然而，当程序结束并且我检查文件mfcc_aligned.hdf时，它恰好是800字节（比我预期的要小得多），并且没有密钥。我为这里发生的事情感到茫然。

提前感谢任何见解！

Answer 1

你试过吗（没有在评论中格式化）：

with h5py.File('mfcc_aligned.hdf', 'r') as hf:
    print hf['train_X'][:]

Answer 2

我的代码没有问题：

In [59]: import h5py
In [60]: training_X = np.arange(12).reshape(3,4)
In [61]: training_Y = np.arange(3).reshape(3,1)
In [62]: with h5py.File("mfcc_aligned.hdf", "w") as aligned_f:
    ...:     # do stuff to create two numpy arrays, training_X and training_Y
    ...:     print(len(training_X)) # this returns the number of elements I expe
    ...: ct in the the numpy arr
    ...:     aligned_f.create_dataset("train_X", data=training_X)
    ...:     aligned_f.create_dataset("train_Y", data=training_Y)
    ...:     
3
In [63]: f = h5py.File("mfcc_aligned.hdf")
In [64]: list(f.keys())
Out[64]: ['train_X', 'train_Y']
In [66]: f['train_X'].value
Out[66]: 
array([[ 0,  1,  2,  3],
       [ 4,  5,  6,  7],
       [ 8,  9, 10, 11]])
In [67]: f['train_Y'][:]
Out[67]: 
array([[0],
       [1],
       [2]])
In [68]: ll mfcc_aligned.hdf
-rw-rw-r-- 1 paul 2204 Mar 31 14:10 mfcc_aligned.hdf

使用h5py无法正确保存HDF文件

2 个答案: