the weights file will not be closed after called load_weights function.

hi, I find that if I load weights from a h5 file, the file will not be closed after called load_weights function, so if I want to save the weights to this file again, it will throw an IO error like follows:

File "g:\Anaconda\lib\site-packages\keras\engine\training.py", line 1124, in fit
   callback_metrics=callback_metrics)
 File "g:\Anaconda\lib\site-packages\keras\engine\training.py", line 862, in _fit_loop
   callbacks.on_epoch_end(epoch, epoch_logs)
 File "g:\Anaconda\lib\site-packages\keras\callbacks.py", line 42, in on_epoch_end
   callback.on_epoch_end(epoch, logs)
 File "g:\Anaconda\lib\site-packages\keras\callbacks.py", line 298, in on_epoch_end
   self.model.save(filepath, overwrite=True)
 File "g:\Anaconda\lib\site-packages\keras\engine\topology.py", line 2423, in save
   save_model(self, filepath, overwrite)
 File "g:\Anaconda\lib\site-packages\keras\models.py", line 48, in save_model
   f = h5py.File(filepath, 'w')
 File "g:\Anaconda\lib\site-packages\h5py\_hl\files.py", line 222, in __init__
   fid = make_fid(name, mode, userblock_size, fapl)
 File "g:\Anaconda\lib\site-packages\h5py\_hl\files.py", line 85, in make_fid
   fid = h5f.create(name, h5f.ACC_TRUNC, fapl=fapl, fcpl=fcpl)
 File "h5f.pyx", line 90, in h5py.h5f.create (h5py\h5f.c:1998)
IOError: Unable to create file (Unable to truncate a file which is already open)

I check the load_weights function code in keras\engine\topology.py is like follows:

    def load_weights(self, filepath, by_name=False):
        import h5py
        f = h5py.File(filepath, mode='r')
        if 'layer_names' not in f.attrs and 'model_weights' in f:
            f = f['model_weights']
        if by_name:
            self.load_weights_from_hdf5_group_by_name(f)
        else:
            self.load_weights_from_hdf5_group(f)

        if hasattr(f, 'close'):
            f.close()

it will call f = f[‘model_weights’] if ‘model_weights’ in f and change f to a HDF5 group, which have no ‘close’ attribute. I think this is a bug.

[fchollet]

Fixed.

[bjtho08]

I made a small addition to the load_weights() method as shown below. I’ll submit a PR momentarily.

def load_weights(self, filepath, by_name=False,
                     skip_mismatch=False, reshape=False):
       """[docstring omitted for clarity]
       """
        
        if h5py is None:
            raise ImportError('`load_weights` requires h5py.')
        with h5py.File(filepath, mode='r') as f:
            if 'layer_names' not in f.attrs and 'model_weights' in f:
                f = f['model_weights']
            if by_name:
                saving.load_weights_from_hdf5_group_by_name(
                    f, self.layers, skip_mismatch=skip_mismatch,
                    reshape=reshape)
            else:
                saving.load_weights_from_hdf5_group(
                    f, self.layers, reshape=reshape)
            if hasattr(f, 'close'):
                f.close()
            elif hasattr(f.file, 'close'):
                f.file.close()