I just recently update keras, and run my old codes. The trained CNN model cannot be saved for some reason, which makes me has to turn back to the old keras at this moment.
model.save(os.path.join(modelDir,'M48DA_ACDCRT.h5'),overwrite=True)
Traceback (most recent call last):
File "<ipython-input-33-a3a0461bebd2>", line 2, in <module>
model.save(os.path.join(modelDir,'M48DA_ACDCRT.h5'),overwrite=True)
File "C:\Users\YL\Anaconda2\lib\site-packages\keras\engine\topology.py", line 2472, in save
save_model(self, filepath, overwrite, include_optimizer)
File "C:\Users\YL\Anaconda2\lib\site-packages\keras\models.py", line 160, in save_model
dtype=val.dtype)
File "C:\Users\YL\Anaconda2\lib\site-packages\h5py\_hl\group.py", line 111, in create_dataset
self[name] = dset
File "h5py\_objects.pyx", line 54, in h5py._objects.with_phil.wrapper (C:\Minonda\conda-bld\h5py_1490029245709\work\h5py\_objects.c:2867)
File "h5py\_objects.pyx", line 55, in h5py._objects.with_phil.wrapper (C:\Minonda\conda-bld\h5py_1490029245709\work\h5py\_objects.c:2825)
File "C:\Users\YL\Anaconda2\lib\site-packages\h5py\_hl\group.py", line 276, in __setitem__
h5o.link(obj.id, self.id, name, lcpl=lcpl, lapl=self._lapl)
File "h5py\_objects.pyx", line 54, in h5py._objects.with_phil.wrapper (C:\Minonda\conda-bld\h5py_1490029245709\work\h5py\_objects.c:2867)
File "h5py\_objects.pyx", line 55, in h5py._objects.with_phil.wrapper (C:\Minonda\conda-bld\h5py_1490029245709\work\h5py\_objects.c:2825)
File "h5py\h5o.pyx", line 202, in h5py.h5o.link (C:\Minonda\conda-bld\h5py_1490029245709\work\h5py\h5o.c:3789)
RuntimeError: Unable to create link (Name already exists)
I also get this problem. Sometimes I save a model and it works fine. Other times I get this error. Then later I run the same code - pretty sure I have not changed anything - and it saves fine again.
I had the same issue after I applied a custom kernel regularizer to my model, but there is no problem using save_weights_only=True.
This issue has been automatically marked as stale because it has not had recent activity. It will be closed after 30 days if no further activity occurs, but feel free to re-open a closed issue if needed.
Most helpful comment
I had the same issue after I applied a custom kernel regularizer to my model, but there is no problem using
save_weights_only=True.