Add Tensorflow backend #596
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
@stavros11 could you please fix conflicts? |
scarrazza
reviewed
Jun 21, 2022
There was a problem hiding this comment.
@stavros11 thanks, I have 2 comments:
- in this branch qibo does not show information about the backend in the first import, only after typing
qibo.set_backend("tensorflow"). - I cannot switch from CPU to GPU, the code prints "Using tensorflow backend on /CPU:0" but performance looks like GPU.
Thank you for reviewing, done.
This is normal after the refactoring because no backend is initialized during import. If you do an operation that requires backend, such as executing a circuit or creating a Hamiltonian, then the
Indeed the device switcher and default device identification was not implemented here. It should work properly after the latest push. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This is an attempt to add the tensorflow backend on top of numpy with as little additional code as possible. It is based on
tensorflow.experimental.numpyand the idea is to useself.np(instead ofnp) in every numpy call used in the numpy backend and then replace this withself.np = tnpfor the tensorflow backend. It seems to work fairly well and it includes GPU, since the tnp ND array is an alias fortf.Tensorand not a numpy array. I have not tested backpropagation but at least according to their docs it is expected to work.One potential disadvantage with this approach is that it requires to have a reference to the numpy module in the numpy backend, by doing
self.np = np. This breaks the pickle serialization of the backend object because Python'spicklecannot pickle modules. Note thatdilldoes not have this limitation (related stackoverflow discussion), however I am not sure how efficient this is, mostly in terms of the pkl file size. Moreover, this breaks parallel.py because multiprocessing uses pickle.