-
-
Notifications
You must be signed in to change notification settings - Fork 8
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Introduce TensorFlow support #4
Introduce TensorFlow support #4
Conversation
65a604a
to
b3f167a
Compare
b3f167a
to
28a39c9
Compare
All right, this seems to be at a state of basic functionality. It's able to convert TFP objects to their meta equivalents and reconstruct/reify them from scratch (e.g. a non-trivial model in TFP). The base objects (i.e. TF objects) corresponding to each meta object are constructed in the current TF graph environment. Since base objects can be constructed as intermediate objects—in order to get otherwise unavailable shape information and such—and not just when requested (e.g. via |
@brandonwillard Will this work in eager mode? |
Good question; haven't tried it out, yet. |
6bf6a34
to
a9706ae
Compare
8e5a43f
to
5671e44
Compare
5671e44
to
5b5a6dd
Compare
fd63c6c
to
c044f6d
Compare
This commit provides basic TensorFlow graph interop. There are still a few design questions and choices to iterate on (e.g. the TF namespace/graph used during intermediate base-object derived meta object construction steps), but the basic class wrappers, helper functions, term representation and unification are working. Closes pymc-devs#3.
a93ced9
to
6daa63d
Compare
This is an attempt to implement #3. First, it abstracts away the Theano specifics/dependencies from the meta objects, then it creates analogous objects for TensorFlow and adds
unification
andterm
/operator
/arguments
generic function implementations for TF objects.