-
Notifications
You must be signed in to change notification settings - Fork 74k
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
tf.Fill has no gradient. (Was: "ValueError: No inputs provided" when creating optimizer, seems like a bug) #686
Comments
Sorry, I don't know what is going wrong here. I'll see if I can get someone to look at it. |
(a) the code is broken: it uses tf.types.float32, which should just be tf.float32. (b) The problem is because when you're doing the tf.select(draws, mx_fill, weights), you're potentially sending the gradients to the result of tf.fill(). Fill has no gradient registered. Fill should have a gradient. Hang on. |
Fixed. Should be pushed here within a day. |
(Thanks for catching this!) |
fixes github issue #686 Change: 112001823
Thanks for the quick turnaround guys, look forward to the update! |
(just to confirm, it got pushed in the above commit, so it's available at HEAD now) |
fixes github issue tensorflow#686 Change: 112001823
…pstream-deven-misc-191030 MLIR unit test updates in the XLA testsuite
I am in the process of implementing a version of BinaryConnect in tensorflow, and ran into this weird crash. I pasted a simple example to demonstrate the crash. Seems that line w = tf.select(draws,mx_fill,-mx_fill) is causing the problem when the optimization graph is being constructed.
The text was updated successfully, but these errors were encountered: