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
Support SperableConv2D and one hot crossentropy #1944
Conversation
62cc9cb
to
3234520
Compare
PR validation pass |
""" | ||
|input_tensor = Input(shape=[3]) | ||
|target_tensor = Input(shape=[3]) | ||
|loss = categorical_crossentropy(target_tensor, input_tensor) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Looks good as some other guys put "input_tensor" at the first position.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
it's the sequence in keras API
p1.almostEqual(p2, 1e-3) should be(true) | ||
} | ||
} | ||
} |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Do we need a serialization test?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
added
Seq(Tensor[Float](4, 24, 24, 3).rand(), filter), | ||
0 | ||
) | ||
} |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Do we need a serialization test?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
we don't need for tf loaders, but have added for the new added operation
Seq(input, filterSize, gradOutput), | ||
0, 1e-3 | ||
) | ||
} |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Do we need a serialization test?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
we don't need for tf loaders, but have added for the new added operation
Seq(size, filter, Tensor[Float](4, 23, 23, 6).rand()), | ||
0 | ||
) | ||
} |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Do we need a serialization test?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
we don't need for tf loaders, but have added for the new added operation
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
if we save model, will tf loaders not be saved ?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
no. It is only used when load tf model, which will generate an operation. I have added for the operations test
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@yiheng it doesn't look so when I ran the SessionSpec unit test, seems loaders are part of the graph meaning it will be serialized as well
Any progress for this? |
@zhichao-li as required by the reviewer, need to add serialization test |
* Revert "Create deploy_googlenet_places365.prototxt" This reverts commit ceaa3f9346faa66e3c4355590d91c4eff5d78e2c. * fix piip python test issue
What changes were proposed in this pull request?
How was this patch tested?
unit test