GRU fix #250
Open
GRU fix #250
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
|
Unlike LSTM, GRU by design does not have separate hidden and forward output. They share the same. See this diagram. The |
|
I mean, you say in the article ot is the output of the layer and h is the hidden state. So it makes sense that you pass the output to the next layer and the hidden state to the next time step. I was wondering if you tried, or have some intuition on which option is better in performance because, computationally, they are very similar. |
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Outdated suggestions cannot be applied.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.
Looking at your formulas in the article I see your implementation of the GRU does not coincide with the code you provide. I don't want you to merge this fork since it would break compatibility. But I leave it here in case you want to discuss the performance of this fixed ConvGRU implementation. It seems you are recycling the hidden states as if it was the forward activation. It is a valid approach, but I see more reasonable to separate between hidden state and forward activation.