[Relay][Frontend] Fix tensorflow frontend lstm forget bias adding order #3410
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
# The first commit's message is: fix tensorflow frontend lstm forget bias adding order # This is the 2nd commit message: remove trailing space for merge
ttyang1018
changed the title
ttyang1018
changed the title
ttyang1018
changed the title
|
@ttyang1018 Thanks for reporting and also providing a fix. Can you help to ass a regression test case for the same ? |
OK. I modified the original TF LSTM test case. Using non-zero forget bias 0.5 instead, and get pass in the unit test. |
tqchen
approved these changes
Jun 27, 2019
|
THanks, @ttyang1018 @srkreddy1238 , this PR is now merged |
wweic
pushed a commit
to wweic/tvm
that referenced
this pull request
Jun 28, 2019
wweic
pushed a commit
to neo-ai/tvm
that referenced
this pull request
Jun 30, 2019
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Outdated suggestions cannot be applied.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.
Recently, I tried to convert my own RNN model trained with TF into TVM.
I found out the TF frontend code with LSTMBlockCell conversion have a bug.
According to TF's implementation, (line 59 - line 76)
https://github.com/tensorflow/tensorflow/blob/master/tensorflow/contrib/rnn/python/ops/lstm_ops.py
forget_gate f should adding forget_bias first, then doing the sigmoid operation. However, in the relay's frontend conversion code, the forget_gate is doing sigmoid first then add with forget_bias.
I have checked with TF_forward test case, since in the LSTM and PTB test case, the forget_bias is set to 0.0, so the test case couldn't detect this bug.
I changed the order on my own, and get the correct prediction on my own model.