Implement MultiHeadAttention Layer #7875
Merged
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
|
@Linchenn Please take a look when you get a chance. Thanks! |
pforderique
added a commit
that referenced
this pull request
Aug 7, 2023
* Implement position embedding * Strip debug ops in jax conversion tests (#7889) INTERNAL This fixes an internal issue with jax tests. See cl/550054296. * Update weights loading (#7872) * Update weights loading * fix tests * remove * fix * fix comments * fix lint * Load python rules in tfjs-converter converters dir (#7892) * Implement MultiHeadAttention Layer (#7875) * Add spec for multi-head attention * Add CachedMultiHeadAttention cache * Fix typos * Lint * Add Transformer Decoder spec * lint * Add Einsum spec * lint * Remove unused type declaration * Move helper functions outside EinsumDense class * Implement Einsum Dense * Address comments * Implement MHA Layer * Add masked softmax support * Fix typo * Check for undef and null * Make buildFromSignature public * Wrap softmax call in tf.tidy * Implement position embedding --------- Co-authored-by: Matthew Soulanille <msoulanille@google.com> Co-authored-by: fengwuyao <131706622+fengwuyao@users.noreply.github.com>
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Outdated suggestions cannot be applied.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.
Implements the MultiHeadAttention layer from Keras attention layers.
NOTE:
Softmaxlayer was changed to support masking. Let me know if having a separate class for this is better. Regardless, the current change should be non-breaking.Depends on #7860.