Fix SuScaledRotaryEmbedding for Phi-3 mini 128k #155
Closed
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
This was referenced Nov 15, 2024
awni
reviewed
Nov 18, 2024
Comment on lines
+35
to
+44
| if longFactorArray.size == 1 { | ||
| // If longFactor is a single value, broadcast it | ||
| self._freqs = MLXArray(longFactor[0]) * freqs | ||
| } else { | ||
| // If longFactor is an array, it should match the frequency dimensions | ||
| precondition( | ||
| longFactorArray.size == freqs.size, | ||
| "longFactor size must match frequency dimensions") | ||
| self._freqs = longFactorArray * freqs | ||
| } |
There was a problem hiding this comment.
I don't see why this condition is necessary. Once you have the longFactorArray it will broadcast correctly if it has size one and if it has the wrong size it will raise an error. So you could replace it like so:
Suggested change
| if longFactorArray.size == 1 { | |
| // If longFactor is a single value, broadcast it | |
| self._freqs = MLXArray(longFactor[0]) * freqs | |
| } else { | |
| // If longFactor is an array, it should match the frequency dimensions | |
| precondition( | |
| longFactorArray.size == freqs.size, | |
| "longFactor size must match frequency dimensions") | |
| self._freqs = longFactorArray * freqs | |
| } | |
| self._freqs = longFactorArray * freqs |
There was a problem hiding this comment.
In that case, it looks like no change is necessary in SuScaledRotaryEmbedding. I'm confused, because I think I was able to reproduce the error in #92, which now no longer seems to occur.
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Outdated suggestions cannot be applied.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.
I've verified that this works with Phi-3 mini 128k 4-bit and Phi-3.5 mini 4-bit. I think this should fix #92 and #127.