We can have a tokenizer anywhere. #2527
Merged
Conversation
This file contains hidden or 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
router/src/server.rs
Outdated
| let tokens: Vec<SimpleToken> = encoding | ||
| .get_ids() | ||
| .iter() | ||
| .zip(encoding.get_offsets()) |
There was a problem hiding this comment.
This zip will fail for Python.
There was a problem hiding this comment.
I "fixed" it by ignoring offsets if not present.
Python non rust will definitely exist in forms where offsets are not available (ByT5) so I'll assume we cannot have them.
In that case we'll send the raw ids and totally ignore the "text" part (since it doesn't have a good definition without offsets in the original query)
router/src/server.rs
Outdated
| let tokens: Vec<SimpleToken> = encoding | ||
| .get_ids() | ||
| .iter() | ||
| .zip(encoding.get_offsets()) |
OlivierDehaene
previously approved these changes
Sep 17, 2024
Narsil
force-pushed
the
omni_tokenizer
branch
4 times, most recently
from
0f0e6a7 to
982f474
Compare
Narsil
force-pushed
the
omni_tokenizer
branch
from
e4f8546 to
905731b
Compare
Comment on lines
+199
to
+202
| try: | ||
| self.embed_tokens = TensorParallelEmbedding(f"{prefix}.embeddings", weights) | ||
| except RuntimeError: | ||
| self.embed_tokens = TensorParallelEmbedding(f"{prefix}.embedding", weights) |
There was a problem hiding this comment.
Confused by this.
Why should it work the second time?
There was a problem hiding this comment.
No s. There's a difference between mamba-hf and mamba ssh here.
Narsil
force-pushed
the
omni_tokenizer
branch
from
8350797 to
bbbd9a6
Compare
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Outdated suggestions cannot be applied.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.
What does this PR do?
We can check that it works with
google/byt5-smallfor instance (And also deadlocks if multiple validation workers are called since they all fight for the GIL as I've owned the gil in each thread.)Fixes # (issue)
Before submitting
Pull Request section?
to it if that's the case.
documentation guidelines, and
here are tips on formatting docstrings.
Who can review?
Anyone in the community is free to review the PR once the tests have passed. Feel free to tag
members/contributors who may be interested in your PR.