Jetstream by default #118
Open
Jetstream by default #118
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
Implementation is slightly different, so a separate test is added.
Most tests work for both, except for the continuous batching one. This allows to remove the old GPT2 based tests, that are quite slow and do not use any sharding or KV cache, so they might not really be representative of most relevant models on TGI.
There are equivalent tests now on the TinyLlama model, that run faster, use the KV cache and sharding. The only test that does not have an equivalence is the continuous batching one, but the test was not working for most other models, so I prefer to remove it anyway, as having it passing was not representative anyway of the current state.
|
The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update. |
Now that the engine is stable and tested, its engine is set as the default one for TGI.
tengomucho
force-pushed
the
jetstream-by-default
branch
from
078dfa4
to
07f74ff
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Outdated suggestions cannot be applied.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?
This makes all the changes to allow having the Jetstream Pytorch engine to be the default backend for TGI on TPUs.
This backend is reliable and performant and give the best throughput on TGI.