fix: binary search for best context length avoiding oom #705
+104
−8
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
Signed-off-by: jerryzhuang <zhuangqhc@gmail.com>
zhuangqh
commented
Nov 19, 2024
ishaansehgal99
approved these changes
Nov 19, 2024
Signed-off-by: jerryzhuang <zhuangqhc@gmail.com>
Signed-off-by: jerryzhuang <zhuangqhc@gmail.com>
ishaansehgal99
approved these changes
Nov 19, 2024
Fei-Guo
approved these changes
Nov 19, 2024
ishaansehgal99
approved these changes
Nov 20, 2024
Closed
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Outdated suggestions cannot be applied.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.
Reason for Change:
fix: binary search for best context length avoiding oom
Issue Fixed:
find_max_available_seq_lenruns to oom when runningon the V100 16GB gpu with 128K context.
Notes for Reviewers:
In the worst case, it costs about 1minutes to find the best
length (running with phi3 medium model and 128k search space).
We set the context length to a safe value to avoid oom.
If the serving server receives a request which token length is longer
than
max_model_len, server will reject this request.example error message:
This model's maximum context length is 2 tokens. However, you requested 19 tokens (9 in the messages, 10 in the completion). Please reduce the length of the messages or completion.