[Frontend] Warn if user max_model_len is greater than derived max_model_len
#7080
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: Jefferson Fialho <jfialho@ibm.com>
Signed-off-by: Jefferson Fialho <jfialho@ibm.com>
|
👋 Hi! Thank you for contributing to the vLLM project. Once the PR is approved and ready to go, please make sure to run full CI as it is required to merge (or just use auto-merge). To run full CI, you can do one of these:
🚀 |
njhill
reviewed
Aug 2, 2024
There was a problem hiding this comment.
Thanks @fialhocoelho! Just have some minor rewording suggestions.
Some text nits Co-authored-by: Nick Hill <nickhill@us.ibm.com>
Some text nits Co-authored-by: Nick Hill <nickhill@us.ibm.com>
Signed-off-by: Jefferson Fialho <jfialho@ibm.com>
Signed-off-by: Jefferson Fialho <jfialho@ibm.com>
DarkLight1337
added
the
ready
|
@fialhocoelho could you merge in the main branch again, it should help with the CI test failures. |
|
@njhill sure |
sfc-gh-mkeralapura
pushed a commit
to sfc-gh-mkeralapura/vllm
that referenced
this pull request
Aug 12, 2024
…model_len` (vllm-project#7080) Signed-off-by: Jefferson Fialho <jfialho@ibm.com> Co-authored-by: Nick Hill <nickhill@us.ibm.com>
kylesayrs
pushed a commit
to neuralmagic/vllm
that referenced
this pull request
Aug 17, 2024
…model_len` (vllm-project#7080) Signed-off-by: Jefferson Fialho <jfialho@ibm.com> Co-authored-by: Nick Hill <nickhill@us.ibm.com>
Alvant
pushed a commit
to compressa-ai/vllm
that referenced
this pull request
Oct 26, 2024
…model_len` (vllm-project#7080) Signed-off-by: Jefferson Fialho <jfialho@ibm.com> Co-authored-by: Nick Hill <nickhill@us.ibm.com> Signed-off-by: Alvant <alvasian@yandex.ru>
KuntaiDu
pushed a commit
to KuntaiDu/vllm
that referenced
this pull request
Nov 20, 2024
…model_len` (vllm-project#7080) Signed-off-by: Jefferson Fialho <jfialho@ibm.com> Co-authored-by: Nick Hill <nickhill@us.ibm.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.
Explanation of Changes
Summary
This update addresses the handling of user-specified
MAX_SEQUENCE_LENGTHvalues that exceed the model's maximum sequence length. The behavior is controlled by the environment variableVLLM_ALLOW_LONG_MAX_MODEL_LEN, which defaults to0.@njhill @Yard1 @robertgshaw2-neuralmagic @maxdebayser I mistakenly closed the other PR while syncing with the main. Here is the link to the previous one: #5911 (to maintain comment traceability).
Motivation
Previously, an error was triggered when user
max_model_lenexceeded the derived value, potentially leading to unintended behavior or CUDA errors. By changing this to a warning, users are alerted without halting execution, allowing flexibility depending on their needs.Notes
VLLM_ALLOW_LONG_MAX_MODEL_LEN=0):MAX_SEQUENCE_LENGTHis greater than the model's maximum sequence length, an error will be raised. This prevents potential incorrect model outputs or CUDA errors that could arise from exceeding the model's capabilities.VLLM_ALLOW_LONG_MAX_MODEL_LEN=1):VLLM_ALLOW_LONG_MAX_MODEL_LENis set to1, the system will allow the user-specifiedMAX_SEQUENCE_LENGTHto exceed the model's maximum. Instead of raising an error, a warning message will be logged, advising users to ensure the value is correct and within the model's context size. This provides flexibility for advanced users who understand the risks and need to override the default limit.