[Bugfix] Add Prefix Caching Warmup Step #3901
Conversation
| if self.cache_config.enable_prefix_caching: | ||
| self.model_runner.warmup_prefix_attn(self.gpu_cache) |
There was a problem hiding this comment.
Is this called only in profiling? Or each inference?
There was a problem hiding this comment.
warm_up_model is not called on the hotpath
vllm/worker/model_runner.py
Outdated
| computed. This thus triggers context_attention_fwd and generates | ||
| the code. | ||
| """ | ||
| NUM_ITERATIONS = 10 |
There was a problem hiding this comment.
I feel one iteration should be good enough?
There was a problem hiding this comment.
I thought so too, but empirically seemed that I needed more than 1 run to make timing stable, so I just picked 10.
This takes <1s so its not impactful to UX, but agree the code is a bit silly
Let me do some more experiments.
vllm/worker/model_runner.py
Outdated
| request_0 = SequenceGroupMetadata( | ||
| request_id="first_request", | ||
| is_prompt=True, | ||
| seq_data={0: SequenceData(prompt_tokens)}, | ||
| sampling_params=SamplingParams(temperature=0), | ||
| block_tables={0: block_table}, | ||
| ) | ||
| self.execute_model([request_0], kv_caches) | ||
|
|
||
| # Prompt forward with block 1 computed. (Triggers | ||
| # context_attention_fwd). | ||
| request_1 = SequenceGroupMetadata( | ||
| request_id="second_request", | ||
| is_prompt=True, | ||
| seq_data={0: SequenceData(prompt_tokens)}, | ||
| sampling_params=SamplingParams(temperature=0), | ||
| block_tables={0: block_table}, | ||
| computed_block_nums=block_table[:NUM_COMPUTED_BLOCKS], | ||
| ) |
There was a problem hiding this comment.
Can we just run request_1? I believe the only goal here is to activate the triton kernel.
There was a problem hiding this comment.
Yeah I think that should work
|
Will address. Thanks Zhuohan |
|
@robertgshaw2-neuralmagic Let me know when this PR is ready for another round of review! |
|
@robertgshaw2-neuralmagic This pr solves my problem. cc. |
|
This pull request has been automatically marked as stale because it has not had any activity within 90 days. It will be automatically closed if no further activity occurs within 30 days. Leave a comment if you feel this pull request should remain open. Thank you! |
|
This pull request has merge conflicts that must be resolved before it can be |
simon-mo
requested review from
alexm-neuralmagic,
comaniac and
njhill
as code owners
Adds warmup step for prefix caching.
Currently, vLLM runs over two out of three attention cases during warmup:
model_runner.profile_run)model_runner.capture_model)context_fwd_attentionruns prefill with KV cache. Currently, this is implemented as a triton function with @triton.jit decorator. As a result, the code is generate at runtime. Sincecontext_fwd_attentiondoes not run on the warmup step, this causes ~3sec delay on the first request that usescontext_fwd_attentioncc @SageMoore @ElizaWszola @tlrmchlsmth
FIX #3846
BEFORE SUBMITTING, PLEASE READ THE CHECKLIST BELOW AND FILL IN THE DESCRIPTION ABOVE
PR Checklist (Click to Expand)
Thank you for your contribution to vLLM! Before submitting the pull request, please ensure the PR meets the following criteria. This helps vLLM maintain the code quality and improve the efficiency of the review process.
PR Title and Classification
Only specific types of PRs will be reviewed. The PR title is prefixed appropriately to indicate the type of change. Please use one of the following:
[Bugfix]for bug fixes.[CI/Build]for build or continuous integration improvements.[Doc]for documentation fixes and improvements.[Model]for adding a new model or improving an existing model. Model name should appear in the title.[Frontend]For changes on the vLLM frontend (e.g., OpenAI API server,LLMclass, etc.)[Kernel]for changes affecting CUDA kernels or other compute kernels.[Core]for changes in the core vLLM logic (e.g.,LLMEngine,AsyncLLMEngine,Scheduler, etc.)[Hardware][Vendor]for hardware-specific changes. Vendor name should appear in the prefix (e.g.,[Hardware][AMD]).[Misc]for PRs that do not fit the above categories. Please use this sparingly.Note: If the PR spans more than one category, please include all relevant prefixes.
Code Quality
The PR need to meet the following code quality standards:
format.shto format your code.docs/source/if the PR modifies the user-facing behaviors of vLLM. It helps vLLM user understand and utilize the new features or changes.Notes for Large Changes
Please keep the changes as concise as possible. For major architectural changes (>500 LOC excluding kernel/data/config/test), we would expect a GitHub issue (RFC) discussing the technical design and justification. Otherwise, we will tag it with
rfc-requiredand might not go through the PR.What to Expect for the Reviews
The goal of the vLLM team is to be a transparent reviewing machine. We would like to make the review process transparent and efficient and make sure no contributor feel confused or frustrated. However, the vLLM team is small, so we need to prioritize some PRs over others. Here is what you can expect from the review process:
action-requiredlabel on the PR if there are changes required. The contributor should address the comments and ping the reviewer to re-review the PR.Thank You
Finally, thank you for taking the time to read these guidelines and for your interest in contributing to vLLM. Your contributions make vLLM a great tool for everyone!