Add EPP scale limits experiment and profiles#1468
Open
madhugoutham wants to merge 5 commits into
Open
Conversation
Add experiment definition and inference-perf profile for evaluating EPP throughput across replica counts (1-4), CPU allocations (1/4/8 cores), and routing strategies (default, optimized-baseline, active-request). 14 setup treatments with QPS ramp from 200 to 5000. Validated with dry-run: 14/14 treatments succeeded. Addresses llm-d/llm-d-router#1290. Signed-off-by: Madhu Goutham Reddy Ambati <mambati@redhat.com>
Add run_epp_scale.sh for standalone EPP scale testing and three inference-perf profiles (baseline, stress, GPU ramp). Signed-off-by: Madhu Goutham Reddy Ambati <mambati@redhat.com>
Add run_epp_scale.sh for standalone EPP scale testing and three inference-perf profiles (baseline, stress, GPU ramp). Signed-off-by: Madhu Goutham Reddy Ambati <mambati@redhat.com>
madhugoutham
requested review from
Vezio,
achandrasekar,
kalantar,
maugustosilva,
mengmeiye and
namasl
as code owners
| streaming: true | ||
| server: | ||
| type: vllm | ||
| model_name: meta-llama/Llama-3.1-8B-Instruct |
Collaborator
There was a problem hiding this comment.
model_name: REPLACE_ENV_LLMDBENCH_DEPLOY_CURRENT_MODEL
Collaborator
There was a problem hiding this comment.
Ah I see... these are rendered files.
Contributor
Author
There was a problem hiding this comment.
Added .yaml.in templates for the baseline and GPU profiles to match repo convention, removed the rendered .yaml files, and wired template rendering into the standalone script.
Convert baseline and GPU profiles to .yaml.in with REPLACE_ENV_*
placeholders. Remove rendered .yaml files. Add sed rendering to
run_epp_scale.sh so --model flag flows into the config.
Signed-off-by: Madhu Goutham Reddy Ambati <mambati@redhat.com>
Add session-affinity-filter as a routing strategy option. Use --set-file instead of --set for plugins config to avoid multiline YAML mangling. Make EPP image repo configurable via EPP_IMAGE_REPO env var. Signed-off-by: Madhu Goutham Reddy Ambati <mambati@redhat.com>
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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.
Add experiment definition and inference-perf profiles for evaluating EPP throughput across replica counts, CPU allocations, and routing strategies. Includes standalone script for EPP scale testing.
Addresses llm-d/llm-d-router#1290.