[plugin][benchmark][dashboard] Enable performance dashboard for ATOM OOT mode#352
Draft
zejunchen-zejun wants to merge 3 commits intomainfrom
Draft
[plugin][benchmark][dashboard] Enable performance dashboard for ATOM OOT mode#352zejunchen-zejun wants to merge 3 commits intomainfrom
zejunchen-zejun wants to merge 3 commits intomainfrom
Conversation
Contributor
There was a problem hiding this comment.
Pull request overview
Adds an out-of-tree (OOT) vLLM benchmarking workflow and updates the benchmark summarization + dashboard pipeline so OOT and native (“ATOM”) results can coexist and be compared without collisions.
Changes:
- Introduces a new scheduled + manual GitHub Actions workflow to build an OOT benchmark image, run a model/config matrix, and publish results to the existing benchmark dashboard.
- Extends the benchmark summarizer and dashboard UI to treat “backend” (ATOM vs OOT) as a first-class dimension for sorting, keys, filters, and trend views.
- Adds OOT-specific model config JSON and a converter script to generate
github-action-benchmarkinput from OOT JSON outputs.
Reviewed changes
Copilot reviewed 7 out of 7 changed files in this pull request and generated 6 comments.
Show a summary per file
| File | Description |
|---|---|
.github/workflows/atom-vllm-oot-benchmark.yaml |
New OOT benchmark workflow: build/push image, run benchmark matrix, summarize/regress, and publish to dashboard. |
.github/scripts/summarize.py |
Adds backend-aware sorting/config keys and reporting so OOT/native results don’t collide. |
.github/scripts/oot_benchmark_to_dashboard.py |
New converter from OOT result JSONs to github-action-benchmark input format. |
.github/scripts/atom_test.sh |
Injects benchmark_backend=ATOM into native benchmark result JSON for downstream tooling. |
.github/scripts/atom_oot_test.sh |
Adjusts OOT server launch defaults and per-model args used by CI/workflows. |
.github/dashboard/index.html |
Updates dashboard parsing, filtering, rendering, and trends to include a backend dimension. |
.github/benchmark/oot_models.json |
New OOT benchmark model matrix definitions (runner/model/args/env). |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
Comment on lines
+82
to
+92
| IFS=';' read -ra SETS <<< "${PARAM_LISTS}" | ||
| MATRIX_JSON="[" | ||
| SEP="" | ||
| for SET in "${SETS[@]}"; do | ||
| IFS=',' read -ra PARAMS <<< "$SET" | ||
| INPUT_LEN="${PARAMS[0]}" | ||
| OUTPUT_LEN="${PARAMS[1]}" | ||
| CONCURRENCY="${PARAMS[2]}" | ||
| RANDOM_RANGE_RATIO="${PARAMS[3]}" | ||
| MATRIX_JSON="${MATRIX_JSON}${SEP}{\"input_length\":${INPUT_LEN},\"output_length\":${OUTPUT_LEN},\"concurrency\":${CONCURRENCY},\"random_range_ratio\":${RANDOM_RANGE_RATIO}}" | ||
| SEP="," |
Comment on lines
+357
to
+359
| rm -rf .oot-benchmark/bench_serving | ||
| mkdir -p .oot-benchmark | ||
| git clone --depth 1 "${BENCH_SERVING_REPO_URL}" .oot-benchmark/bench_serving |
Comment on lines
+137
to
+142
| RUN git clone --depth 1 https://github.com/ROCm/aiter.git /app/aiter-test && \\ | ||
| cd /app/aiter-test && \\ | ||
| git checkout HEAD && \\ | ||
| git submodule sync && git submodule update --init --recursive && \\ | ||
| MAX_JOBS=64 PREBUILD_KERNELS=0 GPU_ARCHS=gfx950 python3 setup.py develop | ||
| RUN echo "=== Aiter version AFTER installation ===" && pip show amd-aiter || true |
Comment on lines
+12
to
+14
| VARIANT_RE = re.compile(r"-(mtp\d*)-") | ||
| DEFAULT_BACKEND = "OOT" | ||
|
|
Comment on lines
180
to
188
| nohup vllm serve "${resolved_model_path}" \ | ||
| --host "${VLLM_HOST}" \ | ||
| --port "${VLLM_PORT}" \ | ||
| --async-scheduling \ | ||
| --load-format fastsafetensors \ | ||
| --max-model-len 16384 \ | ||
| "${extra_arg_array[@]}" \ | ||
| --kv-cache-dtype fp8 \ | ||
| --gpu-memory-utilization 0.9 \ | ||
| > "${VLLM_LOG_FILE}" 2>&1 & |
Comment on lines
95
to
101
| def _config_key(data): | ||
| """Unique identifier for matching a benchmark configuration across runs.""" | ||
| return ( | ||
| _backend_name(data), | ||
| _display_model(data), | ||
| int(data.get("random_input_len", 0)), | ||
| int(data.get("random_output_len", 0)), |
zejunchen-zejun
force-pushed
the
zejun/add_benchmark
branch
from
207aae1 to
ef827ea
Compare
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
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.
No description provided.