Upstream sync 2024 06 23 (#329)

SUMMARY: Upstream sync 2024 06 23 (#329) - not a release candidare SUMMARY: * Merge commits from vllm-project@0f0d8bc to vllm-project@6c916ac Note that vllm-project@0f0d8bc is NOT included in this merge. COMPARE vs UPSTREAM: https://github.com/neuralmagic/nm-vllm/compare/upstream-sync-2024-06-23..6c916ac8a80d1b2f4e0d0113a67767dc254a3598 --------- Signed-off-by: Thomas Parnell <tpa@zurich.ibm.com> Signed-off-by: Lei Wen <wenlei03@qiyi.com> Signed-off-by: Joe Runde <Joseph.Runde@ibm.com> Signed-off-by: kevin <kevin@anyscale.com> Signed-off-by: Rafael Vasquez <rafvasq21@gmail.com> Co-authored-by: Kuntai Du <kuntai@uchicago.edu> Co-authored-by: simon-mo <simon.mo@hey.com> Co-authored-by: Cyrus Leung <tlleungac@connect.ust.hk> Co-authored-by: Tyler Michael Smith <tyler@neuralmagic.com> Co-authored-by: Allen.Dou <allen.dou@hotmail.com> Co-authored-by: mgoin <michael@neuralmagic.com> Co-authored-by: Dipika Sikka <dipikasikka1@gmail.com> Co-authored-by: youkaichao <youkaichao@gmail.com> Co-authored-by: Sanger Steel <sangersteel@gmail.com> Co-authored-by: Thomas Parnell <tpa@zurich.ibm.com> Co-authored-by: Woosuk Kwon <woosuk.kwon@berkeley.edu> Co-authored-by: leiwen83 <leiwen83@users.noreply.github.com> Co-authored-by: Lei Wen <wenlei03@qiyi.com> Co-authored-by: SangBin Cho <rkooo567@gmail.com> Co-authored-by: Alexander Matveev <59768536+alexm-neuralmagic@users.noreply.github.com> Co-authored-by: Nick Hill <nickhill@us.ibm.com> Co-authored-by: zifeitong <zifei.tong@parasail.io> Co-authored-by: Antoni Baum <antoni.baum@protonmail.com> Co-authored-by: Cody Yu <hao.yu.cody@gmail.com> Co-authored-by: Amit Garg <gargamit@microsoft.com> Co-authored-by: Charles Riggins <liqianchen123@foxmail.com> Co-authored-by: Liqian Chen <liqian.chen@deeplang.ai> Co-authored-by: zhyncs <me@zhyncs.com> Co-authored-by: Kunshang Ji <kunshang.ji@intel.com> Co-authored-by: Jiang Li <jiang1.li@intel.com> Co-authored-by: Abhilash Majumder <abhilash.majumder@intel.com> Co-authored-by: Abhilash Majumder <30946547+abhilash1910@users.noreply.github.com> Co-authored-by: Jie Fu (傅杰) <jiefu@tencent.com> Co-authored-by: Bruce Fontaine <bruce@2.7182.net> Co-authored-by: zifeitong <zifeitong@gmail.com> Co-authored-by: sroy745 <142070531+sroy745@users.noreply.github.com> Co-authored-by: Isotr0py <2037008807@qq.com> Co-authored-by: Joe Runde <joe@joerun.de> Co-authored-by: Chang Su <chang.s.su@oracle.com> Co-authored-by: Roger Wang <136131678+ywang96@users.noreply.github.com> Co-authored-by: Kevin H. Luu <kevin@anyscale.com> Co-authored-by: Ronen Schaffer <ronen.schaffer@ibm.com> Co-authored-by: sergey-tinkoff <167607910+sergey-tinkoff@users.noreply.github.com> Co-authored-by: milo157 <43028253+milo157@users.noreply.github.com> Co-authored-by: Shukant Pal <SukantK2002@outlook.com> Co-authored-by: Hongxia Yang <62075498+hongxiayang@users.noreply.github.com> Co-authored-by: DearPlanet <junsong.zhang2021.work@outlook.com> Co-authored-by: Rafael Vasquez <rafvasq21@gmail.com> Co-authored-by: Varun Sundar Rabindranath <varunsundar08@gmail.com> Co-authored-by: Varun Sundar Rabindranath <varun@neuralmagic.com> Co-authored-by: Joshua Rosenkranz <joshua.rosenkranz@gmail.com> Co-authored-by: Davis Wertheimer <Davis.Wertheimer@ibm.com> Co-authored-by: Jinzhen Lin <linjinzhen@hotmail.com> Co-authored-by: Jee Li <pandaleefree@163.com> Co-authored-by: rohithkrn <rohith.nallamaddi@gmail.com> Co-authored-by: Roger Wang <ywang@roblox.com>
neuralmagic · Jun 26, 2024 · b4ad97a · b4ad97a
1 parent bcb2bd7
commit b4ad97a
Show file tree

Hide file tree

Showing 263 changed files with 9,376 additions and 1,617 deletions.
diff --git a/.buildkite/nightly-benchmarks/README.md b/.buildkite/nightly-benchmarks/README.md
@@ -0,0 +1,103 @@
+# vLLM benchmark suite
+
+## Introduction
+
+This directory contains the performance benchmarking CI for vllm.
+The goal is to help developers know the impact of their PRs on the performance of vllm.
+
+This benchmark will be *triggered* upon:
+- A PR being merged into vllm.
+- Every commit for those PRs with `perf-benchmarks` label.
+
+**Benchmarking Coverage**: latency, throughput and fix-qps serving on A100 (the support for more GPUs is comming later), with different models.
+
+**Benchmarking Duration**: about 1hr.
+
+**For benchmarking developers**: please try your best to constraint the duration of benchmarking to less than 1.5 hr so that it won't take forever to run.
+
+
+## Configuring the workload
+
+The benchmarking workload contains three parts:
+- Latency tests in `latency-tests.json`.
+- Throughput tests in `throughput-tests.json`.
+- Serving tests in `serving-tests.json`.
+
+See [descriptions.md](tests/descriptions.md) for detailed descriptions. 
+
+### Latency test
+
+Here is an example of one test inside `latency-tests.json`:
+
+```json
+[
+    {
+        "test_name": "latency_llama8B_tp1",
+        "parameters": {
+            "model": "meta-llama/Meta-Llama-3-8B",
+            "tensor_parallel_size": 1,
+            "load_format": "dummy",
+            "num_iters_warmup": 5,
+            "num_iters": 15
+        }
+    },
+]
+```
+
+In this example:
+-  The `test_name` attributes is a unique identifier for the test. In `latency-tests.json`, it must start with `latency_`.
+-  The `parameters` attribute control the command line arguments to be used for `benchmark_latency.py`. Note that please use underline `_` instead of the dash `-` when specifying the command line arguments, and `run-benchmarks-suite.sh` will convert the underline to dash when feeding the arguments to `benchmark_latency.py`. For example, the corresponding command line arguments for `benchmark_latency.py` will be `--model meta-llama/Meta-Llama-3-8B --tensor-parallel-size 1 --load-format dummy --num-iters-warmup 5 --num-iters 15`
+
+Note that the performance numbers are highly sensitive to the value of the parameters. Please make sure the parameters are set correctly.
+
+WARNING: The benchmarking script will save json results by itself, so please do not configure `--output-json` parameter in the json file.
+
+
+### Throughput test
+The tests are specified in `throughput-tests.json`. The syntax is similar to `latency-tests.json`, except for that the parameters will be fed forward to `benchmark_throughput.py`.
+
+The number of this test is also stable -- a slight change on the value of this number might vary the performance numbers by a lot.
+
+### Serving test
+We test the throughput by using `benchmark_serving.py` with request rate = inf to cover the online serving overhead. The corresponding parameters are in `serving-tests.json`, and here is an example:
+
+```
+[
+    {
+        "test_name": "serving_llama8B_tp1_sharegpt",
+        "qps_list": [1, 4, 16, "inf"],
+        "server_parameters": {
+            "model": "meta-llama/Meta-Llama-3-8B",
+            "tensor_parallel_size": 1,
+            "swap_space": 16,
+            "disable_log_stats": "",
+            "disable_log_requests": "",
+            "load_format": "dummy"
+        },
+        "client_parameters": {
+            "model": "meta-llama/Meta-Llama-3-8B",
+            "backend": "vllm",
+            "dataset_name": "sharegpt",
+            "dataset_path": "./ShareGPT_V3_unfiltered_cleaned_split.json",
+            "num_prompts": 200
+        }
+    },
+]
+```
+
+Inside this example:
+- The `test_name` attribute is also a unique identifier for the test. It must start with `serving_`.
+- The `server-parameters` includes the command line arguments for vLLM server.
+- The `client-parameters` includes the command line arguments for `benchmark_serving.py`.
+- The `qps_list` controls the list of qps for test. It will be used to configure the `--request-rate` parameter in `benchmark_serving.py`
+
+The number of this test is less stable compared to the delay and latency benchmarks (due to randomized sharegpt dataset sampling inside `benchmark_serving.py`), but a large change on this number (e.g. 5% change) still vary the output greatly.
+
+WARNING: The benchmarking script will save json results by itself, so please do not configure `--save-results` or other results-saving-related parameters in `serving-tests.json`.
+
+## Visualizing the results
+The `convert-results-json-to-markdown.py` helps you put the benchmarking results inside a markdown table, by formatting [descriptions.md](tests/descriptions.md) with real benchmarking results.
+You can find the result presented as a table inside the `buildkite/performance-benchmark` job page.
+If you do not see the table, please wait till the benchmark finish running.
+The json version of the table (together with the json version of the benchmark) will be also attached to the markdown file.
+The raw benchmarking results (in the format of json files) are in the `Artifacts` tab of the benchmarking.
diff --git a/.buildkite/nightly-benchmarks/benchmark-pipeline.yaml b/.buildkite/nightly-benchmarks/benchmark-pipeline.yaml
@@ -0,0 +1,62 @@
+steps:
+  - label: "Wait for container to be ready"
+    agents:
+      queue: A100
+    plugins:
+    - kubernetes:
+        podSpec:
+          containers:
+          - image: badouralix/curl-jq
+            command:
+            - sh
+            - .buildkite/nightly-benchmarks/scripts/wait-for-image.sh
+  - wait
+  - label: "A100 Benchmark"
+    agents:
+      queue: A100
+    plugins:
+    - kubernetes:
+        podSpec:
+          priorityClassName: perf-benchmark
+          containers:
+          - image: public.ecr.aws/q9t5s3a7/vllm-ci-test-repo:$BUILDKITE_COMMIT
+            command:
+            - bash .buildkite/nightly-benchmarks/run-benchmarks-suite.sh
+            resources:
+              limits:
+                nvidia.com/gpu: 8
+            volumeMounts:
+            - name: devshm
+              mountPath: /dev/shm
+            env:
+            - name: VLLM_USAGE_SOURCE
+              value: ci-test
+            - name: HF_TOKEN
+              valueFrom:
+                secretKeyRef:
+                  name: hf-token-secret
+                  key: token
+          nodeSelector:
+            nvidia.com/gpu.product: NVIDIA-A100-SXM4-80GB
+          volumes:
+          - name: devshm
+            emptyDir:
+              medium: Memory
+  # - label: "H100: NVIDIA SMI"
+  #   agents:
+  #     queue: H100
+  #   plugins:
+  #   - docker#v5.11.0:
+  #       image: public.ecr.aws/q9t5s3a7/vllm-ci-test-repo:$BUILDKITE_COMMIT
+  #       command:
+  #       - bash
+  #       - .buildkite/nightly-benchmarks/run-benchmarks-suite.sh
+  #       mount-buildkite-agent: true
+  #       propagate-environment: true
+  #       propagate-uid-gid: false
+  #       ipc: host
+  #       gpus: all
+  #       environment:
+  #       - VLLM_USAGE_SOURCE
+  #       - HF_TOKEN
+
diff --git a/.buildkite/nightly-benchmarks/kickoff-pipeline.sh b/.buildkite/nightly-benchmarks/kickoff-pipeline.sh
@@ -1,5 +1,6 @@
 #!/usr/bin/env bash
 
+# NOTE(simon): this script runs inside a buildkite agent with CPU only access.
 set -euo pipefail
 
 # Install system packages
@@ -23,4 +24,4 @@ if [ "$BUILDKITE_PULL_REQUEST" != "false" ]; then
 fi
 
 # Upload sample.yaml
-buildkite-agent pipeline upload .buildkite/nightly-benchmarks/sample.yaml
+buildkite-agent pipeline upload .buildkite/nightly-benchmarks/benchmark-pipeline.yaml