ci: bench: add more ftype, fix triggers and bot comment (ggml-org#6466)

* ci: bench: change trigger path to not spawn on each PR

* ci: bench: add more file type for phi-2: q8_0 and f16.
- do not show the comment by default

* ci: bench: add seed parameter in k6 script

* ci: bench: artefact name perf job

* Add iteration in the commit status, reduce again the autocomment

* ci: bench: add per slot metric in the commit status

* Fix trailing spaces

Author: phymbert

.github/workflows/bench.yml

@@ -24,10 +24,10 @@ on:
   push:
     branches:
       - master
-    paths: ['.github/workflows/bench.yml', '**/CMakeLists.txt', '**/Makefile', '**/*.h', '**/*.hpp', '**/*.c', '**/*.cpp', '**/*.cu', '**/*.swift', '**/*.m', 'examples/server/bench/**.*']
+    paths: ['llama.cpp', 'ggml.c', 'ggml-backend.c', 'ggml-quants.c', '**/*.cu', 'examples/server/*.h*', 'examples/server/*.cpp']
   pull_request_target:
     types: [opened, synchronize, reopened]
-    paths: ['.github/workflows/bench.yml', '**/CMakeLists.txt', '**/Makefile', '**/*.h', '**/*.hpp', '**/*.c', '**/*.cpp', '**/*.cu', '**/*.swift', '**/*.m', 'examples/server/bench/**.*']
+    paths: ['llama.cpp', 'ggml.c', 'ggml-backend.c', 'ggml-quants.c', '**/*.cu', 'examples/server/*.h*', 'examples/server/*.cpp']
   schedule:
     -  cron: '04 2 * * *'
 
@@ -42,6 +42,16 @@ jobs:
       RUNNER_LABEL: Standard_NC4as_T4_v3 # FIXME Do not find a way to not duplicate it
       N_USERS: 8
       DURATION: 10m
+
+    strategy:
+      matrix:
+        model: [phi-2]
+        ftype: [q4_0, q8_0, f16]
+        include:
+          - model: phi-2
+            ftype: q4_0
+            pr_comment_enabled: "true"
+
     if: ${{ github.event.inputs.gpu-series == 'Standard_NC4as_T4_v3' || github.event.schedule || github.event.pull_request || github.head_ref == 'master' || github.ref_name == 'master' || github.event.push.ref == 'refs/heads/master' }}
     steps:
       - name: Clone
@@ -116,7 +126,7 @@ jobs:
               --scenario script.js \
               --duration ${{ github.event.inputs.duration || env.DURATION }} \
               --hf-repo ggml-org/models	 \
-              --hf-file phi-2/ggml-model-q4_0.gguf \
+              --hf-file ${{ matrix.model }}/ggml-model-${{ matrix.ftype }}.gguf \
               --model-path-prefix /models \
               --parallel ${{ env.N_USERS }} \
               -ngl 33 \
@@ -134,7 +144,7 @@ jobs:
 
       - uses: actions/upload-artifact@v4
         with:
-          name: benchmark-results
+          name: bench-server-${{ github.job }}-${{ env.RUNNER_LABEL }}-${{ matrix.model }}-${{ matrix.ftype }}
           compression-level: 9
           path: |
             examples/server/bench/*.jpg
@@ -146,7 +156,7 @@ jobs:
         with:
           authToken: ${{secrets.GITHUB_TOKEN}}
           sha: ${{ inputs.sha || github.event.pull_request.head.sha || github.sha }}
-          context: bench-server-baseline
+          context: bench-server-${{ github.job }}-${{ env.RUNNER_LABEL }}-${{ matrix.model }}-${{ matrix.ftype }}
           description: |
             ${{ env.BENCH_RESULTS }}
           state: 'success'
@@ -203,21 +213,26 @@ jobs:
       - name: Comment PR
         uses: mshick/add-pr-comment@v2
         id: comment_pr
-        if: ${{ github.event.pull_request != '' }}
+        if: ${{ github.event.pull_request != '' && matrix.pr_comment_enabled == 'true' }}
         with:
-          message-id: bench-${{ github.job }}-${{ env.RUNNER_LABEL }}
+          message-id: bench-server-${{ github.job }}-${{ env.RUNNER_LABEL }}-${{ matrix.model }}-${{ matrix.ftype }}
           message: |
-            📈 **llama.cpp server** for _${{ github.job }}_ on _${{ env.RUNNER_LABEL }}_: **${{ env.BENCH_ITERATIONS}} iterations** 🚀
+            <p align="center">
+
+            📈 **llama.cpp server** for _${{ github.job }}_ on _${{ env.RUNNER_LABEL }}_ for `${{ matrix.model }}`-`${{ matrix.ftype }}`: **${{ env.BENCH_ITERATIONS}} iterations** 🚀
+
+            </p>
+
+            <details>
+
+            <summary>Expand details for performance related PR only</summary>
 
             - Concurrent users: ${{ env.N_USERS }}, duration: ${{ github.event.inputs.duration || env.DURATION }}
             - HTTP request          : avg=${{ env.HTTP_REQ_DURATION_AVG }}ms        p(90)=${{ env.HTTP_REQ_DURATION_P_90_ }}ms fails=${{ env.HTTP_REQ_FAILED_PASSES }}, finish reason: stop=${{ env.LLAMACPP_COMPLETIONS_STOP_RATE_PASSES }} truncated=${{ env.LLAMACPP_COMPLETIONS_TRUNCATED_RATE_PASSES }}
             - Prompt processing (pp): avg=${{ env.LLAMACPP_PROMPT_TOKENS_AVG }}tk/s p(90)=${{ env.LLAMACPP_PROMPT_TOKENS_P_90_ }}tk/s **total=${{ env.LLAMACPP_PROMPT_TOKENS_TOTAL_COUNTER_RATE }}tk/s**
             - Token generation  (tg): avg=${{ env.LLAMACPP_TOKENS_SECOND_AVG }}tk/s p(90)=${{ env.LLAMACPP_TOKENS_SECOND_P_90_ }}tk/s **total=${{ env.LLAMACPP_COMPLETION_TOKENS_TOTAL_COUNTER_RATE }}tk/s**
             - ${{ env.BENCH_GRAPH_XLABEL }}
 
-            <details>
-
-            <summary>Time series</summary>
 
             <p align="center">
 

examples/server/bench/bench.py

@@ -16,6 +16,7 @@
 import matplotlib.dates
 import matplotlib.pyplot as plt
 import requests
+from statistics import mean
 
 
 def main(args_in: list[str] | None = None) -> None:
@@ -109,6 +110,7 @@ def main(args_in: list[str] | None = None) -> None:
 
     # Prometheus
     end_time = time.time()
+    prometheus_metrics = {}
     if is_server_listening("0.0.0.0", 9090):
         metrics = ['prompt_tokens_seconds', 'predicted_tokens_seconds',
                    'kv_cache_usage_ratio', 'requests_processing', 'requests_deferred']
@@ -127,6 +129,7 @@ def main(args_in: list[str] | None = None) -> None:
                 values = metric_data['data']['result'][0]['values']
                 timestamps, metric_values = zip(*values)
                 metric_values = [float(value) for value in metric_values]
+                prometheus_metrics[metric] = metric_values
                 timestamps_dt = [datetime.fromtimestamp(int(ts)) for ts in timestamps]
                 plt.figure(figsize=(16, 10), dpi=80)
                 plt.plot(timestamps_dt, metric_values, label=metric)
@@ -176,17 +179,20 @@ def main(args_in: list[str] | None = None) -> None:
 
     # 140 chars max for commit status description
     bench_results = {
+        "i": iterations,
         "req": {
-            "p90": data['metrics']["http_req_duration"]["p(90)"],
-            "avg": data['metrics']["http_req_duration"]["avg"],
+            "p90": round(data['metrics']["http_req_duration"]["p(90)"], 2),
+            "avg": round(data['metrics']["http_req_duration"]["avg"], 2),
         },
         "pp": {
-            "p90": data['metrics']["llamacpp_prompt_tokens"]["p(90)"],
-            "avg": data['metrics']["llamacpp_prompt_tokens"]["avg"],
+            "p90": round(data['metrics']["llamacpp_prompt_tokens"]["p(90)"], 2),
+            "avg": round(data['metrics']["llamacpp_prompt_tokens"]["avg"], 2),
+            "0": round(mean(prometheus_metrics['prompt_tokens_seconds']), 2),
         },
         "tg": {
-            "p90": data['metrics']["llamacpp_tokens_second"]["p(90)"],
-            "avg": data['metrics']["llamacpp_tokens_second"]["avg"],
+            "p90": round(data['metrics']["llamacpp_tokens_second"]["p(90)"], 2),
+            "avg": round(data['metrics']["llamacpp_tokens_second"]["avg"], 2),
+            "0": round(mean(prometheus_metrics['predicted_tokens_seconds']), 2),
         },
     }
     with open("results.github.env", 'a') as github_env:

examples/server/bench/script.js

@@ -87,6 +87,7 @@ export default function () {
         ],
         "model": model,
         "stream": false,
+        "seed": 42,
         "max_tokens": max_tokens
     }