Record machine load with every latency number
Latency measured while another job hogs the cores is not comparable: the scale curve moved from 1.315s to 0.809s at 50,000 sources between two runs of the same code. - bench/systems.py: machine_facts() reports cpus plus the 1/5/15 minute load average. - bench/run.py, bench/run_external.py: both reports carry it, and BASELINE.md labels the latency table a ceiling when the machine was busy. Verification: pytest bench/tests -q -> 60 passed; ruff check bench/ -> all checks passed; mypy (fresh cache) -> no issues in 25 source files.
This commit is contained in:
ZacharyZhang-NY committed
1 parent
61f6506104
commit
c4e2982506
6 files changed
+62
-27
No files matched your search
@@ -1,6 +1,6 @@
|
||||
# BASELINE: what the product retrieves today
|
||||
|
||||
- generated 2026-10-07T05:35:00+00:00 at commit a4953335e77f
|
||||
- generated 2026-10-07T15:33:42+00:00 at commit 29163632af74
|
||||
- corpus 50278 sources, 5308 queries (2928 answerable)
|
||||
- models: dense `sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2`, rerank `BAAI/bge-reranker-base`
|
||||
|
||||
@@ -33,14 +33,20 @@
|
||||
|
||||
## the unmodified product path
|
||||
|
||||
- RetrievalService.search: p50 15.54s, mean 16.68s, max 56.14s over 40 queries
|
||||
- RetrievalService.search: p50 20.16s, mean 20.67s, max 48.90s over 40 queries
|
||||
- top-10 overlap with the fast BM25 path: 0.0%, identical order 0.0%
|
||||
- its top-10 atoms cover only 7.6 distinct sources (min 1): the product ranks atoms and never diversifies, so ten slots fill with fragments of one document and the overlap with a source-level ranking is near zero
|
||||
|
||||
## latency vs corpus size
|
||||
|
||||
| sources | atoms | bm25_only | dense_only |
|
||||
|---|---|---|---|---|
|
||||
| 50000 | 165695 | 0.804/1.576 | 1.446/2.538 |
|
||||
| sources | atoms | bm25_only p50/p95 s | dense_only p50/p95 s |
|
||||
|---|---|---|---|
|
||||
| 500 | 500 | 0.001/0.002 | 1.380/1.576 |
|
||||
| 2000 | 2000 | 0.008/0.016 | 1.398/2.595 |
|
||||
| 10000 | 10000 | 0.052/0.128 | 1.338/2.266 |
|
||||
| 50000 | 165695 | 0.809/2.274 | 1.496/2.464 |
|
||||
|
||||
_latency only: a source prefix drops most gold sources, so recall at these sizes is meaningless_
|
||||
|
||||
## reading
|
||||
|
||||
|
||||
+20
-20
@@ -564,9 +564,9 @@
|
||||
"min": 1
|
||||
},
|
||||
"identical_top10": 0.0,
|
||||
"max": 63.2751505510023,
|
||||
"mean": 21.701527132049886,
|
||||
"p50": 21.831516211001144,
|
||||
"max": 48.895113198996114,
|
||||
"mean": 20.665571158625244,
|
||||
"p50": 20.157494062004844,
|
||||
"queries": 40,
|
||||
"top10_overlap_with_bm25_only": 0.0,
|
||||
"top_k": 10
|
||||
@@ -575,7 +575,7 @@
|
||||
"answerable": 2928,
|
||||
"corpus_manifest_sha256": "8fc1ed861c891d778853d41107630869f9eb267a669c5d1038e389de80114347",
|
||||
"corpus_sources": 50278,
|
||||
"generated_at": "2026-10-07T14:38:12+00:00",
|
||||
"generated_at": "2026-10-07T15:33:42+00:00",
|
||||
"git_commit": "29163632af7418758e79a79a748f68050ca1e0b6",
|
||||
"limit_per_modality": null,
|
||||
"machine": {
|
||||
@@ -602,12 +602,12 @@
|
||||
{
|
||||
"atoms": 500,
|
||||
"bm25_only": {
|
||||
"p50": 0.0007463944966730196,
|
||||
"p95": 0.0026389723490865425
|
||||
"p50": 0.0006466769991675392,
|
||||
"p95": 0.00181096804953995
|
||||
},
|
||||
"dense_only": {
|
||||
"p50": 2.053756675497425,
|
||||
"p95": 3.7252893005985244
|
||||
"p50": 1.3797472870028287,
|
||||
"p95": 1.5762464397041183
|
||||
},
|
||||
"note": "latency only: a source prefix drops most gold sources, so recall at these sizes is meaningless",
|
||||
"queries": 40,
|
||||
@@ -616,12 +616,12 @@
|
||||
{
|
||||
"atoms": 2000,
|
||||
"bm25_only": {
|
||||
"p50": 0.00923213200076134,
|
||||
"p95": 0.023986968196186332
|
||||
"p50": 0.008334371003002161,
|
||||
"p95": 0.015740483100307734
|
||||
},
|
||||
"dense_only": {
|
||||
"p50": 2.2580291965023207,
|
||||
"p95": 4.0939969898456186
|
||||
"p50": 1.3979396294998878,
|
||||
"p95": 2.595364080850413
|
||||
},
|
||||
"note": "latency only: a source prefix drops most gold sources, so recall at these sizes is meaningless",
|
||||
"queries": 40,
|
||||
@@ -630,12 +630,12 @@
|
||||
{
|
||||
"atoms": 10000,
|
||||
"bm25_only": {
|
||||
"p50": 0.11317921599766123,
|
||||
"p95": 0.2961666731029253
|
||||
"p50": 0.05245087000002968,
|
||||
"p95": 0.12780161215196129
|
||||
},
|
||||
"dense_only": {
|
||||
"p50": 2.269751224499487,
|
||||
"p95": 3.6929099251963633
|
||||
"p50": 1.3382894579990534,
|
||||
"p95": 2.2659613330957646
|
||||
},
|
||||
"note": "latency only: a source prefix drops most gold sources, so recall at these sizes is meaningless",
|
||||
"queries": 40,
|
||||
@@ -644,12 +644,12 @@
|
||||
{
|
||||
"atoms": 165695,
|
||||
"bm25_only": {
|
||||
"p50": 1.3151575949996186,
|
||||
"p95": 3.7234880074014645
|
||||
"p50": 0.8085095155001909,
|
||||
"p95": 2.273631770645442
|
||||
},
|
||||
"dense_only": {
|
||||
"p50": 2.3852021460043034,
|
||||
"p95": 4.008884304698221
|
||||
"p50": 1.4964509784986149,
|
||||
"p95": 2.4635968941045574
|
||||
},
|
||||
"note": "latency only: a source prefix drops most gold sources, so recall at these sizes is meaningless",
|
||||
"queries": 40,
|
||||
|
||||
+11
-2
@@ -12,7 +12,6 @@ from __future__ import annotations
|
||||
import argparse
|
||||
import hashlib
|
||||
import json
|
||||
import os
|
||||
import platform
|
||||
import subprocess
|
||||
from collections.abc import Sequence
|
||||
@@ -36,6 +35,7 @@ from bench.systems import (
|
||||
arm_needs,
|
||||
live_product_rankings,
|
||||
load_atoms,
|
||||
machine_facts,
|
||||
)
|
||||
|
||||
ROOT = SETS.parent
|
||||
@@ -213,7 +213,7 @@ def meta(corpus_sources: int, queries: Sequence[Query], limit: int | None = None
|
||||
"generated_at": datetime.now(UTC).isoformat(timespec="seconds"),
|
||||
"git_commit": git_commit(),
|
||||
"python": platform.python_version(),
|
||||
"machine": {"system": platform.system(), "release": platform.release(), "cpus": os.cpu_count()},
|
||||
"machine": machine_facts(),
|
||||
"corpus_manifest_sha256": manifest.get("corpus_manifest_sha256"),
|
||||
"sets": {name: sha256_file(SETS / f"{name}.jsonl") for name in MODALITIES},
|
||||
"queries": len(queries),
|
||||
@@ -251,6 +251,10 @@ def write_markdown(report: dict[str, Any], path: Path) -> None:
|
||||
f"({report['meta']['answerable']} answerable)"
|
||||
),
|
||||
f"- models: dense `{report['meta']['models']['dense']}`, rerank `{report['meta']['models']['rerank']}`",
|
||||
(
|
||||
f"- machine: {report['meta']['machine']['cpus']} cores, load average "
|
||||
f"{report['meta']['machine']['load_average_1_5_15'][0]:.1f} when the report was written"
|
||||
),
|
||||
"",
|
||||
"| arm | recall@1 | recall@5 | recall@10 | nDCG@10 | MRR | hit@10 (95% lower) | p50 s | p95 s |",
|
||||
"|---|---|---|---|---|---|---|---|---|",
|
||||
@@ -321,6 +325,11 @@ def write_markdown(report: dict[str, Any], path: Path) -> None:
|
||||
note = report["scale"][0].get("note")
|
||||
if note:
|
||||
lines += ["", f"_{note}_"]
|
||||
load = report["meta"]["machine"]["load_average_1_5_15"][0]
|
||||
lines.append(
|
||||
f"\n_latency was measured at load average {load:.1f} on {report['meta']['machine']['cpus']} cores; "
|
||||
"a busy machine inflates these figures, so read them as a ceiling_"
|
||||
)
|
||||
lines += [
|
||||
"",
|
||||
"## reading",
|
||||
|
||||
@@ -27,6 +27,7 @@ from bench.ingest_baseline import RUNS, database_path
|
||||
from bench.metrics import aggregate, score_query
|
||||
from bench.run import Query, git_commit, holder_index, load_queries
|
||||
from bench.sets_common import ROOT
|
||||
from bench.systems import machine_facts
|
||||
|
||||
REPORTS = ROOT / "reports"
|
||||
TOP_K = 10
|
||||
@@ -66,6 +67,10 @@ def write_markdown(report: dict[str, Any], path: Path) -> None:
|
||||
"",
|
||||
f"- generated {meta['generated_at']} at commit {meta['git_commit'][:12]}",
|
||||
f"- {meta['queries']} queries ({meta['answerable']} answerable) against {meta['sources']} sources",
|
||||
(
|
||||
f"- machine: {meta['machine']['cpus']} cores, load average "
|
||||
f"{meta['machine']['load_average_1_5_15'][0]:.1f} when the report was written"
|
||||
),
|
||||
(
|
||||
"- extraction is lexical for both stores; Mem0's real add() and Zep's graph builder call a "
|
||||
"generator model, which this round does not have"
|
||||
@@ -139,6 +144,7 @@ def main() -> int:
|
||||
"sources": len(index),
|
||||
"top_k": TOP_K,
|
||||
"extraction": "lexical (see bench/counterpart.py); LLM extraction deferred with the generator",
|
||||
"machine": machine_facts(),
|
||||
},
|
||||
"stores": [score_store(store, queries, holder_of) for store in stores],
|
||||
}
|
||||
|
||||
@@ -10,6 +10,8 @@ from __future__ import annotations
|
||||
|
||||
import hashlib
|
||||
import json
|
||||
import os
|
||||
import platform
|
||||
import re
|
||||
import sqlite3
|
||||
import time
|
||||
@@ -30,6 +32,16 @@ DENSE_MODEL = "sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2"
|
||||
RERANK_MODEL = "BAAI/bge-reranker-base"
|
||||
|
||||
|
||||
def machine_facts() -> dict[str, Any]:
|
||||
"""Hardware plus the load at report time: latency numbers mean nothing without it."""
|
||||
return {
|
||||
"system": platform.system(),
|
||||
"release": platform.release(),
|
||||
"cpus": os.cpu_count(),
|
||||
"load_average_1_5_15": [round(value, 2) for value in os.getloadavg()],
|
||||
}
|
||||
|
||||
|
||||
def local_model_dir(models: Path, repo_id: str) -> Path:
|
||||
"""Where a fetched model lives: one flat directory per repo, named by its repo name only."""
|
||||
return models / repo_id.split("/")[-1]
|
||||
|
||||
@@ -158,6 +158,7 @@ def _report() -> dict:
|
||||
"corpus_sources": 10,
|
||||
"queries": 16,
|
||||
"answerable": 8,
|
||||
"machine": {"system": "Linux", "release": "x", "cpus": 32, "load_average_1_5_15": [1.0, 2.0, 3.0]},
|
||||
"models": {"dense": "d", "rerank": "r"},
|
||||
},
|
||||
"chance": {"recall_at_1": 0.0, "recall_at_5": 0.0, "recall_at_10": 0.0},
|
||||
@@ -191,6 +192,7 @@ def test_the_report_writer_survives_a_scale_note(tmp_path: Path) -> None:
|
||||
text = path.read_text(encoding="utf-8")
|
||||
assert "| 500 | 900 | 0.001/0.002 |" in text
|
||||
assert "_latency only_" in text
|
||||
assert "load average 1.0 on 32 cores" in text
|
||||
assert "## unanswerable queries (50.0% of the suite)" in text
|
||||
|
||||
|
||||
|
||||
Reference in new issue
Block a user