Visual Chronometer

已接入

Visual Chronometer — 指标、依赖与运行命令。

最后更新: 2026年7月22日

本页内容

简介

Visual Chronometer external benchmark task manifest for in-tree PhyFPS prediction.

WorldFoundry 评测路径: 使用仓内 official metric runtime，对 caller 提供的生成产物运行官方指标栈（本地 checkpoint / evaluator）。候选模型输出从 generated-artifact 目录读取。

集成状态

项目	当前状态
Catalog integration	`integrated`
Runner verification	`official_runtime_ready`
Leaderboard	暂不可用于 leaderboard 声明

指标

Task YAML

worldfoundry/data/benchmarks/tasks/external/visual-chronometer.yaml

Primary

visual_chronometer_average

Metric	关注点
`mean_phyfps`	Mean average PhyFPS across evaluated videos.
`inter_video_cv`	Cross-video coefficient of variation over average PhyFPS.
`intra_video_cv`	Mean within-video coefficient of variation over sliding-window clips.
`visual_chronometer_average`	Visual Chronometer 时间感知 probe 的时序一致性 score 族平均值（越低越好）。

环境准备

运行环境

Conda 环境： worldfoundry-unified-cu128。
单独配置： 默认复用 WorldFoundry 统一环境（needs_new_env: false），通常不需要额外 benchmark-only conda env。
WorldFoundry runs Visual Chronometer PhyFPS prediction in-process and normalizes results.csv into the Visual Chronometer metric surface.
Use WORLDFOUNDRY_VISUAL_CHRONOMETER_PREDICT_BACKEND to select the local prediction backend.

测评数据

待测模型输出目录：设置 WORLDFOUNDRY_GENERATED_ARTIFACT_DIR 指向生成视频或 rollout artifact 根目录。
worldfoundry/evaluation/tasks/execution/runners/phyfps_bench_gen/runtime/visual_chronometer
worldfoundry/data/benchmarks/tasks/external/visual-chronometer.yaml

Checkpoint 与资产

Base-model / metric 依赖：visual_chronometer_fps_predictor。
Metric / evaluator checkpoint 默认路径与覆盖变量见 local assets 指南。

关键环境变量

WORLDFOUNDRY_GENERATED_ARTIFACT_DIR
WORLDFOUNDRY_VISUAL_CHRONOMETER_ROOT
WORLDFOUNDRY_VISUAL_CHRONOMETER_PREDICT_BACKEND
WORLDFOUNDRY_VISUAL_CHRONOMETER_RESULTS_PATH
WORLDFOUNDRY_VISUAL_CHRONOMETER_STRICT

环境检查

PYTHONPATH=. ${WORLDFOUNDRY_UNIFIED_PYTHON:-python} -m worldfoundry.evaluation.tasks.execution.orchestration.runtime_preflight --profile visual-chronometer --manifest worldfoundry/data/benchmarks/runtime_profiles/official/visual-chronometer.yaml --output-dir tmp/worldfoundry_preflight/visual-chronometer --json

运行评测

准备好该 benchmark 页面列出的资产和候选产物后，使用下面的公开运行入口。

设置候选产物目录

export WORLDFOUNDRY_GENERATED_ARTIFACT_DIR=/path/to/generated/artifacts

导入已有官方结果

worldfoundry-eval zoo benchmark-run \
  --benchmark-id visual-chronometer \
  --mode normalizer \
  --official-results-path /path/to/official/results-or-report \
  --generated-artifact-dir "${WORLDFOUNDRY_GENERATED_ARTIFACT_DIR}" \
  --output-dir tmp/visual-chronometer/normalizer \
  --json

从生成产物重新打分

worldfoundry-eval zoo benchmark-run \
  --benchmark-id visual-chronometer \
  --mode official-run \
  --generated-artifact-dir "${WORLDFOUNDRY_GENERATED_ARTIFACT_DIR}" \
  --output-dir tmp/visual-chronometer/official-run \
  --json

直接调用仓内 runner

PYTHONPATH=. "${WORLDFOUNDRY_UNIFIED_PYTHON:-python}" worldfoundry/evaluation/tasks/execution/runners/phyfps_bench_gen/run_visual_chronometer_official_runner.py --run-official --generated-artifact-dir "${WORLDFOUNDRY_GENERATED_ARTIFACT_DIR}" --output-dir "${WORLDFOUNDRY_BENCHMARK_OUTPUT_DIR}" --json