物理修复(400 失败点归因,见 docs/failed400_nan_pseudo_convergence_2026_08_17.md): - runner: 阶段 converged 后复查 fort.7,含 NaN/Inf 即否决(fort.9 全零伪收敛, 曾致 261 点误跳过 nl_direct 回退);否决阶段不产出种子,阻断污染传播 - runner: nl_tight 回退——仅能量边际失败时以 CHMAX 收紧 10× 从自身模型续迭代, 残差降幅 ~10×;execute_tlusty_stage 抽取供主链与回退共用 - conv_check: fort.14 为 Eddington 通量 Hλ,积分需乘 4π 再比 σTeff⁴ (旧版 ratio 稳定 0.0796=1/4π,全点系统性假阳性)+ 回归测试 - nst_writer: 单行超 80 字符被 TLUSTY 静默截断,IFALI/JALI/TRAD 等从未生效; 按 75 字符自动换行 - seed_finder: Teff 容忍度改含边界 <=,相邻 5000K 档恢复互为种子 + 回归测试 谱线表与网格: - 默认线表 gfVIS99 → gfATO(全波段 18-23000Å),TaskSpec.linelist 支持工作流 级覆盖,节点按需下载(进程互斥锁防并发重复下载 238MB) - sdB_cno Teff 加密至 5000K 步长,432 → 9216 点;tlusty/synspec 静态二进制更新 统计与部署: - grid 汇总改按 tlusty_status/synspec_status 分项计数,新增 tlusty_failed/ synspec_failed/synspec_pending,前端详情页双视图适配 - deploy/fetch_results 支持跳板机 ProxyJump 与 SSH 主连接复用,fetch 新增 --force; - Docker 构建支持 CARGO_MIRROR/USE_MIRRORS 国内镜像参数;移除 tools/ 拷贝 - 新增 tlusty-synspec-test skill 与 6 篇根因分析/验证文档
122 lines
4.3 KiB
Bash
Executable File
122 lines
4.3 KiB
Bash
Executable File
#!/usr/bin/env bash
|
||
# ============================================================
|
||
# new_test.sh — 脚手架:在 dcts/test/ 下创建规范的四级测试目录
|
||
#
|
||
# 用法:
|
||
# bash new_test.sh <test_id> <type> <grid> <content> [--chain cold|seed|synspec]
|
||
#
|
||
# 示例:
|
||
# bash new_test.sh 20260813_my_div conv t60000_g5.0_he-2_cold baseline --chain cold
|
||
# bash new_test.sh 20260813_seed conv t55000_g5.0_he-2_seed doubleoff --chain seed
|
||
# bash new_test.sh 20260813_spec synspec t50000_g5.0_cno gfato_100_20000 --chain synspec
|
||
#
|
||
# 创建:
|
||
# dcts/test/<test_id>/<type>/<grid>/<content>/{inputs,scripts,run,outputs}
|
||
# scripts/run.sh <- 从 assets/run_{chain}_chain.sh 拷贝
|
||
# README.md <- 骨架
|
||
# ============================================================
|
||
set -euo pipefail
|
||
|
||
# --- 自定位 skill 目录(脚本在 scripts/ 下,skill 根是父目录)---
|
||
SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
|
||
SKILL_DIR="$(cd "$SCRIPT_DIR/.." && pwd)"
|
||
ASSETS="$SKILL_DIR/assets"
|
||
|
||
# --- 找 dcts 根(含 assets/tlusty_static 的目录)---
|
||
# 优先用环境变量;否则从 skill 目录向上找第一个含 assets/tlusty_static 的祖先
|
||
DCTS_ROOT="${DCTS_ROOT:-}"
|
||
if [ -z "$DCTS_ROOT" ]; then
|
||
cur="$SCRIPT_DIR"
|
||
while [ "$cur" != "/" ]; do
|
||
if [ -f "$cur/assets/tlusty_static" ]; then DCTS_ROOT="$cur"; break; fi
|
||
cur="$(dirname "$cur")"
|
||
done
|
||
fi
|
||
TEST_ROOT="$DCTS_ROOT/test"
|
||
if [ -z "$DCTS_ROOT" ] || [ ! -f "$DCTS_ROOT/assets/tlusty_static" ]; then
|
||
echo "❌ 未找到 dcts 根(含 assets/tlusty_static 的目录)。用 DCTS_ROOT=... 环境变量指定。" >&2
|
||
exit 1
|
||
fi
|
||
|
||
# --- 解析参数 ---
|
||
if [ $# -lt 4 ]; then
|
||
sed -n '3,20p' "$0" | sed 's/^# \{0,1\}//' >&2
|
||
exit 1
|
||
fi
|
||
TEST_ID="$1"; TYPE="$2"; GRID="$3"; CONTENT="$4"; shift 4
|
||
CHAIN="cold"
|
||
while [ $# -gt 0 ]; do
|
||
case "$1" in
|
||
--chain) CHAIN="$2"; shift 2;;
|
||
*) echo "未知参数: $1" >&2; exit 1;;
|
||
esac
|
||
done
|
||
|
||
case "$CHAIN" in
|
||
cold) RUN_TEMPLATE="$ASSETS/run_cold_chain.sh"; STAGES="lte → nc → nl";;
|
||
seed) RUN_TEMPLATE="$ASSETS/run_seed_chain.sh"; STAGES="seed_nc → nl";;
|
||
synspec) RUN_TEMPLATE="$ASSETS/run_synspec.sh"; STAGES="SYNSPEC";;
|
||
*) echo "❌ --chain 必须是 cold|seed|synspec,收到: $CHAIN" >&2; exit 1;;
|
||
esac
|
||
|
||
# --- 创建四级目录 ---
|
||
TARGET="$TEST_ROOT/$TEST_ID/$TYPE/$GRID/$CONTENT"
|
||
if [ -e "$TARGET" ]; then
|
||
echo "❌ 目录已存在: $TARGET" >&2
|
||
exit 1
|
||
fi
|
||
mkdir -p "$TARGET/inputs" "$TARGET/scripts" "$TARGET/run" "$TARGET/outputs"
|
||
|
||
# --- 拷 run.sh 模板 ---
|
||
cp "$RUN_TEMPLATE" "$TARGET/scripts/run.sh"
|
||
chmod +x "$TARGET/scripts/run.sh"
|
||
|
||
# --- 建 data 软链到 run/(TLUSTY/SYNSPEC 都要)---
|
||
ln -sfn "$DCTS_ROOT/assets/data" "$TARGET/run/data"
|
||
|
||
# --- README 骨架 ---
|
||
cat > "$TARGET/README.md" <<EOF
|
||
# $TEST_ID / $TYPE / $GRID / $CONTENT
|
||
|
||
## 目的
|
||
TODO: 一句话说清这次测试想验证什么。
|
||
|
||
## 配置
|
||
- 二进制: $DCTS_ROOT/assets/$([ "$CHAIN" = synspec ] && echo synspec || echo tlusty)_static
|
||
- 网格: $GRID
|
||
- 链: $STAGES
|
||
- 变体: TODO(与 baseline 不同的关键参数,如 ITEK/IACC/DPSILG/ORELAX/NITER)
|
||
|
||
## 输入
|
||
TODO: inputs/ 生成方式(代码生成 / 模板修改,留 diff);关键参数。
|
||
- [ ] 见 references/directory-convention.md 的执行前检查清单
|
||
|
||
## 执行
|
||
- 命令: cd "$TARGET" && bash scripts/run.sh
|
||
- 各阶段: TODO
|
||
|
||
## 结果
|
||
TODO: 见 outputs/RESULT.md(用 assets/result_template.md)
|
||
|
||
## 审查结论
|
||
- 数值: TODO(max_relc 是否达标;STOP/雪崩?)
|
||
- 物理: TODO(温度结构三项:表层<3×Teff、各深度∈[10,1e8]、无 NaN)
|
||
- 结论: TODO
|
||
EOF
|
||
|
||
echo "✅ 已创建测试目录: $TARGET"
|
||
echo " 链类型: $CHAIN ($STAGES)"
|
||
echo ""
|
||
echo "下一步:"
|
||
echo " 1. 把输入文件放进 $TARGET/inputs/"
|
||
echo " 2. 执行前校验(TLUSTY 链):"
|
||
echo " python3 $SKILL_DIR/scripts/check_inputs.py $TARGET/inputs --chain $CHAIN --teff <Teff>"
|
||
echo " 3. 跑: cd $TARGET && bash scripts/run.sh"
|
||
echo " 4. 事后审查:"
|
||
if [ "$CHAIN" != "synspec" ]; then
|
||
echo " python3 $SKILL_DIR/scripts/check_fort9.py $TARGET/outputs/nl.9 --teff <Teff>"
|
||
echo " python3 $SKILL_DIR/scripts/check_temperature.py $TARGET/outputs/nl.7 --teff <Teff>"
|
||
else
|
||
echo " grep -ciE 'nan|inf|\\*{3,}' $TARGET/outputs/*.spec # 应为 0"
|
||
fi
|