Use existing loading state for sparse detail

This commit is contained in:
2569718930@qq.com
2026-04-18 14:40:29 +08:00
parent 16033033a7
commit fe6477b09c
18 changed files with 308 additions and 90 deletions
+14 -1
View File
@@ -1,5 +1,7 @@
# 概率训练样本归档说明(中文)
最后更新:`2026-04-18`
## 1. 目的
这份文档说明两件事:
@@ -27,6 +29,12 @@
- `max_so_far`
- `peak_status`
- `prob_snapshot`
- `probability_engine`
- `calibration_mode`
- `calibration_version`
- `raw_mu / raw_sigma`
- `calibrated_mu / calibrated_sigma`
- `shadow_distribution`
- 当天最终 `actual_high`
- 当天最终 `settlement bucket`
@@ -167,7 +175,10 @@
],
"probability_engine": "legacy",
"probability_mode": "emos_shadow",
"calibration_version": "emos-20260320130245"
"calibration_mode": "emos_shadow",
"calibration_version": "emos-20260320130245",
"calibrated_mu": 15.4,
"calibrated_sigma": 1.1
}
```
@@ -176,6 +187,8 @@
- `actual_high`
- `settlement_bucket`
当前前端把这类快照解释为“校准模型概率”。如果 `probability_engine` 为 LGBM 相关值,则显示为 LGBM 校准概率;模型舍入票数和市场价格只用于解释,不直接作为最终概率。
## 7. 现阶段你可以执行的命令
### 7.1 回填历史天气 CSV