Commit f228197d f228197d7051d5e1ab867aa006225a0ec9a34693 by cnb.bofCdSsphPA

Record the first cap32 hybrid score while the larger run continues

Persist the newly finished cap32 hybrid result so the next session can continue the top-two validation run from measured evidence instead of only a running-state checkpoint.

Constraint: cap32 high_energy and the final report are still pending
Rejected: Wait for the full cap32 report before updating docs | Would leave the larger-subset evidence stale across sessions
Confidence: high
Scope-risk: narrow
Directive: Replace the cap32 partial section with the final two-strategy ranking once high_energy eval and report.json land
Tested: Verified /tmp/ab_smoke_seg_cap32_top2/hybrid/fma_reports_smoke/eval.json; verified docs record hybrid=20/0.95/1.0 and high_energy still training
Not-tested: Final cap32 comparison because high_energy has not finished yet
1 parent 5dadbae3
...@@ -14,14 +14,16 @@ ...@@ -14,14 +14,16 @@
14 14
15 当前 fresh evidence: 15 当前 fresh evidence:
16 - `scripts/ab_smoke_segmentation.py ... --work-root /tmp/ab_smoke_seg_cap32_top2` 已启动 16 - `scripts/ab_smoke_segmentation.py ... --work-root /tmp/ab_smoke_seg_cap32_top2` 已启动
17 - 当前 first lane 为: 17 - `hybrid` 已完成首条评测:
18 - `hybrid` 18 - `num_queries = 20`
19 - 当前已进入: 19 - `top1 = 0.95`
20 - `run_demo.py build-index --resume --checkpoint-every-refs 100` 20 - `topk = 1.0`
21 - `high_energy` 已进入训练阶段
21 - `report.json` 尚未落盘 22 - `report.json` 尚未落盘
22 23
23 结论: 24 结论:
24 - 现在已经开始验证 cap24 结论在更大 `subset=32` 上是否继续成立 25 - 现在已经开始验证 cap24 结论在更大 `subset=32` 上是否继续成立
26 - 当前至少可以确认:`hybrid` 在更大子集上仍保持较强表现
25 - 即使当前 session 结束,新 session 也可直接从 handoff 中的 cap32 入口继续盯结果 27 - 即使当前 session 结束,新 session 也可直接从 handoff 中的 cap32 入口继续盯结果
26 28
27 ### Stage: 收尾 cap24 top2 真实 FMA 对照并确认默认策略 29 ### Stage: 收尾 cap24 top2 真实 FMA 对照并确认默认策略
......
...@@ -442,8 +442,8 @@ cd /workspace/acr-engine ...@@ -442,8 +442,8 @@ cd /workspace/acr-engine
442 |---|---| 442 |---|---|
443 | `subset_size` | `32` | 443 | `subset_size` | `32` |
444 | `max_test_queries` | `20` | 444 | `max_test_queries` | `20` |
445 | 首个运行策略 | `hybrid` | 445 | `hybrid` | `num_queries=20`, `top1=0.95`, `topk=1.0` |
446 | 当前阶段 | `run_demo.py build-index --resume --checkpoint-every-refs 100` | 446 | `high_energy` | 训练中 |
447 | `report.json` | 尚未生成 | 447 | `report.json` | 尚未生成 |
448 448
449 恢复检查命令: 449 恢复检查命令:
...@@ -453,7 +453,7 @@ pgrep -af 'ab_smoke_seg_cap32_top2|external_adapters.py smoke-local fma /tmp/ab_ ...@@ -453,7 +453,7 @@ pgrep -af 'ab_smoke_seg_cap32_top2|external_adapters.py smoke-local fma /tmp/ab_
453 ``` 453 ```
454 454
455 优先等待文件: 455 优先等待文件:
456 - `/tmp/ab_smoke_seg_cap32_top2/hybrid/fma_reports_smoke/eval.json` 456 - `/tmp/ab_smoke_seg_cap32_top2/high_energy/fma_reports_smoke/eval.json`
457 - `/tmp/ab_smoke_seg_cap32_top2/report.json` 457 - `/tmp/ab_smoke_seg_cap32_top2/report.json`
458 - `b766c74` Make open-dataset manifests trainable end to end 458 - `b766c74` Make open-dataset manifests trainable end to end
459 - `fa23144` Add a single-page open dataset workflow for training prep 459 - `fa23144` Add a single-page open dataset workflow for training prep
......