Record wider-window FMA smoke progress evidence
Capture a more meaningful follow-up checkpoint after an added wait window so the restart docs show continued forward motion rather than trivial second-to-second sampling. Constraint: Epoch 1 still has not completed, so verification is limited to runtime evidence and manifest revalidation Rejected: Skip the wider-window checkpoint | would miss the chance to prove progress across a longer observation gap Confidence: high Scope-risk: narrow Directive: Keep watching for the first saved model file or transition into build-index/evaluate before changing the project status summary Tested: ps on PID 311629 after 30s wait; validate-splits on /tmp/fma_real_smoke_stopcheck/fma/manifests; find on /tmp/fma_real_smoke_stopcheck/fma_models_smoke Not-tested: End-of-epoch artifacts, build-index, evaluate, final metrics
Showing
3 changed files
with
46 additions
and
0 deletions
| 1 | ## 2026-06-02 真实 FMA smoke fresh evidence 24:11 checkpoint | ||
| 2 | |||
| 3 | 完成项: | ||
| 4 | - 在额外约 30 秒窗口后再次检查真实 FMA smoke,确认 `train.py` elapsed 已推进到 24:11。 | ||
| 5 | - 更新 `docs/session-handoff.md` 与 `docs/changelist-2026-06-02.md`,同步更有间隔意义的 live evidence。 | ||
| 6 | |||
| 7 | 验证结果: | ||
| 8 | - `ps -p 311629 -o pid,etime,%cpu,%mem,cmd` => `ELAPSED=24:11` | ||
| 9 | - 仍未出现 `build-index/evaluate` 相关新进程 | ||
| 10 | - `validate-splits /tmp/fma_real_smoke_stopcheck/fma/manifests` => `ok=true` | ||
| 11 | - `fma_models_smoke/` 仍仅有目录本身 | ||
| 12 | |||
| 13 | 结论: | ||
| 14 | - 真实 FMA 全量 smoke 在更长观察窗口下仍持续推进,没有中断迹象。 | ||
| 15 | - 到该时点仍未产生首个模型文件或下游阶段切换证据。 | ||
| 16 | |||
| 1 | ## 2026-06-02 真实 FMA smoke fresh evidence 22:58 checkpoint | 17 | ## 2026-06-02 真实 FMA smoke fresh evidence 22:58 checkpoint |
| 2 | 18 | ||
| 3 | 完成项: | 19 | 完成项: | ... | ... |
| ... | @@ -200,3 +200,11 @@ cd /workspace/acr-engine | ... | @@ -200,3 +200,11 @@ cd /workspace/acr-engine |
| 200 | - 当前 CPU / 内存观测:`%CPU≈615`, `%MEM≈10.8`。 | 200 | - 当前 CPU / 内存观测:`%CPU≈615`, `%MEM≈10.8`。 |
| 201 | - 仍未出现模型文件,也未切换到 `build-index/evaluate`。 | 201 | - 仍未出现模型文件,也未切换到 `build-index/evaluate`。 |
| 202 | - manifest 复核继续通过,统计保持不变。 | 202 | - manifest 复核继续通过,统计保持不变。 |
| 203 | |||
| 204 | |||
| 205 | ## 12:21 UTC(30 秒窗口)时间推进补充 | ||
| 206 | |||
| 207 | - 最新 live 证据已推进到:`train.py ELAPSED=24:11`。 | ||
| 208 | - 当前 CPU / 内存观测:`%CPU≈615`, `%MEM≈11.3`。 | ||
| 209 | - 30 秒额外等待后,仍未出现模型文件,也未切换到 `build-index/evaluate`。 | ||
| 210 | - manifest 复核继续通过,统计保持不变。 | ... | ... |
| ... | @@ -255,6 +255,28 @@ | ... | @@ -255,6 +255,28 @@ |
| 255 | - 当前依旧处于第 1 个 epoch 内部的持续训练阶段。 | 255 | - 当前依旧处于第 1 个 epoch 内部的持续训练阶段。 |
| 256 | - 到 12:20 UTC 为止,仍没有首个模型文件或下游检索/评测阶段证据。 | 256 | - 到 12:20 UTC 为止,仍没有首个模型文件或下游检索/评测阶段证据。 |
| 257 | 257 | ||
| 258 | ### 30 秒窗口后的 fresh evidence(2026-06-02 12:21 UTC) | ||
| 259 | |||
| 260 | - 经过额外约 30 秒等待后,真实 FMA smoke 继续推进到: | ||
| 261 | - `train.py ELAPSED=24:11` | ||
| 262 | - `%CPU≈615` | ||
| 263 | - `%MEM≈11.3` | ||
| 264 | - 当前进程结构仍未发生阶段切换: | ||
| 265 | - `PID=311494`:`external_adapters.py smoke-local fma ...` | ||
| 266 | - `PID=311629`:`train.py --data /tmp/fma_real_smoke_stopcheck/fma/manifests ...` | ||
| 267 | - 仍未出现 `build-index` / `evaluate` 相关新进程。 | ||
| 268 | - `fma_models_smoke/` 仍只有目录本身,没有模型文件。 | ||
| 269 | - manifest 再次复核仍通过: | ||
| 270 | - `ok=true` | ||
| 271 | - `catalog_references=8000` | ||
| 272 | - `train_queries=6401` | ||
| 273 | - `test_queries=1593` | ||
| 274 | - `val_queries=0` | ||
| 275 | |||
| 276 | 这说明: | ||
| 277 | - 即使拉开更有意义的时间窗口,当前仍能确认训练在前进,而不是僵死。 | ||
| 278 | - 到 12:21 UTC 为止,仍没有首个模型文件或下游检索/评测阶段证据。 | ||
| 279 | |||
| 258 | ### 重启后第一优先级动作 | 280 | ### 重启后第一优先级动作 |
| 259 | 281 | ||
| 260 | 1. 先检查真实 FMA smoke 是否完成: | 282 | 1. 先检查真实 FMA smoke 是否完成: | ... | ... |
-
Please register or sign in to post a comment