保全最后历史周报台账与非训练暂存片段,核验全部原字节和限制说明
Browse files- README.md +3 -0
- closeout-final/weekly/MANIFEST.jsonl +14 -0
- closeout-final/weekly/README.md +7 -0
- closeout-final/weekly/dm-weekly-2026-08-29.md +68 -0
- closeout-final/weekly/dm-weekly-2026-09-05.md +61 -0
- closeout-final/weekly/dm-weekly-2026-09-12.md +83 -0
- closeout-final/weekly/dm-weekly-2026-09-19.md +79 -0
- closeout-final/weekly/dm-weekly-2026-09-26.md +81 -0
- closeout-final/weekly/dm-weekly-2026-10-03.md +77 -0
- closeout-final/weekly/dm-weekly-2026-10-10.md +76 -0
- closeout-final/weekly/dm-weekly-2026-10-17.md +81 -0
- closeout-final/weekly/dm-weekly-2026-10-24.md +80 -0
- closeout-final/weekly/dm-weekly-2026-10-31.md +72 -0
- closeout-final/weekly/dm-weekly-2026-11-07.md +71 -0
- closeout-final/weekly/dm-weekly-2026-11-14.md +71 -0
- closeout-final/weekly/ledger.json.gz +3 -0
- closeout-final/weekly/state.json +49 -0
README.md
CHANGED
|
@@ -36,3 +36,6 @@ One original 179-line `issue-559-ets2-20260806/calibration/telemetry-smoke.jsonl
|
|
| 36 |
TAR shards retain complete documents; no file is truncated or split. Each tar ends with per-member SHA-256 and size. Source bytes and staged public-copy bytes are hashed before transfer, then pinned Hugging Face LFS hashes/sizes and anonymous downloads are checked before original storage deletion. `ARCHIVE_INDEX.json` pins all 50 completed shards.
|
| 37 |
|
| 38 |
Old scripts, endpoints, operational instructions and acceptance gates in these documents are historical context for a closed project. They should not be interpreted as instructions to recreate paid resources or as evidence that those services still exist. Capture implementation: [Diogenes-Archive](https://github.com/CuddlerAI/Diogenes-Archive).
|
|
|
|
|
|
|
|
|
|
|
|
| 36 |
TAR shards retain complete documents; no file is truncated or split. Each tar ends with per-member SHA-256 and size. Source bytes and staged public-copy bytes are hashed before transfer, then pinned Hugging Face LFS hashes/sizes and anonymous downloads are checked before original storage deletion. `ARCHIVE_INDEX.json` pins all 50 completed shards.
|
| 37 |
|
| 38 |
Old scripts, endpoints, operational instructions and acceptance gates in these documents are historical context for a closed project. They should not be interpreted as instructions to recreate paid resources or as evidence that those services still exist. Capture implementation: [Diogenes-Archive](https://github.com/CuddlerAI/Diogenes-Archive).
|
| 39 |
+
|
| 40 |
+
|
| 41 |
+
The [historical weekly supplement](closeout-final/weekly/README.md) preserves the final reviewed source objects with their original limitations and exact member hashes.
|
closeout-final/weekly/MANIFEST.jsonl
ADDED
|
@@ -0,0 +1,14 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{"path": "closeout-final/weekly/dm-weekly-2026-08-29.md", "size": 3192, "sha256": "a699065c1f2061b56b573a21429a983d6df0747f0987a5cd79e6f7150a2f858e", "historical_source_relative_path": "reports/weekly/dm-weekly-2026-08-29.md"}
|
| 2 |
+
{"path": "closeout-final/weekly/dm-weekly-2026-09-05.md", "size": 2814, "sha256": "7b599a2ce29d3b8ff44b15624c6104bafeae9d9039993a045686e9d663db3f09", "historical_source_relative_path": "reports/weekly/dm-weekly-2026-09-05.md"}
|
| 3 |
+
{"path": "closeout-final/weekly/dm-weekly-2026-09-12.md", "size": 4930, "sha256": "67de489dfdf4d04144624e0f9f08ca067218057fb6087a25cab2f3046e459b34", "historical_source_relative_path": "reports/weekly/dm-weekly-2026-09-12.md"}
|
| 4 |
+
{"path": "closeout-final/weekly/dm-weekly-2026-09-19.md", "size": 4563, "sha256": "29209f8d618a22939190251b1c2b2c550dbaf8f329a27a346860036f0e4c668a", "historical_source_relative_path": "reports/weekly/dm-weekly-2026-09-19.md"}
|
| 5 |
+
{"path": "closeout-final/weekly/dm-weekly-2026-09-26.md", "size": 4667, "sha256": "68a60003b35bcf3a41fe2e845db6c24d436017a68a742b1ac65b677c1c637bf3", "historical_source_relative_path": "reports/weekly/dm-weekly-2026-09-26.md"}
|
| 6 |
+
{"path": "closeout-final/weekly/dm-weekly-2026-10-03.md", "size": 4566, "sha256": "f8633941099e586e89087c7e725f175ec738f37edb66298616cc0ecaa97a0cb3", "historical_source_relative_path": "reports/weekly/dm-weekly-2026-10-03.md"}
|
| 7 |
+
{"path": "closeout-final/weekly/dm-weekly-2026-10-10.md", "size": 4454, "sha256": "ae39eda552f22ed7b6a76b746eab0148dd961a85c7b26581297964fef9956275", "historical_source_relative_path": "reports/weekly/dm-weekly-2026-10-10.md"}
|
| 8 |
+
{"path": "closeout-final/weekly/dm-weekly-2026-10-17.md", "size": 4679, "sha256": "0d70394a7c7c0fbf906bee8d809085cf7a751f43416551226cc33b66c3c8482b", "historical_source_relative_path": "reports/weekly/dm-weekly-2026-10-17.md"}
|
| 9 |
+
{"path": "closeout-final/weekly/dm-weekly-2026-10-24.md", "size": 4751, "sha256": "4b632783018de9a609b966fe020efd92422a595418cd95b92c9ddc9bf0a6e4a2", "historical_source_relative_path": "reports/weekly/dm-weekly-2026-10-24.md"}
|
| 10 |
+
{"path": "closeout-final/weekly/dm-weekly-2026-10-31.md", "size": 4417, "sha256": "8b18c82e1e984b14def8aa2a9b5011cf806531956fe0f1903fefcc251252a832", "historical_source_relative_path": "reports/weekly/dm-weekly-2026-10-31.md"}
|
| 11 |
+
{"path": "closeout-final/weekly/dm-weekly-2026-11-07.md", "size": 4303, "sha256": "8a4676091d6040f0c6d0e0b53d8973430d8ea98386fdc19bec45082d34f808d0", "historical_source_relative_path": "reports/weekly/dm-weekly-2026-11-07.md"}
|
| 12 |
+
{"path": "closeout-final/weekly/dm-weekly-2026-11-14.md", "size": 4303, "sha256": "f93b43c9349f89c3ee770efaecbc7ec62ee091e5931676b2472ec058391c53e7", "historical_source_relative_path": "reports/weekly/dm-weekly-2026-11-14.md"}
|
| 13 |
+
{"path": "closeout-final/weekly/ledger.json.gz", "size": 6173867, "sha256": "703f25394b56f382f9b36d0f6a2b7f4f5337ad5d9eae1921990f547f09038135", "historical_source_relative_path": "reports/weekly/ledger.json.gz"}
|
| 14 |
+
{"path": "closeout-final/weekly/state.json", "size": 1412, "sha256": "1c5c5fd37e573d99726227497e818a3a0e1a5c977879101c440506638ae57e32", "historical_source_relative_path": "reports/weekly/state.json"}
|
closeout-final/weekly/README.md
ADDED
|
@@ -0,0 +1,7 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Historical capture stock and weekly reports
|
| 2 |
+
|
| 3 |
+
These fourteen byte-identical files preserve twelve historical weekly reports, the final compressed stock ledger and report state. The ledger has 219,293 session/clip entries; each value records a historical game label, duration and estimated recording timestamp. Its UUID pair is session ID / clip ID, as implemented by `tools/weekly-report/weekly_report.py:ledger_key`; it is not a named-person roster.
|
| 4 |
+
|
| 5 |
+
The reports retain historical corrections, missing components, frozen frames, source-lane blind spots, licensing/eligibility labels and changing counts. These are dated records, not a current rights grant, certification or assertion that a collection is trainable. Their stock counts and hours are not independent of the raw video archives and should not be added to archive byte/member counts. Old console URLs describe the retired reporting interface and are not live service promises. The original compressed ledger bytes are preserved, with its full gzip checksum and all JSON entries examined.
|
| 6 |
+
|
| 7 |
+
All fourteen original files and actual SHA-256 values are listed in `MANIFEST.jsonl`. No source number, label, timestamp or quality flag was rewritten. The parent archive's rights and research-use notice applies.
|
closeout-final/weekly/dm-weekly-2026-08-29.md
ADDED
|
@@ -0,0 +1,68 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# DataMiner 数据周报 · 2026-08-22 ~ 2026-08-29
|
| 2 |
+
|
| 3 |
+
> schema dm-weekly-report/1 · 窗口按对象上传时间(Asia/Shanghai)· 健康判定同 docs/146(console 口径)
|
| 4 |
+
|
| 5 |
+
## 一、本周采集与质检
|
| 6 |
+
|
| 7 |
+
- 总量:**76.87 h** / 438 会话 / 27908 clips
|
| 8 |
+
- 健康异常:**1113**(4.0%),其中补帧越线 1113
|
| 9 |
+
- 音频:缺失 1 · 静音怀疑 4
|
| 10 |
+
- session 验收报告覆盖(全量口径):2/1015
|
| 11 |
+
|
| 12 |
+
| 游戏 | 会话 | clips | 小时 | 异常 | 异常率 | 无音频 | 静音疑 |
|
| 13 |
+
|---|---|---|---|---|---|---|---|
|
| 14 |
+
| cyberpunk-2077-steam | 436 | 27906 | 76.86 | 1112 | 4.0% | 1 | 4 |
|
| 15 |
+
| fallout4-steam-release-test | 2 | 2 | 0.01 | 1 | 50.0% | 0 | 0 |
|
| 16 |
+
|
| 17 |
+
### 逐日
|
| 18 |
+
|
| 19 |
+
| 日期 | clips | 小时 | 异常 |
|
| 20 |
+
|---|---|---|---|
|
| 21 |
+
| 08-22 | 273 | 0.75 | 3 |
|
| 22 |
+
| 08-24 | 5807 | 15.99 | 225 |
|
| 23 |
+
| 08-25 | 6335 | 17.45 | 221 |
|
| 24 |
+
| 08-26 | 5362 | 14.77 | 127 |
|
| 25 |
+
| 08-27 | 5366 | 14.78 | 272 |
|
| 26 |
+
| 08-28 | 4765 | 13.13 | 265 |
|
| 27 |
+
|
| 28 |
+
### 质检 findings
|
| 29 |
+
|
| 30 |
+
- 六件套不完整 ×1: df3801fe(9e8845c5) 缺 ['frames', 'events', 'meta', 'manifest']
|
| 31 |
+
- 整段冻结 clip(dup≥100%)×1: 67f092c1(1a464098)
|
| 32 |
+
|
| 33 |
+
## 二、ML 各线可用数据(累计)
|
| 34 |
+
|
| 35 |
+
| 游戏 | 授权档 | 累计小时 | 累计 clips |
|
| 36 |
+
|---|---|---|---|
|
| 37 |
+
| cyberpunk-2077-steam | 🚫 三方待授权 | 140.4 | 51010 |
|
| 38 |
+
| halo-infinite-steam-release-test | ✅ 伙伴授权可训练(C-1) | 27.9 | 10122 |
|
| 39 |
+
| fallout4-steam-release-test | ⏸ 待 C-1 扩展(微软系) | 2.8 | 963 |
|
| 40 |
+
| wolfenstein2-steam | ⏸ 待 C-1 扩展(微软系) | 0.2 | 67 |
|
| 41 |
+
| atomic-heart-pc | 🚫 三方待授权 | 0.1 | 39 |
|
| 42 |
+
| resident-evil-village-pc | 🚫 三方待授权 | 0.1 | 35 |
|
| 43 |
+
| doom-eternal-pc | ⏸ 待 C-1 扩展(微软系) | 0.1 | 33 |
|
| 44 |
+
| red-dead-redemption-2-pc | 🚫 三方待授权 | 0.1 | 32 |
|
| 45 |
+
| deathloop-pc | ⏸ 待 C-1 扩展(微软系) | 0.1 | 33 |
|
| 46 |
+
| starfield-pc | ⏸ 待 C-1 扩展(微软系) | 0.1 | 29 |
|
| 47 |
+
| call-of-duty-steam-release-test | ⏸ 待 C-1 扩展(微软系) | 0.1 | 22 |
|
| 48 |
+
| wuchang-fallen-feathers-steam | 🚫 三方待授权 | 0.0 | 7 |
|
| 49 |
+
|
| 50 |
+
- 引擎深度交付(verified/):**4 包**
|
| 51 |
+
|
| 52 |
+
### 各线口径
|
| 53 |
+
|
| 54 |
+
- **WM 世界模型线**:halo_partner 采集 + 引擎深度交付包。训练代码随 SG-1 放行; 过闸小时以 halo_qc floor 档普查口径引用
|
| 55 |
+
- **IDM 线**:第一方采集全量带键鼠事件流(可训练性随权利裁决)。方法验证走 D2E; 硬门未过, 先收敛现有语料再买素材(ml/README)
|
| 56 |
+
- **VL 线**:任意第一方 RGB 帧 + gaming-500 bench。游戏识别 SOTA 代理; cohort 数据到位后重新定基
|
| 57 |
+
|
| 58 |
+
### 外部语料(全部禁训)
|
| 59 |
+
|
| 60 |
+
| 语料 | 规模 | 状态 | 用途 |
|
| 61 |
+
|---|---|---|---|
|
| 62 |
+
| D2E-480p | 268.7 h / 29 款 / 原生 raw HID delta | 禁训·方法验证 | IDM 线(zero-shot 止损复核, research/28) |
|
| 63 |
+
| gaming-500-hours | ~500 h / 168 款(HF, 无 license) | 禁训·方法验证 | VL bench 真值 + IDM 早期验证 |
|
| 64 |
+
| VPT (OpenAI) | 6.74 TB / 117,789 对象(R2 镜像) | 禁训 | WM/IDM 参考(research/50 清桶决议未执行, 待 owner 裁决) |
|
| 65 |
+
| CS:GO / GF-Minecraft | R2 镜像 | 禁训 | 参考语料 |
|
| 66 |
+
|
| 67 |
+
---
|
| 68 |
+
*生成于 workflow dataminer-weekly-report · 报告与 state 存 R2 `reports/weekly/`*
|
closeout-final/weekly/dm-weekly-2026-09-05.md
ADDED
|
@@ -0,0 +1,61 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# DataMiner 数据周报 · 2026-08-29 ~ 2026-09-05
|
| 2 |
+
|
| 3 |
+
> schema dm-weekly-report/1 · 窗口按对象上传时间(Asia/Shanghai)· 健康判定同 docs/146(console 口径)
|
| 4 |
+
|
| 5 |
+
## 一、本周采集与质检
|
| 6 |
+
|
| 7 |
+
- 总量:**2.09 h** / 12 会话 / 761 clips
|
| 8 |
+
- 健康异常:**33**(4.3%),其中补帧越线 33
|
| 9 |
+
- 音频:缺失 0 · 静音怀疑 0
|
| 10 |
+
- session 验收报告覆盖(全量口径):2/1015
|
| 11 |
+
|
| 12 |
+
| 游戏 | 会话 | clips | 小时 | 异常 | 异常率 | 无音频 | 静音疑 |
|
| 13 |
+
|---|---|---|---|---|---|---|---|
|
| 14 |
+
| cyberpunk-2077-steam | 12 | 761 | 2.09 | 33 | 4.3% | 0 | 0 |
|
| 15 |
+
|
| 16 |
+
### 逐日
|
| 17 |
+
|
| 18 |
+
| 日期 | clips | 小时 | 异常 |
|
| 19 |
+
|---|---|---|---|
|
| 20 |
+
| 08-29 | 761 | 2.09 | 33 |
|
| 21 |
+
|
| 22 |
+
### 质检 findings
|
| 23 |
+
|
| 24 |
+
- 无
|
| 25 |
+
|
| 26 |
+
## 二、ML 各线可用数据(累计)
|
| 27 |
+
|
| 28 |
+
| 游戏 | 授权档 | 累计小时 | 累计 clips |
|
| 29 |
+
|---|---|---|---|
|
| 30 |
+
| cyberpunk-2077-steam | 🚫 三方待授权 | 142.4 | 51771 |
|
| 31 |
+
| halo-infinite-steam-release-test | ✅ 伙伴授权可训练(C-1) | 27.9 | 10122 |
|
| 32 |
+
| fallout4-steam-release-test | ⏸ 待 C-1 扩展(微软系) | 2.8 | 963 |
|
| 33 |
+
| wolfenstein2-steam | ⏸ 待 C-1 扩展(微软系) | 0.2 | 67 |
|
| 34 |
+
| atomic-heart-pc | 🚫 三方待授权 | 0.1 | 39 |
|
| 35 |
+
| resident-evil-village-pc | 🚫 三方待授权 | 0.1 | 35 |
|
| 36 |
+
| doom-eternal-pc | ⏸ 待 C-1 扩展(微软系) | 0.1 | 33 |
|
| 37 |
+
| red-dead-redemption-2-pc | 🚫 三方待授权 | 0.1 | 32 |
|
| 38 |
+
| deathloop-pc | ⏸ 待 C-1 扩展(微软系) | 0.1 | 33 |
|
| 39 |
+
| starfield-pc | ⏸ 待 C-1 扩展(微软系) | 0.1 | 29 |
|
| 40 |
+
| call-of-duty-steam-release-test | ⏸ 待 C-1 扩展(微软系) | 0.1 | 22 |
|
| 41 |
+
| wuchang-fallen-feathers-steam | 🚫 三方待授权 | 0.0 | 7 |
|
| 42 |
+
|
| 43 |
+
- 引擎深度交付(verified/):**4 包**
|
| 44 |
+
|
| 45 |
+
### 各线口径
|
| 46 |
+
|
| 47 |
+
- **WM 世界模型线**:halo_partner 采集 + 引擎深度交付包。训练代码随 SG-1 放行; 过闸小时以 halo_qc floor 档普查口径引用
|
| 48 |
+
- **IDM 线**:第一方采集全量带键鼠事件流(可训练性随权利裁决)。方法验证走 D2E; 硬门未过, 先收敛现有语料再买素材(ml/README)
|
| 49 |
+
- **VL 线**:任意第一方 RGB 帧 + gaming-500 bench。游戏识别 SOTA 代理; cohort 数据到位后重新定基
|
| 50 |
+
|
| 51 |
+
### 外部语料(全部禁训)
|
| 52 |
+
|
| 53 |
+
| 语料 | 规模 | 状态 | 用途 |
|
| 54 |
+
|---|---|---|---|
|
| 55 |
+
| D2E-480p | 268.7 h / 29 款 / 原生 raw HID delta | 禁训·方法验证 | IDM 线(zero-shot 止损复核, research/28) |
|
| 56 |
+
| gaming-500-hours | ~500 h / 168 款(HF, 无 license) | 禁训·方法验证 | VL bench 真值 + IDM 早期验证 |
|
| 57 |
+
| VPT (OpenAI) | 6.74 TB / 117,789 对象(R2 镜像) | 禁训 | WM/IDM 参考(research/50 清桶决议未执行, 待 owner 裁决) |
|
| 58 |
+
| CS:GO / GF-Minecraft | R2 镜像 | 禁训 | 参考语料 |
|
| 59 |
+
|
| 60 |
+
---
|
| 61 |
+
*生成于 workflow dataminer-weekly-report · 报告与 state 存 R2 `reports/weekly/`*
|
closeout-final/weekly/dm-weekly-2026-09-12.md
ADDED
|
@@ -0,0 +1,83 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# DataMiner 数据周报 · 2026-09-05 ~ 2026-09-12(更正版)
|
| 2 |
+
|
| 3 |
+
> ⚠️ **更正版:09-12 自动卡片因周报车道盲区报 0 clips;本卡片按两条源车道重算并首次回填存量台账(docs/168)**
|
| 4 |
+
|
| 5 |
+
> schema dm-weekly-report/1 · 窗口按对象上传时间(Asia/Shanghai)· 健康判定同 docs/146(console 口径)· 累计 = R2 存量台账(docs/168)
|
| 6 |
+
> 采集员维度看 console:[按采集员](https://console.diogenesmodel.com/output?from=2026-09-05&to=2026-09-11&bucket=day&timeBasis=recorded&groupBy=annotator) · [按游戏](https://console.diogenesmodel.com/output?from=2026-09-05&to=2026-09-11&bucket=day&timeBasis=recorded&groupBy=game)(同窗口, 录制口径)
|
| 7 |
+
|
| 8 |
+
## 一、本周采集与质检
|
| 9 |
+
|
| 10 |
+
- 总量:**58.43 h** / 317 会话 / 21221 clips(上传口径)
|
| 11 |
+
- 录制口径(估算, console 每日进线同族):**58.02 h** / 21071 clips;与上传口径差 -0.41 h(-0.7%)
|
| 12 |
+
- 存量台账:本期入账 109585 · 出账(R2 已删)0 · 现有 109585 clips
|
| 13 |
+
- **补录 88363 clips / 243.24 h**(上传早于本窗、此前各期漏计——累计口径已含, 本周读数不含)
|
| 14 |
+
- 健康异常:**86**(0.4%)——轻度补帧 73 · 重度(>20%) 13 · **整段冻结 0**
|
| 15 |
+
- 音频:缺失 0 · 静音怀疑 29
|
| 16 |
+
- session 验收报告覆盖(两条源车道全量):2/1719(0.1%)
|
| 17 |
+
|
| 18 |
+
| 游戏 | 会话 | clips | 小时 | 异常 | 异常率 | 无音频 | 静音疑 |
|
| 19 |
+
|---|---|---|---|---|---|---|---|
|
| 20 |
+
| red-dead-redemption-2-pc | 112 | 7832 | 21.57 | 11 | 0.1% | 0 | 5 |
|
| 21 |
+
| dying-light-pc | 99 | 6791 | 18.71 | 63 | 0.9% | 0 | 18 |
|
| 22 |
+
| gta5-legacy-pc | 105 | 6597 | 18.15 | 12 | 0.2% | 0 | 6 |
|
| 23 |
+
| kingdom-come-deliverance-2-steam | 1 | 1 | 0.01 | 0 | 0.0% | 0 | 0 |
|
| 24 |
+
|
| 25 |
+
### 逐日
|
| 26 |
+
|
| 27 |
+
| 日期 | clips | 小时 | 异常 |
|
| 28 |
+
|---|---|---|---|
|
| 29 |
+
| 09-05 | 149 | 0.41 | 0 |
|
| 30 |
+
| 09-07 | 6438 | 17.70 | 18 |
|
| 31 |
+
| 09-08 | 4121 | 11.36 | 48 |
|
| 32 |
+
| 09-09 | 4339 | 11.95 | 4 |
|
| 33 |
+
| 09-10 | 5054 | 13.93 | 14 |
|
| 34 |
+
| 09-11 | 1120 | 3.09 | 2 |
|
| 35 |
+
|
| 36 |
+
### 质检 findings
|
| 37 |
+
|
| 38 |
+
- 六件套不完整 ×1: 0ef35e9d(0a660ca2) 缺 ['manifest']
|
| 39 |
+
|
| 40 |
+
## 二、ML 各线可用数据(累计)
|
| 41 |
+
|
| 42 |
+
> 口径:R2 存量台账(有则入账、删则出账),与 console「游戏分布」面板(docs/153)同为存量读数。
|
| 43 |
+
> 「授权档」≠ 可训练:质量闸放行数以 console 回看为准(docs/167 §2.3 实测 0 条 passed)。
|
| 44 |
+
|
| 45 |
+
| 游戏 | 授权档 | 累计小时 | 累计 clips |
|
| 46 |
+
|---|---|---|---|
|
| 47 |
+
| cyberpunk-2077-steam | 🚫 三方待授权 | 202.1 | 73464 |
|
| 48 |
+
| halo-infinite-steam-release-test | ✅ 伙伴授权可训练(C-1) | 27.9 | 10122 |
|
| 49 |
+
| red-dead-redemption-2-pc | 🚫 三方待授权 | 27.2 | 9876 |
|
| 50 |
+
| dying-light-pc | 🚫 三方待授权 | 20.6 | 7465 |
|
| 51 |
+
| gta5-legacy-pc | 🚫 三方待授权 | 20.4 | 7443 |
|
| 52 |
+
| fallout4-steam-release-test | ⏸ 待 C-1 扩展(微软系) | 2.7 | 964 |
|
| 53 |
+
| wolfenstein2-steam | ⏸ 待 C-1 扩展(微软系) | 0.2 | 67 |
|
| 54 |
+
| atomic-heart-pc | 🚫 三方待授权 | 0.1 | 39 |
|
| 55 |
+
| resident-evil-village-pc | 🚫 三方待授权 | 0.1 | 35 |
|
| 56 |
+
| doom-eternal-pc | ⏸ 待 C-1 扩展(微软系) | 0.1 | 33 |
|
| 57 |
+
| deathloop-pc | ⏸ 待 C-1 扩展(微软系) | 0.1 | 33 |
|
| 58 |
+
| starfield-pc | ⏸ 待 C-1 扩展(微软系) | 0.1 | 29 |
|
| 59 |
+
| _missing_target | ? | 0.0 | 1 |
|
| 60 |
+
| wuchang-fallen-feathers-steam | 🚫 三方待授权 | 0.0 | 7 |
|
| 61 |
+
| call-of-duty-steam-release-test | ⏸ 待 C-1 扩展(微软系) | 0.0 | 6 |
|
| 62 |
+
| kingdom-come-deliverance-2-steam | 🚫 三方待授权 | 0.0 | 1 |
|
| 63 |
+
|
| 64 |
+
- 引擎深度交付(verified/):**4 包**,最新一包 2026-08-08(距今 **35 天**)
|
| 65 |
+
|
| 66 |
+
### 各线口径
|
| 67 |
+
|
| 68 |
+
- **WM 世界模型线**:halo_partner 采集 + 引擎深度交付包。训练代码随 SG-1 放行; 过闸小时以 halo_qc floor 档普查口径引用
|
| 69 |
+
- **IDM 线**:第一方采集全量带键鼠事件流(可训练性随权利裁决)。方法验证走 D2E; 硬门未过, 先收敛现有语料再买素材(ml/README)
|
| 70 |
+
- **VL 线**:任意第一方 RGB 帧 + gaming-500 bench。游戏识别 SOTA 代理; cohort 数据到位后重新定基
|
| 71 |
+
|
| 72 |
+
### 外部语料(全部禁训)
|
| 73 |
+
|
| 74 |
+
> 静态常量, 不查桶——每行自带截止日, 过期一眼看得见。
|
| 75 |
+
|
| 76 |
+
| 语料 | 规模 | 状态 | 用途 | 数字截止 |
|
| 77 |
+
|---|---|---|---|---|
|
| 78 |
+
| D2E-480p | 268.7 h / 29 款 / 原生 raw HID delta | 禁训·方法验证 | IDM 线(zero-shot 止损复核, research/28) | 2026-08-13 |
|
| 79 |
+
| gaming-500-hours | ~500 h / 168 款(HF, 无 license) | 禁训·方法验证 | VL bench 真值 + IDM 早期验证 | 2026-08-13 |
|
| 80 |
+
| VPT / CS:GO / GF-Minecraft / D2E 的 R2 镜像 | 已删除(实删 118,773 对象 / 7.82 TB) | 禁训·**镜像已清** | 2026-09-04 清桶完毕, 仅存 vpt-schema-sample;VPT 从此不可复得(research/50 R6) | 2026-09-04 |
|
| 81 |
+
|
| 82 |
+
---
|
| 83 |
+
*生成于 workflow dataminer-weekly-report · 脚本 @41a93de · 报告 / state / 台账存 R2 `reports/weekly/` · console 总览 https://console.diogenesmodel.com/overview*
|
closeout-final/weekly/dm-weekly-2026-09-19.md
ADDED
|
@@ -0,0 +1,79 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# DataMiner 数据周报 · 2026-09-12 ~ 2026-09-19
|
| 2 |
+
|
| 3 |
+
> schema dm-weekly-report/1 · 窗口按对象上传时间(Asia/Shanghai)· 健康判定同 docs/146(console 口径)· 累计 = R2 存量台账(docs/168)
|
| 4 |
+
> 采集员维度看 console:[按采集员](https://console.diogenesmodel.com/output?from=2026-09-12&to=2026-09-18&bucket=day&timeBasis=recorded&groupBy=annotator) · [按游戏](https://console.diogenesmodel.com/output?from=2026-09-12&to=2026-09-18&bucket=day&timeBasis=recorded&groupBy=game)(同窗口, 录制口径)
|
| 5 |
+
|
| 6 |
+
## 一、本周采集与质检
|
| 7 |
+
|
| 8 |
+
- 总量:**70.00 h** / 376 会话 / 25412 clips(上传口径)
|
| 9 |
+
- 录制口径(估算, console 每日进线同族):**71.23 h** / 25860 clips;与上传口径差 +1.23 h(+1.8%)
|
| 10 |
+
- 存量台账:本期入账 25864 · 出账(R2 已删)0 · 现有 135449 clips
|
| 11 |
+
- 健康异常:**90**(0.4%)——轻度补帧 90 · 重度(>20%) 0 · **整段冻结 0**
|
| 12 |
+
- 音频:缺失 0 · 静音怀疑 121
|
| 13 |
+
- session 验收报告覆盖(两条源车道全量):2101/2102(100.0%)
|
| 14 |
+
|
| 15 |
+
| 游戏 | 会话 | clips | 小时 | 异常 | 异常率 | 无音频 | 静音疑 |
|
| 16 |
+
|---|---|---|---|---|---|---|---|
|
| 17 |
+
| red-dead-redemption-2-pc | 142 | 10165 | 28.02 | 24 | 0.2% | 0 | 7 |
|
| 18 |
+
| gta5-legacy-pc | 128 | 8163 | 22.44 | 20 | 0.2% | 0 | 15 |
|
| 19 |
+
| dying-light-pc | 104 | 7080 | 19.50 | 46 | 0.6% | 0 | 99 |
|
| 20 |
+
| kingdom-come-deliverance-2-steam | 2 | 4 | 0.04 | 0 | 0.0% | 0 | 0 |
|
| 21 |
+
|
| 22 |
+
### 逐日
|
| 23 |
+
|
| 24 |
+
| 日期 | clips | 小时 | 异常 |
|
| 25 |
+
|---|---|---|---|
|
| 26 |
+
| 09-14 | 5350 | 14.75 | 14 |
|
| 27 |
+
| 09-15 | 6251 | 17.20 | 8 |
|
| 28 |
+
| 09-16 | 2276 | 6.26 | 2 |
|
| 29 |
+
| 09-17 | 5962 | 16.42 | 42 |
|
| 30 |
+
| 09-18 | 5573 | 15.37 | 24 |
|
| 31 |
+
|
| 32 |
+
### 质检 findings
|
| 33 |
+
|
| 34 |
+
- 无
|
| 35 |
+
|
| 36 |
+
## 二、ML 各线可用数据(累计)
|
| 37 |
+
|
| 38 |
+
> 口径:R2 存量台账(有则入账、删则出账),与 console「游戏分布」面板(docs/153)同为存量读数。
|
| 39 |
+
> 「授权档」≠ 可训练:质量闸放行数以 console 回看为准(docs/167 §2.3 实测 0 条 passed)。
|
| 40 |
+
|
| 41 |
+
| 游戏 | 授权档 | 累计小时 | 累计 clips |
|
| 42 |
+
|---|---|---|---|
|
| 43 |
+
| cyberpunk-2077-steam | 🚫 三方待授权 | 202.1 | 73468 |
|
| 44 |
+
| red-dead-redemption-2-pc | 🚫 三方待授权 | 55.2 | 20041 |
|
| 45 |
+
| gta5-legacy-pc | 🚫 三方待授权 | 43.0 | 15654 |
|
| 46 |
+
| dying-light-pc | 🚫 三方待授权 | 41.1 | 14945 |
|
| 47 |
+
| halo-infinite-steam-release-test | ✅ 伙伴授权可训练(C-1) | 27.9 | 10122 |
|
| 48 |
+
| fallout4-steam-release-test | ⏸ 待 C-1 扩展(微软系) | 2.7 | 964 |
|
| 49 |
+
| wolfenstein2-steam | ⏸ 待 C-1 扩展(微软系) | 0.2 | 67 |
|
| 50 |
+
| atomic-heart-pc | 🚫 三方待授权 | 0.1 | 39 |
|
| 51 |
+
| resident-evil-village-pc | 🚫 三方待授权 | 0.1 | 35 |
|
| 52 |
+
| doom-eternal-pc | ⏸ 待 C-1 扩展(微软系) | 0.1 | 33 |
|
| 53 |
+
| deathloop-pc | ⏸ 待 C-1 扩展(微软系) | 0.1 | 33 |
|
| 54 |
+
| starfield-pc | ⏸ 待 C-1 扩展(微软系) | 0.1 | 29 |
|
| 55 |
+
| kingdom-come-deliverance-2-steam | 🚫 三方待授权 | 0.1 | 5 |
|
| 56 |
+
| _missing_target | ? | 0.0 | 1 |
|
| 57 |
+
| wuchang-fallen-feathers-steam | 🚫 三方待授权 | 0.0 | 7 |
|
| 58 |
+
| call-of-duty-steam-release-test | ⏸ 待 C-1 扩展(微软系) | 0.0 | 6 |
|
| 59 |
+
|
| 60 |
+
- 引擎深度交付(verified/):**4 包**,最新一包 2026-08-08(距今 **42 天**)
|
| 61 |
+
|
| 62 |
+
### 各线口径
|
| 63 |
+
|
| 64 |
+
- **WM 世界模型线**:halo_partner 采集 + 引擎深度交付包。训练代码随 SG-1 放行; 过闸小时以 halo_qc floor 档普查口径引用
|
| 65 |
+
- **IDM 线**:第一方采集全量带键鼠事件流(可训练性随权利裁决)。方法验证走 D2E; 硬门未过, 先收敛现有语料再买素材(ml/README)
|
| 66 |
+
- **VL 线**:任意第一方 RGB 帧 + gaming-500 bench。游戏识别 SOTA 代理; cohort 数据到位后重新定基
|
| 67 |
+
|
| 68 |
+
### 外部语料(全部禁训)
|
| 69 |
+
|
| 70 |
+
> 静态常量, 不查桶——每行自带截止日, 过期一眼看得见。
|
| 71 |
+
|
| 72 |
+
| 语料 | 规模 | 状态 | 用途 | 数字截止 |
|
| 73 |
+
|---|---|---|---|---|
|
| 74 |
+
| D2E-480p | 268.7 h / 29 款 / 原生 raw HID delta | 禁训·方法验证 | IDM 线(zero-shot 止损复核, research/28) | 2026-08-13 |
|
| 75 |
+
| gaming-500-hours | ~500 h / 168 款(HF, 无 license) | 禁训·方法验证 | VL bench 真值 + IDM 早期验证 | 2026-08-13 |
|
| 76 |
+
| VPT / CS:GO / GF-Minecraft / D2E 的 R2 镜像 | 已删除(实删 118,773 对象 / 7.82 TB) | 禁训·**镜像已清** | 2026-09-04 清桶完毕, 仅存 vpt-schema-sample;VPT 从此不可复得(research/50 R6) | 2026-09-04 |
|
| 77 |
+
|
| 78 |
+
---
|
| 79 |
+
*生成于 workflow dataminer-weekly-report · 脚本 @726e25b · 报告 / state / 台账存 R2 `reports/weekly/` · console 总览 https://console.diogenesmodel.com/overview*
|
closeout-final/weekly/dm-weekly-2026-09-26.md
ADDED
|
@@ -0,0 +1,81 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# DataMiner 数据周报 · 2026-09-19 ~ 2026-09-26
|
| 2 |
+
|
| 3 |
+
> schema dm-weekly-report/1 · 窗口按对象上传时间(Asia/Shanghai)· 健康判定同 docs/146(console 口径)· 累计 = R2 存量台账(docs/168)
|
| 4 |
+
> 采集员维度看 console:[按采集员](https://console.diogenesmodel.com/output?from=2026-09-19&to=2026-09-25&bucket=day&timeBasis=recorded&groupBy=annotator) · [按游戏](https://console.diogenesmodel.com/output?from=2026-09-19&to=2026-09-25&bucket=day&timeBasis=recorded&groupBy=game)(同窗口, 录制口径)
|
| 5 |
+
|
| 6 |
+
## 一、本周采集与质检
|
| 7 |
+
|
| 8 |
+
- 总量:**66.51 h** / 368 会话 / 24162 clips(上传口径)
|
| 9 |
+
- 录制口径(估算, console 每日进线同族):**65.28 h** / 23715 clips;与上传口径差 -1.23 h(-1.8%)
|
| 10 |
+
- 存量台账:本期入账 23711 · 出账(R2 已删)0 · 现有 159160 clips
|
| 11 |
+
- 健康异常:**147**(0.6%)——轻度补帧 144 · 重度(>20%) 3 · **整段冻结 0**
|
| 12 |
+
- 音频:缺失 1 · 静音怀疑 65
|
| 13 |
+
- session 验收报告覆盖(两条源车道全量):2461/2461(100.0%)
|
| 14 |
+
|
| 15 |
+
| 游戏 | 会话 | clips | 小时 | 异常 | 异常率 | 无音频 | 静音疑 |
|
| 16 |
+
|---|---|---|---|---|---|---|---|
|
| 17 |
+
| red-dead-redemption-2-pc | 144 | 9950 | 27.41 | 48 | 0.5% | 1 | 10 |
|
| 18 |
+
| gta5-legacy-pc | 129 | 8242 | 22.70 | 46 | 0.6% | 0 | 22 |
|
| 19 |
+
| dying-light-pc | 94 | 5966 | 16.39 | 49 | 0.8% | 0 | 33 |
|
| 20 |
+
| cyberpunk-2077-steam | 1 | 4 | 0.01 | 4 | 100.0% | 0 | 0 |
|
| 21 |
+
|
| 22 |
+
### 逐日
|
| 23 |
+
|
| 24 |
+
| 日期 | clips | 小时 | 异常 |
|
| 25 |
+
|---|---|---|---|
|
| 26 |
+
| 09-19 | 452 | 1.24 | 12 |
|
| 27 |
+
| 09-20 | 6247 | 17.19 | 33 |
|
| 28 |
+
| 09-21 | 4592 | 12.63 | 42 |
|
| 29 |
+
| 09-22 | 4452 | 12.26 | 32 |
|
| 30 |
+
| 09-23 | 3791 | 10.44 | 15 |
|
| 31 |
+
| 09-24 | 3840 | 10.58 | 11 |
|
| 32 |
+
| 09-25 | 788 | 2.18 | 2 |
|
| 33 |
+
|
| 34 |
+
### 质检 findings
|
| 35 |
+
|
| 36 |
+
- 六件套不完整 ×1: f7ca1323(0b4186e6) 缺 ['manifest']
|
| 37 |
+
|
| 38 |
+
## 二、ML 各线可用数据(累计)
|
| 39 |
+
|
| 40 |
+
> 口径:R2 存量台账(有则入账、删则出账),与 console「游戏分布」面板(docs/153)同为存量读数。
|
| 41 |
+
> 「授权档」≠ 可训练:质量闸放行数以 console 回看为准(docs/167 §2.3 实测 0 条 passed)。
|
| 42 |
+
|
| 43 |
+
| 游戏 | 授权档 | 累计小时 | 累计 clips |
|
| 44 |
+
|---|---|---|---|
|
| 45 |
+
| cyberpunk-2077-steam | 🚫 三方待授权 | 202.1 | 73468 |
|
| 46 |
+
| red-dead-redemption-2-pc | 🚫 三方待授权 | 82.6 | 29992 |
|
| 47 |
+
| gta5-legacy-pc | 🚫 三方待授权 | 65.6 | 23848 |
|
| 48 |
+
| dying-light-pc | 🚫 三方待授权 | 56.4 | 20511 |
|
| 49 |
+
| halo-infinite-steam-release-test | ✅ 伙伴授权可训练(C-1) | 27.9 | 10122 |
|
| 50 |
+
| fallout4-steam-release-test | ⏸ 待 C-1 扩展(微软系) | 2.7 | 964 |
|
| 51 |
+
| wolfenstein2-steam | ⏸ 待 C-1 扩展(微软系) | 0.2 | 67 |
|
| 52 |
+
| atomic-heart-pc | 🚫 三方待授权 | 0.1 | 39 |
|
| 53 |
+
| resident-evil-village-pc | 🚫 三方待授权 | 0.1 | 35 |
|
| 54 |
+
| doom-eternal-pc | ⏸ 待 C-1 扩展(微软系) | 0.1 | 33 |
|
| 55 |
+
| deathloop-pc | ⏸ 待 C-1 扩展(微软系) | 0.1 | 33 |
|
| 56 |
+
| starfield-pc | ⏸ 待 C-1 扩展(微软系) | 0.1 | 29 |
|
| 57 |
+
| kingdom-come-deliverance-2-steam | 🚫 三方待授权 | 0.1 | 5 |
|
| 58 |
+
| _missing_target | ? | 0.0 | 1 |
|
| 59 |
+
| wuchang-fallen-feathers-steam | 🚫 三方待授权 | 0.0 | 7 |
|
| 60 |
+
| call-of-duty-steam-release-test | ⏸ 待 C-1 扩展(微软系) | 0.0 | 6 |
|
| 61 |
+
|
| 62 |
+
- 引擎深度交付(verified/):**4 包**,最新一包 2026-08-08(距今 **49 天**)
|
| 63 |
+
|
| 64 |
+
### 各线口径
|
| 65 |
+
|
| 66 |
+
- **WM 世界模型线**:halo_partner 采集 + 引擎深度交付包。训练代码随 SG-1 放行; 过闸小时以 halo_qc floor 档普查口径引用
|
| 67 |
+
- **IDM 线**:第一方采集全量带键鼠事件流(可训练性随权利裁决)。方法验证走 D2E; 硬门未过, 先收敛现有语料再买素材(ml/README)
|
| 68 |
+
- **VL 线**:任意第一方 RGB 帧 + gaming-500 bench。游戏识别 SOTA 代理; cohort 数据到位后重新定基
|
| 69 |
+
|
| 70 |
+
### 外部语料(全部禁训)
|
| 71 |
+
|
| 72 |
+
> 静态常量, 不查桶——每行自带截止日, 过期一眼看得见。
|
| 73 |
+
|
| 74 |
+
| 语料 | 规模 | 状态 | 用途 | 数字截止 |
|
| 75 |
+
|---|---|---|---|---|
|
| 76 |
+
| D2E-480p | 268.7 h / 29 款 / 原生 raw HID delta | 禁训·方法验证 | IDM 线(zero-shot 止损复核, research/28) | 2026-08-13 |
|
| 77 |
+
| gaming-500-hours | ~500 h / 168 款(HF, 无 license) | 禁训·方法验证 | VL bench 真值 + IDM 早期验证 | 2026-08-13 |
|
| 78 |
+
| VPT / CS:GO / GF-Minecraft / D2E 的 R2 镜像 | 已删除(实删 118,773 对象 / 7.82 TB) | 禁训·**镜像已清** | 2026-09-04 清桶完毕, 仅存 vpt-schema-sample;VPT 从此不可复得(research/50 R6) | 2026-09-04 |
|
| 79 |
+
|
| 80 |
+
---
|
| 81 |
+
*生成于 workflow dataminer-weekly-report · 脚本 @b77097e · 报告 / state / 台账存 R2 `reports/weekly/` · console 总览 https://console.diogenesmodel.com/overview*
|
closeout-final/weekly/dm-weekly-2026-10-03.md
ADDED
|
@@ -0,0 +1,77 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# DataMiner 数据周报 · 2026-09-26 ~ 2026-10-03
|
| 2 |
+
|
| 3 |
+
> schema dm-weekly-report/1 · 窗口按对象上传时间(Asia/Shanghai)· 健康判定同 docs/146(console 口径)· 累计 = R2 存量台账(docs/168)
|
| 4 |
+
> 采集员维度看 console:[按采集员](https://console.diogenesmodel.com/output?from=2026-09-26&to=2026-10-02&bucket=day&timeBasis=recorded&groupBy=annotator) · [按游戏](https://console.diogenesmodel.com/output?from=2026-09-26&to=2026-10-02&bucket=day&timeBasis=recorded&groupBy=game)(同窗口, 录制口径)
|
| 5 |
+
|
| 6 |
+
## 一、本周采集与质检
|
| 7 |
+
|
| 8 |
+
- 总量:**28.97 h** / 149 会话 / 10515 clips(上传口径)
|
| 9 |
+
- 录制口径(估算, console 每日进线同族):**28.97 h** / 10515 clips;与上传口径差 -0.00 h(-0.0%)
|
| 10 |
+
- 存量台账:本期入账 10515 · 出账(R2 已删)0 · 现有 169675 clips
|
| 11 |
+
- 健康异常:**39**(0.4%)——轻度补帧 36 · 重度(>20%) 2 · **整段冻结 1**
|
| 12 |
+
- 音频:缺失 0 · 静音怀疑 65
|
| 13 |
+
- session 验收报告覆盖(两条源车道全量):2610/2610(100.0%)
|
| 14 |
+
|
| 15 |
+
| 游戏 | 会话 | clips | 小时 | 异常 | 异常率 | 无音频 | 静音疑 |
|
| 16 |
+
|---|---|---|---|---|---|---|---|
|
| 17 |
+
| red-dead-redemption-2-pc | 85 | 6224 | 17.15 | 29 | 0.5% | 0 | 52 |
|
| 18 |
+
| gta5-legacy-pc | 64 | 4291 | 11.82 | 10 | 0.2% | 0 | 13 |
|
| 19 |
+
|
| 20 |
+
### 逐日
|
| 21 |
+
|
| 22 |
+
| 日期 | clips | 小时 | 异常 |
|
| 23 |
+
|---|---|---|---|
|
| 24 |
+
| 09-28 | 3790 | 10.44 | 16 |
|
| 25 |
+
| 09-29 | 2687 | 7.40 | 8 |
|
| 26 |
+
| 09-30 | 3260 | 8.98 | 12 |
|
| 27 |
+
| 10-01 | 778 | 2.15 | 3 |
|
| 28 |
+
|
| 29 |
+
### 质检 findings
|
| 30 |
+
|
| 31 |
+
- 🚨 本窗 10515 clips, 上窗 24162, 环比 -56%(阈值 -50%)——先确认是真降产还是又漏了车道
|
| 32 |
+
- 整段冻结 clip(dup≥100%)×1: b4803102(6c7c0376)
|
| 33 |
+
|
| 34 |
+
## 二、ML 各线可用数据(累计)
|
| 35 |
+
|
| 36 |
+
> 口径:R2 存量台账(有则入账、删则出账),与 console「游戏分布」面板(docs/153)同为存量读数。
|
| 37 |
+
> 「授权档」≠ 可训练:质量闸放行数以 console 回看为准(docs/167 §2.3 实测 0 条 passed)。
|
| 38 |
+
|
| 39 |
+
| 游戏 | 授权档 | 累计小时 | 累计 clips |
|
| 40 |
+
|---|---|---|---|
|
| 41 |
+
| cyberpunk-2077-steam | 🚫 三方待授权 | 202.1 | 73468 |
|
| 42 |
+
| red-dead-redemption-2-pc | 🚫 三方待授权 | 99.8 | 36216 |
|
| 43 |
+
| gta5-legacy-pc | 🚫 三方待授权 | 77.4 | 28139 |
|
| 44 |
+
| dying-light-pc | 🚫 三方待授权 | 56.4 | 20511 |
|
| 45 |
+
| halo-infinite-steam-release-test | ✅ 伙伴授权可训练(C-1) | 27.9 | 10122 |
|
| 46 |
+
| fallout4-steam-release-test | ⏸ 待 C-1 扩展(微软系) | 2.7 | 964 |
|
| 47 |
+
| wolfenstein2-steam | ⏸ 待 C-1 扩展(微软系) | 0.2 | 67 |
|
| 48 |
+
| atomic-heart-pc | 🚫 三方待授权 | 0.1 | 39 |
|
| 49 |
+
| resident-evil-village-pc | 🚫 三方待授权 | 0.1 | 35 |
|
| 50 |
+
| doom-eternal-pc | ⏸ 待 C-1 扩展(微软系) | 0.1 | 33 |
|
| 51 |
+
| deathloop-pc | ⏸ 待 C-1 扩展(微软系) | 0.1 | 33 |
|
| 52 |
+
| starfield-pc | ⏸ 待 C-1 扩展(微软系) | 0.1 | 29 |
|
| 53 |
+
| kingdom-come-deliverance-2-steam | 🚫 三方待授权 | 0.1 | 5 |
|
| 54 |
+
| _missing_target | ? | 0.0 | 1 |
|
| 55 |
+
| wuchang-fallen-feathers-steam | 🚫 三方待授权 | 0.0 | 7 |
|
| 56 |
+
| call-of-duty-steam-release-test | ⏸ 待 C-1 扩展(微软系) | 0.0 | 6 |
|
| 57 |
+
|
| 58 |
+
- 引擎深度交付(verified/):**4 包**,最新一包 2026-08-08(距今 **56 天**)
|
| 59 |
+
|
| 60 |
+
### 各线口径
|
| 61 |
+
|
| 62 |
+
- **WM 世界模型线**:halo_partner 采集 + 引擎深度交付包。训练代码随 SG-1 放行; 过闸小时以 halo_qc floor 档普查口径引用
|
| 63 |
+
- **IDM 线**:第一方采集全量带键鼠事件流(可训练性随权利裁决)。方法验证走 D2E; 硬门未过, 先收敛现有语料再买素材(ml/README)
|
| 64 |
+
- **VL 线**:任意第一方 RGB 帧 + gaming-500 bench。游戏识别 SOTA 代理; cohort 数据到位后重新定基
|
| 65 |
+
|
| 66 |
+
### 外部语料(全部禁训)
|
| 67 |
+
|
| 68 |
+
> 静态常量, 不查桶——每行自带截止日, 过期一眼看得见。
|
| 69 |
+
|
| 70 |
+
| 语料 | 规模 | 状态 | 用途 | 数字截止 |
|
| 71 |
+
|---|---|---|---|---|
|
| 72 |
+
| D2E-480p | 268.7 h / 29 款 / 原生 raw HID delta | 禁训·方法验证 | IDM 线(zero-shot 止损复核, research/28) | 2026-08-13 |
|
| 73 |
+
| gaming-500-hours | ~500 h / 168 款(HF, 无 license) | 禁训·方法验证 | VL bench 真值 + IDM 早期验证 | 2026-08-13 |
|
| 74 |
+
| VPT / CS:GO / GF-Minecraft / D2E 的 R2 镜像 | 已删除(实删 118,773 对象 / 7.82 TB) | 禁训·**镜像已清** | 2026-09-04 清桶完毕, 仅存 vpt-schema-sample;VPT 从此不可复得(research/50 R6) | 2026-09-04 |
|
| 75 |
+
|
| 76 |
+
---
|
| 77 |
+
*生成于 workflow dataminer-weekly-report · 脚本 @718b294 · 报告 / state / 台账存 R2 `reports/weekly/` · console 总览 https://console.diogenesmodel.com/overview*
|
closeout-final/weekly/dm-weekly-2026-10-10.md
ADDED
|
@@ -0,0 +1,76 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# DataMiner 数据周报 · 2026-10-03 ~ 2026-10-10
|
| 2 |
+
|
| 3 |
+
> schema dm-weekly-report/1 · 窗口按对象上传时间(Asia/Shanghai)· 健康判定同 docs/146(console 口径)· 累计 = R2 存量台账(docs/168)
|
| 4 |
+
> 采集员维度看 console:[按采集员](https://console.diogenesmodel.com/output?from=2026-10-03&to=2026-10-09&bucket=day&timeBasis=recorded&groupBy=annotator) · [按游戏](https://console.diogenesmodel.com/output?from=2026-10-03&to=2026-10-09&bucket=day&timeBasis=recorded&groupBy=game)(同窗口, 录制口径)
|
| 5 |
+
|
| 6 |
+
## 一、本周采集与质检
|
| 7 |
+
|
| 8 |
+
- 总量:**30.89 h** / 161 会话 / 11220 clips(上传口径)
|
| 9 |
+
- 录制口径(估算, console 每日进线同族):**34.17 h** / 12410 clips;与上传口径差 +3.28 h(+10.6%)
|
| 10 |
+
- 存量台账:本期入账 12410 · 出账(R2 已删)0 · 现有 182085 clips
|
| 11 |
+
- 健康异常:**184**(1.6%)——轻度补帧 184 · 重度(>20%) 0 · **整段冻结 0**
|
| 12 |
+
- 音频:缺失 1 · 静音怀疑 15
|
| 13 |
+
- session 验收报告覆盖(两条源车道全量):2787/2787(100.0%)
|
| 14 |
+
|
| 15 |
+
| 游戏 | 会话 | clips | 小时 | 异常 | 异常率 | 无音频 | 静音疑 |
|
| 16 |
+
|---|---|---|---|---|---|---|---|
|
| 17 |
+
| red-dead-redemption-2-pc | 51 | 4128 | 11.39 | 9 | 0.2% | 0 | 4 |
|
| 18 |
+
| far-cry-5-pc | 64 | 4015 | 11.03 | 167 | 4.2% | 1 | 0 |
|
| 19 |
+
| gta5-legacy-pc | 46 | 3077 | 8.47 | 8 | 0.3% | 0 | 11 |
|
| 20 |
+
|
| 21 |
+
### 逐日
|
| 22 |
+
|
| 23 |
+
| 日期 | clips | 小时 | 异常 |
|
| 24 |
+
|---|---|---|---|
|
| 25 |
+
| 10-08 | 5112 | 14.08 | 84 |
|
| 26 |
+
| 10-09 | 6108 | 16.81 | 100 |
|
| 27 |
+
|
| 28 |
+
### 质检 findings
|
| 29 |
+
|
| 30 |
+
- 无
|
| 31 |
+
|
| 32 |
+
## 二、ML 各线可用数据(累计)
|
| 33 |
+
|
| 34 |
+
> 口径:R2 存量台账(有则入账、删则出账),与 console「游戏分布」面板(docs/153)同为存量读数。
|
| 35 |
+
> 「授权档」≠ 可训练:质量闸放行数以 console 回看为准(docs/167 §2.3 实测 0 条 passed)。
|
| 36 |
+
|
| 37 |
+
| 游戏 | 授权档 | 累计小时 | 累计 clips |
|
| 38 |
+
|---|---|---|---|
|
| 39 |
+
| cyberpunk-2077-steam | 🚫 三方待授权 | 202.1 | 73468 |
|
| 40 |
+
| red-dead-redemption-2-pc | 🚫 三方待授权 | 111.6 | 40497 |
|
| 41 |
+
| gta5-legacy-pc | 🚫 三方待授权 | 88.5 | 32177 |
|
| 42 |
+
| dying-light-pc | 🚫 三方待授权 | 56.4 | 20511 |
|
| 43 |
+
| halo-infinite-steam-release-test | ✅ 伙伴授权可训练(C-1) | 27.9 | 10122 |
|
| 44 |
+
| far-cry-5-pc | 🚫 三方待授权 | 11.2 | 4091 |
|
| 45 |
+
| fallout4-steam-release-test | ⏸ 待 C-1 扩展(微软系) | 2.7 | 964 |
|
| 46 |
+
| wolfenstein2-steam | ⏸ 待 C-1 扩展(微软系) | 0.2 | 67 |
|
| 47 |
+
| atomic-heart-pc | 🚫 三方待授权 | 0.1 | 39 |
|
| 48 |
+
| resident-evil-village-pc | 🚫 三方待授权 | 0.1 | 35 |
|
| 49 |
+
| doom-eternal-pc | ⏸ 待 C-1 扩展(微软系) | 0.1 | 33 |
|
| 50 |
+
| deathloop-pc | ⏸ 待 C-1 扩展(微软系) | 0.1 | 33 |
|
| 51 |
+
| starfield-pc | ⏸ 待 C-1 扩展(微软系) | 0.1 | 29 |
|
| 52 |
+
| kingdom-come-deliverance-2-steam | 🚫 三方待授权 | 0.1 | 5 |
|
| 53 |
+
| _missing_target | ? | 0.0 | 1 |
|
| 54 |
+
| wuchang-fallen-feathers-steam | 🚫 三方待授权 | 0.0 | 7 |
|
| 55 |
+
| call-of-duty-steam-release-test | ⏸ 待 C-1 扩展(微软系) | 0.0 | 6 |
|
| 56 |
+
|
| 57 |
+
- 引擎深度交付(verified/):**4 包**,最新一包 2026-08-08(距今 **63 天**)
|
| 58 |
+
|
| 59 |
+
### 各线口径
|
| 60 |
+
|
| 61 |
+
- **WM 世界模型线**:halo_partner 采集 + 引擎深度交付包。训练代码随 SG-1 放行; 过闸小时以 halo_qc floor 档普查口径引用
|
| 62 |
+
- **IDM 线**:第一方采集全量带键鼠事件流(可训练性随权利裁决)。方法验证走 D2E; 硬门未过, 先收敛现有语料再买素材(ml/README)
|
| 63 |
+
- **VL 线**:任意第一方 RGB 帧 + gaming-500 bench。游戏识别 SOTA 代理; cohort 数据到位后重新定基
|
| 64 |
+
|
| 65 |
+
### 外部语料(全部禁训)
|
| 66 |
+
|
| 67 |
+
> 静态常量, 不查桶——每行自带截止日, 过期一眼看得见。
|
| 68 |
+
|
| 69 |
+
| 语料 | 规模 | 状态 | 用途 | 数字截止 |
|
| 70 |
+
|---|---|---|---|---|
|
| 71 |
+
| D2E-480p | 268.7 h / 29 款 / 原生 raw HID delta | 禁训·方法验证 | IDM 线(zero-shot 止损复核, research/28) | 2026-08-13 |
|
| 72 |
+
| gaming-500-hours | ~500 h / 168 款(HF, 无 license) | 禁训·方法验证 | VL bench 真值 + IDM 早期验证 | 2026-08-13 |
|
| 73 |
+
| VPT / CS:GO / GF-Minecraft / D2E 的 R2 镜像 | 已删除(实删 118,773 对象 / 7.82 TB) | 禁训·**镜像已清** | 2026-09-04 清桶完毕, 仅存 vpt-schema-sample;VPT 从此不可复得(research/50 R6) | 2026-09-04 |
|
| 74 |
+
|
| 75 |
+
---
|
| 76 |
+
*生成于 workflow dataminer-weekly-report · 脚本 @718b294 · 报告 / state / 台账存 R2 `reports/weekly/` · console 总览 https://console.diogenesmodel.com/overview*
|
closeout-final/weekly/dm-weekly-2026-10-17.md
ADDED
|
@@ -0,0 +1,81 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# DataMiner 数据周报 · 2026-10-10 ~ 2026-10-17
|
| 2 |
+
|
| 3 |
+
> schema dm-weekly-report/1 · 窗口按对象上传时间(Asia/Shanghai)· 健康判定同 docs/146(console 口径)· 累计 = R2 存量台账(docs/168)
|
| 4 |
+
> 采集员维度看 console:[按采集员](https://console.diogenesmodel.com/output?from=2026-10-10&to=2026-10-16&bucket=day&timeBasis=recorded&groupBy=annotator) · [按游戏](https://console.diogenesmodel.com/output?from=2026-10-10&to=2026-10-16&bucket=day&timeBasis=recorded&groupBy=game)(同窗口, 录制口径)
|
| 5 |
+
|
| 6 |
+
## 一、本周采集与质检
|
| 7 |
+
|
| 8 |
+
- 总量:**79.65 h** / 412 会话 / 28954 clips(上传口径)
|
| 9 |
+
- 录制口径(估算, console 每日进线同族):**80.92 h** / 29415 clips;与上传口径差 +1.27 h(+1.6%)
|
| 10 |
+
- 存量台账:本期入账 29415 · 出账(R2 已删)0 · 现有 211500 clips
|
| 11 |
+
- 健康异常:**244**(0.8%)——轻度补帧 242 · 重度(>20%) 0 · **整段冻结 2**
|
| 12 |
+
- 音频:缺失 3 · 静音怀疑 35
|
| 13 |
+
- session 验收报告覆盖(两条源车道全量):3197/3201(99.9%)
|
| 14 |
+
|
| 15 |
+
| 游戏 | 会话 | clips | 小时 | 异常 | 异常率 | 无音频 | 静音疑 |
|
| 16 |
+
|---|---|---|---|---|---|---|---|
|
| 17 |
+
| red-dead-redemption-2-pc | 148 | 11925 | 32.89 | 25 | 0.2% | 0 | 13 |
|
| 18 |
+
| far-cry-5-pc | 168 | 10278 | 28.16 | 197 | 1.9% | 3 | 0 |
|
| 19 |
+
| gta5-legacy-pc | 96 | 6751 | 18.60 | 22 | 0.3% | 0 | 22 |
|
| 20 |
+
|
| 21 |
+
### 逐日
|
| 22 |
+
|
| 23 |
+
| 日期 | clips | 小时 | 异常 |
|
| 24 |
+
|---|---|---|---|
|
| 25 |
+
| 10-10 | 5052 | 13.90 | 89 |
|
| 26 |
+
| 10-11 | 391 | 1.08 | 0 |
|
| 27 |
+
| 10-12 | 4500 | 12.38 | 39 |
|
| 28 |
+
| 10-13 | 6009 | 16.53 | 47 |
|
| 29 |
+
| 10-14 | 4226 | 11.62 | 33 |
|
| 30 |
+
| 10-15 | 3975 | 10.97 | 8 |
|
| 31 |
+
| 10-16 | 4801 | 13.18 | 28 |
|
| 32 |
+
|
| 33 |
+
### 质检 findings
|
| 34 |
+
|
| 35 |
+
- 整段冻结 clip(dup≥100%)×2: f7df8b63(55c73a3e) 00f98cc6(f48a2e8a)
|
| 36 |
+
|
| 37 |
+
## 二、ML 各线可用数据(累计)
|
| 38 |
+
|
| 39 |
+
> 口径:R2 存量台账(有则入账、删则出账),与 console「游戏分布」面板(docs/153)同为存量读数。
|
| 40 |
+
> 「授权档」≠ 可训练:质量闸放行数以 console 回看为准(docs/167 §2.3 实测 0 条 passed)。
|
| 41 |
+
|
| 42 |
+
| 游戏 | 授权档 | 累计小时 | 累计 clips |
|
| 43 |
+
|---|---|---|---|
|
| 44 |
+
| cyberpunk-2077-steam | 🚫 三方待授权 | 202.1 | 73468 |
|
| 45 |
+
| red-dead-redemption-2-pc | 🚫 三方待授权 | 144.9 | 52578 |
|
| 46 |
+
| gta5-legacy-pc | 🚫 三方待授权 | 108.2 | 39309 |
|
| 47 |
+
| dying-light-pc | 🚫 三方待授权 | 56.4 | 20511 |
|
| 48 |
+
| far-cry-5-pc | 🚫 三方待授权 | 39.2 | 14293 |
|
| 49 |
+
| halo-infinite-steam-release-test | ✅ 伙伴授权可训练(C-1) | 27.9 | 10122 |
|
| 50 |
+
| fallout4-steam-release-test | ⏸ 待 C-1 扩展(微软系) | 2.7 | 964 |
|
| 51 |
+
| wolfenstein2-steam | ⏸ 待 C-1 扩展(微软系) | 0.2 | 67 |
|
| 52 |
+
| atomic-heart-pc | 🚫 三方待授权 | 0.1 | 39 |
|
| 53 |
+
| resident-evil-village-pc | 🚫 三方待授权 | 0.1 | 35 |
|
| 54 |
+
| doom-eternal-pc | ⏸ 待 C-1 扩展(微软系) | 0.1 | 33 |
|
| 55 |
+
| deathloop-pc | ⏸ 待 C-1 扩展(微软系) | 0.1 | 33 |
|
| 56 |
+
| starfield-pc | ⏸ 待 C-1 扩展(微软系) | 0.1 | 29 |
|
| 57 |
+
| kingdom-come-deliverance-2-steam | 🚫 三方待授权 | 0.1 | 5 |
|
| 58 |
+
| _missing_target | ? | 0.0 | 1 |
|
| 59 |
+
| wuchang-fallen-feathers-steam | 🚫 三方待授权 | 0.0 | 7 |
|
| 60 |
+
| call-of-duty-steam-release-test | ⏸ 待 C-1 扩展(微软系) | 0.0 | 6 |
|
| 61 |
+
|
| 62 |
+
- 引擎深度交付(verified/):**4 包**,最新一包 2026-08-08(距今 **70 天**)
|
| 63 |
+
|
| 64 |
+
### 各线口径
|
| 65 |
+
|
| 66 |
+
- **WM 世界模型线**:halo_partner 采集 + 引擎深度交付包。训练代码随 SG-1 放行; 过闸小时以 halo_qc floor 档普查口径引用
|
| 67 |
+
- **IDM 线**:第一方采集全量带键鼠事件流(可训练性随权利裁决)。方法验证走 D2E; 硬门未过, 先收敛现有语料再买素材(ml/README)
|
| 68 |
+
- **VL 线**:任意第一方 RGB 帧 + gaming-500 bench。游戏识别 SOTA 代理; cohort 数据到位后重新定基
|
| 69 |
+
|
| 70 |
+
### 外部语料(全部禁训)
|
| 71 |
+
|
| 72 |
+
> 静态常量, 不查桶——每行自带截止日, 过期一眼看得见。
|
| 73 |
+
|
| 74 |
+
| 语料 | 规模 | 状态 | 用途 | 数字截止 |
|
| 75 |
+
|---|---|---|---|---|
|
| 76 |
+
| D2E-480p | 268.7 h / 29 款 / 原生 raw HID delta | 禁训·方法验证 | IDM 线(zero-shot 止损复核, research/28) | 2026-08-13 |
|
| 77 |
+
| gaming-500-hours | ~500 h / 168 款(HF, 无 license) | 禁训·方法验证 | VL bench 真值 + IDM 早期验证 | 2026-08-13 |
|
| 78 |
+
| VPT / CS:GO / GF-Minecraft / D2E 的 R2 镜像 | 已删除(实删 118,773 对象 / 7.82 TB) | 禁训·**镜像已清** | 2026-09-04 清桶完毕, 仅存 vpt-schema-sample;VPT 从此不可复得(research/50 R6) | 2026-09-04 |
|
| 79 |
+
|
| 80 |
+
---
|
| 81 |
+
*生成于 workflow dataminer-weekly-report · 脚本 @be5be49 · 报告 / state / 台账存 R2 `reports/weekly/` · console 总览 https://console.diogenesmodel.com/overview*
|
closeout-final/weekly/dm-weekly-2026-10-24.md
ADDED
|
@@ -0,0 +1,80 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# DataMiner 数据周报 · 2026-10-17 ~ 2026-10-24
|
| 2 |
+
|
| 3 |
+
> schema dm-weekly-report/1 · 窗口按对象上传时间(Asia/Shanghai)· 健康判定同 docs/146(console 口径)· 累计 = R2 存量台账(docs/168)
|
| 4 |
+
> 采集员维度看 console:[按采集员](https://console.diogenesmodel.com/output?from=2026-10-17&to=2026-10-23&bucket=day&timeBasis=recorded&groupBy=annotator) · [按游戏](https://console.diogenesmodel.com/output?from=2026-10-17&to=2026-10-23&bucket=day&timeBasis=recorded&groupBy=game)(同窗口, 录制口径)
|
| 5 |
+
|
| 6 |
+
## 一、本周采集与质检
|
| 7 |
+
|
| 8 |
+
- 总量:**25.97 h** / 132 会话 / 9444 clips(上传口径)
|
| 9 |
+
- 录制口径(估算, console 每日进线同族):**20.94 h** / 7615 clips;与上传口径差 -5.04 h(-19.4%)
|
| 10 |
+
- 存量台账:本期入账 7793 · 出账(R2 已删)0 · 现有 219293 clips
|
| 11 |
+
- 健康异常:**56**(0.6%)——轻度补帧 54 · 重度(>20%) 1 · **整段冻结 1**
|
| 12 |
+
- 音频:缺失 0 · 静音怀疑 5
|
| 13 |
+
- session 验收报告覆盖(两条源车道全量):3309/3309(100.0%)
|
| 14 |
+
|
| 15 |
+
| 游戏 | 会话 | clips | 小时 | 异常 | 异常率 | 无音频 | 静音疑 |
|
| 16 |
+
|---|---|---|---|---|---|---|---|
|
| 17 |
+
| red-dead-redemption-2-pc | 40 | 3565 | 9.84 | 11 | 0.3% | 0 | 4 |
|
| 18 |
+
| gta5-legacy-pc | 44 | 3045 | 8.38 | 10 | 0.3% | 0 | 1 |
|
| 19 |
+
| far-cry-5-pc | 48 | 2834 | 7.75 | 35 | 1.2% | 0 | 0 |
|
| 20 |
+
|
| 21 |
+
### 逐日
|
| 22 |
+
|
| 23 |
+
| 日期 | clips | 小时 | 异常 |
|
| 24 |
+
|---|---|---|---|
|
| 25 |
+
| 10-17 | 1829 | 5.04 | 9 |
|
| 26 |
+
| 10-19 | 4971 | 13.66 | 36 |
|
| 27 |
+
| 10-20 | 2351 | 6.47 | 11 |
|
| 28 |
+
| 10-21 | 293 | 0.81 | 0 |
|
| 29 |
+
|
| 30 |
+
### 质检 findings
|
| 31 |
+
|
| 32 |
+
- 🚨 本窗 9444 clips, 上窗 28954, 环比 -67%(阈值 -50%)——先确认是真降产还是又漏了车道
|
| 33 |
+
- 六件套不完整 ×1: 9aea1818(b00bffb6) 缺 ['events', 'meta', 'manifest']
|
| 34 |
+
- 整段冻结 clip(dup≥100%)×1: 37f20e84(20995044)
|
| 35 |
+
|
| 36 |
+
## 二、ML 各线可用数据(累计)
|
| 37 |
+
|
| 38 |
+
> 口径:R2 存量台账(有则入账、删则出账),与 console「游戏分布」面板(docs/153)同为存量读数。
|
| 39 |
+
> 「授权档」≠ 可训练:质量闸放行数以 console 回看为准(docs/167 §2.3 实测 0 条 passed)。
|
| 40 |
+
|
| 41 |
+
| 游戏 | 授权档 | 累计小时 | 累计 clips |
|
| 42 |
+
|---|---|---|---|
|
| 43 |
+
| cyberpunk-2077-steam | 🚫 三方待授权 | 202.1 | 73468 |
|
| 44 |
+
| red-dead-redemption-2-pc | 🚫 三方待授权 | 153.9 | 55834 |
|
| 45 |
+
| gta5-legacy-pc | 🚫 三方待授权 | 112.9 | 41012 |
|
| 46 |
+
| dying-light-pc | 🚫 三方待授权 | 56.4 | 20511 |
|
| 47 |
+
| far-cry-5-pc | 🚫 三方待授权 | 46.9 | 17127 |
|
| 48 |
+
| halo-infinite-steam-release-test | ✅ 伙伴授权可训练(C-1) | 27.9 | 10122 |
|
| 49 |
+
| fallout4-steam-release-test | ⏸ 待 C-1 扩展(微软系) | 2.7 | 964 |
|
| 50 |
+
| wolfenstein2-steam | ⏸ 待 C-1 扩展(微软系) | 0.2 | 67 |
|
| 51 |
+
| atomic-heart-pc | 🚫 三方待授权 | 0.1 | 39 |
|
| 52 |
+
| resident-evil-village-pc | 🚫 三方待授权 | 0.1 | 35 |
|
| 53 |
+
| doom-eternal-pc | ⏸ 待 C-1 扩展(微软系) | 0.1 | 33 |
|
| 54 |
+
| deathloop-pc | ⏸ 待 C-1 扩展(微软系) | 0.1 | 33 |
|
| 55 |
+
| starfield-pc | ⏸ 待 C-1 扩展(微软系) | 0.1 | 29 |
|
| 56 |
+
| kingdom-come-deliverance-2-steam | 🚫 三方待授权 | 0.1 | 5 |
|
| 57 |
+
| _missing_target | ? | 0.0 | 1 |
|
| 58 |
+
| wuchang-fallen-feathers-steam | 🚫 三方待授权 | 0.0 | 7 |
|
| 59 |
+
| call-of-duty-steam-release-test | ⏸ 待 C-1 扩展(微软系) | 0.0 | 6 |
|
| 60 |
+
|
| 61 |
+
- 引擎深度交付(verified/):**4 包**,最新一包 2026-08-08(距今 **77 天**)
|
| 62 |
+
|
| 63 |
+
### 各线口径
|
| 64 |
+
|
| 65 |
+
- **WM 世界模型线**:halo_partner 采集 + 引擎深度交付包。训练代码随 SG-1 放行; 过闸小时以 halo_qc floor 档普查口径引用
|
| 66 |
+
- **IDM 线**:第一方采集全量带键鼠事件流(可训练性随权利裁决)。方法验证走 D2E; 硬门未过, 先收敛现有语料再买素材(ml/README)
|
| 67 |
+
- **VL 线**:任意第一方 RGB 帧 + gaming-500 bench。游戏识别 SOTA 代理; cohort 数据到位后重新定基
|
| 68 |
+
|
| 69 |
+
### 外部语料(全部禁训)
|
| 70 |
+
|
| 71 |
+
> 静态常量, 不查桶——每行自带截止日, 过期一眼看得见。
|
| 72 |
+
|
| 73 |
+
| 语料 | 规模 | 状态 | 用途 | 数字截止 |
|
| 74 |
+
|---|---|---|---|---|
|
| 75 |
+
| D2E-480p | 268.7 h / 29 款 / 原生 raw HID delta | 禁训·方法验证 | IDM 线(zero-shot 止损复核, research/28) | 2026-08-13 |
|
| 76 |
+
| gaming-500-hours | ~500 h / 168 款(HF, 无 license) | 禁训·方法验证 | VL bench 真值 + IDM 早期验证 | 2026-08-13 |
|
| 77 |
+
| VPT / CS:GO / GF-Minecraft / D2E 的 R2 镜像 | 已删除(实删 118,773 对象 / 7.82 TB) | 禁训·**镜像已清** | 2026-09-04 清桶完毕, 仅存 vpt-schema-sample;VPT 从此不可复得(research/50 R6) | 2026-09-04 |
|
| 78 |
+
|
| 79 |
+
---
|
| 80 |
+
*生成于 workflow dataminer-weekly-report · 脚本 @9d532a8 · 报告 / state / 台账存 R2 `reports/weekly/` · console 总览 https://console.diogenesmodel.com/overview*
|
closeout-final/weekly/dm-weekly-2026-10-31.md
ADDED
|
@@ -0,0 +1,72 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# DataMiner 数据周报 · 2026-10-24 ~ 2026-10-31
|
| 2 |
+
|
| 3 |
+
> schema dm-weekly-report/1 · 窗口按对象上传时间(Asia/Shanghai)· 健康判定同 docs/146(console 口径)· 累计 = R2 存量台账(docs/168)
|
| 4 |
+
> 采集员维度看 console:[按采集员](https://console.diogenesmodel.com/output?from=2026-10-24&to=2026-10-30&bucket=day&timeBasis=recorded&groupBy=annotator) · [按游戏](https://console.diogenesmodel.com/output?from=2026-10-24&to=2026-10-30&bucket=day&timeBasis=recorded&groupBy=game)(同窗口, 录制口径)
|
| 5 |
+
|
| 6 |
+
## 一、本周采集与质检
|
| 7 |
+
|
| 8 |
+
- 总量:**0.00 h** / 0 会话 / 0 clips(上传口径)
|
| 9 |
+
- 录制口径(估算, console 每日进线同族):**0.00 h** / 0 clips;与上传口径差 +0.00 h(—)
|
| 10 |
+
- 存量台账:本期入账 0 · 出账(R2 已删)0 · 现有 219293 clips
|
| 11 |
+
- 健康异常:**0**(—)——轻度补帧 0 · 重度(>20%) 0 · **整段冻结 0**
|
| 12 |
+
- 音频:缺失 0 · 静音怀疑 0
|
| 13 |
+
- session 验收报告覆盖(两条源车道全量):3309/3309(100.0%)
|
| 14 |
+
|
| 15 |
+
| 游戏 | 会话 | clips | 小时 | 异常 | 异常率 | 无音频 | 静音疑 |
|
| 16 |
+
|---|---|---|---|---|---|---|---|
|
| 17 |
+
|
| 18 |
+
### 逐日
|
| 19 |
+
|
| 20 |
+
| 日期 | clips | 小时 | 异常 |
|
| 21 |
+
|---|---|---|---|
|
| 22 |
+
|
| 23 |
+
### 质检 findings
|
| 24 |
+
|
| 25 |
+
- 🚨 源车道 `raw/v2/dataminer-api-production` 本窗 0 个新对象——真停采, 或取数范围又跟丢了车道
|
| 26 |
+
- 🚨 本窗 0 clips, 上窗 9444, 环比 -100%(阈值 -50%)——先确认是真降产还是又漏了车道
|
| 27 |
+
|
| 28 |
+
## 二、ML 各线可用数据(累计)
|
| 29 |
+
|
| 30 |
+
> 口径:R2 存量台账(有则入账、删则出账),与 console「游戏分布」面板(docs/153)同为存量读数。
|
| 31 |
+
> 「授权档」≠ 可训练:质量闸放行数以 console 回看为准(docs/167 §2.3 实测 0 条 passed)。
|
| 32 |
+
|
| 33 |
+
| 游戏 | 授权档 | 累计小时 | 累计 clips |
|
| 34 |
+
|---|---|---|---|
|
| 35 |
+
| cyberpunk-2077-steam | 🚫 三方待授权 | 202.1 | 73468 |
|
| 36 |
+
| red-dead-redemption-2-pc | 🚫 三方待授权 | 153.9 | 55834 |
|
| 37 |
+
| gta5-legacy-pc | 🚫 三方待授权 | 112.9 | 41012 |
|
| 38 |
+
| dying-light-pc | 🚫 三方待授权 | 56.4 | 20511 |
|
| 39 |
+
| far-cry-5-pc | 🚫 三方待授权 | 46.9 | 17127 |
|
| 40 |
+
| halo-infinite-steam-release-test | ✅ 伙伴授权可训练(C-1) | 27.9 | 10122 |
|
| 41 |
+
| fallout4-steam-release-test | ⏸ 待 C-1 扩展(微软系) | 2.7 | 964 |
|
| 42 |
+
| wolfenstein2-steam | ⏸ 待 C-1 扩展(微软系) | 0.2 | 67 |
|
| 43 |
+
| atomic-heart-pc | 🚫 三方待授权 | 0.1 | 39 |
|
| 44 |
+
| resident-evil-village-pc | 🚫 三方待授权 | 0.1 | 35 |
|
| 45 |
+
| doom-eternal-pc | ⏸ 待 C-1 扩展(微软系) | 0.1 | 33 |
|
| 46 |
+
| deathloop-pc | ⏸ 待 C-1 扩展(微软系) | 0.1 | 33 |
|
| 47 |
+
| starfield-pc | ⏸ 待 C-1 扩展(微软系) | 0.1 | 29 |
|
| 48 |
+
| kingdom-come-deliverance-2-steam | 🚫 三方待授权 | 0.1 | 5 |
|
| 49 |
+
| _missing_target | ? | 0.0 | 1 |
|
| 50 |
+
| wuchang-fallen-feathers-steam | 🚫 三方待授权 | 0.0 | 7 |
|
| 51 |
+
| call-of-duty-steam-release-test | ⏸ 待 C-1 扩展(微软系) | 0.0 | 6 |
|
| 52 |
+
|
| 53 |
+
- 引擎深度交付(verified/):**4 包**,最新一包 2026-08-08(距今 **84 天**)
|
| 54 |
+
|
| 55 |
+
### 各线口径
|
| 56 |
+
|
| 57 |
+
- **WM 世界模型线**:halo_partner 采集 + 引擎深度交付包。训练代码随 SG-1 放行; 过闸小时以 halo_qc floor 档普查口径引用
|
| 58 |
+
- **IDM 线**:第一方采集全量带键鼠事件流(可训练性随权利裁决)。方法验证走 D2E; 硬门未过, 先收敛现有语料再买素材(ml/README)
|
| 59 |
+
- **VL 线**:任意第一方 RGB 帧 + gaming-500 bench。游戏识别 SOTA 代理; cohort 数据到位后重新定基
|
| 60 |
+
|
| 61 |
+
### 外部语料(全部禁训)
|
| 62 |
+
|
| 63 |
+
> 静态常量, 不查桶——每行自带截止日, 过期一眼看得见。
|
| 64 |
+
|
| 65 |
+
| 语料 | 规模 | 状态 | 用途 | 数字截止 |
|
| 66 |
+
|---|---|---|---|---|
|
| 67 |
+
| D2E-480p | 268.7 h / 29 款 / 原生 raw HID delta | 禁训·方法验证 | IDM 线(zero-shot 止损复核, research/28) | 2026-08-13 |
|
| 68 |
+
| gaming-500-hours | ~500 h / 168 款(HF, 无 license) | 禁训·方法验证 | VL bench 真值 + IDM 早期验证 | 2026-08-13 |
|
| 69 |
+
| VPT / CS:GO / GF-Minecraft / D2E 的 R2 镜像 | 已删除(实删 118,773 对象 / 7.82 TB) | 禁训·**镜像已清** | 2026-09-04 清桶完毕, 仅存 vpt-schema-sample;VPT 从此不可复得(research/50 R6) | 2026-09-04 |
|
| 70 |
+
|
| 71 |
+
---
|
| 72 |
+
*生成于 workflow dataminer-weekly-report · 脚本 @9d532a8 · 报告 / state / 台账存 R2 `reports/weekly/` · console 总览 https://console.diogenesmodel.com/overview*
|
closeout-final/weekly/dm-weekly-2026-11-07.md
ADDED
|
@@ -0,0 +1,71 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# DataMiner 数据周报 · 2026-10-31 ~ 2026-11-07
|
| 2 |
+
|
| 3 |
+
> schema dm-weekly-report/1 · 窗口按对象上传时间(Asia/Shanghai)· 健康判定同 docs/146(console 口径)· 累计 = R2 存量台账(docs/168)
|
| 4 |
+
> 采集员维度看 console:[按采集员](https://console.diogenesmodel.com/output?from=2026-10-31&to=2026-11-06&bucket=day&timeBasis=recorded&groupBy=annotator) · [按游戏](https://console.diogenesmodel.com/output?from=2026-10-31&to=2026-11-06&bucket=day&timeBasis=recorded&groupBy=game)(同窗口, 录制口径)
|
| 5 |
+
|
| 6 |
+
## 一、本周采集与质检
|
| 7 |
+
|
| 8 |
+
- 总量:**0.00 h** / 0 会话 / 0 clips(上传口径)
|
| 9 |
+
- 录制口径(估算, console 每日进线同族):**0.00 h** / 0 clips;与上传口径差 +0.00 h(—)
|
| 10 |
+
- 存量台账:本期入账 0 · 出账(R2 已删)0 · 现有 219293 clips
|
| 11 |
+
- 健康异常:**0**(—)——轻度补帧 0 · 重度(>20%) 0 · **整段冻结 0**
|
| 12 |
+
- 音频:缺失 0 · 静音怀疑 0
|
| 13 |
+
- session 验收报告覆盖(两条源车道全量):3309/3309(100.0%)
|
| 14 |
+
|
| 15 |
+
| 游戏 | 会话 | clips | 小时 | 异常 | 异常率 | 无音频 | 静音疑 |
|
| 16 |
+
|---|---|---|---|---|---|---|---|
|
| 17 |
+
|
| 18 |
+
### 逐日
|
| 19 |
+
|
| 20 |
+
| 日期 | clips | 小时 | 异常 |
|
| 21 |
+
|---|---|---|---|
|
| 22 |
+
|
| 23 |
+
### 质检 findings
|
| 24 |
+
|
| 25 |
+
- 🚨 源车道 `raw/v2/dataminer-api-production` 本窗 0 个新对象——真停采, 或取数范围又跟丢了车道
|
| 26 |
+
|
| 27 |
+
## 二、ML 各线可用数据(累计)
|
| 28 |
+
|
| 29 |
+
> 口径:R2 存量台账(有则入账、删则出账),与 console「游戏分布」面板(docs/153)同为存量读数。
|
| 30 |
+
> 「授权档」≠ 可训练:质量闸放行数以 console 回看为准(docs/167 §2.3 实测 0 条 passed)。
|
| 31 |
+
|
| 32 |
+
| 游戏 | 授权档 | 累计小时 | 累计 clips |
|
| 33 |
+
|---|---|---|---|
|
| 34 |
+
| cyberpunk-2077-steam | 🚫 三方待授权 | 202.1 | 73468 |
|
| 35 |
+
| red-dead-redemption-2-pc | 🚫 三方待授权 | 153.9 | 55834 |
|
| 36 |
+
| gta5-legacy-pc | 🚫 三方待授权 | 112.9 | 41012 |
|
| 37 |
+
| dying-light-pc | 🚫 三方待授权 | 56.4 | 20511 |
|
| 38 |
+
| far-cry-5-pc | 🚫 三方待授权 | 46.9 | 17127 |
|
| 39 |
+
| halo-infinite-steam-release-test | ✅ 伙伴授权可训练(C-1) | 27.9 | 10122 |
|
| 40 |
+
| fallout4-steam-release-test | ⏸ 待 C-1 扩展(微软系) | 2.7 | 964 |
|
| 41 |
+
| wolfenstein2-steam | ⏸ 待 C-1 扩展(微软系) | 0.2 | 67 |
|
| 42 |
+
| atomic-heart-pc | 🚫 三方待授权 | 0.1 | 39 |
|
| 43 |
+
| resident-evil-village-pc | 🚫 三方待授权 | 0.1 | 35 |
|
| 44 |
+
| doom-eternal-pc | ⏸ 待 C-1 扩展(微软系) | 0.1 | 33 |
|
| 45 |
+
| deathloop-pc | ⏸ 待 C-1 扩展(微软系) | 0.1 | 33 |
|
| 46 |
+
| starfield-pc | ⏸ 待 C-1 扩展(微软系) | 0.1 | 29 |
|
| 47 |
+
| kingdom-come-deliverance-2-steam | 🚫 三方待授权 | 0.1 | 5 |
|
| 48 |
+
| _missing_target | ? | 0.0 | 1 |
|
| 49 |
+
| wuchang-fallen-feathers-steam | 🚫 三方待授权 | 0.0 | 7 |
|
| 50 |
+
| call-of-duty-steam-release-test | ⏸ 待 C-1 扩展(微软系) | 0.0 | 6 |
|
| 51 |
+
|
| 52 |
+
- 引擎深度交付(verified/):**4 包**,最新一包 2026-08-08(距今 **91 天**)
|
| 53 |
+
|
| 54 |
+
### 各线口径
|
| 55 |
+
|
| 56 |
+
- **WM 世界模型线**:halo_partner 采集 + 引擎深度交付包。训练代码随 SG-1 放行; 过闸小时以 halo_qc floor 档普查口径引用
|
| 57 |
+
- **IDM 线**:第一方采集全量带键鼠事件流(可训练性随权利裁决)。方法验证走 D2E; 硬门未过, 先收敛现有语料再买素材(ml/README)
|
| 58 |
+
- **VL 线**:任意第一方 RGB 帧 + gaming-500 bench。游戏识别 SOTA 代理; cohort 数据到位后重新定基
|
| 59 |
+
|
| 60 |
+
### 外部语料(全部禁训)
|
| 61 |
+
|
| 62 |
+
> 静态常量, 不查桶——每行自带截止日, 过期一眼看得见。
|
| 63 |
+
|
| 64 |
+
| 语料 | 规模 | 状态 | 用途 | 数字截止 |
|
| 65 |
+
|---|---|---|---|---|
|
| 66 |
+
| D2E-480p | 268.7 h / 29 款 / 原生 raw HID delta | 禁训·方法验证 | IDM 线(zero-shot 止损复核, research/28) | 2026-08-13 |
|
| 67 |
+
| gaming-500-hours | ~500 h / 168 款(HF, 无 license) | 禁训·方法验证 | VL bench 真值 + IDM 早期验证 | 2026-08-13 |
|
| 68 |
+
| VPT / CS:GO / GF-Minecraft / D2E 的 R2 镜像 | 已删除(实删 118,773 对象 / 7.82 TB) | 禁训·**镜像已清** | 2026-09-04 清桶完毕, 仅存 vpt-schema-sample;VPT 从此不可复得(research/50 R6) | 2026-09-04 |
|
| 69 |
+
|
| 70 |
+
---
|
| 71 |
+
*生成于 workflow dataminer-weekly-report · 脚本 @9d532a8 · 报告 / state / 台账存 R2 `reports/weekly/` · console 总览 https://console.diogenesmodel.com/overview*
|
closeout-final/weekly/dm-weekly-2026-11-14.md
ADDED
|
@@ -0,0 +1,71 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# DataMiner 数据周报 · 2026-11-07 ~ 2026-11-14
|
| 2 |
+
|
| 3 |
+
> schema dm-weekly-report/1 · 窗口按对象上传时间(Asia/Shanghai)· 健康判定同 docs/146(console 口径)· 累计 = R2 存量台账(docs/168)
|
| 4 |
+
> 采集员维度看 console:[按采集员](https://console.diogenesmodel.com/output?from=2026-11-07&to=2026-11-13&bucket=day&timeBasis=recorded&groupBy=annotator) · [按游戏](https://console.diogenesmodel.com/output?from=2026-11-07&to=2026-11-13&bucket=day&timeBasis=recorded&groupBy=game)(同窗口, 录制口径)
|
| 5 |
+
|
| 6 |
+
## 一、本周采集与质检
|
| 7 |
+
|
| 8 |
+
- 总量:**0.00 h** / 0 会话 / 0 clips(上传口径)
|
| 9 |
+
- 录制口径(估算, console 每日进线同族):**0.00 h** / 0 clips;与上传口径差 +0.00 h(—)
|
| 10 |
+
- 存量台账:本期入账 0 · 出账(R2 已删)0 · 现有 219293 clips
|
| 11 |
+
- 健康异常:**0**(—)——轻度补帧 0 · 重度(>20%) 0 · **整段冻结 0**
|
| 12 |
+
- 音频:缺失 0 · 静音怀疑 0
|
| 13 |
+
- session 验收报告覆盖(两条源车道全量):3309/3309(100.0%)
|
| 14 |
+
|
| 15 |
+
| 游戏 | 会话 | clips | 小时 | 异常 | 异常率 | 无音频 | 静音疑 |
|
| 16 |
+
|---|---|---|---|---|---|---|---|
|
| 17 |
+
|
| 18 |
+
### 逐日
|
| 19 |
+
|
| 20 |
+
| 日期 | clips | 小时 | 异常 |
|
| 21 |
+
|---|---|---|---|
|
| 22 |
+
|
| 23 |
+
### 质检 findings
|
| 24 |
+
|
| 25 |
+
- 🚨 源车道 `raw/v2/dataminer-api-production` 本窗 0 个新对象——真停采, 或取数范围又跟丢了车道
|
| 26 |
+
|
| 27 |
+
## 二、ML 各线可用数据(累计)
|
| 28 |
+
|
| 29 |
+
> 口径:R2 存量台账(有则入账、删则出账),与 console「游戏分布」面板(docs/153)同为存量读数。
|
| 30 |
+
> 「授权档」≠ 可训练:质量闸放行数以 console 回看为准(docs/167 §2.3 实测 0 条 passed)。
|
| 31 |
+
|
| 32 |
+
| 游戏 | 授权档 | 累计小时 | 累计 clips |
|
| 33 |
+
|---|---|---|---|
|
| 34 |
+
| cyberpunk-2077-steam | 🚫 三方待授权 | 202.1 | 73468 |
|
| 35 |
+
| red-dead-redemption-2-pc | 🚫 三方待授权 | 153.9 | 55834 |
|
| 36 |
+
| gta5-legacy-pc | 🚫 三方待授权 | 112.9 | 41012 |
|
| 37 |
+
| dying-light-pc | 🚫 三方待授权 | 56.4 | 20511 |
|
| 38 |
+
| far-cry-5-pc | 🚫 三方待授权 | 46.9 | 17127 |
|
| 39 |
+
| halo-infinite-steam-release-test | ✅ 伙伴授权可训练(C-1) | 27.9 | 10122 |
|
| 40 |
+
| fallout4-steam-release-test | ⏸ 待 C-1 扩展(微软系) | 2.7 | 964 |
|
| 41 |
+
| wolfenstein2-steam | ⏸ 待 C-1 扩展(微软系) | 0.2 | 67 |
|
| 42 |
+
| atomic-heart-pc | 🚫 三方待授权 | 0.1 | 39 |
|
| 43 |
+
| resident-evil-village-pc | 🚫 三方待授权 | 0.1 | 35 |
|
| 44 |
+
| doom-eternal-pc | ⏸ 待 C-1 扩展(微软系) | 0.1 | 33 |
|
| 45 |
+
| deathloop-pc | ⏸ 待 C-1 扩展(微软系) | 0.1 | 33 |
|
| 46 |
+
| starfield-pc | ⏸ 待 C-1 扩展(微软系) | 0.1 | 29 |
|
| 47 |
+
| kingdom-come-deliverance-2-steam | 🚫 三方待授权 | 0.1 | 5 |
|
| 48 |
+
| _missing_target | ? | 0.0 | 1 |
|
| 49 |
+
| wuchang-fallen-feathers-steam | 🚫 三方待授权 | 0.0 | 7 |
|
| 50 |
+
| call-of-duty-steam-release-test | ⏸ 待 C-1 扩展(微软系) | 0.0 | 6 |
|
| 51 |
+
|
| 52 |
+
- 引擎深度交付(verified/):**4 包**,最新一包 2026-08-08(距今 **98 天**)
|
| 53 |
+
|
| 54 |
+
### 各线口径
|
| 55 |
+
|
| 56 |
+
- **WM 世界模型线**:halo_partner 采集 + 引擎深度交付包。训练代码随 SG-1 放行; 过闸小时以 halo_qc floor 档普查口径引用
|
| 57 |
+
- **IDM 线**:第一方采集全量带键鼠事件流(可训练性随权利裁决)。方法验证走 D2E; 硬门未过, 先收敛现有语料再买素材(ml/README)
|
| 58 |
+
- **VL 线**:任意第一方 RGB 帧 + gaming-500 bench。游戏识别 SOTA 代理; cohort 数据到位后重新定基
|
| 59 |
+
|
| 60 |
+
### 外部语料(全部禁训)
|
| 61 |
+
|
| 62 |
+
> 静态常量, 不查桶——每行自带截止日, 过期一眼看得见。
|
| 63 |
+
|
| 64 |
+
| 语料 | 规模 | 状态 | 用途 | 数字截止 |
|
| 65 |
+
|---|---|---|---|---|
|
| 66 |
+
| D2E-480p | 268.7 h / 29 款 / 原生 raw HID delta | 禁训·方法验证 | IDM 线(zero-shot 止损复核, research/28) | 2026-08-13 |
|
| 67 |
+
| gaming-500-hours | ~500 h / 168 款(HF, 无 license) | 禁训·方法验证 | VL bench 真值 + IDM 早期验证 | 2026-08-13 |
|
| 68 |
+
| VPT / CS:GO / GF-Minecraft / D2E 的 R2 镜像 | 已删除(实删 118,773 对象 / 7.82 TB) | 禁训·**镜像已清** | 2026-09-04 清桶完毕, 仅存 vpt-schema-sample;VPT 从此不可复得(research/50 R6) | 2026-09-04 |
|
| 69 |
+
|
| 70 |
+
---
|
| 71 |
+
*生成于 workflow dataminer-weekly-report · 脚本 @9d532a8 · 报告 / state / 台账存 R2 `reports/weekly/` · console 总览 https://console.diogenesmodel.com/overview*
|
closeout-final/weekly/ledger.json.gz
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:703f25394b56f382f9b36d0f6a2b7f4f5337ad5d9eae1921990f547f09038135
|
| 3 |
+
size 6173867
|
closeout-final/weekly/state.json
ADDED
|
@@ -0,0 +1,49 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"schema": "dm-weekly-state/1",
|
| 3 |
+
"seeded_from": "dataminer_production DB, recorded < 2026-08-22 00:00 Asia/Shanghai (docs/152 §3.3)",
|
| 4 |
+
"cumulative_hours": {
|
| 5 |
+
"cyberpunk-2077-steam": 202.13,
|
| 6 |
+
"halo-infinite-steam-release-test": 27.89,
|
| 7 |
+
"fallout4-steam-release-test": 2.73,
|
| 8 |
+
"call-of-duty-steam-release-test": 0.01,
|
| 9 |
+
"starfield-pc": 0.08,
|
| 10 |
+
"wuchang-fallen-feathers-steam": 0.02,
|
| 11 |
+
"red-dead-redemption-2-pc": 153.9,
|
| 12 |
+
"_missing_target": 0.03,
|
| 13 |
+
"doom-eternal-pc": 0.09,
|
| 14 |
+
"wolfenstein2-steam": 0.18,
|
| 15 |
+
"resident-evil-village-pc": 0.1,
|
| 16 |
+
"atomic-heart-pc": 0.11,
|
| 17 |
+
"deathloop-pc": 0.09,
|
| 18 |
+
"gta5-legacy-pc": 112.87,
|
| 19 |
+
"dying-light-pc": 56.44,
|
| 20 |
+
"kingdom-come-deliverance-2-steam": 0.06,
|
| 21 |
+
"far-cry-5-pc": 46.94
|
| 22 |
+
},
|
| 23 |
+
"cumulative_clips": {
|
| 24 |
+
"cyberpunk-2077-steam": 73468,
|
| 25 |
+
"halo-infinite-steam-release-test": 10122,
|
| 26 |
+
"fallout4-steam-release-test": 964,
|
| 27 |
+
"call-of-duty-steam-release-test": 6,
|
| 28 |
+
"starfield-pc": 29,
|
| 29 |
+
"wuchang-fallen-feathers-steam": 7,
|
| 30 |
+
"red-dead-redemption-2-pc": 55834,
|
| 31 |
+
"_missing_target": 1,
|
| 32 |
+
"doom-eternal-pc": 33,
|
| 33 |
+
"wolfenstein2-steam": 67,
|
| 34 |
+
"resident-evil-village-pc": 35,
|
| 35 |
+
"atomic-heart-pc": 39,
|
| 36 |
+
"deathloop-pc": 33,
|
| 37 |
+
"gta5-legacy-pc": 41012,
|
| 38 |
+
"dying-light-pc": 20511,
|
| 39 |
+
"kingdom-come-deliverance-2-steam": 5,
|
| 40 |
+
"far-cry-5-pc": 17127
|
| 41 |
+
},
|
| 42 |
+
"last_window_end": "2026-11-14",
|
| 43 |
+
"cumulative_source": "ledger",
|
| 44 |
+
"last_window": {
|
| 45 |
+
"clips": 0,
|
| 46 |
+
"hours": 0,
|
| 47 |
+
"accepted": 3309
|
| 48 |
+
}
|
| 49 |
+
}
|