Skip to content

wanghai-tech / hikoon-ACR

Go to a project
Toggle navigation
Toggle navigation pinning
Commit 9de8092d 9de8092d5982e71e52d18fd3247b5eb11d2ca4eb by cnb.bofCdSsphPA
Options

Freeze the production encoder before scaling the music index 

Document the production decision to stabilize the embedding space before onboarding a 300k-song catalog, and record the migration rules for future encoder upgrades.

Constraint: 300k-song production rollout makes embedding churn expensive and risky
Rejected: keep iterating encoder before defining a production embedding version | would force repeated full-vector rebuilds and unstable rollout criteria
Confidence: high
Scope-risk: narrow
Directive: Treat encoder changes as versioned index migrations, not in-place model swaps
Tested: reviewed rendered markdown content, docs index link, changelog entry, and git diff for the three touched docs
Not-tested: git push / remote sync outcome depends on repository remote state
1 parent 73d28fae
Expand all
Inline Side-by-side
Showing 3 changed files with 21 additions and 0 deletions
1 ### Stage: production encoder freeze FAQ and rollout guidance
2
3 完成项：
4 - 新增文档：
5 - `docs/production-encoder-freeze-and-embedding-strategy.md`
6 - 文档内容覆盖：
7 - 为什么当前应先冻结 encoder
8 - 当前结构的泛化能力边界
9 - 外置模型权重后如何给其他歌曲直接使用
10 - wav/mp3/flac/ogg 集合如何快速进入 manifest -> build-index -> evaluate 链路
11 - 30 万首生产曲库下 embedding/version/index 的治理建议
12 - encoder 升级后哪些数据必须重建、哪些元数据可以保留
13 - docs 入口已补充：
14 - `docs/README.md` 新增该答疑文档链接
15
16 结论：
17 - 当前阶段先冻结 `encoder v1` 是更稳妥的生产决策
18 - 生产环境应把模型文件、embedding 版本、reference 索引与评测报告解耦管理
19 - 后续新 encoder 应走“离线 shadow build -> A/B -> 切换”的升级路径，而不是直接覆盖旧 embedding 库
20
1 ## 2026-06-02 16:11 UTC / hum_guard fresh eval did not beat hum_focus 21 ## 2026-06-02 16:11 UTC / hum_guard fresh eval did not beat hum_focus
2 22
3 - 对 `/tmp/dualaxis_sweep/hum_guard/eval.json` 做了最新复核 23 - 对 `/tmp/dualaxis_sweep/hum_guard/eval.json` 做了最新复核
......
...@@ -91,6 +91,7 @@ cd /workspace/acr-engine ...@@ -91,6 +91,7 @@ cd /workspace/acr-engine
91 - [数据规范](./dataset-spec.md) 91 - [数据规范](./dataset-spec.md)
92 - [开放数据工作流](./open-dataset-workflow.md) 92 - [开放数据工作流](./open-dataset-workflow.md)
93 - [训练数据与 pgvector 指南](./training-data-and-pgvector-guide.md) 93 - [训练数据与 pgvector 指南](./training-data-and-pgvector-guide.md)
94 - [生产 Encoder 冻结与 Embedding 策略答疑](./production-encoder-freeze-and-embedding-strategy.md)
94 - [数据来源与接入](./dataset-sources-and-licensing.md) 95 - [数据来源与接入](./dataset-sources-and-licensing.md)
95 - [工业评测规范](./industrial-benchmark-spec.md) 96 - [工业评测规范](./industrial-benchmark-spec.md)
96 97
......