Evaluate the existing BERT/MiniLM embedding model (feature/bert-dsl-embeddings): quality + on-device numbers, and a merge path
维护者通常 1 天内回复
还没有人认领这个 Issue。
评估
- 难度
- 5/5
- 预计耗时
- 一周以上
- 新手友好度
- 25/100
- Issue 类型
- 功能
- 描述清晰度
- 基本清楚
- 活跃度
- 停滞
- 技术栈
- kotlin
调研方向
Start by reviewing the feature/bert-dsl-embeddings branch, especially BertEmbeddingModel, MiniLmExportHarness, and the kbert-cli app; compare its eager and StableHLO outputs with sentence-transformers all-MiniLM-L6-v2 on a fixed sentence set. Record parity, retrieval benchmark results, and JVM, Android, and iOS speed and memory measurements. Then document what remains to merge into develop, including the legacy eager BERT removal and whether the export harness can move to commonTest.
由索引模型根据 Issue 内容生成。
描述
Context
feature/bert-dsl-embeddings already carries a substantial embedding stack that
develop doesn't have:
BertEmbeddingModelone-call factory (fromSafeTensors+fromHuggingFace)
inllm-providers, with tests- the DSL BERT in
llm-inference/bert(legacy eager BERT stack removed on the
branch) MiniLmExportHarness(jvmTest) — fixed-seq-bucket StableHLO export for the
on-device embedder cartridge, plus eager reference probes- a
kbert-cliapp inllm-apps
So "do we have a usable embedding model?" is really "how good is the one we
already have, and what does it take to land it?"
Ask — an evaluation pass with recorded numbers
- Correctness / parity: cosine similarity of branch embeddings vs the
reference implementation (sentence-transformersall-MiniLM-L6-v2) over a
fixed sentence set — eager path and StableHLO-exported path both, per
fixed-seq bucket. - Retrieval quality: a small STS/retrieval benchmark (e.g. STS-B subset)
scored against the reference model's numbers, so regressions are visible in
review rather than discovered by users. - Performance: embeddings/sec + resident memory on JVM, Android and iOS —
embedding models are the most credible near-term on-device story (they're
small and latency-tolerant), and the same mobile constraints we documented
for generative models (no native kernels, ART heap cap) apply here and
should be quantified. - Merge path: what remains between the branch and develop (the branch also
removes the legacy eager BERT stack — that's a breaking change worth its own
review), and whether the export harness belongs in commonTest rather than
jvmTest so the parity probes run cross-target.
We can run the on-device measurements (physical Android + iPhone) and contribute
the parity/benchmark test harness.
- 主要语言
- Kotlin
- 星标
- 8
- 派生
- 3
- 平均合并
- 1 天 16 小时
- 30 天内合并 PR
- 48
环境准备
这个项目没有提供开发容器、Dockerfile 或贡献指南，环境需要你自己搭建：先看它的 README，通用步骤见我们的新手贡献指南。
从这里开始
- 先读完整个 Issue，再读项目的贡献指南。
- 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
- Fork 仓库，在一个分支上完成修改。
- 提交 Pull Request，并在描述里引用这个 Issue 编号。
SKaiNET-developers/SKaiNET-transformers 的其他 Issue
-
documentation good first issue size:xs skill:docs sub-issue
难度 2/5 1-3 小时 新手友好度 78/100
SKaiNET-developers/SKaiNET-transformers#478 · 1 条评论 ·
维护者通常 1 天内回复
-
enhancement skill:kotlin-core
难度 4/5 3-5 天 新手友好度 48/100
SKaiNET-developers/SKaiNET-transformers#502 ·
维护者通常 1 天内回复
-
bug
难度 4/5 3-5 天 新手友好度 48/100
SKaiNET-developers/SKaiNET-transformers#483 ·
维护者通常 1 天内回复
-
coding good first issue size:s skill:kotlin-core sub-issue
难度 4/5 1-2 天 新手友好度 64/100
SKaiNET-developers/SKaiNET-transformers#477 · 1 条评论 ·
维护者通常 1 天内回复
-
coding good first issue size:s skill:numerics sub-issue
难度 4/5 3-5 天 新手友好度 55/100
SKaiNET-developers/SKaiNET-transformers#476 · 1 条评论 ·
维护者通常 1 天内回复
查看 SKaiNET-developers/SKaiNET-transformers 的全部 Issue
相似的 Issue
-
难度 2/5 1-3 小时 新手友好度 78/100
AniTrend/anitrend-v2#1344 ·
维护者通常 1 天内回复
-
难度 2/5 1-3 小时 新手友好度 68/100
zhanghai/MaterialFiles#1659 ·
-
难度 2/5 1-3 小时 新手友好度 66/100
Droid-Deck/DroidDeck#1098 ·
维护者通常 1 天内回复
-
Blank screen in Download tab after searching with no results and switching tabs可能已有人在做 @mangod12 今天认领。 未关闭bug good first issue
难度 2/5 1-3 小时 新手友好度 70/100
kiwix/kiwix-android#5211 ·
维护者通常 1 天内回复
-
bug
难度 2/5 1-3 小时 新手友好度 65/100
deniscerri/ytdlnis#1430 ·
维护者通常 1 天内回复