BPVELA-E560M

BPVELA-E560M is the accuracy-first BPVELA release line for Traditional Chinese retrieval and embedding use cases.

繁體中文說明

BPVELA-E560M 是 BPVELA 目前的 accuracy-first 系列，針對繁體中文語意檢索、相似度比對與 retrieval-first RAG 場景做優化。

模型摘要

• 系列版本：v1.0.0
• 基底模型：intfloat/multilingual-e5-large
• 釋出形式：LoRA adapter 加上 SentenceTransformer 組件
• 建議用途：semantic retrieval、retrieval-first RAG、similarity search
• 主要語言：Traditional Chinese / 繁體中文

重要說明

這個 repository 釋出的是 LoRA adapter，不是 merged full checkpoint。使用時需要以 base model 為底，再載入這個 adapter。

驗證摘要

• Taiwan-md pair benchmark：Spearman 0.8400、Pearson 0.9224
• Wrapped retrieval smoke：pass rate 1.0000、retrieval hit rate 1.0000、top-1 rate 0.9667

Query / Passage 格式

這條模型線基於 E5，做檢索時建議保留標準前綴。

• Query：query: 你的問題
• Passage：passage: 文件內容

備註

• bpvela_model_config.yaml 保留了專案內部使用的載入設定。
• 這個公開模型 repo 不需要包含 Taiwan-md corpus 或 FAISS index。
• 公開前請再確認最終 license。

授權說明

• Taiwan-MD 內容授權：CC BY-SA 4.0
• BPVELA 專案程式碼授權：MIT
• 基底模型 intfloat/multilingual-e5-large：Hugging Face 標示為 MIT
• 本 repo 釋出的 adapter 權重與模型卡內容，建議以 CC BY-SA 4.0 方式對外說明

本 repository 公開的是 BPVELA-E560M 的 LoRA adapter 權重、模型卡與相關說明文件，並不包含 intfloat/multilingual-e5-large 的完整基底模型權重。

BPVELA-E560M 的訓練與優化過程使用了 Taiwan-MD 內容；依目前資料來源條件，建議將本 adapter 權重與模型卡內容以 CC BY-SA 4.0 對外說明與散布。

任何再散布、修改版散布、或以本 adapter 為基礎的公開衍生釋出，建議：

• 保留原始出處與適當署名
• 清楚標示修改情形
• 以相同或相容的分享方式提供衍生內容

使用者在載入與使用本 adapter 時，仍需自行遵守上游基底模型 intfloat/multilingual-e5-large 的授權條件。

Summary

• Series version: v1.0.0
• Base model: intfloat/multilingual-e5-large
• Release type: LoRA adapter plus SentenceTransformer modules
• Recommended usage: semantic retrieval, retrieval-first RAG, similarity search
• Language focus: Traditional Chinese

Important

This repository contains a LoRA adapter release, not a merged full checkpoint. Load it on top of the base model.

Validation Snapshot

• Taiwan-md pair benchmark: Spearman 0.8400, Pearson 0.9224
• Wrapped retrieval smoke: pass rate 1.0000, retrieval hit rate 1.0000, top-1 rate 0.9667

Query And Passage Formatting

This line is based on E5. For retrieval, keep the standard E5 prefixes.

• Query: query: your question
• Passage: passage: your document

Loading Example

from sentence_transformers import SentenceTransformer
from sentence_transformers.models import Normalize, Pooling, Transformer
from peft import PeftModel

base_model = "intfloat/multilingual-e5-large"
adapter_repo = "BluePlanetAI/BPVELA-E560M"

transformer = Transformer(base_model)
transformer.auto_model = PeftModel.from_pretrained(
    transformer.auto_model,
    adapter_repo,
    is_trainable=False,
)

pooling = Pooling.load(adapter_repo, subfolder="1_Pooling")
normalize = Normalize.load(adapter_repo, subfolder="2_Normalize")

model = SentenceTransformer(modules=[transformer, pooling, normalize])

emb = model.encode(["query: 台灣颱風災害應變流程"], normalize_embeddings=True)
print(len(emb[0]))

Notes

• bpvela_model_config.yaml is included as the project-side loading reference.
• This public model repo does not need to include the Taiwan-md corpus or FAISS index.
• Release owner should finalize the public license before publishing.

License Notes

• Taiwan-MD content license: CC BY-SA 4.0
• BPVELA project code license: MIT
• Base model intfloat/multilingual-e5-large: marked as MIT on Hugging Face
• The adapter weights and model card content published in this repo are best documented as CC BY-SA 4.0

This repository publishes the BPVELA-E560M LoRA adapter weights, model card, and related documentation only. It does not redistribute the full base-model weights of intfloat/multilingual-e5-large.

Because the training and optimization process uses Taiwan-MD content, the adapter release and model card are best documented for public distribution under CC BY-SA 4.0.

For redistribution, modified redistribution, or public derivative releases based on this adapter, users should:

• preserve attribution to the original release
• clearly indicate modifications
• keep the share-alike expectations for the released derivative materials

Use of this adapter remains subject to the applicable license terms of the upstream base model.