---
license: llama3.2
language:
- vi
base_model:
- meta-llama/Llama-3.2-1B-Instruct
pipeline_tag: sentence-similarity
library_name: transformers
---

# misa-ai/Llama-3.2-1B-Instruct-Embedding-Base

This is a Embedding model for Document Retrieval: It maps sentences & paragraphs to a 2048 dimensional dense vector space and can be used for tasks like clustering or semantic search.

We train the model on a merged training dataset that consists of multiple domains, about 900k triplets in Vietnamese: 

We use [Llama-3.2-1B-Instruct](https://github.com/meta-llama/Llama-3.2-1B-Instruct) as the pre-trained backbone.

This model directed to Document Retrieval.

Details:
- Max support context size: 4096 tokens
- Pooling last token (should use `padding_side = "left"`)
- Language: Vietnamese
- Prompts:
  - Query: "Cho một câu truy vấn tìm kiếm thông tin, hãy truy xuất các tài liệu có liên quan trả lời cho truy vấn đó."
  - Document: ""


### Please cite our manuscript if this dataset is used for your work

- Organization: MISA JSC
- Author: [Sy-The Ho](https://huggingface.co/thehosy)