File size: 1,248 Bytes
3e3cabe
 
 
 
 
 
 
f46fe1a
3e3cabe
 
 
 
 
f46fe1a
b9e8daf
 
 
ddd0981
 
b9e8daf
a03918c
454e1b5
a03918c
454e1b5
a03918c
454e1b5
a03918c
454e1b5
a03918c
454e1b5
a03918c
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
---
license: mit
tags:
- trl
- sft
- retail
- finance
- unsloth
language:
- en
- th
base_model:
- unsloth/Llama-3.2-3B-Instruct
pipeline_tag: text-generation
datasets:
- airesearch/WangchanThaiInstruct
---
# PoonrooSLM Prototype: 3B (พูนรู้ SLM)
## This model is currently on experiment and be able to run on Ollama and LM Studio as the experimental version.

- Lightweight, text-only models that fit onto edge and mobile devices, finetuned version of Llama3.2 with WangchanThaiInstruct Dataset (only Financial and Retail tags)

- Support context length of 128K tokens and are state-of-the-art in their class for on-device use cases like summarization, instruction following, and rewriting tasks running locally at the edge.

- The Model empower developers to build personalized, on-device agentic applications with strong privacy where data never leaves the device. 

- Outperforms the Gemma 2 2B and Phi 3.5-mini (3.8B) models on tasks such as following instructions, summarization, prompt rewriting, and tool-use, while the 1B is competitive with Gemma.

- Multilingual Support in one single model. (Including Thai Language)

- Enhance more model capabilities by using RAG (Retrieval Augmented Generation) + Responsible AI supported