Korean Stock News Qwen 3B LoRA

๋ชจ๋ธ ๊ฐœ์š”

ํ•œ๊ตญ ์ฃผ์‹ ๋‰ด์Šค ๋ถ„์„์„ ์œ„ํ•ด Qwen2.5-3B-Instruct๋ฅผ ํŒŒ์ธํŠœ๋‹ํ•œ LoRA ๋ชจ๋ธ์ž…๋‹ˆ๋‹ค.

์ฃผ์š” ๊ธฐ๋Šฅ

  • ๐Ÿ“ฐ ๋‰ด์Šค ์นดํ…Œ๊ณ ๋ฆฌ ๋ถ„๋ฅ˜ (domestic_direct/global_related/macro_economic/geopolitical/irrelevant)
  • ๐Ÿ“Š ์ฃผ์‹์‹œ์žฅ ์˜ํ–ฅ๋„ ๋ถ„์„
  • ๐Ÿข ๊ด€๋ จ ๊ธฐ์—… ์ถ”์ถœ
  • ๐Ÿ’น ํˆฌ์ž ์ถ”์ฒœ ์ƒ์„ฑ

์„ฑ๋Šฅ

  • Base Model: Qwen2.5-3B-Instruct
  • Training Time: 37๋ถ„ (345๋ฐฐ ์„ฑ๋Šฅ ํ–ฅ์ƒ)
  • Token Accuracy: 79%
  • Training Loss: 0.98

์‚ฌ์šฉ๋ฒ•

from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel

# ๋ฒ ์ด์Šค ๋ชจ๋ธ ๋กœ๋“œ
base_model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen2.5-3B-Instruct")
tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen2.5-3B-Instruct")

# LoRA ์–ด๋Œ‘ํ„ฐ ์ ์šฉ
model = PeftModel.from_pretrained(base_model, "3kd1000/3kd1000/korean-stock-news-qwen-3b-lora")

ํ•™์Šต ํ™˜๊ฒฝ

  • GPU: AMD RX 9070 XT (16GB VRAM)
  • CPU: AMD 9800X3D
  • RAM: 32GB
  • OS: WSL2 Ubuntu 22.04.5 LTS
  • Framework: Transformers + PEFT + TRL

๋ผ์ด์„ ์Šค

Apache 2.0

์ œ์ž‘์ž

์ •์ฃผ์ƒ (jsjung)

Downloads last month
15
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support

Model tree for 3kd1000/korean-stock-news-qwen-3b-lora

Base model

Qwen/Qwen2.5-3B
Adapter
(563)
this model