PGOS Emotion Agent โ Stage 1 (๋๋ถ๋ฅ)
Psychological Growth OS(PGOS) ํ๋ก์ ํธ์ ๊ฐ์ ๋ถ๋ฅ ๋ชจ๋ธ์ ๋๋ค. ๊ณ์ธต์ 2๋จ๊ณ ๋ถ๋ฅ๊ธฐ์ 1๋จ๊ณ ๋๋ถ๋ฅ ๋ชจ๋ธ์ ๋๋ค.
๋ชจ๋ธ ๊ฐ์
์ ๋ ฅ ํ ์คํธ๋ฅผ 6๊ฐ ๊ฐ์ ๋๋ถ๋ฅ๋ก ๋ถ๋ฅํฉ๋๋ค.
| ํด๋์ค | ํฌํจ ๊ฐ์ |
|---|---|
| ๋ถ๋ ธ๊ณ | ๋ถ๋ ธ, ์ข์ , ์ง์ฆ |
| ์ฌํ๊ณ | ์ฌํ, ์ฐ์ธ, ํํ |
| ๋ถ์๊ณ | ๋ถ์, ๊ฑฑ์ |
| ์์ฒ๊ณ | ์์ฒ, ์ต์ธ, ์งํฌ |
| ๋นํฉ๊ณ | ์์น์ฌ, ์ฃ์ฑ ๊ฐ, ๋นํฉ |
| ๊ธฐ์จ๊ณ | ๊ธฐ์จ, ์๋ |
์ฑ๋ฅ
| ์งํ | ๊ฐ |
|---|---|
| Macro F1 | 0.692 |
| Accuracy | 0.684 |
ํ์ต ๋ฐ์ดํฐ
- ์์ฒ: AI Hub 018 ๊ฐ์ฑ๋ํ ๋ง๋ญ์น
- ๋ฐ์ดํฐ: ํด๋์ค๋น ๊ท ๋ฑ ์ํ๋ง + GPT-4o-mini ์ฆ๊ฐ
- ์ด ํ์ต ๋ฐ์ดํฐ: 64,000๊ฐ
๋ฒ ์ด์ค ๋ชจ๋ธ
์ฌ์ฉ ๋ฐฉ๋ฒ
from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch
tokenizer = AutoTokenizer.from_pretrained("nuguri01/pgos-emotion-stage1-major")
model = AutoModelForSequenceClassification.from_pretrained("nuguri01/pgos-emotion-stage1-major")
text = "์ค๋ ๋๋ฌด ํ๊ฐ ๋๊ณ ์ง์ฆ์ด ๋ฐ๋ ค์์"
inputs = tokenizer(text, return_tensors="pt", max_length=128, truncation=True)
with torch.no_grad():
logits = model(**inputs).logits
pred = torch.argmax(logits, dim=-1).item()
labels = ["๋ถ๋
ธ๊ณ", "์ฌํ๊ณ", "๋ถ์๊ณ", "์์ฒ๊ณ", "๋นํฉ๊ณ", "๊ธฐ์จ๊ณ"]
print(f"๋๋ถ๋ฅ: {labels[pred]}")
PGOS ํ์ดํ๋ผ์ธ
์ ๋ ฅ ํ ์คํธ โ Stage 1 (์ด ๋ชจ๋ธ) โ ๋๋ถ๋ฅ ์์ธก โ Stage 2 (์๋ถ๋ฅ ๋ชจ๋ธ) โ ์ธ๋ถ ๊ฐ์ ์์ธก โ Top-3 ๊ฐ์ ๋ฐํ (Hit Rate 0.877)
๊ด๋ จ ๋ชจ๋ธ
- Stage 2: pgos-emotion-stage2-anger
- Stage 2: pgos-emotion-stage2-sadness
- Stage 2: pgos-emotion-stage2-anxiety
- Stage 2: pgos-emotion-stage2-hurt
- Stage 2: pgos-emotion-stage2-embarrass
- Stage 2: pgos-emotion-stage2-joy
ํ๋ก์ ํธ
- ์๋น์ค: Psychological Growth OS (PGOS)
- ์ฒ ํ: Carl Rogers ์ธ๊ฐ์ค์ฌ์น๋ฃ + ACT Hexaflex
- Behavior Agent: nuguri01/pgos-behavior-agent
- Downloads last month
- 17
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐ Ask for provider support