Multiple Choice
Transformers
Safetensors
Korean
qwen3
text-generation
korean
classification
natural-language-inference
function-calling
tool-use
calibrated
Eval Results (legacy)
text-generation-inference
Instructions to use coderecode95/KJH-Judge-4B-ko with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use coderecode95/KJH-Judge-4B-ko with Transformers:
# pip install -U transformers accelerate # Load model directly from transformers import AutoTokenizer, AutoModelForCausalLM tokenizer = AutoTokenizer.from_pretrained("coderecode95/KJH-Judge-4B-ko") model = AutoModelForCausalLM.from_pretrained("coderecode95/KJH-Judge-4B-ko", device_map="auto") - Notebooks
- Google Colab
- Kaggle
Update README.md
Browse files
README.md
CHANGED
|
@@ -1,53 +1,97 @@
|
|
| 1 |
-
---
|
| 2 |
-
license: apache-2.0
|
| 3 |
-
license_link: https://huggingface.co/Qwen/Qwen3-4B/blob/main/LICENSE
|
| 4 |
-
base_model: Qwen/Qwen3-4B
|
| 5 |
-
base_model_relation: finetune
|
| 6 |
-
language:
|
| 7 |
-
|
| 8 |
-
library_name: transformers
|
| 9 |
-
tags:
|
| 10 |
-
|
| 11 |
-
|
| 12 |
-
|
| 13 |
-
|
| 14 |
-
|
| 15 |
-
|
| 16 |
-
|
| 17 |
-
|
| 18 |
-
|
| 19 |
-
|
| 20 |
-
-
|
| 21 |
-
|
| 22 |
-
|
| 23 |
-
|
| 24 |
-
|
| 25 |
-
|
| 26 |
-
|
| 27 |
-
|
| 28 |
-
|
| 29 |
-
|
| 30 |
-
|
| 31 |
-
|
| 32 |
-
|
| 33 |
-
|
| 34 |
-
|
| 35 |
-
|
| 36 |
-
|
| 37 |
-
|
| 38 |
-
|
| 39 |
-
|
| 40 |
-
|
| 41 |
-
|
| 42 |
-
|
| 43 |
-
|
| 44 |
-
|
| 45 |
-
|
| 46 |
-
|
| 47 |
-
|
| 48 |
-
|
| 49 |
-
|
| 50 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 51 |
|
| 52 |
# KJH-Judge-4B-ko
|
| 53 |
|
|
@@ -495,4 +539,4 @@ B. {๋ณด๊ธฐ 2}
|
|
| 495 |
|
| 496 |
- ๋ชจ๋ธ ๊ฐ์ค์น๋ ์๋ณธ Qwen3-4B์ ๊ฐ์ **Apache 2.0**์
๋๋ค ([LICENSE](LICENSE)). ๊ณต๊ฐ ํ์ต ๋ฐ์ดํฐ๋ ๋ชจ๋ CC BY-SA 4.0์
๋๋ค.
|
| 497 |
- ํ์ค ํํ ํ์ , ์ ์ฑ
ํ๋จ ๋ฑ ์ฌ๋์๊ฒ ์ํฅ์ ์ฃผ๋ ๊ฒฐ์ ์ ์ธ ๋๋ ํ์ ์ด ๋ฎ์ ๊ฒฐ๊ณผ๋ฅผ ์ฌ๋์ด ๊ฒํ ํ๋๋ก ํ์ธ์.
|
| 498 |
-
- ๋ฒ๋ฅ ยท๊ธ์ต ํ๋จ์ ๊ทผ๊ฑฐ๋ก ์ฐ์ง ๋ง์ธ์. ์ฃผ์ด์ง ๊ธ ์์์ ๋ด์ฉ์ ์ฐพ๊ณ ๋์กฐํ๋ ๋๊ตฌ์
๋๋ค.
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
license_link: https://huggingface.co/Qwen/Qwen3-4B/blob/main/LICENSE
|
| 4 |
+
base_model: Qwen/Qwen3-4B
|
| 5 |
+
base_model_relation: finetune
|
| 6 |
+
language:
|
| 7 |
+
- ko
|
| 8 |
+
library_name: transformers
|
| 9 |
+
tags:
|
| 10 |
+
- korean
|
| 11 |
+
- classification
|
| 12 |
+
- natural-language-inference
|
| 13 |
+
- function-calling
|
| 14 |
+
- tool-use
|
| 15 |
+
- calibrated
|
| 16 |
+
- multiple-choice
|
| 17 |
+
model-index:
|
| 18 |
+
- name: KJH-Judge-4B-ko
|
| 19 |
+
results:
|
| 20 |
+
- task:
|
| 21 |
+
type: text-classification
|
| 22 |
+
name: Sentiment classification
|
| 23 |
+
dataset:
|
| 24 |
+
name: NSMC
|
| 25 |
+
type: e9t/nsmc
|
| 26 |
+
split: test
|
| 27 |
+
metrics:
|
| 28 |
+
- type: accuracy
|
| 29 |
+
value: 85.7
|
| 30 |
+
name: accuracy (0-shot
|
| 31 |
+
n=1000): null
|
| 32 |
+
- task:
|
| 33 |
+
type: text-classification
|
| 34 |
+
name: Natural language inference
|
| 35 |
+
dataset:
|
| 36 |
+
name: KorNLI (XNLI)
|
| 37 |
+
type: kakaobrain/kor_nli
|
| 38 |
+
config: xnli
|
| 39 |
+
split: test
|
| 40 |
+
metrics:
|
| 41 |
+
- type: accuracy
|
| 42 |
+
value: 74.5
|
| 43 |
+
name: accuracy (0-shot
|
| 44 |
+
n=1000): null
|
| 45 |
+
- task:
|
| 46 |
+
type: text-classification
|
| 47 |
+
name: Paraphrase detection (binary STS)
|
| 48 |
+
dataset:
|
| 49 |
+
name: KLUE-STS
|
| 50 |
+
type: klue/klue
|
| 51 |
+
config: sts
|
| 52 |
+
split: validation
|
| 53 |
+
metrics:
|
| 54 |
+
- type: accuracy
|
| 55 |
+
value: 85.9
|
| 56 |
+
name: accuracy (0-shot
|
| 57 |
+
n=519): null
|
| 58 |
+
- task:
|
| 59 |
+
type: text-classification
|
| 60 |
+
name: Hate speech detection
|
| 61 |
+
dataset:
|
| 62 |
+
name: Korean UnSmile
|
| 63 |
+
type: smilegate-ai/kor_unsmile
|
| 64 |
+
split: valid
|
| 65 |
+
metrics:
|
| 66 |
+
- type: accuracy
|
| 67 |
+
value: 82.1
|
| 68 |
+
name: accuracy (0-shot
|
| 69 |
+
n=1000): null
|
| 70 |
+
- task:
|
| 71 |
+
type: multiple-choice
|
| 72 |
+
name: Knowledge QA
|
| 73 |
+
dataset:
|
| 74 |
+
name: KMMLU
|
| 75 |
+
type: HAERAE-HUB/KMMLU
|
| 76 |
+
split: test
|
| 77 |
+
metrics:
|
| 78 |
+
- type: accuracy
|
| 79 |
+
value: 42.6
|
| 80 |
+
name: accuracy (0-shot
|
| 81 |
+
n=990): null
|
| 82 |
+
- task:
|
| 83 |
+
type: multiple-choice
|
| 84 |
+
name: Korean knowledge QA
|
| 85 |
+
dataset:
|
| 86 |
+
name: HAE-RAE Bench 1.0
|
| 87 |
+
type: HAERAE-HUB/HAE_RAE_BENCH_1.0
|
| 88 |
+
split: test
|
| 89 |
+
metrics:
|
| 90 |
+
- type: accuracy
|
| 91 |
+
value: 56.1
|
| 92 |
+
name: accuracy (0-shot
|
| 93 |
+
n=983): null
|
| 94 |
+
---
|
| 95 |
|
| 96 |
# KJH-Judge-4B-ko
|
| 97 |
|
|
|
|
| 539 |
|
| 540 |
- ๋ชจ๋ธ ๊ฐ์ค์น๋ ์๋ณธ Qwen3-4B์ ๊ฐ์ **Apache 2.0**์
๋๋ค ([LICENSE](LICENSE)). ๊ณต๊ฐ ํ์ต ๋ฐ์ดํฐ๋ ๋ชจ๋ CC BY-SA 4.0์
๋๋ค.
|
| 541 |
- ํ์ค ํํ ํ์ , ์ ์ฑ
ํ๋จ ๋ฑ ์ฌ๋์๊ฒ ์ํฅ์ ์ฃผ๋ ๊ฒฐ์ ์ ์ธ ๋๋ ํ์ ์ด ๋ฎ์ ๊ฒฐ๊ณผ๋ฅผ ์ฌ๋์ด ๊ฒํ ํ๋๋ก ํ์ธ์.
|
| 542 |
+
- ๋ฒ๋ฅ ยท๊ธ์ต ํ๋จ์ ๊ทผ๊ฑฐ๋ก ์ฐ์ง ๋ง์ธ์. ์ฃผ์ด์ง ๊ธ ์์์ ๋ด์ฉ์ ์ฐพ๊ณ ๋์กฐํ๋ ๋๊ตฌ์
๋๋ค.
|