Instructions to use Infinity08/KAWK-500M-Korean-Instruct-v1 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Infinity08/KAWK-500M-Korean-Instruct-v1 with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="Infinity08/KAWK-500M-Korean-Instruct-v1")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Infinity08/KAWK-500M-Korean-Instruct-v1")
model = AutoModelForCausalLM.from_pretrained("Infinity08/KAWK-500M-Korean-Instruct-v1", device_map="auto")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use Infinity08/KAWK-500M-Korean-Instruct-v1 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Infinity08/KAWK-500M-Korean-Instruct-v1"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Infinity08/KAWK-500M-Korean-Instruct-v1",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Infinity08/KAWK-500M-Korean-Instruct-v1

SGLang

How to use Infinity08/KAWK-500M-Korean-Instruct-v1 with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "Infinity08/KAWK-500M-Korean-Instruct-v1" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Infinity08/KAWK-500M-Korean-Instruct-v1",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "Infinity08/KAWK-500M-Korean-Instruct-v1" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Infinity08/KAWK-500M-Korean-Instruct-v1",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Infinity08/KAWK-500M-Korean-Instruct-v1 with Docker Model Runner:
```
docker model run hf.co/Infinity08/KAWK-500M-Korean-Instruct-v1
```

KAWK-500M-Korean-Instruct-v1 / evaluation /20260809T050630Z /summary.csv

Infinity08

Upload KoBEST and KMMLU benchmark results

9a1b839 verified about 1 month ago

Raw

History Blame Contribute Delete

10.8 kB

	model,section,task,metric,value,stderr
	kawk500m-instruct-v1,groups,kobest,sample_len,4561.0,
	kawk500m-instruct-v1,groups,kobest,"acc,none",0.5231308923481692,0.007297298748221897
	kawk500m-instruct-v1,groups,kobest,"acc_norm,none",0.474,0.022352791650914174
	kawk500m-instruct-v1,groups,kobest,"f1,none",0.46754121249081004,
	kawk500m-instruct-v1,groups,kmmlu_stem,sample_len,9900.0,
	kawk500m-instruct-v1,groups,kmmlu_stem,"acc,none",0.13797979797979798,0.003422447552202874
	kawk500m-instruct-v1,groups,kmmlu_other,sample_len,8400.0,
	kawk500m-instruct-v1,groups,kmmlu_other,"acc,none",0.14952380952380953,0.0038749065104715114
	kawk500m-instruct-v1,groups,kmmlu_applied_science,sample_len,11600.0,
	kawk500m-instruct-v1,groups,kmmlu_applied_science,"acc,none",0.12637931034482758,0.0030704612706282304
	kawk500m-instruct-v1,groups,kmmlu_humss,sample_len,5130.0,
	kawk500m-instruct-v1,groups,kmmlu_humss,"acc,none",0.2155945419103314,0.005734774245353025
	kawk500m-instruct-v1,groups,kmmlu,sample_len,35030.0,
	kawk500m-instruct-v1,groups,kmmlu,"acc,none",0.1482729089351984,0.001880917345299668
	kawk500m-instruct-v1,results,kobest_boolq,sample_len,1404.0,
	kawk500m-instruct-v1,results,kobest_boolq,"acc,none",0.5135327635327636,0.013343877331261893
	kawk500m-instruct-v1,results,kobest_boolq,"f1,none",0.4553178437236408,
	kawk500m-instruct-v1,results,kobest_copa,sample_len,1000.0,
	kawk500m-instruct-v1,results,kobest_copa,"acc,none",0.635,0.015231776226264848
	kawk500m-instruct-v1,results,kobest_copa,"f1,none",0.6335455422572189,
	kawk500m-instruct-v1,results,kobest_hellaswag,sample_len,500.0,
	kawk500m-instruct-v1,results,kobest_hellaswag,"acc,none",0.354,0.02140758204791649
	kawk500m-instruct-v1,results,kobest_hellaswag,"f1,none",0.3513485918054787,
	kawk500m-instruct-v1,results,kobest_hellaswag,"acc_norm,none",0.474,0.022352791650914174
	kawk500m-instruct-v1,results,kobest_sentineg,sample_len,397.0,
	kawk500m-instruct-v1,results,kobest_sentineg,"acc,none",0.6045340050377834,0.024570689701367988
	kawk500m-instruct-v1,results,kobest_sentineg,"f1,none",0.592580971990718,
	kawk500m-instruct-v1,results,kobest_wic,sample_len,1260.0,
	kawk500m-instruct-v1,results,kobest_wic,"acc,none",0.4865079365079365,0.014086365971849238
	kawk500m-instruct-v1,results,kobest_wic,"f1,none",0.3561228043986665,
	kawk500m-instruct-v1,results,kmmlu_biology,sample_len,1000.0,
	kawk500m-instruct-v1,results,kmmlu_biology,"acc,none",0.204,0.012749374359024311
	kawk500m-instruct-v1,results,kmmlu_chemical_engineering,sample_len,1000.0,
	kawk500m-instruct-v1,results,kmmlu_chemical_engineering,"acc,none",0.215,0.01299784381903183
	kawk500m-instruct-v1,results,kmmlu_chemistry,sample_len,600.0,
	kawk500m-instruct-v1,results,kmmlu_chemistry,"acc,none",0.21666666666666667,0.016832783728500004
	kawk500m-instruct-v1,results,kmmlu_civil_engineering,sample_len,1000.0,
	kawk500m-instruct-v1,results,kmmlu_civil_engineering,"acc,none",0.084,0.008776162089491113
	kawk500m-instruct-v1,results,kmmlu_computer_science,sample_len,1000.0,
	kawk500m-instruct-v1,results,kmmlu_computer_science,"acc,none",0.086,0.008870325962594761
	kawk500m-instruct-v1,results,kmmlu_ecology,sample_len,1000.0,
	kawk500m-instruct-v1,results,kmmlu_ecology,"acc,none",0.122,0.010354864712936776
	kawk500m-instruct-v1,results,kmmlu_electrical_engineering,sample_len,1000.0,
	kawk500m-instruct-v1,results,kmmlu_electrical_engineering,"acc,none",0.082,0.008680515615523732
	kawk500m-instruct-v1,results,kmmlu_information_technology,sample_len,1000.0,
	kawk500m-instruct-v1,results,kmmlu_information_technology,"acc,none",0.104,0.009658016218524242
	kawk500m-instruct-v1,results,kmmlu_materials_engineering,sample_len,1000.0,
	kawk500m-instruct-v1,results,kmmlu_materials_engineering,"acc,none",0.157,0.011510146979230262
	kawk500m-instruct-v1,results,kmmlu_math,sample_len,300.0,
	kawk500m-instruct-v1,results,kmmlu_math,"acc,none",0.26666666666666666,0.02557404853322572
	kawk500m-instruct-v1,results,kmmlu_mechanical_engineering,sample_len,1000.0,
	kawk500m-instruct-v1,results,kmmlu_mechanical_engineering,"acc,none",0.102,0.009575368801653944
	kawk500m-instruct-v1,results,kmmlu_agricultural_sciences,sample_len,1000.0,
	kawk500m-instruct-v1,results,kmmlu_agricultural_sciences,"acc,none",0.174,0.011994493230973449
	kawk500m-instruct-v1,results,kmmlu_construction,sample_len,1000.0,
	kawk500m-instruct-v1,results,kmmlu_construction,"acc,none",0.111,0.009938701010583716
	kawk500m-instruct-v1,results,kmmlu_fashion,sample_len,1000.0,
	kawk500m-instruct-v1,results,kmmlu_fashion,"acc,none",0.186,0.012310790208412926
	kawk500m-instruct-v1,results,kmmlu_food_processing,sample_len,1000.0,
	kawk500m-instruct-v1,results,kmmlu_food_processing,"acc,none",0.176,0.012048616898597498
	kawk500m-instruct-v1,results,kmmlu_health,sample_len,100.0,
	kawk500m-instruct-v1,results,kmmlu_health,"acc,none",0.25,0.04351941398892446
	kawk500m-instruct-v1,results,kmmlu_interior_architecture_and_design,sample_len,1000.0,
	kawk500m-instruct-v1,results,kmmlu_interior_architecture_and_design,"acc,none",0.111,0.009938701010583716
	kawk500m-instruct-v1,results,kmmlu_marketing,sample_len,1000.0,
	kawk500m-instruct-v1,results,kmmlu_marketing,"acc,none",0.131,0.0106748748448379
	kawk500m-instruct-v1,results,kmmlu_patent,sample_len,100.0,
	kawk500m-instruct-v1,results,kmmlu_patent,"acc,none",0.25,0.04351941398892446
	kawk500m-instruct-v1,results,kmmlu_public_safety,sample_len,1000.0,
	kawk500m-instruct-v1,results,kmmlu_public_safety,"acc,none",0.116,0.010131468138756924
	kawk500m-instruct-v1,results,kmmlu_real_estate,sample_len,200.0,
	kawk500m-instruct-v1,results,kmmlu_real_estate,"acc,none",0.22,0.02936514188266327
	kawk500m-instruct-v1,results,kmmlu_refrigerating_machinery,sample_len,1000.0,
	kawk500m-instruct-v1,results,kmmlu_refrigerating_machinery,"acc,none",0.157,0.011510146979230262
	kawk500m-instruct-v1,results,kmmlu_aviation_engineering_and_maintenance,sample_len,1000.0,
	kawk500m-instruct-v1,results,kmmlu_aviation_engineering_and_maintenance,"acc,none",0.144,0.01110798754893916
	kawk500m-instruct-v1,results,kmmlu_electronics_engineering,sample_len,1000.0,
	kawk500m-instruct-v1,results,kmmlu_electronics_engineering,"acc,none",0.092,0.009144376393151129
	kawk500m-instruct-v1,results,kmmlu_energy_management,sample_len,1000.0,
	kawk500m-instruct-v1,results,kmmlu_energy_management,"acc,none",0.207,0.012818553557844009
	kawk500m-instruct-v1,results,kmmlu_environmental_science,sample_len,1000.0,
	kawk500m-instruct-v1,results,kmmlu_environmental_science,"acc,none",0.101,0.009533618929341046
	kawk500m-instruct-v1,results,kmmlu_gas_technology_and_engineering,sample_len,1000.0,
	kawk500m-instruct-v1,results,kmmlu_gas_technology_and_engineering,"acc,none",0.118,0.010206869264381718
	kawk500m-instruct-v1,results,kmmlu_geomatics,sample_len,1000.0,
	kawk500m-instruct-v1,results,kmmlu_geomatics,"acc,none",0.125,0.010463483381956722
	kawk500m-instruct-v1,results,kmmlu_industrial_engineer,sample_len,1000.0,
	kawk500m-instruct-v1,results,kmmlu_industrial_engineer,"acc,none",0.071,0.008125578442487959
	kawk500m-instruct-v1,results,kmmlu_machine_design_and_manufacturing,sample_len,1000.0,
	kawk500m-instruct-v1,results,kmmlu_machine_design_and_manufacturing,"acc,none",0.123,0.010391293421849803
	kawk500m-instruct-v1,results,kmmlu_maritime_engineering,sample_len,600.0,
	kawk500m-instruct-v1,results,kmmlu_maritime_engineering,"acc,none",0.15666666666666668,0.014851643766757317
	kawk500m-instruct-v1,results,kmmlu_nondestructive_testing,sample_len,1000.0,
	kawk500m-instruct-v1,results,kmmlu_nondestructive_testing,"acc,none",0.15,0.011297239823409416
	kawk500m-instruct-v1,results,kmmlu_railway_and_automotive_engineering,sample_len,1000.0,
	kawk500m-instruct-v1,results,kmmlu_railway_and_automotive_engineering,"acc,none",0.136,0.01084535023047304
	kawk500m-instruct-v1,results,kmmlu_telecommunications_and_wireless_technology,sample_len,1000.0,
	kawk500m-instruct-v1,results,kmmlu_telecommunications_and_wireless_technology,"acc,none",0.105,0.00969892102602496
	kawk500m-instruct-v1,results,kmmlu_accounting,sample_len,100.0,
	kawk500m-instruct-v1,results,kmmlu_accounting,"acc,none",0.24,0.04292346959909278
	kawk500m-instruct-v1,results,kmmlu_criminal_law,sample_len,200.0,
	kawk500m-instruct-v1,results,kmmlu_criminal_law,"acc,none",0.195,0.028085923439997246
	kawk500m-instruct-v1,results,kmmlu_economics,sample_len,130.0,
	kawk500m-instruct-v1,results,kmmlu_economics,"acc,none",0.27692307692307694,0.03939825345266472
	kawk500m-instruct-v1,results,kmmlu_education,sample_len,100.0,
	kawk500m-instruct-v1,results,kmmlu_education,"acc,none",0.28,0.045126085985421296
	kawk500m-instruct-v1,results,kmmlu_korean_history,sample_len,100.0,
	kawk500m-instruct-v1,results,kmmlu_korean_history,"acc,none",0.23,0.04229525846816507
	kawk500m-instruct-v1,results,kmmlu_law,sample_len,1000.0,
	kawk500m-instruct-v1,results,kmmlu_law,"acc,none",0.24,0.013512312258920847
	kawk500m-instruct-v1,results,kmmlu_management,sample_len,1000.0,
	kawk500m-instruct-v1,results,kmmlu_management,"acc,none",0.204,0.012749374359024311
	kawk500m-instruct-v1,results,kmmlu_political_science_and_sociology,sample_len,300.0,
	kawk500m-instruct-v1,results,kmmlu_political_science_and_sociology,"acc,none",0.23,0.024337372337779037
	kawk500m-instruct-v1,results,kmmlu_psychology,sample_len,1000.0,
	kawk500m-instruct-v1,results,kmmlu_psychology,"acc,none",0.224,0.013190830072364589
	kawk500m-instruct-v1,results,kmmlu_social_welfare,sample_len,1000.0,
	kawk500m-instruct-v1,results,kmmlu_social_welfare,"acc,none",0.172,0.011939788882495308
	kawk500m-instruct-v1,results,kmmlu_taxation,sample_len,200.0,
	kawk500m-instruct-v1,results,kmmlu_taxation,"acc,none",0.235,0.03005647949775547
	kawk500m-instruct-v1,results,kobest,sample_len,4561.0,
	kawk500m-instruct-v1,results,kobest,"acc,none",0.5231308923481692,0.007297298748221897
	kawk500m-instruct-v1,results,kobest,"acc_norm,none",0.474,0.022352791650914174
	kawk500m-instruct-v1,results,kobest,"f1,none",0.46754121249081004,
	kawk500m-instruct-v1,results,kmmlu_stem,sample_len,9900.0,
	kawk500m-instruct-v1,results,kmmlu_stem,"acc,none",0.13797979797979798,0.003422447552202874
	kawk500m-instruct-v1,results,kmmlu_other,sample_len,8400.0,
	kawk500m-instruct-v1,results,kmmlu_other,"acc,none",0.14952380952380953,0.0038749065104715114
	kawk500m-instruct-v1,results,kmmlu_applied_science,sample_len,11600.0,
	kawk500m-instruct-v1,results,kmmlu_applied_science,"acc,none",0.12637931034482758,0.0030704612706282304
	kawk500m-instruct-v1,results,kmmlu_humss,sample_len,5130.0,
	kawk500m-instruct-v1,results,kmmlu_humss,"acc,none",0.2155945419103314,0.005734774245353025
	kawk500m-instruct-v1,results,kmmlu,sample_len,35030.0,
	kawk500m-instruct-v1,results,kmmlu,"acc,none",0.1482729089351984,0.001880917345299668