Instructions to use LDCC/KORMo-19B-MoE with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use LDCC/KORMo-19B-MoE with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="LDCC/KORMo-19B-MoE", trust_remote_code=True) messages = [ {"role": "user", "content": "Who are you?"}, ] pipe(messages)# Load model directly from transformers import AutoModelForCausalLM model = AutoModelForCausalLM.from_pretrained("LDCC/KORMo-19B-MoE", trust_remote_code=True, dtype="auto") - Notebooks
- Google Colab
- Kaggle
- Local Apps
- vLLM
How to use LDCC/KORMo-19B-MoE with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "LDCC/KORMo-19B-MoE" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "LDCC/KORMo-19B-MoE", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker
docker model run hf.co/LDCC/KORMo-19B-MoE
- SGLang
How to use LDCC/KORMo-19B-MoE 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 "LDCC/KORMo-19B-MoE" \ --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": "LDCC/KORMo-19B-MoE", "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 "LDCC/KORMo-19B-MoE" \ --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": "LDCC/KORMo-19B-MoE", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }' - Docker Model Runner
How to use LDCC/KORMo-19B-MoE with Docker Model Runner:
docker model run hf.co/LDCC/KORMo-19B-MoE
# Load model directly
from transformers import AutoModelForCausalLM
model = AutoModelForCausalLM.from_pretrained("LDCC/KORMo-19B-MoE", trust_remote_code=True, dtype="auto")Quick Links
KORMo-MoE (2-Experts)
본 모델은 KORMo-Team/KORMo-10B-sft를 기반으로, 두 개의 Expert를 갖는 Mixture of Experts (MoE) 구조로 확장한 버전입니다.
개요
이 모델은 다음 두 개의 한국어 데이터셋을 활용하여 **QLoRA 기반 미세조정(fine-tuning)**을 수행한 후, 2-Expert MoE 모델로 변환한 실험용 연구 모델입니다.
학습 구성
- 미세조정 방식: QLoRA
- Expert 수: 2
- MoE 변환 도구: mergekit
- 베이스 모델: KORMo-10B-sft
vLLM 호환성
본 모델을 vLLM에서 실행하기 위해서는 별도의 패치 적용이 필요합니다. 다음 스크립트를 사용하여 설치를 진행할 수 있습니다.
- vllm==0.10.0
- install_vllm_support.sh
참고 및 감사
본 연구는 KORMo-Team에서 공개한 오픈소스 모델인 KORMo-10B-sft를 기반으로 수행되었습니다. 모델을 공개해주신 KORMo-Team에 감사의 뜻을 전합니다.
- Downloads last month
- 54
Model tree for LDCC/KORMo-19B-MoE
Base model
KORMo-Team/KORMo-10B-sft
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="LDCC/KORMo-19B-MoE", trust_remote_code=True) messages = [ {"role": "user", "content": "Who are you?"}, ] pipe(messages)