Instructions to use DSAiLab/llama2-70b-marlin-4bit with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use DSAiLab/llama2-70b-marlin-4bit with Transformers:

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

pipe = pipeline("text-generation", model="DSAiLab/llama2-70b-marlin-4bit")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("DSAiLab/llama2-70b-marlin-4bit")
model = AutoModelForCausalLM.from_pretrained("DSAiLab/llama2-70b-marlin-4bit", 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=256)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use DSAiLab/llama2-70b-marlin-4bit with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "DSAiLab/llama2-70b-marlin-4bit"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "DSAiLab/llama2-70b-marlin-4bit",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/DSAiLab/llama2-70b-marlin-4bit

SGLang

How to use DSAiLab/llama2-70b-marlin-4bit 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 "DSAiLab/llama2-70b-marlin-4bit" \
    --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": "DSAiLab/llama2-70b-marlin-4bit",
		"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 "DSAiLab/llama2-70b-marlin-4bit" \
        --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": "DSAiLab/llama2-70b-marlin-4bit",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use DSAiLab/llama2-70b-marlin-4bit with Docker Model Runner:
```
docker model run hf.co/DSAiLab/llama2-70b-marlin-4bit
```

LLaMA2-70B-Marlin-4bit

DSAiLab/llama2-70b-marlin-4bit는 Meta의 LLaMA2-70B 모델을 기반으로, GPTQ 양자화와 Marlin 커널을 결합하여 최적화된 4bit 버전입니다.

Marlin 커널은 빠른 추론과 적은 메모리 사용량을 목표로 설계된 고성능 디코딩 커널입니다.

Quantization (GPTQ + Marlin)

Base Model: LLaMA2-70B
Quantization Type: GPTQ 4bit
Kernel: Marlin (CUDA 최적화)
Group Size: 128
Activation Ordering: Enabled
Format: GPTQ with Marlin kernel
지원 프레임워크: vLLM, SGLang

특징

고속 추론: Marlin 커널을 활용해 일반 GPTQ보다 낮은 latency와 빠른 디코딩 속도 제공
메모리 최적화: 동일한 70B 모델을 더 적은 VRAM으로 실행 가능
추론 성능: 정확도 손실 최소화, 특히 대규모 텍스트 생성 및 RAG용도에 적합

Downloads last month: 7

Safetensors

Model size

10B params

Tensor type

I32

F16