Instructions to use beomi/EXAONE-3.5-32B-Instruct-Llamafied with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use beomi/EXAONE-3.5-32B-Instruct-Llamafied with Transformers:

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

pipe = pipeline("text-generation", model="beomi/EXAONE-3.5-32B-Instruct-Llamafied")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("beomi/EXAONE-3.5-32B-Instruct-Llamafied")
model = AutoModelForCausalLM.from_pretrained("beomi/EXAONE-3.5-32B-Instruct-Llamafied")
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 beomi/EXAONE-3.5-32B-Instruct-Llamafied with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "beomi/EXAONE-3.5-32B-Instruct-Llamafied"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "beomi/EXAONE-3.5-32B-Instruct-Llamafied",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/beomi/EXAONE-3.5-32B-Instruct-Llamafied

SGLang

How to use beomi/EXAONE-3.5-32B-Instruct-Llamafied 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 "beomi/EXAONE-3.5-32B-Instruct-Llamafied" \
    --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": "beomi/EXAONE-3.5-32B-Instruct-Llamafied",
		"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 "beomi/EXAONE-3.5-32B-Instruct-Llamafied" \
        --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": "beomi/EXAONE-3.5-32B-Instruct-Llamafied",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use beomi/EXAONE-3.5-32B-Instruct-Llamafied with Docker Model Runner:
```
docker model run hf.co/beomi/EXAONE-3.5-32B-Instruct-Llamafied
```

양자화 관련 질문

by netnk9151 - opened Dec 9, 2024

Discussion

netnk9151

Dec 9, 2024

안녕하세요

Llamafied된 모델이라는 게 기존 모델과 성능적으로 어떤 차이가 있는지 궁금해서 양자화 해서 사용해보려고 하는데 괜찮으실까요?

불쾌하시다면 삭제하겠습니다.

beomi

Owner Dec 9, 2024

안녕하세요,
해당 모델은 Exaone모델이 trust_remote_code를 사용해야한다는 불편함때문에 Llama모델의 이름에 맞춰 파라미터 이름을 변경, 그리고 기존 FP32를 BF16으로 downcasting해서 올린 weight입니다.
성능적으로는 거의 차이가 없을거에요 :)

물론 기존 Exaone에서 GGUF와 AWQ 양자화된 모델을 제공해주고 있어서 호환성을 이미 제공하고 있긴 합니다만, 다른 Llama 기반 하드코딩된 라이브러리에서 좀더 쓰기 편하게 만든것이라 보시면 됩니다.

beomi changed discussion status to closed Dec 9, 2024

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment