Instructions to use aeolian83/Llama-3-8B-Instruct-cp-transfer_0.7 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use aeolian83/Llama-3-8B-Instruct-cp-transfer_0.7 with Transformers:

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

pipe = pipeline("text-generation", model="aeolian83/Llama-3-8B-Instruct-cp-transfer_0.7")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("aeolian83/Llama-3-8B-Instruct-cp-transfer_0.7")
model = AutoModelForCausalLM.from_pretrained("aeolian83/Llama-3-8B-Instruct-cp-transfer_0.7")
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

vLLM

How to use aeolian83/Llama-3-8B-Instruct-cp-transfer_0.7 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "aeolian83/Llama-3-8B-Instruct-cp-transfer_0.7"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "aeolian83/Llama-3-8B-Instruct-cp-transfer_0.7",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/aeolian83/Llama-3-8B-Instruct-cp-transfer_0.7

SGLang

How to use aeolian83/Llama-3-8B-Instruct-cp-transfer_0.7 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 "aeolian83/Llama-3-8B-Instruct-cp-transfer_0.7" \
    --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": "aeolian83/Llama-3-8B-Instruct-cp-transfer_0.7",
		"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 "aeolian83/Llama-3-8B-Instruct-cp-transfer_0.7" \
        --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": "aeolian83/Llama-3-8B-Instruct-cp-transfer_0.7",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use aeolian83/Llama-3-8B-Instruct-cp-transfer_0.7 with Docker Model Runner:
```
docker model run hf.co/aeolian83/Llama-3-8B-Instruct-cp-transfer_0.7
```

Model Card for Model ID

aeolian83/Llama-3-Open-Ko-8B-aeolian83-chatvec 모델은

chat-vector 논문( https://arxiv.org/abs/2310.04799v2 )에 근거하여,

llama3의 pre-trained 모델의 parameter와 instruction 모델의 매개변수의 차이를

beomi님의 Llama-3-Open-Ko-8B에 적용한 모델

이 방법이 가능하다면 llama3의 instruction 모델에

llama3의 pre-trained 모델과 한국어 CP모델인 Llama-3-Open-Ko-8B 모델의 매개변수 차이를

instruction 모델에 넣었을 때 어떻게 되는지 확인하는 모델

매개변수에 가중치를 0.7 줌

Metric

results/all/aeolian83/Llama-3-8B-Instruct-cp-transfer_0.7

	0	5
kobest_boolq (macro_f1)	0.786414	0.775213
kobest_copa (macro_f1)	0.67162	0.719676
kobest_hellaswag (macro_f1)	0.436275	0.423664
kobest_sentineg (macro_f1)	0.503784	0.901741
kohatespeech (macro_f1)	0.260168	0.339337
kohatespeech_apeach (macro_f1)	0.337667	0.611894
kohatespeech_gen_bias (macro_f1)	0.124535	0.505005
korunsmile (f1)	0.410636	0.323403
nsmc (acc)	0.53778	0.81346
pawsx_ko (acc)	0.5485	0.486

Used Model

Base model(weight diff를 구하기 위한 베이스 모델) : meta-llama/Meta-Llama-3-8B
Chat model(weight diff를 제공하는 cp model) : beomi/Llama-3-Open-Ko-8B
Target model(weight diff를 적용해서 instruction 튠을 하고자 하는 모델) : meta-llama/Meta-Llama-3-8B-Instruct

Downloads last month: 1

Safetensors

Model size

8B params

Tensor type

BF16

Paper for aeolian83/Llama-3-8B-Instruct-cp-transfer_0.7

Chat Vector: A Simple Approach to Equip LLMs With New Language Chat Capabilities

Paper • 2310.04799 • Published Oct 7, 2023