Instructions to use lighteternal/Llama3-merge-biomed-8b with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use lighteternal/Llama3-merge-biomed-8b with Transformers:

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

pipe = pipeline("text-generation", model="lighteternal/Llama3-merge-biomed-8b")
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("lighteternal/Llama3-merge-biomed-8b")
model = AutoModelForCausalLM.from_pretrained("lighteternal/Llama3-merge-biomed-8b", 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 lighteternal/Llama3-merge-biomed-8b with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "lighteternal/Llama3-merge-biomed-8b"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "lighteternal/Llama3-merge-biomed-8b",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/lighteternal/Llama3-merge-biomed-8b

SGLang

How to use lighteternal/Llama3-merge-biomed-8b 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 "lighteternal/Llama3-merge-biomed-8b" \
    --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": "lighteternal/Llama3-merge-biomed-8b",
		"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 "lighteternal/Llama3-merge-biomed-8b" \
        --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": "lighteternal/Llama3-merge-biomed-8b",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use lighteternal/Llama3-merge-biomed-8b with Docker Model Runner:
```
docker model run hf.co/lighteternal/Llama3-merge-biomed-8b
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

Llama3-merge-biomed-8b

This is a DARE-TIES Merge of Llama3-8b-Instruct + NousResearch/Hermes-2-Pro-Llama-3-8B + aaditya/Llama3-OpenBioLLM-8B. It is a simple experiment to assess whether combining models with strengths in general language understanding and biomedical knowledge can enhance performance on specialized tasks without compromising general applicability. The results indicate promising outcomes in areas like HendrycksTest tasks related to Biology and Medicine, as well as improvements in complex reasoning as seen in the ARC Challenge and Winogrande benchmarks.

Usage

I recommend using the prompt template of Llama3: https://llama.meta.com/docs/model-cards-and-prompt-formats/meta-llama-3/

Leaderboard metrics according to 🤗 Open LLM Leaderboard

Task	Metric	Ours (%)	Llama38BInstr. (%)	OpenBioLLM8B (%)
ARC Challenge	Accuracy	59.39	57.17	55.38
	Normalized Accuracy	63.65	60.75	58.62
Hellaswag	Accuracy	62.59	59.04	61.83
	Normalized Accuracy	81.53	78.55	80.76
Winogrande	Accuracy	75.93	74.51	70.88
GSM8K	Accuracy	59.36	68.69	10.15
HendrycksTest-Anatomy	Accuracy	72.59	65.19	69.62
HendrycksTest-Clinical Knowledge	Accuracy	77.83	74.72	60.38
HendrycksTest-College Biology	Accuracy	81.94	79.86	79.86
HendrycksTest-College Medicine	Accuracy	69.36	63.58	70.52
HendrycksTest-Medical Genetics	Accuracy	86.00	80.00	80.00
HendrycksTest-Professional Medicine	Accuracy	77.94	71.69	77.94

This is a merge of pre-trained language models created using mergekit.

Merge Details

Merge Method

This model was merged using the DARE TIES merge method using meta-llama/Meta-Llama-3-8B-Instruct as a base.

Models Merged

The following models were included in the merge:

Configuration

The following YAML configuration was used to produce this model:

models:
  - model: meta-llama/Meta-Llama-3-8B-Instruct
    # Base model providing a general foundation without specific parameters

  - model: meta-llama/Meta-Llama-3-8B-Instruct
    parameters:
      density: 0.60  
      weight: 0.5  

  - model: NousResearch/Hermes-2-Pro-Llama-3-8B
    parameters:
      density: 0.55  
      weight: 0.1  

  - model: aaditya/Llama3-OpenBioLLM-8B
    parameters:
      density: 0.55  
      weight: 0.4 

merge_method: dare_ties
base_model: meta-llama/Meta-Llama-3-8B-Instruct
parameters:
  int8_mask: true
dtype: bfloat16