Instructions to use ChenWeiLi/MedLlama-3-8B_DARE with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use ChenWeiLi/MedLlama-3-8B_DARE with Transformers:

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

pipe = pipeline("text-generation", model="ChenWeiLi/MedLlama-3-8B_DARE")

# Load model directly
from transformers import AutoTokenizer, AutoModelForMultimodalLM

tokenizer = AutoTokenizer.from_pretrained("ChenWeiLi/MedLlama-3-8B_DARE")
model = AutoModelForMultimodalLM.from_pretrained("ChenWeiLi/MedLlama-3-8B_DARE")

Inference
Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use ChenWeiLi/MedLlama-3-8B_DARE with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "ChenWeiLi/MedLlama-3-8B_DARE"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "ChenWeiLi/MedLlama-3-8B_DARE",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/ChenWeiLi/MedLlama-3-8B_DARE

SGLang

How to use ChenWeiLi/MedLlama-3-8B_DARE 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 "ChenWeiLi/MedLlama-3-8B_DARE" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "ChenWeiLi/MedLlama-3-8B_DARE",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

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 "ChenWeiLi/MedLlama-3-8B_DARE" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "ChenWeiLi/MedLlama-3-8B_DARE",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use ChenWeiLi/MedLlama-3-8B_DARE with Docker Model Runner:
```
docker model run hf.co/ChenWeiLi/MedLlama-3-8B_DARE
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

medLlama-3-8B_DARE

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 mlabonne/ChimeraLlama-3-8B-v3 as a base.

Models Merged

The following models were included in the merge:

Evaluation

multimedq (0 shot)

Tasks	Version	Filter	Metric	Value		Stderr
- medmcqa	Yaml	none	acc	0.5728	±	0.0076
		none	acc_norm	0.5728	±	0.0076
- medqa_4options	Yaml	none	acc	0.5923	±	0.0138
		none	acc_norm	0.5923	±	0.0138
- anatomy (mmlu)	0	none	acc	0.7111	±	0.0392
- clinical_knowledge (mmlu)	0	none	acc	0.7547	±	0.0265
- college_biology (mmlu)	0	none	acc	0.7917	±	0.0340
- college_medicine (mmlu)	0	none	acc	0.6647	±	0.0360
- medical_genetics (mmlu)	0	none	acc	0.8200	±	0.0386
- professional_medicine (mmlu)	0	none	acc	0.7426	±	0.0266
stem	N/A	none	acc_norm	0.5773	±	0.0067
		none	acc	0.6145	±	0.0057
- pubmedqa	1	none	acc	0.7400	±	0.0196

Groups	Version	Filter	n-shot	Metric	Value		Stderr
stem	N/A	none	0	acc_norm	0.5773	±	0.0067
		none	0	acc	0.6145	±	0.0057

Configuration

The following YAML configuration was used to produce this model:

models:
  - model: mlabonne/ChimeraLlama-3-8B-v3
    # No parameters necessary for base model

  - model: sethuiyer/Medichat-Llama3-8B
    parameters:
      density: 0.53
      weight: 0.5
  - model: johnsnowlabs/JSL-MedLlama-3-8B-v2.0
    parameters:
      density: 0.53
      weight: 0.5
      
merge_method: dare_ties
base_model: mlabonne/ChimeraLlama-3-8B-v3
parameters:
  int8_mask: true
dtype: float16