Instructions to use MBZUAI/LLaVA-Meta-Llama-3-8B-Instruct-FT-S2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use MBZUAI/LLaVA-Meta-Llama-3-8B-Instruct-FT-S2 with Transformers:

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

pipe = pipeline("text-generation", model="MBZUAI/LLaVA-Meta-Llama-3-8B-Instruct-FT-S2")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoModelForCausalLM
model = AutoModelForCausalLM.from_pretrained("MBZUAI/LLaVA-Meta-Llama-3-8B-Instruct-FT-S2", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use MBZUAI/LLaVA-Meta-Llama-3-8B-Instruct-FT-S2 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "MBZUAI/LLaVA-Meta-Llama-3-8B-Instruct-FT-S2"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "MBZUAI/LLaVA-Meta-Llama-3-8B-Instruct-FT-S2",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/MBZUAI/LLaVA-Meta-Llama-3-8B-Instruct-FT-S2

SGLang

How to use MBZUAI/LLaVA-Meta-Llama-3-8B-Instruct-FT-S2 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 "MBZUAI/LLaVA-Meta-Llama-3-8B-Instruct-FT-S2" \
    --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": "MBZUAI/LLaVA-Meta-Llama-3-8B-Instruct-FT-S2",
		"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 "MBZUAI/LLaVA-Meta-Llama-3-8B-Instruct-FT-S2" \
        --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": "MBZUAI/LLaVA-Meta-Llama-3-8B-Instruct-FT-S2",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use MBZUAI/LLaVA-Meta-Llama-3-8B-Instruct-FT-S2 with Docker Model Runner:
```
docker model run hf.co/MBZUAI/LLaVA-Meta-Llama-3-8B-Instruct-FT-S2
```

LLaMA-3-V: Extending the Visual Capabilities of LLaVA with Meta-Llama-3-8B-Instruct

Repository Overview

This repository features LLaVA v1.5 trained with the Meta-Llama-3-8B-Instruct LLM. This integration aims to leverage the strengths of both models to offer advanced vision-language understanding.

Training Strategy

Pretraining: Only Vision-to-Language projector is trained. The rest of the model is frozen.
Fine-tuning: All model parameters including LLM are fine-tuned. Only the vision-backbone (CLIP) is kept frozen.
Note: During both pretraining and fine-tuning, the vision-backbone (CLIP) is augmented with multi-scale features following S2-Wrapper.

Key Components

Base Large Language Model (LLM): Meta-Llama-3-8B-Instruct
Base Large Multimodal Model (LMM): LLaVA-v1.5

Training Data

Pretraining Dataset: LCS-558K
Fine-tuning Dataset: LLaVA-Instruct-665K

Download It As

git lfs install
git clone https://huggingface.co/MBZUAI/LLaVA-Meta-Llama-3-8B-Instruct-FT-S2

Contributions

Contributions are welcome! Please 🌟 our repository LLaVA++ if you find this model useful.

Downloads last month: 20

Safetensors

Model size

8B params

Tensor type

BF16

Collection including MBZUAI/LLaVA-Meta-Llama-3-8B-Instruct-FT-S2

LLaVA++ (LLaMA-3 and Phi-3-Mini)

Collection

Extending Visual Capabilities of LLaVA with LLaMA-3 and Phi-3 • 11 items • Updated Jun 11, 2024 • 23

Paper for MBZUAI/LLaVA-Meta-Llama-3-8B-Instruct-FT-S2

When Do We Not Need Larger Vision Models?

Paper • 2403.13043 • Published Mar 19, 2024 • 26