LLaVa-Next, leveraging mistralai/Mistral-7B-Instruct-v0.2 as LLM

The LLaVA-NeXT model was proposed in LLaVA-NeXT: Improved reasoning, OCR, and world knowledge by Haotian Liu, Chunyuan Li, Yuheng Li, Bo Li, Yuanhan Zhang, Sheng Shen, Yong Jae Lee. LLaVa-NeXT (also called LLaVa-1.6) improves upon LLaVa-1.5 by increasing the input image resolution and training on an improved visual instruction tuning dataset to improve OCR and common sense reasoning.

Disclaimer: The team releasing LLaVa-NeXT did not write a model card for this model so this model card has been written by the Hugging Face team.

Model description

LLaVa combines a pre-trained large language model with a pre-trained vision encoder for multimodal chatbot use cases. LLaVA 1.6 improves on LLaVA 1.5 BY:

• Using Mistral-7B (for this checkpoint) and Nous-Hermes-2-Yi-34B which has better commercial licenses, and bilingual support
• More diverse and high quality data mixture
• Dynamic high resolution

Intended uses & limitations

You can use the raw model for tasks like image captioning, visual question answering, multimodal chatbot use cases. See the model hub to look for other versions on a task that interests you.

How to use

Here's the prompt template for this model but we recomment to use the chat templates to format the prompt with processor.apply_chat_template(). That will apply the correct template for a given checkpoint for you.

"[INST] <image>\nWhat is shown in this image? [/INST]"

To run the model with the pipeline, see the below example:

from transformers import pipeline

pipe = pipeline("image-text-to-text", model="llava-hf/llava-v1.6-mistral-7b-hf")
messages = [
    {
      "role": "user",
      "content": [
          {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/tasks/ai2d-demo.jpg"},
          {"type": "text", "text": "What does the label 15 represent? (1) lava (2) core (3) tunnel (4) ash cloud"},
        ],
    },
]

out = pipe(text=messages, max_new_tokens=20)
print(out)
>>> [{'input_text': [{'role': 'user', 'content': [{'type': 'image', 'url': 'https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/tasks/ai2d-demo.jpg'}, {'type': 'text', 'text': 'What does the label 15 represent? (1) lava (2) core (3) tunnel (4) ash cloud'}]}], 'generated_text': 'Lava'}]

You can also load and use the model like following:

from transformers import LlavaNextProcessor, LlavaNextForConditionalGeneration
import torch
from PIL import Image
import requests

processor = LlavaNextProcessor.from_pretrained("llava-hf/llava-v1.6-mistral-7b-hf")

model = LlavaNextForConditionalGeneration.from_pretrained("llava-hf/llava-v1.6-mistral-7b-hf", torch_dtype=torch.float16, low_cpu_mem_usage=True) 
model.to("cuda:0")

# prepare image and text prompt, using the appropriate prompt template
url = "https://github.com/haotian-liu/LLaVA/blob/1a91fc274d7c35a9b50b3cb29c4247ae5837ce39/images/llava_v1_5_radar.jpg?raw=true"
image = Image.open(requests.get(url, stream=True).raw)

# Define a chat history and use `apply_chat_template` to get correctly formatted prompt
# Each value in "content" has to be a list of dicts with types ("text", "image") 
conversation = [
    {

      "role": "user",
      "content": [
          {"type": "text", "text": "What is shown in this image?"},
          {"type": "image"},
        ],
    },
]
prompt = processor.apply_chat_template(conversation, add_generation_prompt=True)

inputs = processor(images=image, text=prompt, return_tensors="pt").to("cuda:0")

# autoregressively complete prompt
output = model.generate(**inputs, max_new_tokens=100)

print(processor.decode(output[0], skip_special_tokens=True))

From transformers>=v4.48, you can also pass image url or local path to the conversation history, and let the chat template handle the rest. Chat template will load the image for you and return inputs in torch.Tensor which you can pass directly to model.generate()

messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://www.ilankelman.org/stopsigns/australia.jpg"},
            {"type": "text", "text": "What is shown in this image?"},
        ],
    },
]

inputs = processor.apply_chat_template(messages, add_generation_prompt=True, tokenize=True, return_dict=True, return_tensors="pt")
output = model.generate(**inputs, max_new_tokens=50)

Model optimization

4-bit quantization through bitsandbytes library

First make sure to install bitsandbytes, pip install bitsandbytes and make sure to have access to a CUDA compatible GPU device. Simply change the snippet above with:

model = LlavaNextForConditionalGeneration.from_pretrained(
    model_id, 
    torch_dtype=torch.float16, 
    low_cpu_mem_usage=True,
+   load_in_4bit=True
)

Use Flash-Attention 2 to further speed-up generation

First make sure to install flash-attn. Refer to the original repository of Flash Attention regarding that package installation. Simply change the snippet above with:

model = LlavaNextForConditionalGeneration.from_pretrained(
    model_id, 
    torch_dtype=torch.float16, 
    low_cpu_mem_usage=True,
+   use_flash_attention_2=True
).to(0)

BibTeX entry and citation info

@misc{liu2023improved,
      title={Improved Baselines with Visual Instruction Tuning}, 
      author={Haotian Liu and Chunyuan Li and Yuheng Li and Yong Jae Lee},
      year={2023},
      eprint={2310.03744},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}