GIT-Llama-2-7B

GIT-LLM: Generative Image to text Transformer with Large Language Models

Welcome to the GIT-LLM repository. GIT-LLM is an innovative fusion of the GIT Vision and Language model with the linguistic capabilities of the LLM (Language Learning Model). Harnessing the power of both worlds, this model is fine-tuned using the LoRA (Local Re-Attention) method, optimizing it for enhanced performance in diverse vision and language tasks.

Description of the uploaded weight

This model was trained for one epoch using M3IT (excluding videos and Chinese tasks). For more details, please refer to our blog (in Japanese).

Examples

Installation

1. Clone this repository

git clone https://github.com/Ino-Ichan/GIT-LLM
cd GIT-LLM

1. Install Packages

conda create -n git_llm python=3.10 -y
conda activate git_llm
pip install --upgrade pip # enable PEP 660 support

pip install -r requirements.txt
pip install -e .

For Llama 2

First, you request access to the llama-2 models, in huggingface page and facebook website

Please sign-in the huggingface account

huggingface-cli login

Training

Now we support LLaMA, MPT, and OPT as a LLM module.

./scripts/run.sh

Evaluation

You can get the pretrained weight form HuggingFace Hub: Inoichan/GIT-Llama-2-7B<br> See also notebooks.

import requests
from PIL import Image

import torch
from transformers import AutoProcessor
from git_llm.git_llama import GitLlamaForCausalLM

device_id = 0

# prepare a pretrained model
model = GitLlamaForCausalLM.from_pretrained('Inoichan/GIT-Llama-2-7B')
model.eval()
model.to(f"cuda:{device_id}")

# prepare a processor
processor = AutoProcessor.from_pretrained('Inoichan/GIT-Llama-2-7B')

# prepare inputs
url = "https://www.barnorama.com/wp-content/uploads/2016/12/03-Confusing-Pictures.jpg"
image = Image.open(requests.get(url, stream=True).raw)

text = f"##Instruction: Please answer the following question concletely. ##Question: What is unusual about this image? Explain precisely and concletely what he is doing? ##Answer: "

# do preprocessing
inputs = processor(
    text,
    image,
    return_tensors="pt",
    truncation=True,
)
inputs = {k: v.to(f"cuda:{device_id}") for k, v in inputs.items()}

# set eos token
eos_token_id_list = [
    processor.tokenizer.pad_token_id,
    processor.tokenizer.eos_token_id,
]

# do inference
with torch.no_grad():
    out = model.generate(**inputs, max_length=256, do_sample=False, temperature=0., eos_token_id=eos_token_id_list)

# print result
print(processor.tokenizer.batch_decode(out))