%%capture
!pip install -U bitsandbytes

from transformers import AutoProcessor, AutoModelForImageTextToText
import torch
DEVICE = "cuda" if torch.cuda.is_available() else "cpu"

processor = AutoProcessor.from_pretrained("manifestasi/RetinaVLM-300M-DPO")
model = AutoModelForImageTextToText.from_pretrained("manifestasi/RetinaVLM-300M-DPO",
                                                torch_dtype=torch.float16,
                                                _attn_implementation="eager").to(DEVICE)

from PIL import Image
from transformers.image_utils import load_image


# Load images
# image1 = load_image("https://huggingface.co/spaces/HuggingFaceTB/SmolVLM/resolve/main/example_images/rococo.jpg")
image2 = load_image("/kaggle/input/bandaraaa/799269_1200.jpg")

# Create input messages
messages = [
    {
        "role": "user",
        "content": [
            # {"type": "image"},
            {"type": "image"},
            {"type": "text", 
             "text": """
             Instructions :
             you are visual assistant for blind people, please answer politely and short 
             under 100 words.
             Prompt :
             can you direct me to find toilet
             """}
        ]
    },
]

# Prepare inputs
prompt = processor.apply_chat_template(messages, add_generation_prompt=True)
# inputs = processor(text=prompt, return_tensors="pt")
inputs = processor(text=prompt, images=[image2], return_tensors="pt")
inputs = inputs.to(DEVICE)
# Generate outputs
from time import time

tim1 = time()
generated_ids = model.generate(**inputs, max_new_tokens=120)
generated_texts = processor.batch_decode(
    generated_ids,
    skip_special_tokens=True,
)
tim2 = time()
print(f"{(tim2 - tim1)} detik")
print(generated_texts[0].split("Assistant: ")[1])

Training procedure

Visualize in Weights & Biases

This model was trained with DPO, a method introduced in Direct Preference Optimization: Your Language Model is Secretly a Reward Model.

Framework versions

TRL: 0.19.0
Transformers: 4.53.0
Pytorch: 2.6.0+cu124
Datasets: 3.6.0
Tokenizers: 0.21.2

Citations

Cite DPO as:

bibtex

@inproceedings{rafailov2023direct,
    title        = {{Direct Preference Optimization: Your Language Model is Secretly a Reward Model}},
    author       = {Rafael Rafailov and Archit Sharma and Eric Mitchell and Christopher D. Manning and Stefano Ermon and Chelsea Finn},
    year         = 2023,
    booktitle    = {Advances in Neural Information Processing Systems 36: Annual Conference on Neural Information Processing Systems 2023, NeurIPS 2023, New Orleans, LA, USA, December 10 - 16, 2023},
    url          = {http://papers.nips.cc/paper_files/paper/2023/hash/a85b405ed65c6477a4fe8302b5e06ce7-Abstract-Conference.html},
    editor       = {Alice Oh and Tristan Naumann and Amir Globerson and Kate Saenko and Moritz Hardt and Sergey Levine},
}

Cite TRL as:

bibtex

@misc{vonwerra2022trl,
	title        = {{TRL: Transformer Reinforcement Learning}},
	author       = {Leandro von Werra and Younes Belkada and Lewis Tunstall and Edward Beeching and Tristan Thrush and Nathan Lambert and Shengyi Huang and Kashif Rasul and Quentin Gallou{\'e}dec},
	year         = 2020,
	journal      = {GitHub repository},
	publisher    = {GitHub},
	howpublished = {\url{https://github.com/huggingface/trl}}
}