from datasets import load_dataset
from unsloth import FastVisionModel

model, tokenizer = FastVisionModel.from_pretrained(
    model_name = "MMoshtaghi/Qwen2-VL-7B-Instruct-LoRAAdpt-MathOCR",
    load_in_4bit = True,
)
FastVisionModel.for_inference(model) # Enable for inference!

dataset = load_dataset("unsloth/LaTeX_OCR", split = "train")
image = dataset[0]["image"]
instruction = "Write the LaTeX representation for this image."

messages = [
    {"role": "user", "content": [
        {"type": "image"},
        {"type": "text", "text": instruction}
    ]}
]
input_text = tokenizer.apply_chat_template(messages, add_generation_prompt = True)
inputs = tokenizer(
    image,
    input_text,
    add_special_tokens = False,
    return_tensors = "pt",
).to("cuda")

from transformers import TextStreamer
text_streamer = TextStreamer(tokenizer, skip_prompt = True)
_ = model.generate(**inputs, streamer = text_streamer, max_new_tokens = 128,
                   use_cache = True, temperature = 1.5, min_p = 0.1)

Framework versions

Developed by: MMoshtaghi
License: apache-2.0
Finetuned from model: unsloth/qwen2-vl-7b-instruct-unsloth-bnb-4bit
Finetuned on dataset: <a href="https://huggingface.co/datasets/unsloth/LaTeX_OCR" target="_blank" rel="noopener noreferrer" class="text-blue-400 hover:text-blue-300 underline underline-offset-2 transition-colors">unsloth/LaTeX_OCR
PEFT method: <a href="https://huggingface.co/papers/2305.14314" target="_blank" rel="noopener noreferrer" class="text-blue-400 hover:text-blue-300 underline underline-offset-2 transition-colors">Quantized LoRA

TRL: 0.13.0
Transformers: 4.47.1
Pytorch: 2.5.1+cu121
Datasets: 3.2.0
Tokenizers: 0.21.0
Unsloth: 2025.1.5

Training procedure

(Log-in required!)

Visualize in Weights & Biases

Citations

This VLM model was trained 2x faster with Unsloth and Huggingface's TRL library.