finetuned_paligemma_vqav2_small

This model is a fine-tuned version of google/paligemma-3b-pt-224 using the QLoRA technique on a small chunk of vqav2 dataset by Merve.

How to Use?

import torch
import requests

from PIL import Image
from transformers import AutoProcessor, PaliGemmaForConditionalGeneration

pretrained_model_id = "google/paligemma-3b-pt-224"
finetuned_model_id = "pyimagesearch/finetuned_paligemma_vqav2_small"

processor = AutoProcessor.from_pretrained(pretrained_model_id)
finetuned_model = PaliGemmaForConditionalGeneration.from_pretrained(finetuned_model_id)

prompt = "What is behind the cat?"
image_file = "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/cat.png?download=true"
raw_image = Image.open(requests.get(image_file, stream=True).raw)

inputs = processor(raw_image.convert("RGB"), prompt, return_tensors="pt")
output = finetuned_model.generate(**inputs, max_new_tokens=20)

print(processor.decode(output[0], skip_special_tokens=True)[len(prompt):])
# gramophone

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 2e-05
• train_batch_size: 4
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 2
• num_epochs: 2

Training results

Framework versions

• PEFT 0.13.0
• Transformers 4.46.0.dev0
• Pytorch 2.4.1+cu121
• Datasets 3.0.1
• Tokenizers 0.20.0