app.py

from huggingface_hub import login
import os
from peft import PeftModel, PeftConfig
from transformers import AutoProcessor, PaliGemmaForConditionalGeneration
from PIL import Image
import requests
import torch
import io
import base64
import cv2

access_token = os.environ["HF_TOKEN"]
login(token=access_token)

device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
dtype = torch.bfloat16

config = PeftConfig.from_pretrained("anushettypsl/paligemma_vqav2")
# base_model = AutoModelForCausalLM.from_pretrained("google/paligemma-3b-pt-448")
base_model = PaliGemmaForConditionalGeneration.from_pretrained("google/paligemma-3b-pt-448")
model = PeftModel.from_pretrained(base_model, "anushettypsl/paligemma_vqav2", device_map=device)
processor = AutoProcessor.from_pretrained("google/paligemma-3b-pt-448", device_map=device)
model.to(device)

image = cv2.imread('/content/15_BC_G2_6358_40x_2_jpg.rf.97595fa4965f66ad45be8fd055331933.jpg')

# Convert the image to base64 encoding
image_bytes = cv2.imencode('.jpg', image)[1]
base64_string = base64.b64encode(image_bytes).decode('utf-8')

input_image = Image.open(io.BytesIO(image_bytes)).convert('RGB')

model_inputs = processor(
    text=input_text, images=input_image, return_tensors="pt").to(device)
input_len = model_inputs["input_ids"].shape[-1]
model.to(device)
with torch.inference_mode():
    generation = model.generate(
        **model_inputs, max_new_tokens=100, do_sample=False)
    generation = generation[0][input_len:]
    decoded = processor.decode(generation, skip_special_tokens=True)
    print(decoded)