app.py

import os
import torch
from PIL import Image
from transformers import Blip2Processor, Blip2ForConditionalGeneration
import gradio as gr

# ---------------------------------------------------------
# ENV FIX (OPSİYONEL AMA TEMİZ)
# ---------------------------------------------------------
os.environ["OMP_NUM_THREADS"] = "1"

# ---------------------------------------------------------
# DEVICE
# ---------------------------------------------------------
device = "cuda" if torch.cuda.is_available() else "cpu"
dtype = torch.float16 if device == "cuda" else torch.float32

print(f"🚀 Device: {device} | dtype: {dtype}")

# ---------------------------------------------------------
# MODEL
# ---------------------------------------------------------
MODEL_NAME = "Salesforce/blip2-flan-t5-xl"
# Space çok çöküyorsa şuna düş:
# MODEL_NAME = "Salesforce/blip2-flan-t5-base"

print("⏳ Model yükleniyor...")

processor = Blip2Processor.from_pretrained(
    MODEL_NAME,
    use_fast=True
)

model = Blip2ForConditionalGeneration.from_pretrained(
    MODEL_NAME,
    torch_dtype=dtype,
    device_map="auto" if device == "cuda" else None
)

model.to(device)
model.eval()

print("✅ Model hazır!")

# ---------------------------------------------------------
# INFERENCE FUNCTION
# ---------------------------------------------------------
def generate_caption(image: Image.Image):
    if image is None:
        return "❌ Lütfen bir görsel yükleyin."

    image = image.convert("RGB")

    inputs = processor(image, return_tensors="pt").to(device)

    with torch.no_grad():
        output = model.generate(
            **inputs,
            max_new_tokens=40
        )

    caption = processor.decode(
        output[0],
        skip_special_tokens=True
    )

    return caption

# ---------------------------------------------------------
# GRADIO UI + API
# ---------------------------------------------------------
demo = gr.Interface(
    fn=generate_caption,
    inputs=gr.Image(type="pil", label="📷 Görsel Yükle"),
    outputs=gr.Textbox(label="📝 Üretilen Açıklama"),
    api_name="generate_caption",  # 🔴 API İSMİ
    title="BLIP-2 Image Captioning",
    description="BLIP-2 FLAN-T5 ile Image → Text"
)

demo.launch(
    server_name="0.0.0.0",
    server_port=7860,
    show_error=True
)