File size: 2,305 Bytes
a26e606
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
from diffusers import StableDiffusionPipeline
import torch
from PIL import Image
import io

class ImageGenerator:
    def __init__(self, model_name="CompVis/stable-diffusion-v1-4"):
        # Explicit GPU detection and setup
        if torch.cuda.is_available():
            self.device = torch.device("cuda")
            print(f"Image Generator: Using GPU - {torch.cuda.get_device_name(0)}")
            print(f"GPU Memory: {torch.cuda.get_device_properties(0).total_memory / 1024**3:.2f} GB")
        else:
            self.device = torch.device("cpu")
            print("Image Generator: Using CPU")
        
        print(f"Loading model {model_name}...")
        self.pipe = StableDiffusionPipeline.from_pretrained(
            model_name,
            torch_dtype=torch.float16 if self.device.type == "cuda" else torch.float32,
            safety_checker=None,  # Disable safety checker for better performance
            variant="fp16" if self.device.type == "cuda" else None  # Use fp16 weights on GPU
        ).to(self.device)
        print(f"Model loaded and moved to {self.device}")
    
    def generate_image(self, prompt, num_inference_steps=30, guidance_scale=7.0):
        """
        Generate an image based on the given prompt
        
        Args:
            prompt (str): The text prompt to generate from
            num_inference_steps (int): Number of denoising steps
            guidance_scale (float): Scale for classifier-free guidance
            
        Returns:
            PIL.Image: Generated image
        """
        try:
            print(f"Generating image on {self.device}...")
            
            # Add quality prompts
            enhanced_prompt = f"{prompt}, high quality, detailed, 4k, professional photography"
            
            image = self.pipe(
                enhanced_prompt,
                num_inference_steps=num_inference_steps,
                guidance_scale=guidance_scale,
                negative_prompt="blurry, low quality, distorted, deformed",
                width=512,  # Reduced resolution for faster generation
                height=512  # Reduced resolution for faster generation
            ).images[0]
            
            return image
        except Exception as e:
            return f"Error generating image: {str(e)}"