Spaces:
Sleeping
Sleeping
File size: 3,201 Bytes
2480131 | 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 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 | from PIL import Image
import os
import ollama
import torch
from diffusers import DiffusionPipeline, AutoPipelineForImage2Image, LCMScheduler, AutoPipelineForText2Image
import time
model_id = "simianluo/lcm_dreamshaper_v7"
class GenerationSession:
def __init__(self, model_id):
self.model_id = model_id
self.txt2img_pipeline = None
self.img2img_pipeline = None
self.current_image = None
self.current_prompt = None
self._initialize_pipelines()
def _initialize_pipelines(self):
print("initializing pipelines...")
self.txt2img_pipeline = DiffusionPipeline.from_pretrained(
model_id,
torch_dtype = torch.float16,
safety_checker = None
)
self.txt2img_pipeline.scheduler = LCMScheduler.from_config(self.txt2img_pipeline.scheduler.config)
self.txt2img_pipeline.to("cuda")
self.txt2img_pipeline.enable_attention_slicing()
self.txt2img_pipeline.enable_vae_slicing()
# self.txt2img_pipeline.unet = torch.compile(
# self.txt2img_pipeline.unet,
# mode = "reduce-overhead",
# fullgraph = True
#)
print("Text 2 image pipeline loaded and compiled.")
self.img2img_pipeline = AutoPipelineForImage2Image.from_pipe(self.txt2img_pipeline)
print("Image 2 image pipeline loaded (shared weights).")
def GeneratingBaseImage(self, prompt: str, negative_prompt: str = "Blurry, low quality, static and distorted image") -> str:
start = time.time()
image = self.txt2img_pipeline(
prompt = prompt,
negative_prompt= negative_prompt,
num_inference_steps = 4,
guidance_scale = 1.0,
height = 512,
width = 512
).images
print(f"Text to image generated in [{time.time() - start:.2f}s]")
return image
def GeneratingVariationImage(self, prompt: str, reference_image: Image.Image, strength: float = 0.5, negative_prompt: str = "Blurry, low quality, static and distorted image") -> str:
start = time.time()
image = self.img2img_pipeline(
prompt = prompt,
image = reference_image,
strength = strength,
num_inference_steps = 4,
guidance_scale = 1.0,
negative_prompt = negative_prompt
).images
print(f"Image to image generated in [{time.time() - start:.2f}s]")
return image
def Generate(self, new_prompt: str, strength: float = 0.5):
if self.current_image is None:
self.current_image = self.GeneratingBaseImage(new_prompt)
else:
self.current_image = self.GeneratingVariationImage(new_prompt, self.current_image, strength)
self.current_prompt = new_prompt
return self.current_image
def reset(self):
self.current_image = None
self.current_prompt = None
print("Session reset. Ready for new generation.")
|