Spaces:
Sleeping
Sleeping
| from diffusers import StableDiffusionXLPipeline, StableDiffusionXLImg2ImgPipeline | |
| from diffusers.utils import load_image | |
| import torch | |
| import time | |
| import utilities as u | |
| import card_generator as card | |
| from PIL import Image | |
| start_time = time.time() | |
| torch.backends.cuda.matmul.allow_tf32 = True | |
| model_path = ("../../models/stable-diffusion/SDXLFaetastic_v24.safetensors") | |
| lora_path = "../../models/stable-diffusion/Loras/blank-card-template-5.safetensors" | |
| detail_lora_path = "../../models/stable-diffusion/Loras/add-detail-xl.safetensors" | |
| mimic_lora_path = "../../models/stable-diffusion/Loras/EnvyMimicXL01.safetensors" | |
| card_pre_prompt = " blank magic card,high resolution, detailed intricate high quality border, textbox, high quality magnum opus drawing of a " | |
| negative_prompts = "text, words, numbers, letters" | |
| image_list = [] | |
| def generate_image(num_img, prompt, item, user_input_template, mimic = None) : | |
| prompt = card_pre_prompt + item + ' ' + prompt | |
| print(prompt) | |
| image_path = f"card_templates/{user_input_template}" | |
| init_image = load_image(image_path).convert("RGB") | |
| pipe = StableDiffusionXLImg2ImgPipeline.from_single_file(model_path, | |
| custom_pipeline="low_stable_diffusion", | |
| torch_dtype=torch.float16, | |
| variant="fp16", | |
| local_files_only = True).to("cuda") | |
| # Load LoRAs for controlling image | |
| pipe.load_lora_weights(lora_path, weight_name = "blank-card-template-5.safetensors",adapter_name = 'blank-card-template') | |
| pipe.load_lora_weights(detail_lora_path, weight_name = "add-detail-xl.safetensors", adapter_name = "add-detail-xl") | |
| # If mimic keyword has been detected, load the mimic LoRA and set adapter values | |
| if mimic: | |
| pipe.load_lora_weights(mimic_lora_path, weight_name = "EnvyMimicXL01.safetensors", adapter_name = "EnvyMimicXL") | |
| pipe.set_adapters(['blank-card-template', "add-detail-xl", "EnvyMimicXL"], adapter_weights = [0.9,0.9,1.0]) | |
| else : | |
| pipe.set_adapters(['blank-card-template', "add-detail-xl"], adapter_weights = [0.9,0.9]) | |
| pipe.enable_vae_slicing() | |
| for x in range(num_img): | |
| img_start = time.time() | |
| image = pipe(prompt=prompt, | |
| strength = .9, | |
| guidance_scale = 5, | |
| image= init_image, | |
| negative_promt = negative_prompts, | |
| num_inference_steps=50, | |
| height = 1024, width = 768).images[0] | |
| image = image.save(str(x) + f"{item}.png") | |
| output_image_path = str(x) + f"{item}.png" | |
| img_time = time.time() - img_start | |
| img_its = 50/img_time | |
| print(f"image gen time = {img_time} and {img_its} it/s") | |
| print(f"Memory after image {x} = {torch.cuda.memory_allocated()}") | |
| image_list.append(output_image_path) | |
| # Delete the image variable to keep VRAM open to load the LLM | |
| del image | |
| print(f"Memory after del {torch.cuda.memory_allocated()}") | |
| print(image_list) | |
| total_time = time.time() - start_time | |
| print(f"Total Time to generate{x} images = {total_time} ") | |
| del pipe | |
| u.reclaim_mem() | |
| return image_list | |