app.py

import subprocess
import sys

def install(package):
    subprocess.check_call([sys.executable, "-m", "pip", "install", package])

libraries = ["transformers", "diffusers", "accelerate", "ftfy", "safetensors"]

for library in libraries:
    try:
        install(library)
        print(f"{library} 安装成功")
    except subprocess.CalledProcessError as e:
        print(f"安装 {library} 时出错: {e}")
from prompt_to_prompt import prompt_to_prompt
from prompt_to_image import prompt_to_image

# 用户提供的简短文本描述
user_description = "a cute cat"
# 将简短描述转换为适合Stable Diffusion的提示词
generated_prompt = prompt_to_prompt(user_description)
print(f"Generated Prompt: {generated_prompt}")
# 使用提示词生成图像
image = prompt_to_image(generated_prompt)
# 保存生成的图像
image.save("generated_image.png")
print("Image saved as generated_image.png") 
from diffusers import StableDiffusionPipeline
import torch

# 加载Stable Diffusion模型
pipe = StableDiffusionPipeline.from_pretrained("runwayml/stable-diffusion-v1-5", torch_dtype=torch.float16)
# 将模型移动到GPU上（如果有可用的GPU）
pipe = pipe.to("cuda")

def prompt_to_image(prompt):
    """
    使用提示词生成图像
    :param prompt: 适合Stable Diffusion的提示词
    :return: 生成的图像
    """
    # 生成图像
    image = pipe(prompt).images[0]
    return image   
from transformers import AutoModelForCausalLM, AutoTokenizer

# 选择一个大语言模型，这里以GPT - Neo 125M为例
model_name = "EleutherAI/gpt-neo-125M"
# 加载分词器
tokenizer = AutoTokenizer.from_pretrained(model_name)
# 加载模型
model = AutoModelForCausalLM.from_pretrained(model_name)

def prompt_to_prompt(user_description):
    """
    将用户的简短描述转换为适合Stable Diffusion的提示词
    :param user_description: 用户提供的简短文本描述
    :return: 适合Stable Diffusion的提示词
    """
    # 将用户描述编码为模型可以接受的输入
    input_ids = tokenizer.encode(user_description, return_tensors='pt')
    # 生成文本
    output = model.generate(input_ids, max_length=50)
    # 将生成的文本解码为字符串
    generated_prompt = tokenizer.decode(output[0], skip_special_tokens=True)
    return generated_prompt