prepare python environment

[Sascha]

Create virtual environment

1. Create a project folder
2. Cd to the project folder
3. Create virtual environment: conda create -n ai_painting python=3.10
4. Activate virtual environment: conda activate ai_painting

Install librarys

1. Install CPU torch pip install torch trochvision trochaudio
2. Install AI painting libraries pip install diffusers["torch"] transformers accelerate
3. Install web framework:pip install "fastapi[standard]"

Test Installing


Create a python file:


import torch
from diffusers import StableDiffusionPipeline
import gc # Garbage collector

print(" Setting up optimized AI system...")

# Force CPU and clear memory
device = "cpu"
torch.cuda.empty_cache() if torch.cuda.is_available() else None
gc.collect()

print(" Loading model with memory optimization...")

try:
# Load with memory optimizations
model_id = "runwayml/stable-diffusion-v1-5"

# Load to CPU with specific settings for stability
pipe = StableDiffusionPipeline.from_pretrained(
model_id,
torch_dtype=torch.float32, # Use float32 for CPU stability
use_safetensors=True
)
pipe = pipe.to(device)

# Disable the safety checker to save memory
pipe.safety_checker = None
pipe.requires_safety_checker = False

print(" Model loaded! Generating smaller test image...")

# Generate a smaller image to save memory
prompt = "a simple red apple on a table"

with torch.no_grad():
image = pipe(
prompt,
num_inference_steps=20, # Fewer steps = less memory
guidance_scale=7.5,
width=256, # Smaller image
height=256 # Smaller image
).images[0]

image.save("test_small.jpg")
print(" Success! Image saved as 'test_small.jpg'")

except Exception as e:
print(f" Error: {e}")

Источник:

Пожалуйста Авторизируйтесь или Зарегистрируйтесь для просмотра скрытого текста.