Pixart 900m 1024 Ft V0.6

🚀 pixart-900m-1024-ft

本项目是基于 ptx0/pixart-900m-1024-ft-large 的全秩微调模型。它能够根据输入的文本描述生成高质量的图像，在图像生成领域具有广泛的应用前景。

🚀 快速开始

你可以按照以下步骤进行推理：

import torch
from diffusers import DiffusionPipeline

model_id = 'pixart-900m-1024-ft'
prompt = 'ethnographic photography of teddy bear at a picnic, ears tucked behind a cozy hoodie looking darkly off to the stormy picnic skies'
negative_prompt = 'blurry, cropped, ugly'
pipeline = DiffusionPipeline.from_pretrained(model_id)
pipeline.to('cuda' if torch.cuda.is_available() else 'mps' if torch.backends.mps.is_available() else 'cpu')

prompt = "ethnographic photography of teddy bear at a picnic, ears tucked behind a cozy hoodie looking darkly off to the stormy picnic skies"
negative_prompt = "blurry, cropped, ugly"

pipeline = DiffusionPipeline.from_pretrained(model_id)
pipeline.to('cuda' if torch.cuda.is_available() else 'mps' if torch.backends.mps.is_available() else 'cpu')
image = pipeline(
    prompt=prompt,
    negative_prompt='blurry, cropped, ugly',
    num_inference_steps=25,
    generator=torch.Generator(device='cuda' if torch.cuda.is_available() else 'mps' if torch.backends.mps.is_available() else 'cpu').manual_seed(1641421826),
    width=1152,
    height=768,
    guidance_scale=4.5,
    guidance_rescale=0.0,
).images[0]
image.save("output.png", format="PNG")

✨ 主要特性

• 基于 ptx0/pixart-900m-1024-ft-large 进行全秩微调。
• 文本编码器未进行训练，可复用基础模型的文本编码器进行推理。

📚 详细文档

验证设置

• CFG: 4.5
• CFG Rescale: 0.0
• Steps: 25
• Sampler: None
• Seed: 42
• Resolutions: 1024x1024,1344x768,916x1152

注意：验证设置不一定与训练设置相同。

你可以在以下图库中找到一些示例图像：

训练设置

数据集

photo-concept-bucket

💻 使用示例

基础用法

import torch
from diffusers import DiffusionPipeline

model_id = 'pixart-900m-1024-ft'
prompt = 'ethnographic photography of teddy bear at a picnic, ears tucked behind a cozy hoodie looking darkly off to the stormy picnic skies'
negative_prompt = 'blurry, cropped, ugly'
pipeline = DiffusionPipeline.from_pretrained(model_id)
pipeline.to('cuda' if torch.cuda.is_available() else 'mps' if torch.backends.mps.is_available() else 'cpu')

prompt = "ethnographic photography of teddy bear at a picnic, ears tucked behind a cozy hoodie looking darkly off to the stormy picnic skies"
negative_prompt = "blurry, cropped, ugly"

pipeline = DiffusionPipeline.from_pretrained(model_id)
pipeline.to('cuda' if torch.cuda.is_available() else 'mps' if torch.backends.mps.is_available() else 'cpu')
image = pipeline(
    prompt=prompt,
    negative_prompt='blurry, cropped, ugly',
    num_inference_steps=25,
    generator=torch.Generator(device='cuda' if torch.cuda.is_available() else 'mps' if torch.backends.mps.is_available() else 'cpu').manual_seed(1641421826),
    width=1152,
    height=768,
    guidance_scale=4.5,
    guidance_rescale=0.0,
).images[0]
image.save("output.png", format="PNG")

高级用法

你可以根据需要调整 prompt、negative_prompt、num_inference_steps、width、height、guidance_scale 和 guidance_rescale 等参数，以获得不同风格和质量的图像。

import torch
from diffusers import DiffusionPipeline

model_id = 'pixart-900m-1024-ft'
# 自定义提示
prompt = 'A beautiful sunset over the ocean'
negative_prompt = 'ugly, blurry'
pipeline = DiffusionPipeline.from_pretrained(model_id)
pipeline.to('cuda' if torch.cuda.is_available() else 'mps' if torch.backends.mps.is_available() else 'cpu')

image = pipeline(
    prompt=prompt,
    negative_prompt=negative_prompt,
    num_inference_steps=30,  # 增加推理步数
    generator=torch.Generator(device='cuda' if torch.cuda.is_available() else 'mps' if torch.backends.mps.is_available() else 'cpu').manual_seed(42),
    width=1280,  # 调整图像宽度
    height=720,  # 调整图像高度
    guidance_scale=5.0,  # 调整引导比例
    guidance_rescale=0.1,  # 调整引导重新缩放比例
).images[0]
image.save("custom_output.png", format="PNG")

📄 许可证

本项目采用 creativeml-openrail-m 许可证。