Compile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.

Compile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.

RuntimeError: CUDA error: invalid device ordinal
CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.
For debugging consider passing CUDA_LAUNCH_BLOCKING=1
Compile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.

原因：

os.environ['CUDA_VISIBLE_DEVICES'] = '2'

然后绑定显卡：

import torch
from transformers import StableDiffusionUpscalePipeline
 
# 指定模型名称和数据类型
model_name = "your_model_name"
torch_dtype = torch.float16  # 或者 torch.float32, 取决于你的需求
 
# 加载模型并将其移动到第二个GPU（cuda:1）
pipe = StableDiffusionUpscalePipeline.from_pretrained(model_name, torch_dtype=torch_dtype).to("cuda:1")

 
# 现在 pipe 已经在第二个显卡上
代码解读
解决方法：
注释掉：

os.environ['CUDA_VISIBLE_DEVICES'] = '2'