1. 报错信息


WARNING[XFORMERS]: xFormers can't load C++/CUDA extensions. xFormers was built for:

    PyTorch 2.4.1+cu121 with CUDA 1201 (you have 2.7.1+cu126)

    Python  3.9.20 (you have 3.9.23)

  Please reinstall xformers (see https://github.com/facebookresearch/xformers#installing-xformers)

  Memory-efficient attention, SwiGLU, sparse and more won't be available.

  Set XFORMERS_MORE_DETAILS=1 for more details

Traceback (most recent call last):

  File "<frozen importlib._bootstrap>", line 1030, in gcdimport

  File "<frozen importlib._bootstrap>", line 1007, in findand_load

  File "<frozen importlib._bootstrap>", line 972, in findand_load_unlocked

  File "<frozen importlib._bootstrap>", line 228, in callwith_frames_removed

  File "<frozen importlib._bootstrap>", line 1030, in gcdimport

  File "<frozen importlib._bootstrap>", line 1007, in findand_load

  File "<frozen importlib._bootstrap>", line 986, in findand_load_unlocked

  File "<frozen importlib._bootstrap>", line 680, in loadunlocked

  File "<frozen importlib._bootstrap_external>", line 850, in exec_module

  File "<frozen importlib._bootstrap>", line 228, in callwith_frames_removed

AttributeError: Cannot set attribute 'src' directly. Use '_unsafe_update_src()' and manually clear .hash of all callersinstead.

The above exception was the direct cause of the following exception:

Traceback (most recent call last):

RuntimeError: Failed to import diffusers.models.autoencoders.autoencoder_kl because of the following error (look up to see its traceback):

Cannot set attribute 'src' directly. Use ' _unsafe_update_src() ' and manually clear .hash of all callersinstead.

运行时错误：由于以下错误，无法导入diffusers.models.autoencoders.autoencoder_kl（查找以查看其回溯）：无法直接设置属性“src”。使用“_unsafe_update_src（）”并手动清除所有调用者的.hash。

2. 解决方案

warning先提示xformer版本不兼容，参照官方文档进行重新安装，我这里是cuda12.6版本：

# [linux & win] cuda 12.6 version
pip3 install -U xformers --index-url https://download.pytorch.org/whl/cu126

显然由于某些原因还需要换源：


# [linux & win] cuda 12.6 version
pip3 install -U xformers --index-url https://mirrors.nju.edu.cn/whl/cu126

南大源似乎并没有更新xformers，这里使用的官方指令安装，不然会提示安装torch2.4版本的：


ERROR: Could not find a version that satisfies the requirement torch==2.4.1 (from xformers) (from versions: none)
ERROR: No matching distribution found for torch==2.4.1

使用官方地址顺利更新到了xformers的0.0.31版本，问题解决。