ReAotor自定义节点在RTX50显卡安装最新pytorch2.8 cuda12.8的情况下报错

### Feature Idea

ReActor加载初始化错误
[ReActor] - STATUS - Running v0.6.0-a1 in ComfyUI
Torch version: 2.8.0.dev20250325+cu128
2025-03-30 22:16:43.304855: F tensorflow/python/lib/core/py_exception_registry.cc:28] Check failed: singleton_ == nullpt
r PyExceptionRegistry::Init() already called

测试正常，非tensorflow本身的问题
Python 3.10.16 | packaged by Anaconda, Inc. | (main, Dec 11 2024, 16:19:12) [MSC v.1929 64 bit (AMD64)] on win32
Type "help", "copyright", "credits" or "license" for more information.
>>> import os
>>> os.environ["CUDA_VISIBLE_DEVICES"] = "-1"
>>> import tensorflow as tf
2025-03-30 22:34:27.949432: I tensorflow/core/util/port.cc:113] oneDNN custom operations are on. You may see slightly different numerical results due to floating-point round-off errors from different computation orders. To turn them off, se
t the environment variable `TF_ENABLE_ONEDNN_OPTS=0`.
2025-03-30 22:34:28.493710: I tensorflow/core/util/port.cc:113] oneDNN custom operations are on. You may see slightly di
fferent numerical results due to floating-point round-off errors from different computation orders. To turn them off, se
t the environment variable `TF_ENABLE_ONEDNN_OPTS=0`.
>>>
>>> try:
... print(tf.__version__)
... print(tf.reduce_sum(tf.constant([1, 2, 3]))) # 应输出 6 ... except Exception as e: ... print("错误:", e)
... 2.16.2 2025-03-30 22:34:31.459697: I tensorflow/core/platform/cpu_feature_guard.cc:210] This TensorFlow binary is optimized to
use available CPU instructions in performance-critical operations.
To enable the following instructions: AVX2 AVX_VNNI AVX_VNNI_INT8 AVX_NE_CONVERT FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags.
tf.Tensor(6, shape=(), dtype=int32) >>>
>>>

### Existing Solutions

_No response_

### Other

_No response_