sdk4.3.0+bi150上使用tensorflow框架训练resnet50报错：AttributeError: module 'tensorflow.python.framework.ops' has no attribute 'register_tensor_conversion_function'

## Bug Report - 故障报告
### Decription - 问题描述
sdk4.3.0+bi150上使用tensorflow框架训练resnet50报错：AttributeError: module 'tensorflow.python.framework.ops' has no attribute 'register_tensor_conversion_function'
### Error Log - 报错信息
bash run_train_resnet50_imagenette.sh
报错：
bash run_train_resnet50_imagenette.sh
```
2025-09-25 09:37:55.357769: E external/local_xla/xla/stream_executor/cuda/cuda_fft.cc:495] Unable to register cuFFT factory: Attempting to register factory for plugin cuFFT when one has already been registered
2025-09-25 09:37:55.387471: E external/local_xla/xla/stream_executor/cuda/cuda_dnn.cc:10626] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered
2025-09-25 09:37:55.387503: E external/local_xla/xla/stream_executor/cuda/cuda_blas.cc:1499] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered
WARNING:tensorflow:Deprecation warnings have been disabled. Set TF_ENABLE_DEPRECATION_WARNINGS=1 to re-enable them.
Traceback (most recent call last):
  File "/data/dk_125/sm_train/deepsparkhub/cv/classification/resnet50/tensorflow/tf_cnn_benchmarks.py", line 30, in <module>
    import benchmark_cnn
  File "/data/dk_125/sm_train/deepsparkhub/cv/classification/resnet50/tensorflow/benchmark_cnn.py", line 50, in <module>
    import variable_mgr
  File "/data/dk_125/sm_train/deepsparkhub/cv/classification/resnet50/tensorflow/variable_mgr.py", line 30, in <module>
    import variable_mgr_util
  File "/data/dk_125/sm_train/deepsparkhub/cv/classification/resnet50/tensorflow/variable_mgr_util.py", line 317, in <module>
    ops.register_tensor_conversion_function(
AttributeError: module 'tensorflow.python.framework.ops' has no attribute 'register_tensor_conversion_function'
```

DeepSpark/DeepSparkHub

内容风险标识

评论 (1)

DeepSpark/DeepSparkHub .gitee-modal { width: 500px !important; }

内容风险标识