is that backbone style modle support variation dropout layers to get feature importance ?

[immusferr]

Is that backbone style modle support variation dropout layers to get feature importance I am current using backbone to build my model , everything seems well. However, when I going to export the feature importance of variation layers , I got : Traceback (most recent call last): File "/home/data/miniconda/envs/py3/lib/python3.8/runpy.py", line 194, in _run_module_as_main return _run_code(code, main_globals, None, File "/home/data/miniconda/envs/py3/lib/python3.8/runpy.py", line 87, in _run_code exec(code, run_globals) File "/ofs/llm-for-dptd/huxinlei/general-prediction-server/easy_rec/python/tools/feature_selection.py", line 307, in fs.process() File "/ofs/llm-for-dptd/huxinlei/general-prediction-server/easy_rec/python/tools/feature_selection.py", line 59, in process feature_dim_dropout_p_map, embedding_wise_variational_dropout = self._feature_dim_dropout_ratio( File "/ofs/llm-for-dptd/huxinlei/general-prediction-server/easy_rec/python/tools/feature_selection.py", line 110, in _feature_dim_dropout_ratio logit_p = reader.get_tensor(logit_p_name) File "/home/luban/.local/lib/python3.8/site-packages/tensorflow/python/training/py_checkpoint_reader.py", line 71, in get_tensor error_translator(e) File "/home/luban/.local/lib/python3.8/site-packages/tensorflow/python/training/py_checkpoint_reader.py", line 31, in error_translator raise errors_impl.NotFoundError(None, None, error_message) tensorflow.python.framework.errors_impl.NotFoundError: Key logit_p_wide not found in checkpoint