Python 3.9.21 Crash with TensorFlow 2.10.1 Windows - AMD 9950X3D

TensorFlow
,
1

Hi,

Everything has been working in my code for a year now, last week I swapped my CPU from a AMD 7950x to a 9950x3D (with 3D cache). I have rebuilt my conda environment to create an env for Windows 10, with Python 3.9.21, TF 2.10.1 (using Keras 2.10) and now I get a hard crash in python with no exception caught.
The code is basically reading in persisted LSTMs created around a month ago, and looks like this

  • the model_file_toload is just the path to the h5 model, and this code was working perfectly before.

      try:
      self.model = tf.keras.models.load_model(model_file_toload)
      logger.info(f"Loaded Model with included compilation:{model_file_toload}")
  except Exception as e:
      self.model = None
      logger.error(f"Error loading Model:{model_file_toload} - {e}")

It doesnt even catch the exception it just hard crashes, as I am running outside any IDE (ie no VSCode, no debugger), the traceback looks like this

Thread 0x00002b20 (most recent call first):
File “U:\Miniconda3\envs\ML_CPU_WIN_9950_2025\lib\site-packages\tensorflow\python\ops\gen_linalg_ops.py”, line 2133 in qr
File “U:\Miniconda3\envs\ML_CPU_WIN_9950_2025\lib\site-packages\keras\initializers\initializers_v2.py”, line 764 in _generate_init_val
File “U:\Miniconda3\envs\ML_CPU_WIN_9950_2025\lib\site-packages\keras\initializers\initializers_v2.py”, line 750 in call
File “U:\Miniconda3\envs\ML_CPU_WIN_9950_2025\lib\site-packages\tensorflow\python\ops\resource_variable_ops.py”, line 1812 in _init_from_args
File “U:\Miniconda3\envs\ML_CPU_WIN_9950_2025\lib\site-packages\tensorflow\python\ops\resource_variable_ops.py”, line 1659 in init
File “U:\Miniconda3\envs\ML_CPU_WIN_9950_2025\lib\site-packages\tensorflow\python\ops\variables.py”, line 268 in call
File “U:\Miniconda3\envs\ML_CPU_WIN_9950_2025\lib\site-packages\tensorflow\python\util\traceback_utils.py”, line 150 in error_handler
File “U:\Miniconda3\envs\ML_CPU_WIN_9950_2025\lib\site-packages\tensorflow\python\ops\variable_scope.py”, line 2705 in default_variable_creator
File “U:\Miniconda3\envs\ML_CPU_WIN_9950_2025\lib\site-packages\tensorflow\python\ops\variables.py”, line 202 in
File “U:\Miniconda3\envs\ML_CPU_WIN_9950_2025\lib\site-packages\tensorflow\python\ops\variables.py”, line 209 in _variable_v1_call
File “U:\Miniconda3\envs\ML_CPU_WIN_9950_2025\lib\site-packages\tensorflow\python\ops\variables.py”, line 264 in call
File “U:\Miniconda3\envs\ML_CPU_WIN_9950_2025\lib\site-packages\tensorflow\python\util\traceback_utils.py”, line 150 in error_handler
File “U:\Miniconda3\envs\ML_CPU_WIN_9950_2025\lib\site-packages\keras\engine\base_layer_utils.py”, line 134 in make_variable
File “U:\Miniconda3\envs\ML_CPU_WIN_9950_2025\lib\site-packages\tensorflow\python\trackable\base.py”, line 489 in _add_variable_with_custom_getter
File “U:\Miniconda3\envs\ML_CPU_WIN_9950_2025\lib\site-packages\keras\engine\base_layer.py”, line 705 in add_weight
File “U:\Miniconda3\envs\ML_CPU_WIN_9950_2025\lib\site-packages\keras\layers\rnn\lstm.py”, line 199 in build
File “U:\Miniconda3\envs\ML_CPU_WIN_9950_2025\lib\site-packages\keras\utils\tf_utils.py”, line 354 in wrapper
File “U:\Miniconda3\envs\ML_CPU_WIN_9950_2025\lib\site-packages\keras\layers\rnn\base_rnn.py”, line 458 in build
File “U:\Miniconda3\envs\ML_CPU_WIN_9950_2025\lib\site-packages\keras\engine\base_layer.py”, line 2949 in _maybe_build
File “U:\Miniconda3\envs\ML_CPU_WIN_9950_2025\lib\site-packages\keras\engine\base_layer.py”, line 2402 in _infer_output_signature
File “U:\Miniconda3\envs\ML_CPU_WIN_9950_2025\lib\site-packages\keras\engine\base_layer.py”, line 2345 in _keras_tensor_symbolic_call
File “U:\Miniconda3\envs\ML_CPU_WIN_9950_2025\lib\site-packages\keras\engine\base_layer.py”, line 2498 in _functional_construction_call
File “U:\Miniconda3\envs\ML_CPU_WIN_9950_2025\lib\site-packages\keras\engine\base_layer.py”, line 1011 in call
File “U:\Miniconda3\envs\ML_CPU_WIN_9950_2025\lib\site-packages\keras\utils\traceback_utils.py”, line 65 in error_handler
File “U:\Miniconda3\envs\ML_CPU_WIN_9950_2025\lib\site-packages\keras\layers\rnn\base_rnn.py”, line 553 in call
File “U:\Miniconda3\envs\ML_CPU_WIN_9950_2025\lib\site-packages\keras\engine\sequential.py”, line 234 in add
File “U:\Miniconda3\envs\ML_CPU_WIN_9950_2025\lib\site-packages\keras\utils\traceback_utils.py”, line 65 in error_handler
File “U:\Miniconda3\envs\ML_CPU_WIN_9950_2025\lib\site-packages\tensorflow\python\trackable\base.py”, line 205 in _method_wrapper
File “U:\Miniconda3\envs\ML_CPU_WIN_9950_2025\lib\site-packages\keras\engine\sequential.py”, line 475 in from_config
File “U:\Miniconda3\envs\ML_CPU_WIN_9950_2025\lib\site-packages\keras\utils\generic_utils.py”, line 734 in deserialize_keras_object
File “U:\Miniconda3\envs\ML_CPU_WIN_9950_2025\lib\site-packages\keras\layers\serialization.py”, line 249 in deserialize
File “U:\Miniconda3\envs\ML_CPU_WIN_9950_2025\lib\site-packages\keras\saving\model_config.py”, line 55 in model_from_config
File “U:\Miniconda3\envs\ML_CPU_WIN_9950_2025\lib\site-packages\keras\saving\hdf5_format.py”, line 200 in load_model_from_hdf5
File “U:\Miniconda3\envs\ML_CPU_WIN_9950_2025\lib\site-packages\keras\saving\save.py”, line 241 in load_model
File “U:\Miniconda3\envs\ML_CPU_WIN_9950_2025\lib\site-packages\keras\utils\traceback_utils.py”, line 65 in error_handler
File “D:\mldev\project\TransformerStockPrices\ann_models\ANNModelSM.py”, line 734 in loadGoldenModel
File “D:\mldev\project\TransformerStockPrices\main.py”, line 2258 in runticker
File “D:\mldev\project\TransformerStockPrices\main.py”, line 876 in run_golden_model
File “D:\mldev\project\TransformerStockPrices\main.py”, line 786 in run_models_from_folder
File “D:\mldev\project\TransformerStockPrices\main.py”, line 1854 in main
File “D:\mldev\project\TransformerStockPrices\main.py”, line 2787 in

Any suggestions on how to fix this ? Thanks.

3

Hi @M_O1, Welcome to the forum!

Please try creating a new environment and upgrade the tensorflow/ keras versions and check if you’re able to work with it.