Model Loading Error with Bi-directional LSTM layer about tensorflow HOT 3 OPEN

@nilesh-sengupta Please use the appropriate loading function based on the saving format. For example, tf.keras.models.load_model for HDF5 and tf.saved_model.load for SavedModel.
Thank you!

from tensorflow.

@sushreebarsa The model is in .keras format, it is saved using model.save('model_name.keras'), loading it using tf.saved_model.load gives the following exception:
Traceback (most recent call last):
File "/usr/local/bin/mldnn/dnn_1.6/predictor.py", line 26, in
model = tf.saved_model.load('deep_neural_network.keras')
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/root/miniconda3/envs/model_training/lib/python3.11/site-packages/tensorflow/python/saved_model/load.py", line 912, in load
result = load_partial(export_dir, None, tags, options)["root"]
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/root/miniconda3/envs/model_training/lib/python3.11/site-packages/tensorflow/python/saved_model/load.py", line 1016, in load_partial
loader_impl.parse_saved_model_with_debug_info(export_dir))
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/root/miniconda3/envs/model_training/lib/python3.11/site-packages/tensorflow/python/saved_model/loader_impl.py", line 59, in parse_saved_model_with_debug_info
saved_model = parse_saved_model(export_dir)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/root/miniconda3/envs/model_training/lib/python3.11/site-packages/tensorflow/python/saved_model/loader_impl.py", line 119, in parse_saved_model
raise IOError(
OSError: SavedModel file does not exist at: deep_neural_network.keras/{saved_model.pbtxt|saved_model.pb}

The previous error is encountered while using tf.keras.models.load_model

from tensorflow.

@nilesh-sengupta
Could you try to re-save your Keras model as a SavedModel:

tf.saved_model.save(model, 'path/to/save/model')  # Use a distinct path to avoid overwriting

Since your model is in the .keras format, please use tf.keras.models.load_model to load it:

model = tf.saved_model.load('path/to/save/model')

This function is specifically designed to handle models saved in the Keras format.
Thank you!

from tensorflow.