Error in loading model from the saved model in the disk

[ramkumarkoppu]

# Model file path gpt2_finetuned_model_file = os.path.join("/content/drive/MyDrive/Colab_Data", "gpt2_finetuned_model") if not os.path.exists(gpt2_finetuned_model_file): # Train the model. start = time.time() gpt2_lm.fit(train_ds, epochs=EPOCHS, callbacks=[gpu_memory_callback]) end = time.time() gpt2_lm_memory_usage = gpu_memory_callback.memory_usage print(f"Fine-tuning took {end - start}secs and {gpt2_lm_memory_usage}GB memory to run")
gpt2_lm.save(gpt2_finetuned_model_file)
else: print(f"Loading model from {gpt2_finetuned_model_file}")
gpt2_lm = tf.keras.models.load_model(gpt2_finetuned_model_file)

Loading model throws error:

Loading model from /content/drive/MyDrive/Colab_Data/gpt2_finetuned_model
WARNING:tensorflow:The following Variables were used in a Lambda layer's call (tf.linalg.matmul_1), but are not present in its tracked objects:   <tf.Variable 'token_embedding/embeddings:0' shape=(50257, 768) dtype=float32>. This is a strong indication that the Lambda layer should be rewritten as a subclassed Layer.
---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
[<ipython-input-15-65f40446ec1f>](https://t85rd88mxlp-496ff2e9c6d22116-0-colab.googleusercontent.com/outputframe.html?vrz=colab-20230628-060137-RC00_543986831#) in <cell line: 1>()
      9 else:
     10     print(f"Loading model from {gpt2_finetuned_model_file}")
---> 11     gpt2_lm = tf.keras.models.load_model(gpt2_finetuned_model_file)

2 frames
[/usr/local/lib/python3.10/dist-packages/tensorflow/python/saved_model/load.py](https://t85rd88mxlp-496ff2e9c6d22116-0-colab.googleusercontent.com/outputframe.html?vrz=colab-20230628-060137-RC00_543986831#) in _add_object_graph_edges(self, proto, node_id)
    348 
    349     for reference in proto.children:
--> 350       setter(obj, reference.local_name, self._nodes[reference.node_id])
    351       # Note: if an object has an attribute `__call__` add a class method
    352       # that allows `obj()` syntax to work. This is done per-instance to

AttributeError: can't set attribute 'current_loss_scale'

[jbischof]

Hi @ramkumarkoppu can you file an issue with a minimal colab gist reproducing the error? Thank you!

[ramkumarkoppu]

Created issue #1112, thank you.