No module named 'keras.saving.hdf5_format'

[MobinRoohi]

Hello,

when running python trainer_finetune.py, I first hit the following error:

Traceback (most recent call last):
  File "/content/PAED/trainer_finetune.py", line 28, in <module>
    from dataset import ExtDataTr, VAEData, get_dataloader
  File "/content/PAED/dataset.py", line 18, in <module>
    from encoding import ExtractEncoder
  File "/content/PAED/encoding.py", line 11, in <module>
    from transformer_base import run_summarization
  File "/content/PAED/transformer_base/run_summarization.py", line 37, in <module>
    from cst_trainer import CustomTrainer
  File "/content/PAED/cst_trainer.py", line 32, in <module>
    from transformers.trainer_pt_utils import IterableDatasetShard, get_model_param_count, nested_detach
ImportError: cannot import name 'get_model_param_count' from 'transformers.trainer_pt_utils' (/usr/local/lib/python3.10/dist-packages/transformers/trainer_pt_utils.py)

I temporarily resolved this by removing the get_model_param_count. After fixing a further error rooted in wrapper.py where the if statement at the end was empty, I finally, get the following error:

Traceback (most recent call last):
  File "/usr/local/lib/python3.10/dist-packages/transformers/utils/import_utils.py", line 1031, in _get_module
    return importlib.import_module("." + module_name, self.__name__)
  File "/usr/lib/python3.10/importlib/__init__.py", line 126, in import_module
    return _bootstrap._gcd_import(name[level:], package, level)
  File "<frozen importlib._bootstrap>", line 1050, in _gcd_import
  File "<frozen importlib._bootstrap>", line 1027, in _find_and_load
  File "<frozen importlib._bootstrap>", line 1006, in _find_and_load_unlocked
  File "<frozen importlib._bootstrap>", line 688, in _load_unlocked
  File "<frozen importlib._bootstrap_external>", line 883, in exec_module
  File "<frozen importlib._bootstrap>", line 241, in _call_with_frames_removed
  File "/usr/local/lib/python3.10/dist-packages/transformers/models/gpt2/modeling_tf_gpt2.py", line 30, in <module>
    from ...modeling_tf_utils import (
  File "/usr/local/lib/python3.10/dist-packages/transformers/modeling_tf_utils.py", line 39, in <module>
    from keras.saving.hdf5_format import save_attributes_to_hdf5_group
ModuleNotFoundError: No module named 'keras.saving.hdf5_format'

The above exception was the direct cause of the following exception:

Traceback (most recent call last):
  File "/content/PAED/trainer_finetune.py", line 735, in <module>
    main(
  File "/content/PAED/trainer_finetune.py", line 599, in main
    gen_synthetic(save_dir, path_train, path_dev, path_test)
  File "/content/PAED/trainer_finetune.py", line 584, in gen_synthetic
    generator.generate(labels_dev + labels_test, path_out=path_synthetic)
  File "/content/PAED/wrapper.py", line 207, in generate
    pipe = model.make_pipe()
  File "/content/PAED/modeling.py", line 52, in make_pipe
    pipe = pipeline(
  File "/usr/local/lib/python3.10/dist-packages/transformers/pipelines/__init__.py", line 702, in pipeline
    framework, model = infer_framework_load_model(
  File "/usr/local/lib/python3.10/dist-packages/transformers/pipelines/base.py", line 233, in infer_framework_load_model
    _class = getattr(transformers_module, f"TF{architecture}", None)
  File "/usr/local/lib/python3.10/dist-packages/transformers/utils/import_utils.py", line 1022, in __getattr__
    value = getattr(module, name)
  File "/usr/local/lib/python3.10/dist-packages/transformers/utils/import_utils.py", line 1021, in __getattr__
    module = self._get_module(self._class_to_module[name])
  File "/usr/local/lib/python3.10/dist-packages/transformers/utils/import_utils.py", line 1033, in _get_module
    raise RuntimeError(
RuntimeError: Failed to import transformers.models.gpt2.modeling_tf_gpt2 because of the following error (look up to see its traceback):
No module named 'keras.saving.hdf5_format'

The versions of Python and requirements I am using align with the ones you have specified:

Python 3.10.12
Transformers version: 4.22.2
Tensorflow version: 2.17.1
Torch version: 2.5.1+cu121
fastavro==1.4.0
fire==0.4.0
nltk==3.6.6
lxml==4.6.5
editdistance==0.5.3
seqeval==1.2.2

and the latest versions of the following libraries that pip installs:

datasets
pydantic
pyyaml
tqdm
spacy
jsonlines
scikit-learn

Could you please provide guidance on resolving the compatibility issues and ensure the script runs as expected? Let me know if additional details are needed to reproduce the errors.

Thanks!