How can I convert a TensorFlow model on Kaggle to ONNX format? Or How can I generate a model in TensorFlow Lite format with fixed shape such as batch_size?

I am trying to convert biggan, distributed at the URL below, to ONNX format in order to use it in an embedded environment.

When I perform the following operation, the following error message is output. How can I convert it to ONNX format?　Ultimately, I would like to convert to a tflite model with the input shape batch_size fixed to 1. I will convert to ONNX with tf2onnx, set batch_size to 1 in ONNX, then run onnx2tf to write out the TensorFlow Lite format.

### create conda environment
$ conda create -n biggan_convert python=3.10
$ conda activate biggan_convert
$ pip install tensorflow==2.9 tf_keras==2.14.1 numpy==1.26.4 onnx onnxruntime onnx2tf==1.26.9 onnx_graphsurgeon psutil sng4onnx tf2onnx==1.13.0
$ conda install six
$ conda install requests

### download model from kaggle
$ curl -L -o ~/Downloads/model.tar.gz\
  

### expand
$ mkdir saved_model
$ tar zxvf ~/Downloads/model.tar.gz -C saved_model
$ ls saved_model/
saved_model.pb  tfhub_module.pb  variables

### convert to onnx
$ python3 -m tf2onnx.convert --saved-model saved_model --output biggan
.onnx
...
RuntimeError: MetaGraphDef associated with tags 'serve' could not be found in SavedModel, with available tags '[set()]'. To inspect available tag-sets in the SavedModel, please use the SavedModel CLI: `saved_model_cli`.

$ saved_model_cli show --dir ./saved_model --all
2025-04-01 17:23:14.550812: I tensorflow/core/util/util:169] oneDNN custom operations are on. You may see slightly different numerical results due to floating-point round-off errors from different computation orders. To turn them off, set the environment variable `TF_ENABLE_ONEDNN_OPTS=0`.
2025-04-01 17:23:14.552998: W tensorflow/stream_executor/platform/default/dso_loader:64] Could not load dynamic library 'libcudart.so.11.0'; dlerror: libcudart.so.11.0: cannot open shared object file: No such file or directory; LD_LIBRARY_PATH: /usr/local/cuda/lib64:
2025-04-01 17:23:14.553009: I tensorflow/stream_executor/cuda/cudart_stub:29] Ignore above cudart dlerror if you do not have a GPU set up on your machine.

MetaGraphDef with tag-set: '' contains the following SignatureDefs:

signature_def['default']:
  The given SavedModel SignatureDef contains the following input(s):
    inputs['truncation'] tensor_info:
        dtype: DT_FLOAT
        shape: ()
        name: truncation:0
    inputs['y'] tensor_info:
        dtype: DT_FLOAT
        shape: (-1, 1000)
        name: y:0
    inputs['z'] tensor_info:
        dtype: DT_FLOAT
        shape: (-1, 120)
        name: z:0
  The given SavedModel SignatureDef contains the following output(s):
    outputs['default'] tensor_info:
        dtype: DT_FLOAT
        shape: (-1, 128, 128, 3)
        name: G_trunc_output:0
  Method name is: