Convert Keras 1 models to Keras 2

keras_1_to_2.py

#!/usr/bin/env python3

'''
This will convert Keras 1 model weights and configurations
to their Keras 2 equivalent.  Please note that all optimizer
configuration is ignored with this script and models will 
need to be recompiled.
'''

import h5py
import sys
import json
import argparse
import tensorflow as tf
from keras.models import model_from_config
from pathlib import Path

# import warnings
# warnings.filterwarnings('ignore')

# sys.argv = ['program', '-o', '/path/to/output/directory', 
# '-m', '/path/to/model/file.hdf5', '/path/to/model/directory']

parser = argparse.ArgumentParser()
parser.add_argument('-m','--models', nargs='+', type=Path, help='List of model files/directories to fix.')
parser.add_argument('-x','--ext', default='hdf5', help='Model extension (if using directories)')
parser.add_argument('-o','--output', type=Path, help='Output directory.')
args = parser.parse_args()

for p in args.models:
    models = []
    if p.is_dir():
        models.extend([_ for _ in p.glob('*.{}'.format(args.ext))])
    else:
        if p.exists():
            models.append(p)
        else:
            print('Missing file: {}'.format(p))

args.models = models


def fix_model_file(fp,od=Path('.')):
    assert fp.exists()
    if not od.is_dir():
        od.mkdir(parents=True, exist_ok=True)
    op = str(od / fp.name)
    fp = str(fp)
    
    with h5py.File(fp) as h5:
        config = json.loads(h5.attrs.get("model_config")
                            .decode('utf-8')
                            .replace('input_dtype','dtype'))
    with tf.Session('') as sess:
        model = model_from_config(config)
        model.load_weights(fp)
        model.save(op)
        del model
    del sess
    print(op)

if __name__ == '__main__':
    for model in args.models:
        fix_model_file(model, od=args.output)