coremltools: Keras model converted to ML Program doesn't load in Xcode

🐞Describe the bug

If you convert a Keras model to ML Program, it doesn’t work at all.

• Shows a warning while converting (see Trace below)
• Shows an error in Xcode:

Trace

Running TensorFlow Graph Passes: 100%|████████████████████████████████████████████████████████████████████████████| 5/5 [00:00<00:00, 98.66 passes/s]
Converting Frontend ==> MIL Ops: 100%|████████████████████████████████████████████████████████████████████████████| 4/4 [00:00<00:00, 13842.59 ops/s]
Running MIL Common passes: 100%|█████████████████████████████████████████████████████████████████████████████| 31/31 [00:00<00:00, 59398.55 passes/s]
Running MIL FP16ComputePrecision pass: 100%|████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 1912.59 passes/s]
Running MIL Clean up passes: 100%|█████████████████████████████████████████████████████████████████████████████| 8/8 [00:00<00:00, 11869.27 passes/s]
/Users/x/miniforge3/envs/tf/lib/python3.9/site-packages/coremltools/models/model.py:122: RuntimeWarning: You will not be able to run predict() on this Core ML model. Underlying exception message was: Error compiling model: "Error reading protobuf spec. validator error: Model specification version field missing or corrupt.".
  _warnings.warn(

To Reproduce

from tensorflow import keras
from tensorflow.keras import layers

import coremltools as ct

input = keras.Input(shape=[10])
output = layers.Dense(10)(input)

model = keras.Model(input, output)
model.compile(optimizer='adam', loss='mse')

mlmodel = ct.convert(model, minimum_deployment_target=ct.target.iOS15)
mlmodel.save('model.mlpackage')

System environment (please complete the following information):

• coremltools version (e.g., 3.0b5): 5.0b4
• OS (e.g., MacOS, Linux): macOS
• macOS version (if applicable): 11.6
• XCode version (if applicable): 13.0 RC
• How you install python (anaconda, virtualenv, system): anaconda
• python version (e.g. 3.7): 3.9.4
• any other relevant information:
  • TensorFlow version: 2.5.0