[Python] How to export tflite model maker model_spec for offline usage?

I'm using a tflite model maker efficientdet_lite3 model for object detection. The basic set up goes like this:

import numpy as np
import os
from tflite_model_maker.config import ExportFormat, QuantizationConfig
from tflite_model_maker import model_spec
from tflite_model_maker import object_detector
from tflite_support import metadata
import tensorflow as tf

assert tf.__version__.startswith('2')
tf.get_logger().setLevel('ERROR')
from absl import logging
logging.set_verbosity(logging.ERROR)


train_data = object_detector.DataLoader.from_pascal_voc(
'my_data/train',
'my_data/train',
['obj']
)

val_data = object_detector.DataLoader.from_pascal_voc(
'my_data/validate',
'my_data/validate',
['obj']
)

spec = model_spec.get('efficientdet_lite3')

model = object_detector.create(train_data, model_spec=spec, batch_size=4, train_whole_model=True, epochs=200, validation_data=val_data)


And it works fine on a machine with internet access and even yields some usable results. But on a machine without internet access I am getting an urlopen error which I assume is caused by the fact that model spec is being fetched from the internet.

Now this model itself can be found and downloaded here: https://www.kaggle.com/models/tenso...eworks/tensorFlow2/variations/lite3-detection But I have no idea how do create a custom model_spec object for training.

Can you please help me setting up this project for a completely offline usage? This is important for me because the training data is on the offline computer connected to a closed network and transfering this data every time to other computer is a painful process. It would've been much easier to transfer the necessary tflite data to the offline machine.

I've tried using this spec: https://www.tensorflow.org/lite/api...l_maker/object_detector/EfficientDetLite3Spec and pointing it to a local saved_model.pb file but I get all sorts of weird errors like:

ERROR:absl:hub.KerasLayer is trainable but has zero trainable weights.

ValueError: Could not find matching concrete function to call loaded from the SavedModel. Got:
Positional arguments (1 total):
* <tf.Tensor 'imgs:0' shape=(None, 512, 512, 3) dtype=float32>
Keyword arguments: {}

Expected these arguments to match one of the following 1 option(s):

Continue reading...