Converting model to tflite without optimization option about tflite-support HOT 1 CLOSED

You're right. Quantization won'y apply if you don't explicitly set up the optimization option.

And the tflite model (a flatbuffers file) is in a different format of the saved model. From the Keras documentation:
"_A Keras model consists of multiple components:

1. An architecture, or configuration, which specifies what layers the model contain, and how they're connected.
2. A set of weights values (the "state of the model").
3. An optimizer (defined by compiling the model).
4. A set of losses and metrics (defined by compiling the model or calling add_loss() or add_metric())._"

And the tflite model strips off information about training (item 4) and compilation (item 3). Therefore the size will be smaller than Keras model.

from tflite-support.