What is the difference between loading frozen or non-frozen model

[junyi]

Model Optimizer supports loading frozen or non-frozen models

python3 mo_tf.py --input_model <INPUT_MODEL>.pb

python3 mo_tf.py --input_model <INFERENCE_GRAPH>.pb 
                 --input_checkpoint <INPUT_CHECKPOINT>

My question is what is the difference between each other, will it impact the accuracy or inference time ?

[Severine_H_Intel · Accepted solution]

Dear Junyi, 

it will not impact accuracy or performance time. It only allows easing the user experience in the TF mode importation. Before, you would have to freeze yourself the model with a TF script. Now, it is included.

Best, 

Severine