Save & Load a TF saved model. Get TF graph (GraphDef) from a saved model.

saved_model.md

TF Saved Model

Save a saved_model

#!/usr/bin/env python3

import tensorflow as tf
import numpy as np

# Get model
model = tf.keras.applications.ResNet50()

# Random input
x = np.random.rand(1,224,224,3).astype(np.float32)

# Run once
result = model(x)

# Save
tf.saved_model.save(model, "./resnet/1/")

Load a saved_model. Get Graph Def.

import tensorflow as tf
from tensorflow.core.protobuf import saved_model_pb2

with tf.io.gfile.GFile("saved_model.pb", mode='rb') as f:
    model = f.read()
    
saved_model = saved_model_pb2.SavedModel()
saved_model.ParseFromString(model)

graph = saved_model.meta_graphs[0].graph_def
print(graph)