Can't Load Models in Tensorflow

[chigkim]

Thanks so much for quick response on FigmaAI/zeplin-ml/issues/1!
I'm having a problem loading the models.
If I try to load using tf.keras.models.load_model("rico/models/mobilenetv2-50k/saved-model/saved_model"), it throws error below.

WARNING:tensorflow:SavedModel saved prior to TF 2.5 detected when loading Keras model. Please ensure that you are saving the model with model.save() or tf.keras.models.save_model(), *NOT* tf.saved_model.save(). To confirm, there should be a file named "keras_metadata.pb" in the SavedModel directory.
ValueError: Unable to create a Keras model from SavedModel at design-system-detector/rico/models/mobilenetv2-50k/saved-model/saved_model. This SavedModel was exported with `tf.saved_model.save`, and lacks the Keras metadata file. Please save your Keras model by calling `model.save` or `tf.keras.models.save_model`.
 Note that you can still load this SavedModel with `tf.saved_model.load`.

If I load with tf.saved_model.load, it just returns _UserObject, and I can't run model.predict.
AttributeError: '_UserObject' object has no attribute 'predict'
Is that possible for you to resave models, so people can load them in recent Tensorflow?
Thanks for your help!

[dusskapark]

When you try to load that model, I recommend setting the version of TensorFlow to 2.5 or below 2022년 11월 11일 (금) 오전 12:13, chigkim ***@***.***>님이 작성:

…

Thanks so much for quick response on FigmaAI/zeplin-ml#1 <FigmaAI/zeplin-ml#1>! I'm having a problem loading the models. If I try to load using tf.keras.models.load_model("rico/models/mobilenetv2-50k/saved-model/saved_model"), it throws error below. WARNING:tensorflow:SavedModel saved prior to TF 2.5 detected when loading Keras model. Please ensure that you are saving the model with model.save() or tf.keras.models.save_model(), *NOT* tf.saved_model.save(). To confirm, there should be a file named "keras_metadata.pb" in the SavedModel directory. ValueError: Unable to create a Keras model from SavedModel at design-system-detector/rico/models/mobilenetv2-50k/saved-model/saved_model. This SavedModel was exported with `tf.saved_model.save`, and lacks the Keras metadata file. Please save your Keras model by calling `model.save` or `tf.keras.models.save_model`. Note that you can still load this SavedModel with `tf.saved_model.load`. If I load with tf.saved_model.load, it just returns _UserObject, and I can't run model.predict. AttributeError: '_UserObject' object has no attribute 'predict' Is that possible for you to resave models, so people can load them in recent Tensorflow? Thanks for your help! — Reply to this email directly, view it on GitHub <#1>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AA7346KAHRVC2HF6WRWE54DWHUGH7ANCNFSM6AAAAAAR4VF6V4> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[chigkim]

Actually I tried loading it with Tensorflow 2.5, 2.4, 2.3, 2.2 with no luck. I get this error.

raise JSONDecodeError("Expecting value", s, err.value) from None
json.decoder.JSONDecodeError: Expecting value: line 1 column 1 (char 0)

Do you have any code snippet I can look on Colab or Kaggle for inferencing or fine tuning on Tensorflow, not Tensorflow.js?
Thank you so much for your help

[dusskapark]

This is a quite old memory, but I think it would be helpful to refer to the links below: https://gist.github.com/dusskapark/c59eae6cae2b822bf7d133a2860c4c51 Convert_RICO_for_Object_Detection_based_on_CLAY.ipynb gist.github.com https://gist.github.com/dusskapark/a332f8b0701449b679e4c0d766009b7c CLAY-tensorflow2-object-detection.ipynb gist.github.com

…

On Nov 11, 2022, at 10:50 AM, chigkim ***@***.***> wrote: Actually I tried loading it with Tensorflow 2.5, 2.4, 2.3, 2.2 with no luck. I get this error. raise JSONDecodeError("Expecting value", s, err.value) from None json.decoder.JSONDecodeError: Expecting value: line 1 column 1 (char 0) Do you have any code snippet I can look on Colab or Kaggle for inferencing or fine tuning on Tensorflow, not Tensorflow.js? Thank you so much for your help — Reply to this email directly, view it on GitHub <#1 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AA7346JUS5UU5MDBSU7OHY3WHWQ77ANCNFSM6AAAAAAR4VF6V4>. You are receiving this because you commented.

[chigkim]

Thank you!!!
I was able to get it going in Tensorflow 2.10.0.
In case other people are interested, here's a little snippet to predict one image file.

import tensorflow as tf
from tensorflow.keras.utils import load_img, img_to_array
import numpy as np

model_dir = "design-system-detector/rico/models/mobilenetv2-50k/saved-model/saved_model"
model = tf.saved_model.load(model_dir)
model = model.signatures['serving_default']
img = load_img("rico/combined/100.jpg")
img = img_to_array(img, dtype=np.uint8)[np.newaxis,...]
input_tensor = tf.convert_to_tensor(img)
output_dict = model(input_tensor=input_tensor)
classes = np.array(output_dict['detection_classes'])
boxes = np.array(output_dict['detection_boxes'])
scores = np.array(output_dict['detection_scores'])
print(classes.shape, classes.dtype)
print(boxes.shape, boxes.dtype)
print(scores.shape, scores.dtype)