issue to make TTS model from .pb file to TFlight AttributeError: 'AutoTrackable' object has no attribute 'inputs'

Question

Am trying to make TTS model for arabic with tacotron and I have checkpoint successfully generated with training and I stop the train on this point .

I save the file to .pb file and when I try to convert it to tflite this error appear any help ?

Code to convert checkpoint to pb

import os
import tensorflow as tf

trained_checkpoint_prefix = '/home/kayandeveloper/tacotron/logs-tacotron/model.ckpt-4200'
export_dir = os.path.join('/home/kayandeveloper/tacotron/', '0')

graph = tf.Graph()
with tf.compat.v1.Session(graph=graph) as sess:
    # Restore from checkpoint
    loader = tf.compat.v1.train.import_meta_graph(trained_checkpoint_prefix + '.meta')
    loader.restore(sess, trained_checkpoint_prefix)

    # Export checkpoint to SavedModel
    builder = tf.compat.v1.saved_model.builder.SavedModelBuilder(export_dir)
    builder.add_meta_graph_and_variables(sess,
                                         [tf.saved_model.TRAINING, tf.saved_model.SERVING],
                                         strip_default_attrs=True)
    builder.save()    

Code that I use to convert pb file to tflight

import tensorflow as tf

# Convert the model
converter = tf.lite.TFLiteConverter.from_keras_model_file('/home/kayandeveloper/tacotron/0') # path to the SavedModel directory
tflite_model = converter.convert()

# Save the model.
with open('/home/kayandeveloper/tacotron/araby.tflite', 'wb') as f:
  f.write(tflite_model)

Error log :

2022-02-20 10:30:43.207065: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:983] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2022-02-20 10:30:43.207567: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1639] Found device 0 with properties: 
name: Tesla T4 major: 7 minor: 5 memoryClockRate(GHz): 1.59
pciBusID: 0000:00:04.0
2022-02-20 10:30:43.207651: I tensorflow/stream_executor/platform/default/dso_loader.cc:48] Successfully opened dynamic library libcudart.so.11.0
2022-02-20 10:30:43.207699: I tensorflow/stream_executor/platform/default/dso_loader.cc:48] Successfully opened dynamic library libcublas.so.11
2022-02-20 10:30:43.207722: I tensorflow/stream_executor/platform/default/dso_loader.cc:48] Successfully opened dynamic library libcufft.so.10
2022-02-20 10:30:43.207742: I tensorflow/stream_executor/platform/default/dso_loader.cc:48] Successfully opened dynamic library libcurand.so.10
2022-02-20 10:30:43.207761: I tensorflow/stream_executor/platform/default/dso_loader.cc:48] Successfully opened dynamic library libcusolver.so.10
2022-02-20 10:30:43.207789: I tensorflow/stream_executor/platform/default/dso_loader.cc:48] Successfully opened dynamic library libcusparse.so.11
2022-02-20 10:30:43.207810: I tensorflow/stream_executor/platform/default/dso_loader.cc:48] Successfully opened dynamic library libcudnn.so.8
2022-02-20 10:30:43.207912: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:983] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2022-02-20 10:30:43.208728: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:983] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2022-02-20 10:30:43.209109: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1767] Adding visible gpu devices: 0
2022-02-20 10:30:43.209168: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1180] Device interconnect StreamExecutor with strength 1 edge matrix:
2022-02-20 10:30:43.209181: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1186]      0 
2022-02-20 10:30:43.209190: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1199] 0:   N 
2022-02-20 10:30:43.209328: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:983] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2022-02-20 10:30:43.209784: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:983] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2022-02-20 10:30:43.210145: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1325] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 14160 MB memory) -> physical GPU (device: 0, name: Tesla T4, pci bus id: 0000:00:04.0, compute capability: 7.5)
---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
/tmp/ipykernel_27138/2295115983.py in <module>
      2 
      3 # Convert the model
----> 4 converter = tf.lite.TFLiteConverter.from_keras_model_file('/home/kayandeveloper/tacotron/0') # path to the SavedModel directory
      5 tflite_model = converter.convert()
      6 

/opt/conda/lib/python3.7/site-packages/tensorflow_core/lite/python/lite.py in from_keras_model_file(cls, model_file, input_arrays, input_shapes, output_arrays, custom_objects)
    825       input_tensors = _get_tensors_from_tensor_names(sess.graph, input_arrays)
    826     else:
--> 827       input_tensors = keras_model.inputs
    828 
    829     if output_arrays:

AttributeError: 'AutoTrackable' object has no attribute 'inputs'

Answer 1

try to replace the : ( keras_model.inputs )

with : ( keras_model.Input )

capital i ( I )