PINTO0309
commented
2 years ago There is no information on your environment and no way to investigate. I will not support any use of containers other than those I provide. I am not a handyman.
docker run --gpus all -it --rm \
-v `pwd`:/home/user/workdir \
ghcr.io/pinto0309/openvino2tensorflow:latest
$INTEL_OPENVINO_DIR/deployment_tools/model_optimizer/mo.py \
--input_model mnist-8.onnx \
--data_type FP32 \
--output_dir openvino/FP32 \
--model_name mnist_8
openvino2tensorflow \
--model_path openvino/FP32/mnist_8.xml \
--output_saved_model \
--output_pb \
--output_no_quant_float32_tflite \
--non_verboseTensorFlow/Keras model building process starts ======================================
Layer structure =====================================================================
layer_type: Input
layer_id: 0
tf_layers_dict: KerasTensor(type_spec=TensorSpec(shape=(1, 28, 28, 1), dtype=tf.float32, name='Input3'), name='Input3', description="created by layer 'Input3'")
====================================================================================
layer_type: Const
layer_id: 1
tf_layers_dict_shape: (8, 1, 5, 5)
====================================================================================
layer_type: Convolution
layer_id: 2
input_layer0: layer_id=0: KerasTensor(type_spec=TensorSpec(shape=(1, 28, 28, 1), dtype=tf.float32, name='Input3'), name='Input3', description="created by layer 'Input3'")
input_layer1_shape: layer_id=1: (8, 1, 5, 5)
tf_layers_dict: KerasTensor(type_spec=TensorSpec(shape=(1, 28, 28, 8), dtype=tf.float32, name=None), name='conv2d/Conv2D:0', description="created by layer 'conv2d'")
====================================================================================
layer_type: Const
layer_id: 3
tf_layers_dict_shape: (1, 8, 1, 1)
====================================================================================
layer_type: Add
layer_id: 4
input_layer0: layer_id=2: KerasTensor(type_spec=TensorSpec(shape=(1, 28, 28, 8), dtype=tf.float32, name=None), name='conv2d/Conv2D:0', description="created by layer 'conv2d'")
input_layer1_shape: layer_id=3: (1, 8, 1, 1)
tf_layers_dict: KerasTensor(type_spec=TensorSpec(shape=(1, 28, 28, 8), dtype=tf.float32, name=None), name='tf.math.add/Add:0', description="created by layer 'tf.math.add'")
====================================================================================
layer_type: ReLU
layer_id: 5
input_layer0: layer_id=4: KerasTensor(type_spec=TensorSpec(shape=(1, 28, 28, 8), dtype=tf.float32, name=None), name='tf.math.add/Add:0', description="created by layer 'tf.math.add'")
tf_layers_dict: KerasTensor(type_spec=TensorSpec(shape=(1, 28, 28, 8), dtype=tf.float32, name=None), name='tf.nn.relu/Relu:0', description="created by layer 'tf.nn.relu'")
====================================================================================
layer_type: MaxPool
layer_id: 6
input_layer0: layer_id=5: KerasTensor(type_spec=TensorSpec(shape=(1, 28, 28, 8), dtype=tf.float32, name=None), name='tf.nn.relu/Relu:0', description="created by layer 'tf.nn.relu'")
tf_layers_dict: KerasTensor(type_spec=TensorSpec(shape=(1, 14, 14, 8), dtype=tf.float32, name=None), name='tf.nn.max_pool2d/MaxPool2d:0', description="created by layer 'tf.nn.max_pool2d'")
====================================================================================
layer_type: Const
layer_id: 7
tf_layers_dict_shape: (16, 8, 5, 5)
====================================================================================
layer_type: Convolution
layer_id: 8
input_layer0: layer_id=6: KerasTensor(type_spec=TensorSpec(shape=(1, 14, 14, 8), dtype=tf.float32, name=None), name='tf.nn.max_pool2d/MaxPool2d:0', description="created by layer 'tf.nn.max_pool2d'")
input_layer1_shape: layer_id=7: (16, 8, 5, 5)
tf_layers_dict: KerasTensor(type_spec=TensorSpec(shape=(1, 14, 14, 16), dtype=tf.float32, name=None), name='conv2d_1/Conv2D:0', description="created by layer 'conv2d_1'")
====================================================================================
layer_type: Const
layer_id: 9
tf_layers_dict_shape: (1, 16, 1, 1)
====================================================================================
layer_type: Add
layer_id: 10
input_layer0: layer_id=8: KerasTensor(type_spec=TensorSpec(shape=(1, 14, 14, 16), dtype=tf.float32, name=None), name='conv2d_1/Conv2D:0', description="created by layer 'conv2d_1'")
input_layer1_shape: layer_id=9: (1, 16, 1, 1)
tf_layers_dict: KerasTensor(type_spec=TensorSpec(shape=(1, 14, 14, 16), dtype=tf.float32, name=None), name='tf.math.add_1/Add:0', description="created by layer 'tf.math.add_1'")
====================================================================================
layer_type: ReLU
layer_id: 11
input_layer0: layer_id=10: KerasTensor(type_spec=TensorSpec(shape=(1, 14, 14, 16), dtype=tf.float32, name=None), name='tf.math.add_1/Add:0', description="created by layer 'tf.math.add_1'")
tf_layers_dict: KerasTensor(type_spec=TensorSpec(shape=(1, 14, 14, 16), dtype=tf.float32, name=None), name='tf.nn.relu_1/Relu:0', description="created by layer 'tf.nn.relu_1'")
====================================================================================
layer_type: MaxPool
layer_id: 12
input_layer0: layer_id=11: KerasTensor(type_spec=TensorSpec(shape=(1, 14, 14, 16), dtype=tf.float32, name=None), name='tf.nn.relu_1/Relu:0', description="created by layer 'tf.nn.relu_1'")
tf_layers_dict: KerasTensor(type_spec=TensorSpec(shape=(1, 4, 4, 16), dtype=tf.float32, name=None), name='tf.nn.max_pool2d_1/MaxPool2d:0', description="created by layer 'tf.nn.max_pool2d_1'")
====================================================================================
layer_type: Const
layer_id: 13
tf_layers_dict_shape: (2,)
====================================================================================
layer_type: Reshape
layer_id: 14
input_layer0: layer_id=12: KerasTensor(type_spec=TensorSpec(shape=(1, 4, 4, 16), dtype=tf.float32, name=None), name='tf.nn.max_pool2d_1/MaxPool2d:0', description="created by layer 'tf.nn.max_pool2d_1'")
input_layer1_shape: layer_id=13: (2,)
tf_layers_dict: KerasTensor(type_spec=TensorSpec(shape=(1, 256), dtype=tf.float32, name=None), name='tf.reshape/Reshape:0', description="created by layer 'tf.reshape'")
====================================================================================
layer_type: Const
layer_id: 15
tf_layers_dict_shape: (10, 256)
====================================================================================
layer_type: MatMul
layer_id: 16
input_layer0: layer_id=14: KerasTensor(type_spec=TensorSpec(shape=(1, 256), dtype=tf.float32, name=None), name='tf.reshape/Reshape:0', description="created by layer 'tf.reshape'")
input_layer1_shape: layer_id=15: (10, 256)
tf_layers_dict: KerasTensor(type_spec=TensorSpec(shape=(1, 10), dtype=tf.float32, name=None), name='tf.linalg.matmul/MatMul:0', description="created by layer 'tf.linalg.matmul'")
====================================================================================
layer_type: Const
layer_id: 17
tf_layers_dict_shape: (1, 10)
====================================================================================
layer_type: Add
layer_id: 18
input_layer0: layer_id=16: KerasTensor(type_spec=TensorSpec(shape=(1, 10), dtype=tf.float32, name=None), name='tf.linalg.matmul/MatMul:0', description="created by layer 'tf.linalg.matmul'")
input_layer1_shape: layer_id=17: (1, 10)
tf_layers_dict: KerasTensor(type_spec=TensorSpec(shape=(1, 10), dtype=tf.float32, name=None), name='tf.math.add_2/Add:0', description="created by layer 'tf.math.add_2'")
====================================================================================
layer_type: Result
layer_id: 19
input_layer0: layer_id=18: KerasTensor(type_spec=TensorSpec(shape=(1, 10), dtype=tf.float32, name=None), name='tf.math.add_2/Add:0', description="created by layer 'tf.math.add_2'")
tf_layers_dict: KerasTensor(type_spec=TensorSpec(shape=(1, 10), dtype=tf.float32, name=None), name='tf.identity/Identity:0', description="created by layer 'tf.identity'")
====================================================================================
TensorFlow/Keras model building process complete!
saved_model output started ==========================================================
WARNING:absl:Function `_wrapped_model` contains input name(s) Input3 with unsupported characters which will be renamed to input3 in the SavedModel.
saved_model output complete!
.pb output started ==================================================================
.pb output complete! - saved_model/model_float32.pb
tflite Float32 convertion started ===================================================
WARNING:absl:Function `_wrapped_model` contains input name(s) Input3 with unsupported characters which will be renamed to input3 in the SavedModel.
Estimated count of arithmetic ops: 1.591 M ops, equivalently 0.796 M MACs
WARNING:absl:Buffer deduplication procedure will be skipped when flatbuffer library is not properly loaded
tflite Float32 convertion complete! - saved_model/model_float32.tflite
All the conversion process is finished! =============================================
hayoyo12
hamedgorji
Issue Type
Others
OS
Ubuntu
OS architecture
x86_64
Programming Language
Python
Framework
OpenVINO, ONNX, TensorFlow
Download URL for ONNX / OpenVINO IR
https://github.com/onnx/models/blob/main/vision/classification/mnist/model/mnist-8.onnx
I used pre-trained mnist-8.onnx model. And I converted the onnx model to OpenVINO IR. So, Now I am trying to convert the OpenVINO IR model to pb model by using openvino2tensorflow
Convert Script
Description
While I reset the environment, I needed to re-install onnx-tf, tensorflow-cpu, openvino, openvino2tensorflow.
(all PYPI: onnx-tf==1.9.0, tensorflow-cpu==2.6.2, openvino-dev==2022.1.0[onnx, tensorflow2], openvino2tensorflow==1.29.3 or 1.29.4, tensorflow-datasets==4.5.2)
Before, there was no problem to convert mnist-8.xml (OpenVINO IR) to tensorflow model using openvino2tensorflow. However, after the reset up environment, it gets 'KeyError' problem. I cannot find what is the 'KeyError'. And I want to know why it suddenly showed up............
It does not show only for the mnist model, other models do too :(. Firstly, I thought it showed up only after the maxpool node, but sometimes it showed up at weird layers...:( Is this issue can solve to change input shape? like 6-7 in the README? https://github.com/PINTO0309/openvino2tensorflow
I tried to search the issue, but I cannot find any related issue. I really want to understand the problem. Please help me.
Relevant Log Output
Source code for simple inference testing code
It is same as convert script.