Cannot load .pb file input model: SavedModel format load failure: '

I had converted my .h5 file to .pb file by using load_model and model.save as following

model = load_model("model_190-1.00.h5")
model.summary()
model.save("saved_model.pb")
converted = load_model(r"C:\Users\Hsien\Desktop\NCS2\OCT")
converted.summary()

Then I used model.summary() to make sure my original .h5 file and converted .pb file share the same structure.

But I got an error message when I tried to use mo_tf.py to transfer .pb file to IR format

hcl-lab@hcllab-SYS-5049A-TR:/opt/intel/openvino_2021.4.689/deployment_tools/model_optimizer$ sudo python mo_tf.py --saved_model_dir /home/hcl-lab/NCS2/OCT --input_shape [1,256,256,3] --data_type FP16
Model Optimizer arguments:
Common parameters:
        - Path to the Input Model:      None
        - Path for generated IR:        /opt/intel/openvino_2021.4.689/deployment_tools/model_optimizer/.
        - IR output name:       saved_model
        - Log level:    ERROR
        - Batch:        Not specified, inherited from the model
        - Input layers:         Not specified, inherited from the model
        - Output layers:        Not specified, inherited from the model
        - Input shapes:         [1,256,256,3]
        - Mean values:  Not specified
        - Scale values:         Not specified
        - Scale factor:         Not specified
        - Precision of IR:      FP16
        - Enable fusing:        True
        - Enable grouped convolutions fusing:   True
        - Move mean values to preprocess section:       None
        - Reverse input channels:       False
TensorFlow specific parameters:
        - Input model in text protobuf format:  False
        - Path to model dump for TensorBoard:   None
        - List of shared libraries with TensorFlow custom layers implementation:        None
        - Update the configuration file with input/output node names:   None
        - Use configuration file used to generate the model with Object Detection API:  None
        - Use the config file:  None
        - Inference Engine found in:    /opt/intel/openvino_2021.4.689/python/python3.6/openvino
Inference Engine version:       2021.4.1-3926-14e67d86634-releases/2021/4
Model Optimizer version:        2021.4.1-3926-14e67d86634-releases/2021/4
2021-11-24 02:43:53.158753: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcudart.so.11.0
/usr/local/lib/python3.6/dist-packages/tensorflow/python/autograph/impl/api.py:22: DeprecationWarning: the imp module is deprecated in favour of importlib; see the module's documentation for alternative uses
  import imp
2021-11-24 02:43:54.273204: I tensorflow/compiler/jit/xla_cpu_device.cc:41] Not creating XLA devices, tf_xla_enable_xla_devices not set
2021-11-24 02:43:54.273881: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcuda.so.1
2021-11-24 02:43:54.306060: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1720] Found device 0 with properties:
pciBusID: 0000:1b:00.0 name: RTX A5000 computeCapability: 8.6
coreClock: 1.695GHz coreCount: 64 deviceMemorySize: 23.69GiB deviceMemoryBandwidth: 715.34GiB/s
2021-11-24 02:43:54.306553: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1720] Found device 1 with properties:
pciBusID: 0000:1c:00.0 name: RTX A4000 computeCapability: 8.6
coreClock: 1.56GHz coreCount: 48 deviceMemorySize: 15.74GiB deviceMemoryBandwidth: 417.29GiB/s
2021-11-24 02:43:54.306574: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcudart.so.11.0
2021-11-24 02:43:54.308779: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcublas.so.11
2021-11-24 02:43:54.308821: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcublasLt.so.11
2021-11-24 02:43:54.309547: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcufft.so.10
2021-11-24 02:43:54.309755: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcurand.so.10
2021-11-24 02:43:54.311122: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcusolver.so.10
2021-11-24 02:43:54.311697: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcusparse.so.11
2021-11-24 02:43:54.311815: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcudnn.so.8
2021-11-24 02:43:54.314219: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1862] Adding visible gpu devices: 0, 1
2021-11-24 02:43:54.314443: I tensorflow/core/platform/cpu_feature_guard.cc:142] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations:  AVX2 AVX512F FMA
To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.
2021-11-24 02:43:54.315010: I tensorflow/compiler/jit/xla_gpu_device.cc:99] Not creating XLA devices, tf_xla_enable_xla_devices not set
2021-11-24 02:43:54.438124: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1720] Found device 0 with properties:
pciBusID: 0000:1b:00.0 name: RTX A5000 computeCapability: 8.6
coreClock: 1.695GHz coreCount: 64 deviceMemorySize: 23.69GiB deviceMemoryBandwidth: 715.34GiB/s
2021-11-24 02:43:54.438626: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1720] Found device 1 with properties:
pciBusID: 0000:1c:00.0 name: RTX A4000 computeCapability: 8.6
coreClock: 1.56GHz coreCount: 48 deviceMemorySize: 15.74GiB deviceMemoryBandwidth: 417.29GiB/s
2021-11-24 02:43:54.438656: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcudart.so.11.0
2021-11-24 02:43:54.438675: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcublas.so.11
2021-11-24 02:43:54.438683: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcublasLt.so.11
2021-11-24 02:43:54.438690: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcufft.so.10
2021-11-24 02:43:54.438698: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcurand.so.10
2021-11-24 02:43:54.438706: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcusolver.so.10
2021-11-24 02:43:54.438714: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcusparse.so.11
2021-11-24 02:43:54.438722: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcudnn.so.8
2021-11-24 02:43:54.440675: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1862] Adding visible gpu devices: 0, 1
2021-11-24 02:43:54.440721: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcudart.so.11.0
2021-11-24 02:43:55.014738: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1261] Device interconnect StreamExecutor with strength 1 edge matrix:
2021-11-24 02:43:55.014773: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1267]      0 1
2021-11-24 02:43:55.014779: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1280] 0:   N N
2021-11-24 02:43:55.014782: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1280] 1:   N N
2021-11-24 02:43:55.017101: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1406] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 22272 MB memory) -> physical GPU (device: 0, name: RTX A5000, pci bus id: 0000:1b:00.0, compute capability: 8.6)
2021-11-24 02:43:55.018332: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1406] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:1 with 14704 MB memory) -> physical GPU (device: 1, name: RTX A4000, pci bus id: 0000:1c:00.0, compute capability: 8.6)
[ FRAMEWORK ERROR ]  Cannot load input model: SavedModel format load failure: '_UserObject' object has no attribute 'add_slot'

I tried to do the same procedures both in Windows 10 and Ubuntu 18.04 with openvino_2021.4.689 and openvino_2021.4.752.

(I have put the .pb related files in the proper folder.)

Is it possible for me to do the wrong way to convert the two files (.h5 to .pb)?

CodePudding user response：

You need to save the saved_model.pb file inside the saved_model folder, because the --saved_model_dir argument must provide a path to the SavedModel directory.

For instance, your current location is C:\Users\Hsien\Desktop\NCS2\OCT, move the model to C:\Users\Hsien\Desktop\NCS2\saved_model.

CodePudding user response：

Thanks for Intel devs' help.

Finally I found where the shoe pinches.

I should use tf.saved_model.save rather than model.save when transferring .h5 file to .pb file.

The former will create only saved_model.pb file while the latter will create keras_metadata.pb and saved_model.pb files at the same time, which will lead to SavedModel format load failure with mo_tf.py.