convert coreml to onnx error on apply_prelu???

[sharepointer]

onnx_model = onnxmltools.convert_coreml(coreml_model)

File "/usr/local/lib/python3.6/dist-packages/onnxmltools/convert/main.py", line 18, in convert_coreml
custom_conversion_functions, custom_shape_calculators)
File "/usr/local/lib/python3.6/dist-packages/onnxmltools/convert/coreml/convert.py", line 80, in convert
onnx_model = convert_topology(topology, name, doc_string, target_opset, targeted_onnx)
File "/usr/local/lib/python3.6/dist-packages/onnxconverter_common/topology.py", line 729, in convert_topology
registration.get_converter(operator.type)(scope, operator, container)
File "/usr/local/lib/python3.6/dist-packages/onnxmltools/convert/coreml/operator_converters/neural_network/Activation.py", line 27, in convert_activation
apply_prelu(scope, inputs, outputs, container, operator_name=attrs['name'], slope=[params.PReLU.alpha])
File "/usr/local/lib/python3.6/dist-packages/onnxconverter_common/onnx_ops.py", line 440, in apply_prelu
s_shape = slope.shape
AttributeError: 'list' object has no attribute 'shape'

[sharepointer]

error on apply_prelu???

[jiafatom]

Can you provide the model for us to debug?

[KeunwooPark]

I had the same issue. I was trying to convert Openpose to an ONNX model. I used their body_25 models. The below is my code.

# Convert Caffe model to CoreML
coreml_model = coremltools.converters.caffe.convert((args.caffemodel, args.protofile))
# Convert the Core ML model into ONNX
onnx_model = onnxmltools.convert_coreml(coreml_model)
# Save as protobuf
onnxmltools.utils.save_model(onnx_model, args.output)

And here is the error I got.

Traceback (most recent call last):
  File "caffe2onnx.py", line 27, in <module>
    main(args)
  File "caffe2onnx.py", line 21, in main
    onnx_model = onnxmltools.convert_coreml(coreml_model)
  File "[home]/.virtualenvs/caffe2pytorch/lib/python3.5/site-packages/onnxmltools/convert/main.py", line 18, in convert_coreml
    custom_conversion_functions, custom_shape_calculators)
  File "[home]/.virtualenvs/caffe2pytorch/lib/python3.5/site-packages/onnxmltools/convert/coreml/convert.py", line 80, in convert
    onnx_model = convert_topology(topology, name, doc_string, target_opset, targeted_onnx)
  File "[home]/.virtualenvs/caffe2pytorch/lib/python3.5/site-packages/onnxconverter_common/topology.py", line 729, in convert_topology
    registration.get_converter(operator.type)(scope, operator, container)
  File "[home]/.virtualenvs/caffe2pytorch/lib/python3.5/site-packages/onnxmltools/convert/coreml/operator_converters/neural_network/Activation.py", line 27, in convert_activation
    apply_prelu(scope, inputs, outputs, container, operator_name=attrs['name'], slope=[params.PReLU.alpha])
  File "[home]/.virtualenvs/caffe2pytorch/lib/python3.5/site-packages/onnxconverter_common/onnx_ops.py", line 522, in apply_prelu
    s_shape = slope.shape
AttributeError: 'list' object has no attribute 'shape'

[OuHangKresnik]

I just encounter the same issue here.

[orsveri]

Hello! I was trying to convert openpose model from Caffe to ONNX too and got the same error.

SCRIPT:

import os
import coremltools
import onnxmltools

# Update your input name and path for your caffe model
dir = "/media/sveta/DATASTORE/AI_ML_DL/Projects/PoseEstimation/CMU-PCL_OpenPose/openpose/models/pose/body_25/"
proto_file = 'pose_deploy.prototxt'
input_caffe_path = 'pose_iter_584000.caffemodel'

# Update the output name and path for intermediate coreml model, or leave as is
output_coreml_model = 'temp/model.mlmodel'

# Change this path to the output name and path for the onnx model
output_onnx_model = 'pose_body25_model.onnx'

# Convert Caffe model to CoreML
coreml_model = coremltools.converters.caffe.convert((os.path.join(dir, input_caffe_path), os.path.join(dir, proto_file)))

# Save CoreML model
coreml_model.save(output_coreml_model)

# Load a Core ML model
coreml_model = coremltools.utils.load_spec(output_coreml_model)

# Convert the Core ML model into ONNX
onnx_model = onnxmltools.convert_coreml(coreml_model)

# Save as protobuf
onnxmltools.utils.save_model(onnx_model, output_onnx_model)

print("Successfully done!")

ERROR:

Traceback (most recent call last):
  File "/media/sveta/DATASTORE/AI_ML_DL/Projects/PoseEstimation/CMU-PCL_OpenPose/convert_openpose/openpose2onnx.py", line 33, in <module>
    onnx_model = onnxmltools.convert_coreml(coreml_model)
  File "/home/sveta/.local/envs/pytorch_env/lib/python3.6/site-packages/onnxmltools/convert/main.py", line 18, in convert_coreml
    custom_conversion_functions, custom_shape_calculators)
  File "/home/sveta/.local/envs/pytorch_env/lib/python3.6/site-packages/onnxmltools/convert/coreml/convert.py", line 80, in convert
    onnx_model = convert_topology(topology, name, doc_string, target_opset, targeted_onnx)
  File "/home/sveta/.local/envs/pytorch_env/lib/python3.6/site-packages/onnxconverter_common/topology.py", line 737, in convert_topology
    registration.get_converter(operator.type)(scope, operator, container)
  File "/home/sveta/.local/envs/pytorch_env/lib/python3.6/site-packages/onnxmltools/convert/coreml/operator_converters/neural_network/Activation.py", line 27, in convert_activation
    apply_prelu(scope, inputs, outputs, container, operator_name=attrs['name'], slope=[params.PReLU.alpha])
  File "/home/sveta/.local/envs/pytorch_env/lib/python3.6/site-packages/onnxconverter_common/onnx_ops.py", line 522, in apply_prelu
    s_shape = slope.shape
AttributeError: 'list' object has no attribute 'shape'

MODEL:
prototxt
caffemodel

[milesxu]

I have got exactly the same error.

[Rajrup]

Encountered the same error.
@jiafatom Can you please help us in coverting openpose model?

[parameciumvance]

Also met the error.
Why is this issue not receiving a response since?

[Zhivago-Sizov]

the list object [slope] which passed to the func apply_prelu in onnxconverter_common/onnx_ops.py do not have shape attribute cause it's a list, without dive in too much, I make a quick fix simply turn the list into a NumPy array

# onnxconverter_common/onnx_ops.py
def apply_prelu(scope, input_name, output_name, container, operator_name=None, slope=None):
    slope=np.array(slope[0].floatValue)
    ...

[jiafatom]

@vuvalini thanks for your change. Generally please make changes on onnxmltools, not on onnxconverter_common. Because onnxconverter_common is shared by other converters. This will break other converters if the change is made.