Closed PaulCahuana closed 1 year ago
Hey guys, I have the same issue when trying to convert this OpenSee model using the OpenVINO Workbench.
Hello @PaulCahuana, @campos537,
Thank you for reaching OpenVINO!
This operation is not from the standard ONNX opset while from the Microsoft ONNX RT extended opset. Currently supported fused ops are listed here
We should discuss this internally how to properly proceed with that. Stay tuned!
CC @mlukasze @tomdol
Ref. 92500
Hello @PaulCahuana, @campos537,
Thank you for reaching OpenVINO!
This operation is not from the standard ONNX opset while from the Microsoft ONNX RT extended opset. Currently supported fused ops are listed here
We should discuss this internally how to properly proceed with that. Stay tuned!
CC @mlukasze @tomdol
Sure. Thanks for all!
Hello @PaulCahuana!
The support of FusedConv
was added in https://github.com/openvinotoolkit/openvino/pull/13553
The model lm_model3_opt.onnx
can be loaded and inference now.
Tested via benchmark_app
(./benchmark_app -shape [1,3,224,224] -m lm_model3_opt.onnx
).
Please let me know, if the change conver all your cases and if we can close the issue.
Wow, that is really great! This model is awesome!
Detailing Information presented as follows
Microsoft ONNX Runtime is an open source inference accelerator focused on ONNX models. It is the platform Vitis AI has integrated with to provide first-class ONNX model support, which can be exported from a wide variety of training frameworks. It incorporates very easy to use runtime APIs in Python and C++ and can support models without requiring the separate compilation phase that TVM requires. Included in ONNXRuntime is a partitioner that can automatically partition between the CPU and FPGA further enhancing the ease of model deployment. Finally, it also incorporates the Vitis AI quantizer in a way that does not require separate quantization setup.
com.microsoft.FusedConv
The fused convolution operator schema is the same as Convolution besides it includes an attribute activation.
This version of the operator has been available since version 1 of the 'com.microsoft' operator set. Attributes
activation : string activation_params : list of floats auto_pad : string dilations : list of ints group : int kernel_shape : list of ints pads : list of ints strides : list of ints
Inputs (2 - 4)
X : T W : T B (optional) : T Z (optional) : T
Outputs
Y : T
Type Constraints
T : tensor(float16), tensor(float), tensor(double) Constrain input and output types to float tensors
Closing this, as PR with support for FusedConv op has been merged. Feel free to reopen to ask any questions related to this topic.
System information
Detailed description
I was trying to load the model with openvino but I got the error: "RuntimeError: Check 'unknown_operators.empty()' failed at frontends/onnx/frontend/src/core/graph.cpp:133: OpenVINO does not support the following ONNX operations: com.microsoft.FusedConv"
Steps to reproduce
from openvino.runtime import Core ie = Core() model= ie.read_model(set_config_value("lm_model3_opt.onnx"))
So, do you have plans to add the "com.microsoft.FusedConv" layer?