mantasp / barracuda-release Goto Github PK

I am trying to convert a ONNX model to the Barracuda format. My model uses NOTSET for the attribute auto_pad:
https://github.com/onnx/onnx/blob/master/docs/Operators.md#conv

According to the ONNX documentation that means that the padding is manually specified in the attribute pad. I assume that what it would be needed to copy the values in pad as done when auto_pad is not present, right?

So, a fix would be to include auto_pad == "NOTSET" in the check for absence of auto padding?

Hi,
I am trying to convert a keras model into a format that I can use in barracuda. Can you provide an example? I tried many solutions but I got errors (like in the title) when I am trying the conversation.

Dear Sir,
I am having some issues to convert a pb model to nn type format. My concern is that I need a specific version of python and tensorFlow. I am currently using python 3.5 and TensorFlow 1.8 but I got an error "data" referenced before ... when trying to use tensorflow to barracuda script. I wonder if it is because I need a specific version of python and tf? I am on Windows
Best regards
ps. trained a MobileNetV2 with 1 added dense layer that I used for transfer learning.

I get this error google.protobuf.message.DecodeError: Error parsing message. When trying to convert it to Barracuda.

the model is a pb file and i trained the model with Teachable Machine.

Hi all!
I convert a pytorch model to onnx with pytorch version 1.0,0, when load the onnx file using modelloder, error comes:

_model = ModelLoader.Load(myModelPath)

error:

NotSupportedException: Format version not supported: 241
Unity.Barracuda.ModelLoader.Load (System.IO.BinaryReader fileReader, System.Boolean verbose, System.Boolean applyPatching, System.Boolean skipWeights) (at D:/ThreeDPoseTracke_yunran_taobao_v2/ThreeDPoseTracker_yunran_taobao_v2/Library/PackageCache/[email protected]/Barracuda/Runtime/Core/ModelLoader.cs:70)
Unity.Barracuda.ModelLoader.Load (System.String filepath, System.Boolean verbose, System.Boolean skipWeights) (at D:/ThreeDPoseTracke_yunran_taobao_v2/ThreeDPoseTracker_yunran_taobao_v2/Library/PackageCache/[email protected]/Barracuda/Runtime/Core/ModelLoader.cs:40)
TmpBarracudaRunner.Start () (at Assets/Scripts/TmpBarracudaRunner.cs:232)

Does this mean my pytorch version is too new? But I guess pytorch 1.0.0 is a quite normal version for now. Could you please help me with this? Thanks in advance.

Hi, mantasp.
When I try to run my model on barracuda 2.4 with shaders above 2.2, a GroupMemoryBarrierWithGroupSync error appears on mobile devices. Maybe GL_MAX_COMPUTE_WORK_GROUP_INVOCATIONS not counted.