Can't load the training model

[Trick3r]

When I have tried to loading the model, I got this error
"{"NodeDef mentions attr 'explicit_paddings' not in Op<name=Conv2D; signature=input:T, filter:T -> output:T; attr=T:type,allowed=[DT_HALF, DT_BFLOAT16, DT_FLOAT, DT_DOUBLE]; attr=strides:list(int); attr=use_cudnn_on_gpu:bool,default=true; attr=padding:string,allowed=["SAME", "VALID"]; attr=data_format:string,default="NHWC",allowed=["NHWC", "NCHW"]; attr=dilations:list(int),default=[1, 1, 1, 1]>; NodeDef: {{node conv1/convolution}} = Conv2D[T=DT_FLOAT, data_format="NHWC", dilations=[1, 1, 1, 1], explicit_paddings=[], padding="SAME", strides=[1, 1, 1, 1], use_cudnn_on_gpu=true](conv1_input, conv1/kernel/read). (Check whether your GraphDef-interpreting binary is up to date with your GraphDef-generating binary.)."}"

What am I doing wrong? here is the code for loading model
folderWithMNIST = @"D:\Tensorflow_C#\MNISTpng-master\MNISTpng\testing";

the rest of the code is same as you use

[sdg002]

Hi @Trick3r ,
The resolution in migueldeicaza/TensorFlowSharp#444
might help you. In my case, the native Tensorlfow DLL had to be manually updated to get the C# to work.

[Trick3r]

@sdg002 ,
Thx for your reply, but when I have do it like you said in that forum, I got the same error, it seem install the tensor gpu version 1.13.1 doesn't help

[sdg002]

My suggestion would be

1. Use 1.14 of TF while training
2. In your C# app, copy over the native tensorflow.dll from TF site and rename it as libtensorflow.dll in the Bin of your project. You will have to be careful because libtensorflow.dll from Nuget will keep updating the Bin folder. This is because TensorFlowSharp is delivering an older version of the native tensorflow.dll.

[Trick3r]

@sdg002 ok, thx, but could you please tell me, the model in the project that you have posted in Github, what version you use to train this model?, because I have try to run your code with your model and got the error

[geometrikal]

Please see my comment here for an easier way to get the dll:

migueldeicaza/TensorFlowSharp#444 (comment)