converting VariableV2 on tensorflow to UFF

Hi All,

I am converting the Inception V3 model to use tensorrt by using the guidelines mentioned in the tensorrt documentation. The operation VariableV2 of tensorflow is not supported yet. Is there a plan to add support for that or are there any guidelines that i can follow to do the conversion myself. I get the following errors:

Warning: No conversion function registered for layer: VariableV2 yet.
Converting as custom op VariableV2 InceptionV3/Logits/Conv2d_1c_1x1/weights
name: “InceptionV3/Logits/Conv2d_1c_1x1/weights”
op: “VariableV2”
attr {
key: “_class”
value {
list {
s: “loc:@InceptionV3/Logits/Conv2d_1c_1x1/weights”
}
}
}
attr {
key: “container”
value {
s: “”
}
}
attr {
key: “dtype”
value {
type: DT_FLOAT
}
}
attr {
key: “shape”
value {
shape {
dim {
size: 1
}
dim {
size: 1
}
dim {
size: 2048
}
dim {
size: 3
}
}
}
}
attr {
key: “shared_name”
value {
s: “”
}
}

You need to freeze your variables to constants.

Thanks a ton Thomas. That helped out a lot. I have another question:
Is there an easy way to use tfrecord files for inference with tensorrt? Any related example would help as well.

Probably not… TensorRT expects a gpu buffer as input. TensorFlow support is still… rough.