Failed to parse UFF

AI & Data Science Deep Learning (Training & Inference) TensorRT
1

Hi there,

Please find the process and error below and any assistance offered will greatly benefit on my progress.

THE TERMINAL INPUT I GAVE:

nvidia@tegra-ubuntu:~/build/tf_to_trt_image_classification/scripts$ python convert_plan.py frozen_model.pb frozen_model.plan input 224 224 final_result 1 0 float

THE OUTPUT:

=== Automatically deduced input nodes ===
[name: “Placeholder”
op: “Placeholder”
attr {
key: “dtype”
value {
type: DT_FLOAT
}
}
attr {
key: “shape”
value {
shape {
dim {
size: -1
}
dim {
size: 299
}
dim {
size: 299
}
dim {
size: 3
}
}
}
}
]

Using output node final_result
Converting to UFF graph
Warning: No conversion function registered for layer: PlaceholderWithDefault yet.
Converting as custom op PlaceholderWithDefault input/BottleneckInputPlaceholder
name: “input/BottleneckInputPlaceholder”
op: “PlaceholderWithDefault”
input: “module_apply_default/hub_output/feature_vector/SpatialSqueeze”
attr {
key: “dtype”
value {
type: DT_FLOAT
}
}
attr {
key: “shape”
value {
shape {
dim {
size: -1
}
dim {
size: 2048
}
}
}
}

Warning: keepdims is ignored by the UFF Parser and defaults to True
No. nodes: 794
UFF Output written to data/tmp.uff
UFFParser: Validator error: input/BottleneckInputPlaceholder: Unsupported operation _PlaceholderWithDefault
Failed to parse UFF

Kind regards
Sri

2

Hello,

it looks like _PlaceholderWithDefault is a unsupported TRT operation.

For a list of supported ops: https://docs.nvidia.com/deeplearning/sdk/tensorrt-developer-guide/index.html#support_op

For any unsupported layers/operations plugins can be used.

Custom layers can be added to TensorRT network in Python using plugin nodes. https://docs.nvidia.com/deeplearning/sdk/tensorrt-developer-guide/index.html#add_custom_layer_python.

Checkout
/workspace/tensorrt/samples/python/uff_custom_plugin [python]
/workspace/tensorrt/samples/samplePlugin [c]

in your TRT installation/container for examples on howto work with the Plugin API.

1 Like