How to resolve unsupported layer issue for my model

[varun-tangoit]

Hi,

Im currently working on facenet,mtcnn model while i need to optimize my model with TensorRT. I have .pb file for model, when i try to convert with uff model i have some error kind of;

uff_model = uff.from_tensorflow_frozen_model("20180518-115854.pb", ["embeddings"])

Warning: keepdims is ignored by the UFF Parser and defaults to True
Warning: No conversion function registered for layer: QueueDequeueUpToV2 yet.
Converting as custom op QueueDequeueUpToV2 batch_join
name: "batch_join"
op: "QueueDequeueUpToV2"
input: "batch_join/fifo_queue"
input: "batch_size"
attr {
key: "component_types"
value {
list {
type: DT_FLOAT
type: DT_INT64
}
}
}
attr {
key: "timeout_ms"
value {
i: -1
}
}

Warning: No conversion function registered for layer: FIFOQueueV2 yet.
Converting as custom op FIFOQueueV2 batch_join/fifo_queue
name: "batch_join/fifo_queue"
op: "FIFOQueueV2"
attr {
key: "capacity"
value {
i: 1440
}
}
attr {
key: "component_types"
value {
list {
type: DT_FLOAT
type: DT_INT64
}
}
}
attr {
key: "container"
value {
s: ""
}
}
attr {
key: "shapes"
value {
list {
shape {
dim {
size: 160
}
dim {
size: 160
}
dim {
size: 3
}
}
shape {
}
}
}
}
attr {
key: "shared_name"
value {
s: ""
}
}

Traceback (most recent call last):
File "", line 1, in
File "/usr/lib/python3.5/dist-packages/uff/converters/tensorflow/conversion_helpers.py", line 149, in from_tensorflow_frozen_model
return from_tensorflow(graphdef, output_nodes, preprocessor, **kwargs)
File "/usr/lib/python3.5/dist-packages/uff/converters/tensorflow/conversion_helpers.py", line 120, in from_tensorflow
name="main")
File "/usr/lib/python3.5/dist-packages/uff/converters/tensorflow/converter.py", line 76, in convert_tf2uff_graph
uff_graph, input_replacements)
File "/usr/lib/python3.5/dist-packages/uff/converters/tensorflow/converter.py", line 63, in convert_tf2uff_node
op, name, tf_node, inputs, uff_graph, tf_nodes=tf_nodes)
File "/usr/lib/python3.5/dist-packages/uff/converters/tensorflow/converter.py", line 38, in convert_layer
fields = cls.parse_tf_attrs(tf_node.attr)
File "/usr/lib/python3.5/dist-packages/uff/converters/tensorflow/converter.py", line 209, in parse_tf_attrs
for key, val in attrs.items()}
File "/usr/lib/python3.5/dist-packages/uff/converters/tensorflow/converter.py", line 209, in
for key, val in attrs.items()}
File "/usr/lib/python3.5/dist-packages/uff/converters/tensorflow/converter.py", line 204, in parse_tf_attr_value
return cls.convert_tf2uff_field(code, val)
File "/usr/lib/python3.5/dist-packages/uff/converters/tensorflow/converter.py", line 189, in convert_tf2uff_field
'type': 'dtype', 'list': 'list'}[code]
KeyError: 'shape'

I can't proceed further, im stuck on this, pls give me some suggestion how to resolve this issue

[ningjieliu]

The same error with you. Have you solved this problem?

[varun-tangoit]

Nope.. If you have resolved, pls let me know.

[ningjieliu]

I tried to use insightface to replace facenet and it works. The technology support in nvidia said that facenet has some un-support layers so that it cannot directly convert to uff unless you have the caffemodel. It seems that the caffemodel has plug-in API.

[varun-tangoit]

Can you please help me with more detail or where do refer and resolve this issue.. im stuck on this issue. Thanks in advance

[ningjieliu]

In fact, I use the resnet_v1_50 model in this github: https://github.com/auroua/InsightFace_TF
and then convert .ckpt file to .pd file. Finally use uff.from_tensorflow() to uff. I just entirely give up the facenet. Hope it can help you.

[varun-tangoit]

Thanks for the help. i will try.. Once you convert tensorRT please tell me how much fps would reached on jetson tx2.. i need to know performance increased or not for facenet model.

[deaffella]

In fact, I use the resnet_v1_50 model in this github: https://github.com/auroua/InsightFace_TF
and then convert .ckpt file to .pd file. Finally use uff.from_tensorflow() to uff. I just entirely give up the facenet. Hope it can help you.

Hi!
Have you solve this issue? I want to convert facenet to trt model too but I ran into a lot of problems