zmr0469
October 19, 2018, 10:42am
1
Tensorflow:1.10.0
TensorRT:TensorRT-5.0.0.10.Red-Hat.x86_64-gnu.cuda-9.0.cudnn7.3.tar.gz
System:Linux n22-145-156 4.9.0-0.bpo.5-amd64 #1 SMP Debian 4.9.65-3+deb9u2~bpo8+13 (2018-05-15) x86_64 GNU/Linux
When i run convert-to-uff got exception: “uff.model.exceptions.UffException: func had bad type or is not present”
Could any have a look at this problem? Thanks
zmr0469
October 19, 2018, 10:45am
2
Before got this exception, i got below warning “Warning: No conversion function registered for layer: ExpandDims yet”.
When i tried print some debug log in line 87 of uff/model/descriptor.py, i got below last printed-out information:
CODE:
def constraint(self, name, func, error, priority=None):
def _constraint(node, fields, extra_fields, shared_mem):
print(“========name”)
print(name)
print(“========func”)
print(func)
print(“========node”)
print(node)
print(“========fields”)
print(fields)
print(“========extra_fields”)
print(extra_fields)
res = func(node, fields, extra_fields, shared_mem)
if not res:
raise error
if priority is None:
self._priority += 1
priority = self._priority
self._constraints[name] = _Constraint(_constraint, priority)
return self
LOG:
========fields
{‘func’: s: “pow”
}
========extra_fields
{}
========name
field_func
========func
<function DescriptorOp._field_enum.._check_field at 0x7f931a955e18>
========node
id: “student/IAF/IAF_5/dilated_convolution/layer_3/dense/pow”
inputs: “student/IAF/IAF_5/dilated_convolution/layer_3/dense/Reshape”
operation: “Unary”
fields {
key: “func”
value {
s: “pow”
}
}
========fields
{‘func’: s: “pow”
}
========extra_fields
{}
NVES
October 19, 2018, 6:07pm
3
Hello,
can you clarify why you have TensorRT Red-Hat, but running on a Debian-based host?
RensorRT:TensorRT-5.0.0.10.Red-Hat.x86_64-gnu.cuda-9.0.cudnn7.3.tar.gz
System:Linux n22-145-156 4.9.0-0.bpo.5-amd64 #1 SMP Debian 4.9.65-3+deb9u2~bpo8+13 (2018-05-15) x86_64 GNU/Linux
NVES:
Hello,
can you clarify why you have TensorRT Red-Hat, but running on a Debian-based host?
RensorRT:TensorRT-5.0.0.10.Red-Hat.x86_64-gnu.cuda-9.0.cudnn7.3.tar.gz
System:Linux n22-145-156 4.9.0-0.bpo.5-amd64 #1 SMP Debian 4.9.65-3+deb9u2~bpo8+13 (2018-05-15) x86_64 GNU/Linux
I am not sure which version to use, but from the website, it seems that this version can match my needs (others are all about using under Ubuntu), if not, could you please help to tell me which version should i use (I am using cuda 9.1)? Thanks a lot.