get a problem…
After I convert a tensorflow pb file to uff file, then I want to use trt C++ API to create a engine with this uff file. But it shows error:
ERROR: UFFParser: Parser error: sub: Invalid scale mode, nbWeights: 3
ERROR: tensorrtNet: Fail to parse
how to fix it? here https://devtalk.nvidia.com/default/topic/1028602/tensorrt/tensorrt-no-conversion-function-for-realdiv/ is a similar problem, but it does not get resolved.
envs:
python2.7
tensorflow1.4
centos7.2
tensorrt5.0.2.6 source code
nvidia GPU p40
NVES_K
2
Can you share the code and .pb file with us so we can take a look at what’s going on? DM me directly if you don’t want to post it to the forum.
@KevinSchlichter I send a Private Message to you. Please check it.
@KevinSchlichter After I remove reshape op, it works. thank you
Hi. I am encountering similar problem. Could you please tell me how to fix it?
Specifically, my operation is
residuals = feature_map - clusters
, where clusters is a constant with shape[1, 1, 32, C] and feature_map is with shape[1, n, 1, C]