Just wondering when and how this will be released? Nvidia press emails make it sound like it’s already available for Xavier, but I don’t see it in the SDK Manager… Will it come with the next JetPack release or before that?
dear all,
How do I check my TensorRT version that came with JetPack 4.2?
I cannot issue
> nm -D /usr/lib/<ARCH>/libnvinfer.so.<VERSION> | grep tensorrt_version
> nm -D /usr/lib/aarch64-linux-gnu/libnvinfer.so.5.0.3 | grep tensorrt_version
There is no libnvinfer* on my ubuntu 18.04 (my HP Omen desktop 870)
$ nm -D /usr/lib/x86_64-linux-gnu/libn*
libnautilus-extension.so.1 libnetsnmphelpers.so.30.0.3 libnorm.so.1.0.0
libnautilus-extension.so.1.4.0 libnetsnmpmibs.so.30 libnotify.so.4
libndp.so.0 libnetsnmpmibs.so.30.0.3 libnotify.so.4.0.0
libndp.so.0.1.0 libnetsnmp.so.30 libnpth.so.0
libndr-krb5pac.so.0 libnetsnmp.so.30.0.3 libnpth.so.0.1.1
libndr-krb5pac.so.0.0.1 libnetsnmptrapd.so.30 libnsl.a
libndr-nbt.so.0 libnetsnmptrapd.so.30.0.3 libnsl.so
libndr-nbt.so.0.0.1 libnettle.so.6 libnspr4.so
libndr.so.0 libnettle.so.6.4 libnss3.so
libndr.so.0.1.0 libnfnetlink.so.0 libnss_compat.so
libndr-standard.so.0 libnfnetlink.so.0.2.0 libnss_dns.so
libndr-standard.so.0.0.1 libnghttp2.so.14 libnss_files.so
libneon-gnutls.so.27 libnghttp2.so.14.15.2 libnss_hesiod.so
libneon-gnutls.so.27.3.2 libnice.so.10 libnss_nisplus.so
libnetapi.so.0 libnice.so.10.7.0 libnss_nis.so
libnetfilter_conntrack.so.3 libnma.so.0 libnssutil3.so
libnetfilter_conntrack.so.3.6.0 libnma.so.0.0.0 libnuma.so.1
libnetsnmpagent.so.30 libnm.so.0 libnuma.so.1.0.0
libnetsnmpagent.so.30.0.3 libnm.so.0.1.0
libnetsnmphelpers.so.30 libnorm.so.1
$ nm -D /usr/lib/aarch64-linux-gnu/
gstreamer-1.0/ ldscripts/ libgstnvdsmeta.so tegra/
I already do install JetPack 4.2 successfully (installed on the host and Xavier).
Where is that TensorRT?
Do I need to download TensorRT 5.0 for Ubuntu 18.04 and Cuda 10.0 from https://developer.nvidia.com/nvidia-tensorrt-5x-download and install it manually?
Thank you very much in advance.
warmest regards,
suryadi
Hello All,
Jetpack 4.2 uses TensorRT 5.0.6.
I want to update TRT on Xavier to use python calibration example.
Moreover, there are performance improvements upgrading my TRT.
How can I safely upgrade my TRT to the latest version? Does downloading deb package and installing it without waiting for a newer JetPack version work?
Thanks,
Hi cmehrshad, the versions provided through the TensorRT webpage are not compatible with JetPack and aarch64 architecture. An upgrade to TensorRT 5.1 will be provided in the next release of JetPack and the NVIDIA SDK Manager.
Thanks for the explanation.
Is there any python calibration example available for the current TRT version (5.0.0-5.0.6)?