We’ve wasted days trying to work through the horrors of this over-engineered bug-ridden jetpack flashing process.
There’s got to be a better way of doing this.
If I want to install a raspbian image on a raspberry pi, I just burn it to a microSD card, stick it in, and turn it on. DONE!
Something similar has to be possible for getting the jetpack onto the tx2.
If not, why not?
Do such images exist?
What is the procedure for transferring them to the TX2?
Hi alexryan, the procedure for flashing the L4T BSP needs to be performed from a host PC running Ubuntu 16.04 natively, using the provided micro-USB cable while the Jetson has been booted into recovery mode.
The post-flashing packages that JetPack installs (like CUDA toolkit, cuDNN, TensorRT, VisionWorks, ect) are typically installed by JetPack over the network via SSH, however if desired you can manually copy these packages from the jetpack_downloads/ folder on your host PC to your Jetson and manually install those.
If you continue to have issues, can you post more information about the behavior you are encountering along with the configuration and OS of your host PC?
Somebody should write a proper step by step jetpack install tutorial showing the potential pitfalls and how to setup and test interweb connection using wifi / bluetooth. I could list at least 5 major problems that i personally encountered. Even with help from guys on this forum, few of the solutions were obvious. One particularly nasty obstacle sticks in my mind - the hidden menu enabling installing cuda CDNn etc without re-installing the OS.
Maybe it’s some kind of ‘rights of passage’ cos everything afterwards seems to be a lot easier!