Problem installing node-red-contrib-tf-model on Jetson Nano

[dcaputo-harmoni]

I am having an issue installing on my Jetson Nano. Specifically, I get the following error upon install:

npm WARN [email protected] requires a peer of @tensorflow/tfjs-node@^1.4.0 but none is installed. You must install peer dependencies yourself.

This happens despite installing tensorflow previously using:

JOBS=MAX npm install -g --production --unsafe-perm @tensorflow/[email protected]

After which I see the error below, which I understand is OK to disregard:

CPU-linux-1.7.3.tar.gz
* Downloading libtensorflow
(node:19) UnhandledPromiseRejectionWarning: Error: Unsupported system: cpu-linux-arm64

Any ideas as to what might be wrong? I am not running Jetpack but rather running node-red in an ubuntu 18.04 docker container with the latest CUDA/cuDNN versions 10.2.89/8.0.0.145 and L4T config/drivers version 32.4.2.

[yhwang]

one question: are you able to find the @tensorflow/tfjs-node that you install globally? If yes, can you go there, following the instruction of using custom tensorflow shared lib and do the npm install:

For the Jetson Nano, you need to provide a file named custom-binary.json under the scripts directory with the following contents:
{
  "tf-lib": "https://s3.us.cloud-object-storage.appdomain.cloud/tfjs-cos/libtensorflow-gpu-linux-arm64-1.15.0.tar.gz"
}

My guess is that you may use the tensorflow shared lib that I built on Jetpack still. Let me know if you can do that.