building tensorflow lite from source on tx2 failed

Hi, I try to build tensorflow lite to evaluate the performance of resnet on tx2. But I encountered a problem and have not found ways to solve them.
I download the master branch code of tensorflow from github, and use the following command to build the tensorflow lite convertor:

bazel build //tensorflow/lite/python:tflite_convert --local_ram_resources=“HOST_RAM*.9” --local_cpu_resources=4 --verbose_failures --config=nonccl

The local resources flags are used to restrict resources occupied by bazel to avoid insufficient ram problem. The ‘config=nonccl’ flag is to set a non-nccl building, as there is no nccl on tx2. However, a nccl-related building error appeared.

//…many similar NVLink errors, listing the last few below along with final error log.
nvlink error : entry function ‘_Z28ncclAllReduceLLKernel_sum_i88ncclColl’ with max regcount of 80 calls function ‘_Z25ncclReduceScatter_max_u64P14CollectiveArgs’ with regcount of 96
nvlink error : entry function ‘_Z29ncclAllReduceLLKernel_sum_i328ncclColl’ with max regcount of 80 calls function ‘_Z25ncclReduceScatter_max_u64P14CollectiveArgs’ with regcount of 96
nvlink error : entry function ‘_Z29ncclAllReduceLLKernel_sum_f168ncclColl’ with max regcount of 80 calls function ‘_Z25ncclReduceScatter_max_u64P14CollectiveArgs’ with regcount of 96
nvlink error : entry function ‘_Z29ncclAllReduceLLKernel_sum_u328ncclColl’ with max regcount of 80 calls function ‘_Z25ncclReduceScatter_max_u64P14CollectiveArgs’ with regcount of 96
nvlink error : entry function ‘_Z29ncclAllReduceLLKernel_sum_f328ncclColl’ with max regcount of 80 calls function ‘_Z25ncclReduceScatter_max_u64P14CollectiveArgs’ with regcount of 96
nvlink error : entry function ‘_Z29ncclAllReduceLLKernel_sum_u648ncclColl’ with max regcount of 80 calls function ‘_Z25ncclReduceScatter_max_u64P14CollectiveArgs’ with regcount of 96
nvlink error : entry function ‘_Z28ncclAllReduceLLKernel_sum_u88ncclColl’ with max regcount of 80 calls function ‘_Z25ncclReduceScatter_max_u64P14CollectiveArgs’ with regcount of 96

It seems the error is related to some nccl kernels. I dont know why nccl-related kernels were still compiled.

I have solved this problem. Xla still depends on the nccl and set the ‘–config=nonccl’ can not prevent xla from building nccl. My sulution is turn off the xla support.

Hi,

YES. NCCL is not available on the Jetson platform.
Here are the steps shared by others for your reference:
[url]https://devtalk.nvidia.com/default/topic/1055131/jetson-agx-xavier/building-tensorflow-1-13-on-jetson-xavier/[/url]

Thanks.