grapeot edited a comment on issue #16042: Error when calling get_backend_symbol URL: https://github.com/apache/incubator-mxnet/issues/16042#issuecomment-538795957 @Caenorst I can see a misunderstanding here that you meant to build MXNet with TensorRT support but @Zheweiqiu thought it meant to build the model. I hit this problem with exactly the same error messages using the default python package, and am willing to dig deeper by rebuilding MXNet with TensorRT support. ~~However, I don't see instructions in the [doc](https://github.com/apache/incubator-mxnet/blob/1.5.1/docs/install/build_from_source.md), or the [make options](https://github.com/apache/incubator-mxnet/blob/master/make/config.mk). Wondering do you mind pointing us to the proper doc? Thanks for your time!~~ Update: found the switch USE_TENSORRT in [Cmake file](https://github.com/apache/incubator-mxnet/blob/1.5.1/CMakeLists.txt). Update: found this auto build [script](https://github.com/apache/incubator-mxnet/blob/master/ci/docker/runtime_functions.sh#L679). Update: was able to successfully build MXNet with TensorRT support by following the aforementioned script (used CUDA 10.1, CuDNN 7.6, TensorRT 6.0). But the code got stuck for half an hour with 100% CPU on one core, before it throws a ton of error messages on some nodes have a cycle. Anyway the original error message disappeared. But I was trying on a different model. @Zheweiqiu you may want to try this approach. It may work on your model.
---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected] With regards, Apache Git Services
