编译MXNet ResNet_18模型时，英特尔warp shuffle不支持width != warp_size

Question

编译英特尔集成显卡的MXNet ResNet_18模型时，使用以下命令：

 target = tvm.target.intel_graphics() 
 relay.build_module.build(mod, target, params=params)

获取错误：

File "./source/intrin_rule_opencl.cc", line 80
TVMError: 
---------------------------------------------------------------
An internal invariant was violated during the execution of TVM.
Please read TVM's error reporting guidelines.
More details can be found here: https://discuss.tvm.ai/t/error-reporting/7793.
---------------------------------------------------------------
  Check failed: (analyzer.CanProve(call->args[3] == call->args[4])) is false: Intel warp shuffle dose not support width != warp_size

我正在使用其模型动物园中的MXNet ResNet_18模型。

Answer 1

这个问题似乎与TVM有关，与MXNet无关。另外，请注意MXNet不支持“英特尔集成显卡”。请在https://github.com/apache/tvm/issues或https://github.com/dmlc/gluon-cv/issues中打开一个问题