Hi all,
I encountered an issue while building a model on LLVM using vm.relay.build(mod, target='llvm', target_host='llvm', params={}). The model contains MergeComposite to merge nn.conv2d + add.
I am able to reproduce the error using the code snippet test_merge_composite_llvm.py. In the code snippet, I created nn.conv2d + add IRModule and apply MergeComposite. You can do
# Run without MergeComposite. No error
python3 test_merge_composite_llvm.py
# Run with MergeComposite. Has error
python3 test_merge_composite_llvm.py -do_merge
The error message is as blow
....
11: tvm::relay::backend::MemoizedExprTranslator<tvm::runtime::Array<tvm::te::Tensor, void> >::VisitExpr(tvm::RelayExpr const&)
10: tvm::relay::ExprFunctor<tvm::runtime::Array<tvm::te::Tensor, void> (tvm::RelayExpr const&)>::VisitExpr(tvm::RelayExpr const&)
9: _ZZN3tvm5relay11ExprFunctorIFNS_7runtime5ArrayINS_2te6TensorEvEERKNS_
8: tvm::relay::tec::ScheduleBuilder::VisitExpr_(tvm::relay::CallNode const*)
7: tvm::relay::backend::MemoizedExprTranslator<tvm::runtime::Array<tvm::te::Tensor, void> >::VisitExpr(tvm::RelayExpr const&)
6: tvm::relay::ExprFunctor<tvm::runtime::Array<tvm::te::Tensor, void> (tvm::RelayExpr const&)>::VisitExpr(tvm::RelayExpr const&)
5: _ZZN3tvm5relay11ExprFunctorIFNS_7runtime5ArrayINS_2te6TensorEvEERKNS_
4: tvm::relay::tec::ScheduleBuilder::VisitExpr_(tvm::relay::CallNode const*)
3: tvm::relay::backend::MemoizedExprTranslator<tvm::runtime::Array<tvm::te::Tensor, void> >::VisitExpr(tvm::RelayExpr const&)
2: tvm::relay::ExprFunctor<tvm::runtime::Array<tvm::te::Tensor, void> (tvm::RelayExpr const&)>::VisitExpr(tvm::RelayExpr const&)
1: _ZZN3tvm5relay11ExprFunctorIFNS_7runtime5ArrayINS_2te6TensorEvEERKNS_
0: tvm::relay::tec::ScheduleBuilder::VisitExpr_(tvm::relay::ConstantNode const*)
File "../src/relay/backend/te_compiler_cache.cc", line 208
TVMError:
---------------------------------------------------------------
An error occurred during the execution of TVM.
For more information, please see: https://tvm.apache.org/docs/errors.html