[MSFT] Workaround for BNNS compilation

[KJlaccHoeUM9l]

For two models (aec, cruse) it was not possible to apply BNNS. The script for compiling the model stops working with an error when executing the partition_for_bnns function. This is due to the fact that the check_dense function does not take into account the fact that the operator can be commutative (an internal assumption, which in our case was not fulfilled).

These changes are a workaround for this problem.

The following error occurred while compiling models:

Traceback (most recent call last):
  File "/Users/agladyshev/workspace/tvm-samples/deploy_flows/flow3c/compile_local_cpu_no_tune.py", line 48, in <module>
    mod = partition_for_bnns(mod)
  File "/Users/agladyshev/workspace/tvm/python/tvm/relay/op/contrib/bnns.py", line 73, in partition_for_bnns
    return seq(mod)
  File "/Users/agladyshev/workspace/tvm/python/tvm/ir/transform.py", line 127, in __call__
    return _ffi_transform_api.RunPass(self, mod)
  File "/Users/agladyshev/workspace/tvm/python/tvm/_ffi/_ctypes/packed_func.py", line 237, in __call__
    raise get_last_ffi_error()
AttributeError: Traceback (most recent call last):
  [bt] (8) 9   libtvm.dylib                        0x000000011a94b498 tvm::runtime::TVMRetValue tvm::runtime::PackedFunc::operator()<tvm::RelayExpr const&>(tvm::RelayExpr const&) const + 232
  [bt] (7) 8   libtvm.dylib                        0x00000001186f8d51 std::__1::function<void (tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*)>::operator()(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const + 65
  [bt] (6) 7   libtvm.dylib                        0x00000001186f8f8a std::__1::__function::__value_func<void (tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*)>::operator()(tvm::runtime::TVMArgs&&, tvm::runtime::TVMRetValue*&&) const + 106
  [bt] (5) 6   libtvm.dylib                        0x000000011b57e738 std::__1::__function::__func<TVMFuncCreateFromCFunc::$_2, std::__1::allocator<TVMFuncCreateFromCFunc::$_2>, void (tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*)>::operator()(tvm::runtime::TVMArgs&&, tvm::runtime::TVMRetValue*&&) + 72
  [bt] (4) 5   libtvm.dylib                        0x000000011b57fc67 std::__1::__function::__alloc_func<TVMFuncCreateFromCFunc::$_2, std::__1::allocator<TVMFuncCreateFromCFunc::$_2>, void (tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*)>::operator()(tvm::runtime::TVMArgs&&, tvm::runtime::TVMRetValue*&&) + 71
  [bt] (3) 4   libtvm.dylib                        0x000000011b57fcb7 void std::__1::__invoke_void_return_wrapper<void>::__call<TVMFuncCreateFromCFunc::$_2&, tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*>(TVMFuncCreateFromCFunc::$_2&, tvm::runtime::TVMArgs&&, tvm::runtime::TVMRetValue*&&) + 71
  [bt] (2) 3   libtvm.dylib                        0x000000011b57fd53 decltype(std::__1::forward<TVMFuncCreateFromCFunc::$_2&>(fp)(std::__1::forward<tvm::runtime::TVMArgs>(fp0), std::__1::forward<tvm::runtime::TVMRetValue*>(fp0))) std::__1::__invoke<TVMFuncCreateFromCFunc::$_2&, tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*>(TVMFuncCreateFromCFunc::$_2&, tvm::runtime::TVMArgs&&, tvm::runtime::TVMRetValue*&&) + 115
  [bt] (1) 2   libtvm.dylib                        0x000000011b57fe24 TVMFuncCreateFromCFunc::$_2::operator()(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const + 180
  [bt] (0) 1   libtvm.dylib                        0x000000011b5d8495 tvm::runtime::Backtrace() + 37
  File "/Users/agladyshev/workspace/tvm/python/tvm/runtime/object.py", line 63, in __getattr__
    return _ffi_node_api.NodeGetAttr(self, name)
  File "/Users/agladyshev/workspace/tvm/python/tvm/_ffi/_ctypes/packed_func.py", line 237, in __call__
    raise get_last_ffi_error()
  [bt] (8) 9   libtvm.dylib                        0x0000000118b90c87 std::__1::__function::__alloc_func<void (*)(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*), std::__1::allocator<void (*)(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*)>, void (tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*)>::operator()(tvm::runtime::TVMArgs&&, tvm::runtime::TVMRetValue*&&) + 71
  [bt] (7) 8   libtvm.dylib                        0x0000000118b90cd7 void std::__1::__invoke_void_return_wrapper<void>::__call<void (*&)(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*), tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*>(void (*&)(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*), tvm::runtime::TVMArgs&&, tvm::runtime::TVMRetValue*&&) + 71
  [bt] (6) 7   libtvm.dylib                        0x0000000118b90d73 decltype(std::__1::forward<void (*&)(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*)>(fp)(std::__1::forward<tvm::runtime::TVMArgs>(fp0), std::__1::forward<tvm::runtime::TVMRetValue*>(fp0))) std::__1::__invoke<void (*&)(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*), tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*>(void (*&)(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*), tvm::runtime::TVMArgs&&, tvm::runtime::TVMRetValue*&&) + 115
  [bt] (5) 6   libtvm.dylib                        0x0000000118b7fe11 tvm::NodeGetAttr(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) + 657
  [bt] (4) 5   libtvm.dylib                        0x0000000118b7e136 tvm::ReflectionVTable::GetAttr(tvm::runtime::Object*, tvm::runtime::String const&) const + 1014
  [bt] (3) 4   libtvm.dylib                        0x000000011839f195 tvm::runtime::detail::LogFatal::~LogFatal() + 21
  [bt] (2) 3   libtvm.dylib                        0x00000001183a141d tvm::runtime::detail::LogFatal::~LogFatal() + 29
  [bt] (1) 2   libtvm.dylib                        0x00000001183a14cc tvm::runtime::detail::LogFatal::Entry::Finalize() + 156
  [bt] (0) 1   libtvm.dylib                        0x000000011b5d8495 tvm::runtime::Backtrace() + 37
  File "/Users/agladyshev/workspace/tvm/src/node/reflection.cc", line 110
  File "/Users/agladyshev/workspace/tvm/python/tvm/_ffi/_ctypes/packed_func.py", line 81, in cfun
    rv = local_pyfunc(*pyargs)
  File "/Users/agladyshev/workspace/tvm/python/tvm/relay/op/contrib/bnns.py", line 266, in check_dense
    while call.op.name != "nn.dense":
  File "/Users/agladyshev/workspace/tvm/python/tvm/runtime/object.py", line 65, in __getattr__
    raise AttributeError("%s has no attribute %s" % (str(type(self)), name))
AttributeError: relay.TupleGetItem object has no attributed op
During handling of the above exception, another exception occurred:
AttributeError: <class 'tvm.relay.expr.TupleGetItem'> has no attribute op

[echuraev]

Also, I'm not sure that it is a good fix for the problem. Maybe I don't understand the whole problem.
I think that this problem should be fixed somewhere in BNNS part, for example in codegen.cc or in fixing existing pattern for BNNS in bnns.py.

Are you sure that your change won't brake something in other tvm components?

[echuraev]

Remove the comment and redundant brackets around op_node->name == "multiply"