Segmentation fault when using Relay's Python API

Troubleshooting
1

Hi all,

I am trying to construct Relay expressions using the Python API. I am using the following code

‘’’ import tvm from tvm import relay from tvm.relay import var, Var, Function, Let, TensorType, Module from tvm.relay.prelude import Prelude

from network import Network

class GatherNDTest():
    def __init__(self, name="f", **kwargs):
        self.mod = Module()
        self.p = Prelude(self.mod)
        self.inputs = []
        self.f = relay.GlobalVar(name)
        self.mod[self.f] = relay.Function(self.inputs, self.build_impl(**kwargs), None)

    def input(self, i):
        self.inputs.append(i)
        return i

    def build_impl(self, vocab_size, embedding_size):
        # Input declarations
        embeddings_type = TensorType(shape = (vocab_size, embedding_size), dtype = "float32")
        embeddings = self.input(Var("embeddings", embeddings_type))

        word_index_type = TensorType(shape = (), dtype = "int32")
        word_index = self.input(Var("word_index", word_index_type))

        index_expr = tvm.relay.gather_nd(embeddings, word_index)
        return index_expr

mod = GatherNDTest(name = "gather_test", vocab_size = 200, embedding_size = 400).mod
print(mod)

‘’’

The backtrace from GDB is here (https://pastebin.com/cvbtn1tm). I am using commit e3eff20d8a91c2b986548f780493a8e30b0c2b7a on master.

In any case, I cannot figure out how to pass arguments to gather_nd, so I can index the embeddings tensor to get a single word embedding. Thanks!

2

Try to change the shape of word_index?

3

Do you mean making it a tensor instead of the scalar it is now? I tried changing the shape to (1, ) instead of the empty tuple it is now and I get the following error.

‘’’ Traceback (most recent call last):

  File "test.py", line 31, in <module>
    mod = GatherNDTest(name = "gather_test", vocab_size = 200, embedding_size = 400).mod

  File "test.py", line 14, in __init__
    self.mod[self.f] = relay.Function(self.inputs, self.build_impl(**kwargs), None)

  File "/home/ppf/data/ppf/projects/projects/rnn_compilers/incubator-tvm/python/tvm/relay/module.py", line 85, in __setitem__
    return self._add(var, val)

  File "/home/ppf/data/ppf/projects/projects/rnn_compilers/incubator-tvm/python/tvm/relay/module.py", line 94, in _add
    _module.Module_Add(self, var, val, update)

  File "/home/ppf/data/ppf/projects/projects/rnn_compilers/incubator-tvm/python/tvm/_ffi/_ctypes/function.py", line 207, in __call__
    raise get_last_ffi_error()

tvm._ffi.base.TVMError: Traceback (most recent call last):
  [bt] (8) /home/ppf/data/ppf/projects/projects/rnn_compilers/incubator-tvm/build/libtvm.so(tvm::NodeFunctor<void (tvm::runtime::ObjectRef const&, tvm::relay::ExprFunctor<void (tvm::relay::Expr const&)>*)>::operator()(tvm::runtime::ObjectRef const&, tvm::relay::ExprFunctor<void (tvm::relay::Expr const&)>*) const+0x151) [0x7fd7121430c3]
  [bt] (7) /home/ppf/data/ppf/projects/projects/rnn_compilers/incubator-tvm/build/libtvm.so(tvm::relay::ExprFunctor<void (tvm::relay::Expr const&)>::InitVTable()::{lambda(tvm::runtime::ObjectRef const&, tvm::relay::ExprFunctor<void (tvm::relay::Expr const&)>*)#5}::_FUN(tvm::runtime::ObjectRef const&, tvm::relay::ExprFunctor<void (tvm::relay::Expr const&)>*)+0x25) [0x7fd71214279e]
  [bt] (6) /home/ppf/data/ppf/projects/projects/rnn_compilers/incubator-tvm/build/libtvm.so(tvm::relay::ExprFunctor<void (tvm::relay::Expr const&)>::InitVTable()::{lambda(tvm::runtime::ObjectRef const&, tvm::relay::ExprFunctor<void (tvm::relay::Expr const&)>*)#5}::operator()(tvm::runtime::ObjectRef const&, tvm::relay::ExprFunctor<void (tvm::relay::Expr const&)>*) const+0x3e) [0x7fd712142772]
  [bt] (5) /home/ppf/data/ppf/projects/projects/rnn_compilers/incubator-tvm/build/libtvm.so(tvm::relay::WellFormedChecker::VisitExpr_(tvm::relay::FunctionNode const*)+0xd5) [0x7fd7122a9b29]
  [bt] (4) /home/ppf/data/ppf/projects/projects/rnn_compilers/incubator-tvm/build/libtvm.so(tvm::relay::WellFormedChecker::CheckWellFormed(tvm::relay::Expr const&)+0x23) [0x7fd7122a9d05]
  [bt] (3) /home/ppf/data/ppf/projects/projects/rnn_compilers/incubator-tvm/build/libtvm.so(tvm::relay::WellFormedChecker::VisitExpr(tvm::relay::Expr const&)+0x4f) [0x7fd7122a9cdf]
  [bt] (2) /home/ppf/data/ppf/projects/projects/rnn_compilers/incubator-tvm/build/libtvm.so(tvm::relay::ExprVisitor::VisitExpr(tvm::relay::Expr const&)+0xa3) [0x7fd7123a3f53]
  [bt] (1) /home/ppf/data/ppf/projects/projects/rnn_compilers/incubator-tvm/build/libtvm.so(tvm::relay::ExprFunctor<void (tvm::relay::Expr const&)>::VisitExpr(tvm::relay::Expr const&)+0x98) [0x7fd712142108]
  [bt] (0) /home/ppf/data/ppf/projects/projects/rnn_compilers/incubator-tvm/build/libtvm.so(dmlc::LogMessageFatal::~LogMessageFatal()+0x4a) [0x7fd711a5527c]
  File "/home/ppf/Documents/projects/rnn_compilers/incubator-tvm/include/tvm/relay/expr_functor.h", line 92
TVMError: Check failed: n.defined():

‘’’