L0phTg commented on PR #15544:
URL: https://github.com/apache/tvm/pull/15544#issuecomment-1676291622

   case 1:
   ```python
   @I.ir_module
   class Module:
       @R.function
       def func(
           A: R.Tensor(("m", "m + n"), dtype="float32")
       ) -> R.Tensor(dtype="float32", ndim=2):
           m = T.int64()
           n = T.int64()
           k = T.int64()
           with R.dataflow():
               lv: R.Tensor((m, k), dtype="float32") = R.match_cast(
                   A, R.Tensor((m, k), dtype="float32")
               )
               gv: R.Tensor((m, k), dtype="float32") = lv
               R.output(gv)
           return gv
   ```
   after pipeline:
   ```python
   mod_after_pipeline = relax.pipeline.get_pipeline()(mod)
   ```
   will cause:
   ```bash
    6: tvm::relax::GraphCreator::VisitBindingBlock(tvm::relax::BindingBlock 
const&)
     5: 
tvm::relax::ExprVisitor::VisitBindingBlock_(tvm::relax::DataflowBlockNode 
const*)
     4: tvm::relax::ExprVisitor::VisitBinding(tvm::relax::Binding const&)
     3: tvm::relax::GraphCreator::VisitBinding_(tvm::relax::VarBindingNode 
const*)
     2: tvm::relax::PostOrderVisit(tvm::RelayExpr const&, std::function)
     1: std::_Function_handler::_M_invoke(std::_Any_data const&, tvm::RelayExpr 
const&)
     0: tvm::relax::GraphCreator::VisitLeaf(tvm::RelayExpr const&, 
tvm::relay::IndexedForwardGraph::Node*, tvm::relay::OpPatternKind const&)
     File 
"/media/l0phtg/movable/tools-git/tvm/src/relax/transform/fuse_ops.cc", line 269
   TVMError: The leaf Expr is supposed to be defined before, but got: lv used 
before definition.
   ```


-- 
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.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to