Small bug fixes in eager mode (#691)

python/torch_mlir/eager_mode/ir_building.py

@@ -194,6 +194,9 @@ def get_func_op_with_name(module: ir.Module, name: str) -> Optional[FuncOp]:
         name_attr = ir.StringAttr.get(name)
     for op in module.body.operations:
         if isinstance(op, FuncOp) and op.name == name_attr:
+            # Add name of torch op as debug_module_name so that
+            # run_pipeline_with_repro_report can use it.
+            module.operation.attributes["torch.debug_module_name"] = name_attr
             return op
 
     return None

python/torch_mlir/eager_mode/torch_mlir_dispatch.py

@@ -105,6 +105,13 @@ def build_script_function(
                 inp.setDebugName(arg.name)
         # If arg is a constant, inline (at the top of the graph).
         else:
+            if val == []:
+                # Some ops have empty list default values for args
+                # (such as aten::max_pool2d_with_indices with int[2] stride=[]
+                # but graph.insertConstant doesnt' recognize [] as an empty list IValue.
+                # This might be an upstream bug but there doesn't seem to be a way to
+                # build a prim::ListConstruct list that's empty.
+                val = None
             inp = graph.insertConstant(val)
             inp.node().moveBefore(node)
 
@@ -219,7 +226,7 @@ def try_torch_mlir_eager(op, args, kwargs, backend):
     else:
         raise RuntimeError(f"op {op} has no name")
 
-    if op_name == "detach":
+    if "detach" in op_name:
         # We don't handle detach as it only pertains to autograd graph construction, which is handled by pytorch.
         raise UnsupportedByTorchMlirEagerMode("detaching")
 

python/torch_mlir/eager_mode/torch_mlir_tensor.py

@@ -41,7 +41,9 @@ class TorchMLIRTensor(torch.Tensor):
             dtype=elem.dtype,
             layout=elem.layout,
             device=elem.device,
-            requires_grad=kwargs.get("requires_grad", False) or elem.requires_grad,
+            # Only float tensors can have gradients.
+            requires_grad=elem.dtype in {torch.float, torch.float32, torch.float64}
+            and (kwargs.get("requires_grad", False) or elem.requires_grad),
         )
         r.elem = elem.detach() if r.requires_grad else elem
         return r