[dynamo] Graph break on random_ op

Fixes https://github.com/pytorch/pytorch/issues/121621

ghstack-source-id: 098b44305ae2aaab334e6973f6e94f937e61f9a0
Pull Request resolved: https://github.com/pytorch/pytorch/pull/130222

test/dynamo/test_repros.py

@@ -5299,6 +5299,17 @@ def forward(self, s0 : torch.SymInt, s1 : torch.SymInt, L_x_ : torch.Tensor):
         # the second call causes a failure
         m()
 
+    # https://github.com/pytorch/pytorch/issues/121621
+    def test_tensor_random(self):
+        def random_op(tensor, params):
+            res = tensor.random_(**params)
+            return res
+
+        random_op = torch.compile(random_op)
+        params = {"from": -10, "to": 10}
+        tensor = torch.randn([2, 3])
+        res = random_op(tensor, params)
+
 
 instantiate_parametrized_tests(ReproTests)
 

torch/_dynamo/symbolic_convert.py

@@ -1495,16 +1495,27 @@ class InstructionTranslatorBase(
             null = self.pop()
             assert isinstance(null, NullVariable)
 
-        if (
-            isinstance(fn, GetAttrVariable)
-            and isinstance(fn.obj, TensorVariable)
-            and fn.name == "view"
-            and isinstance(argsvars, (ConstantVariable, TensorVariable))
-        ):
-            # Hack to handle special case in some bert models.  Converts
-            # x.view(*shape) into x.view(shape), which is correct for view()
-            # but not generally.  See test_transpose_for_scores().
-            argsvars = TupleVariable([argsvars])
+        if isinstance(fn, GetAttrVariable) and isinstance(fn.obj, TensorVariable):
+            if fn.name == "view" and isinstance(
+                argsvars, (ConstantVariable, TensorVariable)
+            ):
+                # Hack to handle special case in some bert models.  Converts
+                # x.view(*shape) into x.view(shape), which is correct for view()
+                # but not generally.  See test_transpose_for_scores().
+                argsvars = TupleVariable([argsvars])
+            elif (
+                fn.name == "random_"
+                and isinstance(argsvars, TupleVariable)
+                and len(argsvars.items) == 0
+                and isinstance(kwargsvars, ConstDictVariable)
+                and ConstantVariable.create("from") in kwargsvars
+            ):
+                # `from`` is python keyword. Adding random_ with `from` in the
+                # Fx graph causes syntax error. Even if we convert the kwargs to
+                # args, aot_autograd/inductor while lowering generates
+                # aten.random.from, again causing syntax errors. Since this
+                # usecase is uncommon, graph break.
+                unimplemented("random_ op is called with from keyword")
 
         if not isinstance(
             argsvars, BaseListVariable