Exported Programs with Dynamic Shapes¶
The following script shows the exported program for many short cases and various way to retrieve an ONNX model equivalent to the original model. The tested scenarios are described at Tested Scenarios.
<<<
import inspect
import textwrap
import pandas
from experimental_experiment.torch_interpreter.eval import discover, run_exporter
from experimental_experiment.ext_test_case import unit_test_going
cases = discover()
print()
print(":ref:`Summary <led-summary-exported-program>`")
print()
sorted_cases = sorted(cases.items())
if unit_test_going():
sorted_cases = sorted_cases[:3]
for name, cls_model in sorted_cases:
print(f"* :ref:`{name} <led-model-case-export-{name}>`")
print()
obs = []
for name, cls_model in sorted(cases.items()):
print()
print(f".. _led-model-case-export-{name}:")
print()
print(name)
print("=" * len(name))
print()
print("forward")
print("+++++++")
print()
print("::")
print()
print(
textwrap.indent(textwrap.dedent(inspect.getsource(cls_model.forward)), " ")
)
print()
for exporter in (
"export-strict",
"export-strict-decall",
"export-nostrict",
"export-nostrict-decall",
"export-jit",
"export-jit-decall",
"export-tracing",
):
expname = exporter.replace("export-", "")
print()
print(expname)
print("+" * len(expname))
print()
res = run_exporter(exporter, cls_model, True, quiet=True)
case_ref = f":ref:`{name} <led-model-case-export-{name}>`"
expo = exporter.split("-", maxsplit=1)[-1]
if "exported" in res:
print("::")
print()
print(textwrap.indent(str(res["exported"].graph), " "))
print()
obs.append(dict(case=case_ref, error="", exporter=expo))
else:
print("**FAILED**")
print()
print("::")
print()
print(textwrap.indent(str(res["error"]), " "))
print()
obs.append(dict(case=case_ref, error="FAIL", exporter=expo))
print()
print(".. _led-summary-exported-program:")
print()
print("Summary")
print("+++++++")
print()
df = pandas.DataFrame(obs)
piv = df.pivot(index="case", columns="exporter", values="error")
print(piv.to_markdown(tablefmt="rst"))
print()
>>>
AtenAsStrided¶
forward¶
def forward(self, x):
y = torch.as_strided(x, (2, 2, 8, 4), (128, 8, 16, 1))
return y
strict¶
graph():
%x : [num_users=1] = placeholder[target=x]
%as_strided : [num_users=1] = call_function[target=torch.ops.aten.as_strided.default](args = (%x, [2, 2, 8, 4], [128, 8, 16, 1]), kwargs = {})
return (as_strided,)
strict-decall¶
graph():
%x : [num_users=1] = placeholder[target=x]
%as_strided : [num_users=1] = call_function[target=torch.ops.aten.as_strided.default](args = (%x, [2, 2, 8, 4], [128, 8, 16, 1]), kwargs = {})
return (as_strided,)
nostrict¶
graph():
%x : [num_users=1] = placeholder[target=x]
%as_strided : [num_users=1] = call_function[target=torch.ops.aten.as_strided.default](args = (%x, [2, 2, 8, 4], [128, 8, 16, 1]), kwargs = {})
return (as_strided,)
nostrict-decall¶
graph():
%x : [num_users=1] = placeholder[target=x]
%as_strided : [num_users=1] = call_function[target=torch.ops.aten.as_strided.default](args = (%x, [2, 2, 8, 4], [128, 8, 16, 1]), kwargs = {})
return (as_strided,)
jit¶
graph():
%x : [num_users=1] = placeholder[target=x]
%as_strided : [num_users=1] = call_function[target=torch.ops.aten.as_strided.default](args = (%x, [2, 2, 8, 4], [128, 8, 16, 1]), kwargs = {})
return (as_strided,)
jit-decall¶
graph():
%x : [num_users=1] = placeholder[target=x]
%as_strided : [num_users=1] = call_function[target=torch.ops.aten.as_strided.default](args = (%x, [2, 2, 8, 4], [128, 8, 16, 1]), kwargs = {})
return (as_strided,)
tracing¶
graph():
%x : [num_users=1] = placeholder[target=x]
%as_strided : [num_users=1] = call_function[target=torch.as_strided](args = (%x, (2, 2, 8, 4), (128, 8, 16, 1)), kwargs = {})
return as_strided
AtenInterpolate¶
forward¶
def forward(self, x):
y = torch.nn.functional.interpolate(
x,
scale_factor=2.0,
mode="bilinear",
recompute_scale_factor=False,
)
return y
strict¶
graph():
%x : [num_users=1] = placeholder[target=x]
%upsample_bilinear2d : [num_users=1] = call_function[target=torch.ops.aten.upsample_bilinear2d.vec](args = (%x, None, False, [2.0, 2.0]), kwargs = {})
return (upsample_bilinear2d,)
strict-decall¶
graph():
%x : [num_users=1] = placeholder[target=x]
%_to_copy : [num_users=4] = call_function[target=torch.ops.aten._to_copy.default](args = (%x,), kwargs = {dtype: torch.float32})
%arange : [num_users=2] = call_function[target=torch.ops.aten.arange.start_step](args = (0, 6), kwargs = {layout: torch.strided, device: cpu, pin_memory: False})
%_to_copy_1 : [num_users=1] = call_function[target=torch.ops.aten._to_copy.default](args = (%arange,), kwargs = {dtype: torch.float32})
%_assert_tensor_metadata : [num_users=0] = call_function[target=torch.ops.aten._assert_tensor_metadata.default](args = (%arange, None, None, torch.int64), kwargs = {})
%add_5 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%_to_copy_1, 0.5), kwargs = {})
%mul_2 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%add_5, 0.5), kwargs = {})
%sub_1 : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%mul_2, 0.5), kwargs = {})
%clamp : [num_users=1] = call_function[target=torch.ops.aten.clamp.default](args = (%sub_1, 0.0), kwargs = {})
%view : [num_users=3] = call_function[target=torch.ops.aten.view.default](args = (%clamp, [6, 1]), kwargs = {})
%_to_copy_2 : [num_users=4] = call_function[target=torch.ops.aten._to_copy.default](args = (%view,), kwargs = {dtype: torch.int64})
%_assert_tensor_metadata_1 : [num_users=0] = call_function[target=torch.ops.aten._assert_tensor_metadata.default](args = (%view, None, None, torch.float32), kwargs = {})
%add_6 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%_to_copy_2, 1), kwargs = {})
%clamp_1 : [num_users=2] = call_function[target=torch.ops.aten.clamp.default](args = (%add_6, None, 2), kwargs = {})
%arange_1 : [num_users=2] = call_function[target=torch.ops.aten.arange.start_step](args = (0, 8), kwargs = {layout: torch.strided, device: cpu, pin_memory: False})
%_to_copy_3 : [num_users=1] = call_function[target=torch.ops.aten._to_copy.default](args = (%arange_1,), kwargs = {dtype: torch.float32})
%_assert_tensor_metadata_2 : [num_users=0] = call_function[target=torch.ops.aten._assert_tensor_metadata.default](args = (%arange_1, None, None, torch.int64), kwargs = {})
%add_7 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%_to_copy_3, 0.5), kwargs = {})
%mul_3 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%add_7, 0.5), kwargs = {})
%sub_2 : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%mul_3, 0.5), kwargs = {})
%clamp_2 : [num_users=1] = call_function[target=torch.ops.aten.clamp.default](args = (%sub_2, 0.0), kwargs = {})
%view_1 : [num_users=3] = call_function[target=torch.ops.aten.view.default](args = (%clamp_2, [8]), kwargs = {})
%_to_copy_4 : [num_users=4] = call_function[target=torch.ops.aten._to_copy.default](args = (%view_1,), kwargs = {dtype: torch.int64})
%_assert_tensor_metadata_3 : [num_users=0] = call_function[target=torch.ops.aten._assert_tensor_metadata.default](args = (%view_1, None, None, torch.float32), kwargs = {})
%add_8 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%_to_copy_4, 1), kwargs = {})
%clamp_3 : [num_users=2] = call_function[target=torch.ops.aten.clamp.default](args = (%add_8, None, 3), kwargs = {})
%index : [num_users=2] = call_function[target=torch.ops.aten.index.Tensor](args = (%_to_copy, [None, None, %_to_copy_2, %_to_copy_4]), kwargs = {})
%index_1 : [num_users=1] = call_function[target=torch.ops.aten.index.Tensor](args = (%_to_copy, [None, None, %_to_copy_2, %clamp_3]), kwargs = {})
%index_2 : [num_users=2] = call_function[target=torch.ops.aten.index.Tensor](args = (%_to_copy, [None, None, %clamp_1, %_to_copy_4]), kwargs = {})
%index_3 : [num_users=1] = call_function[target=torch.ops.aten.index.Tensor](args = (%_to_copy, [None, None, %clamp_1, %clamp_3]), kwargs = {})
%sub_7 : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%view_1, %_to_copy_4), kwargs = {})
%clamp_4 : [num_users=1] = call_function[target=torch.ops.aten.clamp.default](args = (%sub_7, 0.0, 1.0), kwargs = {})
%_to_copy_5 : [num_users=2] = call_function[target=torch.ops.aten._to_copy.default](args = (%clamp_4,), kwargs = {dtype: torch.float32})
%sub_8 : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%index_1, %index), kwargs = {})
%mul_14 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%sub_8, %_to_copy_5), kwargs = {})
%add_39 : [num_users=2] = call_function[target=torch.ops.aten.add.Tensor](args = (%index, %mul_14), kwargs = {})
%sub_12 : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%index_3, %index_2), kwargs = {})
%mul_21 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%sub_12, %_to_copy_5), kwargs = {})
%add_55 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%index_2, %mul_21), kwargs = {})
%sub_16 : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%view, %_to_copy_2), kwargs = {})
%clamp_5 : [num_users=1] = call_function[target=torch.ops.aten.clamp.default](args = (%sub_16, 0.0, 1.0), kwargs = {})
%_to_copy_6 : [num_users=1] = call_function[target=torch.ops.aten._to_copy.default](args = (%clamp_5,), kwargs = {dtype: torch.float32})
%sub_17 : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%add_55, %add_39), kwargs = {})
%mul_28 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%sub_17, %_to_copy_6), kwargs = {})
%add_71 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%add_39, %mul_28), kwargs = {})
%_to_copy_7 : [num_users=1] = call_function[target=torch.ops.aten._to_copy.default](args = (%add_71,), kwargs = {dtype: torch.float32})
return (_to_copy_7,)
nostrict¶
graph():
%x : [num_users=1] = placeholder[target=x]
%upsample_bilinear2d : [num_users=1] = call_function[target=torch.ops.aten.upsample_bilinear2d.vec](args = (%x, None, False, [2.0, 2.0]), kwargs = {})
return (upsample_bilinear2d,)
nostrict-decall¶
graph():
%x : [num_users=1] = placeholder[target=x]
%_to_copy : [num_users=4] = call_function[target=torch.ops.aten._to_copy.default](args = (%x,), kwargs = {dtype: torch.float32})
%arange : [num_users=2] = call_function[target=torch.ops.aten.arange.start_step](args = (0, 6), kwargs = {layout: torch.strided, device: cpu, pin_memory: False})
%_to_copy_1 : [num_users=1] = call_function[target=torch.ops.aten._to_copy.default](args = (%arange,), kwargs = {dtype: torch.float32})
%_assert_tensor_metadata : [num_users=0] = call_function[target=torch.ops.aten._assert_tensor_metadata.default](args = (%arange, None, None, torch.int64), kwargs = {})
%add_5 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%_to_copy_1, 0.5), kwargs = {})
%mul_2 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%add_5, 0.5), kwargs = {})
%sub_1 : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%mul_2, 0.5), kwargs = {})
%clamp : [num_users=1] = call_function[target=torch.ops.aten.clamp.default](args = (%sub_1, 0.0), kwargs = {})
%view : [num_users=3] = call_function[target=torch.ops.aten.view.default](args = (%clamp, [6, 1]), kwargs = {})
%_to_copy_2 : [num_users=4] = call_function[target=torch.ops.aten._to_copy.default](args = (%view,), kwargs = {dtype: torch.int64})
%_assert_tensor_metadata_1 : [num_users=0] = call_function[target=torch.ops.aten._assert_tensor_metadata.default](args = (%view, None, None, torch.float32), kwargs = {})
%add_6 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%_to_copy_2, 1), kwargs = {})
%clamp_1 : [num_users=2] = call_function[target=torch.ops.aten.clamp.default](args = (%add_6, None, 2), kwargs = {})
%arange_1 : [num_users=2] = call_function[target=torch.ops.aten.arange.start_step](args = (0, 8), kwargs = {layout: torch.strided, device: cpu, pin_memory: False})
%_to_copy_3 : [num_users=1] = call_function[target=torch.ops.aten._to_copy.default](args = (%arange_1,), kwargs = {dtype: torch.float32})
%_assert_tensor_metadata_2 : [num_users=0] = call_function[target=torch.ops.aten._assert_tensor_metadata.default](args = (%arange_1, None, None, torch.int64), kwargs = {})
%add_7 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%_to_copy_3, 0.5), kwargs = {})
%mul_3 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%add_7, 0.5), kwargs = {})
%sub_2 : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%mul_3, 0.5), kwargs = {})
%clamp_2 : [num_users=1] = call_function[target=torch.ops.aten.clamp.default](args = (%sub_2, 0.0), kwargs = {})
%view_1 : [num_users=3] = call_function[target=torch.ops.aten.view.default](args = (%clamp_2, [8]), kwargs = {})
%_to_copy_4 : [num_users=4] = call_function[target=torch.ops.aten._to_copy.default](args = (%view_1,), kwargs = {dtype: torch.int64})
%_assert_tensor_metadata_3 : [num_users=0] = call_function[target=torch.ops.aten._assert_tensor_metadata.default](args = (%view_1, None, None, torch.float32), kwargs = {})
%add_8 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%_to_copy_4, 1), kwargs = {})
%clamp_3 : [num_users=2] = call_function[target=torch.ops.aten.clamp.default](args = (%add_8, None, 3), kwargs = {})
%index : [num_users=2] = call_function[target=torch.ops.aten.index.Tensor](args = (%_to_copy, [None, None, %_to_copy_2, %_to_copy_4]), kwargs = {})
%index_1 : [num_users=1] = call_function[target=torch.ops.aten.index.Tensor](args = (%_to_copy, [None, None, %_to_copy_2, %clamp_3]), kwargs = {})
%index_2 : [num_users=2] = call_function[target=torch.ops.aten.index.Tensor](args = (%_to_copy, [None, None, %clamp_1, %_to_copy_4]), kwargs = {})
%index_3 : [num_users=1] = call_function[target=torch.ops.aten.index.Tensor](args = (%_to_copy, [None, None, %clamp_1, %clamp_3]), kwargs = {})
%sub_7 : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%view_1, %_to_copy_4), kwargs = {})
%clamp_4 : [num_users=1] = call_function[target=torch.ops.aten.clamp.default](args = (%sub_7, 0.0, 1.0), kwargs = {})
%_to_copy_5 : [num_users=2] = call_function[target=torch.ops.aten._to_copy.default](args = (%clamp_4,), kwargs = {dtype: torch.float32})
%sub_8 : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%index_1, %index), kwargs = {})
%mul_14 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%sub_8, %_to_copy_5), kwargs = {})
%add_39 : [num_users=2] = call_function[target=torch.ops.aten.add.Tensor](args = (%index, %mul_14), kwargs = {})
%sub_12 : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%index_3, %index_2), kwargs = {})
%mul_21 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%sub_12, %_to_copy_5), kwargs = {})
%add_55 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%index_2, %mul_21), kwargs = {})
%sub_16 : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%view, %_to_copy_2), kwargs = {})
%clamp_5 : [num_users=1] = call_function[target=torch.ops.aten.clamp.default](args = (%sub_16, 0.0, 1.0), kwargs = {})
%_to_copy_6 : [num_users=1] = call_function[target=torch.ops.aten._to_copy.default](args = (%clamp_5,), kwargs = {dtype: torch.float32})
%sub_17 : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%add_55, %add_39), kwargs = {})
%mul_28 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%sub_17, %_to_copy_6), kwargs = {})
%add_71 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%add_39, %mul_28), kwargs = {})
%_to_copy_7 : [num_users=1] = call_function[target=torch.ops.aten._to_copy.default](args = (%add_71,), kwargs = {dtype: torch.float32})
return (_to_copy_7,)
jit¶
graph():
%x : [num_users=1] = placeholder[target=x]
%upsample_bilinear2d : [num_users=1] = call_function[target=torch.ops.aten.upsample_bilinear2d.vec](args = (%x, None, False, [2.0, 2.0]), kwargs = {})
return (upsample_bilinear2d,)
jit-decall¶
graph():
%x : [num_users=3] = placeholder[target=x]
%sym_size_int_8 : [num_users=2] = call_function[target=torch.ops.aten.sym_size.int](args = (%x, 2), kwargs = {})
%sym_size_int_9 : [num_users=2] = call_function[target=torch.ops.aten.sym_size.int](args = (%x, 3), kwargs = {})
%mul : [num_users=2] = call_function[target=operator.mul](args = (%sym_size_int_8, 2), kwargs = {})
%mul_1 : [num_users=2] = call_function[target=operator.mul](args = (%sym_size_int_9, 2), kwargs = {})
%_to_copy : [num_users=4] = call_function[target=torch.ops.aten._to_copy.default](args = (%x,), kwargs = {dtype: torch.float32})
%arange : [num_users=2] = call_function[target=torch.ops.aten.arange.start_step](args = (0, %mul), kwargs = {layout: torch.strided, device: cpu, pin_memory: False})
%_to_copy_1 : [num_users=1] = call_function[target=torch.ops.aten._to_copy.default](args = (%arange,), kwargs = {dtype: torch.float32})
%_assert_tensor_metadata : [num_users=0] = call_function[target=torch.ops.aten._assert_tensor_metadata.default](args = (%arange, None, None, torch.int64), kwargs = {})
%add_9 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%_to_copy_1, 0.5), kwargs = {})
%mul_9 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%add_9, 0.5), kwargs = {})
%sub_8 : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%mul_9, 0.5), kwargs = {})
%clamp : [num_users=1] = call_function[target=torch.ops.aten.clamp.default](args = (%sub_8, 0.0), kwargs = {})
%view : [num_users=3] = call_function[target=torch.ops.aten.view.default](args = (%clamp, [%mul, 1]), kwargs = {})
%_to_copy_2 : [num_users=4] = call_function[target=torch.ops.aten._to_copy.default](args = (%view,), kwargs = {dtype: torch.int64})
%_assert_tensor_metadata_1 : [num_users=0] = call_function[target=torch.ops.aten._assert_tensor_metadata.default](args = (%view, None, None, torch.float32), kwargs = {})
%add_24 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%_to_copy_2, 1), kwargs = {})
%sub_14 : [num_users=1] = call_function[target=operator.sub](args = (%sym_size_int_8, 1), kwargs = {})
%clamp_1 : [num_users=2] = call_function[target=torch.ops.aten.clamp.default](args = (%add_24, None, %sub_14), kwargs = {})
%arange_1 : [num_users=2] = call_function[target=torch.ops.aten.arange.start_step](args = (0, %mul_1), kwargs = {layout: torch.strided, device: cpu, pin_memory: False})
%_to_copy_3 : [num_users=1] = call_function[target=torch.ops.aten._to_copy.default](args = (%arange_1,), kwargs = {dtype: torch.float32})
%_assert_tensor_metadata_2 : [num_users=0] = call_function[target=torch.ops.aten._assert_tensor_metadata.default](args = (%arange_1, None, None, torch.int64), kwargs = {})
%add_35 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%_to_copy_3, 0.5), kwargs = {})
%mul_20 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%add_35, 0.5), kwargs = {})
%sub_20 : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%mul_20, 0.5), kwargs = {})
%clamp_2 : [num_users=1] = call_function[target=torch.ops.aten.clamp.default](args = (%sub_20, 0.0), kwargs = {})
%view_1 : [num_users=3] = call_function[target=torch.ops.aten.view.default](args = (%clamp_2, [%mul_1]), kwargs = {})
%_to_copy_4 : [num_users=4] = call_function[target=torch.ops.aten._to_copy.default](args = (%view_1,), kwargs = {dtype: torch.int64})
%_assert_tensor_metadata_3 : [num_users=0] = call_function[target=torch.ops.aten._assert_tensor_metadata.default](args = (%view_1, None, None, torch.float32), kwargs = {})
%add_48 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%_to_copy_4, 1), kwargs = {})
%sub_26 : [num_users=1] = call_function[target=operator.sub](args = (%sym_size_int_9, 1), kwargs = {})
%clamp_3 : [num_users=2] = call_function[target=torch.ops.aten.clamp.default](args = (%add_48, None, %sub_26), kwargs = {})
%index : [num_users=2] = call_function[target=torch.ops.aten.index.Tensor](args = (%_to_copy, [None, None, %_to_copy_2, %_to_copy_4]), kwargs = {})
%index_1 : [num_users=1] = call_function[target=torch.ops.aten.index.Tensor](args = (%_to_copy, [None, None, %_to_copy_2, %clamp_3]), kwargs = {})
%index_2 : [num_users=2] = call_function[target=torch.ops.aten.index.Tensor](args = (%_to_copy, [None, None, %clamp_1, %_to_copy_4]), kwargs = {})
%index_3 : [num_users=1] = call_function[target=torch.ops.aten.index.Tensor](args = (%_to_copy, [None, None, %clamp_1, %clamp_3]), kwargs = {})
%sub_44 : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%view_1, %_to_copy_4), kwargs = {})
%clamp_4 : [num_users=1] = call_function[target=torch.ops.aten.clamp.default](args = (%sub_44, 0.0, 1.0), kwargs = {})
%_to_copy_5 : [num_users=2] = call_function[target=torch.ops.aten._to_copy.default](args = (%clamp_4,), kwargs = {dtype: torch.float32})
%sub_48 : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%index_1, %index), kwargs = {})
%mul_51 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%sub_48, %_to_copy_5), kwargs = {})
%add_89 : [num_users=2] = call_function[target=torch.ops.aten.add.Tensor](args = (%index, %mul_51), kwargs = {})
%sub_61 : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%index_3, %index_2), kwargs = {})
%mul_64 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%sub_61, %_to_copy_5), kwargs = {})
%add_105 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%index_2, %mul_64), kwargs = {})
%sub_74 : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%view, %_to_copy_2), kwargs = {})
%clamp_5 : [num_users=1] = call_function[target=torch.ops.aten.clamp.default](args = (%sub_74, 0.0, 1.0), kwargs = {})
%_to_copy_6 : [num_users=1] = call_function[target=torch.ops.aten._to_copy.default](args = (%clamp_5,), kwargs = {dtype: torch.float32})
%sub_78 : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%add_105, %add_89), kwargs = {})
%mul_80 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%sub_78, %_to_copy_6), kwargs = {})
%add_130 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%add_89, %mul_80), kwargs = {})
%_to_copy_7 : [num_users=1] = call_function[target=torch.ops.aten._to_copy.default](args = (%add_130,), kwargs = {dtype: torch.float32})
return (_to_copy_7,)
tracing¶
graph():
%x : [num_users=1] = placeholder[target=x]
%interpolate : [num_users=1] = call_function[target=torch.nn.functional.interpolate](args = (%x,), kwargs = {size: None, scale_factor: 2.0, mode: bilinear, align_corners: None, recompute_scale_factor: False, antialias: False})
return interpolate
AtenNonZero¶
forward¶
def forward(self, x):
y = torch.nonzero(x)
return y
strict¶
graph():
%x : [num_users=1] = placeholder[target=x]
%nonzero : [num_users=2] = call_function[target=torch.ops.aten.nonzero.default](args = (%x,), kwargs = {})
%sym_size_int_2 : [num_users=2] = call_function[target=torch.ops.aten.sym_size.int](args = (%nonzero, 0), kwargs = {})
%sym_constrain_range_for_size_default : [num_users=0] = call_function[target=torch.ops.aten.sym_constrain_range_for_size.default](args = (%sym_size_int_2,), kwargs = {})
%ge_1 : [num_users=1] = call_function[target=operator.ge](args = (%sym_size_int_2, 0), kwargs = {})
%_assert_scalar_default : [num_users=0] = call_function[target=torch.ops.aten._assert_scalar.default](args = (%ge_1, Runtime assertion failed for expression u0 >= 0 on node 'ge_1'), kwargs = {})
return (nonzero,)
strict-decall¶
graph():
%x : [num_users=1] = placeholder[target=x]
%nonzero : [num_users=2] = call_function[target=torch.ops.aten.nonzero.default](args = (%x,), kwargs = {})
%sym_size_int_2 : [num_users=2] = call_function[target=torch.ops.aten.sym_size.int](args = (%nonzero, 0), kwargs = {})
%sym_constrain_range_for_size_default : [num_users=0] = call_function[target=torch.ops.aten.sym_constrain_range_for_size.default](args = (%sym_size_int_2,), kwargs = {})
%ge_1 : [num_users=1] = call_function[target=operator.ge](args = (%sym_size_int_2, 0), kwargs = {})
%_assert_scalar_default : [num_users=0] = call_function[target=torch.ops.aten._assert_scalar.default](args = (%ge_1, Runtime assertion failed for expression u0 >= 0 on node 'ge_1'), kwargs = {})
return (nonzero,)
nostrict¶
graph():
%x : [num_users=1] = placeholder[target=x]
%nonzero : [num_users=2] = call_function[target=torch.ops.aten.nonzero.default](args = (%x,), kwargs = {})
%sym_size_int_1 : [num_users=2] = call_function[target=torch.ops.aten.sym_size.int](args = (%nonzero, 0), kwargs = {})
%sym_constrain_range_for_size_default : [num_users=0] = call_function[target=torch.ops.aten.sym_constrain_range_for_size.default](args = (%sym_size_int_1,), kwargs = {})
%ge : [num_users=1] = call_function[target=operator.ge](args = (%sym_size_int_1, 0), kwargs = {})
%_assert_scalar_default : [num_users=0] = call_function[target=torch.ops.aten._assert_scalar.default](args = (%ge, Runtime assertion failed for expression u0 >= 0 on node 'ge'), kwargs = {})
return (nonzero,)
nostrict-decall¶
graph():
%x : [num_users=1] = placeholder[target=x]
%nonzero : [num_users=2] = call_function[target=torch.ops.aten.nonzero.default](args = (%x,), kwargs = {})
%sym_size_int_2 : [num_users=2] = call_function[target=torch.ops.aten.sym_size.int](args = (%nonzero, 0), kwargs = {})
%sym_constrain_range_for_size_default : [num_users=0] = call_function[target=torch.ops.aten.sym_constrain_range_for_size.default](args = (%sym_size_int_2,), kwargs = {})
%ge_1 : [num_users=1] = call_function[target=operator.ge](args = (%sym_size_int_2, 0), kwargs = {})
%_assert_scalar_default : [num_users=0] = call_function[target=torch.ops.aten._assert_scalar.default](args = (%ge_1, Runtime assertion failed for expression u0 >= 0 on node 'ge_1'), kwargs = {})
return (nonzero,)
jit¶
graph():
%x : [num_users=1] = placeholder[target=x]
%nonzero : [num_users=2] = call_function[target=torch.ops.aten.nonzero.default](args = (%x,), kwargs = {})
%sym_size_int_2 : [num_users=2] = call_function[target=torch.ops.aten.sym_size.int](args = (%nonzero, 0), kwargs = {})
%sym_constrain_range_for_size_default : [num_users=0] = call_function[target=torch.ops.aten.sym_constrain_range_for_size.default](args = (%sym_size_int_2,), kwargs = {})
%ge : [num_users=1] = call_function[target=operator.ge](args = (%sym_size_int_2, 0), kwargs = {})
%_assert_scalar_default : [num_users=0] = call_function[target=torch.ops.aten._assert_scalar.default](args = (%ge, Runtime assertion failed for expression u0 >= 0 on node 'ge'), kwargs = {})
return (nonzero,)
jit-decall¶
graph():
%x : [num_users=1] = placeholder[target=x]
%nonzero : [num_users=2] = call_function[target=torch.ops.aten.nonzero.default](args = (%x,), kwargs = {})
%sym_size_int_3 : [num_users=2] = call_function[target=torch.ops.aten.sym_size.int](args = (%nonzero, 0), kwargs = {})
%sym_constrain_range_for_size_default : [num_users=0] = call_function[target=torch.ops.aten.sym_constrain_range_for_size.default](args = (%sym_size_int_3,), kwargs = {})
%ge_1 : [num_users=1] = call_function[target=operator.ge](args = (%sym_size_int_3, 0), kwargs = {})
%_assert_scalar_default : [num_users=0] = call_function[target=torch.ops.aten._assert_scalar.default](args = (%ge_1, Runtime assertion failed for expression u0 >= 0 on node 'ge_1'), kwargs = {})
return (nonzero,)
tracing¶
graph():
%x : [num_users=1] = placeholder[target=x]
%nonzero : [num_users=1] = call_function[target=torch.nonzero](args = (%x,), kwargs = {})
return nonzero
AtenNonZeroTuple¶
forward¶
def forward(self, x):
y = torch.nonzero(x, as_tuple=True)
return y[0], y[1]
strict¶
graph():
%x : [num_users=1] = placeholder[target=x]
%nonzero_numpy : [num_users=2] = call_function[target=torch.ops.aten.nonzero_numpy.default](args = (%x,), kwargs = {})
%getitem_2 : [num_users=2] = call_function[target=operator.getitem](args = (%nonzero_numpy, 0), kwargs = {})
%sym_size_int_2 : [num_users=2] = call_function[target=torch.ops.aten.sym_size.int](args = (%getitem_2, 0), kwargs = {})
%sym_constrain_range_for_size_default : [num_users=0] = call_function[target=torch.ops.aten.sym_constrain_range_for_size.default](args = (%sym_size_int_2,), kwargs = {})
%ge_1 : [num_users=1] = call_function[target=operator.ge](args = (%sym_size_int_2, 0), kwargs = {})
%_assert_scalar_default : [num_users=0] = call_function[target=torch.ops.aten._assert_scalar.default](args = (%ge_1, Runtime assertion failed for expression u0 >= 0 on node 'ge_1'), kwargs = {})
%getitem_1 : [num_users=1] = call_function[target=operator.getitem](args = (%nonzero_numpy, 1), kwargs = {})
return (getitem_2, getitem_1)
strict-decall¶
graph():
%x : [num_users=1] = placeholder[target=x]
%nonzero : [num_users=3] = call_function[target=torch.ops.aten.nonzero.default](args = (%x,), kwargs = {})
%sym_size_int_2 : [num_users=2] = call_function[target=torch.ops.aten.sym_size.int](args = (%nonzero, 0), kwargs = {})
%sym_constrain_range_for_size_default : [num_users=0] = call_function[target=torch.ops.aten.sym_constrain_range_for_size.default](args = (%sym_size_int_2,), kwargs = {})
%ge_2 : [num_users=1] = call_function[target=operator.ge](args = (%sym_size_int_2, 0), kwargs = {})
%_assert_scalar_default : [num_users=0] = call_function[target=torch.ops.aten._assert_scalar.default](args = (%ge_2, Runtime assertion failed for expression u0 >= 0 on node 'ge_2'), kwargs = {})
%slice_1 : [num_users=1] = call_function[target=torch.ops.aten.slice.Tensor](args = (%nonzero, 1, 0, 1), kwargs = {})
%slice_2 : [num_users=1] = call_function[target=torch.ops.aten.slice.Tensor](args = (%nonzero, 1, 1, 2), kwargs = {})
%squeeze : [num_users=1] = call_function[target=torch.ops.aten.squeeze.dims](args = (%slice_1, [1]), kwargs = {})
%squeeze_1 : [num_users=1] = call_function[target=torch.ops.aten.squeeze.dims](args = (%slice_2, [1]), kwargs = {})
return (squeeze, squeeze_1)
nostrict¶
graph():
%x : [num_users=1] = placeholder[target=x]
%nonzero_numpy : [num_users=2] = call_function[target=torch.ops.aten.nonzero_numpy.default](args = (%x,), kwargs = {})
%getitem_2 : [num_users=2] = call_function[target=operator.getitem](args = (%nonzero_numpy, 0), kwargs = {})
%sym_size_int_1 : [num_users=2] = call_function[target=torch.ops.aten.sym_size.int](args = (%getitem_2, 0), kwargs = {})
%sym_constrain_range_for_size_default : [num_users=0] = call_function[target=torch.ops.aten.sym_constrain_range_for_size.default](args = (%sym_size_int_1,), kwargs = {})
%ge : [num_users=1] = call_function[target=operator.ge](args = (%sym_size_int_1, 0), kwargs = {})
%_assert_scalar_default : [num_users=0] = call_function[target=torch.ops.aten._assert_scalar.default](args = (%ge, Runtime assertion failed for expression u0 >= 0 on node 'ge'), kwargs = {})
%getitem_1 : [num_users=1] = call_function[target=operator.getitem](args = (%nonzero_numpy, 1), kwargs = {})
return (getitem_2, getitem_1)
nostrict-decall¶
graph():
%x : [num_users=1] = placeholder[target=x]
%nonzero : [num_users=3] = call_function[target=torch.ops.aten.nonzero.default](args = (%x,), kwargs = {})
%sym_size_int_2 : [num_users=2] = call_function[target=torch.ops.aten.sym_size.int](args = (%nonzero, 0), kwargs = {})
%sym_constrain_range_for_size_default : [num_users=0] = call_function[target=torch.ops.aten.sym_constrain_range_for_size.default](args = (%sym_size_int_2,), kwargs = {})
%ge_2 : [num_users=1] = call_function[target=operator.ge](args = (%sym_size_int_2, 0), kwargs = {})
%_assert_scalar_default : [num_users=0] = call_function[target=torch.ops.aten._assert_scalar.default](args = (%ge_2, Runtime assertion failed for expression u0 >= 0 on node 'ge_2'), kwargs = {})
%slice_1 : [num_users=1] = call_function[target=torch.ops.aten.slice.Tensor](args = (%nonzero, 1, 0, 1), kwargs = {})
%slice_2 : [num_users=1] = call_function[target=torch.ops.aten.slice.Tensor](args = (%nonzero, 1, 1, 2), kwargs = {})
%squeeze : [num_users=1] = call_function[target=torch.ops.aten.squeeze.dims](args = (%slice_1, [1]), kwargs = {})
%squeeze_1 : [num_users=1] = call_function[target=torch.ops.aten.squeeze.dims](args = (%slice_2, [1]), kwargs = {})
return (squeeze, squeeze_1)
jit¶
graph():
%x : [num_users=1] = placeholder[target=x]
%nonzero_numpy : [num_users=2] = call_function[target=torch.ops.aten.nonzero_numpy.default](args = (%x,), kwargs = {})
%getitem_2 : [num_users=2] = call_function[target=operator.getitem](args = (%nonzero_numpy, 0), kwargs = {})
%sym_size_int_2 : [num_users=2] = call_function[target=torch.ops.aten.sym_size.int](args = (%getitem_2, 0), kwargs = {})
%sym_constrain_range_for_size_default : [num_users=0] = call_function[target=torch.ops.aten.sym_constrain_range_for_size.default](args = (%sym_size_int_2,), kwargs = {})
%ge : [num_users=1] = call_function[target=operator.ge](args = (%sym_size_int_2, 0), kwargs = {})
%_assert_scalar_default : [num_users=0] = call_function[target=torch.ops.aten._assert_scalar.default](args = (%ge, Runtime assertion failed for expression u0 >= 0 on node 'ge'), kwargs = {})
%getitem_1 : [num_users=1] = call_function[target=operator.getitem](args = (%nonzero_numpy, 1), kwargs = {})
return (getitem_2, getitem_1)
jit-decall¶
graph():
%x : [num_users=1] = placeholder[target=x]
%nonzero : [num_users=3] = call_function[target=torch.ops.aten.nonzero.default](args = (%x,), kwargs = {})
%sym_size_int_3 : [num_users=2] = call_function[target=torch.ops.aten.sym_size.int](args = (%nonzero, 0), kwargs = {})
%sym_constrain_range_for_size_default : [num_users=0] = call_function[target=torch.ops.aten.sym_constrain_range_for_size.default](args = (%sym_size_int_3,), kwargs = {})
%ge_2 : [num_users=1] = call_function[target=operator.ge](args = (%sym_size_int_3, 0), kwargs = {})
%_assert_scalar_default : [num_users=0] = call_function[target=torch.ops.aten._assert_scalar.default](args = (%ge_2, Runtime assertion failed for expression u0 >= 0 on node 'ge_2'), kwargs = {})
%slice_1 : [num_users=1] = call_function[target=torch.ops.aten.slice.Tensor](args = (%nonzero, 1, 0, 1), kwargs = {})
%slice_2 : [num_users=1] = call_function[target=torch.ops.aten.slice.Tensor](args = (%nonzero, 1, 1, 2), kwargs = {})
%squeeze : [num_users=1] = call_function[target=torch.ops.aten.squeeze.dims](args = (%slice_1, [1]), kwargs = {})
%squeeze_1 : [num_users=1] = call_function[target=torch.ops.aten.squeeze.dims](args = (%slice_2, [1]), kwargs = {})
return (squeeze, squeeze_1)
tracing¶
graph():
%x : [num_users=1] = placeholder[target=x]
%nonzero : [num_users=2] = call_function[target=torch.nonzero](args = (%x,), kwargs = {as_tuple: True})
%getitem : [num_users=1] = call_function[target=operator.getitem](args = (%nonzero, 0), kwargs = {})
%getitem_1 : [num_users=1] = call_function[target=operator.getitem](args = (%nonzero, 1), kwargs = {})
return (getitem, getitem_1)
AtenRollPos¶
forward¶
def forward(self, x):
return torch.roll(x, 1, -1)
strict¶
graph():
%x : [num_users=1] = placeholder[target=x]
%roll : [num_users=1] = call_function[target=torch.ops.aten.roll.default](args = (%x, [1], [-1]), kwargs = {})
return (roll,)
strict-decall¶
graph():
%x : [num_users=1] = placeholder[target=x]
%arange : [num_users=1] = call_function[target=torch.ops.aten.arange.start_step](args = (0, 4), kwargs = {layout: torch.strided, device: cpu, pin_memory: False})
%add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%arange, 3), kwargs = {})
%fmod : [num_users=1] = call_function[target=torch.ops.aten.fmod.Scalar](args = (%add, 4), kwargs = {})
%index_select : [num_users=1] = call_function[target=torch.ops.aten.index_select.default](args = (%x, 2, %fmod), kwargs = {})
return (index_select,)
nostrict¶
graph():
%x : [num_users=1] = placeholder[target=x]
%roll : [num_users=1] = call_function[target=torch.ops.aten.roll.default](args = (%x, [1], [-1]), kwargs = {})
return (roll,)
nostrict-decall¶
graph():
%x : [num_users=1] = placeholder[target=x]
%arange : [num_users=1] = call_function[target=torch.ops.aten.arange.start_step](args = (0, 4), kwargs = {layout: torch.strided, device: cpu, pin_memory: False})
%add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%arange, 3), kwargs = {})
%fmod : [num_users=1] = call_function[target=torch.ops.aten.fmod.Scalar](args = (%add, 4), kwargs = {})
%index_select : [num_users=1] = call_function[target=torch.ops.aten.index_select.default](args = (%x, 2, %fmod), kwargs = {})
return (index_select,)
jit¶
graph():
%x : [num_users=1] = placeholder[target=x]
%roll : [num_users=1] = call_function[target=torch.ops.aten.roll.default](args = (%x, [1], [-1]), kwargs = {})
return (roll,)
jit-decall¶
graph():
%x : [num_users=2] = placeholder[target=x]
%sym_size_int_5 : [num_users=4] = call_function[target=torch.ops.aten.sym_size.int](args = (%x, 2), kwargs = {})
%sub : [num_users=1] = call_function[target=operator.sub](args = (%sym_size_int_5, 1), kwargs = {})
%mod : [num_users=1] = call_function[target=operator.mod](args = (%sub, %sym_size_int_5), kwargs = {})
%arange : [num_users=1] = call_function[target=torch.ops.aten.arange.start_step](args = (0, %sym_size_int_5), kwargs = {layout: torch.strided, device: cpu, pin_memory: False})
%add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%arange, %mod), kwargs = {})
%fmod : [num_users=1] = call_function[target=torch.ops.aten.fmod.Scalar](args = (%add, %sym_size_int_5), kwargs = {})
%index_select : [num_users=1] = call_function[target=torch.ops.aten.index_select.default](args = (%x, 2, %fmod), kwargs = {})
return (index_select,)
tracing¶
graph():
%x : [num_users=1] = placeholder[target=x]
%roll : [num_users=1] = call_function[target=torch.roll](args = (%x, 1, -1), kwargs = {})
return roll
AtenRollRelu¶
forward¶
def forward(self, x):
return torch.relu(torch.roll(x, -1, -1))
strict¶
graph():
%x : [num_users=1] = placeholder[target=x]
%roll : [num_users=1] = call_function[target=torch.ops.aten.roll.default](args = (%x, [-1], [-1]), kwargs = {})
%relu : [num_users=1] = call_function[target=torch.ops.aten.relu.default](args = (%roll,), kwargs = {})
return (relu,)
strict-decall¶
graph():
%x : [num_users=1] = placeholder[target=x]
%arange : [num_users=1] = call_function[target=torch.ops.aten.arange.start_step](args = (0, 4), kwargs = {layout: torch.strided, device: cpu, pin_memory: False})
%add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%arange, 1), kwargs = {})
%fmod : [num_users=1] = call_function[target=torch.ops.aten.fmod.Scalar](args = (%add, 4), kwargs = {})
%index_select : [num_users=1] = call_function[target=torch.ops.aten.index_select.default](args = (%x, 2, %fmod), kwargs = {})
%relu : [num_users=1] = call_function[target=torch.ops.aten.relu.default](args = (%index_select,), kwargs = {})
return (relu,)
nostrict¶
graph():
%x : [num_users=1] = placeholder[target=x]
%roll : [num_users=1] = call_function[target=torch.ops.aten.roll.default](args = (%x, [-1], [-1]), kwargs = {})
%relu : [num_users=1] = call_function[target=torch.ops.aten.relu.default](args = (%roll,), kwargs = {})
return (relu,)
nostrict-decall¶
graph():
%x : [num_users=1] = placeholder[target=x]
%arange : [num_users=1] = call_function[target=torch.ops.aten.arange.start_step](args = (0, 4), kwargs = {layout: torch.strided, device: cpu, pin_memory: False})
%add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%arange, 1), kwargs = {})
%fmod : [num_users=1] = call_function[target=torch.ops.aten.fmod.Scalar](args = (%add, 4), kwargs = {})
%index_select : [num_users=1] = call_function[target=torch.ops.aten.index_select.default](args = (%x, 2, %fmod), kwargs = {})
%relu : [num_users=1] = call_function[target=torch.ops.aten.relu.default](args = (%index_select,), kwargs = {})
return (relu,)
jit¶
graph():
%x : [num_users=1] = placeholder[target=x]
%roll : [num_users=1] = call_function[target=torch.ops.aten.roll.default](args = (%x, [-1], [-1]), kwargs = {})
%relu : [num_users=1] = call_function[target=torch.ops.aten.relu.default](args = (%roll,), kwargs = {})
return (relu,)
jit-decall¶
graph():
%x : [num_users=2] = placeholder[target=x]
%sym_size_int_5 : [num_users=4] = call_function[target=torch.ops.aten.sym_size.int](args = (%x, 2), kwargs = {})
%sub : [num_users=1] = call_function[target=operator.sub](args = (%sym_size_int_5, -1), kwargs = {})
%mod : [num_users=1] = call_function[target=operator.mod](args = (%sub, %sym_size_int_5), kwargs = {})
%arange : [num_users=1] = call_function[target=torch.ops.aten.arange.start_step](args = (0, %sym_size_int_5), kwargs = {layout: torch.strided, device: cpu, pin_memory: False})
%add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%arange, %mod), kwargs = {})
%fmod : [num_users=1] = call_function[target=torch.ops.aten.fmod.Scalar](args = (%add, %sym_size_int_5), kwargs = {})
%index_select : [num_users=1] = call_function[target=torch.ops.aten.index_select.default](args = (%x, 2, %fmod), kwargs = {})
%relu : [num_users=1] = call_function[target=torch.ops.aten.relu.default](args = (%index_select,), kwargs = {})
return (relu,)
tracing¶
graph():
%x : [num_users=1] = placeholder[target=x]
%roll : [num_users=1] = call_function[target=torch.roll](args = (%x, -1, -1), kwargs = {})
%relu : [num_users=1] = call_function[target=torch.relu](args = (%roll,), kwargs = {})
return relu
BuildInIsInstance¶
forward¶
def forward(self, x, lx: list | torch.Tensor):
if isinstance(lx, list):
t = lx[0] * lx[1].sum(axis=1, keepdim=True)
return torch.sigmoid(self.linear(x)) - self.buff + t
return torch.sigmoid(self.linear(x)) - self.buff + lx
strict¶
graph():
%linear_weight : [num_users=1] = get_attr[target=linear.weight]
%linear_bias : [num_users=1] = get_attr[target=linear.bias]
%buff : [num_users=1] = get_attr[target=buff]
%x : [num_users=1] = placeholder[target=x]
%lx_0 : [num_users=1] = placeholder[target=lx_0]
%lx_1 : [num_users=1] = placeholder[target=lx_1]
%sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%lx_1, [1], True), kwargs = {})
%mul : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%lx_0, %sum_1), kwargs = {})
%linear : [num_users=1] = call_function[target=torch.ops.aten.linear.default](args = (%x, %linear_weight, %linear_bias), kwargs = {})
%sigmoid : [num_users=1] = call_function[target=torch.ops.aten.sigmoid.default](args = (%linear,), kwargs = {})
%sub : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%sigmoid, %buff), kwargs = {})
%add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%sub, %mul), kwargs = {})
return (add,)
strict-decall¶
graph():
%linear_weight : [num_users=1] = get_attr[target=linear.weight]
%linear_bias : [num_users=1] = get_attr[target=linear.bias]
%buff : [num_users=1] = get_attr[target=buff]
%x : [num_users=1] = placeholder[target=x]
%lx_0 : [num_users=1] = placeholder[target=lx_0]
%lx_1 : [num_users=1] = placeholder[target=lx_1]
%sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%lx_1, [1], True), kwargs = {})
%mul_1 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%lx_0, %sum_1), kwargs = {})
%permute : [num_users=1] = call_function[target=torch.ops.aten.permute.default](args = (%linear_weight, [1, 0]), kwargs = {})
%addmm : [num_users=1] = call_function[target=torch.ops.aten.addmm.default](args = (%linear_bias, %x, %permute), kwargs = {})
%sigmoid : [num_users=1] = call_function[target=torch.ops.aten.sigmoid.default](args = (%addmm,), kwargs = {})
%sub_4 : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%sigmoid, %buff), kwargs = {})
%add_15 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%sub_4, %mul_1), kwargs = {})
return (add_15,)
nostrict¶
graph():
%linear_weight : [num_users=1] = get_attr[target=linear.weight]
%linear_bias : [num_users=1] = get_attr[target=linear.bias]
%buff : [num_users=1] = get_attr[target=buff]
%x : [num_users=1] = placeholder[target=x]
%lx_0 : [num_users=1] = placeholder[target=lx_0]
%lx_1 : [num_users=1] = placeholder[target=lx_1]
%sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%lx_1, [1], True), kwargs = {})
%mul : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%lx_0, %sum_1), kwargs = {})
%linear : [num_users=1] = call_function[target=torch.ops.aten.linear.default](args = (%x, %linear_weight, %linear_bias), kwargs = {})
%sigmoid : [num_users=1] = call_function[target=torch.ops.aten.sigmoid.default](args = (%linear,), kwargs = {})
%sub : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%sigmoid, %buff), kwargs = {})
%add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%sub, %mul), kwargs = {})
return (add,)
nostrict-decall¶
graph():
%linear_weight : [num_users=1] = get_attr[target=linear.weight]
%linear_bias : [num_users=1] = get_attr[target=linear.bias]
%buff : [num_users=1] = get_attr[target=buff]
%x : [num_users=1] = placeholder[target=x]
%lx_0 : [num_users=1] = placeholder[target=lx_0]
%lx_1 : [num_users=1] = placeholder[target=lx_1]
%sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%lx_1, [1], True), kwargs = {})
%mul_1 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%lx_0, %sum_1), kwargs = {})
%permute : [num_users=1] = call_function[target=torch.ops.aten.permute.default](args = (%linear_weight, [1, 0]), kwargs = {})
%addmm : [num_users=1] = call_function[target=torch.ops.aten.addmm.default](args = (%linear_bias, %x, %permute), kwargs = {})
%sigmoid : [num_users=1] = call_function[target=torch.ops.aten.sigmoid.default](args = (%addmm,), kwargs = {})
%sub_4 : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%sigmoid, %buff), kwargs = {})
%add_15 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%sub_4, %mul_1), kwargs = {})
return (add_15,)
jit¶
graph():
%linear_bias : [num_users=1] = get_attr[target=linear.bias]
%linear_weight : [num_users=1] = get_attr[target=linear.weight]
%buff : [num_users=1] = get_attr[target=buff]
%x : [num_users=1] = placeholder[target=x]
%lx_0 : [num_users=1] = placeholder[target=lx_0]
%lx_1 : [num_users=1] = placeholder[target=lx_1]
%sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%lx_1, [1], True), kwargs = {})
%mul : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%lx_0, %sum_1), kwargs = {})
%linear : [num_users=1] = call_function[target=torch.ops.aten.linear.default](args = (%x, %linear_weight, %linear_bias), kwargs = {})
%sigmoid : [num_users=1] = call_function[target=torch.ops.aten.sigmoid.default](args = (%linear,), kwargs = {})
%sub : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%sigmoid, %buff), kwargs = {})
%add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%sub, %mul), kwargs = {})
return (add,)
jit-decall¶
graph():
%linear_bias : [num_users=1] = get_attr[target=linear.bias]
%linear_weight : [num_users=1] = get_attr[target=linear.weight]
%buff : [num_users=1] = get_attr[target=buff]
%x : [num_users=1] = placeholder[target=x]
%lx_0 : [num_users=1] = placeholder[target=lx_0]
%lx_1 : [num_users=1] = placeholder[target=lx_1]
%sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%lx_1, [1], True), kwargs = {})
%mul_1 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%lx_0, %sum_1), kwargs = {})
%permute : [num_users=1] = call_function[target=torch.ops.aten.permute.default](args = (%linear_weight, [1, 0]), kwargs = {})
%addmm : [num_users=1] = call_function[target=torch.ops.aten.addmm.default](args = (%linear_bias, %x, %permute), kwargs = {})
%sigmoid : [num_users=1] = call_function[target=torch.ops.aten.sigmoid.default](args = (%addmm,), kwargs = {})
%sub_4 : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%sigmoid, %buff), kwargs = {})
%add_15 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%sub_4, %mul_1), kwargs = {})
return (add_15,)
tracing¶
- Traceback (most recent call last):
- File “/home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/graph_module.py”, line 393, in __call__
- File “/home/xadupre/vv/this312/lib/python3.12/site-packages/torch/nn/modules/module.py”, line 1751, in _wrapped_call_impl
- File “/home/xadupre/vv/this312/lib/python3.12/site-packages/torch/nn/modules/module.py”, line 1762, in _call_impl
- File “<eval_with_key>.12938 from /home/xadupre/github/experimental-experiment/experimental_experiment/torch_interpreter/eval/model_cases.py:784 in forward”, line 9, in forward
- add = sub + lx; sub = lx = None
~~~~^~~~
TypeError: unsupported operand type(s) for +: ‘Tensor’ and ‘list’
- Call using an FX-traced Module, line 9 of the traced Module’s generated forward function:
sub = sigmoid - buff; sigmoid = buff = None add = sub + lx; sub = lx = None
- ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ <— HERE
return add
FAILED
unsupported operand type(s) for +: 'Tensor' and 'list'
BuildInLen¶
forward¶
def forward(self, x, lx: list):
t = lx[0] * lx[1].sum(axis=1, keepdim=True)
if len(lx) > 2:
t = t + lx[2].sum(axis=1, keepdim=True)
return torch.sigmoid(self.linear(x)) - self.buff + t
strict¶
FAILED
Trying to flatten user inputs with exported input tree spec:
TreeSpec(tuple, None, [TreeSpec(tuple, None, [*,
TreeSpec(list, None, [*,
*])]),
TreeSpec(dict, [], [])])
but actually got inputs with tree spec of:
TreeSpec(tuple, None, [TreeSpec(tuple, None, [*,
TreeSpec(list, None, [*,
*,
*])]),
TreeSpec(dict, [], [])]).
Please check that the inputs have the same number of args and kwargs as the ones you used when tracing.
strict-decall¶
FAILED
Trying to flatten user inputs with exported input tree spec:
TreeSpec(tuple, None, [TreeSpec(tuple, None, [*,
TreeSpec(list, None, [*,
*])]),
TreeSpec(dict, [], [])])
but actually got inputs with tree spec of:
TreeSpec(tuple, None, [TreeSpec(tuple, None, [*,
TreeSpec(list, None, [*,
*,
*])]),
TreeSpec(dict, [], [])]).
Please check that the inputs have the same number of args and kwargs as the ones you used when tracing.
nostrict¶
FAILED
Trying to flatten user inputs with exported input tree spec:
TreeSpec(tuple, None, [TreeSpec(tuple, None, [*,
TreeSpec(list, None, [*,
*])]),
TreeSpec(dict, [], [])])
but actually got inputs with tree spec of:
TreeSpec(tuple, None, [TreeSpec(tuple, None, [*,
TreeSpec(list, None, [*,
*,
*])]),
TreeSpec(dict, [], [])]).
Please check that the inputs have the same number of args and kwargs as the ones you used when tracing.
nostrict-decall¶
FAILED
Trying to flatten user inputs with exported input tree spec:
TreeSpec(tuple, None, [TreeSpec(tuple, None, [*,
TreeSpec(list, None, [*,
*])]),
TreeSpec(dict, [], [])])
but actually got inputs with tree spec of:
TreeSpec(tuple, None, [TreeSpec(tuple, None, [*,
TreeSpec(list, None, [*,
*,
*])]),
TreeSpec(dict, [], [])]).
Please check that the inputs have the same number of args and kwargs as the ones you used when tracing.
jit¶
FAILED
Trying to flatten user inputs with exported input tree spec:
TreeSpec(tuple, None, [TreeSpec(tuple, None, [*,
TreeSpec(list, None, [*,
*])]),
TreeSpec(dict, [], [])])
but actually got inputs with tree spec of:
TreeSpec(tuple, None, [TreeSpec(tuple, None, [*,
TreeSpec(list, None, [*,
*,
*])]),
TreeSpec(dict, [], [])]).
Please check that the inputs have the same number of args and kwargs as the ones you used when tracing.
jit-decall¶
FAILED
Trying to flatten user inputs with exported input tree spec:
TreeSpec(tuple, None, [TreeSpec(tuple, None, [*,
TreeSpec(list, None, [*,
*])]),
TreeSpec(dict, [], [])])
but actually got inputs with tree spec of:
TreeSpec(tuple, None, [TreeSpec(tuple, None, [*,
TreeSpec(list, None, [*,
*,
*])]),
TreeSpec(dict, [], [])]).
Please check that the inputs have the same number of args and kwargs as the ones you used when tracing.
tracing¶
FAILED
len(.) expects an integer, len needs to be replaced. You should use _len.
ComplexPolar¶
forward¶
def forward(self, x, angle):
return torch.polar(x, angle)
strict¶
graph():
%x : [num_users=1] = placeholder[target=x]
%angle : [num_users=1] = placeholder[target=angle]
%polar : [num_users=1] = call_function[target=torch.ops.aten.polar.default](args = (%x, %angle), kwargs = {})
return (polar,)
strict-decall¶
graph():
%x : [num_users=1] = placeholder[target=x]
%angle : [num_users=1] = placeholder[target=angle]
%polar : [num_users=1] = call_function[target=torch.ops.aten.polar.default](args = (%x, %angle), kwargs = {})
return (polar,)
nostrict¶
graph():
%x : [num_users=1] = placeholder[target=x]
%angle : [num_users=1] = placeholder[target=angle]
%polar : [num_users=1] = call_function[target=torch.ops.aten.polar.default](args = (%x, %angle), kwargs = {})
return (polar,)
nostrict-decall¶
graph():
%x : [num_users=1] = placeholder[target=x]
%angle : [num_users=1] = placeholder[target=angle]
%polar : [num_users=1] = call_function[target=torch.ops.aten.polar.default](args = (%x, %angle), kwargs = {})
return (polar,)
jit¶
graph():
%x : [num_users=1] = placeholder[target=x]
%angle : [num_users=1] = placeholder[target=angle]
%polar : [num_users=1] = call_function[target=torch.ops.aten.polar.default](args = (%x, %angle), kwargs = {})
return (polar,)
jit-decall¶
graph():
%x : [num_users=1] = placeholder[target=x]
%angle : [num_users=1] = placeholder[target=angle]
%polar : [num_users=1] = call_function[target=torch.ops.aten.polar.default](args = (%x, %angle), kwargs = {})
return (polar,)
tracing¶
graph():
%x : [num_users=1] = placeholder[target=x]
%angle : [num_users=1] = placeholder[target=angle]
%polar : [num_users=1] = call_function[target=torch.polar](args = (%x, %angle), kwargs = {})
return polar
ControlFlowCond¶
forward¶
def forward(self, x):
def true_fn(x):
return torch.sin(x)
def false_fn(x):
return torch.cos(x)
return torch.cond(x.sum() > 0, true_fn, false_fn, [x])
strict¶
graph():
%x : [num_users=2] = placeholder[target=x]
%sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.default](args = (%x,), kwargs = {})
%gt : [num_users=1] = call_function[target=torch.ops.aten.gt.Scalar](args = (%sum_1, 0), kwargs = {})
%true_graph_0 : [num_users=1] = get_attr[target=true_graph_0]
%false_graph_0 : [num_users=1] = get_attr[target=false_graph_0]
%cond : [num_users=1] = call_function[target=torch.ops.higher_order.cond](args = (%gt, %true_graph_0, %false_graph_0, [%x]), kwargs = {})
%getitem : [num_users=1] = call_function[target=operator.getitem](args = (%cond, 0), kwargs = {})
return (getitem,)
strict-decall¶
graph():
%x : [num_users=2] = placeholder[target=x]
%sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%x, []), kwargs = {})
%gt : [num_users=1] = call_function[target=torch.ops.aten.gt.Scalar](args = (%sum_1, 0), kwargs = {})
%true_graph_0 : [num_users=1] = get_attr[target=true_graph_0]
%false_graph_0 : [num_users=1] = get_attr[target=false_graph_0]
%cond : [num_users=1] = call_function[target=torch.ops.higher_order.cond](args = (%gt, %true_graph_0, %false_graph_0, [%x]), kwargs = {})
%getitem : [num_users=1] = call_function[target=operator.getitem](args = (%cond, 0), kwargs = {})
return (getitem,)
nostrict¶
FAILED
Constraints violated (batch)! For more information, run with TORCH_LOGS="+dynamic".
- Not all values of batch = L['args'][0][0].size()[0] in the specified range are valid because batch was inferred to be a constant (5).
Suggested fixes:
batch = 5
nostrict-decall¶
FAILED
Constraints violated (batch)! For more information, run with TORCH_LOGS="+dynamic".
- Not all values of batch = L['args'][0][0].size()[0] in the specified range are valid because batch was inferred to be a constant (5).
Suggested fixes:
batch = 5
jit¶
FAILED
Detected that you are using FX to torch.jit.trace a dynamo-optimized function. This is not supported at the moment.
jit-decall¶
FAILED
Detected that you are using FX to torch.jit.trace a dynamo-optimized function. This is not supported at the moment.
tracing¶
graph():
%x : [num_users=2] = placeholder[target=x]
%sum_1 : [num_users=1] = call_method[target=sum](args = (%x,), kwargs = {})
%gt : [num_users=1] = call_function[target=operator.gt](args = (%sum_1, 0), kwargs = {})
%_cb_cond_true_fn_0 : [num_users=1] = get_attr[target=_cb_cond_true_fn_0]
%_cb_cond_false_fn_0 : [num_users=1] = get_attr[target=_cb_cond_false_fn_0]
%condcc : [num_users=1] = call_function[target=torch.ops.higher_order.cond](args = (%gt, %_cb_cond_true_fn_0, %_cb_cond_false_fn_0, [%x]), kwargs = {})
return condcc
ControlFlowCond2Inputs¶
forward¶
def forward(self, x, y):
def true_fn(x, y):
return torch.sin(x), torch.cos(x) + y
def false_fn(x, y):
return torch.cos(x), torch.sin(x) + y
return torch.cond(x.sum() > 0, true_fn, false_fn, [x, y])
strict¶
graph():
%x : [num_users=2] = placeholder[target=x]
%y : [num_users=1] = placeholder[target=y]
%sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.default](args = (%x,), kwargs = {})
%gt : [num_users=1] = call_function[target=torch.ops.aten.gt.Scalar](args = (%sum_1, 0), kwargs = {})
%true_graph_0 : [num_users=1] = get_attr[target=true_graph_0]
%false_graph_0 : [num_users=1] = get_attr[target=false_graph_0]
%cond : [num_users=2] = call_function[target=torch.ops.higher_order.cond](args = (%gt, %true_graph_0, %false_graph_0, [%x, %y]), kwargs = {})
%getitem : [num_users=1] = call_function[target=operator.getitem](args = (%cond, 0), kwargs = {})
%getitem_1 : [num_users=1] = call_function[target=operator.getitem](args = (%cond, 1), kwargs = {})
return (getitem, getitem_1)
strict-decall¶
graph():
%x : [num_users=2] = placeholder[target=x]
%y : [num_users=1] = placeholder[target=y]
%sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%x, []), kwargs = {})
%gt : [num_users=1] = call_function[target=torch.ops.aten.gt.Scalar](args = (%sum_1, 0), kwargs = {})
%true_graph_0 : [num_users=1] = get_attr[target=true_graph_0]
%false_graph_0 : [num_users=1] = get_attr[target=false_graph_0]
%cond : [num_users=2] = call_function[target=torch.ops.higher_order.cond](args = (%gt, %true_graph_0, %false_graph_0, [%x, %y]), kwargs = {})
%getitem : [num_users=1] = call_function[target=operator.getitem](args = (%cond, 0), kwargs = {})
%getitem_1 : [num_users=1] = call_function[target=operator.getitem](args = (%cond, 1), kwargs = {})
return (getitem, getitem_1)
nostrict¶
FAILED
Constraints violated (batch)! For more information, run with TORCH_LOGS="+dynamic".
- Not all values of batch = L['args'][0][0].size()[0] in the specified range are valid because batch was inferred to be a constant (5).
- Not all values of batch = L['args'][0][1].size()[0] in the specified range are valid because batch was inferred to be a constant (5).
Suggested fixes:
batch = 5
nostrict-decall¶
FAILED
Constraints violated (batch)! For more information, run with TORCH_LOGS="+dynamic".
- Not all values of batch = L['args'][0][0].size()[0] in the specified range are valid because batch was inferred to be a constant (5).
- Not all values of batch = L['args'][0][1].size()[0] in the specified range are valid because batch was inferred to be a constant (5).
Suggested fixes:
batch = 5
jit¶
FAILED
Detected that you are using FX to torch.jit.trace a dynamo-optimized function. This is not supported at the moment.
jit-decall¶
FAILED
Detected that you are using FX to torch.jit.trace a dynamo-optimized function. This is not supported at the moment.
tracing¶
graph():
%x : [num_users=2] = placeholder[target=x]
%y : [num_users=1] = placeholder[target=y]
%sum_1 : [num_users=1] = call_method[target=sum](args = (%x,), kwargs = {})
%gt : [num_users=1] = call_function[target=operator.gt](args = (%sum_1, 0), kwargs = {})
%_cb_cond_true_fn_0 : [num_users=1] = get_attr[target=_cb_cond_true_fn_0]
%_cb_cond_false_fn_0 : [num_users=1] = get_attr[target=_cb_cond_false_fn_0]
%condcc : [num_users=1] = call_function[target=torch.ops.higher_order.cond](args = (%gt, %_cb_cond_true_fn_0, %_cb_cond_false_fn_0, [%x, %y]), kwargs = {})
return condcc
ControlFlowCond2Outputs¶
forward¶
def forward(self, x):
def true_fn(x):
return torch.sin(x), torch.cos(x)
def false_fn(x):
return torch.cos(x), torch.sin(x)
return torch.cond(x.sum() > 0, true_fn, false_fn, [x])
strict¶
graph():
%x : [num_users=2] = placeholder[target=x]
%sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.default](args = (%x,), kwargs = {})
%gt : [num_users=1] = call_function[target=torch.ops.aten.gt.Scalar](args = (%sum_1, 0), kwargs = {})
%true_graph_0 : [num_users=1] = get_attr[target=true_graph_0]
%false_graph_0 : [num_users=1] = get_attr[target=false_graph_0]
%cond : [num_users=2] = call_function[target=torch.ops.higher_order.cond](args = (%gt, %true_graph_0, %false_graph_0, [%x]), kwargs = {})
%getitem : [num_users=1] = call_function[target=operator.getitem](args = (%cond, 0), kwargs = {})
%getitem_1 : [num_users=1] = call_function[target=operator.getitem](args = (%cond, 1), kwargs = {})
return (getitem, getitem_1)
strict-decall¶
graph():
%x : [num_users=2] = placeholder[target=x]
%sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%x, []), kwargs = {})
%gt : [num_users=1] = call_function[target=torch.ops.aten.gt.Scalar](args = (%sum_1, 0), kwargs = {})
%true_graph_0 : [num_users=1] = get_attr[target=true_graph_0]
%false_graph_0 : [num_users=1] = get_attr[target=false_graph_0]
%cond : [num_users=2] = call_function[target=torch.ops.higher_order.cond](args = (%gt, %true_graph_0, %false_graph_0, [%x]), kwargs = {})
%getitem : [num_users=1] = call_function[target=operator.getitem](args = (%cond, 0), kwargs = {})
%getitem_1 : [num_users=1] = call_function[target=operator.getitem](args = (%cond, 1), kwargs = {})
return (getitem, getitem_1)
nostrict¶
FAILED
Constraints violated (batch)! For more information, run with TORCH_LOGS="+dynamic".
- Not all values of batch = L['args'][0][0].size()[0] in the specified range are valid because batch was inferred to be a constant (5).
Suggested fixes:
batch = 5
nostrict-decall¶
FAILED
Constraints violated (batch)! For more information, run with TORCH_LOGS="+dynamic".
- Not all values of batch = L['args'][0][0].size()[0] in the specified range are valid because batch was inferred to be a constant (5).
Suggested fixes:
batch = 5
jit¶
FAILED
Detected that you are using FX to torch.jit.trace a dynamo-optimized function. This is not supported at the moment.
jit-decall¶
FAILED
Detected that you are using FX to torch.jit.trace a dynamo-optimized function. This is not supported at the moment.
tracing¶
graph():
%x : [num_users=2] = placeholder[target=x]
%sum_1 : [num_users=1] = call_method[target=sum](args = (%x,), kwargs = {})
%gt : [num_users=1] = call_function[target=operator.gt](args = (%sum_1, 0), kwargs = {})
%_cb_cond_true_fn_0 : [num_users=1] = get_attr[target=_cb_cond_true_fn_0]
%_cb_cond_false_fn_0 : [num_users=1] = get_attr[target=_cb_cond_false_fn_0]
%condcc : [num_users=1] = call_function[target=torch.ops.higher_order.cond](args = (%gt, %_cb_cond_true_fn_0, %_cb_cond_false_fn_0, [%x]), kwargs = {})
return condcc
ControlFlowCondConstant¶
forward¶
def forward(self, x):
def true_fn(x):
return torch.sin(x) - torch.ones(x.shape, dtype=x.dtype)
def false_fn(x):
return torch.cos(x) + torch.ones((1, 1024), dtype=x.dtype)
return torch.cond(x.sum() > 0, true_fn, false_fn, [x])
strict¶
graph():
%x : [num_users=2] = placeholder[target=x]
%sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.default](args = (%x,), kwargs = {})
%gt : [num_users=1] = call_function[target=torch.ops.aten.gt.Scalar](args = (%sum_1, 0), kwargs = {})
%true_graph_0 : [num_users=1] = get_attr[target=true_graph_0]
%false_graph_0 : [num_users=1] = get_attr[target=false_graph_0]
%cond : [num_users=1] = call_function[target=torch.ops.higher_order.cond](args = (%gt, %true_graph_0, %false_graph_0, [%x]), kwargs = {})
%getitem : [num_users=1] = call_function[target=operator.getitem](args = (%cond, 0), kwargs = {})
return (getitem,)
strict-decall¶
graph():
%x : [num_users=2] = placeholder[target=x]
%sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%x, []), kwargs = {})
%gt : [num_users=1] = call_function[target=torch.ops.aten.gt.Scalar](args = (%sum_1, 0), kwargs = {})
%true_graph_0 : [num_users=1] = get_attr[target=true_graph_0]
%false_graph_0 : [num_users=1] = get_attr[target=false_graph_0]
%cond : [num_users=1] = call_function[target=torch.ops.higher_order.cond](args = (%gt, %true_graph_0, %false_graph_0, [%x]), kwargs = {})
%getitem : [num_users=1] = call_function[target=operator.getitem](args = (%cond, 0), kwargs = {})
return (getitem,)
nostrict¶
graph():
%x : [num_users=2] = placeholder[target=x]
%sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.default](args = (%x,), kwargs = {})
%gt : [num_users=1] = call_function[target=torch.ops.aten.gt.Scalar](args = (%sum_1, 0), kwargs = {})
%true_graph_0 : [num_users=1] = get_attr[target=true_graph_0]
%false_graph_0 : [num_users=1] = get_attr[target=false_graph_0]
%cond : [num_users=1] = call_function[target=torch.ops.higher_order.cond](args = (%gt, %true_graph_0, %false_graph_0, [%x]), kwargs = {})
%getitem : [num_users=1] = call_function[target=operator.getitem](args = (%cond, 0), kwargs = {})
return (getitem,)
nostrict-decall¶
graph():
%x : [num_users=2] = placeholder[target=x]
%sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%x, []), kwargs = {})
%gt : [num_users=1] = call_function[target=torch.ops.aten.gt.Scalar](args = (%sum_1, 0), kwargs = {})
%true_graph_0 : [num_users=1] = get_attr[target=true_graph_0]
%false_graph_0 : [num_users=1] = get_attr[target=false_graph_0]
%cond : [num_users=1] = call_function[target=torch.ops.higher_order.cond](args = (%gt, %true_graph_0, %false_graph_0, [%x]), kwargs = {})
%getitem : [num_users=1] = call_function[target=operator.getitem](args = (%cond, 0), kwargs = {})
return (getitem,)
jit¶
FAILED
Detected that you are using FX to torch.jit.trace a dynamo-optimized function. This is not supported at the moment.
jit-decall¶
FAILED
Detected that you are using FX to torch.jit.trace a dynamo-optimized function. This is not supported at the moment.
tracing¶
graph():
%x : [num_users=2] = placeholder[target=x]
%sum_1 : [num_users=1] = call_method[target=sum](args = (%x,), kwargs = {})
%gt : [num_users=1] = call_function[target=operator.gt](args = (%sum_1, 0), kwargs = {})
%_cb_cond_true_fn_0 : [num_users=1] = get_attr[target=_cb_cond_true_fn_0]
%_cb_cond_false_fn_0 : [num_users=1] = get_attr[target=_cb_cond_false_fn_0]
%condcc : [num_users=1] = call_function[target=torch.ops.higher_order.cond](args = (%gt, %_cb_cond_true_fn_0, %_cb_cond_false_fn_0, [%x]), kwargs = {})
return condcc
ControlFlowCondNestedModule¶
forward¶
def forward(self, x):
def true_fn(x):
return self.submodule(x)
def false_fn(x):
return x - self.weight
y = torch.cond(x.sum() > 0, true_fn, false_fn, [x])
return y
strict¶
graph():
%submodule_weight : [num_users=1] = get_attr[target=submodule.weight]
%weight : [num_users=1] = get_attr[target=weight]
%x : [num_users=2] = placeholder[target=x]
%sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.default](args = (%x,), kwargs = {})
%gt : [num_users=1] = call_function[target=torch.ops.aten.gt.Scalar](args = (%sum_1, 0), kwargs = {})
%true_graph_0 : [num_users=1] = get_attr[target=true_graph_0]
%false_graph_0 : [num_users=1] = get_attr[target=false_graph_0]
%cond : [num_users=1] = call_function[target=torch.ops.higher_order.cond](args = (%gt, %true_graph_0, %false_graph_0, [%x, %submodule_weight, %weight]), kwargs = {})
%getitem : [num_users=1] = call_function[target=operator.getitem](args = (%cond, 0), kwargs = {})
return (getitem,)
strict-decall¶
graph():
%weight : [num_users=1] = get_attr[target=weight]
%submodule_weight : [num_users=1] = get_attr[target=submodule.weight]
%x : [num_users=2] = placeholder[target=x]
%sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%x, []), kwargs = {})
%gt : [num_users=1] = call_function[target=torch.ops.aten.gt.Scalar](args = (%sum_1, 0), kwargs = {})
%true_graph_0 : [num_users=1] = get_attr[target=true_graph_0]
%false_graph_0 : [num_users=1] = get_attr[target=false_graph_0]
%cond : [num_users=1] = call_function[target=torch.ops.higher_order.cond](args = (%gt, %true_graph_0, %false_graph_0, (%x, %submodule_weight, %weight)), kwargs = {})
%getitem : [num_users=1] = call_function[target=operator.getitem](args = (%cond, 0), kwargs = {})
return (getitem,)
nostrict¶
graph():
%weight : [num_users=1] = get_attr[target=weight]
%submodule_weight : [num_users=1] = get_attr[target=submodule.weight]
%x : [num_users=2] = placeholder[target=x]
%sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.default](args = (%x,), kwargs = {})
%gt : [num_users=1] = call_function[target=torch.ops.aten.gt.Scalar](args = (%sum_1, 0), kwargs = {})
%true_graph_0 : [num_users=1] = get_attr[target=true_graph_0]
%false_graph_0 : [num_users=1] = get_attr[target=false_graph_0]
%cond : [num_users=1] = call_function[target=torch.ops.higher_order.cond](args = (%gt, %true_graph_0, %false_graph_0, [%x, %submodule_weight, %weight]), kwargs = {})
%getitem : [num_users=1] = call_function[target=operator.getitem](args = (%cond, 0), kwargs = {})
return (getitem,)
nostrict-decall¶
graph():
%weight : [num_users=1] = get_attr[target=weight]
%submodule_weight : [num_users=1] = get_attr[target=submodule.weight]
%x : [num_users=2] = placeholder[target=x]
%sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%x, []), kwargs = {})
%gt : [num_users=1] = call_function[target=torch.ops.aten.gt.Scalar](args = (%sum_1, 0), kwargs = {})
%true_graph_0 : [num_users=1] = get_attr[target=true_graph_0]
%false_graph_0 : [num_users=1] = get_attr[target=false_graph_0]
%cond : [num_users=1] = call_function[target=torch.ops.higher_order.cond](args = (%gt, %true_graph_0, %false_graph_0, (%x, %submodule_weight, %weight)), kwargs = {})
%getitem : [num_users=1] = call_function[target=operator.getitem](args = (%cond, 0), kwargs = {})
return (getitem,)
jit¶
FAILED
Detected that you are using FX to torch.jit.trace a dynamo-optimized function. This is not supported at the moment.
jit-decall¶
FAILED
Detected that you are using FX to torch.jit.trace a dynamo-optimized function. This is not supported at the moment.
tracing¶
graph():
%x : [num_users=2] = placeholder[target=x]
%sum_1 : [num_users=1] = call_method[target=sum](args = (%x,), kwargs = {})
%gt : [num_users=1] = call_function[target=operator.gt](args = (%sum_1, 0), kwargs = {})
%_cb_cond_true_fn_0 : [num_users=1] = get_attr[target=_cb_cond_true_fn_0]
%_cb_cond_false_fn_0 : [num_users=1] = get_attr[target=_cb_cond_false_fn_0]
%condcc : [num_users=1] = call_function[target=torch.ops.higher_order.cond](args = (%gt, %_cb_cond_true_fn_0, %_cb_cond_false_fn_0, [%x]), kwargs = {})
return condcc
ControlFlowCondNonZero¶
forward¶
def forward(self, input_ids, image_features, vocab_size):
def then_branch(input_ids, image_features, vocab_size):
input_shape = input_ids.size()
input_ids = input_ids.view(-1, input_shape[-1])
condition = (input_ids < 0) & (input_ids > -int(1e9))
positions = torch.nonzero(condition, as_tuple=True)
input_ids = input_ids.clamp_min(0).clamp_max(vocab_size)
return (input_ids, positions[0], positions[1])
def else_branch(input_ids, image_features, vocab_size):
r = torch.where(torch.zeros((1, 1), dtype=torch.bool))
return (input_ids, r[0], r[1])
a, b, c = torch.cond(
image_features.numel() > 0,
then_branch,
else_branch,
[input_ids, image_features, vocab_size],
)
return a, b, c
strict¶
FAILED
Expected true_fn_output and false_fn_output to have same metadata but found:
pair[1] differ in 'shape: torch.Size([u0]) vs torch.Size([u1])', where lhs is FakeTensor(..., size=(u0,), dtype=torch.int64) and rhs is FakeTensor(..., size=(u1,), dtype=torch.int64)
pair[2] differ in 'shape: torch.Size([u0]) vs torch.Size([u1])', where lhs is FakeTensor(..., size=(u0,), dtype=torch.int64) and rhs is FakeTensor(..., size=(u1,), dtype=torch.int64)
from user code:
File "/home/xadupre/github/experimental-experiment/experimental_experiment/torch_interpreter/eval/model_cases.py", line 339, in forward
a, b, c = torch.cond(
File "/home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_higher_order_ops/cond.py", line 138, in cond
return cond_op(pred, true_fn, false_fn, operands)
Set TORCH_LOGS="+dynamo" and TORCHDYNAMO_VERBOSE=1 for more information
strict-decall¶
FAILED
Expected true_fn_output and false_fn_output to have same metadata but found:
pair[1] differ in 'shape: torch.Size([u0]) vs torch.Size([u1])', where lhs is FakeTensor(..., size=(u0,), dtype=torch.int64) and rhs is FakeTensor(..., size=(u1,), dtype=torch.int64)
pair[2] differ in 'shape: torch.Size([u0]) vs torch.Size([u1])', where lhs is FakeTensor(..., size=(u0,), dtype=torch.int64) and rhs is FakeTensor(..., size=(u1,), dtype=torch.int64)
from user code:
File "/home/xadupre/github/experimental-experiment/experimental_experiment/torch_interpreter/eval/model_cases.py", line 339, in forward
a, b, c = torch.cond(
File "/home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_higher_order_ops/cond.py", line 138, in cond
return cond_op(pred, true_fn, false_fn, operands)
Set TORCH_LOGS="+dynamo" and TORCHDYNAMO_VERBOSE=1 for more information
nostrict¶
FAILED
Expect operands to be a tuple of possibly nested dict/list/tuple that only consists of tensor leaves, but got [FakeTensor(..., size=(s1, 12), dtype=torch.int64), FakeTensor(..., size=(s2, s3)), 1025].
nostrict-decall¶
FAILED
Expect operands to be a tuple of possibly nested dict/list/tuple that only consists of tensor leaves, but got [FakeTensor(..., size=(s1, 12), dtype=torch.int64), FakeTensor(..., size=(s2, s3)), 1025].
jit¶
FAILED
Type 'Tuple[Tensor, Tensor, int]' cannot be traced. Only Tensors and (possibly nested) Lists, Dicts, and Tuples of Tensors can be traced
jit-decall¶
FAILED
Type 'Tuple[Tensor, Tensor, int]' cannot be traced. Only Tensors and (possibly nested) Lists, Dicts, and Tuples of Tensors can be traced
tracing¶
graph():
%input_ids : [num_users=1] = placeholder[target=input_ids]
%image_features : [num_users=2] = placeholder[target=image_features]
%vocab_size : [num_users=1] = placeholder[target=vocab_size]
%numel : [num_users=1] = call_method[target=numel](args = (%image_features,), kwargs = {})
%gt : [num_users=1] = call_function[target=operator.gt](args = (%numel, 0), kwargs = {})
%_cb_cond_then_branch_0 : [num_users=1] = get_attr[target=_cb_cond_then_branch_0]
%_cb_cond_else_branch_0 : [num_users=1] = get_attr[target=_cb_cond_else_branch_0]
%condcc : [num_users=3] = call_function[target=torch.ops.higher_order.cond](args = (%gt, %_cb_cond_then_branch_0, %_cb_cond_else_branch_0, [%input_ids, %image_features, %vocab_size]), kwargs = {})
%getitem : [num_users=1] = call_function[target=operator.getitem](args = (%condcc, 0), kwargs = {})
%getitem_1 : [num_users=1] = call_function[target=operator.getitem](args = (%condcc, 1), kwargs = {})
%getitem_2 : [num_users=1] = call_function[target=operator.getitem](args = (%condcc, 2), kwargs = {})
return (getitem, getitem_1, getitem_2)
ControlFlowNestCond¶
forward¶
def forward(self, x):
def true_fn2(x):
def true_fn1(x):
return torch.sin(x)
def false_fn1(x):
return torch.cos(x)
return torch.cond(x.sum() < 0, true_fn1, false_fn1, [x])
def false_fn2(x):
return -x
return torch.cond(x.sum() > 0, true_fn2, false_fn2, [x])
strict¶
graph():
%x : [num_users=2] = placeholder[target=x]
%sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.default](args = (%x,), kwargs = {})
%gt : [num_users=1] = call_function[target=torch.ops.aten.gt.Scalar](args = (%sum_1, 0), kwargs = {})
%true_graph_0 : [num_users=1] = get_attr[target=true_graph_0]
%false_graph_0 : [num_users=1] = get_attr[target=false_graph_0]
%cond : [num_users=1] = call_function[target=torch.ops.higher_order.cond](args = (%gt, %true_graph_0, %false_graph_0, [%x]), kwargs = {})
%getitem : [num_users=1] = call_function[target=operator.getitem](args = (%cond, 0), kwargs = {})
return (getitem,)
strict-decall¶
graph():
%x : [num_users=2] = placeholder[target=x]
%sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%x, []), kwargs = {})
%gt : [num_users=1] = call_function[target=torch.ops.aten.gt.Scalar](args = (%sum_1, 0), kwargs = {})
%true_graph_0 : [num_users=1] = get_attr[target=true_graph_0]
%false_graph_0 : [num_users=1] = get_attr[target=false_graph_0]
%cond : [num_users=1] = call_function[target=torch.ops.higher_order.cond](args = (%gt, %true_graph_0, %false_graph_0, [%x]), kwargs = {})
%getitem : [num_users=1] = call_function[target=operator.getitem](args = (%cond, 0), kwargs = {})
return (getitem,)
nostrict¶
graph():
%x : [num_users=2] = placeholder[target=x]
%sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.default](args = (%x,), kwargs = {})
%gt : [num_users=1] = call_function[target=torch.ops.aten.gt.Scalar](args = (%sum_1, 0), kwargs = {})
%true_graph_0 : [num_users=1] = get_attr[target=true_graph_0]
%false_graph_0 : [num_users=1] = get_attr[target=false_graph_0]
%cond : [num_users=1] = call_function[target=torch.ops.higher_order.cond](args = (%gt, %true_graph_0, %false_graph_0, [%x]), kwargs = {})
%getitem : [num_users=1] = call_function[target=operator.getitem](args = (%cond, 0), kwargs = {})
return (getitem,)
nostrict-decall¶
graph():
%x : [num_users=2] = placeholder[target=x]
%sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%x, []), kwargs = {})
%gt : [num_users=1] = call_function[target=torch.ops.aten.gt.Scalar](args = (%sum_1, 0), kwargs = {})
%true_graph_0 : [num_users=1] = get_attr[target=true_graph_0]
%false_graph_0 : [num_users=1] = get_attr[target=false_graph_0]
%cond : [num_users=1] = call_function[target=torch.ops.higher_order.cond](args = (%gt, %true_graph_0, %false_graph_0, [%x]), kwargs = {})
%getitem : [num_users=1] = call_function[target=operator.getitem](args = (%cond, 0), kwargs = {})
return (getitem,)
jit¶
FAILED
Detected that you are using FX to torch.jit.trace a dynamo-optimized function. This is not supported at the moment.
jit-decall¶
FAILED
Detected that you are using FX to torch.jit.trace a dynamo-optimized function. This is not supported at the moment.
tracing¶
graph():
%x : [num_users=2] = placeholder[target=x]
%sum_1 : [num_users=1] = call_method[target=sum](args = (%x,), kwargs = {})
%gt : [num_users=1] = call_function[target=operator.gt](args = (%sum_1, 0), kwargs = {})
%_cb_cond_true_fn2_0 : [num_users=1] = get_attr[target=_cb_cond_true_fn2_0]
%_cb_cond_false_fn2_0 : [num_users=1] = get_attr[target=_cb_cond_false_fn2_0]
%condcc : [num_users=1] = call_function[target=torch.ops.higher_order.cond](args = (%gt, %_cb_cond_true_fn2_0, %_cb_cond_false_fn2_0, [%x]), kwargs = {})
return condcc
ControlFlowScan¶
forward¶
def forward(self, x):
init = torch.zeros_like(x[0])
carry, out = torch.ops.higher_order.scan(
ControlFlowScan.add, [init], [x], reverse=False, additional_inputs=[]
)
return carry
strict¶
graph():
%x : [num_users=2] = placeholder[target=x]
%select : [num_users=1] = call_function[target=torch.ops.aten.select.int](args = (%x, 0, 0), kwargs = {})
%zeros_like : [num_users=1] = call_function[target=torch.ops.aten.zeros_like.default](args = (%select,), kwargs = {pin_memory: False})
%scan_combine_graph_0 : [num_users=1] = get_attr[target=scan_combine_graph_0]
%scan : [num_users=2] = call_function[target=torch.ops.higher_order.scan](args = (%scan_combine_graph_0, [%zeros_like], [%x], False, []), kwargs = {})
%getitem : [num_users=1] = call_function[target=operator.getitem](args = (%scan, 0), kwargs = {})
%getitem_1 : [num_users=0] = call_function[target=operator.getitem](args = (%scan, 1), kwargs = {})
return (getitem,)
strict-decall¶
graph():
%x : [num_users=2] = placeholder[target=x]
%select : [num_users=1] = call_function[target=torch.ops.aten.select.int](args = (%x, 0, 0), kwargs = {})
%full_like : [num_users=1] = call_function[target=torch.ops.aten.full_like.default](args = (%select, 0), kwargs = {pin_memory: False, memory_format: torch.preserve_format})
%scan_combine_graph_0 : [num_users=1] = get_attr[target=scan_combine_graph_0]
%scan : [num_users=1] = call_function[target=torch.ops.higher_order.scan](args = (%scan_combine_graph_0, [%full_like], [%x], False, []), kwargs = {})
%getitem : [num_users=1] = call_function[target=operator.getitem](args = (%scan, 0), kwargs = {})
return (getitem,)
nostrict¶
graph():
%x : [num_users=2] = placeholder[target=x]
%select : [num_users=1] = call_function[target=torch.ops.aten.select.int](args = (%x, 0, 0), kwargs = {})
%zeros_like : [num_users=1] = call_function[target=torch.ops.aten.zeros_like.default](args = (%select,), kwargs = {pin_memory: False})
%scan_combine_graph_0 : [num_users=1] = get_attr[target=scan_combine_graph_0]
%scan : [num_users=2] = call_function[target=torch.ops.higher_order.scan](args = (%scan_combine_graph_0, [%zeros_like], [%x], False, []), kwargs = {})
%getitem : [num_users=1] = call_function[target=operator.getitem](args = (%scan, 0), kwargs = {})
%getitem_1 : [num_users=0] = call_function[target=operator.getitem](args = (%scan, 1), kwargs = {})
return (getitem,)
nostrict-decall¶
graph():
%x : [num_users=2] = placeholder[target=x]
%select : [num_users=1] = call_function[target=torch.ops.aten.select.int](args = (%x, 0, 0), kwargs = {})
%full_like : [num_users=1] = call_function[target=torch.ops.aten.full_like.default](args = (%select, 0), kwargs = {pin_memory: False, memory_format: torch.preserve_format})
%scan_combine_graph_0 : [num_users=1] = get_attr[target=scan_combine_graph_0]
%scan : [num_users=1] = call_function[target=torch.ops.higher_order.scan](args = (%scan_combine_graph_0, [%full_like], [%x], False, []), kwargs = {})
%getitem : [num_users=1] = call_function[target=operator.getitem](args = (%scan, 0), kwargs = {})
return (getitem,)
jit¶
FAILED
could not find kernel for HigherOrderOperator scan at dispatch key DispatchKey.??? (resolved from DispatchKey.???)
jit-decall¶
FAILED
could not find kernel for HigherOrderOperator scan at dispatch key DispatchKey.??? (resolved from DispatchKey.???)
tracing¶
FAILED
Unable to symbolically trace HigherOrderOperators
ControlFlowScan2Carried¶
forward¶
def forward(self, x):
init1 = torch.zeros_like(x[0])
init2 = torch.ones_like(x[0])
carry1, carry2, out1, out2 = torch.ops.higher_order.scan(
ControlFlowScan2Carried.add,
[init1, init2],
[x, x * 2],
# dim=0, # 01/31/2025, not supported anymore
reverse=False,
additional_inputs=[],
)
return carry1, carry2, out1, out2
strict¶
graph():
%x : [num_users=4] = placeholder[target=x]
%select : [num_users=1] = call_function[target=torch.ops.aten.select.int](args = (%x, 0, 0), kwargs = {})
%zeros_like : [num_users=1] = call_function[target=torch.ops.aten.zeros_like.default](args = (%select,), kwargs = {pin_memory: False})
%select_1 : [num_users=1] = call_function[target=torch.ops.aten.select.int](args = (%x, 0, 0), kwargs = {})
%ones_like : [num_users=1] = call_function[target=torch.ops.aten.ones_like.default](args = (%select_1,), kwargs = {pin_memory: False})
%mul : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%x, 2), kwargs = {})
%scan_combine_graph_0 : [num_users=1] = get_attr[target=scan_combine_graph_0]
%scan : [num_users=4] = call_function[target=torch.ops.higher_order.scan](args = (%scan_combine_graph_0, [%zeros_like, %ones_like], [%x, %mul], False, []), kwargs = {})
%getitem : [num_users=1] = call_function[target=operator.getitem](args = (%scan, 0), kwargs = {})
%getitem_1 : [num_users=1] = call_function[target=operator.getitem](args = (%scan, 1), kwargs = {})
%getitem_2 : [num_users=1] = call_function[target=operator.getitem](args = (%scan, 2), kwargs = {})
%getitem_3 : [num_users=1] = call_function[target=operator.getitem](args = (%scan, 3), kwargs = {})
return (getitem, getitem_1, getitem_2, getitem_3)
strict-decall¶
graph():
%x : [num_users=4] = placeholder[target=x]
%select : [num_users=1] = call_function[target=torch.ops.aten.select.int](args = (%x, 0, 0), kwargs = {})
%full_like : [num_users=1] = call_function[target=torch.ops.aten.full_like.default](args = (%select, 0), kwargs = {pin_memory: False, memory_format: torch.preserve_format})
%select_1 : [num_users=1] = call_function[target=torch.ops.aten.select.int](args = (%x, 0, 0), kwargs = {})
%full_like_1 : [num_users=1] = call_function[target=torch.ops.aten.full_like.default](args = (%select_1, 1), kwargs = {pin_memory: False, memory_format: torch.preserve_format})
%mul : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%x, 2), kwargs = {})
%scan_combine_graph_0 : [num_users=1] = get_attr[target=scan_combine_graph_0]
%scan : [num_users=4] = call_function[target=torch.ops.higher_order.scan](args = (%scan_combine_graph_0, [%full_like, %full_like_1], [%x, %mul], False, []), kwargs = {})
%getitem : [num_users=1] = call_function[target=operator.getitem](args = (%scan, 0), kwargs = {})
%getitem_1 : [num_users=1] = call_function[target=operator.getitem](args = (%scan, 1), kwargs = {})
%getitem_2 : [num_users=1] = call_function[target=operator.getitem](args = (%scan, 2), kwargs = {})
%getitem_3 : [num_users=1] = call_function[target=operator.getitem](args = (%scan, 3), kwargs = {})
return (getitem, getitem_1, getitem_2, getitem_3)
nostrict¶
graph():
%x : [num_users=4] = placeholder[target=x]
%select : [num_users=1] = call_function[target=torch.ops.aten.select.int](args = (%x, 0, 0), kwargs = {})
%zeros_like : [num_users=1] = call_function[target=torch.ops.aten.zeros_like.default](args = (%select,), kwargs = {pin_memory: False})
%select_1 : [num_users=1] = call_function[target=torch.ops.aten.select.int](args = (%x, 0, 0), kwargs = {})
%ones_like : [num_users=1] = call_function[target=torch.ops.aten.ones_like.default](args = (%select_1,), kwargs = {pin_memory: False})
%mul : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%x, 2), kwargs = {})
%scan_combine_graph_0 : [num_users=1] = get_attr[target=scan_combine_graph_0]
%scan : [num_users=4] = call_function[target=torch.ops.higher_order.scan](args = (%scan_combine_graph_0, [%zeros_like, %ones_like], [%x, %mul], False, []), kwargs = {})
%getitem : [num_users=1] = call_function[target=operator.getitem](args = (%scan, 0), kwargs = {})
%getitem_1 : [num_users=1] = call_function[target=operator.getitem](args = (%scan, 1), kwargs = {})
%getitem_2 : [num_users=1] = call_function[target=operator.getitem](args = (%scan, 2), kwargs = {})
%getitem_3 : [num_users=1] = call_function[target=operator.getitem](args = (%scan, 3), kwargs = {})
return (getitem, getitem_1, getitem_2, getitem_3)
nostrict-decall¶
graph():
%x : [num_users=4] = placeholder[target=x]
%select : [num_users=1] = call_function[target=torch.ops.aten.select.int](args = (%x, 0, 0), kwargs = {})
%full_like : [num_users=1] = call_function[target=torch.ops.aten.full_like.default](args = (%select, 0), kwargs = {pin_memory: False, memory_format: torch.preserve_format})
%select_1 : [num_users=1] = call_function[target=torch.ops.aten.select.int](args = (%x, 0, 0), kwargs = {})
%full_like_1 : [num_users=1] = call_function[target=torch.ops.aten.full_like.default](args = (%select_1, 1), kwargs = {pin_memory: False, memory_format: torch.preserve_format})
%mul : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%x, 2), kwargs = {})
%scan_combine_graph_0 : [num_users=1] = get_attr[target=scan_combine_graph_0]
%scan : [num_users=4] = call_function[target=torch.ops.higher_order.scan](args = (%scan_combine_graph_0, [%full_like, %full_like_1], [%x, %mul], False, []), kwargs = {})
%getitem : [num_users=1] = call_function[target=operator.getitem](args = (%scan, 0), kwargs = {})
%getitem_1 : [num_users=1] = call_function[target=operator.getitem](args = (%scan, 1), kwargs = {})
%getitem_2 : [num_users=1] = call_function[target=operator.getitem](args = (%scan, 2), kwargs = {})
%getitem_3 : [num_users=1] = call_function[target=operator.getitem](args = (%scan, 3), kwargs = {})
return (getitem, getitem_1, getitem_2, getitem_3)
jit¶
FAILED
could not find kernel for HigherOrderOperator scan at dispatch key DispatchKey.??? (resolved from DispatchKey.???)
jit-decall¶
FAILED
could not find kernel for HigherOrderOperator scan at dispatch key DispatchKey.??? (resolved from DispatchKey.???)
tracing¶
FAILED
Unable to symbolically trace HigherOrderOperators
ControlFlowScanCDist¶
forward¶
def forward(self, x):
carry, out = torch.ops.higher_order.scan(
ControlFlowScanCDist.dist,
[x],
[x],
# dim=0, # 01/31/2025, not supported anymore
reverse=False,
additional_inputs=[],
)
return out
strict¶
graph():
%x : [num_users=1] = placeholder[target=x]
%scan_combine_graph_0 : [num_users=1] = get_attr[target=scan_combine_graph_0]
%scan : [num_users=2] = call_function[target=torch.ops.higher_order.scan](args = (%scan_combine_graph_0, [%x], [%x], False, []), kwargs = {})
%getitem : [num_users=0] = call_function[target=operator.getitem](args = (%scan, 0), kwargs = {})
%getitem_1 : [num_users=1] = call_function[target=operator.getitem](args = (%scan, 1), kwargs = {})
return (getitem_1,)
strict-decall¶
graph():
%x : [num_users=1] = placeholder[target=x]
%scan_combine_graph_0 : [num_users=1] = get_attr[target=scan_combine_graph_0]
%scan : [num_users=1] = call_function[target=torch.ops.higher_order.scan](args = (%scan_combine_graph_0, [%x], [%x], False, []), kwargs = {})
%getitem_1 : [num_users=1] = call_function[target=operator.getitem](args = (%scan, 1), kwargs = {})
return (getitem_1,)
nostrict¶
graph():
%x : [num_users=1] = placeholder[target=x]
%scan_combine_graph_0 : [num_users=1] = get_attr[target=scan_combine_graph_0]
%scan : [num_users=2] = call_function[target=torch.ops.higher_order.scan](args = (%scan_combine_graph_0, [%x], [%x], False, []), kwargs = {})
%getitem : [num_users=0] = call_function[target=operator.getitem](args = (%scan, 0), kwargs = {})
%getitem_1 : [num_users=1] = call_function[target=operator.getitem](args = (%scan, 1), kwargs = {})
return (getitem_1,)
nostrict-decall¶
graph():
%x : [num_users=1] = placeholder[target=x]
%scan_combine_graph_0 : [num_users=1] = get_attr[target=scan_combine_graph_0]
%scan : [num_users=1] = call_function[target=torch.ops.higher_order.scan](args = (%scan_combine_graph_0, [%x], [%x], False, []), kwargs = {})
%getitem_1 : [num_users=1] = call_function[target=operator.getitem](args = (%scan, 1), kwargs = {})
return (getitem_1,)
jit¶
FAILED
could not find kernel for HigherOrderOperator scan at dispatch key DispatchKey.??? (resolved from DispatchKey.???)
jit-decall¶
FAILED
could not find kernel for HigherOrderOperator scan at dispatch key DispatchKey.??? (resolved from DispatchKey.???)
tracing¶
FAILED
Unable to symbolically trace HigherOrderOperators
ControlFlowScanCDist2¶
forward¶
def forward(self, x):
z = torch.tensor([0], dtype=torch.float32)
y = x.clone()
out = torch.ops.higher_order.scan(
ControlFlowScanCDist2.dist,
[z],
[x],
# dim=0, # 01/31/2025, not supported anymore
reverse=False,
additional_inputs=[y],
)
return out[1]
strict¶
graph():
%lifted_tensor_0 : [num_users=1] = get_attr[target=lifted_tensor_0]
%x : [num_users=2] = placeholder[target=x]
%lift_fresh_copy : [num_users=1] = call_function[target=torch.ops.aten.lift_fresh_copy.default](args = (%lifted_tensor_0,), kwargs = {})
%detach_ : [num_users=1] = call_function[target=torch.ops.aten.detach_.default](args = (%lift_fresh_copy,), kwargs = {})
%clone : [num_users=1] = call_function[target=torch.ops.aten.clone.default](args = (%x,), kwargs = {})
%scan_combine_graph_0 : [num_users=1] = get_attr[target=scan_combine_graph_0]
%scan : [num_users=2] = call_function[target=torch.ops.higher_order.scan](args = (%scan_combine_graph_0, [%detach_], [%x], False, [%clone]), kwargs = {})
%getitem : [num_users=0] = call_function[target=operator.getitem](args = (%scan, 0), kwargs = {})
%getitem_1 : [num_users=1] = call_function[target=operator.getitem](args = (%scan, 1), kwargs = {})
return (getitem_1,)
strict-decall¶
graph():
%lifted_tensor_0 : [num_users=1] = get_attr[target=lifted_tensor_0]
%x : [num_users=2] = placeholder[target=x]
%clone : [num_users=1] = call_function[target=torch.ops.aten.clone.default](args = (%lifted_tensor_0,), kwargs = {})
%clone_1 : [num_users=1] = call_function[target=torch.ops.aten.clone.default](args = (%x,), kwargs = {})
%scan_combine_graph_0 : [num_users=1] = get_attr[target=scan_combine_graph_0]
%scan : [num_users=1] = call_function[target=torch.ops.higher_order.scan](args = (%scan_combine_graph_0, [%clone], [%x], False, [%clone_1]), kwargs = {})
%getitem_1 : [num_users=1] = call_function[target=operator.getitem](args = (%scan, 1), kwargs = {})
return (getitem_1,)
nostrict¶
graph():
%lifted_tensor_0 : [num_users=1] = get_attr[target=lifted_tensor_0]
%x : [num_users=2] = placeholder[target=x]
%lift_fresh_copy : [num_users=1] = call_function[target=torch.ops.aten.lift_fresh_copy.default](args = (%lifted_tensor_0,), kwargs = {})
%detach_ : [num_users=1] = call_function[target=torch.ops.aten.detach_.default](args = (%lift_fresh_copy,), kwargs = {})
%clone : [num_users=1] = call_function[target=torch.ops.aten.clone.default](args = (%x,), kwargs = {})
%scan_combine_graph_0 : [num_users=1] = get_attr[target=scan_combine_graph_0]
%scan : [num_users=2] = call_function[target=torch.ops.higher_order.scan](args = (%scan_combine_graph_0, [%detach_], [%x], False, [%clone]), kwargs = {})
%getitem : [num_users=0] = call_function[target=operator.getitem](args = (%scan, 0), kwargs = {})
%getitem_1 : [num_users=1] = call_function[target=operator.getitem](args = (%scan, 1), kwargs = {})
return (getitem_1,)
nostrict-decall¶
graph():
%lifted_tensor_0 : [num_users=1] = get_attr[target=lifted_tensor_0]
%x : [num_users=2] = placeholder[target=x]
%clone : [num_users=1] = call_function[target=torch.ops.aten.clone.default](args = (%lifted_tensor_0,), kwargs = {})
%clone_1 : [num_users=1] = call_function[target=torch.ops.aten.clone.default](args = (%x,), kwargs = {})
%scan_combine_graph_0 : [num_users=1] = get_attr[target=scan_combine_graph_0]
%scan : [num_users=1] = call_function[target=torch.ops.higher_order.scan](args = (%scan_combine_graph_0, [%clone], [%x], False, [%clone_1]), kwargs = {})
%getitem_1 : [num_users=1] = call_function[target=operator.getitem](args = (%scan, 1), kwargs = {})
return (getitem_1,)
jit¶
- /home/xadupre/github/experimental-experiment/experimental_experiment/torch_interpreter/eval/model_cases.py:449: TracerWarning: torch.tensor results are registered as constants in the trace. You can safely ignore this warning if you use this function to create tensors out of constant variables that would be the same every time you call this function. In any other case, this might cause the trace to be incorrect.
z = torch.tensor([0], dtype=torch.float32)
FAILED
could not find kernel for HigherOrderOperator scan at dispatch key DispatchKey.??? (resolved from DispatchKey.???)
jit-decall¶
FAILED
could not find kernel for HigherOrderOperator scan at dispatch key DispatchKey.??? (resolved from DispatchKey.???)
tracing¶
FAILED
[CustomProxy(clone)] can only be of (<class 'torch.Tensor'>, <class 'int'>, <class 'torch.SymInt'>) but got (<class 'experimental_experiment.torch_interpreter.tracing.CustomProxy'>,)
ControlFlowScanCDistXY¶
forward¶
def forward(self, x, y):
carry, out = torch.ops.higher_order.scan(
ControlFlowScanCDistXY.dist,
[y],
[x],
# dim=0, # 01/31/2025, not supported anymore
reverse=False,
additional_inputs=[],
)
return out
strict¶
graph():
%x : [num_users=1] = placeholder[target=x]
%y : [num_users=1] = placeholder[target=y]
%scan_combine_graph_0 : [num_users=1] = get_attr[target=scan_combine_graph_0]
%scan : [num_users=2] = call_function[target=torch.ops.higher_order.scan](args = (%scan_combine_graph_0, [%y], [%x], False, []), kwargs = {})
%getitem : [num_users=0] = call_function[target=operator.getitem](args = (%scan, 0), kwargs = {})
%getitem_1 : [num_users=1] = call_function[target=operator.getitem](args = (%scan, 1), kwargs = {})
return (getitem_1,)
strict-decall¶
graph():
%x : [num_users=1] = placeholder[target=x]
%y : [num_users=1] = placeholder[target=y]
%scan_combine_graph_0 : [num_users=1] = get_attr[target=scan_combine_graph_0]
%scan : [num_users=1] = call_function[target=torch.ops.higher_order.scan](args = (%scan_combine_graph_0, [%y], [%x], False, []), kwargs = {})
%getitem_1 : [num_users=1] = call_function[target=operator.getitem](args = (%scan, 1), kwargs = {})
return (getitem_1,)
nostrict¶
graph():
%x : [num_users=1] = placeholder[target=x]
%y : [num_users=1] = placeholder[target=y]
%scan_combine_graph_0 : [num_users=1] = get_attr[target=scan_combine_graph_0]
%scan : [num_users=2] = call_function[target=torch.ops.higher_order.scan](args = (%scan_combine_graph_0, [%y], [%x], False, []), kwargs = {})
%getitem : [num_users=0] = call_function[target=operator.getitem](args = (%scan, 0), kwargs = {})
%getitem_1 : [num_users=1] = call_function[target=operator.getitem](args = (%scan, 1), kwargs = {})
return (getitem_1,)
nostrict-decall¶
graph():
%x : [num_users=1] = placeholder[target=x]
%y : [num_users=1] = placeholder[target=y]
%scan_combine_graph_0 : [num_users=1] = get_attr[target=scan_combine_graph_0]
%scan : [num_users=1] = call_function[target=torch.ops.higher_order.scan](args = (%scan_combine_graph_0, [%y], [%x], False, []), kwargs = {})
%getitem_1 : [num_users=1] = call_function[target=operator.getitem](args = (%scan, 1), kwargs = {})
return (getitem_1,)
jit¶
FAILED
could not find kernel for HigherOrderOperator scan at dispatch key DispatchKey.??? (resolved from DispatchKey.???)
jit-decall¶
FAILED
could not find kernel for HigherOrderOperator scan at dispatch key DispatchKey.??? (resolved from DispatchKey.???)
tracing¶
FAILED
Unable to symbolically trace HigherOrderOperators
CreateFromShape¶
forward¶
def forward(self, x):
y = torch.ones((x.shape[0], x.shape[1] + 1))
return y
strict¶
graph():
%x : [num_users=2] = placeholder[target=x]
%sym_size_int_2 : [num_users=1] = call_function[target=torch.ops.aten.sym_size.int](args = (%x, 0), kwargs = {})
%sym_size_int_3 : [num_users=1] = call_function[target=torch.ops.aten.sym_size.int](args = (%x, 1), kwargs = {})
%add : [num_users=1] = call_function[target=operator.add](args = (%sym_size_int_3, 1), kwargs = {})
%ones : [num_users=1] = call_function[target=torch.ops.aten.ones.default](args = ([%sym_size_int_2, %add],), kwargs = {device: cpu, pin_memory: False})
return (ones,)
strict-decall¶
graph():
%x : [num_users=2] = placeholder[target=x]
%sym_size_int_2 : [num_users=1] = call_function[target=torch.ops.aten.sym_size.int](args = (%x, 0), kwargs = {})
%sym_size_int_3 : [num_users=1] = call_function[target=torch.ops.aten.sym_size.int](args = (%x, 1), kwargs = {})
%add : [num_users=1] = call_function[target=operator.add](args = (%sym_size_int_3, 1), kwargs = {})
%full : [num_users=1] = call_function[target=torch.ops.aten.full.default](args = ([%sym_size_int_2, %add], 1), kwargs = {dtype: torch.float32, layout: torch.strided, device: cpu, pin_memory: False})
return (full,)
nostrict¶
graph():
%x : [num_users=2] = placeholder[target=x]
%sym_size_int_2 : [num_users=1] = call_function[target=torch.ops.aten.sym_size.int](args = (%x, 0), kwargs = {})
%sym_size_int_3 : [num_users=1] = call_function[target=torch.ops.aten.sym_size.int](args = (%x, 1), kwargs = {})
%add : [num_users=1] = call_function[target=operator.add](args = (%sym_size_int_3, 1), kwargs = {})
%ones : [num_users=1] = call_function[target=torch.ops.aten.ones.default](args = ([%sym_size_int_2, %add],), kwargs = {device: cpu, pin_memory: False})
return (ones,)
nostrict-decall¶
graph():
%x : [num_users=2] = placeholder[target=x]
%sym_size_int_2 : [num_users=1] = call_function[target=torch.ops.aten.sym_size.int](args = (%x, 0), kwargs = {})
%sym_size_int_3 : [num_users=1] = call_function[target=torch.ops.aten.sym_size.int](args = (%x, 1), kwargs = {})
%add : [num_users=1] = call_function[target=operator.add](args = (%sym_size_int_3, 1), kwargs = {})
%full : [num_users=1] = call_function[target=torch.ops.aten.full.default](args = ([%sym_size_int_2, %add], 1), kwargs = {dtype: torch.float32, layout: torch.strided, device: cpu, pin_memory: False})
return (full,)
jit¶
graph():
%lifted_tensor_5 : [num_users=1] = get_attr[target=lifted_tensor_5]
%x : [num_users=2] = placeholder[target=x]
%sym_size_int_2 : [num_users=1] = call_function[target=torch.ops.aten.sym_size.int](args = (%x, 0), kwargs = {})
%sym_size_int_3 : [num_users=1] = call_function[target=torch.ops.aten.sym_size.int](args = (%x, 1), kwargs = {})
%scalar_tensor : [num_users=1] = call_function[target=torch.ops.aten.scalar_tensor.default](args = (%sym_size_int_3,), kwargs = {dtype: torch.int64})
%add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%scalar_tensor, %lifted_tensor_5), kwargs = {})
%_to_copy : [num_users=1] = call_function[target=torch.ops.aten._to_copy.default](args = (%add,), kwargs = {dtype: torch.int32})
%_local_scalar_dense : [num_users=3] = call_function[target=torch.ops.aten._local_scalar_dense.default](args = (%_to_copy,), kwargs = {})
%sym_constrain_range_for_size_default : [num_users=0] = call_function[target=torch.ops.aten.sym_constrain_range_for_size.default](args = (%_local_scalar_dense,), kwargs = {})
%ge : [num_users=1] = call_function[target=operator.ge](args = (%_local_scalar_dense, 0), kwargs = {})
%_assert_scalar_default : [num_users=0] = call_function[target=torch.ops.aten._assert_scalar.default](args = (%ge, Runtime assertion failed for expression u0 >= 0 on node 'ge'), kwargs = {})
%ones : [num_users=1] = call_function[target=torch.ops.aten.ones.default](args = ([%sym_size_int_2, %_local_scalar_dense],), kwargs = {device: cpu, pin_memory: False})
return (ones,)
jit-decall¶
graph():
%lifted_tensor_5 : [num_users=1] = get_attr[target=lifted_tensor_5]
%x : [num_users=2] = placeholder[target=x]
%sym_size_int_2 : [num_users=1] = call_function[target=torch.ops.aten.sym_size.int](args = (%x, 0), kwargs = {})
%sym_size_int_3 : [num_users=1] = call_function[target=torch.ops.aten.sym_size.int](args = (%x, 1), kwargs = {})
%scalar_tensor : [num_users=1] = call_function[target=torch.ops.aten.scalar_tensor.default](args = (%sym_size_int_3,), kwargs = {dtype: torch.int64})
%add : [num_users=2] = call_function[target=torch.ops.aten.add.Tensor](args = (%scalar_tensor, %lifted_tensor_5), kwargs = {})
%_to_copy : [num_users=1] = call_function[target=torch.ops.aten._to_copy.default](args = (%add,), kwargs = {dtype: torch.int32})
%_assert_tensor_metadata : [num_users=0] = call_function[target=torch.ops.aten._assert_tensor_metadata.default](args = (%add, None, None, torch.int64), kwargs = {})
%_local_scalar_dense : [num_users=3] = call_function[target=torch.ops.aten._local_scalar_dense.default](args = (%_to_copy,), kwargs = {})
%sym_constrain_range_for_size_default : [num_users=0] = call_function[target=torch.ops.aten.sym_constrain_range_for_size.default](args = (%_local_scalar_dense,), kwargs = {})
%ge_3 : [num_users=1] = call_function[target=operator.ge](args = (%_local_scalar_dense, 0), kwargs = {})
%_assert_scalar_default : [num_users=0] = call_function[target=torch.ops.aten._assert_scalar.default](args = (%ge_3, Runtime assertion failed for expression u0 >= 0 on node 'ge_3'), kwargs = {})
%full : [num_users=1] = call_function[target=torch.ops.aten.full.default](args = ([%sym_size_int_2, %_local_scalar_dense], 1), kwargs = {dtype: torch.float32, layout: torch.strided, device: cpu, pin_memory: False})
return (full,)
tracing¶
FAILED
ones(): argument 'size' (position 1) must be tuple of ints, but found element of type CustomProxy at pos 0
CreateFromShapeThroughFunction¶
forward¶
def forward(self, x):
dy1 = CreateFromShapeThroughFunction.add_one(x.shape[1])
y = torch.ones((x.shape[0], dy1))
return y
strict¶
graph():
%x : [num_users=2] = placeholder[target=x]
%sym_size_int_2 : [num_users=1] = call_function[target=torch.ops.aten.sym_size.int](args = (%x, 0), kwargs = {})
%sym_size_int_3 : [num_users=1] = call_function[target=torch.ops.aten.sym_size.int](args = (%x, 1), kwargs = {})
%add : [num_users=1] = call_function[target=operator.add](args = (%sym_size_int_3, 1), kwargs = {})
%ones : [num_users=1] = call_function[target=torch.ops.aten.ones.default](args = ([%sym_size_int_2, %add],), kwargs = {device: cpu, pin_memory: False})
return (ones,)
strict-decall¶
graph():
%x : [num_users=2] = placeholder[target=x]
%sym_size_int_2 : [num_users=1] = call_function[target=torch.ops.aten.sym_size.int](args = (%x, 0), kwargs = {})
%sym_size_int_3 : [num_users=1] = call_function[target=torch.ops.aten.sym_size.int](args = (%x, 1), kwargs = {})
%add : [num_users=1] = call_function[target=operator.add](args = (%sym_size_int_3, 1), kwargs = {})
%full : [num_users=1] = call_function[target=torch.ops.aten.full.default](args = ([%sym_size_int_2, %add], 1), kwargs = {dtype: torch.float32, layout: torch.strided, device: cpu, pin_memory: False})
return (full,)
nostrict¶
graph():
%x : [num_users=2] = placeholder[target=x]
%sym_size_int_2 : [num_users=1] = call_function[target=torch.ops.aten.sym_size.int](args = (%x, 0), kwargs = {})
%sym_size_int_3 : [num_users=1] = call_function[target=torch.ops.aten.sym_size.int](args = (%x, 1), kwargs = {})
%add : [num_users=1] = call_function[target=operator.add](args = (%sym_size_int_3, 1), kwargs = {})
%ones : [num_users=1] = call_function[target=torch.ops.aten.ones.default](args = ([%sym_size_int_2, %add],), kwargs = {device: cpu, pin_memory: False})
return (ones,)
nostrict-decall¶
graph():
%x : [num_users=2] = placeholder[target=x]
%sym_size_int_2 : [num_users=1] = call_function[target=torch.ops.aten.sym_size.int](args = (%x, 0), kwargs = {})
%sym_size_int_3 : [num_users=1] = call_function[target=torch.ops.aten.sym_size.int](args = (%x, 1), kwargs = {})
%add : [num_users=1] = call_function[target=operator.add](args = (%sym_size_int_3, 1), kwargs = {})
%full : [num_users=1] = call_function[target=torch.ops.aten.full.default](args = ([%sym_size_int_2, %add], 1), kwargs = {dtype: torch.float32, layout: torch.strided, device: cpu, pin_memory: False})
return (full,)
jit¶
graph():
%lifted_tensor_6 : [num_users=1] = get_attr[target=lifted_tensor_6]
%x : [num_users=2] = placeholder[target=x]
%sym_size_int_2 : [num_users=1] = call_function[target=torch.ops.aten.sym_size.int](args = (%x, 0), kwargs = {})
%sym_size_int_3 : [num_users=1] = call_function[target=torch.ops.aten.sym_size.int](args = (%x, 1), kwargs = {})
%scalar_tensor : [num_users=1] = call_function[target=torch.ops.aten.scalar_tensor.default](args = (%sym_size_int_3,), kwargs = {dtype: torch.int64})
%add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%scalar_tensor, %lifted_tensor_6), kwargs = {})
%_to_copy : [num_users=1] = call_function[target=torch.ops.aten._to_copy.default](args = (%add,), kwargs = {dtype: torch.int32})
%_local_scalar_dense : [num_users=3] = call_function[target=torch.ops.aten._local_scalar_dense.default](args = (%_to_copy,), kwargs = {})
%sym_constrain_range_for_size_default : [num_users=0] = call_function[target=torch.ops.aten.sym_constrain_range_for_size.default](args = (%_local_scalar_dense,), kwargs = {})
%ge : [num_users=1] = call_function[target=operator.ge](args = (%_local_scalar_dense, 0), kwargs = {})
%_assert_scalar_default : [num_users=0] = call_function[target=torch.ops.aten._assert_scalar.default](args = (%ge, Runtime assertion failed for expression u0 >= 0 on node 'ge'), kwargs = {})
%ones : [num_users=1] = call_function[target=torch.ops.aten.ones.default](args = ([%sym_size_int_2, %_local_scalar_dense],), kwargs = {device: cpu, pin_memory: False})
return (ones,)
jit-decall¶
graph():
%lifted_tensor_6 : [num_users=1] = get_attr[target=lifted_tensor_6]
%x : [num_users=2] = placeholder[target=x]
%sym_size_int_2 : [num_users=1] = call_function[target=torch.ops.aten.sym_size.int](args = (%x, 0), kwargs = {})
%sym_size_int_3 : [num_users=1] = call_function[target=torch.ops.aten.sym_size.int](args = (%x, 1), kwargs = {})
%scalar_tensor : [num_users=1] = call_function[target=torch.ops.aten.scalar_tensor.default](args = (%sym_size_int_3,), kwargs = {dtype: torch.int64})
%add : [num_users=2] = call_function[target=torch.ops.aten.add.Tensor](args = (%scalar_tensor, %lifted_tensor_6), kwargs = {})
%_to_copy : [num_users=1] = call_function[target=torch.ops.aten._to_copy.default](args = (%add,), kwargs = {dtype: torch.int32})
%_assert_tensor_metadata : [num_users=0] = call_function[target=torch.ops.aten._assert_tensor_metadata.default](args = (%add, None, None, torch.int64), kwargs = {})
%_local_scalar_dense : [num_users=3] = call_function[target=torch.ops.aten._local_scalar_dense.default](args = (%_to_copy,), kwargs = {})
%sym_constrain_range_for_size_default : [num_users=0] = call_function[target=torch.ops.aten.sym_constrain_range_for_size.default](args = (%_local_scalar_dense,), kwargs = {})
%ge_3 : [num_users=1] = call_function[target=operator.ge](args = (%_local_scalar_dense, 0), kwargs = {})
%_assert_scalar_default : [num_users=0] = call_function[target=torch.ops.aten._assert_scalar.default](args = (%ge_3, Runtime assertion failed for expression u0 >= 0 on node 'ge_3'), kwargs = {})
%full : [num_users=1] = call_function[target=torch.ops.aten.full.default](args = ([%sym_size_int_2, %_local_scalar_dense], 1), kwargs = {dtype: torch.float32, layout: torch.strided, device: cpu, pin_memory: False})
return (full,)
tracing¶
FAILED
ones(): argument 'size' (position 1) must be tuple of ints, but found element of type CustomProxy at pos 0
CropLastDimensionWithTensorContent¶
forward¶
def forward(self, x, shape):
return x[..., : shape[0]]
strict¶
FAILED
When `dynamic_shapes` is specified as a dict, its top-level keys must be the arg names ['x', 'shape'] of `inputs`, but here they are ['x']. Alternatively, you could also ignore arg names entirely and specify `dynamic_shapes` as a list/tuple matching `inputs`. For more information about this error, see: https://pytorch.org/docs/main/generated/exportdb/index.html#dynamic-shapes-validation
strict-decall¶
FAILED
When `dynamic_shapes` is specified as a dict, its top-level keys must be the arg names ['x', 'shape'] of `inputs`, but here they are ['x']. Alternatively, you could also ignore arg names entirely and specify `dynamic_shapes` as a list/tuple matching `inputs`. For more information about this error, see: https://pytorch.org/docs/main/generated/exportdb/index.html#dynamic-shapes-validation
nostrict¶
FAILED
When `dynamic_shapes` is specified as a dict, its top-level keys must be the arg names ['x', 'shape'] of `inputs`, but here they are ['x']. Alternatively, you could also ignore arg names entirely and specify `dynamic_shapes` as a list/tuple matching `inputs`. For more information about this error, see: https://pytorch.org/docs/main/generated/exportdb/index.html#dynamic-shapes-validation
nostrict-decall¶
FAILED
When `dynamic_shapes` is specified as a dict, its top-level keys must be the arg names ['x', 'shape'] of `inputs`, but here they are ['x']. Alternatively, you could also ignore arg names entirely and specify `dynamic_shapes` as a list/tuple matching `inputs`. For more information about this error, see: https://pytorch.org/docs/main/generated/exportdb/index.html#dynamic-shapes-validation
jit¶
- def forward(self, arg0_1: “f32[s0, s1, s2]”, arg1_1: “i64[1]”):
# No stacktrace found for following nodes select: “i64[]” = torch.ops.aten.select.int(arg1_1, 0, 0); arg1_1 = None _to_copy: “i32[]” = torch.ops.aten._to_copy.default(select, dtype = torch.int32); select = None _local_scalar_dense: “Sym(u0)” = torch.ops.aten._local_scalar_dense.default(_to_copy); _to_copy = None slice_1 = torch.ops.aten.slice.Tensor(arg0_1, 2, 0, _local_scalar_dense); arg0_1 = _local_scalar_dense = slice_1 = None
FAILED
Could not guard on data-dependent expression u0 < 0 (unhinted: u0 < 0). (Size-like symbols: none)
Caused by: (_decomp/decompositions.py:734 in slice_forward)
For more information, run with TORCH_LOGS="dynamic"
For extended logs when we create symbols, also add TORCHDYNAMO_EXTENDED_DEBUG_CREATE_SYMBOL="u0"
If you suspect the guard was triggered from C++, add TORCHDYNAMO_EXTENDED_DEBUG_CPP=1
For more debugging help, see https://docs.google.com/document/d/1HSuTTVvYH1pTew89Rtpeu84Ht3nQEFTYhAX3Ypa_xJs/edit?usp=sharing
For C++ stack trace, run with TORCHDYNAMO_EXTENDED_DEBUG_CPP=1
The following call raised this error:
File "<string>", line 1, in <lambda>
While executing %slice_tensor : [num_users=1] = call_function[target=torch.ops.aten.slice.Tensor](args = (%x, 2, 0, %_local_scalar_dense_default, 1), kwargs = {})
GraphModule: class GraphModule(torch.nn.Module):
def forward(self, x, shape):
# No stacktrace found for following nodes
select_int = torch.ops.aten.select.int(shape, 0, 0); shape = None
_to_copy_default = torch.ops.aten._to_copy.default(select_int, dtype = torch.int32); select_int = None
_local_scalar_dense_default = torch.ops.aten._local_scalar_dense.default(_to_copy_default); _to_copy_default = None
slice_tensor = torch.ops.aten.slice.Tensor(x, 2, 0, _local_scalar_dense_default, 1); x = _local_scalar_dense_default = None
return slice_tensor
Original traceback:
None
jit-decall¶
- def forward(self, arg0_1: “f32[s0, s1, s2]”, arg1_1: “i64[1]”):
# No stacktrace found for following nodes select: “i64[]” = torch.ops.aten.select.int(arg1_1, 0, 0); arg1_1 = None _to_copy: “i32[]” = torch.ops.aten._to_copy.default(select, dtype = torch.int32); select = None _local_scalar_dense: “Sym(u0)” = torch.ops.aten._local_scalar_dense.default(_to_copy); _to_copy = None slice_1 = torch.ops.aten.slice.Tensor(arg0_1, 2, 0, _local_scalar_dense); arg0_1 = _local_scalar_dense = slice_1 = None
FAILED
Could not guard on data-dependent expression u0 < 0 (unhinted: u0 < 0). (Size-like symbols: none)
Caused by: (_decomp/decompositions.py:734 in slice_forward)
For more information, run with TORCH_LOGS="dynamic"
For extended logs when we create symbols, also add TORCHDYNAMO_EXTENDED_DEBUG_CREATE_SYMBOL="u0"
If you suspect the guard was triggered from C++, add TORCHDYNAMO_EXTENDED_DEBUG_CPP=1
For more debugging help, see https://docs.google.com/document/d/1HSuTTVvYH1pTew89Rtpeu84Ht3nQEFTYhAX3Ypa_xJs/edit?usp=sharing
For C++ stack trace, run with TORCHDYNAMO_EXTENDED_DEBUG_CPP=1
The following call raised this error:
File "<string>", line 1, in <lambda>
While executing %slice_tensor : [num_users=1] = call_function[target=torch.ops.aten.slice.Tensor](args = (%x, 2, 0, %_local_scalar_dense_default, 1), kwargs = {})
GraphModule: class GraphModule(torch.nn.Module):
def forward(self, x, shape):
# No stacktrace found for following nodes
select_int = torch.ops.aten.select.int(shape, 0, 0); shape = None
_to_copy_default = torch.ops.aten._to_copy.default(select_int, dtype = torch.int32); select_int = None
_local_scalar_dense_default = torch.ops.aten._local_scalar_dense.default(_to_copy_default); _to_copy_default = None
slice_tensor = torch.ops.aten.slice.Tensor(x, 2, 0, _local_scalar_dense_default, 1); x = _local_scalar_dense_default = None
return slice_tensor
Original traceback:
None
tracing¶
graph():
%x : [num_users=1] = placeholder[target=x]
%shape : [num_users=1] = placeholder[target=shape]
%getitem : [num_users=1] = call_function[target=operator.getitem](args = (%shape, 0), kwargs = {})
%getitem_1 : [num_users=1] = call_function[target=operator.getitem](args = (%x, (Ellipsis, slice(None, getitem, None))), kwargs = {})
return getitem_1
CropLastDimensionWithTensorShape¶
forward¶
def forward(self, x, y):
return x[..., : y.shape[0]]
strict¶
graph():
%x : [num_users=1] = placeholder[target=x]
%y : [num_users=1] = placeholder[target=y]
%sym_size_int_2 : [num_users=1] = call_function[target=torch.ops.aten.sym_size.int](args = (%y, 0), kwargs = {})
%slice_1 : [num_users=1] = call_function[target=torch.ops.aten.slice.Tensor](args = (%x, 2, 0, %sym_size_int_2), kwargs = {})
return (slice_1,)
strict-decall¶
graph():
%x : [num_users=1] = placeholder[target=x]
%y : [num_users=1] = placeholder[target=y]
%sym_size_int_4 : [num_users=1] = call_function[target=torch.ops.aten.sym_size.int](args = (%y, 0), kwargs = {})
%slice_1 : [num_users=1] = call_function[target=torch.ops.aten.slice.Tensor](args = (%x, 2, 0, %sym_size_int_4), kwargs = {})
return (slice_1,)
nostrict¶
graph():
%x : [num_users=1] = placeholder[target=x]
%y : [num_users=1] = placeholder[target=y]
%sym_size_int_2 : [num_users=1] = call_function[target=torch.ops.aten.sym_size.int](args = (%y, 0), kwargs = {})
%slice_1 : [num_users=1] = call_function[target=torch.ops.aten.slice.Tensor](args = (%x, 2, 0, %sym_size_int_2), kwargs = {})
return (slice_1,)
nostrict-decall¶
graph():
%x : [num_users=1] = placeholder[target=x]
%y : [num_users=1] = placeholder[target=y]
%sym_size_int_4 : [num_users=1] = call_function[target=torch.ops.aten.sym_size.int](args = (%y, 0), kwargs = {})
%slice_1 : [num_users=1] = call_function[target=torch.ops.aten.slice.Tensor](args = (%x, 2, 0, %sym_size_int_4), kwargs = {})
return (slice_1,)
jit¶
graph():
%x : [num_users=1] = placeholder[target=x]
%y : [num_users=1] = placeholder[target=y]
%sym_size_int_4 : [num_users=1] = call_function[target=torch.ops.aten.sym_size.int](args = (%y, 0), kwargs = {})
%slice_1 : [num_users=1] = call_function[target=torch.ops.aten.slice.Tensor](args = (%x, 2, 0, %sym_size_int_4), kwargs = {})
return (slice_1,)
jit-decall¶
graph():
%x : [num_users=1] = placeholder[target=x]
%y : [num_users=1] = placeholder[target=y]
%sym_size_int_7 : [num_users=1] = call_function[target=torch.ops.aten.sym_size.int](args = (%y, 0), kwargs = {})
%slice_1 : [num_users=1] = call_function[target=torch.ops.aten.slice.Tensor](args = (%x, 2, 0, %sym_size_int_7), kwargs = {})
return (slice_1,)
tracing¶
graph():
%x : [num_users=1] = placeholder[target=x]
%y : [num_users=1] = placeholder[target=y]
%getattr_1 : [num_users=1] = call_function[target=builtins.getattr](args = (%y, shape), kwargs = {})
%getitem : [num_users=1] = call_function[target=operator.getitem](args = (%getattr_1, 0), kwargs = {})
%getitem_1 : [num_users=1] = call_function[target=operator.getitem](args = (%x, (Ellipsis, slice(None, getitem, None))), kwargs = {})
return getitem_1
InplaceAdd¶
forward¶
def forward(self, x):
x += self.bias
return x
strict¶
graph():
%bias : [num_users=1] = get_attr[target=bias]
%x : [num_users=1] = placeholder[target=x]
%add_ : [num_users=1] = call_function[target=torch.ops.aten.add_.Tensor](args = (%x, %bias), kwargs = {})
return (add_,)
strict-decall¶
graph():
%bias : [num_users=1] = get_attr[target=bias]
%x : [num_users=2] = placeholder[target=x]
%add_3 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%x, %bias), kwargs = {})
%copy__default : [num_users=1] = call_function[target=torch.ops.aten.copy_.default](args = (%x, %add_3), kwargs = {})
return (copy__default,)
nostrict¶
graph():
%bias : [num_users=1] = get_attr[target=bias]
%x : [num_users=1] = placeholder[target=x]
%add_ : [num_users=1] = call_function[target=torch.ops.aten.add_.Tensor](args = (%x, %bias), kwargs = {})
return (add_,)
nostrict-decall¶
graph():
%bias : [num_users=1] = get_attr[target=bias]
%x : [num_users=2] = placeholder[target=x]
%add_3 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%x, %bias), kwargs = {})
%copy__default : [num_users=1] = call_function[target=torch.ops.aten.copy_.default](args = (%x, %add_3), kwargs = {})
return (copy__default,)
jit¶
graph():
%lifted_tensor_3 : [num_users=1] = get_attr[target=lifted_tensor_3]
%x : [num_users=1] = placeholder[target=x]
%add_ : [num_users=1] = call_function[target=torch.ops.aten.add_.Tensor](args = (%x, %lifted_tensor_3), kwargs = {})
return (add_,)
jit-decall¶
graph():
%lifted_tensor_3 : [num_users=1] = get_attr[target=lifted_tensor_3]
%x : [num_users=2] = placeholder[target=x]
%add_3 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%x, %lifted_tensor_3), kwargs = {})
%copy__default : [num_users=1] = call_function[target=torch.ops.aten.copy_.default](args = (%x, %add_3), kwargs = {})
return (copy__default,)
tracing¶
graph():
%x : [num_users=1] = placeholder[target=x]
%bias : [num_users=1] = get_attr[target=bias]
%add : [num_users=1] = call_function[target=operator.add](args = (%x, %bias), kwargs = {})
return add
InplaceAdd2¶
forward¶
def forward(self, x):
x.add_(self.bias)
return x
strict¶
graph():
%bias : [num_users=1] = get_attr[target=bias]
%x : [num_users=1] = placeholder[target=x]
%add_ : [num_users=1] = call_function[target=torch.ops.aten.add_.Tensor](args = (%x, %bias), kwargs = {})
return (add_,)
strict-decall¶
graph():
%bias : [num_users=1] = get_attr[target=bias]
%x : [num_users=2] = placeholder[target=x]
%add_3 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%x, %bias), kwargs = {})
%copy__default : [num_users=1] = call_function[target=torch.ops.aten.copy_.default](args = (%x, %add_3), kwargs = {})
return (copy__default,)
nostrict¶
graph():
%bias : [num_users=1] = get_attr[target=bias]
%x : [num_users=1] = placeholder[target=x]
%add_ : [num_users=1] = call_function[target=torch.ops.aten.add_.Tensor](args = (%x, %bias), kwargs = {})
return (add_,)
nostrict-decall¶
graph():
%bias : [num_users=1] = get_attr[target=bias]
%x : [num_users=2] = placeholder[target=x]
%add_3 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%x, %bias), kwargs = {})
%copy__default : [num_users=1] = call_function[target=torch.ops.aten.copy_.default](args = (%x, %add_3), kwargs = {})
return (copy__default,)
jit¶
graph():
%lifted_tensor_3 : [num_users=1] = get_attr[target=lifted_tensor_3]
%x : [num_users=1] = placeholder[target=x]
%add_ : [num_users=1] = call_function[target=torch.ops.aten.add_.Tensor](args = (%x, %lifted_tensor_3), kwargs = {})
return (add_,)
jit-decall¶
graph():
%lifted_tensor_3 : [num_users=1] = get_attr[target=lifted_tensor_3]
%x : [num_users=2] = placeholder[target=x]
%add_3 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%x, %lifted_tensor_3), kwargs = {})
%copy__default : [num_users=1] = call_function[target=torch.ops.aten.copy_.default](args = (%x, %add_3), kwargs = {})
return (copy__default,)
tracing¶
graph():
%x : [num_users=1] = placeholder[target=x]
%bias : [num_users=1] = get_attr[target=bias]
%add_ : [num_users=1] = call_method[target=add_](args = (%x, %bias), kwargs = {})
return add_
InplaceAdd_Mul¶
forward¶
def forward(self, x):
x.add_(self.bias)
return x * 2
strict¶
graph():
%bias : [num_users=1] = get_attr[target=bias]
%x : [num_users=1] = placeholder[target=x]
%add_ : [num_users=1] = call_function[target=torch.ops.aten.add_.Tensor](args = (%x, %bias), kwargs = {})
%mul : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%add_, 2), kwargs = {})
return (mul,)
strict-decall¶
graph():
%bias : [num_users=1] = get_attr[target=bias]
%x : [num_users=2] = placeholder[target=x]
%add_3 : [num_users=2] = call_function[target=torch.ops.aten.add.Tensor](args = (%x, %bias), kwargs = {})
%mul_4 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%add_3, 2), kwargs = {})
%copy__default : [num_users=0] = call_function[target=torch.ops.aten.copy_.default](args = (%x, %add_3), kwargs = {})
return (mul_4,)
nostrict¶
graph():
%bias : [num_users=1] = get_attr[target=bias]
%x : [num_users=1] = placeholder[target=x]
%add_ : [num_users=1] = call_function[target=torch.ops.aten.add_.Tensor](args = (%x, %bias), kwargs = {})
%mul : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%add_, 2), kwargs = {})
return (mul,)
nostrict-decall¶
graph():
%bias : [num_users=1] = get_attr[target=bias]
%x : [num_users=2] = placeholder[target=x]
%add_3 : [num_users=2] = call_function[target=torch.ops.aten.add.Tensor](args = (%x, %bias), kwargs = {})
%mul_4 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%add_3, 2), kwargs = {})
%copy__default : [num_users=0] = call_function[target=torch.ops.aten.copy_.default](args = (%x, %add_3), kwargs = {})
return (mul_4,)
jit¶
graph():
%lifted_tensor_2 : [num_users=1] = get_attr[target=lifted_tensor_2]
%lifted_tensor_4 : [num_users=1] = get_attr[target=lifted_tensor_4]
%x_1 : [num_users=1] = placeholder[target=x_1]
%add_ : [num_users=1] = call_function[target=torch.ops.aten.add_.Tensor](args = (%x_1, %lifted_tensor_4), kwargs = {})
%mul : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%add_, %lifted_tensor_2), kwargs = {})
return (mul,)
jit-decall¶
graph():
%lifted_tensor_2 : [num_users=1] = get_attr[target=lifted_tensor_2]
%lifted_tensor_4 : [num_users=1] = get_attr[target=lifted_tensor_4]
%x_1 : [num_users=2] = placeholder[target=x_1]
%add_3 : [num_users=2] = call_function[target=torch.ops.aten.add.Tensor](args = (%x_1, %lifted_tensor_4), kwargs = {})
%mul_4 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%add_3, %lifted_tensor_2), kwargs = {})
%copy__default : [num_users=0] = call_function[target=torch.ops.aten.copy_.default](args = (%x_1, %add_3), kwargs = {})
return (mul_4,)
tracing¶
graph():
%x : [num_users=1] = placeholder[target=x]
%bias : [num_users=1] = get_attr[target=bias]
%add_ : [num_users=1] = call_method[target=add_](args = (%x, %bias), kwargs = {})
%mul : [num_users=1] = call_function[target=operator.mul](args = (%add_, 2), kwargs = {})
return mul
InplaceCloneAdd¶
forward¶
def forward(self, x):
x = x.clone()
x.add_(self.bias)
return x
strict¶
graph():
%bias : [num_users=1] = get_attr[target=bias]
%x : [num_users=1] = placeholder[target=x]
%clone : [num_users=1] = call_function[target=torch.ops.aten.clone.default](args = (%x,), kwargs = {})
%add_ : [num_users=1] = call_function[target=torch.ops.aten.add_.Tensor](args = (%clone, %bias), kwargs = {})
return (add_,)
strict-decall¶
graph():
%bias : [num_users=1] = get_attr[target=bias]
%x : [num_users=1] = placeholder[target=x]
%clone : [num_users=1] = call_function[target=torch.ops.aten.clone.default](args = (%x,), kwargs = {})
%add_6 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%clone, %bias), kwargs = {})
return (add_6,)
nostrict¶
graph():
%bias : [num_users=1] = get_attr[target=bias]
%x : [num_users=1] = placeholder[target=x]
%clone : [num_users=1] = call_function[target=torch.ops.aten.clone.default](args = (%x,), kwargs = {})
%add_ : [num_users=1] = call_function[target=torch.ops.aten.add_.Tensor](args = (%clone, %bias), kwargs = {})
return (add_,)
nostrict-decall¶
graph():
%bias : [num_users=1] = get_attr[target=bias]
%x : [num_users=1] = placeholder[target=x]
%clone : [num_users=1] = call_function[target=torch.ops.aten.clone.default](args = (%x,), kwargs = {})
%add_6 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%clone, %bias), kwargs = {})
return (add_6,)
jit¶
graph():
%lifted_tensor_3 : [num_users=1] = get_attr[target=lifted_tensor_3]
%x_1 : [num_users=1] = placeholder[target=x_1]
%clone : [num_users=1] = call_function[target=torch.ops.aten.clone.default](args = (%x_1,), kwargs = {})
%add_ : [num_users=1] = call_function[target=torch.ops.aten.add_.Tensor](args = (%clone, %lifted_tensor_3), kwargs = {})
return (add_,)
jit-decall¶
graph():
%lifted_tensor_3 : [num_users=1] = get_attr[target=lifted_tensor_3]
%x_1 : [num_users=1] = placeholder[target=x_1]
%clone : [num_users=1] = call_function[target=torch.ops.aten.clone.default](args = (%x_1,), kwargs = {})
%add_6 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%clone, %lifted_tensor_3), kwargs = {})
return (add_6,)
tracing¶
graph():
%x : [num_users=1] = placeholder[target=x]
%clone : [num_users=1] = call_method[target=clone](args = (%x,), kwargs = {})
%bias : [num_users=1] = get_attr[target=bias]
%add_ : [num_users=1] = call_method[target=add_](args = (%clone, %bias), kwargs = {})
return add_
InplaceSetItemEllipsis_1¶
forward¶
def forward(self, index, update):
copy = self.params.clone()
copy[..., index] = update
return copy
strict¶
FAILED
When `dynamic_shapes` is specified as a dict, its top-level keys must be the arg names ['index', 'update'] of `inputs`, but here they are ['x']. Alternatively, you could also ignore arg names entirely and specify `dynamic_shapes` as a list/tuple matching `inputs`. For more information about this error, see: https://pytorch.org/docs/main/generated/exportdb/index.html#dynamic-shapes-validation
strict-decall¶
FAILED
When `dynamic_shapes` is specified as a dict, its top-level keys must be the arg names ['index', 'update'] of `inputs`, but here they are ['x']. Alternatively, you could also ignore arg names entirely and specify `dynamic_shapes` as a list/tuple matching `inputs`. For more information about this error, see: https://pytorch.org/docs/main/generated/exportdb/index.html#dynamic-shapes-validation
nostrict¶
FAILED
When `dynamic_shapes` is specified as a dict, its top-level keys must be the arg names ['index', 'update'] of `inputs`, but here they are ['x']. Alternatively, you could also ignore arg names entirely and specify `dynamic_shapes` as a list/tuple matching `inputs`. For more information about this error, see: https://pytorch.org/docs/main/generated/exportdb/index.html#dynamic-shapes-validation
nostrict-decall¶
FAILED
When `dynamic_shapes` is specified as a dict, its top-level keys must be the arg names ['index', 'update'] of `inputs`, but here they are ['x']. Alternatively, you could also ignore arg names entirely and specify `dynamic_shapes` as a list/tuple matching `inputs`. For more information about this error, see: https://pytorch.org/docs/main/generated/exportdb/index.html#dynamic-shapes-validation
jit¶
graph():
%lifted_tensor_5 : [num_users=1] = get_attr[target=lifted_tensor_5]
%index : [num_users=1] = placeholder[target=index]
%update : [num_users=1] = placeholder[target=update]
%clone : [num_users=1] = call_function[target=torch.ops.aten.clone.default](args = (%lifted_tensor_5,), kwargs = {})
%index_put_ : [num_users=1] = call_function[target=torch.ops.aten.index_put_.default](args = (%clone, [None, None, %index], %update), kwargs = {})
return (index_put_,)
jit-decall¶
graph():
%lifted_tensor_5 : [num_users=1] = get_attr[target=lifted_tensor_5]
%index : [num_users=1] = placeholder[target=index]
%update : [num_users=1] = placeholder[target=update]
%clone : [num_users=1] = call_function[target=torch.ops.aten.clone.default](args = (%lifted_tensor_5,), kwargs = {})
%index_put : [num_users=1] = call_function[target=torch.ops.aten.index_put.default](args = (%clone, [None, None, %index], %update), kwargs = {})
return (index_put,)
tracing¶
graph():
%index : [num_users=1] = placeholder[target=index]
%update : [num_users=1] = placeholder[target=update]
%_tensor_constant0 : [num_users=1] = get_attr[target=_tensor_constant0]
%setitem : [num_users=1] = call_function[target=operator.setitem](args = (%_tensor_constant0, (Ellipsis, %index), %update), kwargs = {})
return setitem
InplaceSetItemEllipsis_2¶
forward¶
def forward(self, index, update):
copy = self.params.clone()
copy[..., index] = update
return copy
strict¶
FAILED
When `dynamic_shapes` is specified as a dict, its top-level keys must be the arg names ['index', 'update'] of `inputs`, but here they are ['x']. Alternatively, you could also ignore arg names entirely and specify `dynamic_shapes` as a list/tuple matching `inputs`. For more information about this error, see: https://pytorch.org/docs/main/generated/exportdb/index.html#dynamic-shapes-validation
strict-decall¶
FAILED
When `dynamic_shapes` is specified as a dict, its top-level keys must be the arg names ['index', 'update'] of `inputs`, but here they are ['x']. Alternatively, you could also ignore arg names entirely and specify `dynamic_shapes` as a list/tuple matching `inputs`. For more information about this error, see: https://pytorch.org/docs/main/generated/exportdb/index.html#dynamic-shapes-validation
nostrict¶
FAILED
When `dynamic_shapes` is specified as a dict, its top-level keys must be the arg names ['index', 'update'] of `inputs`, but here they are ['x']. Alternatively, you could also ignore arg names entirely and specify `dynamic_shapes` as a list/tuple matching `inputs`. For more information about this error, see: https://pytorch.org/docs/main/generated/exportdb/index.html#dynamic-shapes-validation
nostrict-decall¶
FAILED
When `dynamic_shapes` is specified as a dict, its top-level keys must be the arg names ['index', 'update'] of `inputs`, but here they are ['x']. Alternatively, you could also ignore arg names entirely and specify `dynamic_shapes` as a list/tuple matching `inputs`. For more information about this error, see: https://pytorch.org/docs/main/generated/exportdb/index.html#dynamic-shapes-validation
jit¶
graph():
%lifted_tensor_5 : [num_users=1] = get_attr[target=lifted_tensor_5]
%index : [num_users=1] = placeholder[target=index]
%update : [num_users=1] = placeholder[target=update]
%clone : [num_users=1] = call_function[target=torch.ops.aten.clone.default](args = (%lifted_tensor_5,), kwargs = {})
%index_put_ : [num_users=1] = call_function[target=torch.ops.aten.index_put_.default](args = (%clone, [None, None, %index], %update), kwargs = {})
return (index_put_,)
jit-decall¶
graph():
%lifted_tensor_5 : [num_users=1] = get_attr[target=lifted_tensor_5]
%index : [num_users=1] = placeholder[target=index]
%update : [num_users=1] = placeholder[target=update]
%clone : [num_users=1] = call_function[target=torch.ops.aten.clone.default](args = (%lifted_tensor_5,), kwargs = {})
%index_put : [num_users=1] = call_function[target=torch.ops.aten.index_put.default](args = (%clone, [None, None, %index], %update), kwargs = {})
return (index_put,)
tracing¶
graph():
%index : [num_users=1] = placeholder[target=index]
%update : [num_users=1] = placeholder[target=update]
%_tensor_constant0 : [num_users=1] = get_attr[target=_tensor_constant0]
%setitem : [num_users=1] = call_function[target=operator.setitem](args = (%_tensor_constant0, (Ellipsis, %index), %update), kwargs = {})
return setitem
InplaceSetItemMask¶
forward¶
def forward(self, x):
mask = x.to(bool)
x[mask] = 2
return x
strict¶
graph():
%lifted_tensor_0 : [num_users=1] = get_attr[target=lifted_tensor_0]
%x : [num_users=2] = placeholder[target=x]
%to : [num_users=1] = call_function[target=torch.ops.aten.to.dtype](args = (%x, torch.bool), kwargs = {})
%lift_fresh_copy : [num_users=1] = call_function[target=torch.ops.aten.lift_fresh_copy.default](args = (%lifted_tensor_0,), kwargs = {})
%index_put_ : [num_users=1] = call_function[target=torch.ops.aten.index_put_.default](args = (%x, [%to], %lift_fresh_copy), kwargs = {})
return (index_put_,)
strict-decall¶
graph():
%lifted_tensor_0 : [num_users=1] = get_attr[target=lifted_tensor_0]
%x : [num_users=4] = placeholder[target=x]
%_to_copy : [num_users=1] = call_function[target=torch.ops.aten._to_copy.default](args = (%x,), kwargs = {dtype: torch.bool})
%_assert_tensor_metadata : [num_users=0] = call_function[target=torch.ops.aten._assert_tensor_metadata.default](args = (%x, None, None, torch.float32), kwargs = {})
%clone : [num_users=1] = call_function[target=torch.ops.aten.clone.default](args = (%lifted_tensor_0,), kwargs = {})
%index_put : [num_users=1] = call_function[target=torch.ops.aten.index_put.default](args = (%x, [%_to_copy], %clone), kwargs = {})
%copy__default : [num_users=1] = call_function[target=torch.ops.aten.copy_.default](args = (%x, %index_put), kwargs = {})
return (copy__default,)
nostrict¶
graph():
%lifted_tensor_0 : [num_users=1] = get_attr[target=lifted_tensor_0]
%x : [num_users=2] = placeholder[target=x]
%to : [num_users=1] = call_function[target=torch.ops.aten.to.dtype](args = (%x, torch.bool), kwargs = {})
%lift_fresh_copy : [num_users=1] = call_function[target=torch.ops.aten.lift_fresh_copy.default](args = (%lifted_tensor_0,), kwargs = {})
%index_put_ : [num_users=1] = call_function[target=torch.ops.aten.index_put_.default](args = (%x, [%to], %lift_fresh_copy), kwargs = {})
return (index_put_,)
nostrict-decall¶
graph():
%lifted_tensor_0 : [num_users=1] = get_attr[target=lifted_tensor_0]
%x : [num_users=4] = placeholder[target=x]
%_to_copy : [num_users=1] = call_function[target=torch.ops.aten._to_copy.default](args = (%x,), kwargs = {dtype: torch.bool})
%_assert_tensor_metadata : [num_users=0] = call_function[target=torch.ops.aten._assert_tensor_metadata.default](args = (%x, None, None, torch.float32), kwargs = {})
%clone : [num_users=1] = call_function[target=torch.ops.aten.clone.default](args = (%lifted_tensor_0,), kwargs = {})
%index_put : [num_users=1] = call_function[target=torch.ops.aten.index_put.default](args = (%x, [%_to_copy], %clone), kwargs = {})
%copy__default : [num_users=1] = call_function[target=torch.ops.aten.copy_.default](args = (%x, %index_put), kwargs = {})
return (copy__default,)
jit¶
graph():
%lifted_tensor_2 : [num_users=1] = get_attr[target=lifted_tensor_2]
%x : [num_users=2] = placeholder[target=x]
%to : [num_users=1] = call_function[target=torch.ops.aten.to.dtype](args = (%x, torch.bool), kwargs = {})
%index_put_ : [num_users=1] = call_function[target=torch.ops.aten.index_put_.default](args = (%x, [%to], %lifted_tensor_2), kwargs = {})
return (index_put_,)
jit-decall¶
graph():
%lifted_tensor_2 : [num_users=1] = get_attr[target=lifted_tensor_2]
%x : [num_users=4] = placeholder[target=x]
%_to_copy : [num_users=1] = call_function[target=torch.ops.aten._to_copy.default](args = (%x,), kwargs = {dtype: torch.bool})
%_assert_tensor_metadata : [num_users=0] = call_function[target=torch.ops.aten._assert_tensor_metadata.default](args = (%x, None, None, torch.float32), kwargs = {})
%index_put : [num_users=1] = call_function[target=torch.ops.aten.index_put.default](args = (%x, [%_to_copy], %lifted_tensor_2), kwargs = {})
%copy__default : [num_users=1] = call_function[target=torch.ops.aten.copy_.default](args = (%x, %index_put), kwargs = {})
return (copy__default,)
tracing¶
graph():
%x : [num_users=2] = placeholder[target=x]
%to : [num_users=1] = call_method[target=to](args = (%x, torch.bool), kwargs = {})
%setitem : [num_users=1] = call_function[target=operator.setitem](args = (%x, %to, 2), kwargs = {})
return setitem
InplaceSetItemSquare¶
forward¶
def forward(self, x):
x[:2, :3] = 1
return x
strict¶
graph():
%lifted_tensor_0 : [num_users=1] = get_attr[target=lifted_tensor_0]
%x : [num_users=2] = placeholder[target=x]
%lift_fresh_copy : [num_users=1] = call_function[target=torch.ops.aten.lift_fresh_copy.default](args = (%lifted_tensor_0,), kwargs = {})
%slice_1 : [num_users=1] = call_function[target=torch.ops.aten.slice.Tensor](args = (%x, 0, 0, 2), kwargs = {})
%slice_2 : [num_users=1] = call_function[target=torch.ops.aten.slice.Tensor](args = (%slice_1, 1, 0, 3), kwargs = {})
%fill_ : [num_users=0] = call_function[target=torch.ops.aten.fill_.Tensor](args = (%slice_2, %lift_fresh_copy), kwargs = {})
return (x,)
strict-decall¶
graph():
%lifted_tensor_0 : [num_users=1] = get_attr[target=lifted_tensor_0]
%x : [num_users=4] = placeholder[target=x]
%clone : [num_users=1] = call_function[target=torch.ops.aten.clone.default](args = (%lifted_tensor_0,), kwargs = {})
%slice_1 : [num_users=1] = call_function[target=torch.ops.aten.slice.Tensor](args = (%x, 0, 0, 2), kwargs = {})
%slice_2 : [num_users=1] = call_function[target=torch.ops.aten.slice.Tensor](args = (%slice_1, 1, 0, 3), kwargs = {})
%copy : [num_users=1] = call_function[target=torch.ops.aten.copy.default](args = (%slice_2, %clone), kwargs = {})
%slice_3 : [num_users=1] = call_function[target=torch.ops.aten.slice.Tensor](args = (%x, 0, 0, 2), kwargs = {})
%slice_scatter : [num_users=1] = call_function[target=torch.ops.aten.slice_scatter.default](args = (%slice_3, %copy, 1, 0, 3), kwargs = {})
%slice_scatter_1 : [num_users=1] = call_function[target=torch.ops.aten.slice_scatter.default](args = (%x, %slice_scatter, 0, 0, 2), kwargs = {})
%copy__default : [num_users=1] = call_function[target=torch.ops.aten.copy_.default](args = (%x, %slice_scatter_1), kwargs = {})
return (copy__default,)
nostrict¶
FAILED
Constraints violated (batch)! For more information, run with TORCH_LOGS="+dynamic".
- Not all values of batch = L['args'][0][0].size()[0] in the specified range satisfy the generated guard L['args'][0][0].size()[0] != 2.
nostrict-decall¶
FAILED
Constraints violated (batch)! For more information, run with TORCH_LOGS="+dynamic".
- Not all values of batch = L['args'][0][0].size()[0] in the specified range satisfy the generated guard L['args'][0][0].size()[0] != 2.
jit¶
graph():
%lifted_tensor_2 : [num_users=1] = get_attr[target=lifted_tensor_2]
%x : [num_users=2] = placeholder[target=x]
%slice_1 : [num_users=1] = call_function[target=torch.ops.aten.slice.Tensor](args = (%x, 0, 0, 2), kwargs = {})
%slice_2 : [num_users=1] = call_function[target=torch.ops.aten.slice.Tensor](args = (%slice_1, 1, 0, 3), kwargs = {})
%fill_ : [num_users=0] = call_function[target=torch.ops.aten.fill_.Tensor](args = (%slice_2, %lifted_tensor_2), kwargs = {})
return (x,)
jit-decall¶
graph():
%lifted_tensor_2 : [num_users=1] = get_attr[target=lifted_tensor_2]
%x : [num_users=4] = placeholder[target=x]
%slice_1 : [num_users=1] = call_function[target=torch.ops.aten.slice.Tensor](args = (%x, 0, 0, 2), kwargs = {})
%slice_2 : [num_users=1] = call_function[target=torch.ops.aten.slice.Tensor](args = (%slice_1, 1, 0, 3), kwargs = {})
%copy : [num_users=1] = call_function[target=torch.ops.aten.copy.default](args = (%slice_2, %lifted_tensor_2), kwargs = {})
%slice_3 : [num_users=1] = call_function[target=torch.ops.aten.slice.Tensor](args = (%x, 0, 0, 2), kwargs = {})
%slice_scatter : [num_users=1] = call_function[target=torch.ops.aten.slice_scatter.default](args = (%slice_3, %copy, 1, 0, 3), kwargs = {})
%slice_scatter_1 : [num_users=1] = call_function[target=torch.ops.aten.slice_scatter.default](args = (%x, %slice_scatter, 0, 0, 2), kwargs = {})
%copy__default : [num_users=1] = call_function[target=torch.ops.aten.copy_.default](args = (%x, %slice_scatter_1), kwargs = {})
return (copy__default,)
tracing¶
graph():
%x : [num_users=1] = placeholder[target=x]
%setitem : [num_users=1] = call_function[target=operator.setitem](args = (%x, (slice(None, 2, None), slice(None, 3, None)), 1), kwargs = {})
return setitem
InplaceSetItemSquareAdd¶
forward¶
def forward(self, x):
x[:2, :3] = 1
return x + 2
strict¶
graph():
%lifted_tensor_0 : [num_users=1] = get_attr[target=lifted_tensor_0]
%x : [num_users=2] = placeholder[target=x]
%lift_fresh_copy : [num_users=1] = call_function[target=torch.ops.aten.lift_fresh_copy.default](args = (%lifted_tensor_0,), kwargs = {})
%slice_1 : [num_users=1] = call_function[target=torch.ops.aten.slice.Tensor](args = (%x, 0, 0, 2), kwargs = {})
%slice_2 : [num_users=1] = call_function[target=torch.ops.aten.slice.Tensor](args = (%slice_1, 1, 0, 3), kwargs = {})
%fill_ : [num_users=0] = call_function[target=torch.ops.aten.fill_.Tensor](args = (%slice_2, %lift_fresh_copy), kwargs = {})
%add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%x, 2), kwargs = {})
return (add,)
strict-decall¶
graph():
%lifted_tensor_0 : [num_users=1] = get_attr[target=lifted_tensor_0]
%x : [num_users=4] = placeholder[target=x]
%clone : [num_users=1] = call_function[target=torch.ops.aten.clone.default](args = (%lifted_tensor_0,), kwargs = {})
%slice_1 : [num_users=1] = call_function[target=torch.ops.aten.slice.Tensor](args = (%x, 0, 0, 2), kwargs = {})
%slice_2 : [num_users=1] = call_function[target=torch.ops.aten.slice.Tensor](args = (%slice_1, 1, 0, 3), kwargs = {})
%copy : [num_users=1] = call_function[target=torch.ops.aten.copy.default](args = (%slice_2, %clone), kwargs = {})
%slice_3 : [num_users=1] = call_function[target=torch.ops.aten.slice.Tensor](args = (%x, 0, 0, 2), kwargs = {})
%slice_scatter : [num_users=1] = call_function[target=torch.ops.aten.slice_scatter.default](args = (%slice_3, %copy, 1, 0, 3), kwargs = {})
%slice_scatter_1 : [num_users=2] = call_function[target=torch.ops.aten.slice_scatter.default](args = (%x, %slice_scatter, 0, 0, 2), kwargs = {})
%add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%slice_scatter_1, 2), kwargs = {})
%copy__default : [num_users=0] = call_function[target=torch.ops.aten.copy_.default](args = (%x, %slice_scatter_1), kwargs = {})
return (add,)
nostrict¶
FAILED
Constraints violated (batch)! For more information, run with TORCH_LOGS="+dynamic".
- Not all values of batch = L['args'][0][0].size()[0] in the specified range satisfy the generated guard L['args'][0][0].size()[0] != 2.
nostrict-decall¶
FAILED
Constraints violated (batch)! For more information, run with TORCH_LOGS="+dynamic".
- Not all values of batch = L['args'][0][0].size()[0] in the specified range satisfy the generated guard L['args'][0][0].size()[0] != 2.
jit¶
graph():
%lifted_tensor_2 : [num_users=1] = get_attr[target=lifted_tensor_2]
%lifted_tensor_3 : [num_users=1] = get_attr[target=lifted_tensor_3]
%x : [num_users=2] = placeholder[target=x]
%slice_1 : [num_users=1] = call_function[target=torch.ops.aten.slice.Tensor](args = (%x, 0, 0, 2), kwargs = {})
%slice_2 : [num_users=1] = call_function[target=torch.ops.aten.slice.Tensor](args = (%slice_1, 1, 0, 3), kwargs = {})
%fill_ : [num_users=0] = call_function[target=torch.ops.aten.fill_.Tensor](args = (%slice_2, %lifted_tensor_3), kwargs = {})
%add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%x, %lifted_tensor_2), kwargs = {})
return (add,)
jit-decall¶
graph():
%lifted_tensor_2 : [num_users=1] = get_attr[target=lifted_tensor_2]
%lifted_tensor_3 : [num_users=1] = get_attr[target=lifted_tensor_3]
%x : [num_users=4] = placeholder[target=x]
%slice_1 : [num_users=1] = call_function[target=torch.ops.aten.slice.Tensor](args = (%x, 0, 0, 2), kwargs = {})
%slice_2 : [num_users=1] = call_function[target=torch.ops.aten.slice.Tensor](args = (%slice_1, 1, 0, 3), kwargs = {})
%copy : [num_users=1] = call_function[target=torch.ops.aten.copy.default](args = (%slice_2, %lifted_tensor_3), kwargs = {})
%slice_3 : [num_users=1] = call_function[target=torch.ops.aten.slice.Tensor](args = (%x, 0, 0, 2), kwargs = {})
%slice_scatter : [num_users=1] = call_function[target=torch.ops.aten.slice_scatter.default](args = (%slice_3, %copy, 1, 0, 3), kwargs = {})
%slice_scatter_1 : [num_users=2] = call_function[target=torch.ops.aten.slice_scatter.default](args = (%x, %slice_scatter, 0, 0, 2), kwargs = {})
%add_12 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%slice_scatter_1, %lifted_tensor_2), kwargs = {})
%copy__default : [num_users=0] = call_function[target=torch.ops.aten.copy_.default](args = (%x, %slice_scatter_1), kwargs = {})
return (add_12,)
tracing¶
graph():
%x : [num_users=1] = placeholder[target=x]
%setitem : [num_users=1] = call_function[target=operator.setitem](args = (%x, (slice(None, 2, None), slice(None, 3, None)), 1), kwargs = {})
%add : [num_users=1] = call_function[target=operator.add](args = (%setitem, 2), kwargs = {})
return add
InplaceSetItemSquareAdd2¶
forward¶
def forward(self, x):
x[:2, :3] = 1
return x + 2, x + 3
strict¶
graph():
%lifted_tensor_0 : [num_users=1] = get_attr[target=lifted_tensor_0]
%x : [num_users=3] = placeholder[target=x]
%lift_fresh_copy : [num_users=1] = call_function[target=torch.ops.aten.lift_fresh_copy.default](args = (%lifted_tensor_0,), kwargs = {})
%slice_1 : [num_users=1] = call_function[target=torch.ops.aten.slice.Tensor](args = (%x, 0, 0, 2), kwargs = {})
%slice_2 : [num_users=1] = call_function[target=torch.ops.aten.slice.Tensor](args = (%slice_1, 1, 0, 3), kwargs = {})
%fill_ : [num_users=0] = call_function[target=torch.ops.aten.fill_.Tensor](args = (%slice_2, %lift_fresh_copy), kwargs = {})
%add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%x, 2), kwargs = {})
%add_1 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%x, 3), kwargs = {})
return (add, add_1)
strict-decall¶
graph():
%lifted_tensor_0 : [num_users=1] = get_attr[target=lifted_tensor_0]
%x : [num_users=4] = placeholder[target=x]
%clone : [num_users=1] = call_function[target=torch.ops.aten.clone.default](args = (%lifted_tensor_0,), kwargs = {})
%slice_1 : [num_users=1] = call_function[target=torch.ops.aten.slice.Tensor](args = (%x, 0, 0, 2), kwargs = {})
%slice_2 : [num_users=1] = call_function[target=torch.ops.aten.slice.Tensor](args = (%slice_1, 1, 0, 3), kwargs = {})
%copy : [num_users=1] = call_function[target=torch.ops.aten.copy.default](args = (%slice_2, %clone), kwargs = {})
%slice_3 : [num_users=1] = call_function[target=torch.ops.aten.slice.Tensor](args = (%x, 0, 0, 2), kwargs = {})
%slice_scatter : [num_users=1] = call_function[target=torch.ops.aten.slice_scatter.default](args = (%slice_3, %copy, 1, 0, 3), kwargs = {})
%slice_scatter_1 : [num_users=3] = call_function[target=torch.ops.aten.slice_scatter.default](args = (%x, %slice_scatter, 0, 0, 2), kwargs = {})
%add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%slice_scatter_1, 2), kwargs = {})
%add_4 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%slice_scatter_1, 3), kwargs = {})
%copy__default : [num_users=0] = call_function[target=torch.ops.aten.copy_.default](args = (%x, %slice_scatter_1), kwargs = {})
return (add, add_4)
nostrict¶
FAILED
Constraints violated (batch)! For more information, run with TORCH_LOGS="+dynamic".
- Not all values of batch = L['args'][0][0].size()[0] in the specified range satisfy the generated guard L['args'][0][0].size()[0] != 2.
nostrict-decall¶
FAILED
Constraints violated (batch)! For more information, run with TORCH_LOGS="+dynamic".
- Not all values of batch = L['args'][0][0].size()[0] in the specified range satisfy the generated guard L['args'][0][0].size()[0] != 2.
jit¶
graph():
%lifted_tensor_2 : [num_users=1] = get_attr[target=lifted_tensor_2]
%lifted_tensor_3 : [num_users=1] = get_attr[target=lifted_tensor_3]
%lifted_tensor_4 : [num_users=1] = get_attr[target=lifted_tensor_4]
%x : [num_users=3] = placeholder[target=x]
%slice_1 : [num_users=1] = call_function[target=torch.ops.aten.slice.Tensor](args = (%x, 0, 0, 2), kwargs = {})
%slice_2 : [num_users=1] = call_function[target=torch.ops.aten.slice.Tensor](args = (%slice_1, 1, 0, 3), kwargs = {})
%fill_ : [num_users=0] = call_function[target=torch.ops.aten.fill_.Tensor](args = (%slice_2, %lifted_tensor_4), kwargs = {})
%add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%x, %lifted_tensor_3), kwargs = {})
%add_1 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%x, %lifted_tensor_2), kwargs = {})
return (add, add_1)
jit-decall¶
graph():
%lifted_tensor_2 : [num_users=1] = get_attr[target=lifted_tensor_2]
%lifted_tensor_3 : [num_users=1] = get_attr[target=lifted_tensor_3]
%lifted_tensor_4 : [num_users=1] = get_attr[target=lifted_tensor_4]
%x : [num_users=4] = placeholder[target=x]
%slice_1 : [num_users=1] = call_function[target=torch.ops.aten.slice.Tensor](args = (%x, 0, 0, 2), kwargs = {})
%slice_2 : [num_users=1] = call_function[target=torch.ops.aten.slice.Tensor](args = (%slice_1, 1, 0, 3), kwargs = {})
%copy : [num_users=1] = call_function[target=torch.ops.aten.copy.default](args = (%slice_2, %lifted_tensor_4), kwargs = {})
%slice_3 : [num_users=1] = call_function[target=torch.ops.aten.slice.Tensor](args = (%x, 0, 0, 2), kwargs = {})
%slice_scatter : [num_users=1] = call_function[target=torch.ops.aten.slice_scatter.default](args = (%slice_3, %copy, 1, 0, 3), kwargs = {})
%slice_scatter_1 : [num_users=3] = call_function[target=torch.ops.aten.slice_scatter.default](args = (%x, %slice_scatter, 0, 0, 2), kwargs = {})
%add_12 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%slice_scatter_1, %lifted_tensor_3), kwargs = {})
%add_16 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%slice_scatter_1, %lifted_tensor_2), kwargs = {})
%copy__default : [num_users=0] = call_function[target=torch.ops.aten.copy_.default](args = (%x, %slice_scatter_1), kwargs = {})
return (add_12, add_16)
tracing¶
graph():
%x : [num_users=1] = placeholder[target=x]
%setitem : [num_users=2] = call_function[target=operator.setitem](args = (%x, (slice(None, 2, None), slice(None, 3, None)), 1), kwargs = {})
%add : [num_users=1] = call_function[target=operator.add](args = (%setitem, 2), kwargs = {})
%add_1 : [num_users=1] = call_function[target=operator.add](args = (%setitem, 3), kwargs = {})
return (add, add_1)
SignatureFloat1¶
forward¶
def forward(self, x, alpha: float = 2.0):
return torch.sigmoid(self.linear(x)) - self.buff * alpha
strict¶
FAILED
Expected input at *args[1] to be equal to 1.5, but got 2.5
strict-decall¶
FAILED
Expected input at *args[1] to be equal to 1.5, but got 2.5
nostrict¶
FAILED
Expected input at *args[1] to be equal to 1.5, but got 2.5
nostrict-decall¶
FAILED
Expected input at *args[1] to be equal to 1.5, but got 2.5
jit¶
FAILED
Type 'Tuple[Tensor, float]' cannot be traced. Only Tensors and (possibly nested) Lists, Dicts, and Tuples of Tensors can be traced
jit-decall¶
FAILED
Type 'Tuple[Tensor, float]' cannot be traced. Only Tensors and (possibly nested) Lists, Dicts, and Tuples of Tensors can be traced
tracing¶
graph():
%x : [num_users=1] = placeholder[target=x]
%alpha : float [num_users=1] = placeholder[target=alpha](default=2.0)
%linear : [num_users=1] = call_module[target=linear](args = (%x,), kwargs = {})
%sigmoid : [num_users=1] = call_function[target=torch.sigmoid](args = (%linear,), kwargs = {})
%buff : [num_users=1] = get_attr[target=buff]
%mul : [num_users=1] = call_method[target=mul](args = (%buff, %alpha), kwargs = {})
%sub : [num_users=1] = call_function[target=operator.sub](args = (%sigmoid, %mul), kwargs = {})
return sub
SignatureInt1¶
forward¶
def forward(self, x, i: int = 2):
return torch.sigmoid(self.linear(x)) - self.buff + x[:, i : i + 1]
strict¶
FAILED
Expected input at *args[1] to be equal to 1, but got 2
strict-decall¶
FAILED
Expected input at *args[1] to be equal to 1, but got 2
nostrict¶
FAILED
Expected input at *args[1] to be equal to 1, but got 2
nostrict-decall¶
FAILED
Expected input at *args[1] to be equal to 1, but got 2
jit¶
FAILED
Type 'Tuple[Tensor, int]' cannot be traced. Only Tensors and (possibly nested) Lists, Dicts, and Tuples of Tensors can be traced
jit-decall¶
FAILED
Type 'Tuple[Tensor, int]' cannot be traced. Only Tensors and (possibly nested) Lists, Dicts, and Tuples of Tensors can be traced
tracing¶
graph():
%x : [num_users=2] = placeholder[target=x]
%i : int [num_users=2] = placeholder[target=i](default=2)
%linear : [num_users=1] = call_module[target=linear](args = (%x,), kwargs = {})
%sigmoid : [num_users=1] = call_function[target=torch.sigmoid](args = (%linear,), kwargs = {})
%buff : [num_users=1] = get_attr[target=buff]
%sub : [num_users=1] = call_function[target=operator.sub](args = (%sigmoid, %buff), kwargs = {})
%add : [num_users=1] = call_function[target=operator.add](args = (%i, 1), kwargs = {})
%getitem : [num_users=1] = call_function[target=operator.getitem](args = (%x, (slice(None, None, None), slice(i, add, None))), kwargs = {})
%add_1 : [num_users=1] = call_function[target=operator.add](args = (%sub, %getitem), kwargs = {})
return add_1
SignatureInt2¶
forward¶
def forward(self, x, i: int = 2):
return torch.sigmoid(self.linear(x)) - self.buff + x[:, i]
strict¶
FAILED
Constraints violated (batch)! For more information, run with TORCH_LOGS="+dynamic".
- Not all values of batch = L['x'].size()[0] in the specified range satisfy the generated guard L['x'].size()[0] != 9223372036854775807.
strict-decall¶
FAILED
Constraints violated (batch)! For more information, run with TORCH_LOGS="+dynamic".
- Not all values of batch = L['x'].size()[0] in the specified range satisfy the generated guard L['x'].size()[0] != 9223372036854775807.
nostrict¶
FAILED
Constraints violated (batch)! For more information, run with TORCH_LOGS="+dynamic".
- Not all values of batch = L['args'][0][0].size()[0] in the specified range satisfy the generated guard L['args'][0][0].size()[0] != 9223372036854775807.
nostrict-decall¶
FAILED
Constraints violated (batch)! For more information, run with TORCH_LOGS="+dynamic".
- Not all values of batch = L['args'][0][0].size()[0] in the specified range satisfy the generated guard L['args'][0][0].size()[0] != 9223372036854775807.
jit¶
FAILED
Type 'Tuple[Tensor, int]' cannot be traced. Only Tensors and (possibly nested) Lists, Dicts, and Tuples of Tensors can be traced
jit-decall¶
FAILED
Type 'Tuple[Tensor, int]' cannot be traced. Only Tensors and (possibly nested) Lists, Dicts, and Tuples of Tensors can be traced
tracing¶
graph():
%x : [num_users=2] = placeholder[target=x]
%i : int [num_users=1] = placeholder[target=i](default=2)
%linear : [num_users=1] = call_module[target=linear](args = (%x,), kwargs = {})
%sigmoid : [num_users=1] = call_function[target=torch.sigmoid](args = (%linear,), kwargs = {})
%buff : [num_users=1] = get_attr[target=buff]
%sub : [num_users=1] = call_function[target=operator.sub](args = (%sigmoid, %buff), kwargs = {})
%getitem : [num_users=1] = call_function[target=operator.getitem](args = (%x, (slice(None, None, None), %i)), kwargs = {})
%add : [num_users=1] = call_function[target=operator.add](args = (%sub, %getitem), kwargs = {})
return add
SignatureListFixedLength¶
forward¶
def forward(self, x, lx: list):
return (
torch.sigmoid(self.linear(x)) - self.buff + lx[0] * lx[1].sum(axis=1, keepdim=True)
)
strict¶
graph():
%linear_weight : [num_users=1] = get_attr[target=linear.weight]
%linear_bias : [num_users=1] = get_attr[target=linear.bias]
%buff : [num_users=1] = get_attr[target=buff]
%x : [num_users=1] = placeholder[target=x]
%lx_0 : [num_users=1] = placeholder[target=lx_0]
%lx_1 : [num_users=1] = placeholder[target=lx_1]
%linear : [num_users=1] = call_function[target=torch.ops.aten.linear.default](args = (%x, %linear_weight, %linear_bias), kwargs = {})
%sigmoid : [num_users=1] = call_function[target=torch.ops.aten.sigmoid.default](args = (%linear,), kwargs = {})
%sub : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%sigmoid, %buff), kwargs = {})
%sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%lx_1, [1], True), kwargs = {})
%mul : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%lx_0, %sum_1), kwargs = {})
%add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%sub, %mul), kwargs = {})
return (add,)
strict-decall¶
graph():
%linear_weight : [num_users=1] = get_attr[target=linear.weight]
%linear_bias : [num_users=1] = get_attr[target=linear.bias]
%buff : [num_users=1] = get_attr[target=buff]
%x : [num_users=1] = placeholder[target=x]
%lx_0 : [num_users=1] = placeholder[target=lx_0]
%lx_1 : [num_users=1] = placeholder[target=lx_1]
%permute : [num_users=1] = call_function[target=torch.ops.aten.permute.default](args = (%linear_weight, [1, 0]), kwargs = {})
%addmm : [num_users=1] = call_function[target=torch.ops.aten.addmm.default](args = (%linear_bias, %x, %permute), kwargs = {})
%sigmoid : [num_users=1] = call_function[target=torch.ops.aten.sigmoid.default](args = (%addmm,), kwargs = {})
%sub_2 : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%sigmoid, %buff), kwargs = {})
%sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%lx_1, [1], True), kwargs = {})
%mul_4 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%lx_0, %sum_1), kwargs = {})
%add_15 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%sub_2, %mul_4), kwargs = {})
return (add_15,)
nostrict¶
graph():
%linear_weight : [num_users=1] = get_attr[target=linear.weight]
%linear_bias : [num_users=1] = get_attr[target=linear.bias]
%buff : [num_users=1] = get_attr[target=buff]
%x : [num_users=1] = placeholder[target=x]
%lx_0 : [num_users=1] = placeholder[target=lx_0]
%lx_1 : [num_users=1] = placeholder[target=lx_1]
%linear : [num_users=1] = call_function[target=torch.ops.aten.linear.default](args = (%x, %linear_weight, %linear_bias), kwargs = {})
%sigmoid : [num_users=1] = call_function[target=torch.ops.aten.sigmoid.default](args = (%linear,), kwargs = {})
%sub : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%sigmoid, %buff), kwargs = {})
%sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%lx_1, [1], True), kwargs = {})
%mul : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%lx_0, %sum_1), kwargs = {})
%add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%sub, %mul), kwargs = {})
return (add,)
nostrict-decall¶
graph():
%linear_weight : [num_users=1] = get_attr[target=linear.weight]
%linear_bias : [num_users=1] = get_attr[target=linear.bias]
%buff : [num_users=1] = get_attr[target=buff]
%x : [num_users=1] = placeholder[target=x]
%lx_0 : [num_users=1] = placeholder[target=lx_0]
%lx_1 : [num_users=1] = placeholder[target=lx_1]
%permute : [num_users=1] = call_function[target=torch.ops.aten.permute.default](args = (%linear_weight, [1, 0]), kwargs = {})
%addmm : [num_users=1] = call_function[target=torch.ops.aten.addmm.default](args = (%linear_bias, %x, %permute), kwargs = {})
%sigmoid : [num_users=1] = call_function[target=torch.ops.aten.sigmoid.default](args = (%addmm,), kwargs = {})
%sub_2 : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%sigmoid, %buff), kwargs = {})
%sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%lx_1, [1], True), kwargs = {})
%mul_4 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%lx_0, %sum_1), kwargs = {})
%add_15 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%sub_2, %mul_4), kwargs = {})
return (add_15,)
jit¶
graph():
%linear_bias : [num_users=1] = get_attr[target=linear.bias]
%linear_weight : [num_users=1] = get_attr[target=linear.weight]
%buff : [num_users=1] = get_attr[target=buff]
%x : [num_users=1] = placeholder[target=x]
%lx_0 : [num_users=1] = placeholder[target=lx_0]
%lx_1 : [num_users=1] = placeholder[target=lx_1]
%linear : [num_users=1] = call_function[target=torch.ops.aten.linear.default](args = (%x, %linear_weight, %linear_bias), kwargs = {})
%sigmoid : [num_users=1] = call_function[target=torch.ops.aten.sigmoid.default](args = (%linear,), kwargs = {})
%sub : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%sigmoid, %buff), kwargs = {})
%sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%lx_1, [1], True), kwargs = {})
%mul : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%lx_0, %sum_1), kwargs = {})
%add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%sub, %mul), kwargs = {})
return (add,)
jit-decall¶
graph():
%linear_bias : [num_users=1] = get_attr[target=linear.bias]
%linear_weight : [num_users=1] = get_attr[target=linear.weight]
%buff : [num_users=1] = get_attr[target=buff]
%x : [num_users=1] = placeholder[target=x]
%lx_0 : [num_users=1] = placeholder[target=lx_0]
%lx_1 : [num_users=1] = placeholder[target=lx_1]
%permute : [num_users=1] = call_function[target=torch.ops.aten.permute.default](args = (%linear_weight, [1, 0]), kwargs = {})
%addmm : [num_users=1] = call_function[target=torch.ops.aten.addmm.default](args = (%linear_bias, %x, %permute), kwargs = {})
%sigmoid : [num_users=1] = call_function[target=torch.ops.aten.sigmoid.default](args = (%addmm,), kwargs = {})
%sub_2 : [num_users=1] = call_function[target=torch.ops.aten.sub.Tensor](args = (%sigmoid, %buff), kwargs = {})
%sum_1 : [num_users=1] = call_function[target=torch.ops.aten.sum.dim_IntList](args = (%lx_1, [1], True), kwargs = {})
%mul_4 : [num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%lx_0, %sum_1), kwargs = {})
%add_15 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%sub_2, %mul_4), kwargs = {})
return (add_15,)
tracing¶
graph():
%x : [num_users=1] = placeholder[target=x]
%lx : list [num_users=2] = placeholder[target=lx]
%linear : [num_users=1] = call_module[target=linear](args = (%x,), kwargs = {})
%sigmoid : [num_users=1] = call_function[target=torch.sigmoid](args = (%linear,), kwargs = {})
%buff : [num_users=1] = get_attr[target=buff]
%sub : [num_users=1] = call_function[target=operator.sub](args = (%sigmoid, %buff), kwargs = {})
%getitem : [num_users=1] = call_function[target=operator.getitem](args = (%lx, 0), kwargs = {})
%getitem_1 : [num_users=1] = call_function[target=operator.getitem](args = (%lx, 1), kwargs = {})
%sum_1 : [num_users=1] = call_method[target=sum](args = (%getitem_1,), kwargs = {axis: 1, keepdim: True})
%mul : [num_users=1] = call_function[target=operator.mul](args = (%getitem, %sum_1), kwargs = {})
%add : [num_users=1] = call_function[target=operator.add](args = (%sub, %mul), kwargs = {})
return add
SignatureListFixedWithNone¶
forward¶
def forward(self, lx):
x = lx[0]
if lx[1] is not None:
x += lx[1]
if lx[2] is not None:
x += lx[2]
return x
strict¶
FAILED
Detected mismatch between the structure of `inputs` and `dynamic_shapes`: `inputs['lx']` has 3 elements, but `dynamic_shapes['lx']` has 2 elements
For more information about this error, see: https://pytorch.org/docs/main/generated/exportdb/index.html#dynamic-shapes-validation
strict-decall¶
FAILED
Detected mismatch between the structure of `inputs` and `dynamic_shapes`: `inputs['lx']` has 3 elements, but `dynamic_shapes['lx']` has 2 elements
For more information about this error, see: https://pytorch.org/docs/main/generated/exportdb/index.html#dynamic-shapes-validation
nostrict¶
FAILED
Detected mismatch between the structure of `inputs` and `dynamic_shapes`: `inputs['lx']` has 3 elements, but `dynamic_shapes['lx']` has 2 elements
For more information about this error, see: https://pytorch.org/docs/main/generated/exportdb/index.html#dynamic-shapes-validation
nostrict-decall¶
FAILED
Detected mismatch between the structure of `inputs` and `dynamic_shapes`: `inputs['lx']` has 3 elements, but `dynamic_shapes['lx']` has 2 elements
For more information about this error, see: https://pytorch.org/docs/main/generated/exportdb/index.html#dynamic-shapes-validation
jit¶
FAILED
Type 'Tuple[List[Optional[Tensor]]]' cannot be traced. Only Tensors and (possibly nested) Lists, Dicts, and Tuples of Tensors can be traced
jit-decall¶
FAILED
Type 'Tuple[List[Optional[Tensor]]]' cannot be traced. Only Tensors and (possibly nested) Lists, Dicts, and Tuples of Tensors can be traced
tracing¶
FAILED
Unable to clone type <class 'NoneType'>, x=None into numpy
SignatureListVariableLength¶
forward¶
def forward(self, x, lx: list):
t = torch.cat(lx, dim=1).sum(axis=1, keepdim=True)
return torch.sigmoid(self.linear(x)) - self.buff + t
strict¶
FAILED
Trying to flatten user inputs with exported input tree spec:
TreeSpec(tuple, None, [TreeSpec(tuple, None, [*,
TreeSpec(list, None, [*,
*])]),
TreeSpec(dict, [], [])])
but actually got inputs with tree spec of:
TreeSpec(tuple, None, [TreeSpec(tuple, None, [*,
TreeSpec(list, None, [*,
*,
*])]),
TreeSpec(dict, [], [])]).
Please check that the inputs have the same number of args and kwargs as the ones you used when tracing.
strict-decall¶
FAILED
Trying to flatten user inputs with exported input tree spec:
TreeSpec(tuple, None, [TreeSpec(tuple, None, [*,
TreeSpec(list, None, [*,
*])]),
TreeSpec(dict, [], [])])
but actually got inputs with tree spec of:
TreeSpec(tuple, None, [TreeSpec(tuple, None, [*,
TreeSpec(list, None, [*,
*,
*])]),
TreeSpec(dict, [], [])]).
Please check that the inputs have the same number of args and kwargs as the ones you used when tracing.
nostrict¶
FAILED
Trying to flatten user inputs with exported input tree spec:
TreeSpec(tuple, None, [TreeSpec(tuple, None, [*,
TreeSpec(list, None, [*,
*])]),
TreeSpec(dict, [], [])])
but actually got inputs with tree spec of:
TreeSpec(tuple, None, [TreeSpec(tuple, None, [*,
TreeSpec(list, None, [*,
*,
*])]),
TreeSpec(dict, [], [])]).
Please check that the inputs have the same number of args and kwargs as the ones you used when tracing.
nostrict-decall¶
FAILED
Trying to flatten user inputs with exported input tree spec:
TreeSpec(tuple, None, [TreeSpec(tuple, None, [*,
TreeSpec(list, None, [*,
*])]),
TreeSpec(dict, [], [])])
but actually got inputs with tree spec of:
TreeSpec(tuple, None, [TreeSpec(tuple, None, [*,
TreeSpec(list, None, [*,
*,
*])]),
TreeSpec(dict, [], [])]).
Please check that the inputs have the same number of args and kwargs as the ones you used when tracing.
jit¶
FAILED
Trying to flatten user inputs with exported input tree spec:
TreeSpec(tuple, None, [TreeSpec(tuple, None, [*,
TreeSpec(list, None, [*,
*])]),
TreeSpec(dict, [], [])])
but actually got inputs with tree spec of:
TreeSpec(tuple, None, [TreeSpec(tuple, None, [*,
TreeSpec(list, None, [*,
*,
*])]),
TreeSpec(dict, [], [])]).
Please check that the inputs have the same number of args and kwargs as the ones you used when tracing.
jit-decall¶
FAILED
Trying to flatten user inputs with exported input tree spec:
TreeSpec(tuple, None, [TreeSpec(tuple, None, [*,
TreeSpec(list, None, [*,
*])]),
TreeSpec(dict, [], [])])
but actually got inputs with tree spec of:
TreeSpec(tuple, None, [TreeSpec(tuple, None, [*,
TreeSpec(list, None, [*,
*,
*])]),
TreeSpec(dict, [], [])]).
Please check that the inputs have the same number of args and kwargs as the ones you used when tracing.
tracing¶
graph():
%x : [num_users=1] = placeholder[target=x]
%lx : list [num_users=1] = placeholder[target=lx]
%cat : [num_users=1] = call_function[target=torch.cat](args = (%lx, 1), kwargs = {})
%sum_1 : [num_users=1] = call_method[target=sum](args = (%cat,), kwargs = {axis: 1, keepdim: True})
%linear : [num_users=1] = call_module[target=linear](args = (%x,), kwargs = {})
%sigmoid : [num_users=1] = call_function[target=torch.sigmoid](args = (%linear,), kwargs = {})
%buff : [num_users=1] = get_attr[target=buff]
%sub : [num_users=1] = call_function[target=operator.sub](args = (%sigmoid, %buff), kwargs = {})
%add : [num_users=1] = call_function[target=operator.add](args = (%sub, %sum_1), kwargs = {})
return add
SignatureShapeAsIndex¶
forward¶
def forward(self, x, y):
t = torch.sigmoid(self.linear(x)) + x
return t[:, : y.shape[1]]
strict¶
graph():
%linear_weight : [num_users=1] = get_attr[target=linear.weight]
%linear_bias : [num_users=1] = get_attr[target=linear.bias]
%x : [num_users=2] = placeholder[target=x]
%y : [num_users=1] = placeholder[target=y]
%sym_size_int_3 : [num_users=1] = call_function[target=torch.ops.aten.sym_size.int](args = (%y, 1), kwargs = {})
%linear : [num_users=1] = call_function[target=torch.ops.aten.linear.default](args = (%x, %linear_weight, %linear_bias), kwargs = {})
%sigmoid : [num_users=1] = call_function[target=torch.ops.aten.sigmoid.default](args = (%linear,), kwargs = {})
%add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%sigmoid, %x), kwargs = {})
%slice_1 : [num_users=1] = call_function[target=torch.ops.aten.slice.Tensor](args = (%add, 0, 0, 9223372036854775807), kwargs = {})
%slice_2 : [num_users=1] = call_function[target=torch.ops.aten.slice.Tensor](args = (%slice_1, 1, 0, %sym_size_int_3), kwargs = {})
return (slice_2,)
strict-decall¶
graph():
%linear_weight : [num_users=1] = get_attr[target=linear.weight]
%linear_bias : [num_users=1] = get_attr[target=linear.bias]
%x : [num_users=2] = placeholder[target=x]
%y : [num_users=1] = placeholder[target=y]
%sym_size_int_5 : [num_users=1] = call_function[target=torch.ops.aten.sym_size.int](args = (%y, 1), kwargs = {})
%permute : [num_users=1] = call_function[target=torch.ops.aten.permute.default](args = (%linear_weight, [1, 0]), kwargs = {})
%addmm : [num_users=1] = call_function[target=torch.ops.aten.addmm.default](args = (%linear_bias, %x, %permute), kwargs = {})
%sigmoid : [num_users=1] = call_function[target=torch.ops.aten.sigmoid.default](args = (%addmm,), kwargs = {})
%add_6 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%sigmoid, %x), kwargs = {})
%slice_1 : [num_users=1] = call_function[target=torch.ops.aten.slice.Tensor](args = (%add_6, 0, 0, 9223372036854775807), kwargs = {})
%slice_2 : [num_users=1] = call_function[target=torch.ops.aten.slice.Tensor](args = (%slice_1, 1, 0, %sym_size_int_5), kwargs = {})
return (slice_2,)
nostrict¶
graph():
%linear_weight : [num_users=1] = get_attr[target=linear.weight]
%linear_bias : [num_users=1] = get_attr[target=linear.bias]
%buff : [num_users=0] = get_attr[target=buff]
%x : [num_users=2] = placeholder[target=x]
%y : [num_users=1] = placeholder[target=y]
%sym_size_int_3 : [num_users=1] = call_function[target=torch.ops.aten.sym_size.int](args = (%y, 1), kwargs = {})
%linear : [num_users=1] = call_function[target=torch.ops.aten.linear.default](args = (%x, %linear_weight, %linear_bias), kwargs = {})
%sigmoid : [num_users=1] = call_function[target=torch.ops.aten.sigmoid.default](args = (%linear,), kwargs = {})
%add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%sigmoid, %x), kwargs = {})
%slice_1 : [num_users=1] = call_function[target=torch.ops.aten.slice.Tensor](args = (%add, 0, 0, 9223372036854775807), kwargs = {})
%slice_2 : [num_users=1] = call_function[target=torch.ops.aten.slice.Tensor](args = (%slice_1, 1, 0, %sym_size_int_3), kwargs = {})
return (slice_2,)
nostrict-decall¶
graph():
%linear_weight : [num_users=1] = get_attr[target=linear.weight]
%linear_bias : [num_users=1] = get_attr[target=linear.bias]
%buff : [num_users=0] = get_attr[target=buff]
%x : [num_users=2] = placeholder[target=x]
%y : [num_users=1] = placeholder[target=y]
%sym_size_int_5 : [num_users=1] = call_function[target=torch.ops.aten.sym_size.int](args = (%y, 1), kwargs = {})
%permute : [num_users=1] = call_function[target=torch.ops.aten.permute.default](args = (%linear_weight, [1, 0]), kwargs = {})
%addmm : [num_users=1] = call_function[target=torch.ops.aten.addmm.default](args = (%linear_bias, %x, %permute), kwargs = {})
%sigmoid : [num_users=1] = call_function[target=torch.ops.aten.sigmoid.default](args = (%addmm,), kwargs = {})
%add_6 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%sigmoid, %x), kwargs = {})
%slice_1 : [num_users=1] = call_function[target=torch.ops.aten.slice.Tensor](args = (%add_6, 0, 0, 9223372036854775807), kwargs = {})
%slice_2 : [num_users=1] = call_function[target=torch.ops.aten.slice.Tensor](args = (%slice_1, 1, 0, %sym_size_int_5), kwargs = {})
return (slice_2,)
jit¶
graph():
%linear_bias : [num_users=1] = get_attr[target=linear.bias]
%linear_weight : [num_users=1] = get_attr[target=linear.weight]
%x : [num_users=2] = placeholder[target=x]
%y : [num_users=1] = placeholder[target=y]
%sym_size_int_3 : [num_users=1] = call_function[target=torch.ops.aten.sym_size.int](args = (%y, 1), kwargs = {})
%linear : [num_users=1] = call_function[target=torch.ops.aten.linear.default](args = (%x, %linear_weight, %linear_bias), kwargs = {})
%sigmoid : [num_users=1] = call_function[target=torch.ops.aten.sigmoid.default](args = (%linear,), kwargs = {})
%add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%sigmoid, %x), kwargs = {})
%slice_1 : [num_users=1] = call_function[target=torch.ops.aten.slice.Tensor](args = (%add, 0, 0, 9223372036854775807), kwargs = {})
%slice_2 : [num_users=1] = call_function[target=torch.ops.aten.slice.Tensor](args = (%slice_1, 1, 0, %sym_size_int_3), kwargs = {})
return (slice_2,)
jit-decall¶
graph():
%linear_bias : [num_users=1] = get_attr[target=linear.bias]
%linear_weight : [num_users=1] = get_attr[target=linear.weight]
%x : [num_users=2] = placeholder[target=x]
%y : [num_users=1] = placeholder[target=y]
%sym_size_int_6 : [num_users=1] = call_function[target=torch.ops.aten.sym_size.int](args = (%y, 1), kwargs = {})
%permute : [num_users=1] = call_function[target=torch.ops.aten.permute.default](args = (%linear_weight, [1, 0]), kwargs = {})
%addmm : [num_users=1] = call_function[target=torch.ops.aten.addmm.default](args = (%linear_bias, %x, %permute), kwargs = {})
%sigmoid : [num_users=1] = call_function[target=torch.ops.aten.sigmoid.default](args = (%addmm,), kwargs = {})
%add_6 : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%sigmoid, %x), kwargs = {})
%slice_1 : [num_users=1] = call_function[target=torch.ops.aten.slice.Tensor](args = (%add_6, 0, 0, 9223372036854775807), kwargs = {})
%slice_2 : [num_users=1] = call_function[target=torch.ops.aten.slice.Tensor](args = (%slice_1, 1, 0, %sym_size_int_6), kwargs = {})
return (slice_2,)
tracing¶
graph():
%x : [num_users=2] = placeholder[target=x]
%y : [num_users=1] = placeholder[target=y]
%linear : [num_users=1] = call_module[target=linear](args = (%x,), kwargs = {})
%sigmoid : [num_users=1] = call_function[target=torch.sigmoid](args = (%linear,), kwargs = {})
%add : [num_users=1] = call_function[target=operator.add](args = (%sigmoid, %x), kwargs = {})
%getattr_1 : [num_users=1] = call_function[target=builtins.getattr](args = (%y, shape), kwargs = {})
%getitem : [num_users=1] = call_function[target=operator.getitem](args = (%getattr_1, 1), kwargs = {})
%getitem_1 : [num_users=1] = call_function[target=operator.getitem](args = (%add, (slice(None, None, None), slice(None, getitem, None))), kwargs = {})
return getitem_1
TypeBFloat16¶
forward¶
def forward(self, x):
xb = x.to(torch.bfloat16)
return (xb + xb).to(torch.float32)
strict¶
graph():
%x : [num_users=1] = placeholder[target=x]
%to : [num_users=1] = call_function[target=torch.ops.aten.to.dtype](args = (%x, torch.bfloat16), kwargs = {})
%add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%to, %to), kwargs = {})
%to_1 : [num_users=1] = call_function[target=torch.ops.aten.to.dtype](args = (%add, torch.float32), kwargs = {})
return (to_1,)
strict-decall¶
graph():
%x : [num_users=2] = placeholder[target=x]
%_to_copy : [num_users=1] = call_function[target=torch.ops.aten._to_copy.default](args = (%x,), kwargs = {dtype: torch.bfloat16})
%_assert_tensor_metadata : [num_users=0] = call_function[target=torch.ops.aten._assert_tensor_metadata.default](args = (%x, None, None, torch.float32), kwargs = {})
%add_3 : [num_users=2] = call_function[target=torch.ops.aten.add.Tensor](args = (%_to_copy, %_to_copy), kwargs = {})
%_to_copy_1 : [num_users=1] = call_function[target=torch.ops.aten._to_copy.default](args = (%add_3,), kwargs = {dtype: torch.float32})
%_assert_tensor_metadata_1 : [num_users=0] = call_function[target=torch.ops.aten._assert_tensor_metadata.default](args = (%add_3, None, None, torch.bfloat16), kwargs = {})
return (_to_copy_1,)
nostrict¶
graph():
%x : [num_users=1] = placeholder[target=x]
%to : [num_users=1] = call_function[target=torch.ops.aten.to.dtype](args = (%x, torch.bfloat16), kwargs = {})
%add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%to, %to), kwargs = {})
%to_1 : [num_users=1] = call_function[target=torch.ops.aten.to.dtype](args = (%add, torch.float32), kwargs = {})
return (to_1,)
nostrict-decall¶
graph():
%x : [num_users=2] = placeholder[target=x]
%_to_copy : [num_users=1] = call_function[target=torch.ops.aten._to_copy.default](args = (%x,), kwargs = {dtype: torch.bfloat16})
%_assert_tensor_metadata : [num_users=0] = call_function[target=torch.ops.aten._assert_tensor_metadata.default](args = (%x, None, None, torch.float32), kwargs = {})
%add_3 : [num_users=2] = call_function[target=torch.ops.aten.add.Tensor](args = (%_to_copy, %_to_copy), kwargs = {})
%_to_copy_1 : [num_users=1] = call_function[target=torch.ops.aten._to_copy.default](args = (%add_3,), kwargs = {dtype: torch.float32})
%_assert_tensor_metadata_1 : [num_users=0] = call_function[target=torch.ops.aten._assert_tensor_metadata.default](args = (%add_3, None, None, torch.bfloat16), kwargs = {})
return (_to_copy_1,)
jit¶
graph():
%x : [num_users=1] = placeholder[target=x]
%to : [num_users=1] = call_function[target=torch.ops.aten.to.dtype](args = (%x, torch.bfloat16), kwargs = {})
%add : [num_users=1] = call_function[target=torch.ops.aten.add.Tensor](args = (%to, %to), kwargs = {})
%to_1 : [num_users=1] = call_function[target=torch.ops.aten.to.dtype](args = (%add, torch.float32), kwargs = {})
return (to_1,)
jit-decall¶
graph():
%x : [num_users=2] = placeholder[target=x]
%_to_copy : [num_users=1] = call_function[target=torch.ops.aten._to_copy.default](args = (%x,), kwargs = {dtype: torch.bfloat16})
%_assert_tensor_metadata : [num_users=0] = call_function[target=torch.ops.aten._assert_tensor_metadata.default](args = (%x, None, None, torch.float32), kwargs = {})
%add_3 : [num_users=2] = call_function[target=torch.ops.aten.add.Tensor](args = (%_to_copy, %_to_copy), kwargs = {})
%_to_copy_1 : [num_users=1] = call_function[target=torch.ops.aten._to_copy.default](args = (%add_3,), kwargs = {dtype: torch.float32})
%_assert_tensor_metadata_1 : [num_users=0] = call_function[target=torch.ops.aten._assert_tensor_metadata.default](args = (%add_3, None, None, torch.bfloat16), kwargs = {})
return (_to_copy_1,)
tracing¶
graph():
%x : [num_users=1] = placeholder[target=x]
%to : [num_users=1] = call_method[target=to](args = (%x, torch.bfloat16), kwargs = {})
%add : [num_users=1] = call_function[target=operator.add](args = (%to, %to), kwargs = {})
%to_1 : [num_users=1] = call_method[target=to](args = (%add, torch.float32), kwargs = {})
return to_1
Summary¶
case |
jit |
jit-decall |
nostrict |
nostrict-decall |
strict |
strict-decall |
tracing |
|---|---|---|---|---|---|---|---|
FAIL |
|||||||
FAIL |
FAIL |
FAIL |
FAIL |
FAIL |
FAIL |
FAIL |
|
FAIL |
FAIL |
FAIL |
FAIL |
||||
FAIL |
FAIL |
FAIL |
FAIL |
||||
FAIL |
FAIL |
FAIL |
FAIL |
||||
FAIL |
FAIL |
||||||
FAIL |
FAIL |
||||||
FAIL |
FAIL |
FAIL |
FAIL |
FAIL |
FAIL |
||
FAIL |
FAIL |
||||||
FAIL |
FAIL |
FAIL |
|||||
FAIL |
FAIL |
FAIL |
|||||
FAIL |
FAIL |
FAIL |
|||||
FAIL |
FAIL |
FAIL |
|||||
FAIL |
FAIL |
FAIL |
|||||
FAIL |
|||||||
FAIL |
|||||||
FAIL |
FAIL |
FAIL |
FAIL |
FAIL |
FAIL |
||
FAIL |
FAIL |
FAIL |
FAIL |
||||
FAIL |
FAIL |
FAIL |
FAIL |
||||
FAIL |
FAIL |
||||||
FAIL |
FAIL |
||||||
FAIL |
FAIL |
||||||
FAIL |
FAIL |
FAIL |
FAIL |
FAIL |
FAIL |
||
FAIL |
FAIL |
FAIL |
FAIL |
FAIL |
FAIL |
||
FAIL |
FAIL |
FAIL |
FAIL |
FAIL |
FAIL |
||
FAIL |
FAIL |
FAIL |
FAIL |
FAIL |
FAIL |
FAIL |
|
FAIL |
FAIL |
FAIL |
FAIL |
FAIL |
FAIL |
||