Exported ONNX with Static Shapes¶

The following script shows the exported program for many short cases and various way to retrieve the torch.fx.Graph equivalent to the original model.

<<<

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
from experimental_experiment.helpers import pretty_onnx

cases = discover()
print()
print(":ref:`Summary <lo-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} <lo-model-case-export-{name}>`")
print()

obs = []
for name, cls_model in sorted(cases.items()):
    print()
    print(f".. _lo-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 (
        "custom-fallback",
        "custom-tracing",
        "dynamo-ir",
        "script",
    ):
        expname = exporter.replace("export-", "")
        print()
        print(expname)
        print("+" * len(expname))
        print()
        res = run_exporter(exporter, cls_model, False, quiet=True)
        case_ref = f":ref:`{name} <lo-model-case-export-{name}>`"
        if "exported" in res:
            print("::")
            print()
            print(textwrap.indent(pretty_onnx(res["onnx"]), "    "))
            print()
            obs.append(dict(case=case_ref, error="", exporter=exporter))
        else:
            print("**FAILED**")
            print()
            print("::")
            print()
            print(textwrap.indent(str(res["error"]), "    "))
            print()
            obs.append(dict(case=case_ref, error="FAIL", exporter=exporter))

print()
print(".. _lo-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()

>>>

Summary

AtenAsStrided¶

forward¶

def forward(self, x):
    y = torch.as_strided(x, (2, 2, 8, 4), (128, 8, 16, 1))
    return y

custom-fallback¶

opset: domain='' version=18
doc_string: large_model=False, inline=False, external_threshold=102...
input: name='x' type=dtype('float32') shape=[2, 2, 8, 8]
init: name='init7_s1_-1' type=int64 shape=(1,) -- array([-1])         -- Opset.make_node.1/Shape
init: name='init7_s128_' type=int64 shape=(128,)                      -- Opset.make_node.0
init: name='init7_s4_2_2_8_4' type=int64 shape=(4,) -- array([2, 2, 8, 4])-- Opset.make_node.1/Shape
Reshape(x, init7_s1_-1) -> _onx_reshape0
  Gather(_onx_reshape0, init7_s128_) -> _onx_gather0
    Reshape(_onx_gather0, init7_s4_2_2_8_4) -> output_0
output: name='output_0' type=dtype('float32') shape=[2, 2, 8, 4]

custom-tracing¶

opset: domain='' version=18
doc_string: large_model=False, inline=False, external_threshold=102...
input: name='x' type=dtype('float32') shape=[2, 2, 8, 8]
init: name='init7_s1_-1' type=int64 shape=(1,) -- array([-1])         -- Opset.make_node.1/Shape
init: name='init7_s128_' type=int64 shape=(128,)                      -- Opset.make_node.0
init: name='init7_s4_2_2_8_4' type=int64 shape=(4,) -- array([2, 2, 8, 4])-- Opset.make_node.1/Shape
Reshape(x, init7_s1_-1) -> _onx_reshape0
  Gather(_onx_reshape0, init7_s128_) -> _onx_gather0
    Reshape(_onx_gather0, init7_s4_2_2_8_4) -> output
output: name='output' type=dtype('float32') shape=[2, 2, 8, 4]

dynamo-ir¶

opset: domain='pkg.onnxscript.torch_lib.common' version=1
opset: domain='' version=18
opset: domain='pkg.onnxscript.torch_lib' version=1
input: name='x' type=dtype('float32') shape=[2, 2, 8, 8]
Constant(value_ints=[-1]) -> neg_1
Constant(value=[2, 2, 8, ...) -> val_0
Constant(value=[128, 8, 1...) -> val_1
Constant(value=[4]) -> rank_tensor
Constant(value_int=0) -> indices
SequenceEmpty() -> one_seq
Constant(value_int=4) -> rank_0
Loop(rank_0, , one_seq, indices, body=G1) -> one_seq_16, indices_17
Constant(value_ints=[-1]) -> tmp_18
  Reshape(x, tmp_18) -> self_flatten
Constant(value_int=0) -> storage_offset
  CastLike(storage_offset, indices_17) -> storage_offset_cast
  Add(indices_17, storage_offset_cast) -> indices_19
    Gather(self_flatten, indices_19) -> as_strided
output: name='as_strided' type=dtype('float32') shape=[2, 2, 8, 4]
----- subgraph ---- Loop - n6_2 - att.body=G1 -- level=1 -- i,cond_in,one_seq_1,indices_2 -> cond_out,one_seq_15,indices_13
input: name='i' type=dtype('int64') shape=None
input: name='cond_in' type=dtype('bool') shape=None
input: name='one_seq_1' type='NOTENSOR' shape=None
input: name='indices_2' type='NOTENSOR' shape=None
Constant(value_floats=[1.0]) -> tmp_14
  SequenceInsert(one_seq_1, tmp_14) -> one_seq_15
Constant(value=4) -> rank_3_cast
  Sub(rank_3_cast, i) -> tmp
Constant(value=1) -> int64_1_cast
  Sub(tmp, int64_1_cast) -> j
Reshape(j, neg_1) -> j_tensor
Gather(val_0, j_tensor, axis=0) -> size_dim_j
Slice(val_0, j_tensor, rank_tensor) -> size_after_j
  Expand(indices_2, size_after_j) -> indices_4
Gather(val_1, j_tensor, axis=0) -> stride_dim_j
Constant(value=0) -> int64_0_cast
Constant(value=1) -> int64_1_5_cast
  Range(int64_0_cast, size_dim_j, int64_1_5_cast) -> tmp_6
  Mul(tmp_6, stride_dim_j) -> add_value
Constant(value=0) -> int64_0_7_cast
  Equal(i, int64_0_7_cast) -> cond
    If(cond, then_branch=G2, else_branch=G3) -> shape_11
    Reshape(add_value, shape_11) -> add_value_12
    Add(indices_4, add_value_12) -> indices_13
Identity(cond_in) -> cond_out
output: name='cond_out' type=dtype('bool') shape=None
output: name='one_seq_15' type='NOTENSOR' shape=None
output: name='indices_13' type='NOTENSOR' shape=None
----- subgraph ---- If - n20 - att.then_branch=G2 -- level=2 --  -> shape
Identity(size_dim_j) -> shape
output: name='shape' type=dtype('int64') shape=[1]
----- subgraph ---- If - n20 - att.else_branch=G3 -- level=2 --  -> shape_10
Cast(size_dim_j, to=1) -> tmp_8
ConcatFromSequence(one_seq_1, axis=0) -> ones
  Concat(tmp_8, ones, axis=0) -> shape_9
    Cast(shape_9, to=7) -> shape_10
output: name='shape_10' type=dtype('int64') shape=None
----- subgraph ---- If - n20 - att.then_branch=G2 -- level=1 --  -> shape
Identity(size_dim_j) -> shape
output: name='shape' type=dtype('int64') shape=[1]
----- subgraph ---- If - n20 - att.else_branch=G3 -- level=1 --  -> shape_10
Cast(size_dim_j, to=1) -> tmp_8
ConcatFromSequence(one_seq_1, axis=0) -> ones
  Concat(tmp_8, ones, axis=0) -> shape_9
    Cast(shape_9, to=7) -> shape_10
output: name='shape_10' type=dtype('int64') shape=None

script¶

opset: domain='' version=17
input: name='onnx::Reshape_0' type=dtype('float32') shape=[2, 2, 8, 8]
Constant(value=[-1]) -> /Constant_output_0
  Reshape(onnx::Reshape_0, /Constant_output_0, allowzero=0) -> /Reshape_output_0
Constant(value=[[[[0, 1, ...) -> /Constant_1_output_0
  Gather(/Reshape_output_0, /Constant_1_output_0) -> 4
output: name='4' type=dtype('float32') shape=[2, 2, 8, 4]

AtenInterpolate¶

forward¶

def forward(self, x):
    y = torch.nn.functional.interpolate(
        x,
        scale_factor=2.0,
        mode="bilinear",
        recompute_scale_factor=False,
    )
    return y

custom-fallback¶

opset: domain='' version=18
doc_string: large_model=False, inline=False, external_threshold=102...
input: name='x' type=dtype('float32') shape=[2, 2, 3, 4]
init: name='init7_s2_2_2' type=int64 shape=(2,) -- array([2, 2])      -- _aten_upsample_output_size.batch_channel
init: name='init7_s2_6_8' type=int64 shape=(2,) -- array([6, 8])      -- _aten_upsample_output_size.rsize
Concat(init7_s2_2_2, init7_s2_6_8, axis=0) -> _onx_concat0
Resize(x, , , _onx_concat0, coordinate_transformation_mode=b'pytorch_half_pixel', mode=b'linear', nearest_mode=b'floor') -> output_0
output: name='output_0' type=dtype('float32') shape=[2, 2, 6, 8]

custom-tracing¶

opset: domain='' version=18
doc_string: large_model=False, inline=False, external_threshold=102...
input: name='x' type=dtype('float32') shape=[2, 2, 3, 4]
init: name='init7_s2_2_2' type=int64 shape=(2,) -- array([2, 2])      -- _aten_upsample_output_size.batch_channel
init: name='init7_s2_6_8' type=int64 shape=(2,) -- array([6, 8])      -- _aten_upsample_output_size.rsize
Concat(init7_s2_2_2, init7_s2_6_8, axis=0) -> _onx_concat0
Resize(x, , , _onx_concat0, coordinate_transformation_mode=b'pytorch_half_pixel', mode=b'linear', nearest_mode=b'floor') -> output
output: name='output' type=dtype('float32') shape=['d_output_0', 'd_output_1', 'd_output_2', 'd_output_3']

dynamo-ir¶

opset: domain='pkg.onnxscript.torch_lib.common' version=1
opset: domain='' version=18
input: name='x' type=dtype('float32') shape=[2, 2, 3, 4]
Constant(value_floats=[1.0,1.0,2.0,2.0]) -> val_0
Resize(x, , val_0, keep_aspect_ratio_policy=b'stretch', antialias=0, extrapolation_value=0.00, exclude_outside=0, nearest_mode=b'floor', coordinate_transformation_mode=b'pytorch_half_pixel', cubic_coeff_a=-0.75, mode=b'linear') -> upsample_bilinear2d
output: name='upsample_bilinear2d' type=dtype('float32') shape=[2, 2, 6, 8]

script¶

opset: domain='' version=17
input: name='x' type=dtype('float32') shape=[2, 2, 3, 4]
Constant(value=[1.0, 1.0,...) -> /Constant_output_0
Resize(x, , /Constant_output_0, coordinate_transformation_mode=b'half_pixel', cubic_coeff_a=-0.75, mode=b'linear', nearest_mode=b'floor') -> 5
output: name='5' type=dtype('float32') shape=[2, 2, 6, 8]

AtenNonZero¶

forward¶

def forward(self, x):
    y = torch.nonzero(x)
    return y

custom-fallback¶

opset: domain='' version=18
doc_string: large_model=False, inline=False, external_threshold=102...
input: name='x' type=dtype('float32') shape=[3, 4]
NonZero(x) -> _onx_nonzero0
  Transpose(_onx_nonzero0, perm=[1,0]) -> output_0
output: name='output_0' type=dtype('int64') shape=['u1', 2]

custom-tracing¶

opset: domain='' version=18
doc_string: large_model=False, inline=False, external_threshold=102...
input: name='x' type=dtype('float32') shape=[3, 4]
NonZero(x) -> _onx_nonzero0
  Transpose(_onx_nonzero0, perm=[1,0]) -> output
output: name='output' type=dtype('int64') shape=['d_output_0', 'd_output_1']

dynamo-ir¶

opset: domain='pkg.onnxscript.torch_lib.common' version=1
opset: domain='pkg.onnxscript.torch_lib' version=1
opset: domain='' version=18
input: name='x' type=dtype('float32') shape=[3, 4]
NonZero(x) -> tmp
  Transpose(tmp, perm=[1,0]) -> nonzero
output: name='nonzero' type=dtype('int64') shape=['', 2]

script¶

opset: domain='' version=17
input: name='onnx::NonZero_0' type=dtype('float32') shape=[3, 4]
NonZero(onnx::NonZero_0) -> /NonZero_output_0
  Transpose(/NonZero_output_0, perm=[1,0]) -> 2
output: name='2' type=dtype('int64') shape=['Transpose2_dim_0', 2]

AtenNonZeroTuple¶

forward¶

def forward(self, x):
    y = torch.nonzero(x, as_tuple=True)
    return y[0], y[1]

custom-fallback¶

opset: domain='' version=18
doc_string: large_model=False, inline=False, external_threshold=102...
input: name='x' type=dtype('float32') shape=[3, 4]
init: name='init7_s_0' type=int64 shape=() -- array([0])              -- DynamoInterpreter.getitem.1##shape_type_compute._cast_inputs.1(ge)
init: name='init7_s_1' type=int64 shape=() -- array([1])              -- DynamoInterpreter.getitem.1
NonZero(x) -> _onx_nonzero0
  SplitToSequence(_onx_nonzero0, axis=0, keepdims=0) -> nonzero_numpy
    SequenceAt(nonzero_numpy, init7_s_0) -> output_0
SequenceAt(nonzero_numpy, init7_s_1) -> output_1
output: name='output_0' type=dtype('int64') shape=['u1']
output: name='output_1' type=dtype('int64') shape=['u1']

custom-tracing¶

opset: domain='' version=18
doc_string: large_model=False, inline=False, external_threshold=102...
input: name='x' type=dtype('float32') shape=[3, 4]
init: name='init7_s_0' type=int64 shape=() -- array([0])              -- DynamoInterpreter.getitem.1
init: name='init7_s_1' type=int64 shape=() -- array([1])              -- DynamoInterpreter.getitem.1
NonZero(x) -> _onx_nonzero0
  SplitToSequence(_onx_nonzero0, axis=0, keepdims=0) -> nonzero
    SequenceAt(nonzero, init7_s_0) -> output_0
SequenceAt(nonzero, init7_s_1) -> output_1
output: name='output_0' type=dtype('int64') shape=['d_output_0_0']
output: name='output_1' type=dtype('int64') shape=['d_output_1_0']

dynamo-ir¶

opset: domain='pkg.onnxscript.torch_lib.common' version=1
opset: domain='pkg.onnxscript.torch_lib' version=1
opset: domain='' version=18
input: name='x' type=dtype('float32') shape=[3, 4]
Constant(value_ints=[1]) -> unbind_axis
NonZero(x) -> tmp
  Transpose(tmp, perm=[1,0]) -> nonzero
    Split(nonzero, axis=1, num_outputs=2) -> unbind_split_0, unbind_split_1
  Squeeze(unbind_split_0, unbind_axis) -> unbind_split_0_squeeze
    Identity(unbind_split_0_squeeze) -> getitem
  Squeeze(unbind_split_1, unbind_axis) -> unbind_split_1_squeeze
    Identity(unbind_split_1_squeeze) -> getitem_1
output: name='getitem' type=dtype('int64') shape=['']
output: name='getitem_1' type=dtype('int64') shape=['']

script¶

opset: domain='' version=17
input: name='onnx::NonZero_0' type=dtype('float32') shape=[3, 4]
Constant(value=[1, 1]) -> /Constant_output_0
NonZero(onnx::NonZero_0) -> /NonZero_output_0
  Transpose(/NonZero_output_0, perm=[1,0]) -> /Transpose_output_0
  Split(/Transpose_output_0, /Constant_output_0, axis=1) -> /Split_output_0, /Split_output_1
Constant(value=[1]) -> /Constant_1_output_0
  Squeeze(/Split_output_0, /Constant_1_output_0) -> 7
Constant(value=[1]) -> /Constant_2_output_0
  Squeeze(/Split_output_1, /Constant_2_output_0) -> 9
output: name='7' type=dtype('int64') shape=['Squeeze7_dim_0']
output: name='9' type=dtype('int64') shape=['Squeeze7_dim_0']

AtenRollPos¶

forward¶

def forward(self, x):
    return torch.roll(x, 1, -1)

custom-fallback¶

opset: domain='' version=18
doc_string: large_model=False, inline=False, external_threshold=102...
input: name='x' type=dtype('float32') shape=[2, 3, 4]
init: name='init7_s1_-1' type=int64 shape=(1,) -- array([-1])         -- Opset.make_node.1/Shape##Opset.make_node.1/Shape##Opset.make_node.1/Shape##Opset.make_node.1/Shape
init: name='init7_s1_4' type=int64 shape=(1,) -- array([4])           -- Opset.make_node.1/Shape
init: name='init7_s1_0' type=int64 shape=(1,) -- array([0])           -- Opset.make_node.1/Shape
Slice(x, init7_s1_-1, init7_s1_4, init7_s1_-1) -> _onx_slice0
Slice(x, init7_s1_0, init7_s1_-1, init7_s1_-1) -> _onx_slice02
  Concat(_onx_slice0, _onx_slice02, axis=-1) -> output_0
output: name='output_0' type=dtype('float32') shape=[2, 3, 4]

custom-tracing¶

opset: domain='' version=18
doc_string: large_model=False, inline=False, external_threshold=102...
input: name='x' type=dtype('float32') shape=[2, 3, 4]
init: name='init7_s1_-1' type=int64 shape=(1,) -- array([-1])         -- Opset.make_node.1/Shape##Opset.make_node.1/Shape##Opset.make_node.1/Shape##Opset.make_node.1/Shape
init: name='init7_s1_4' type=int64 shape=(1,) -- array([4])           -- Opset.make_node.1/Shape
init: name='init7_s1_0' type=int64 shape=(1,) -- array([0])           -- Opset.make_node.1/Shape
Slice(x, init7_s1_-1, init7_s1_4, init7_s1_-1) -> _onx_slice0
Slice(x, init7_s1_0, init7_s1_-1, init7_s1_-1) -> _onx_slice02
  Concat(_onx_slice0, _onx_slice02, axis=-1) -> output
output: name='output' type=dtype('float32') shape=[2, 3, 4]

dynamo-ir¶

opset: domain='pkg.onnxscript.torch_lib.common' version=1
opset: domain='' version=18
opset: domain='pkg.onnxscript.torch_lib' version=1
input: name='x' type=dtype('float32') shape=[2, 3, 4]
Constant(value_ints=[0]) -> tmp_4
Constant(value=[-1]) -> dim_tensor
Constant(value=[1]) -> shift_tensor
Constant(value=[3]) -> slice_length_3
  Slice(x, tmp_4, slice_length_3, dim_tensor) -> suffix
Constant(value=[24]) -> tmp_6
  Slice(x, slice_length_3, tmp_6, dim_tensor) -> prefix
    Concat(prefix, suffix, axis=-1) -> roll
output: name='roll' type=dtype('float32') shape=[2, 3, 4]

script¶

opset: domain='' version=17
input: name='onnx::Slice_0' type=dtype('float32') shape=[2, 3, 4]
Constant(value=[-1]) -> /Constant_output_0
Constant(value=[-1]) -> /Constant_1_output_0
Constant(value=[922337203...) -> /Constant_2_output_0
  Slice(onnx::Slice_0, /Constant_1_output_0, /Constant_2_output_0, /Constant_output_0) -> /Slice_output_0
Constant(value=[-1]) -> /Constant_3_output_0
Constant(value=[0]) -> /Constant_4_output_0
Constant(value=[-1]) -> /Constant_5_output_0
  Slice(onnx::Slice_0, /Constant_4_output_0, /Constant_5_output_0, /Constant_3_output_0) -> /Slice_1_output_0
    Concat(/Slice_output_0, /Slice_1_output_0, axis=-1) -> 9
output: name='9' type=dtype('float32') shape=[2, 3, 4]

AtenRollRelu¶

forward¶

def forward(self, x):
    return torch.relu(torch.roll(x, -1, -1))

custom-fallback¶

opset: domain='' version=18
doc_string: large_model=False, inline=False, external_threshold=102...
input: name='x' type=dtype('float32') shape=[2, 3, 4]
init: name='init7_s1_1' type=int64 shape=(1,) -- array([1])           -- Opset.make_node.1/Shape##Opset.make_node.1/Shape
init: name='init7_s1_4' type=int64 shape=(1,) -- array([4])           -- Opset.make_node.1/Shape
init: name='init7_s1_-1' type=int64 shape=(1,) -- array([-1])         -- Opset.make_node.1/Shape##Opset.make_node.1/Shape
init: name='init7_s1_0' type=int64 shape=(1,) -- array([0])           -- Opset.make_node.1/Shape
Slice(x, init7_s1_1, init7_s1_4, init7_s1_-1) -> _onx_slice0
Slice(x, init7_s1_0, init7_s1_1, init7_s1_-1) -> _onx_slice02
  Concat(_onx_slice0, _onx_slice02, axis=-1) -> _onx_concat0
    Relu(_onx_concat0) -> output_0
output: name='output_0' type=dtype('float32') shape=[2, 3, 4]

custom-tracing¶

opset: domain='' version=18
doc_string: large_model=False, inline=False, external_threshold=102...
input: name='x' type=dtype('float32') shape=[2, 3, 4]
init: name='init7_s1_1' type=int64 shape=(1,) -- array([1])           -- Opset.make_node.1/Shape##Opset.make_node.1/Shape
init: name='init7_s1_4' type=int64 shape=(1,) -- array([4])           -- Opset.make_node.1/Shape
init: name='init7_s1_-1' type=int64 shape=(1,) -- array([-1])         -- Opset.make_node.1/Shape##Opset.make_node.1/Shape
init: name='init7_s1_0' type=int64 shape=(1,) -- array([0])           -- Opset.make_node.1/Shape
Slice(x, init7_s1_1, init7_s1_4, init7_s1_-1) -> _onx_slice0
Slice(x, init7_s1_0, init7_s1_1, init7_s1_-1) -> _onx_slice02
  Concat(_onx_slice0, _onx_slice02, axis=-1) -> _onx_concat0
    Relu(_onx_concat0) -> output
output: name='output' type=dtype('float32') shape=[2, 3, 4]

dynamo-ir¶

opset: domain='pkg.onnxscript.torch_lib.common' version=1
opset: domain='' version=18
opset: domain='pkg.onnxscript.torch_lib' version=1
input: name='x' type=dtype('float32') shape=[2, 3, 4]
Constant(value_ints=[0]) -> tmp_4
Constant(value=[-1]) -> dim_tensor
Constant(value=[-1]) -> shift_tensor
Constant(value=[1]) -> slice_length_3
  Slice(x, tmp_4, slice_length_3, dim_tensor) -> suffix
Constant(value=[24]) -> tmp_6
  Slice(x, slice_length_3, tmp_6, dim_tensor) -> prefix
    Concat(prefix, suffix, axis=-1) -> roll
      Relu(roll) -> relu
output: name='relu' type=dtype('float32') shape=[2, 3, 4]

script¶

opset: domain='' version=17
input: name='onnx::Slice_0' type=dtype('float32') shape=[2, 3, 4]
Constant(value=[-1]) -> /Constant_output_0
Constant(value=[1]) -> /Constant_1_output_0
Constant(value=[922337203...) -> /Constant_2_output_0
  Slice(onnx::Slice_0, /Constant_1_output_0, /Constant_2_output_0, /Constant_output_0) -> /Slice_output_0
Constant(value=[-1]) -> /Constant_3_output_0
Constant(value=[0]) -> /Constant_4_output_0
Constant(value=[1]) -> /Constant_5_output_0
  Slice(onnx::Slice_0, /Constant_4_output_0, /Constant_5_output_0, /Constant_3_output_0) -> /Slice_1_output_0
    Concat(/Slice_output_0, /Slice_1_output_0, axis=-1) -> /Concat_output_0
      Relu(/Concat_output_0) -> 10
output: name='10' type=dtype('float32') shape=[2, 3, 4]

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

custom-fallback¶

opset: domain='' version=18
doc_string: large_model=False, inline=False, external_threshold=102...
input: name='x' type=dtype('float32') shape=[4, 3]
input: name='lx_0' type=dtype('float32') shape=[4, 1]
input: name='lx_1' type=dtype('float32') shape=[4, 2]
init: name='b_buff' type=float32 shape=(1,) -- array([0.5], dtype=float32)-- DynamoInterpret.placeholder.0
init: name='init7_s1_1' type=int64 shape=(1,) -- array([1])           -- Opset.make_node.1/Shape
init: name='linear.weight' type=float32 shape=(1, 3) -- array([-0.443,  0.327, -0.031], dtype=float32)-- DynamoInterpret.placeholder.1/P(linear.weight)
init: name='linear.bias' type=float32 shape=(1,) -- array([-0.457], dtype=float32)-- DynamoInterpret.placeholder.1/P(linear.bias)
Gemm(x, linear.weight, linear.bias, transB=1) -> linear
  Sigmoid(linear) -> sigmoid
    Sub(sigmoid, b_buff) -> sub
ReduceSum(lx_1, init7_s1_1, keepdims=1) -> sum_1
  Mul(lx_0, sum_1) -> mul
    Add(sub, mul) -> output_0
output: name='output_0' type=dtype('float32') shape=[4, 1]

custom-tracing¶

FAILED

Type is unknown for result 'lx', known_types={'x': 1, 'linear.weight': 1, 'linear.bias': 1, '_sub_Linear_input_1': 1, '_sub_Linear__onx_transpose0': 1, '_sub_Linear__onx_matmul0': 1, '_sub_Linear_linear': 1, '_sub_Linear_output': 1, 'linear': 1, 'sigmoid': 1, 'buff': 1, 'sub': 1}
--DEBUG--
[GraphBuilder-LFS] Message starts, there are 3 initializers, 8 nodes, 2 inputs, 2 outputs.
--PARAMETERS--
dynamic_examples=
--SHAPE--
dynamic_examples=
dynamic_objects=
dynamic_objects_rev=
dynamic_dimensions_source={}
dynamic_alias={}
dynamic_shapes=None
_known_value_shape={}
_known_types={'_sub_Linear__onx_matmul0': 1,
 '_sub_Linear__onx_transpose0': 1,
 '_sub_Linear_input_1': 1,
 '_sub_Linear_linear': 1,
 '_sub_Linear_output': 1,
 'buff': 1,
 'linear': 1,
 'linear.bias': 1,
 'linear.weight': 1,
 'sigmoid': 1,
 'sub': 1,
 'x': 1}
_known_shapes={'_sub_Linear__onx_matmul0': (4, 1),
 '_sub_Linear__onx_transpose0': (3, 1),
 '_sub_Linear_input_1': (4, 3),
 '_sub_Linear_linear': (4, 1),
 '_sub_Linear_output': (4, 1),
 'buff': (1,),
 'linear': (4, 1),
 'linear.bias': (1,),
 'linear.weight': (1, 3),
 'sigmoid': (4, 1),
 'sub': (4, 1),
 'x': (4, 3)}
_known_constants=['_sub_Linear__onx_transpose0', 'buff', 'linear.bias', 'linear.weight']
_known_ranks={}
--TORCH-USERS--
add -> {output}
buff -> {sub}
linear -> {sigmoid}
lx -> {add}
sigmoid -> {sub}
sub -> {add}
x -> {linear}
--TORCH-SHAPES--
x: ('run_node', (('example_value', torch.float32, torch.Size([4, 3])), '')) --- 1:2:(4, 3):
lx: ('run_node', ('', '')) --- :::
linear: ('run_node', ('', '')) --- 1:2:(4, 1):
sigmoid: ('run_node', ('', '')) --- 1:2:(4, 1):
buff: ('run_node', ('', '')) --- 1:1:(1,):
sub: ('run_node', ('', '')) --- 1:2:(4, 1):
add: ('run_node', ('', '')) --- :::
--ONNX--
-- process.graph_module --
graph():
    %x : [num_users=1] = placeholder[target=x]
    %lx : list | torch.Tensor [num_users=1] = 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 = {})
    %add : [num_users=1] = call_function[target=operator.add](args = (%sub, %lx), kwargs = {})
    return add
-- process.progress --
node 6/8 target=<built-in function add>
--
[GraphBuilder-LFS.make_tensor_input] x[1:4x3]
[GraphBuilder-LFS.make_tensor_input] lx[0:]
[GraphBuilder-LFS.make_initializer] linear.weight[torch.float32:torch.float32:[0.18872031569480896, -0.22813834249973297, 0.44105055928230286]] - SOURCE: GraphBuilder.make_nodes/fromlinear.weight##DynamoInterpret.get_attr.1/P(linear.weight)
[GraphBuilder-LFS.make_initializer] linear.bias[torch.float32:torch.float32:[-0.26602017879486084]] - SOURCE: GraphBuilder.make_nodes/fromlinear.bias##DynamoInterpret.get_attr.1/P(linear.bias)
[GraphBuilder-LFS.make_initializer] buff[torch.float32:torch.float32:[0.5]] - SOURCE: DynamoInterpret.get_attr.0
[GraphBuilder-LFS.make_node] .make_nodes     [#:#   ] Identity:['x']->['_sub_Linear_input_1']
[GraphBuilder-LFS.make_node] linear          [#:#   ] Transpose:['linear.weight']->['_sub_Linear__onx_transpose0']
[GraphBuilder-LFS.make_node] Opset           [##:#  ] MatMul:['_sub_Linear_input_1', '_sub_Linear__onx_transpose0']->['_sub_Linear__onx_matmul0']
[GraphBuilder-LFS.make_node] Opset2          [##:#  ] Add:['_sub_Linear__onx_matmul0', 'linear.bias']->['_sub_Linear_linear']
[GraphBuilder-LFS.make_node] .output         [#:#   ] Identity:['_sub_Linear_linear']->['_sub_Linear_output']
[GraphBuilder-LFS.make_node] .make_nodes2    [#:#   ] Identity:['_sub_Linear_output']->['linear']
[GraphBuilder-LFS.make_node] Opset3          [#:#   ] Sigmoid:['linear']->['sigmoid']
[GraphBuilder-LFS.make_node] sub             [##:#  ] Sub:['sigmoid', 'buff']->['sub']
[GraphBuilder-LFS] Message completed, there are 3 initializers, 8 nodes, 2 inputs, 2 outputs..

dynamo-ir¶

opset: domain='pkg.onnxscript.torch_lib.common' version=1
opset: domain='' version=18
input: name='x' type=dtype('float32') shape=[4, 3]
input: name='lx_0' type=dtype('float32') shape=[4, 1]
input: name='lx_1' type=dtype('float32') shape=[4, 2]
init: name='linear.bias' type=float32 shape=(1,) -- array([0.339], dtype=float32)
init: name='buff' type=float32 shape=(1,) -- array([0.5], dtype=float32)
Constant(value=[1]) -> val_3
  ReduceSum(lx_1, val_3, noop_with_empty_axes=0, keepdims=1) -> sum_1
    Mul(lx_0, sum_1) -> mul
Constant(value=[[0.486663...) -> t
  Gemm(x, t, linear.bias, beta=1.00, transB=0, alpha=1.00, transA=0) -> addmm
    Sigmoid(addmm) -> sigmoid
      Sub(sigmoid, buff) -> sub
      Add(sub, mul) -> add
output: name='add' type=dtype('float32') shape=[4, 1]

script¶

opset: domain='' version=17
input: name='onnx::Gemm_0' type=dtype('float32') shape=[4, 3]
input: name='onnx::Mul_1' type=dtype('float32') shape=[4, 1]
input: name='onnx::ReduceSum_2' type=dtype('float32') shape=[4, 2]
init: name='buff' type=float32 shape=(1,) -- array([0.5], dtype=float32)
init: name='linear.weight' type=float32 shape=(1, 3) -- array([-0.3  ,  0.118,  0.517], dtype=float32)
init: name='linear.bias' type=float32 shape=(1,) -- array([-0.087], dtype=float32)
Constant(value=[1]) -> onnx::ReduceSum_6
  ReduceSum(onnx::ReduceSum_2, onnx::ReduceSum_6, keepdims=1) -> /ReduceSum_output_0
    Mul(onnx::Mul_1, /ReduceSum_output_0) -> /Mul_output_0
Gemm(onnx::Gemm_0, linear.weight, linear.bias, alpha=1.00, beta=1.00, transB=1) -> /linear/Gemm_output_0
  Sigmoid(/linear/Gemm_output_0) -> /Sigmoid_output_0
    Sub(/Sigmoid_output_0, buff) -> /Sub_output_0
      Add(/Sub_output_0, /Mul_output_0) -> 12
output: name='12' type=dtype('float32') shape=[4, 1]

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

custom-fallback¶

opset: domain='' version=18
doc_string: large_model=False, inline=False, external_threshold=102...
input: name='x' type=dtype('float32') shape=[4, 3]
input: name='lx_0' type=dtype('float32') shape=[4, 1]
input: name='lx_1' type=dtype('float32') shape=[4, 2]
init: name='b_buff' type=float32 shape=(1,) -- array([0.5], dtype=float32)-- DynamoInterpret.placeholder.0
init: name='init7_s1_1' type=int64 shape=(1,) -- array([1])           -- Opset.make_node.1/Shape
init: name='linear.weight' type=float32 shape=(1, 3) -- array([ 0.376, -0.151,  0.388], dtype=float32)-- DynamoInterpret.placeholder.1/P(linear.weight)
init: name='linear.bias' type=float32 shape=(1,) -- array([0.313], dtype=float32)-- DynamoInterpret.placeholder.1/P(linear.bias)
Gemm(x, linear.weight, linear.bias, transB=1) -> linear
  Sigmoid(linear) -> sigmoid
    Sub(sigmoid, b_buff) -> sub
ReduceSum(lx_1, init7_s1_1, keepdims=1) -> sum_1
  Mul(lx_0, sum_1) -> mul
    Add(sub, mul) -> output_0
output: name='output_0' type=dtype('float32') shape=[4, 1]

custom-tracing¶

FAILED

len(.) expects an integer, len needs to be replaced. You should use _len.

dynamo-ir¶

opset: domain='pkg.onnxscript.torch_lib.common' version=1
opset: domain='' version=18
input: name='x' type=dtype('float32') shape=[4, 3]
input: name='lx_0' type=dtype('float32') shape=[4, 1]
input: name='lx_1' type=dtype('float32') shape=[4, 2]
init: name='linear.bias' type=float32 shape=(1,) -- array([-0.334], dtype=float32)
init: name='buff' type=float32 shape=(1,) -- array([0.5], dtype=float32)
Constant(value=[1]) -> val_3
  ReduceSum(lx_1, val_3, noop_with_empty_axes=0, keepdims=1) -> sum_1
    Mul(lx_0, sum_1) -> mul
Constant(value=[[0.441906...) -> t
  Gemm(x, t, linear.bias, beta=1.00, transB=0, alpha=1.00, transA=0) -> addmm
    Sigmoid(addmm) -> sigmoid
      Sub(sigmoid, buff) -> sub
      Add(sub, mul) -> add
output: name='add' type=dtype('float32') shape=[4, 1]

script¶

opset: domain='' version=17
input: name='onnx::Gemm_0' type=dtype('float32') shape=[4, 3]
input: name='onnx::Mul_1' type=dtype('float32') shape=[4, 1]
input: name='onnx::ReduceSum_2' type=dtype('float32') shape=[4, 2]
init: name='buff' type=float32 shape=(1,) -- array([0.5], dtype=float32)
init: name='linear.weight' type=float32 shape=(1, 3) -- array([ 0.211, -0.33 ,  0.426], dtype=float32)
init: name='linear.bias' type=float32 shape=(1,) -- array([0.356], dtype=float32)
Constant(value=[1]) -> onnx::ReduceSum_6
  ReduceSum(onnx::ReduceSum_2, onnx::ReduceSum_6, keepdims=1) -> /ReduceSum_output_0
    Mul(onnx::Mul_1, /ReduceSum_output_0) -> /Mul_output_0
Gemm(onnx::Gemm_0, linear.weight, linear.bias, alpha=1.00, beta=1.00, transB=1) -> /linear/Gemm_output_0
  Sigmoid(/linear/Gemm_output_0) -> /Sigmoid_output_0
    Sub(/Sigmoid_output_0, buff) -> /Sub_output_0
      Add(/Sub_output_0, /Mul_output_0) -> 12
output: name='12' type=dtype('float32') shape=[4, 1]

ComplexPolar¶

forward¶

def forward(self, x, angle):
    return torch.polar(x, angle)

custom-fallback¶

FAILED

[ONNXRuntimeError] : 10 : INVALID_GRAPH : This is an invalid model. Type Error: Type 'tensor(complex64)' of input parameter (_onx_cast02) of operator (Mul) in node (polar5) is invalid.

custom-tracing¶

FAILED

[ONNXRuntimeError] : 10 : INVALID_GRAPH : This is an invalid model. Type Error: Type 'tensor(complex64)' of input parameter (_onx_cast02) of operator (Mul) in node (polar5) is invalid.

dynamo-ir¶

opset: domain='pkg.onnxscript.torch_lib.common' version=1
opset: domain='pkg.onnxscript.torch_lib' version=1
opset: domain='' version=18
input: name='x' type=dtype('float32') shape=[4, 4]
input: name='angle' type=dtype('float32') shape=[4, 4]
Constant(value=[-1]) -> int64_m1_1d
Cos(angle) -> tmp
  Mul(x, tmp) -> tmp_0
  Unsqueeze(tmp_0, int64_m1_1d) -> real
Sin(angle) -> tmp_1
  Mul(x, tmp_1) -> tmp_2
Constant(value=[-1]) -> int64_m1_1d_3
  Unsqueeze(tmp_2, int64_m1_1d_3) -> imag
    Concat(real, imag, axis=-1) -> polar
output: name='polar' type=dtype('float32') shape=[4, 4, 2]

script¶

FAILED

Exporting the operator 'aten::polar' to ONNX opset version 17 is not supported. Please feel free to request support or submit a pull request on PyTorch GitHub: https://github.com/pytorch/pytorch/issues.

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])

custom-fallback¶

opset: domain='' version=18
opset: domain='local_functions' version=1
doc_string: large_model=False, inline=False, external_threshold=102...
input: name='x' type=dtype('float32') shape=[5, 3]
init: name='init1_s_' type=float32 shape=() -- array([0.], dtype=float32)-- shape_type_compute._cast_inputs.1(gt_Scalar)
ReduceSum(x, keepdims=0) -> sum_1
  Greater(sum_1, init1_s_) -> gt
    If(gt, else_branch=G1, then_branch=G2) -> output_0
output: name='output_0' type=dtype('float32') shape=[5, 3]
----- subgraph ---- If - aten_cond - att.else_branch=G1 -- level=1 --  -> cond#0
false_graph_0[local_functions](x) -> cond#0
output: name='cond#0' type='NOTENSOR' shape=None
----- subgraph ---- If - aten_cond - att.then_branch=G2 -- level=1 --  -> cond#0
true_graph_0[local_functions](x) -> cond#0
output: name='cond#0' type='NOTENSOR' shape=None
----- function name=true_graph_0 domain=local_functions
----- doc_string: function_options=FunctionOptions(export_as_function=Tru...
opset: domain='' version=18
input: 'x'
Sin(x) -> output_0
output: name='output_0' type=? shape=?
----- function name=false_graph_0 domain=local_functions
----- doc_string: function_options=FunctionOptions(export_as_function=Tru...
opset: domain='' version=18
opset: domain='local_functions' version=1
input: 'x'
Cos(x) -> output_0
output: name='output_0' type=? shape=?

custom-tracing¶

FAILED

aten_meth_sum() missing 1 required positional argument: 'axis'

dynamo-ir¶

opset: domain='pkg.onnxscript.torch_lib.common' version=1
opset: domain='' version=18
opset: domain='pkg.torch.__subgraph__' version=1
input: name='x' type=dtype('float32') shape=[5, 3]
Constant(value=0.0) -> scalar_tensor_default
ReduceSum(x, noop_with_empty_axes=0, keepdims=0) -> sum_1
  Greater(sum_1, scalar_tensor_default) -> gt
    If(gt, then_branch=G1, else_branch=G2) -> getitem
output: name='getitem' type=dtype('float32') shape=[5, 3]
----- subgraph ---- If - node_If_4 - att.then_branch=G1 -- level=1 --  -> sin_true_graph_0
Sin(x) -> sin_true_graph_0
output: name='sin_true_graph_0' type=dtype('float32') shape=[5, 3]
----- subgraph ---- If - node_If_4 - att.else_branch=G2 -- level=1 --  -> cos_false_graph_0
Cos(x) -> cos_false_graph_0
output: name='cos_false_graph_0' type=dtype('float32') shape=[5, 3]

script¶

FAILED

Detected that you are using FX to torch.jit.trace a dynamo-optimized function. This is not supported at the moment.

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])

custom-fallback¶

opset: domain='' version=18
opset: domain='local_functions' version=1
doc_string: large_model=False, inline=False, external_threshold=102...
input: name='x' type=dtype('float32') shape=[5, 3]
input: name='y' type=dtype('float32') shape=[5, 3]
init: name='init1_s_' type=float32 shape=() -- array([0.], dtype=float32)-- shape_type_compute._cast_inputs.1(gt_Scalar)
ReduceSum(x, keepdims=0) -> sum_1
  Greater(sum_1, init1_s_) -> gt
    If(gt, else_branch=G1, then_branch=G2) -> output_0, output_1
output: name='output_0' type=dtype('float32') shape=[5, 3]
output: name='output_1' type=dtype('float32') shape=[5, 3]
----- subgraph ---- If - aten_cond - att.else_branch=G1 -- level=1 --  -> cond#0,cond#1
false_graph_0[local_functions](x, y) -> cond#0, cond#1
output: name='cond#0' type='NOTENSOR' shape=None
output: name='cond#1' type='NOTENSOR' shape=None
----- subgraph ---- If - aten_cond - att.then_branch=G2 -- level=1 --  -> cond#0,cond#1
true_graph_0[local_functions](x, y) -> cond#0, cond#1
output: name='cond#0' type='NOTENSOR' shape=None
output: name='cond#1' type='NOTENSOR' shape=None
----- function name=true_graph_0 domain=local_functions
----- doc_string: function_options=FunctionOptions(export_as_function=Tru...
opset: domain='' version=18
input: 'x'
input: 'y'
Cos(x) -> cos
  Add(cos, y) -> output_1
Sin(x) -> output_0
output: name='output_0' type=? shape=?
output: name='output_1' type=? shape=?
----- function name=false_graph_0 domain=local_functions
----- doc_string: function_options=FunctionOptions(export_as_function=Tru...
opset: domain='' version=18
opset: domain='local_functions' version=1
input: 'x'
input: 'y'
Cos(x) -> output_0
Sin(x) -> sin
  Add(sin, y) -> output_1
output: name='output_0' type=? shape=?
output: name='output_1' type=? shape=?

custom-tracing¶

FAILED

aten_meth_sum() missing 1 required positional argument: 'axis'

dynamo-ir¶

FAILED

Failed to convert the exported program to an ONNX model. [96mThis is step 3/3[0m of exporting the model to ONNX. Next steps:
- If there is a missing ONNX function, implement it and register it to the registry.
- If there is an internal error during ONNX conversion, debug the error and summit a PR to PyTorch.
- Create an error report with `torch.onnx.export(..., report=True)`, and save the ExportedProgram as a pt2 file. Create an issue in the PyTorch GitHub repository against the [96m*onnx*[0m component. Attach the error report and the pt2 model.

## Exception summary

<class 'IndexError'>: tuple index out of range
⬆️
<class 'torch.onnx._internal.exporter._errors.ConversionError'>: Error when translating node %getitem_1 : [num_users=1] = call_function[target=operator.getitem](args = (%cond, 1), kwargs = {}). See the stack trace for more information.

(Refer to the full stack trace above for more information.)

script¶

FAILED

Detected that you are using FX to torch.jit.trace a dynamo-optimized function. This is not supported at the moment.

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])

custom-fallback¶

opset: domain='' version=18
opset: domain='local_functions' version=1
doc_string: large_model=False, inline=False, external_threshold=102...
input: name='x' type=dtype('float32') shape=[5, 3]
init: name='init1_s_' type=float32 shape=() -- array([0.], dtype=float32)-- shape_type_compute._cast_inputs.1(gt_Scalar)
ReduceSum(x, keepdims=0) -> sum_1
  Greater(sum_1, init1_s_) -> gt
    If(gt, else_branch=G1, then_branch=G2) -> output_0, output_1
output: name='output_0' type=dtype('float32') shape=[5, 3]
output: name='output_1' type=dtype('float32') shape=[5, 3]
----- subgraph ---- If - aten_cond - att.else_branch=G1 -- level=1 --  -> cond#0,cond#1
false_graph_0[local_functions](x) -> cond#0, cond#1
output: name='cond#0' type='NOTENSOR' shape=None
output: name='cond#1' type='NOTENSOR' shape=None
----- subgraph ---- If - aten_cond - att.then_branch=G2 -- level=1 --  -> cond#0,cond#1
true_graph_0[local_functions](x) -> cond#0, cond#1
output: name='cond#0' type='NOTENSOR' shape=None
output: name='cond#1' type='NOTENSOR' shape=None
----- function name=true_graph_0 domain=local_functions
----- doc_string: function_options=FunctionOptions(export_as_function=Tru...
opset: domain='' version=18
input: 'x'
Cos(x) -> output_1
Sin(x) -> output_0
output: name='output_0' type=? shape=?
output: name='output_1' type=? shape=?
----- function name=false_graph_0 domain=local_functions
----- doc_string: function_options=FunctionOptions(export_as_function=Tru...
opset: domain='' version=18
opset: domain='local_functions' version=1
input: 'x'
Cos(x) -> output_0
Sin(x) -> output_1
output: name='output_0' type=? shape=?
output: name='output_1' type=? shape=?

custom-tracing¶

FAILED

aten_meth_sum() missing 1 required positional argument: 'axis'

dynamo-ir¶

FAILED

Failed to convert the exported program to an ONNX model. [96mThis is step 3/3[0m of exporting the model to ONNX. Next steps:
- If there is a missing ONNX function, implement it and register it to the registry.
- If there is an internal error during ONNX conversion, debug the error and summit a PR to PyTorch.
- Create an error report with `torch.onnx.export(..., report=True)`, and save the ExportedProgram as a pt2 file. Create an issue in the PyTorch GitHub repository against the [96m*onnx*[0m component. Attach the error report and the pt2 model.

## Exception summary

<class 'IndexError'>: tuple index out of range
⬆️
<class 'torch.onnx._internal.exporter._errors.ConversionError'>: Error when translating node %getitem_1 : [num_users=1] = call_function[target=operator.getitem](args = (%cond, 1), kwargs = {}). See the stack trace for more information.

(Refer to the full stack trace above for more information.)

script¶

FAILED

Detected that you are using FX to torch.jit.trace a dynamo-optimized function. This is not supported at the moment.

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])

custom-fallback¶

opset: domain='' version=18
opset: domain='local_functions' version=1
doc_string: large_model=False, inline=False, external_threshold=102...
input: name='x' type=dtype('float32') shape=[1024, 1024]
init: name='init1_s_' type=float32 shape=() -- array([0.], dtype=float32)-- shape_type_compute._cast_inputs.1(gt_Scalar)
ReduceSum(x, keepdims=0) -> sum_1
  Greater(sum_1, init1_s_) -> gt
    If(gt, else_branch=G1, then_branch=G2) -> output_0
output: name='output_0' type=dtype('float32') shape=[1024, 1024]
----- subgraph ---- If - aten_cond - att.else_branch=G1 -- level=1 --  -> cond#0
false_graph_0[local_functions](x) -> cond#0
output: name='cond#0' type='NOTENSOR' shape=None
----- subgraph ---- If - aten_cond - att.then_branch=G2 -- level=1 --  -> cond#0
true_graph_0[local_functions](x) -> cond#0
output: name='cond#0' type='NOTENSOR' shape=None
----- function name=true_graph_0 domain=local_functions
----- doc_string: function_options=FunctionOptions(export_as_function=Tru...
opset: domain='' version=18
input: 'x'
Constant(value=[1024, 102...) -> init7_s2_1024_1024
  ConstantOfShape(init7_s2_1024_1024, value=[1.0]) -> ones
Sin(x) -> sin
  Sub(sin, ones) -> output_0
output: name='output_0' type=? shape=?
----- function name=false_graph_0 domain=local_functions
----- doc_string: function_options=FunctionOptions(export_as_function=Tru...
opset: domain='' version=18
opset: domain='local_functions' version=1
input: 'x'
Constant(value=[1, 1024]) -> init7_s2_1_1024
  ConstantOfShape(init7_s2_1_1024, value=[1.0]) -> ones
Cos(x) -> cos
  Add(cos, ones) -> output_0
output: name='output_0' type=? shape=?

custom-tracing¶

FAILED

aten_meth_sum() missing 1 required positional argument: 'axis'

dynamo-ir¶

opset: domain='pkg.onnxscript.torch_lib.common' version=1
opset: domain='' version=18
opset: domain='pkg.torch.__subgraph__' version=1
input: name='x' type=dtype('float32') shape=[1024, 1024]
Constant(value=0.0) -> scalar_tensor_default
ReduceSum(x, noop_with_empty_axes=0, keepdims=0) -> sum_1
  Greater(sum_1, scalar_tensor_default) -> gt
    If(gt, then_branch=G1, else_branch=G2) -> getitem
output: name='getitem' type=dtype('float32') shape=[1024, 1024]
----- subgraph ---- If - node_If_4 - att.then_branch=G1 -- level=1 --  -> sub_true_graph_0
Constant(value=[1024, 102...) -> val_1
Sin(x) -> sin
Constant(value=1.0) -> val_3
  Expand(val_3, val_1) -> ones
  Sub(sin, ones) -> sub_true_graph_0
output: name='sub_true_graph_0' type=dtype('float32') shape=[1024, 1024]
----- subgraph ---- If - node_If_4 - att.else_branch=G2 -- level=1 --  -> add_false_graph_0
Constant(value=[[1.0, 1.0...) -> ones_2
Cos(x) -> cos
  Add(cos, ones_2) -> add_false_graph_0
output: name='add_false_graph_0' type=dtype('float32') shape=[1024, 1024]

script¶

FAILED

Detected that you are using FX to torch.jit.trace a dynamo-optimized function. This is not supported at the moment.

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

custom-fallback¶

opset: domain='' version=18
opset: domain='local_functions.0' version=1
opset: domain='local_functions' version=1
doc_string: large_model=False, inline=False, external_threshold=102...
input: name='x' type=dtype('int64') shape=[2]
init: name='init7_s_0' type=int64 shape=() -- array([0])              -- shape_type_compute._cast_inputs.1(gt_Scalar)
init: name='submodule.weight' type=float32 shape=(1,) -- array([100.], dtype=float32)-- DynamoInterpret.placeholder.1/P(submodule.weight)
init: name='weight' type=float32 shape=(1,) -- array([42.], dtype=float32)-- DynamoInterpret.placeholder.1/P(weight)
ReduceSum(x, keepdims=0) -> sum_1
  Greater(sum_1, init7_s_0) -> gt
    If(gt, else_branch=G1, then_branch=G2) -> output_0
output: name='output_0' type=dtype('float32') shape=[2]
----- subgraph ---- If - aten_cond - att.else_branch=G1 -- level=1 --  -> cond#0
false_graph_0[local_functions](x, submodule.weight, weight) -> cond#0
output: name='cond#0' type='NOTENSOR' shape=None
----- subgraph ---- If - aten_cond - att.then_branch=G2 -- level=1 --  -> cond#0
true_graph_0[local_functions](x, submodule.weight, weight) -> cond#0
output: name='cond#0' type='NOTENSOR' shape=None
----- function name=true_graph_0 domain=local_functions.0
----- doc_string: function_options=FunctionOptions(export_as_function=Tru...
opset: domain='' version=18
input: 'p_submodule_weight'
input: 'x'
Cast(x, to=1) -> _onx_cast0
  Mul(_onx_cast0, p_submodule_weight) -> output_0
output: name='output_0' type=? shape=?
----- function name=false_graph_0 domain=local_functions.0
----- doc_string: function_options=FunctionOptions(export_as_function=Tru...
opset: domain='' version=18
opset: domain='local_functions.0' version=1
input: 'p_submodule_weight'
input: 'x'
Cast(x, to=1) -> _onx_cast0
  Div(_onx_cast0, p_submodule_weight) -> output_0
output: name='output_0' type=? shape=?
----- function name=true_graph_0 domain=local_functions
----- doc_string: function_options=FunctionOptions(export_as_function=Tru...
opset: domain='' version=18
opset: domain='local_functions.0' version=1
input: 'x'
input: 'p_submodule_weight'
input: 'p_weight'
Abs(x) -> abs_1
  ReduceSum(abs_1, keepdims=0) -> sum_1
Constant(value=100) -> init7_s_100
  Greater(sum_1, init7_s_100) -> gt
    If(gt, else_branch=G3, then_branch=G4) -> output_0
output: name='output_0' type=? shape=?
----- subgraph ---- If - aten_cond - att.else_branch=G3 -- level=1 --  -> cond#0
false_graph_0[local_functions.0](p_submodule_weight, x) -> cond#0
output: name='cond#0' type='NOTENSOR' shape=None
----- subgraph ---- If - aten_cond - att.then_branch=G4 -- level=1 --  -> cond#0
true_graph_0[local_functions.0](p_submodule_weight, x) -> cond#0
output: name='cond#0' type='NOTENSOR' shape=None
----- function name=false_graph_0 domain=local_functions
----- doc_string: function_options=FunctionOptions(export_as_function=Tru...
opset: domain='' version=18
opset: domain='local_functions.0' version=1
opset: domain='local_functions' version=1
input: 'x'
input: 'p_submodule_weight'
input: 'p_weight'
Cast(x, to=1) -> _onx_cast0
  Sub(_onx_cast0, p_weight) -> output_0
output: name='output_0' type=? shape=?

custom-tracing¶

FAILED

aten_meth_sum() missing 1 required positional argument: 'axis'

dynamo-ir¶

opset: domain='pkg.onnxscript.torch_lib.common' version=1
opset: domain='' version=18
opset: domain='pkg.torch.__subgraph__' version=1
opset: domain='pkg.onnxscript.torch_lib' version=1
input: name='x' type=dtype('int64') shape=[2]
init: name='weight' type=float32 shape=(1,) -- array([42.], dtype=float32)
init: name='submodule.weight' type=float32 shape=(1,) -- array([100.], dtype=float32)
Constant(value=0) -> val_0
ReduceSum(x, noop_with_empty_axes=0, keepdims=0) -> sum_1
  Greater(sum_1, val_0) -> gt
    If(gt, then_branch=G1, else_branch=G2) -> getitem
output: name='getitem' type=dtype('float32') shape=[2]
----- subgraph ---- If - node_If_3 - att.then_branch=G1 -- level=1 --  -> getitem_true_graph_0
Abs(x) -> abs_1
  ReduceSum(abs_1, noop_with_empty_axes=0, keepdims=0) -> sum_1_2
Constant(value=100) -> val_0_2
  Greater(sum_1_2, val_0_2) -> gt_2
    If(gt_2, then_branch=G3, else_branch=G4) -> getitem_true_graph_0
output: name='getitem_true_graph_0' type='NOTENSOR' shape=None
----- subgraph ---- If - node_If_4 - att.then_branch=G3 -- level=2 --  -> mul_true_graph_0__true_graph_0
Cast(x, to=1) -> convert_element_type_default
Mul(convert_element_type_default, submodule.weight) -> mul_true_graph_0__true_graph_0
output: name='mul_true_graph_0__true_graph_0' type=dtype('float32') shape=[2]
----- subgraph ---- If - node_If_4 - att.else_branch=G4 -- level=2 --  -> div_true_graph_0__false_graph_0
Cast(x, to=1) -> convert_element_type_default_2
Div(convert_element_type_default_2, submodule.weight) -> div_true_graph_0__false_graph_0
output: name='div_true_graph_0__false_graph_0' type=dtype('float32') shape=[2]
----- subgraph ---- If - node_If_4 - att.then_branch=G3 -- level=1 --  -> mul_true_graph_0__true_graph_0
Cast(x, to=1) -> convert_element_type_default
Mul(convert_element_type_default, submodule.weight) -> mul_true_graph_0__true_graph_0
output: name='mul_true_graph_0__true_graph_0' type=dtype('float32') shape=[2]
----- subgraph ---- If - node_If_4 - att.else_branch=G4 -- level=1 --  -> div_true_graph_0__false_graph_0
Cast(x, to=1) -> convert_element_type_default_2
Div(convert_element_type_default_2, submodule.weight) -> div_true_graph_0__false_graph_0
output: name='div_true_graph_0__false_graph_0' type=dtype('float32') shape=[2]
----- subgraph ---- If - node_If_3 - att.else_branch=G2 -- level=1 --  -> sub_false_graph_0
Cast(x, to=1) -> convert_element_type_default_3
Sub(convert_element_type_default_3, weight) -> sub_false_graph_0
output: name='sub_false_graph_0' type=dtype('float32') shape=[2]

script¶

FAILED

Detected that you are using FX to torch.jit.trace a dynamo-optimized function. This is not supported at the moment.

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])

custom-fallback¶

opset: domain='' version=18
opset: domain='local_functions.0' version=1
opset: domain='local_functions' version=1
doc_string: large_model=False, inline=False, external_threshold=102...
input: name='x' type=dtype('float32') shape=[5, 3]
init: name='init1_s_' type=float32 shape=() -- array([0.], dtype=float32)-- shape_type_compute._cast_inputs.1(gt_Scalar)
ReduceSum(x, keepdims=0) -> sum_1
  Greater(sum_1, init1_s_) -> gt
    If(gt, else_branch=G1, then_branch=G2) -> output_0
output: name='output_0' type=dtype('float32') shape=[5, 3]
----- subgraph ---- If - aten_cond - att.else_branch=G1 -- level=1 --  -> cond#0
false_graph_0[local_functions](x) -> cond#0
output: name='cond#0' type='NOTENSOR' shape=None
----- subgraph ---- If - aten_cond - att.then_branch=G2 -- level=1 --  -> cond#0
true_graph_0[local_functions](x) -> cond#0
output: name='cond#0' type='NOTENSOR' shape=None
----- function name=true_graph_0 domain=local_functions.0
----- doc_string: function_options=FunctionOptions(export_as_function=Tru...
opset: domain='' version=18
input: 'x'
Sin(x) -> output_0
output: name='output_0' type=? shape=?
----- function name=false_graph_0 domain=local_functions.0
----- doc_string: function_options=FunctionOptions(export_as_function=Tru...
opset: domain='' version=18
opset: domain='local_functions.0' version=1
input: 'x'
Cos(x) -> output_0
output: name='output_0' type=? shape=?
----- function name=true_graph_0 domain=local_functions
----- doc_string: function_options=FunctionOptions(export_as_function=Tru...
opset: domain='' version=18
opset: domain='local_functions.0' version=1
input: 'x'
Constant(value=0.0) -> init1_s_
ReduceSum(x, keepdims=0) -> sum_1
  Less(sum_1, init1_s_) -> lt
    If(lt, else_branch=G3, then_branch=G4) -> output_0
output: name='output_0' type=? shape=?
----- subgraph ---- If - aten_cond - att.else_branch=G3 -- level=1 --  -> cond#0
false_graph_0[local_functions.0](x) -> cond#0
output: name='cond#0' type='NOTENSOR' shape=None
----- subgraph ---- If - aten_cond - att.then_branch=G4 -- level=1 --  -> cond#0
true_graph_0[local_functions.0](x) -> cond#0
output: name='cond#0' type='NOTENSOR' shape=None
----- function name=false_graph_0 domain=local_functions
----- doc_string: function_options=FunctionOptions(export_as_function=Tru...
opset: domain='' version=18
opset: domain='local_functions.0' version=1
opset: domain='local_functions' version=1
input: 'x'
Neg(x) -> output_0
output: name='output_0' type=? shape=?

custom-tracing¶

FAILED

aten_meth_sum() missing 1 required positional argument: 'axis'

dynamo-ir¶

opset: domain='pkg.onnxscript.torch_lib.common' version=1
opset: domain='' version=18
opset: domain='pkg.torch.__subgraph__' version=1
input: name='x' type=dtype('float32') shape=[5, 3]
Constant(value=0.0) -> scalar_tensor_default
ReduceSum(x, noop_with_empty_axes=0, keepdims=0) -> sum_1
  Greater(sum_1, scalar_tensor_default) -> gt
    If(gt, then_branch=G1, else_branch=G2) -> getitem
output: name='getitem' type=dtype('float32') shape=[5, 3]
----- subgraph ---- If - node_If_4 - att.then_branch=G1 -- level=1 --  -> getitem_true_graph_0
Constant(value=0.0) -> scalar_tensor_default_2
ReduceSum(x, noop_with_empty_axes=0, keepdims=0) -> sum_1_2
  Less(sum_1_2, scalar_tensor_default_2) -> lt
    If(lt, then_branch=G3, else_branch=G4) -> getitem_true_graph_0
output: name='getitem_true_graph_0' type='NOTENSOR' shape=None
----- subgraph ---- If - node_If_4_2 - att.then_branch=G3 -- level=2 --  -> sin_true_graph_0__true_graph_0
Sin(x) -> sin_true_graph_0__true_graph_0
output: name='sin_true_graph_0__true_graph_0' type=dtype('float32') shape=[5, 3]
----- subgraph ---- If - node_If_4_2 - att.else_branch=G4 -- level=2 --  -> cos_true_graph_0__false_graph_0
Cos(x) -> cos_true_graph_0__false_graph_0
output: name='cos_true_graph_0__false_graph_0' type=dtype('float32') shape=[5, 3]
----- subgraph ---- If - node_If_4_2 - att.then_branch=G3 -- level=1 --  -> sin_true_graph_0__true_graph_0
Sin(x) -> sin_true_graph_0__true_graph_0
output: name='sin_true_graph_0__true_graph_0' type=dtype('float32') shape=[5, 3]
----- subgraph ---- If - node_If_4_2 - att.else_branch=G4 -- level=1 --  -> cos_true_graph_0__false_graph_0
Cos(x) -> cos_true_graph_0__false_graph_0
output: name='cos_true_graph_0__false_graph_0' type=dtype('float32') shape=[5, 3]
----- subgraph ---- If - node_If_4 - att.else_branch=G2 -- level=1 --  -> neg_false_graph_0
Neg(x) -> neg_false_graph_0
output: name='neg_false_graph_0' type=dtype('float32') shape=[5, 3]

script¶

FAILED

Detected that you are using FX to torch.jit.trace a dynamo-optimized function. This is not supported at the moment.

ControlFlowScan¶

forward¶

def forward(self, x):
    init = torch.zeros_like(x[0])
    carry, out = torch.ops.higher_order.scan(
        ControlFlowScan.add, [init], [x], dim=0, reverse=False, additional_inputs=[]
    )
    return carry

custom-fallback¶

opset: domain='' version=18
opset: domain='local_functions' version=1
doc_string: large_model=False, inline=False, external_threshold=102...
input: name='x' type=dtype('float32') shape=[3, 3]
init: name='init7_s1_3' type=int64 shape=(1,) -- array([3])           -- Opset.make_node.1/Shape
ConstantOfShape(init7_s1_3, value=[0.0]) -> zeros_like
  Scan(zeros_like, x, body=G1, num_scan_inputs=1, scan_input_directions=[0], scan_output_axes=[0], scan_output_directions=[0]) -> output_0, scan#1
output: name='output_0' type=dtype('float32') shape=[3]
----- subgraph ---- Scan - aten_scan - att.body=G1 -- level=1 -- init_0_zeros_like,scan_0_x -> output_0,output_1
input: name='init_0_zeros_like' type='NOTENSOR' shape=None
input: name='scan_0_x' type='NOTENSOR' shape=None
scan_combine_graph_0[local_functions](init_0_zeros_like, scan_0_x) -> output_0, output_1
output: name='output_0' type='NOTENSOR' shape=None
output: name='output_1' type='NOTENSOR' shape=None
----- function name=scan_combine_graph_0 domain=local_functions
----- doc_string: function_options=FunctionOptions(export_as_function=Tru...
opset: domain='' version=18
input: 'arg0_1'
input: 'arg1_1'
Add(arg0_1, arg1_1) -> output_0
  Identity(output_0) -> output_1
output: name='output_0' type=? shape=?
output: name='output_1' type=? shape=?

custom-tracing¶

FAILED

Unable to symbolically trace HigherOrderOperators

dynamo-ir¶

FAILED

Failed to convert the exported program to an ONNX model. [96mThis is step 3/3[0m of exporting the model to ONNX. Next steps:
- If there is a missing ONNX function, implement it and register it to the registry.
- If there is an internal error during ONNX conversion, debug the error and summit a PR to PyTorch.
- Create an error report with `torch.onnx.export(..., report=True)`, and save the ExportedProgram as a pt2 file. Create an issue in the PyTorch GitHub repository against the [96m*onnx*[0m component. Attach the error report and the pt2 model.

## Exception summary

<class 'torch.onnx._internal.exporter._errors.DispatchError'>: No ONNX function found for <torch._higher_order_ops.scan.ScanOp object at 0x7fa8cc5714f0>. Failure message: No decompositions registered for the real-valued input
⬆️
<class 'torch.onnx._internal.exporter._errors.ConversionError'>: Error when translating node %scan : [num_users=1] = call_function[target=torch.ops.higher_order.scan](args = (%scan_combine_graph_0, [%zeros_like], [%x], 0, False, []), kwargs = {}). See the stack trace for more information.

(Refer to the full stack trace above for more information.)

script¶

FAILED

could not find kernel for HigherOrderOperator scan at dispatch key DispatchKey.??? (resolved from DispatchKey.???)

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,
        reverse=False,
        additional_inputs=[],
    )
    return carry1, carry2, out1, out2

custom-fallback¶

opset: domain='' version=18
opset: domain='local_functions' version=1
doc_string: large_model=False, inline=False, external_threshold=102...
input: name='x' type=dtype('float32') shape=[3, 4]
init: name='init7_s1_4' type=int64 shape=(1,) -- array([4])           -- Opset.make_node.1/Shape##Opset.make_node.1/Shape
init: name='init1_s_' type=float32 shape=() -- array([2.], dtype=float32)-- shape_type_compute._cast_inputs.1(mul_Tensor)
init: name='init7_s1_1' type=int64 shape=(1,) -- array([1])           -- Opset.make_node.1/Shape
ConstantOfShape(init7_s1_4, value=[1.0]) -> ones_like
ConstantOfShape(init7_s1_4, value=[0.0]) -> zeros_like
Reshape(init1_s_, init7_s1_1) -> _onx_reshape0
  Mul(x, _onx_reshape0) -> _onx_mul0
  Scan(zeros_like, ones_like, x, _onx_mul0, body=G1, num_scan_inputs=2, scan_input_directions=[0,0], scan_output_axes=[0,0], scan_output_directions=[0,0]) -> output_0, output_1, output_2, output_3
output: name='output_0' type=dtype('float32') shape=[4]
output: name='output_1' type=dtype('float32') shape=[4]
output: name='output_2' type=dtype('float32') shape=[3, 4]
output: name='output_3' type=dtype('float32') shape=[3, 4]
----- subgraph ---- Scan - aten_scan - att.body=G1 -- level=1 -- init_0_zeros_like,init_1_ones_like,scan_0_x,scan_1_mul -> output_0,output_1,output_2,output_3
input: name='init_0_zeros_like' type='NOTENSOR' shape=None
input: name='init_1_ones_like' type='NOTENSOR' shape=None
input: name='scan_0_x' type='NOTENSOR' shape=None
input: name='scan_1_mul' type='NOTENSOR' shape=None
scan_combine_graph_0[local_functions](init_0_zeros_like, init_1_ones_like, scan_0_x, scan_1_mul) -> output_0, output_1, output_2, output_3
output: name='output_0' type='NOTENSOR' shape=None
output: name='output_1' type='NOTENSOR' shape=None
output: name='output_2' type='NOTENSOR' shape=None
output: name='output_3' type='NOTENSOR' shape=None
----- function name=scan_combine_graph_0 domain=local_functions
----- doc_string: function_options=FunctionOptions(export_as_function=Tru...
opset: domain='' version=18
input: 'arg0_1'
input: 'arg1_1'
input: 'arg2_1'
input: 'arg3_1'
Add(arg0_1, arg2_1) -> output_0
  Identity(output_0) -> output_2
Mul(arg1_1, arg3_1) -> output_1
  Identity(output_1) -> output_3
output: name='output_0' type=? shape=?
output: name='output_1' type=? shape=?
output: name='output_2' type=? shape=?
output: name='output_3' type=? shape=?

custom-tracing¶

FAILED

Unable to symbolically trace HigherOrderOperators

dynamo-ir¶

FAILED

Failed to convert the exported program to an ONNX model. [96mThis is step 3/3[0m of exporting the model to ONNX. Next steps:
- If there is a missing ONNX function, implement it and register it to the registry.
- If there is an internal error during ONNX conversion, debug the error and summit a PR to PyTorch.
- Create an error report with `torch.onnx.export(..., report=True)`, and save the ExportedProgram as a pt2 file. Create an issue in the PyTorch GitHub repository against the [96m*onnx*[0m component. Attach the error report and the pt2 model.

## Exception summary

<class 'torch.onnx._internal.exporter._errors.DispatchError'>: No ONNX function found for <torch._higher_order_ops.scan.ScanOp object at 0x7fa8cc5714f0>. Failure message: No decompositions registered for the real-valued input
⬆️
<class 'torch.onnx._internal.exporter._errors.ConversionError'>: Error when translating node %scan : [num_users=4] = call_function[target=torch.ops.higher_order.scan](args = (%scan_combine_graph_0, [%zeros_like, %ones_like], [%x, %mul], 0, False, []), kwargs = {}). See the stack trace for more information.

(Refer to the full stack trace above for more information.)

script¶

FAILED

could not find kernel for HigherOrderOperator scan at dispatch key DispatchKey.??? (resolved from DispatchKey.???)

ControlFlowScanCDist¶

forward¶

def forward(self, x):
    carry, out = torch.ops.higher_order.scan(
        ControlFlowScanCDist.dist,
        [x],
        [x],
        dim=0,
        reverse=False,
        additional_inputs=[],
    )
    return out

custom-fallback¶

opset: domain='' version=18
opset: domain='local_functions' version=1
doc_string: large_model=False, inline=False, external_threshold=102...
input: name='x' type=dtype('float32') shape=[3, 4]
Scan(x, x, body=G1, num_scan_inputs=1, scan_input_directions=[0], scan_output_axes=[0], scan_output_directions=[0]) -> scan#0, output_0
output: name='output_0' type=dtype('float32') shape=[3, 3]
----- subgraph ---- Scan - aten_scan - att.body=G1 -- level=1 -- init_0_x,scan_0_x -> output_0,output_1
input: name='init_0_x' type='NOTENSOR' shape=None
input: name='scan_0_x' type='NOTENSOR' shape=None
scan_combine_graph_0[local_functions](init_0_x, scan_0_x) -> output_0, output_1
output: name='output_0' type='NOTENSOR' shape=None
output: name='output_1' type='NOTENSOR' shape=None
----- function name=scan_combine_graph_0 domain=local_functions
----- doc_string: function_options=FunctionOptions(export_as_function=Tru...
opset: domain='' version=18
input: 'arg0_1'
input: 'arg1_1'
Constant(value=[1, -1]) -> init7_s2_1_-1
  Reshape(arg1_1, init7_s2_1_-1) -> reshape
    Sub(arg0_1, reshape) -> sub
      Mul(sub, sub) -> mul
Constant(value=[1]) -> init7_s1_1
  ReduceSum(mul, init7_s1_1, keepdims=0) -> sum_1
    Sqrt(sum_1) -> output_1
Identity(arg0_1) -> output_0
output: name='output_0' type=? shape=?
output: name='output_1' type=? shape=?

custom-tracing¶

FAILED

Unable to symbolically trace HigherOrderOperators

dynamo-ir¶

FAILED

Failed to convert the exported program to an ONNX model. [96mThis is step 3/3[0m of exporting the model to ONNX. Next steps:
- If there is a missing ONNX function, implement it and register it to the registry.
- If there is an internal error during ONNX conversion, debug the error and summit a PR to PyTorch.
- Create an error report with `torch.onnx.export(..., report=True)`, and save the ExportedProgram as a pt2 file. Create an issue in the PyTorch GitHub repository against the [96m*onnx*[0m component. Attach the error report and the pt2 model.

## Exception summary

<class 'torch.onnx._internal.exporter._errors.DispatchError'>: No ONNX function found for <torch._higher_order_ops.scan.ScanOp object at 0x7fa8cc5714f0>. Failure message: No decompositions registered for the real-valued input
⬆️
<class 'torch.onnx._internal.exporter._errors.ConversionError'>: Error when translating node %scan : [num_users=1] = call_function[target=torch.ops.higher_order.scan](args = (%scan_combine_graph_0, [%x], [%x], 0, False, []), kwargs = {}). See the stack trace for more information.

(Refer to the full stack trace above for more information.)

script¶

FAILED

could not find kernel for HigherOrderOperator scan at dispatch key DispatchKey.??? (resolved from DispatchKey.???)

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,
        reverse=False,
        additional_inputs=[y],
    )
    return out[1]

custom-fallback¶

opset: domain='' version=18
opset: domain='local_functions' version=1
doc_string: large_model=False, inline=False, external_threshold=102...
input: name='x' type=dtype('float32') shape=[3, 4]
init: name='c_lifted_tensor_0' type=float32 shape=(1,) -- array([0.], dtype=float32)-- DynamoInterpret.placeholder.0
Identity(x) -> hidden_input_scan_0_clone
Scan(c_lifted_tensor_0, x, body=G1, num_scan_inputs=1, scan_input_directions=[0], scan_output_axes=[0], scan_output_directions=[0]) -> scan#0, output_0
output: name='output_0' type=dtype('float32') shape=[3, 3]
----- subgraph ---- Scan - aten_scan - att.body=G1 -- level=1 -- init_0_detach_,scan_0_x -> output_0,output_1
input: name='init_0_detach_' type='NOTENSOR' shape=None
input: name='scan_0_x' type='NOTENSOR' shape=None
scan_combine_graph_0[local_functions](init_0_detach_, scan_0_x, hidden_input_scan_0_clone) -> output_0, output_1
output: name='output_0' type='NOTENSOR' shape=None
output: name='output_1' type='NOTENSOR' shape=None
----- function name=scan_combine_graph_0 domain=local_functions
----- doc_string: function_options=FunctionOptions(export_as_function=Tru...
opset: domain='' version=18
input: 'arg0_1'
input: 'arg1_1'
input: 'arg2_1'
Constant(value=[1, -1]) -> init7_s2_1_-1
  Reshape(arg1_1, init7_s2_1_-1) -> reshape
    Sub(arg2_1, reshape) -> sub
      Mul(sub, sub) -> mul
Constant(value=[1]) -> init7_s1_1
  ReduceSum(mul, init7_s1_1, keepdims=0) -> sum_1
    Sqrt(sum_1) -> output_1
Identity(arg0_1) -> output_0
output: name='output_0' type=? shape=?
output: name='output_1' type=? shape=?

custom-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'>,)

dynamo-ir¶

FAILED

Failed to convert the exported program to an ONNX model. [96mThis is step 3/3[0m of exporting the model to ONNX. Next steps:
- If there is a missing ONNX function, implement it and register it to the registry.
- If there is an internal error during ONNX conversion, debug the error and summit a PR to PyTorch.
- Create an error report with `torch.onnx.export(..., report=True)`, and save the ExportedProgram as a pt2 file. Create an issue in the PyTorch GitHub repository against the [96m*onnx*[0m component. Attach the error report and the pt2 model.

## Exception summary

<class 'torch.onnx._internal.exporter._errors.DispatchError'>: No ONNX function found for <torch._higher_order_ops.scan.ScanOp object at 0x7fa8cc5714f0>. Failure message: No decompositions registered for the real-valued input
⬆️
<class 'torch.onnx._internal.exporter._errors.ConversionError'>: Error when translating node %scan : [num_users=1] = call_function[target=torch.ops.higher_order.scan](args = (%scan_combine_graph_0, [%lift_fresh_copy], [%x], 0, False, [%clone]), kwargs = {}). See the stack trace for more information.

(Refer to the full stack trace above for more information.)

script¶

FAILED

could not find kernel for HigherOrderOperator scan at dispatch key DispatchKey.??? (resolved from DispatchKey.???)

ControlFlowScanCDistXY¶

forward¶

def forward(self, x, y):
    carry, out = torch.ops.higher_order.scan(
        ControlFlowScanCDistXY.dist,
        [y],
        [x],
        dim=0,
        reverse=False,
        additional_inputs=[],
    )
    return out

custom-fallback¶

opset: domain='' version=18
opset: domain='local_functions' version=1
doc_string: large_model=False, inline=False, external_threshold=102...
input: name='x' type=dtype('float32') shape=[3, 4]
input: name='y' type=dtype('float32') shape=[5, 4]
Scan(y, x, body=G1, num_scan_inputs=1, scan_input_directions=[0], scan_output_axes=[0], scan_output_directions=[0]) -> scan#0, output_0
output: name='output_0' type=dtype('float32') shape=[3, 5]
----- subgraph ---- Scan - aten_scan - att.body=G1 -- level=1 -- init_0_y,scan_0_x -> output_0,output_1
input: name='init_0_y' type='NOTENSOR' shape=None
input: name='scan_0_x' type='NOTENSOR' shape=None
scan_combine_graph_0[local_functions](init_0_y, scan_0_x) -> output_0, output_1
output: name='output_0' type='NOTENSOR' shape=None
output: name='output_1' type='NOTENSOR' shape=None
----- function name=scan_combine_graph_0 domain=local_functions
----- doc_string: function_options=FunctionOptions(export_as_function=Tru...
opset: domain='' version=18
input: 'arg0_1'
input: 'arg1_1'
Constant(value=[1, -1]) -> init7_s2_1_-1
  Reshape(arg1_1, init7_s2_1_-1) -> reshape
    Sub(arg0_1, reshape) -> sub
      Mul(sub, sub) -> mul
Constant(value=[1]) -> init7_s1_1
  ReduceSum(mul, init7_s1_1, keepdims=0) -> sum_1
    Sqrt(sum_1) -> output_1
Identity(arg0_1) -> output_0
output: name='output_0' type=? shape=?
output: name='output_1' type=? shape=?

custom-tracing¶

FAILED

Unable to symbolically trace HigherOrderOperators

dynamo-ir¶

FAILED

Failed to convert the exported program to an ONNX model. [96mThis is step 3/3[0m of exporting the model to ONNX. Next steps:
- If there is a missing ONNX function, implement it and register it to the registry.
- If there is an internal error during ONNX conversion, debug the error and summit a PR to PyTorch.
- Create an error report with `torch.onnx.export(..., report=True)`, and save the ExportedProgram as a pt2 file. Create an issue in the PyTorch GitHub repository against the [96m*onnx*[0m component. Attach the error report and the pt2 model.

## Exception summary

<class 'torch.onnx._internal.exporter._errors.DispatchError'>: No ONNX function found for <torch._higher_order_ops.scan.ScanOp object at 0x7fa8cc5714f0>. Failure message: No decompositions registered for the real-valued input
⬆️
<class 'torch.onnx._internal.exporter._errors.ConversionError'>: Error when translating node %scan : [num_users=1] = call_function[target=torch.ops.higher_order.scan](args = (%scan_combine_graph_0, [%y], [%x], 0, False, []), kwargs = {}). See the stack trace for more information.

(Refer to the full stack trace above for more information.)

script¶

FAILED

could not find kernel for HigherOrderOperator scan at dispatch key DispatchKey.??? (resolved from DispatchKey.???)

CreateFromShape¶

forward¶

def forward(self, x):
    y = torch.ones((x.shape[0], x.shape[1] + 1))
    return y

custom-fallback¶

opset: domain='' version=18
doc_string: large_model=False, inline=False, external_threshold=102...
input: name='x' type=dtype('float32') shape=[4, 4]
init: name='init7_s2_4_5' type=int64 shape=(2,) -- array([4, 5])      -- Opset.make_node.1/Shape
ConstantOfShape(init7_s2_4_5, value=[1.0]) -> output_0
output: name='output_0' type=dtype('float32') shape=[4, 5]

custom-tracing¶

FAILED

ones(): argument 'size' (position 1) must be tuple of ints, but found element of type CustomProxy at pos 0

dynamo-ir¶

opset: domain='pkg.onnxscript.torch_lib.common' version=1
opset: domain='' version=18
input: name='x' type=dtype('float32') shape=[4, 4]
Constant(value=[[1.0, 1.0...) -> ones
output: name='ones' type=dtype('float32') shape=[4, 5]

script¶

FAILED

Input mismatch, inputs[0]=(T1r2,) but names=[], model=CreateFromShape, export='script'

CreateFromShapeThroughFunction¶

forward¶

def forward(self, x):
    dy1 = CreateFromShapeThroughFunction.add_one(x.shape[1])
    y = torch.ones((x.shape[0], dy1))
    return y

custom-fallback¶

opset: domain='' version=18
doc_string: large_model=False, inline=False, external_threshold=102...
input: name='x' type=dtype('float32') shape=[4, 4]
init: name='init7_s2_4_5' type=int64 shape=(2,) -- array([4, 5])      -- Opset.make_node.1/Shape
ConstantOfShape(init7_s2_4_5, value=[1.0]) -> output_0
output: name='output_0' type=dtype('float32') shape=[4, 5]

custom-tracing¶

FAILED

ones(): argument 'size' (position 1) must be tuple of ints, but found element of type CustomProxy at pos 0

dynamo-ir¶

opset: domain='pkg.onnxscript.torch_lib.common' version=1
opset: domain='' version=18
input: name='x' type=dtype('float32') shape=[4, 4]
Constant(value=[[1.0, 1.0...) -> ones
output: name='ones' type=dtype('float32') shape=[4, 5]

script¶

FAILED

Input mismatch, inputs[0]=(T1r2,) but names=[], model=CreateFromShapeThroughFunction, export='script'

CropLastDimensionWithTensorContent¶

forward¶

def forward(self, x, shape):
    return x[..., : shape[0]]

custom-fallback¶

FAILED

None of the following options [ExportOptions(), ExportOptions(strict=False), ExportOptions(decomposition_table='default'), ExportOptions(strict=False, decomposition_table='default'), ExportOptions(dynamo=True), ExportOptions(decomposition_table='default', dynamo=True), ExportOptions(jit=True)] worked, args=(T1r3,T7r1), kwargs=None, exception=
-----
[(ExportOptions(),
  Unsupported('Dynamic slicing on data-dependent value is not supported\n\nfrom user code:\n   File "/home/xadupre/github/experimental-experiment/experimental_experiment/torch_interpreter/eval/model_cases.py", line 699, in forward\n    return x[..., : shape[0]]\n\nSet TORCH_LOGS="+dynamo" and TORCHDYNAMO_VERBOSE=1 for more information\n')),
 (ExportOptions(strict=False),
  GuardOnDataDependentSymNode('Could not extract specialized integer from data-dependent expression u0 (unhinted: u0).  (Size-like symbols: none)\n\nCaused by: (_export/non_strict_utils.py:581 in __torch_function__)\nFor more information, run with TORCH_LOGS="dynamic"\nFor extended logs when we create symbols, also add TORCHDYNAMO_EXTENDED_DEBUG_CREATE_SYMBOL="u0"\nIf you suspect the guard was triggered from C++, add TORCHDYNAMO_EXTENDED_DEBUG_CPP=1\nFor more debugging help, see https://docs.google.com/document/d/1HSuTTVvYH1pTew89Rtpeu84Ht3nQEFTYhAX3Ypa_xJs/edit?usp=sharing\n\nFor C++ stack trace, run with TORCHDYNAMO_EXTENDED_DEBUG_CPP=1\n\nThe following call raised this error:\n  File "/home/xadupre/github/experimental-experiment/experimental_experiment/torch_interpreter/eval/model_cases.py", line 699, in forward\n    return x[..., : shape[0]]\n')),
 (ExportOptions(decomposition_table='default'),
  Unsupported('Dynamic slicing on data-dependent value is not supported\n\nfrom user code:\n   File "/home/xadupre/github/experimental-experiment/experimental_experiment/torch_interpreter/eval/model_cases.py", line 699, in forward\n    return x[..., : shape[0]]\n\nSet TORCH_LOGS="+dynamo" and TORCHDYNAMO_VERBOSE=1 for more information\n')),
 (ExportOptions(strict=False, decomposition_table='default'),
  GuardOnDataDependentSymNode('Could not extract specialized integer from data-dependent expression u0 (unhinted: u0).  (Size-like symbols: none)\n\nCaused by: (_export/non_strict_utils.py:581 in __torch_function__)\nFor more information, run with TORCH_LOGS="dynamic"\nFor extended logs when we create symbols, also add TORCHDYNAMO_EXTENDED_DEBUG_CREATE_SYMBOL="u0"\nIf you suspect the guard was triggered from C++, add TORCHDYNAMO_EXTENDED_DEBUG_CPP=1\nFor more debugging help, see https://docs.google.com/document/d/1HSuTTVvYH1pTew89Rtpeu84Ht3nQEFTYhAX3Ypa_xJs/edit?usp=sharing\n\nFor C++ stack trace, run with TORCHDYNAMO_EXTENDED_DEBUG_CPP=1\n\nThe following call raised this error:\n  File "/home/xadupre/github/experimental-experiment/experimental_experiment/torch_interpreter/eval/model_cases.py", line 699, in forward\n    return x[..., : shape[0]]\n')),
 (ExportOptions(dynamo=True),
  Unsupported('Dynamic slicing on data-dependent value is not supported\n\nfrom user code:\n   File "/home/xadupre/github/experimental-experiment/experimental_experiment/torch_interpreter/eval/model_cases.py", line 699, in forward\n    return x[..., : shape[0]]\n\nSet TORCH_LOGS="+dynamo" and TORCHDYNAMO_VERBOSE=1 for more information\n')),
 (ExportOptions(decomposition_table='default', dynamo=True),
  Unsupported('Dynamic slicing on data-dependent value is not supported\n\nfrom user code:\n   File "/home/xadupre/github/experimental-experiment/experimental_experiment/torch_interpreter/eval/model_cases.py", line 699, in forward\n    return x[..., : shape[0]]\n\nSet TORCH_LOGS="+dynamo" and TORCHDYNAMO_VERBOSE=1 for more information\n')),
 (ExportOptions(jit=True),
  GuardOnDataDependentSymNode('Could not guard on data-dependent expression u0 < 0 (unhinted: u0 < 0).  (Size-like symbols: none)\n\nCaused by: (_decomp/decompositions.py:733 in slice_forward)\nFor more information, run with TORCH_LOGS="dynamic"\nFor extended logs when we create symbols, also add TORCHDYNAMO_EXTENDED_DEBUG_CREATE_SYMBOL="u0"\nIf you suspect the guard was triggered from C++, add TORCHDYNAMO_EXTENDED_DEBUG_CPP=1\nFor more debugging help, see https://docs.google.com/document/d/1HSuTTVvYH1pTew89Rtpeu84Ht3nQEFTYhAX3Ypa_xJs/edit?usp=sharing\n\nFor C++ stack trace, run with TORCHDYNAMO_EXTENDED_DEBUG_CPP=1\n\nThe following call raised this error:\n  File "<string>", line 1, in <lambda>\n\n\nWhile executing %slice_tensor : [num_users=1] = call_function[target=torch.ops.aten.slice.Tensor](args = (%x, 2, 0, %_local_scalar_dense_default, 1), kwargs = {})\nOriginal traceback:\nNone'))]

custom-tracing¶

opset: domain='' version=18
doc_string: large_model=False, inline=False, external_threshold=102...
input: name='x' type=dtype('float32') shape=[3, 4, 4]
input: name='shape' type=dtype('int64') shape=[1]
init: name='init7_s1_0' type=int64 shape=(1,) -- array([0])           -- Opset.make_node.1/Shape##Opset.make_node.1/Shape##Opset.make_node.1/Shape
init: name='getitem_1_axis' type=int64 shape=(1,) -- array([-1])      -- DynamoInterpreter._getitem_slice.axis.1
init: name='getitem_1_step' type=int64 shape=(1,) -- array([1])       -- DynamoInterpreter._getitem_slice.3
Gather(shape, init7_s1_0) -> _onx_gather0
  Squeeze(_onx_gather0, init7_s1_0) -> getitem
    Unsqueeze(getitem, init7_s1_0) -> _onx_unsqueeze0
      Slice(x, init7_s1_0, _onx_unsqueeze0, getitem_1_axis, getitem_1_step) -> output
output: name='output' type=dtype('float32') shape=['d_output_0', 'd_output_1', 'd_output_2']

dynamo-ir¶

FAILED

Failed to export the model with torch.export. [96mThis is step 1/3[0m of exporting the model to ONNX. Next steps:
- Modify the model code for `torch.export.export` to succeed. Refer to https://pytorch.org/docs/stable/generated/exportdb/index.html for more information.
- Debug `torch.export.export` and summit a PR to PyTorch.
- Create an issue in the PyTorch GitHub repository against the [96m*torch.export*[0m component and attach the full error stack as well as reproduction scripts.

## Exception summary

<class 'torch.fx.experimental.symbolic_shapes.GuardOnDataDependentSymNode'>: Could not extract specialized integer from data-dependent expression u0 (unhinted: u0).  (Size-like symbols: none)

Caused by: (_export/non_strict_utils.py:581 in __torch_function__)
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 "/home/xadupre/github/experimental-experiment/experimental_experiment/torch_interpreter/eval/model_cases.py", line 699, in forward
    return x[..., : shape[0]]


(Refer to the full stack trace above for more information.)

script¶

opset: domain='' version=17
input: name='onnx::Slice_0' type=dtype('float32') shape=[3, 4, 4]
input: name='onnx::Gather_1' type=dtype('int64') shape=[1]
Constant(value=0) -> /Constant_output_0
  Gather(onnx::Gather_1, /Constant_output_0, axis=0) -> /Gather_output_0
Constant(value=[2]) -> /Constant_1_output_0
Constant(value=[0]) -> /Constant_2_output_0
Constant(value=[0]) -> /Constant_3_output_0
  Unsqueeze(/Gather_output_0, /Constant_3_output_0) -> /Unsqueeze_output_0
Constant(value=[1]) -> /Constant_4_output_0
  Slice(onnx::Slice_0, /Constant_2_output_0, /Unsqueeze_output_0, /Constant_1_output_0, /Constant_4_output_0) -> 14
output: name='14' type=dtype('float32') shape=[3, 4, 'Slice14_dim_2']

CropLastDimensionWithTensorShape¶

forward¶

def forward(self, x, y):
    return x[..., : y.shape[0]]

custom-fallback¶

opset: domain='' version=18
doc_string: large_model=False, inline=False, external_threshold=102...
input: name='x' type=dtype('float32') shape=[3, 4, 4]
input: name='y' type=dtype('float32') shape=[2]
init: name='init7_s1_0' type=int64 shape=(1,) -- array([0])           -- Opset.make_node.1/Shape
init: name='init7_s1_2' type=int64 shape=(1,) -- array([2])           -- Opset.make_node.1/Shape##Opset.make_node.1/Shape
Slice(x, init7_s1_0, init7_s1_2, init7_s1_2) -> output_0
output: name='output_0' type=dtype('float32') shape=[3, 4, 2]

custom-tracing¶

opset: domain='' version=18
doc_string: large_model=False, inline=False, external_threshold=102...
input: name='x' type=dtype('float32') shape=[3, 4, 4]
input: name='y' type=dtype('float32') shape=[2]
init: name='init7_s1_0' type=int64 shape=(1,) -- array([0])           -- Opset.make_node.1/Shape##Opset.make_node.1/Shape##Opset.make_node.1/Shape
init: name='getitem_1_axis' type=int64 shape=(1,) -- array([-1])      -- DynamoInterpreter._getitem_slice.axis.1
init: name='getitem_1_step' type=int64 shape=(1,) -- array([1])       -- DynamoInterpreter._getitem_slice.3
Shape(y) -> getattr_1
  Gather(getattr_1, init7_s1_0) -> _onx_gather0
    Squeeze(_onx_gather0, init7_s1_0) -> getitem
      Unsqueeze(getitem, init7_s1_0) -> _onx_unsqueeze0
        Slice(x, init7_s1_0, _onx_unsqueeze0, getitem_1_axis, getitem_1_step) -> output
output: name='output' type=dtype('float32') shape=['d_output_0', 'd_output_1', 'd_output_2']

dynamo-ir¶

opset: domain='pkg.onnxscript.torch_lib.common' version=1
opset: domain='' version=18
input: name='x' type=dtype('float32') shape=[3, 4, 4]
input: name='y' type=dtype('float32') shape=[2]
Constant(value_ints=[1]) -> val_11
Constant(value=[0]) -> val_3
Constant(value=[2]) -> val_7
Constant(value=[2]) -> val_10
  Slice(x, val_3, val_7, val_10, val_11) -> slice_1
output: name='slice_1' type=dtype('float32') shape=[3, 4, 2]

script¶

FAILED

Input mismatch, inputs[0]=(T1r3,T1r1) but names=['onnx::Slice_0'], model=CropLastDimensionWithTensorShape, export='script'

InplaceAdd¶

forward¶

def forward(self, x):
    x += self.bias
    return x

custom-fallback¶

opset: domain='' version=18
doc_string: large_model=False, inline=False, external_threshold=102...
input: name='x' type=dtype('float32') shape=[3, 4]
init: name='c_bias' type=float32 shape=(1, 4) -- array([1., 1., 1., 1.], dtype=float32)-- DynamoInterpret.placeholder.0
Add(x, c_bias) -> output_0
output: name='output_0' type=dtype('float32') shape=[3, 4]

custom-tracing¶

opset: domain='' version=18
doc_string: large_model=False, inline=False, external_threshold=102...
input: name='x' type=dtype('float32') shape=[3, 4]
init: name='bias' type=float32 shape=(1, 4) -- array([1., 1., 1., 1.], dtype=float32)-- DynamoInterpret.get_attr.0
Add(x, bias) -> output
output: name='output' type=dtype('float32') shape=[3, 4]

dynamo-ir¶

opset: domain='pkg.onnxscript.torch_lib.common' version=1
opset: domain='' version=18
input: name='x' type=dtype('float32') shape=[3, 4]
init: name='bias' type=float32 shape=(1, 4) -- array([1., 1., 1., 1.], dtype=float32)
Add(x, bias) -> add
output: name='add' type=dtype('float32') shape=[3, 4]

script¶

opset: domain='' version=17
input: name='onnx::Add_0' type=dtype('float32') shape=[3, 4]
Constant(value=[[1.0, 1.0...) -> /Constant_output_0
  Add(onnx::Add_0, /Constant_output_0) -> 2
output: name='2' type=dtype('float32') shape=[3, 4]

InplaceAdd_¶

forward¶

def forward(self, x):
    x.add_(self.bias)
    return x

custom-fallback¶

opset: domain='' version=18
doc_string: large_model=False, inline=False, external_threshold=102...
input: name='x' type=dtype('float32') shape=[3, 4]
init: name='c_bias' type=float32 shape=(1, 4) -- array([1., 1., 1., 1.], dtype=float32)-- DynamoInterpret.placeholder.0
Add(x, c_bias) -> output_0
output: name='output_0' type=dtype('float32') shape=[3, 4]

custom-tracing¶

FAILED

Unable to interpret method 'aten_meth_add_', args=(x, bias), kwargs={}, dispatcher=None
--DEBUG--
[GraphBuilder-AYQ] Message starts, there are 1 initializers, 0 nodes, 1 inputs, 1 outputs.
--PARAMETERS--
dynamic_examples=
--SHAPE--
dynamic_examples=
dynamic_objects=
dynamic_objects_rev=
dynamic_dimensions_source={}
dynamic_alias={}
dynamic_shapes=None
_known_value_shape={}
_known_types={'bias': 1, 'x': 1}
_known_shapes={'bias': (1, 4), 'x': (3, 4)}
_known_constants=['bias']
_known_ranks={}
--TORCH-USERS--
add_ -> {output}
bias -> {add_}
x -> {add_}
--TORCH-SHAPES--
x: ('run_node', (('example_value', torch.float32, torch.Size([3, 4])), '')) --- 1:2:(3, 4):
bias: ('run_node', ('', '')) --- 1:2:(1, 4):
add_: ('run_node', ('', '')) --- :::
--ONNX--
-- process.graph_module --
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_
-- process.progress --
node 2/4 target=add_
--
[GraphBuilder-AYQ.make_tensor_input] x[1:3x4]
[GraphBuilder-AYQ.make_initializer] bias[torch.float32:torch.float32:[1.0, 1.0, 1.0, 1.0]] - SOURCE: DynamoInterpret.get_attr.0
[GraphBuilder-AYQ] Message completed, there are 1 initializers, 0 nodes, 1 inputs, 1 outputs.

dynamo-ir¶

opset: domain='pkg.onnxscript.torch_lib.common' version=1
opset: domain='' version=18
input: name='x' type=dtype('float32') shape=[3, 4]
init: name='bias' type=float32 shape=(1, 4) -- array([1., 1., 1., 1.], dtype=float32)
Add(x, bias) -> add
output: name='add' type=dtype('float32') shape=[3, 4]

script¶

opset: domain='' version=17
input: name='onnx::Add_0' type=dtype('float32') shape=[3, 4]
Constant(value=[[1.0, 1.0...) -> /Constant_output_0
  Add(onnx::Add_0, /Constant_output_0) -> 2
output: name='2' type=dtype('float32') shape=[3, 4]

InplaceAdd_Mul¶

forward¶

def forward(self, x):
    x.add_(self.bias)
    return x * 2

custom-fallback¶

opset: domain='' version=18
doc_string: large_model=False, inline=False, external_threshold=102...
input: name='x' type=dtype('float32') shape=[3, 4]
init: name='c_bias' type=float32 shape=(1, 4) -- array([1., 1., 1., 1.], dtype=float32)-- DynamoInterpret.placeholder.0
init: name='init1_s_' type=float32 shape=() -- array([2.], dtype=float32)-- shape_type_compute._cast_inputs.1(mul_Tensor)
init: name='init7_s1_1' type=int64 shape=(1,) -- array([1])           -- Opset.make_node.1/Shape
Add(x, c_bias) -> add_
Reshape(init1_s_, init7_s1_1) -> _onx_reshape0
  Mul(add_, _onx_reshape0) -> output_0
output: name='output_0' type=dtype('float32') shape=[3, 4]

custom-tracing¶

FAILED

Unable to interpret method 'aten_meth_add_', args=(x, bias), kwargs={}, dispatcher=None
--DEBUG--
[GraphBuilder-AGM] Message starts, there are 1 initializers, 0 nodes, 1 inputs, 1 outputs.
--PARAMETERS--
dynamic_examples=
--SHAPE--
dynamic_examples=
dynamic_objects=
dynamic_objects_rev=
dynamic_dimensions_source={}
dynamic_alias={}
dynamic_shapes=None
_known_value_shape={}
_known_types={'bias': 1, 'x': 1}
_known_shapes={'bias': (1, 4), 'x': (3, 4)}
_known_constants=['bias']
_known_ranks={}
--TORCH-USERS--
add_ -> {mul}
bias -> {add_}
x -> {add_}
--TORCH-SHAPES--
x: ('run_node', (('example_value', torch.float32, torch.Size([3, 4])), '')) --- 1:2:(3, 4):
bias: ('run_node', ('', '')) --- 1:2:(1, 4):
add_: ('run_node', ('', '')) --- :::
--ONNX--
-- process.graph_module --
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
-- process.progress --
node 2/5 target=add_
--
[GraphBuilder-AGM.make_tensor_input] x[1:3x4]
[GraphBuilder-AGM.make_initializer] bias[torch.float32:torch.float32:[1.0, 1.0, 1.0, 1.0]] - SOURCE: DynamoInterpret.get_attr.0
[GraphBuilder-AGM] Message completed, there are 1 initializers, 0 nodes, 1 inputs, 1 outputs.

dynamo-ir¶

opset: domain='pkg.onnxscript.torch_lib.common' version=1
opset: domain='' version=18
input: name='x' type=dtype('float32') shape=[3, 4]
init: name='bias' type=float32 shape=(1, 4) -- array([1., 1., 1., 1.], dtype=float32)
Add(x, bias) -> add
Constant(value=2.0) -> scalar_tensor_default
  Mul(add, scalar_tensor_default) -> mul
output: name='mul' type=dtype('float32') shape=[3, 4]

script¶

opset: domain='' version=17
input: name='onnx::Add_0' type=dtype('float32') shape=[3, 4]
Constant(value=[[1.0, 1.0...) -> /Constant_output_0
  Add(onnx::Add_0, /Constant_output_0) -> /Add_output_0
Constant(value=2.0) -> /Constant_1_output_0
  Mul(/Add_output_0, /Constant_1_output_0) -> 4
output: name='4' type=dtype('float32') shape=[3, 4]

InplaceCloneAdd ¶

forward¶

def forward(self, x):
    x = x.clone()
    x.add_(self.bias)
    return x

custom-fallback¶

opset: domain='' version=18
doc_string: large_model=False, inline=False, external_threshold=102...
input: name='x' type=dtype('float32') shape=[3, 4]
init: name='c_bias' type=float32 shape=(1, 4) -- array([1., 1., 1., 1.], dtype=float32)-- DynamoInterpret.placeholder.0
Add(x, c_bias) -> output_0
output: name='output_0' type=dtype('float32') shape=[3, 4]

custom-tracing¶

FAILED

Unable to interpret method 'aten_meth_add_', args=(clone, bias), kwargs={}, dispatcher=None
--DEBUG--
[GraphBuilder-PWA] Message starts, there are 1 initializers, 1 nodes, 1 inputs, 1 outputs.
--PARAMETERS--
dynamic_examples=
--SHAPE--
dynamic_examples=
dynamic_objects=
dynamic_objects_rev=
dynamic_dimensions_source={}
dynamic_alias={}
dynamic_shapes=None
_known_value_shape={}
_known_types={'bias': 1, 'clone': 1, 'x': 1}
_known_shapes={'bias': (1, 4), 'clone': (3, 4), 'x': (3, 4)}
_known_constants=['bias']
_known_ranks={}
--TORCH-USERS--
add_ -> {output}
bias -> {add_}
clone -> {add_}
x -> {clone}
--TORCH-SHAPES--
x: ('run_node', (('example_value', torch.float32, torch.Size([3, 4])), '')) --- 1:2:(3, 4):
clone: ('run_node', ('', '')) --- 1:2:(3, 4):
bias: ('run_node', ('', '')) --- 1:2:(1, 4):
add_: ('run_node', ('', '')) --- :::
--ONNX--
-- process.graph_module --
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_
-- process.progress --
node 3/5 target=add_
--
[GraphBuilder-PWA.make_tensor_input] x[1:3x4]
[GraphBuilder-PWA.make_initializer] bias[torch.float32:torch.float32:[1.0, 1.0, 1.0, 1.0]] - SOURCE: DynamoInterpret.get_attr.0
[GraphBuilder-PWA.make_node] .clone          [#:#   ] Identity:['x']->['clone']
[GraphBuilder-PWA] Message completed, there are 1 initializers, 1 nodes, 1 inputs, 1 outputs.

dynamo-ir¶

opset: domain='pkg.onnxscript.torch_lib.common' version=1
opset: domain='' version=18
input: name='x' type=dtype('float32') shape=[3, 4]
init: name='bias' type=float32 shape=(1, 4) -- array([1., 1., 1., 1.], dtype=float32)
Add(x, bias) -> add
output: name='add' type=dtype('float32') shape=[3, 4]

script¶

opset: domain='' version=17
input: name='onnx::Add_0' type=dtype('float32') shape=[3, 4]
Constant(value=[[1.0, 1.0...) -> /Constant_output_0
  Add(onnx::Add_0, /Constant_output_0) -> 2
output: name='2' type=dtype('float32') shape=[3, 4]

InplaceSetItemEllipsis_1¶

forward¶

def forward(self, index, update):
    copy = self.params.clone()
    copy[..., index] = update
    return copy

custom-fallback¶

opset: domain='' version=18
doc_string: large_model=False, inline=False, external_threshold=102...
input: name='index' type=dtype('int64') shape=[4]
input: name='update' type=dtype('float32') shape=[8192, 4]
init: name='c_params' type=float32 shape=(1, 8192, 4)                 -- DynamoInterpret.placeholder.0
init: name='init7_s1_0' type=int64 shape=(1,) -- array([0])           -- Opset.make_node.1/Shape
init: name='init7_s3_-1_1_1' type=int64 shape=(3,) -- array([-1,  1,  1])-- Opset.make_node.1/Shape
init: name='init7_s8192_' type=int64 shape=(8192,)                    -- Opset.make_node.0
init: name='init7_s3_1_-1_1' type=int64 shape=(3,) -- array([ 1, -1,  1])-- Opset.make_node.1/Shape
init: name='init7_s3_1_1_-1' type=int64 shape=(3,) -- array([ 1,  1, -1])-- Opset.make_node.1/Shape
init: name='init7_s3_1_8192_4' type=int64 shape=(3,) -- array([   1, 8192,    4])-- Opset.make_node.1/Shape##Opset.make_node.1/Shape
init: name='init7_s1_32768' type=int64 shape=(1,) -- array([32768])   -- Opset.make_node.1/Shape
init: name='init7_s1_4' type=int64 shape=(1,) -- array([4])           -- Opset.make_node.1/Shape
init: name='init7_s1_-1' type=int64 shape=(1,) -- array([-1])         -- Opset.make_node.1/Shape##Opset.make_node.1/Shape##Opset.make_node.1/Shape
init: name='init7_s32768_' type=int64 shape=(32768,)                  -- Opset.make_node.0
Expand(update, init7_s3_1_8192_4) -> _onx_expand0
  Reshape(_onx_expand0, init7_s1_-1) -> _onx_reshape05
Reshape(init7_s1_0, init7_s3_-1_1_1) -> _onx_reshape0
  Mul(_onx_reshape0, init7_s1_32768) -> _onx_mul0
Reshape(init7_s8192_, init7_s3_1_-1_1) -> _onx_reshape02
  Mul(_onx_reshape02, init7_s1_4) -> _onx_mul02
Reshape(index, init7_s3_1_1_-1) -> _onx_reshape03
  Add(_onx_mul0, _onx_reshape03) -> add-_onx_mul0
    Add(add-_onx_mul0, _onx_mul02) -> _onx_add02
      Reshape(_onx_add02, init7_s1_-1) -> _onx_reshape04
        GatherElements(init7_s32768_, _onx_reshape04) -> _onx_gatherelements0
Reshape(c_params, init7_s1_-1) -> _onx_reshape06
  ScatterElements(_onx_reshape06, _onx_gatherelements0, _onx_reshape05) -> _onx_scatterelements0
    Reshape(_onx_scatterelements0, init7_s3_1_8192_4) -> output_0
output: name='output_0' type=dtype('float32') shape=[1, 8192, 4]

custom-tracing¶

FAILED

setitem not implemented for indices=(Ellipsis, 'index')
--DEBUG--
[GraphBuilder-ZBK] Message starts, there are 1 initializers, 0 nodes, 2 inputs, 2 outputs.
--PARAMETERS--
dynamic_examples=
--SHAPE--
dynamic_examples=
dynamic_objects=
dynamic_objects_rev=
dynamic_dimensions_source={}
dynamic_alias={}
dynamic_shapes=None
_known_value_shape={}
_known_types={'_tensor_constant0': 1, 'index': 7, 'update': 1}
_known_shapes={'_tensor_constant0': (1, 8192, 4), 'index': (4,), 'update': (8192, 4)}
_known_constants=['_tensor_constant0']
_known_ranks={}
--TORCH-USERS--
_tensor_constant0 -> {setitem}
index -> {setitem}
setitem -> {output}
update -> {setitem}
--TORCH-SHAPES--
index: ('run_node', (('example_value', torch.int64, torch.Size([4])), '')) --- 7:1:(4,):
update: ('run_node', (('example_value', torch.float32, torch.Size([8192, 4])), '')) --- 1:2:(8192, 4):
_tensor_constant0: ('run_node', ('', '')) --- 1:3:(1, 8192, 4):
setitem: ('run_node', ('', '')) --- :::
--ONNX--
-- process.graph_module --
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
-- process.progress --
node 3/5 target=<built-in function setitem>
--
[GraphBuilder-ZBK.make_tensor_input] index[7:4]
[GraphBuilder-ZBK.make_tensor_input] update[1:8192x4]
[GraphBuilder-ZBK.make_initializer] _tensor_constant0[torch.float32:torch.float32] - SOURCE: DynamoInterpret.get_attr.0
[GraphBuilder-ZBK] Message completed, there are 1 initializers, 0 nodes, 2 inputs, 2 outputs.

dynamo-ir¶

FAILED

Failed to decompose the FX graph for ONNX compatibility. [96mThis is step 2/3[0m of exporting the model to ONNX. Next steps:
- Create an issue in the PyTorch GitHub repository against the [96m*torch.export*[0m component and attach the full error stack as well as reproduction scripts.
- Create an error report with `torch.onnx.export(..., report=True)`, and save the ExportedProgram as a pt2 file. Create an issue in the PyTorch GitHub repository against the [96m*onnx*[0m component. Attach the error report and the pt2 model.

## Exception summary

<class 'RuntimeError'>: false INTERNAL ASSERT FAILED at "/pytorch/build/aten/src/ATen/RegisterFunctionalization_1.cpp":5941, please report a bug to PyTorch. mutating a non-functional tensor with a functional tensor is not allowed. Please ensure that all of your inputs are wrapped inside of a functionalize() call.

While executing %index_put_ : [num_users=1] = call_function[target=torch.ops.aten.index_put_.default](args = (%clone, [None, None, %index], %update), kwargs = {})
Original traceback:
File "/home/xadupre/github/experimental-experiment/experimental_experiment/torch_interpreter/eval/model_cases.py", line 116, in forward
    copy[..., index] = update

(Refer to the full stack trace above for more information.)

script¶

FAILED

[ONNXRuntimeError] : 10 : INVALID_GRAPH : This is an invalid model. In Node, ("/Add", Add, "", -1) : ("","",) -> ("/Add_output_0",) , Error Node (/Add)'s input 0 is marked single but has an empty string in the graph

InplaceSetItemEllipsis_2¶

forward¶

def forward(self, index, update):
    copy = self.params.clone()
    copy[..., index] = update
    return copy

custom-fallback¶

opset: domain='' version=18
doc_string: large_model=False, inline=False, external_threshold=102...
input: name='index' type=dtype('int64') shape=[4]
input: name='update' type=dtype('float32') shape=[8192, 4]
init: name='c_params' type=float32 shape=(1, 8192, 6)                 -- DynamoInterpret.placeholder.0
init: name='init7_s1_0' type=int64 shape=(1,) -- array([0])           -- Opset.make_node.1/Shape
init: name='init7_s3_-1_1_1' type=int64 shape=(3,) -- array([-1,  1,  1])-- Opset.make_node.1/Shape
init: name='init7_s8192_' type=int64 shape=(8192,)                    -- Opset.make_node.0
init: name='init7_s3_1_-1_1' type=int64 shape=(3,) -- array([ 1, -1,  1])-- Opset.make_node.1/Shape
init: name='init7_s3_1_1_-1' type=int64 shape=(3,) -- array([ 1,  1, -1])-- Opset.make_node.1/Shape
init: name='init7_s3_1_8192_4' type=int64 shape=(3,) -- array([   1, 8192,    4])-- Opset.make_node.1/Shape
init: name='init7_s1_49152' type=int64 shape=(1,) -- array([49152])   -- Opset.make_node.1/Shape
init: name='init7_s1_6' type=int64 shape=(1,) -- array([6])           -- Opset.make_node.1/Shape
init: name='init7_s1_-1' type=int64 shape=(1,) -- array([-1])         -- Opset.make_node.1/Shape##Opset.make_node.1/Shape##Opset.make_node.1/Shape
init: name='init7_s49152_' type=int64 shape=(49152,)                  -- Opset.make_node.0
init: name='init7_s3_1_8192_6' type=int64 shape=(3,) -- array([   1, 8192,    6])-- Opset.make_node.1/Shape
Expand(update, init7_s3_1_8192_4) -> _onx_expand0
  Reshape(_onx_expand0, init7_s1_-1) -> _onx_reshape05
Reshape(init7_s1_0, init7_s3_-1_1_1) -> _onx_reshape0
  Mul(_onx_reshape0, init7_s1_49152) -> _onx_mul0
Reshape(init7_s8192_, init7_s3_1_-1_1) -> _onx_reshape02
  Mul(_onx_reshape02, init7_s1_6) -> _onx_mul02
Reshape(index, init7_s3_1_1_-1) -> _onx_reshape03
  Add(_onx_mul0, _onx_reshape03) -> add-_onx_mul0
    Add(add-_onx_mul0, _onx_mul02) -> _onx_add02
      Reshape(_onx_add02, init7_s1_-1) -> _onx_reshape04
        GatherElements(init7_s49152_, _onx_reshape04) -> _onx_gatherelements0
Reshape(c_params, init7_s1_-1) -> _onx_reshape06
  ScatterElements(_onx_reshape06, _onx_gatherelements0, _onx_reshape05) -> _onx_scatterelements0
    Reshape(_onx_scatterelements0, init7_s3_1_8192_6) -> output_0
output: name='output_0' type=dtype('float32') shape=[1, 8192, 6]

custom-tracing¶

FAILED

setitem not implemented for indices=(Ellipsis, 'index')
--DEBUG--
[GraphBuilder-EWU] Message starts, there are 1 initializers, 0 nodes, 2 inputs, 2 outputs.
--PARAMETERS--
dynamic_examples=
--SHAPE--
dynamic_examples=
dynamic_objects=
dynamic_objects_rev=
dynamic_dimensions_source={}
dynamic_alias={}
dynamic_shapes=None
_known_value_shape={}
_known_types={'_tensor_constant0': 1, 'index': 7, 'update': 1}
_known_shapes={'_tensor_constant0': (1, 8192, 6), 'index': (4,), 'update': (8192, 4)}
_known_constants=['_tensor_constant0']
_known_ranks={}
--TORCH-USERS--
_tensor_constant0 -> {setitem}
index -> {setitem}
setitem -> {output}
update -> {setitem}
--TORCH-SHAPES--
index: ('run_node', (('example_value', torch.int64, torch.Size([4])), '')) --- 7:1:(4,):
update: ('run_node', (('example_value', torch.float32, torch.Size([8192, 4])), '')) --- 1:2:(8192, 4):
_tensor_constant0: ('run_node', ('', '')) --- 1:3:(1, 8192, 6):
setitem: ('run_node', ('', '')) --- :::
--ONNX--
-- process.graph_module --
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
-- process.progress --
node 3/5 target=<built-in function setitem>
--
[GraphBuilder-EWU.make_tensor_input] index[7:4]
[GraphBuilder-EWU.make_tensor_input] update[1:8192x4]
[GraphBuilder-EWU.make_initializer] _tensor_constant0[torch.float32:torch.float32] - SOURCE: DynamoInterpret.get_attr.0
[GraphBuilder-EWU] Message completed, there are 1 initializers, 0 nodes, 2 inputs, 2 outputs.

dynamo-ir¶

FAILED

Failed to decompose the FX graph for ONNX compatibility. [96mThis is step 2/3[0m of exporting the model to ONNX. Next steps:
- Create an issue in the PyTorch GitHub repository against the [96m*torch.export*[0m component and attach the full error stack as well as reproduction scripts.
- Create an error report with `torch.onnx.export(..., report=True)`, and save the ExportedProgram as a pt2 file. Create an issue in the PyTorch GitHub repository against the [96m*onnx*[0m component. Attach the error report and the pt2 model.

## Exception summary

<class 'RuntimeError'>: false INTERNAL ASSERT FAILED at "/pytorch/build/aten/src/ATen/RegisterFunctionalization_1.cpp":5941, please report a bug to PyTorch. mutating a non-functional tensor with a functional tensor is not allowed. Please ensure that all of your inputs are wrapped inside of a functionalize() call.

While executing %index_put_ : [num_users=1] = call_function[target=torch.ops.aten.index_put_.default](args = (%clone, [None, None, %index], %update), kwargs = {})
Original traceback:
File "/home/xadupre/github/experimental-experiment/experimental_experiment/torch_interpreter/eval/model_cases.py", line 134, in forward
    copy[..., index] = update

(Refer to the full stack trace above for more information.)

script¶

FAILED

[ONNXRuntimeError] : 10 : INVALID_GRAPH : This is an invalid model. In Node, ("/Add", Add, "", -1) : ("","",) -> ("/Add_output_0",) , Error Node (/Add)'s input 0 is marked single but has an empty string in the graph

InplaceSetItemMask¶

forward¶

def forward(self, x):
    mask = x.to(bool)
    x[mask] = 2
    return x

custom-fallback¶

opset: domain='' version=18
doc_string: large_model=False, inline=False, external_threshold=102...
input: name='x' type=dtype('float32') shape=[2, 3, 3]
init: name='c_lifted_tensor_0' type=float32 shape=() -- array([2.], dtype=float32)-- DynamoInterpret.placeholder.0
Cast(x, to=9) -> to
  Where(to, c_lifted_tensor_0, x) -> output_0
output: name='output_0' type=dtype('float32') shape=[2, 3, 3]

custom-tracing¶

FAILED

setitem not implemented for indices=to
--DEBUG--
[GraphBuilder-ROQ] Message starts, there are 0 initializers, 1 nodes, 1 inputs, 1 outputs.
--PARAMETERS--
dynamic_examples=
--SHAPE--
dynamic_examples=
dynamic_objects=
dynamic_objects_rev=
dynamic_dimensions_source={}
dynamic_alias={}
dynamic_shapes=None
_known_value_shape={}
_known_types={'to': 9, 'x': 1}
_known_shapes={'to': (2, 3, 3), 'x': (2, 3, 3)}
_known_constants=[]
_known_ranks={}
--TORCH-USERS--
setitem -> {output}
to -> {setitem}
x -> {setitem, to}
--TORCH-SHAPES--
x: ('run_node', (('example_value', torch.float32, torch.Size([2, 3, 3])), '')) --- 1:3:(2, 3, 3):
to: ('run_node', ('', '')) --- 9:3:(2, 3, 3):
setitem: ('run_node', ('', '')) --- :::
--ONNX--
-- process.graph_module --
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
-- process.progress --
node 2/4 target=<built-in function setitem>
--
[GraphBuilder-ROQ.make_tensor_input] x[1:2x3x3]
[GraphBuilder-ROQ.make_node] .to             [#:#   ] Cast>9:['x']->['to']
[GraphBuilder-ROQ] Message completed, there are 0 initializers, 1 nodes, 1 inputs, 1 outputs.

dynamo-ir¶

FAILED

[ONNXRuntimeError] : 2 : INVALID_ARGUMENT : Non-zero status code returned while running If node. Name:'n7_2' Status Message: Non-zero status code returned while running ScatterElements node. Name:'n15_n0_7' Status Message: Indices and updates must have the same rank

script¶

opset: domain='' version=17
input: name='onnx::Cast_0' type=dtype('float32') shape=[2, 3, 3]
Cast(onnx::Cast_0, to=9) -> /Cast_output_0
  Cast(/Cast_output_0, to=9) -> /Cast_1_output_0
Constant(value=2.0) -> /Constant_output_0
  Where(/Cast_1_output_0, /Constant_output_0, onnx::Cast_0) -> 4
output: name='4' type=dtype('float32') shape=[2, 3, 3]

InplaceSetItemSquare¶

forward¶

def forward(self, x):
    x[:2, :3] = 1
    return x

custom-fallback¶

/home/xadupre/vv/this312/lib/python3.12/site-packages/torch/export/_unlift.py:75: UserWarning: Attempted to insert a get_attr Node with no underlying reference in the owning GraphModule! Call GraphModule.add_submodule to add the necessary submodule, GraphModule.add_parameter to add the necessary Parameter, or nn.Module.register_buffer to add the necessary buffer: getattr_node = gm.graph.get_attr(lifted_node)
/home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/graph.py:1801: UserWarning: Node lifted_tensor_0 target lifted_tensor_0 lifted_tensor_0 of does not reference an nn.Module, nn.Parameter, or buffer, which is what ‘get_attr’ Nodes typically target: warnings.warn(

opset: domain='' version=18
doc_string: large_model=False, inline=False, external_threshold=102...
input: name='x' type=dtype('float32') shape=[5, 5]
init: name='c_lifted_tensor_0' type=float32 shape=() -- array([1.], dtype=float32)-- DynamoInterpret.placeholder.0
init: name='init7_s1_0' type=int64 shape=(1,) -- array([0])           -- Opset.make_node.1/Shape##Opset.make_node.1/Shape##Opset.make_node.1/Shape##Opset.make_node.1/Shape##Opset.make_node.1/Shape
init: name='init7_s1_2' type=int64 shape=(1,) -- array([2])           -- Opset.make_node.1/Shape##Opset.make_node.1/Shape
init: name='init7_s1_3' type=int64 shape=(1,) -- array([3])           -- Opset.make_node.1/Shape
init: name='init7_s1_1' type=int64 shape=(1,) -- array([1])           -- Opset.make_node.1/Shape
init: name='init7_s3_0_1_2' type=int64 shape=(3,) -- array([0, 1, 2]) -- Opset.make_node.1/Shape
init: name='init7_s2_-1_1' type=int64 shape=(2,) -- array([-1,  1])   -- Opset.make_node.1/Shape##Opset.make_node.1/Shape
init: name='init7_s2_0_1' type=int64 shape=(2,) -- array([0, 1])      -- Opset.make_node.1/Shape
Concat(init7_s1_0, init7_s1_1, axis=0) -> SliceSlicePattern_init7_s1_1_axis
Concat(init7_s1_0, init7_s1_0, axis=0) -> SliceSlicePattern_init7_s1_0_start
Concat(init7_s1_2, init7_s1_3, axis=0) -> SliceSlicePattern_init7_s1_3_end
  Slice(x, SliceSlicePattern_init7_s1_0_start, SliceSlicePattern_init7_s1_3_end, SliceSlicePattern_init7_s1_1_axis) -> slice_2
    Shape(slice_2) -> _onx_shape0
      Expand(c_lifted_tensor_0, _onx_shape0) -> fill
        Transpose(fill, perm=[1,0]) -> _onx_transpose02
Slice(x, init7_s1_0, init7_s1_2, init7_s1_0) -> slice_3
  Transpose(slice_3, perm=[1,0]) -> _onx_transpose0
Reshape(init7_s3_0_1_2, init7_s2_-1_1) -> _onx_reshape0
  ScatterND(_onx_transpose0, _onx_reshape0, _onx_transpose02) -> _onx_scatternd0
    Transpose(_onx_scatternd0, perm=[1,0]) -> slice_scatter
Reshape(init7_s2_0_1, init7_s2_-1_1) -> _onx_reshape02
  ScatterND(x, _onx_reshape02, slice_scatter) -> output_0
    Identity(output_0) -> output_1
output: name='output_1' type=dtype('float32') shape=[5, 5]

custom-tracing¶

FAILED

setitem not implemented for indices=(slice(None, 2, None), slice(None, 3, None))
--DEBUG--
[GraphBuilder-GUA] Message starts, there are 0 initializers, 0 nodes, 1 inputs, 1 outputs.
--PARAMETERS--
dynamic_examples=
--SHAPE--
dynamic_examples=
dynamic_objects=
dynamic_objects_rev=
dynamic_dimensions_source={}
dynamic_alias={}
dynamic_shapes=None
_known_value_shape={}
_known_types={'x': 1}
_known_shapes={'x': (5, 5)}
_known_constants=[]
_known_ranks={}
--TORCH-USERS--
setitem -> {output}
x -> {setitem}
--TORCH-SHAPES--
x: ('run_node', (('example_value', torch.float32, torch.Size([5, 5])), '')) --- 1:2:(5, 5):
setitem: ('run_node', ('', '')) --- :::
--ONNX--
-- process.graph_module --
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
-- process.progress --
node 1/3 target=<built-in function setitem>
--
[GraphBuilder-GUA.make_tensor_input] x[1:5x5]
[GraphBuilder-GUA] Message completed, there are 0 initializers, 0 nodes, 1 inputs, 1 outputs.

dynamo-ir¶

opset: domain='pkg.onnxscript.torch_lib.common' version=1
opset: domain='' version=18
opset: domain='pkg.onnxscript.torch_lib' version=1
input: name='x' type=dtype('float32') shape=[5, 5]
Constant(value=[[0], [1],...) -> val_43
Constant(value=[0]) -> val_26
Constant(value=[2]) -> val_29
Constant(value=[0]) -> val_32
Constant(value_ints=[1]) -> val_33
  Slice(x, val_26, val_29, val_32, val_33) -> slice_3
    Transpose(slice_3, perm=[1,0]) -> val_45
Constant(value=[[1.0, 1.0...) -> val_44
  ScatterND(val_45, val_43, val_44, reduction=b'none') -> val_46
    Transpose(val_46, perm=[1,0]) -> slice_scatter
Constant(value=[[0], [1]]) -> val_55
  ScatterND(x, val_55, slice_scatter, reduction=b'none') -> slice_scatter_1
output: name='slice_scatter_1' type=dtype('float32') shape=[5, 5]

script¶

opset: domain='' version=17
input: name='onnx::Slice_0' type=dtype('float32') shape=[5, 5]
Constant(value=[0]) -> /Constant_output_0
Constant(value=[0]) -> /Constant_1_output_0
Constant(value=[2]) -> /Constant_2_output_0
Constant(value=[1]) -> /Constant_3_output_0
  Slice(onnx::Slice_0, /Constant_1_output_0, /Constant_2_output_0, /Constant_output_0, /Constant_3_output_0) -> /Slice_output_0
Constant(value=[1]) -> /Constant_4_output_0
Constant(value=[0]) -> /Constant_5_output_0
Constant(value=[3]) -> /Constant_6_output_0
Constant(value=[1]) -> /Constant_7_output_0
  Slice(/Slice_output_0, /Constant_5_output_0, /Constant_6_output_0, /Constant_4_output_0, /Constant_7_output_0) -> /Slice_1_output_0
    Shape(/Slice_1_output_0) -> /Shape_output_0
Constant(value=[[1.0]]) -> /Constant_8_output_0
  Expand(/Constant_8_output_0, /Shape_output_0) -> /Expand_output_0
Constant(value=[0, 1, 2]) -> /Constant_9_output_0
Constant(value=[[0], [1]]) -> onnx::Expand_27
Constant(value=[2, 3]) -> /Constant_10_output_0
Constant(value=[2]) -> /Constant_11_output_0
  ConstantOfShape(/Constant_11_output_0, value=[1]) -> /ConstantOfShape_output_0
Constant(value=-1) -> /Constant_12_output_0
  Mul(/ConstantOfShape_output_0, /Constant_12_output_0) -> /Mul_output_0
  Equal(/Constant_10_output_0, /Mul_output_0) -> /Equal_output_0
  Where(/Equal_output_0, /ConstantOfShape_output_0, /Constant_10_output_0) -> /Where_output_0
  Expand(onnx::Expand_27, /Where_output_0) -> /Expand_1_output_0
Constant(value=[-1]) -> /Constant_13_output_0
  Unsqueeze(/Expand_1_output_0, /Constant_13_output_0) -> /Unsqueeze_output_0
Constant(value=[2]) -> /Constant_14_output_0
  ConstantOfShape(/Constant_14_output_0, value=[1]) -> /ConstantOfShape_1_output_0
Constant(value=-1) -> /Constant_15_output_0
  Mul(/ConstantOfShape_1_output_0, /Constant_15_output_0) -> /Mul_1_output_0
  Equal(/Constant_10_output_0, /Mul_1_output_0) -> /Equal_1_output_0
  Where(/Equal_1_output_0, /ConstantOfShape_1_output_0, /Constant_10_output_0) -> /Where_1_output_0
  Expand(/Constant_9_output_0, /Where_1_output_0) -> /Expand_2_output_0
Constant(value=[-1]) -> /Constant_16_output_0
  Unsqueeze(/Expand_2_output_0, /Constant_16_output_0) -> /Unsqueeze_1_output_0
    Concat(/Unsqueeze_output_0, /Unsqueeze_1_output_0, axis=-1) -> /Concat_output_0
Shape(onnx::Slice_0) -> /Shape_1_output_0
Constant(value=[0]) -> /Constant_17_output_0
Constant(value=[2]) -> /Constant_18_output_0
Constant(value=[922337203...) -> /Constant_19_output_0
  Slice(/Shape_1_output_0, /Constant_18_output_0, /Constant_19_output_0, /Constant_17_output_0) -> /Slice_2_output_0
  Concat(/Constant_10_output_0, /Slice_2_output_0, axis=0) -> /Concat_1_output_0
    Reshape(/Expand_output_0, /Concat_1_output_0, allowzero=0) -> /Reshape_output_0
      ScatterND(onnx::Slice_0, /Concat_output_0, /Reshape_output_0) -> 58
output: name='58' type=dtype('float32') shape=[5, 5]

InplaceSetItemSquareAdd¶

forward¶

def forward(self, x):
    x[:2, :3] = 1
    return x + 2

custom-fallback¶

/home/xadupre/vv/this312/lib/python3.12/site-packages/torch/export/_unlift.py:75: UserWarning: Attempted to insert a get_attr Node with no underlying reference in the owning GraphModule! Call GraphModule.add_submodule to add the necessary submodule, GraphModule.add_parameter to add the necessary Parameter, or nn.Module.register_buffer to add the necessary buffer: getattr_node = gm.graph.get_attr(lifted_node)
/home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/graph.py:1801: UserWarning: Node lifted_tensor_0 target lifted_tensor_0 lifted_tensor_0 of does not reference an nn.Module, nn.Parameter, or buffer, which is what ‘get_attr’ Nodes typically target: warnings.warn(

opset: domain='' version=18
doc_string: large_model=False, inline=False, external_threshold=102...
input: name='x' type=dtype('float32') shape=[5, 5]
init: name='c_lifted_tensor_0' type=float32 shape=() -- array([1.], dtype=float32)-- DynamoInterpret.placeholder.0
init: name='init7_s1_0' type=int64 shape=(1,) -- array([0])           -- Opset.make_node.1/Shape##Opset.make_node.1/Shape##Opset.make_node.1/Shape##Opset.make_node.1/Shape##Opset.make_node.1/Shape
init: name='init7_s1_2' type=int64 shape=(1,) -- array([2])           -- Opset.make_node.1/Shape##Opset.make_node.1/Shape
init: name='init7_s1_3' type=int64 shape=(1,) -- array([3])           -- Opset.make_node.1/Shape
init: name='init7_s1_1' type=int64 shape=(1,) -- array([1])           -- Opset.make_node.1/Shape##Opset.make_node.1/Shape
init: name='init7_s3_0_1_2' type=int64 shape=(3,) -- array([0, 1, 2]) -- Opset.make_node.1/Shape
init: name='init7_s2_-1_1' type=int64 shape=(2,) -- array([-1,  1])   -- Opset.make_node.1/Shape##Opset.make_node.1/Shape
init: name='init7_s2_0_1' type=int64 shape=(2,) -- array([0, 1])      -- Opset.make_node.1/Shape
init: name='init1_s_' type=float32 shape=() -- array([2.], dtype=float32)-- shape_type_compute._cast_inputs.0
Concat(init7_s1_0, init7_s1_1, axis=0) -> SliceSlicePattern_init7_s1_1_axis
Concat(init7_s1_0, init7_s1_0, axis=0) -> SliceSlicePattern_init7_s1_0_start
Concat(init7_s1_2, init7_s1_3, axis=0) -> SliceSlicePattern_init7_s1_3_end
  Slice(x, SliceSlicePattern_init7_s1_0_start, SliceSlicePattern_init7_s1_3_end, SliceSlicePattern_init7_s1_1_axis) -> slice_2
    Shape(slice_2) -> _onx_shape0
      Expand(c_lifted_tensor_0, _onx_shape0) -> fill
        Transpose(fill, perm=[1,0]) -> _onx_transpose02
Slice(x, init7_s1_0, init7_s1_2, init7_s1_0) -> slice_3
  Transpose(slice_3, perm=[1,0]) -> _onx_transpose0
Reshape(init7_s3_0_1_2, init7_s2_-1_1) -> _onx_reshape0
  ScatterND(_onx_transpose0, _onx_reshape0, _onx_transpose02) -> _onx_scatternd0
    Transpose(_onx_scatternd0, perm=[1,0]) -> slice_scatter
Reshape(init7_s2_0_1, init7_s2_-1_1) -> _onx_reshape02
  ScatterND(x, _onx_reshape02, slice_scatter) -> output_0
Reshape(init1_s_, init7_s1_1) -> _onx_reshape03
  Add(output_0, _onx_reshape03) -> output_1
output: name='output_1' type=dtype('float32') shape=[5, 5]

custom-tracing¶

FAILED

setitem not implemented for indices=(slice(None, 2, None), slice(None, 3, None))
--DEBUG--
[GraphBuilder-FGQ] Message starts, there are 0 initializers, 0 nodes, 1 inputs, 1 outputs.
--PARAMETERS--
dynamic_examples=
--SHAPE--
dynamic_examples=
dynamic_objects=
dynamic_objects_rev=
dynamic_dimensions_source={}
dynamic_alias={}
dynamic_shapes=None
_known_value_shape={}
_known_types={'x': 1}
_known_shapes={'x': (5, 5)}
_known_constants=[]
_known_ranks={}
--TORCH-USERS--
setitem -> {add}
x -> {setitem}
--TORCH-SHAPES--
x: ('run_node', (('example_value', torch.float32, torch.Size([5, 5])), '')) --- 1:2:(5, 5):
setitem: ('run_node', ('', '')) --- :::
--ONNX--
-- process.graph_module --
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
-- process.progress --
node 1/4 target=<built-in function setitem>
--
[GraphBuilder-FGQ.make_tensor_input] x[1:5x5]
[GraphBuilder-FGQ] Message completed, there are 0 initializers, 0 nodes, 1 inputs, 1 outputs.

dynamo-ir¶

opset: domain='pkg.onnxscript.torch_lib.common' version=1
opset: domain='' version=18
opset: domain='pkg.onnxscript.torch_lib' version=1
input: name='x' type=dtype('float32') shape=[5, 5]
Constant(value=[[0], [1],...) -> val_43
Constant(value=[0]) -> val_26
Constant(value=[2]) -> val_29
Constant(value=[0]) -> val_32
Constant(value_ints=[1]) -> val_33
  Slice(x, val_26, val_29, val_32, val_33) -> slice_3
    Transpose(slice_3, perm=[1,0]) -> val_45
Constant(value=[[1.0, 1.0...) -> val_44
  ScatterND(val_45, val_43, val_44, reduction=b'none') -> val_46
    Transpose(val_46, perm=[1,0]) -> slice_scatter
Constant(value=[[0], [1]]) -> val_55
  ScatterND(x, val_55, slice_scatter, reduction=b'none') -> slice_scatter_1
Constant(value=2.0) -> scalar_tensor_default
  Add(slice_scatter_1, scalar_tensor_default) -> add
output: name='add' type=dtype('float32') shape=[5, 5]

script¶

opset: domain='' version=17
input: name='onnx::Slice_0' type=dtype('float32') shape=[5, 5]
Constant(value=[0]) -> /Constant_output_0
Constant(value=[0]) -> /Constant_1_output_0
Constant(value=[2]) -> /Constant_2_output_0
Constant(value=[1]) -> /Constant_3_output_0
  Slice(onnx::Slice_0, /Constant_1_output_0, /Constant_2_output_0, /Constant_output_0, /Constant_3_output_0) -> /Slice_output_0
Constant(value=[1]) -> /Constant_4_output_0
Constant(value=[0]) -> /Constant_5_output_0
Constant(value=[3]) -> /Constant_6_output_0
Constant(value=[1]) -> /Constant_7_output_0
  Slice(/Slice_output_0, /Constant_5_output_0, /Constant_6_output_0, /Constant_4_output_0, /Constant_7_output_0) -> /Slice_1_output_0
    Shape(/Slice_1_output_0) -> /Shape_output_0
Constant(value=[[1.0]]) -> /Constant_8_output_0
  Expand(/Constant_8_output_0, /Shape_output_0) -> /Expand_output_0
Constant(value=[0, 1, 2]) -> /Constant_9_output_0
Constant(value=[[0], [1]]) -> onnx::Expand_27
Constant(value=[2, 3]) -> /Constant_10_output_0
Constant(value=[2]) -> /Constant_11_output_0
  ConstantOfShape(/Constant_11_output_0, value=[1]) -> /ConstantOfShape_output_0
Constant(value=-1) -> /Constant_12_output_0
  Mul(/ConstantOfShape_output_0, /Constant_12_output_0) -> /Mul_output_0
  Equal(/Constant_10_output_0, /Mul_output_0) -> /Equal_output_0
  Where(/Equal_output_0, /ConstantOfShape_output_0, /Constant_10_output_0) -> /Where_output_0
  Expand(onnx::Expand_27, /Where_output_0) -> /Expand_1_output_0
Constant(value=[-1]) -> /Constant_13_output_0
  Unsqueeze(/Expand_1_output_0, /Constant_13_output_0) -> /Unsqueeze_output_0
Constant(value=[2]) -> /Constant_14_output_0
  ConstantOfShape(/Constant_14_output_0, value=[1]) -> /ConstantOfShape_1_output_0
Constant(value=-1) -> /Constant_15_output_0
  Mul(/ConstantOfShape_1_output_0, /Constant_15_output_0) -> /Mul_1_output_0
  Equal(/Constant_10_output_0, /Mul_1_output_0) -> /Equal_1_output_0
  Where(/Equal_1_output_0, /ConstantOfShape_1_output_0, /Constant_10_output_0) -> /Where_1_output_0
  Expand(/Constant_9_output_0, /Where_1_output_0) -> /Expand_2_output_0
Constant(value=[-1]) -> /Constant_16_output_0
  Unsqueeze(/Expand_2_output_0, /Constant_16_output_0) -> /Unsqueeze_1_output_0
    Concat(/Unsqueeze_output_0, /Unsqueeze_1_output_0, axis=-1) -> /Concat_output_0
Shape(onnx::Slice_0) -> /Shape_1_output_0
Constant(value=[0]) -> /Constant_17_output_0
Constant(value=[2]) -> /Constant_18_output_0
Constant(value=[922337203...) -> /Constant_19_output_0
  Slice(/Shape_1_output_0, /Constant_18_output_0, /Constant_19_output_0, /Constant_17_output_0) -> /Slice_2_output_0
  Concat(/Constant_10_output_0, /Slice_2_output_0, axis=0) -> /Concat_1_output_0
    Reshape(/Expand_output_0, /Concat_1_output_0, allowzero=0) -> /Reshape_output_0
      ScatterND(onnx::Slice_0, /Concat_output_0, /Reshape_output_0) -> /ScatterND_output_0
Constant(value=2.0) -> /Constant_20_output_0
  Add(/ScatterND_output_0, /Constant_20_output_0) -> 60
output: name='60' type=dtype('float32') shape=[5, 5]

InplaceSetItemSquareAdd2¶

forward¶

def forward(self, x):
    x[:2, :3] = 1
    return x + 2, x + 3

custom-fallback¶

/home/xadupre/vv/this312/lib/python3.12/site-packages/torch/export/_unlift.py:75: UserWarning: Attempted to insert a get_attr Node with no underlying reference in the owning GraphModule! Call GraphModule.add_submodule to add the necessary submodule, GraphModule.add_parameter to add the necessary Parameter, or nn.Module.register_buffer to add the necessary buffer: getattr_node = gm.graph.get_attr(lifted_node)
/home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/graph.py:1801: UserWarning: Node lifted_tensor_0 target lifted_tensor_0 lifted_tensor_0 of does not reference an nn.Module, nn.Parameter, or buffer, which is what ‘get_attr’ Nodes typically target: warnings.warn(

opset: domain='' version=18
doc_string: large_model=False, inline=False, external_threshold=102...
input: name='x' type=dtype('float32') shape=[5, 5]
init: name='c_lifted_tensor_0' type=float32 shape=() -- array([1.], dtype=float32)-- DynamoInterpret.placeholder.0
init: name='init7_s1_0' type=int64 shape=(1,) -- array([0])           -- Opset.make_node.1/Shape##Opset.make_node.1/Shape##Opset.make_node.1/Shape##Opset.make_node.1/Shape##Opset.make_node.1/Shape
init: name='init7_s1_2' type=int64 shape=(1,) -- array([2])           -- Opset.make_node.1/Shape##Opset.make_node.1/Shape
init: name='init7_s1_3' type=int64 shape=(1,) -- array([3])           -- Opset.make_node.1/Shape
init: name='init7_s1_1' type=int64 shape=(1,) -- array([1])           -- Opset.make_node.1/Shape##Opset.make_node.1/Shape##Opset.make_node.1/Shape
init: name='init7_s3_0_1_2' type=int64 shape=(3,) -- array([0, 1, 2]) -- Opset.make_node.1/Shape
init: name='init7_s2_-1_1' type=int64 shape=(2,) -- array([-1,  1])   -- Opset.make_node.1/Shape##Opset.make_node.1/Shape
init: name='init7_s2_0_1' type=int64 shape=(2,) -- array([0, 1])      -- Opset.make_node.1/Shape
init: name='init1_s_' type=float32 shape=() -- array([2.], dtype=float32)-- shape_type_compute._cast_inputs.0
init: name='init1_s_2' type=float32 shape=() -- array([3.], dtype=float32)-- shape_type_compute._cast_inputs.0
Concat(init7_s1_0, init7_s1_1, axis=0) -> SliceSlicePattern_init7_s1_1_axis
Concat(init7_s1_0, init7_s1_0, axis=0) -> SliceSlicePattern_init7_s1_0_start
Concat(init7_s1_2, init7_s1_3, axis=0) -> SliceSlicePattern_init7_s1_3_end
  Slice(x, SliceSlicePattern_init7_s1_0_start, SliceSlicePattern_init7_s1_3_end, SliceSlicePattern_init7_s1_1_axis) -> slice_2
    Shape(slice_2) -> _onx_shape0
      Expand(c_lifted_tensor_0, _onx_shape0) -> fill
        Transpose(fill, perm=[1,0]) -> _onx_transpose02
Slice(x, init7_s1_0, init7_s1_2, init7_s1_0) -> slice_3
  Transpose(slice_3, perm=[1,0]) -> _onx_transpose0
Reshape(init7_s3_0_1_2, init7_s2_-1_1) -> _onx_reshape0
  ScatterND(_onx_transpose0, _onx_reshape0, _onx_transpose02) -> _onx_scatternd0
    Transpose(_onx_scatternd0, perm=[1,0]) -> slice_scatter
Reshape(init7_s2_0_1, init7_s2_-1_1) -> _onx_reshape02
  ScatterND(x, _onx_reshape02, slice_scatter) -> output_0
Reshape(init1_s_, init7_s1_1) -> _onx_reshape03
  Add(output_0, _onx_reshape03) -> output_1
Reshape(init1_s_2, init7_s1_1) -> _onx_reshape04
  Add(output_0, _onx_reshape04) -> output_2
output: name='output_1' type=dtype('float32') shape=[5, 5]
output: name='output_2' type=dtype('float32') shape=[5, 5]

custom-tracing¶

FAILED

setitem not implemented for indices=(slice(None, 2, None), slice(None, 3, None))
--DEBUG--
[GraphBuilder-LDO] Message starts, there are 0 initializers, 0 nodes, 1 inputs, 1 outputs.
--PARAMETERS--
dynamic_examples=
--SHAPE--
dynamic_examples=
dynamic_objects=
dynamic_objects_rev=
dynamic_dimensions_source={}
dynamic_alias={}
dynamic_shapes=None
_known_value_shape={}
_known_types={'x': 1}
_known_shapes={'x': (5, 5)}
_known_constants=[]
_known_ranks={}
--TORCH-USERS--
setitem -> {add, add_1}
x -> {setitem}
--TORCH-SHAPES--
x: ('run_node', (('example_value', torch.float32, torch.Size([5, 5])), '')) --- 1:2:(5, 5):
setitem: ('run_node', ('', '')) --- :::
--ONNX--
-- process.graph_module --
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)
-- process.progress --
node 1/5 target=<built-in function setitem>
--
[GraphBuilder-LDO.make_tensor_input] x[1:5x5]
[GraphBuilder-LDO] Message completed, there are 0 initializers, 0 nodes, 1 inputs, 1 outputs.

dynamo-ir¶

opset: domain='pkg.onnxscript.torch_lib.common' version=1
opset: domain='' version=18
opset: domain='pkg.onnxscript.torch_lib' version=1
input: name='x' type=dtype('float32') shape=[5, 5]
Constant(value=[[0], [1],...) -> val_43
Constant(value=[0]) -> val_26
Constant(value=[2]) -> val_29
Constant(value=[0]) -> val_32
Constant(value_ints=[1]) -> val_33
  Slice(x, val_26, val_29, val_32, val_33) -> slice_3
    Transpose(slice_3, perm=[1,0]) -> val_45
Constant(value=[[1.0, 1.0...) -> val_44
  ScatterND(val_45, val_43, val_44, reduction=b'none') -> val_46
    Transpose(val_46, perm=[1,0]) -> slice_scatter
Constant(value=[[0], [1]]) -> val_55
  ScatterND(x, val_55, slice_scatter, reduction=b'none') -> slice_scatter_1
Constant(value=2.0) -> scalar_tensor_default
  Add(slice_scatter_1, scalar_tensor_default) -> add
Constant(value=3.0) -> scalar_tensor_default_1
  Add(slice_scatter_1, scalar_tensor_default_1) -> add_1
output: name='add' type=dtype('float32') shape=[5, 5]
output: name='add_1' type=dtype('float32') shape=[5, 5]

script¶

opset: domain='' version=17
input: name='onnx::Slice_0' type=dtype('float32') shape=[5, 5]
Constant(value=[0]) -> /Constant_output_0
Constant(value=[0]) -> /Constant_1_output_0
Constant(value=[2]) -> /Constant_2_output_0
Constant(value=[1]) -> /Constant_3_output_0
  Slice(onnx::Slice_0, /Constant_1_output_0, /Constant_2_output_0, /Constant_output_0, /Constant_3_output_0) -> /Slice_output_0
Constant(value=[1]) -> /Constant_4_output_0
Constant(value=[0]) -> /Constant_5_output_0
Constant(value=[3]) -> /Constant_6_output_0
Constant(value=[1]) -> /Constant_7_output_0
  Slice(/Slice_output_0, /Constant_5_output_0, /Constant_6_output_0, /Constant_4_output_0, /Constant_7_output_0) -> /Slice_1_output_0
    Shape(/Slice_1_output_0) -> /Shape_output_0
Constant(value=[[1.0]]) -> /Constant_8_output_0
  Expand(/Constant_8_output_0, /Shape_output_0) -> /Expand_output_0
Constant(value=[0, 1, 2]) -> /Constant_9_output_0
Constant(value=[[0], [1]]) -> onnx::Expand_27
Constant(value=[2, 3]) -> /Constant_10_output_0
Constant(value=[2]) -> /Constant_11_output_0
  ConstantOfShape(/Constant_11_output_0, value=[1]) -> /ConstantOfShape_output_0
Constant(value=-1) -> /Constant_12_output_0
  Mul(/ConstantOfShape_output_0, /Constant_12_output_0) -> /Mul_output_0
  Equal(/Constant_10_output_0, /Mul_output_0) -> /Equal_output_0
  Where(/Equal_output_0, /ConstantOfShape_output_0, /Constant_10_output_0) -> /Where_output_0
  Expand(onnx::Expand_27, /Where_output_0) -> /Expand_1_output_0
Constant(value=[-1]) -> /Constant_13_output_0
  Unsqueeze(/Expand_1_output_0, /Constant_13_output_0) -> /Unsqueeze_output_0
Constant(value=[2]) -> /Constant_14_output_0
  ConstantOfShape(/Constant_14_output_0, value=[1]) -> /ConstantOfShape_1_output_0
Constant(value=-1) -> /Constant_15_output_0
  Mul(/ConstantOfShape_1_output_0, /Constant_15_output_0) -> /Mul_1_output_0
  Equal(/Constant_10_output_0, /Mul_1_output_0) -> /Equal_1_output_0
  Where(/Equal_1_output_0, /ConstantOfShape_1_output_0, /Constant_10_output_0) -> /Where_1_output_0
  Expand(/Constant_9_output_0, /Where_1_output_0) -> /Expand_2_output_0
Constant(value=[-1]) -> /Constant_16_output_0
  Unsqueeze(/Expand_2_output_0, /Constant_16_output_0) -> /Unsqueeze_1_output_0
    Concat(/Unsqueeze_output_0, /Unsqueeze_1_output_0, axis=-1) -> /Concat_output_0
Shape(onnx::Slice_0) -> /Shape_1_output_0
Constant(value=[0]) -> /Constant_17_output_0
Constant(value=[2]) -> /Constant_18_output_0
Constant(value=[922337203...) -> /Constant_19_output_0
  Slice(/Shape_1_output_0, /Constant_18_output_0, /Constant_19_output_0, /Constant_17_output_0) -> /Slice_2_output_0
  Concat(/Constant_10_output_0, /Slice_2_output_0, axis=0) -> /Concat_1_output_0
    Reshape(/Expand_output_0, /Concat_1_output_0, allowzero=0) -> /Reshape_output_0
      ScatterND(onnx::Slice_0, /Concat_output_0, /Reshape_output_0) -> /ScatterND_output_0
Constant(value=2.0) -> /Constant_20_output_0
  Add(/ScatterND_output_0, /Constant_20_output_0) -> 60
Constant(value=3.0) -> /Constant_21_output_0
  Add(/ScatterND_output_0, /Constant_21_output_0) -> 62
output: name='60' type=dtype('float32') shape=[5, 5]
output: name='62' type=dtype('float32') shape=[5, 5]

SignatureFloat1¶

forward¶

def forward(self, x, alpha: float = 2.0):
    return torch.sigmoid(self.linear(x)) - self.buff * alpha

custom-fallback¶

opset: domain='' version=18
doc_string: large_model=False, inline=False, external_threshold=102...
input: name='x' type=dtype('float32') shape=[4, 3]
input: name='alpha' type=dtype('float32') shape=[1]
init: name='b_buff' type=float32 shape=(1,) -- array([0.5], dtype=float32)-- DynamoInterpret.placeholder.0
init: name='init1_s_' type=float32 shape=() -- array([1.5], dtype=float32)-- shape_type_compute._cast_inputs.1(mul_Tensor)
init: name='init7_s1_1' type=int64 shape=(1,) -- array([1])           -- Opset.make_node.1/Shape
init: name='linear.weight' type=float32 shape=(1, 3) -- array([0.113, 0.051, 0.52 ], dtype=float32)-- DynamoInterpret.placeholder.1/P(linear.weight)
init: name='linear.bias' type=float32 shape=(1,) -- array([-0.549], dtype=float32)-- DynamoInterpret.placeholder.1/P(linear.bias)
Gemm(x, linear.weight, linear.bias, transB=1) -> linear
  Sigmoid(linear) -> sigmoid
Reshape(init1_s_, init7_s1_1) -> _onx_reshape0
  Mul(b_buff, _onx_reshape0) -> _onx_mul0
    Sub(sigmoid, _onx_mul0) -> output_0
output: name='output_0' type=dtype('float32') shape=[4, 1]

custom-tracing¶

FAILED

Unable to interpret method 'aten_meth_mul', args=(buff, alpha), kwargs={}, dispatcher=None
--DEBUG--
[GraphBuilder-QGU] Message starts, there are 3 initializers, 7 nodes, 2 inputs, 2 outputs.
--PARAMETERS--
dynamic_examples=
--SHAPE--
dynamic_examples=
dynamic_objects=
dynamic_objects_rev=
dynamic_dimensions_source={}
dynamic_alias={}
dynamic_shapes=None
_known_value_shape={}
_known_types={'_sub_Linear__onx_matmul0': 1,
 '_sub_Linear__onx_transpose0': 1,
 '_sub_Linear_input_1': 1,
 '_sub_Linear_linear': 1,
 '_sub_Linear_output': 1,
 'alpha': 1,
 'buff': 1,
 'linear': 1,
 'linear.bias': 1,
 'linear.weight': 1,
 'sigmoid': 1,
 'x': 1}
_known_shapes={'_sub_Linear__onx_matmul0': (4, 1),
 '_sub_Linear__onx_transpose0': (3, 1),
 '_sub_Linear_input_1': (4, 3),
 '_sub_Linear_linear': (4, 1),
 '_sub_Linear_output': (4, 1),
 'alpha': (1,),
 'buff': (1,),
 'linear': (4, 1),
 'linear.bias': (1,),
 'linear.weight': (1, 3),
 'sigmoid': (4, 1),
 'x': (4, 3)}
_known_constants=['_sub_Linear__onx_transpose0', 'buff', 'linear.bias', 'linear.weight']
_known_ranks={}
--TORCH-USERS--
alpha -> {mul}
buff -> {mul}
linear -> {sigmoid}
mul -> {sub}
sigmoid -> {sub}
x -> {linear}
--TORCH-SHAPES--
x: ('run_node', (('example_value', torch.float32, torch.Size([4, 3])), '')) --- 1:2:(4, 3):
alpha: ('run_node', ('', '')) --- 1:1:(1,):
linear: ('run_node', ('', '')) --- 1:2:(4, 1):
sigmoid: ('run_node', ('', '')) --- 1:2:(4, 1):
buff: ('run_node', ('', '')) --- 1:1:(1,):
mul: ('run_node', ('', '')) --- :::
--ONNX--
-- process.graph_module --
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
-- process.progress --
node 5/8 target=mul
--
[GraphBuilder-QGU.make_tensor_input] x[1:4x3]
[GraphBuilder-QGU.make_tensor_input] alpha[1:1]
[GraphBuilder-QGU.make_initializer] linear.weight[torch.float32:torch.float32:[-0.015062670223414898, 0.3196617662906647, 0.5069112181663513]] - SOURCE: GraphBuilder.make_nodes/fromlinear.weight##DynamoInterpret.get_attr.1/P(linear.weight)
[GraphBuilder-QGU.make_initializer] linear.bias[torch.float32:torch.float32:[0.262298583984375]] - SOURCE: GraphBuilder.make_nodes/fromlinear.bias##DynamoInterpret.get_attr.1/P(linear.bias)
[GraphBuilder-QGU.make_initializer] buff[torch.float32:torch.float32:[0.5]] - SOURCE: DynamoInterpret.get_attr.0
[GraphBuilder-QGU.make_node] .make_nodes     [#:#   ] Identity:['x']->['_sub_Linear_input_1']
[GraphBuilder-QGU.make_node] linear          [#:#   ] Transpose:['linear.weight']->['_sub_Linear__onx_transpose0']
[GraphBuilder-QGU.make_node] Opset           [##:#  ] MatMul:['_sub_Linear_input_1', '_sub_Linear__onx_transpose0']->['_sub_Linear__onx_matmul0']
[GraphBuilder-QGU.make_node] Opset2          [##:#  ] Add:['_sub_Linear__onx_matmul0', 'linear.bias']->['_sub_Linear_linear']
[GraphBuilder-QGU.make_node] .output         [#:#   ] Identity:['_sub_Linear_linear']->['_sub_Linear_output']
[GraphBuilder-QGU.make_node] .make_nodes2    [#:#   ] Identity:['_sub_Linear_output']->['linear']
[GraphBuilder-QGU.make_node] Opset3          [#:#   ] Sigmoid:['linear']->['sigmoid']
[GraphBuilder-QGU] Message completed, there are 3 initializers, 7 nodes, 2 inputs, 2 outputs.

dynamo-ir¶

FAILED

Input mismatch, inputs[0]=(T1r2,float) but names=['x'], model=SignatureFloat1, export='dynamo-ir'

script¶

FAILED

[ONNXRuntimeError] : 2 : INVALID_ARGUMENT : Unexpected input data type. Actual: (tensor(float)) , expected: (tensor(double))

SignatureInt1¶

forward¶

def forward(self, x, i: int = 2):
    return torch.sigmoid(self.linear(x)) - self.buff + x[:, i : i + 1]

custom-fallback¶

opset: domain='' version=18
doc_string: large_model=False, inline=False, external_threshold=102...
input: name='x' type=dtype('float32') shape=[4, 3]
input: name='i' type=dtype('int64') shape=[1]
init: name='b_buff' type=float32 shape=(1,) -- array([0.5], dtype=float32)-- DynamoInterpret.placeholder.0
init: name='init7_s1_1' type=int64 shape=(1,) -- array([1])           -- Opset.make_node.1/Shape##Opset.make_node.1/Shape
init: name='init7_s1_2' type=int64 shape=(1,) -- array([2])           -- Opset.make_node.1/Shape
init: name='linear.weight' type=float32 shape=(1, 3) -- array([ 0.08 , -0.448,  0.258], dtype=float32)-- DynamoInterpret.placeholder.1/P(linear.weight)
init: name='linear.bias' type=float32 shape=(1,) -- array([-0.028], dtype=float32)-- DynamoInterpret.placeholder.1/P(linear.bias)
Gemm(x, linear.weight, linear.bias, transB=1) -> linear
  Sigmoid(linear) -> sigmoid
    Sub(sigmoid, b_buff) -> sub
Slice(x, init7_s1_1, init7_s1_2, init7_s1_1) -> slice_2
  Add(sub, slice_2) -> output_0
output: name='output_0' type=dtype('float32') shape=[4, 1]

custom-tracing¶

opset: domain='' version=18
doc_string: large_model=False, inline=False, external_threshold=102...
input: name='x' type=dtype('float32') shape=[4, 3]
input: name='i' type=dtype('int64') shape=[1]
init: name='linear.weight' type=float32 shape=(1, 3) -- array([ 0.452, -0.168, -0.495], dtype=float32)-- GraphBuilder.make_nodes/fromlinear.weight##DynamoInterpret.get_attr.1/P(linear.weight)
init: name='linear.bias' type=float32 shape=(1,) -- array([0.232], dtype=float32)-- GraphBuilder.make_nodes/fromlinear.bias##DynamoInterpret.get_attr.1/P(linear.bias)
init: name='buff' type=float32 shape=(1,) -- array([0.5], dtype=float32)-- DynamoInterpret.get_attr.0
init: name='init7_s_1' type=int64 shape=() -- array([1])              -- shape_type_compute._cast_inputs.1(add)
init: name='init7_s1_1' type=int64 shape=(1,) -- array([1])           -- Opset.make_node.1/Shape
init: name='getitem_axis' type=int64 shape=(2,) -- array([0, 1])      -- DynamoInterpreter._getitem_slice.axis.1
init: name='init7_s1_4' type=int64 shape=(1,) -- array([4])           -- Opset.make_node.1/Shape
init: name='init7_s1_0' type=int64 shape=(1,) -- array([0])           -- Opset.make_node.1/Shape
init: name='getitem_step' type=int64 shape=(2,) -- array([1, 1])      -- DynamoInterpreter._getitem_slice.3
Concat(init7_s1_0, i, axis=0) -> _onx_concat02
Gemm(x, linear.weight, linear.bias, transB=1) -> _sub_Linear_linear
  Sigmoid(_sub_Linear_linear) -> sigmoid
    Sub(sigmoid, buff) -> sub
Reshape(init7_s_1, init7_s1_1) -> _onx_reshape0
  Add(i, _onx_reshape0) -> _onx_add0
    Concat(init7_s1_4, _onx_add0, axis=0) -> _onx_concat0
  Slice(x, _onx_concat02, _onx_concat0, getitem_axis, getitem_step) -> getitem
    Add(sub, getitem) -> output
output: name='output' type=dtype('float32') shape=['d_output_0', 'd_output_1']

dynamo-ir¶

FAILED

Input mismatch, inputs[0]=(T1r2,int) but names=['x'], model=SignatureInt1, export='dynamo-ir'

script¶

FAILED

[ONNXRuntimeError] : 1 : FAIL : Non-zero status code returned while running Slice node. Name:'/Slice' Status Message: slice.cc:195 FillVectorsFromInput Starts must be a 1-D array

SignatureInt2¶

forward¶

def forward(self, x, i: int = 2):
    return torch.sigmoid(self.linear(x)) - self.buff + x[:, i]

custom-fallback¶

opset: domain='' version=18
doc_string: large_model=False, inline=False, external_threshold=102...
input: name='x' type=dtype('float32') shape=[4, 3]
input: name='i' type=dtype('int64') shape=[1]
init: name='b_buff' type=float32 shape=(1,) -- array([0.5], dtype=float32)-- DynamoInterpret.placeholder.0
init: name='init7_s_1' type=int64 shape=() -- array([1])              -- Opset.make_node.1/Shape
init: name='linear.weight' type=float32 shape=(1, 3) -- array([-0.283,  0.411,  0.159], dtype=float32)-- DynamoInterpret.placeholder.1/P(linear.weight)
init: name='linear.bias' type=float32 shape=(1,) -- array([0.344], dtype=float32)-- DynamoInterpret.placeholder.1/P(linear.bias)
Gather(x, init7_s_1, axis=1) -> select
Gemm(x, linear.weight, linear.bias, transB=1) -> linear
  Sigmoid(linear) -> sigmoid
    Sub(sigmoid, b_buff) -> sub
  Add(sub, select) -> output_0
output: name='output_0' type=dtype('float32') shape=[4, 4]

custom-tracing¶

FAILED

One index is given as an integer i but this requires to append a node 'Squeeze' after this one and this is not yet implemented. You can replace the integer by `i:i+1`
--DEBUG--
[GraphBuilder-UBO] Message starts, there are 4 initializers, 8 nodes, 2 inputs, 2 outputs.
--PARAMETERS--
dynamic_examples=
--SHAPE--
dynamic_examples=
dynamic_objects=
dynamic_objects_rev=
dynamic_dimensions_source={}
dynamic_alias={}
dynamic_shapes=None
_known_value_shape={}
_known_types={'_sub_Linear__onx_matmul0': 1,
 '_sub_Linear__onx_transpose0': 1,
 '_sub_Linear_input_1': 1,
 '_sub_Linear_linear': 1,
 '_sub_Linear_output': 1,
 'buff': 1,
 'getitem_axis': 7,
 'i': 7,
 'linear': 1,
 'linear.bias': 1,
 'linear.weight': 1,
 'sigmoid': 1,
 'sub': 1,
 'x': 1}
_known_shapes={'_sub_Linear__onx_matmul0': (4, 1),
 '_sub_Linear__onx_transpose0': (3, 1),
 '_sub_Linear_input_1': (4, 3),
 '_sub_Linear_linear': (4, 1),
 '_sub_Linear_output': (4, 1),
 'buff': (1,),
 'getitem_axis': (2,),
 'i': (1,),
 'linear': (4, 1),
 'linear.bias': (1,),
 'linear.weight': (1, 3),
 'sigmoid': (4, 1),
 'sub': (4, 1),
 'x': (4, 3)}
_known_constants=['_sub_Linear__onx_transpose0',
 'buff',
 'getitem_axis',
 'linear.bias',
 'linear.weight']
_known_ranks={}
--TORCH-USERS--
buff -> {sub}
getitem -> {add}
i -> {getitem}
linear -> {sigmoid}
sigmoid -> {sub}
sub -> {add}
x -> {linear, getitem}
--TORCH-SHAPES--
x: ('run_node', (('example_value', torch.float32, torch.Size([4, 3])), '')) --- 1:2:(4, 3):
i: ('run_node', ('', '')) --- 7:1:(1,):
linear: ('run_node', ('', '')) --- 1:2:(4, 1):
sigmoid: ('run_node', ('', '')) --- 1:2:(4, 1):
buff: ('run_node', ('', '')) --- 1:1:(1,):
sub: ('run_node', ('', '')) --- 1:2:(4, 1):
getitem: ('run_node', ('', '')) --- :::
--ONNX--
-- process.graph_module --
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
-- process.progress --
node 6/9 target=<built-in function getitem>
--
[GraphBuilder-UBO.make_tensor_input] x[1:4x3]
[GraphBuilder-UBO.make_tensor_input] i[7:1]
[GraphBuilder-UBO.make_initializer] linear.weight[torch.float32:torch.float32:[-0.44255390763282776, 0.37168198823928833, -0.3502069413661957]] - SOURCE: GraphBuilder.make_nodes/fromlinear.weight##DynamoInterpret.get_attr.1/P(linear.weight)
[GraphBuilder-UBO.make_initializer] linear.bias[torch.float32:torch.float32:[0.5427876114845276]] - SOURCE: GraphBuilder.make_nodes/fromlinear.bias##DynamoInterpret.get_attr.1/P(linear.bias)
[GraphBuilder-UBO.make_initializer] buff[torch.float32:torch.float32:[0.5]] - SOURCE: DynamoInterpret.get_attr.0
[GraphBuilder-UBO.make_initializer] getitem_axis[int64:int64:[0, 1]] - SOURCE: DynamoInterpreter._getitem_slice.axis.1
[GraphBuilder-UBO.make_node] .make_nodes     [#:#   ] Identity:['x']->['_sub_Linear_input_1']
[GraphBuilder-UBO.make_node] linear          [#:#   ] Transpose:['linear.weight']->['_sub_Linear__onx_transpose0']
[GraphBuilder-UBO.make_node] Opset           [##:#  ] MatMul:['_sub_Linear_input_1', '_sub_Linear__onx_transpose0']->['_sub_Linear__onx_matmul0']
[GraphBuilder-UBO.make_node] Opset2          [##:#  ] Add:['_sub_Linear__onx_matmul0', 'linear.bias']->['_sub_Linear_linear']
[GraphBuilder-UBO.make_node] .output         [#:#   ] Identity:['_sub_Linear_linear']->['_sub_Linear_output']
[GraphBuilder-UBO.make_node] .make_nodes2    [#:#   ] Identity:['_sub_Linear_output']->['linear']
[GraphBuilder-UBO.make_node] Opset3          [#:#   ] Sigmoid:['linear']->['sigmoid']
[GraphBuilder-UBO.make_node] sub             [##:#  ] Sub:['sigmoid', 'buff']->['sub']
[GraphBuilder-UBO] Message completed, there are 4 initializers, 8 nodes, 2 inputs, 2 outputs.

dynamo-ir¶

FAILED

Input mismatch, inputs[0]=(T1r2,int) but names=['x'], model=SignatureInt2, export='dynamo-ir'

script¶

opset: domain='' version=17
input: name='onnx::Gemm_0' type=dtype('float32') shape=[4, 3]
input: name='onnx::Gather_1' type=dtype('int64') shape=None
init: name='buff' type=float32 shape=(1,) -- array([0.5], dtype=float32)
init: name='linear.weight' type=float32 shape=(1, 3) -- array([-0.454, -0.57 , -0.019], dtype=float32)
init: name='linear.bias' type=float32 shape=(1,) -- array([-0.154], dtype=float32)
Gather(onnx::Gemm_0, onnx::Gather_1, axis=1) -> /Gather_output_0
Gemm(onnx::Gemm_0, linear.weight, linear.bias, alpha=1.00, beta=1.00, transB=1) -> /linear/Gemm_output_0
  Sigmoid(/linear/Gemm_output_0) -> /Sigmoid_output_0
    Sub(/Sigmoid_output_0, buff) -> /Sub_output_0
  Add(/Sub_output_0, /Gather_output_0) -> 9
output: name='9' type=dtype('float32') shape=[4, 4]

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)
    )

custom-fallback¶

opset: domain='' version=18
doc_string: large_model=False, inline=False, external_threshold=102...
input: name='x' type=dtype('float32') shape=[4, 3]
input: name='lx_0' type=dtype('float32') shape=[4, 1]
input: name='lx_1' type=dtype('float32') shape=[4, 2]
init: name='b_buff' type=float32 shape=(1,) -- array([0.5], dtype=float32)-- DynamoInterpret.placeholder.0
init: name='init7_s1_1' type=int64 shape=(1,) -- array([1])           -- Opset.make_node.1/Shape
init: name='linear.weight' type=float32 shape=(1, 3) -- array([ 0.309, -0.131,  0.478], dtype=float32)-- DynamoInterpret.placeholder.1/P(linear.weight)
init: name='linear.bias' type=float32 shape=(1,) -- array([0.189], dtype=float32)-- DynamoInterpret.placeholder.1/P(linear.bias)
Gemm(x, linear.weight, linear.bias, transB=1) -> linear
  Sigmoid(linear) -> sigmoid
    Sub(sigmoid, b_buff) -> sub
ReduceSum(lx_1, init7_s1_1, keepdims=1) -> sum_1
  Mul(lx_0, sum_1) -> mul
    Add(sub, mul) -> output_0
output: name='output_0' type=dtype('float32') shape=[4, 1]

custom-tracing¶

opset: domain='' version=18
doc_string: large_model=False, inline=False, external_threshold=102...
input: name='x' type=dtype('float32') shape=[4, 3]
input: name='lx' type='NOTENSOR' shape=None
init: name='linear.weight' type=float32 shape=(1, 3) -- array([ 0.21 , -0.209, -0.159], dtype=float32)-- GraphBuilder.make_nodes/fromlinear.weight##DynamoInterpret.get_attr.1/P(linear.weight)
init: name='linear.bias' type=float32 shape=(1,) -- array([-0.308], dtype=float32)-- GraphBuilder.make_nodes/fromlinear.bias##DynamoInterpret.get_attr.1/P(linear.bias)
init: name='buff' type=float32 shape=(1,) -- array([0.5], dtype=float32)-- DynamoInterpret.get_attr.0
init: name='init7_s_0' type=int64 shape=() -- array([0])              -- DynamoInterpreter.getitem.1
init: name='init7_s_1' type=int64 shape=() -- array([1])              -- DynamoInterpreter.getitem.1
init: name='init7_s1_1' type=int64 shape=(1,) -- array([1])           -- Opset.make_node.1/Shape
Gemm(x, linear.weight, linear.bias, transB=1) -> _sub_Linear_linear
  Sigmoid(_sub_Linear_linear) -> sigmoid
    Sub(sigmoid, buff) -> sub
SequenceAt(lx, init7_s_0) -> getitem
SequenceAt(lx, init7_s_1) -> getitem_1
  ReduceSum(getitem_1, init7_s1_1, keepdims=1) -> sum_1
  Mul(getitem, sum_1) -> mul
    Add(sub, mul) -> output
output: name='output' type=dtype('float32') shape=[4, 1]

dynamo-ir¶

opset: domain='pkg.onnxscript.torch_lib.common' version=1
opset: domain='' version=18
input: name='x' type=dtype('float32') shape=[4, 3]
input: name='lx_0' type=dtype('float32') shape=[4, 1]
input: name='lx_1' type=dtype('float32') shape=[4, 2]
init: name='linear.bias' type=float32 shape=(1,) -- array([0.256], dtype=float32)
init: name='buff' type=float32 shape=(1,) -- array([0.5], dtype=float32)
Constant(value=[[0.183093...) -> t
  Gemm(x, t, linear.bias, beta=1.00, transB=0, alpha=1.00, transA=0) -> addmm
    Sigmoid(addmm) -> sigmoid
      Sub(sigmoid, buff) -> sub
Constant(value=[1]) -> val_3
  ReduceSum(lx_1, val_3, noop_with_empty_axes=0, keepdims=1) -> sum_1
    Mul(lx_0, sum_1) -> mul
      Add(sub, mul) -> add
output: name='add' type=dtype('float32') shape=[4, 1]

script¶

opset: domain='' version=17
input: name='onnx::Gemm_0' type=dtype('float32') shape=[4, 3]
input: name='onnx::Mul_1' type=dtype('float32') shape=[4, 1]
input: name='onnx::ReduceSum_2' type=dtype('float32') shape=[4, 2]
init: name='buff' type=float32 shape=(1,) -- array([0.5], dtype=float32)
init: name='linear.weight' type=float32 shape=(1, 3) -- array([ 0.285, -0.179,  0.475], dtype=float32)
init: name='linear.bias' type=float32 shape=(1,) -- array([-0.059], dtype=float32)
Constant(value=[1]) -> onnx::ReduceSum_9
  ReduceSum(onnx::ReduceSum_2, onnx::ReduceSum_9, keepdims=1) -> /ReduceSum_output_0
    Mul(onnx::Mul_1, /ReduceSum_output_0) -> /Mul_output_0
Gemm(onnx::Gemm_0, linear.weight, linear.bias, alpha=1.00, beta=1.00, transB=1) -> /linear/Gemm_output_0
  Sigmoid(/linear/Gemm_output_0) -> /Sigmoid_output_0
    Sub(/Sigmoid_output_0, buff) -> /Sub_output_0
      Add(/Sub_output_0, /Mul_output_0) -> 12
output: name='12' type=dtype('float32') shape=[4, 1]

SignatureListFixedWithNone¶

forward¶

def forward(self, lx):
    print(lx)
    print(lx[1])
    x = lx[0]
    if lx[1] is not None:
        x += lx[1]
    if lx[2] is not None:
        x += lx[2]
    return x

custom-fallback¶

FAILED

Input mismatch, inputs[0]=(#3[T1r2,T1r2,None],) but names=['lx_0', 'lx_1'], model=SignatureListFixedWithNone, export='custom-fallback'

custom-tracing¶

CustomProxy(lx) CustomProxy(getitem) FAILED

Unable to create an input 'lx' with type #3[T1r2,T1r2,None]
--DEBUG--
[GraphBuilder-FZK] Message starts, there are 0 initializers, 0 nodes, 0 inputs, 0 outputs.
--PARAMETERS--
dynamic_examples=
--SHAPE--
dynamic_examples=
dynamic_objects=
dynamic_objects_rev=
dynamic_dimensions_source={}
dynamic_alias={}
dynamic_shapes=None
_known_value_shape={}
_known_types={}
_known_shapes={}
_known_constants=[]
_known_ranks={}
--TORCH-USERS--
lx -> {getitem_1, getitem_5, getitem_4, getitem, getitem_2, getitem_3}
--TORCH-SHAPES--
lx: ('run_node', ('', '')) --- :::
--ONNX--
-- process.graph_module --
graph():
    %lx : [num_users=6] = placeholder[target=lx]
    %getitem : [num_users=0] = call_function[target=operator.getitem](args = (%lx, 1), kwargs = {})
    %getitem_1 : [num_users=1] = call_function[target=operator.getitem](args = (%lx, 0), kwargs = {})
    %getitem_2 : [num_users=0] = call_function[target=operator.getitem](args = (%lx, 1), kwargs = {})
    %getitem_3 : [num_users=1] = call_function[target=operator.getitem](args = (%lx, 1), kwargs = {})
    %add : [num_users=1] = call_function[target=operator.add](args = (%getitem_1, %getitem_3), kwargs = {})
    %getitem_4 : [num_users=0] = call_function[target=operator.getitem](args = (%lx, 2), kwargs = {})
    %getitem_5 : [num_users=1] = call_function[target=operator.getitem](args = (%lx, 2), kwargs = {})
    %add_1 : [num_users=1] = call_function[target=operator.add](args = (%add, %getitem_5), kwargs = {})
    return add_1
-- process.progress --
node 0/10 target=lx
--
[GraphBuilder-FZK] Message completed, there are 0 initializers, 0 nodes, 0 inputs, 0 outputs.

dynamo-ir¶

FAILED

Input mismatch, inputs[0]=(#3[T1r2,T1r2,None],) but names=['lx_0', 'lx_1'], model=SignatureListFixedWithNone, export='dynamo-ir'

script¶

FAILED

Input mismatch, inputs[0]=(#3[T1r2,T1r2,None],) but names=['inp.1', 'inp'], model=SignatureListFixedWithNone, export='script'

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

custom-fallback¶

opset: domain='' version=18
doc_string: large_model=False, inline=False, external_threshold=102...
input: name='x' type=dtype('float32') shape=[4, 3]
input: name='lx_0' type=dtype('float32') shape=[4, 1]
input: name='lx_1' type=dtype('float32') shape=[4, 2]
init: name='b_buff' type=float32 shape=(1,) -- array([0.5], dtype=float32)-- DynamoInterpret.placeholder.0
init: name='init7_s1_1' type=int64 shape=(1,) -- array([1])           -- Opset.make_node.1/Shape
init: name='linear.weight' type=float32 shape=(1, 3) -- array([-0.234,  0.473, -0.379], dtype=float32)-- DynamoInterpret.placeholder.1/P(linear.weight)
init: name='linear.bias' type=float32 shape=(1,) -- array([0.4], dtype=float32)-- DynamoInterpret.placeholder.1/P(linear.bias)
Concat(lx_0, lx_1, axis=1) -> cat
  ReduceSum(cat, init7_s1_1, keepdims=1) -> sum_1
Gemm(x, linear.weight, linear.bias, transB=1) -> linear
  Sigmoid(linear) -> sigmoid
    Sub(sigmoid, b_buff) -> sub
    Add(sub, sum_1) -> output_0
output: name='output_0' type=dtype('float32') shape=[4, 1]

custom-tracing¶

FAILED

Type is unknown for result 'l', known_types={'x': 1}
--DEBUG--
[GraphBuilder-CKM] Message starts, there are 0 initializers, 0 nodes, 2 inputs, 2 outputs.
--PARAMETERS--
dynamic_examples=
--SHAPE--
dynamic_examples=
dynamic_objects=
dynamic_objects_rev=
dynamic_dimensions_source={}
dynamic_alias={}
dynamic_shapes=None
_known_value_shape={}
_known_types={'x': 1}
_known_shapes={'x': (4, 3)}
_known_constants=[]
_known_ranks={}
--TORCH-USERS--
cat -> {sum_1}
lx -> {cat}
x -> {linear}
--TORCH-SHAPES--
x: ('run_node', (('example_value', torch.float32, torch.Size([4, 3])), '')) --- 1:2:(4, 3):
lx: ('run_node', ('', '')) --- :::
cat: ('run_node', ('', '')) --- :::
--ONNX--
-- process.graph_module --
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
-- process.progress --
node 2/10 target=<built-in method cat of type object at 0x7fa9d1ef7ba0>
--
[GraphBuilder-CKM.make_tensor_input] x[1:4x3]
[GraphBuilder-CKM.make_tensor_input] lx[0:]
[GraphBuilder-CKM] Message completed, there are 0 initializers, 0 nodes, 2 inputs, 2 outputs..

dynamo-ir¶

opset: domain='pkg.onnxscript.torch_lib.common' version=1
opset: domain='' version=18
input: name='x' type=dtype('float32') shape=[4, 3]
input: name='lx_0' type=dtype('float32') shape=[4, 1]
input: name='lx_1' type=dtype('float32') shape=[4, 2]
init: name='linear.bias' type=float32 shape=(1,) -- array([0.258], dtype=float32)
init: name='buff' type=float32 shape=(1,) -- array([0.5], dtype=float32)
Concat(lx_0, lx_1, axis=1) -> cat
Constant(value=[1]) -> val_3
  ReduceSum(cat, val_3, noop_with_empty_axes=0, keepdims=1) -> sum_1
Constant(value=[[0.061144...) -> t
  Gemm(x, t, linear.bias, beta=1.00, transB=0, alpha=1.00, transA=0) -> addmm
    Sigmoid(addmm) -> sigmoid
      Sub(sigmoid, buff) -> sub
    Add(sub, sum_1) -> add
output: name='add' type=dtype('float32') shape=[4, 1]

script¶

opset: domain='' version=17
input: name='onnx::Gemm_0' type=dtype('float32') shape=[4, 3]
input: name='onnx::Concat_1' type=dtype('float32') shape=[4, 1]
input: name='onnx::Concat_2' type=dtype('float32') shape=[4, 2]
init: name='buff' type=float32 shape=(1,) -- array([0.5], dtype=float32)
init: name='linear.weight' type=float32 shape=(1, 3) -- array([-0.45 , -0.575,  0.551], dtype=float32)
init: name='linear.bias' type=float32 shape=(1,) -- array([-0.343], dtype=float32)
Concat(onnx::Concat_1, onnx::Concat_2, axis=1) -> /Concat_output_0
Constant(value=[1]) -> onnx::ReduceSum_7
  ReduceSum(/Concat_output_0, onnx::ReduceSum_7, keepdims=1) -> /ReduceSum_output_0
Gemm(onnx::Gemm_0, linear.weight, linear.bias, alpha=1.00, beta=1.00, transB=1) -> /linear/Gemm_output_0
  Sigmoid(/linear/Gemm_output_0) -> /Sigmoid_output_0
    Sub(/Sigmoid_output_0, buff) -> /Sub_output_0
    Add(/Sub_output_0, /ReduceSum_output_0) -> 12
output: name='12' type=dtype('float32') shape=[4, 1]

SignatureShapeAsIndex¶

forward¶

def forward(self, x, y):
    t = torch.sigmoid(self.linear(x)) + x
    return t[:, : y.shape[1]]

custom-fallback¶

opset: domain='' version=18
doc_string: large_model=False, inline=False, external_threshold=102...
input: name='x' type=dtype('float32') shape=[4, 3]
input: name='y' type=dtype('float32') shape=[4, 2]
init: name='init7_s1_0' type=int64 shape=(1,) -- array([0])           -- Opset.make_node.1/Shape
init: name='init7_s1_2' type=int64 shape=(1,) -- array([2])           -- Opset.make_node.1/Shape
init: name='init7_s1_1' type=int64 shape=(1,) -- array([1])           -- Opset.make_node.1/Shape
init: name='linear.weight' type=float32 shape=(1, 3) -- array([-0.269, -0.268, -0.107], dtype=float32)-- DynamoInterpret.placeholder.1/P(linear.weight)
init: name='linear.bias' type=float32 shape=(1,) -- array([-0.276], dtype=float32)-- DynamoInterpret.placeholder.1/P(linear.bias)
Gemm(x, linear.weight, linear.bias, transB=1) -> linear
  Sigmoid(linear) -> sigmoid
    Add(sigmoid, x) -> add
      Slice(add, init7_s1_0, init7_s1_2, init7_s1_1) -> output_0
output: name='output_0' type=dtype('float32') shape=[4, 2]

custom-tracing¶

opset: domain='' version=18
doc_string: large_model=False, inline=False, external_threshold=102...
input: name='x' type=dtype('float32') shape=[4, 3]
input: name='y' type=dtype('float32') shape=[4, 2]
init: name='linear.weight' type=float32 shape=(1, 3) -- array([ 0.046,  0.272, -0.34 ], dtype=float32)-- GraphBuilder.make_nodes/fromlinear.weight##DynamoInterpret.get_attr.1/P(linear.weight)
init: name='linear.bias' type=float32 shape=(1,) -- array([0.022], dtype=float32)-- GraphBuilder.make_nodes/fromlinear.bias##DynamoInterpret.get_attr.1/P(linear.bias)
init: name='init7_s1_1' type=int64 shape=(1,) -- array([1])           -- Opset.make_node.1/Shape
init: name='init7_s1_0' type=int64 shape=(1,) -- array([0])           -- Opset.make_node.1/Shape##Opset.make_node.1/Shape
init: name='getitem_1_axis' type=int64 shape=(2,) -- array([0, 1])    -- DynamoInterpreter._getitem_slice.axis.1
init: name='init7_s1_4' type=int64 shape=(1,) -- array([4])           -- Opset.make_node.1/Shape
init: name='getitem_1_start' type=int64 shape=(2,) -- array([0, 0])   -- DynamoInterpreter._getitem_slice.2
init: name='getitem_1_step' type=int64 shape=(2,) -- array([1, 1])    -- DynamoInterpreter._getitem_slice.3
Gemm(x, linear.weight, linear.bias, transB=1) -> _sub_Linear_linear
  Sigmoid(_sub_Linear_linear) -> sigmoid
    Add(sigmoid, x) -> add
Shape(y) -> getattr_1
  Gather(getattr_1, init7_s1_1) -> _onx_gather0
    Squeeze(_onx_gather0, init7_s1_0) -> getitem
      Unsqueeze(getitem, init7_s1_0) -> _onx_unsqueeze0
        Concat(init7_s1_4, _onx_unsqueeze0, axis=0) -> _onx_concat0
      Slice(add, getitem_1_start, _onx_concat0, getitem_1_axis, getitem_1_step) -> output
output: name='output' type=dtype('float32') shape=['d_output_0', 'd_output_1']

dynamo-ir¶

opset: domain='pkg.onnxscript.torch_lib.common' version=1
opset: domain='' version=18
input: name='x' type=dtype('float32') shape=[4, 3]
input: name='y' type=dtype('float32') shape=[4, 2]
init: name='linear.bias' type=float32 shape=(1,) -- array([0.063], dtype=float32)
Constant(value_ints=[1]) -> val_23
Constant(value=[[0.370703...) -> t
  Gemm(x, t, linear.bias, beta=1.00, transB=0, alpha=1.00, transA=0) -> addmm
    Sigmoid(addmm) -> sigmoid
      Add(sigmoid, x) -> add
Constant(value=[0]) -> val_14
Constant(value=[2]) -> val_18
Constant(value=[1]) -> val_22
  Slice(add, val_14, val_18, val_22, val_23) -> slice_2
output: name='slice_2' type=dtype('float32') shape=[4, 2]

script¶

FAILED

Input mismatch, inputs[0]=(T1r2,T1r2) but names=['onnx::Gemm_0'], model=SignatureShapeAsIndex, export='script'

TypeBFloat16¶

forward¶

def forward(self, x):
    xb = x.to(torch.bfloat16)
    return (xb + xb).to(torch.float32)

custom-fallback¶

opset: domain='' version=18
doc_string: large_model=False, inline=False, external_threshold=102...
input: name='x' type=dtype('float32') shape=[4, 4]
Add(x, x) -> add-x
  Cast(add-x, to=16) -> add
    Cast(add, to=1) -> output_0
output: name='output_0' type=dtype('float32') shape=[4, 4]

custom-tracing¶

opset: domain='' version=18
doc_string: large_model=False, inline=False, external_threshold=102...
input: name='x' type=dtype('float32') shape=[4, 4]
Add(x, x) -> add-x
  Cast(add-x, to=16) -> add
    Cast(add, to=1) -> output
output: name='output' type=dtype('float32') shape=[4, 4]

dynamo-ir¶

FAILED

[ONNXRuntimeError] : 9 : NOT_IMPLEMENTED : Could not find an implementation for Add(14) node with name 'node_Add_1'

script¶

FAILED

[ONNXRuntimeError] : 9 : NOT_IMPLEMENTED : Could not find an implementation for Add(14) node with name '/Add'

Summary¶

case	custom-fallback	custom-tracing	dynamo-ir	script
AtenAsStrided
AtenInterpolate
AtenNonZero
AtenNonZeroTuple
AtenRollPos
AtenRollRelu
BuildInIsInstance		FAIL
BuildInLen		FAIL
ComplexPolar	FAIL	FAIL		FAIL
ControlFlowCond		FAIL		FAIL
ControlFlowCond2Inputs		FAIL	FAIL	FAIL
ControlFlowCond2Outputs		FAIL	FAIL	FAIL
ControlFlowCondConstant		FAIL		FAIL
ControlFlowCondNestedModule		FAIL		FAIL
ControlFlowNestCond		FAIL		FAIL
ControlFlowScan		FAIL	FAIL	FAIL
ControlFlowScan2Carried		FAIL	FAIL	FAIL
ControlFlowScanCDist		FAIL	FAIL	FAIL
ControlFlowScanCDist2		FAIL	FAIL	FAIL
ControlFlowScanCDistXY		FAIL	FAIL	FAIL
CreateFromShape		FAIL		FAIL
CreateFromShapeThroughFunction		FAIL		FAIL
CropLastDimensionWithTensorContent	FAIL		FAIL
CropLastDimensionWithTensorShape				FAIL
InplaceAdd
InplaceAdd_		FAIL
InplaceAdd_Mul		FAIL
InplaceCloneAdd_		FAIL
InplaceSetItemEllipsis_1		FAIL	FAIL	FAIL
InplaceSetItemEllipsis_2		FAIL	FAIL	FAIL
InplaceSetItemMask		FAIL	FAIL
InplaceSetItemSquare		FAIL
InplaceSetItemSquareAdd		FAIL
InplaceSetItemSquareAdd2		FAIL
SignatureFloat1		FAIL	FAIL	FAIL
SignatureInt1			FAIL	FAIL
SignatureInt2		FAIL	FAIL
SignatureListFixedLength
SignatureListFixedWithNone	FAIL	FAIL	FAIL	FAIL
SignatureListVariableLength		FAIL
SignatureShapeAsIndex				FAIL
TypeBFloat16			FAIL	FAIL