xadupre
commented
1 year ago What is the error message? Does it work with x = op.Cast(op.Constant(value_floats=[1.0]), to=FLOAT16.dtype)?
Open xiaowuhu opened 1 year ago
xadupre
commented
1 year ago What is the error message? Does it work with x = op.Cast(op.Constant(value_floats=[1.0]), to=FLOAT16.dtype)?
pranavsharma
commented
1 year ago Is this an issue with onnxruntime or onnx script? Please file the issue in an appropriate place.
justinchuby
commented
1 year ago Updated issue with error message. I think this is an issue with onnx shape inferencing. @gramalingam I wonder if this is part of the fix we have since 1.14?
justinchuby
commented
1 year ago The error manifests itself in test like so:
ONNX Runtime raises [ONNXRuntimeError] : 1 : FAIL : Node (_0x7b6efe0_n21) Op (Loop) [TypeInferenceError] Graph attribute inferencing failed: Node (_0x7b6efe0_n12) Op (If) [TypeInferenceError] Mismatched tensor element type: source=10 target=1 when executing test ops_test.TestOutputConsistencyFullGraphCPU.test_output_match_opinfo__ops_aten_embedding_bag_cpu_float16 in ONNX Script TorchLib.
To recreate this report, use
CREATE_REPRODUCTION_REPORT=1 python -m pytest onnxscript/tests/function_libs/torch_lib/ops_test.py -k test_output_match_opinfo__ops_aten_embedding_bag_cpu_float16import google.protobuf.text_format
import numpy as np
from numpy import array, float16, float32, float64, int32, int64
import onnx
import onnxruntime as ort
# Run n times
N = 1
onnx_model_text = """
ir_version: 8
producer_name: "pytorch"
producer_version: "2.1.0"
graph {
node {
output: "_val_3"
name: "Constant_0"
op_type: "Constant"
attribute {
name: "value"
t {
dims: 1
data_type: 7
raw_data: "\377\377\377\377\377\377\377\377"
}
type: TENSOR
}
doc_string: ""
}
node {
input: "input_1"
input: "_val_3"
output: "_val_4"
name: "Reshape_1"
op_type: "Reshape"
attribute {
name: "allowzero"
i: 0
type: INT
}
doc_string: ""
}
node {
input: "_val_4"
output: "_val_5"
name: "Shape_2"
op_type: "Shape"
attribute {
name: "start"
i: 0
type: INT
}
doc_string: ""
}
node {
output: "_val_6"
name: "Constant_3"
op_type: "Constant"
attribute {
name: "value"
t {
data_type: 7
raw_data: "\001\000\000\000\000\000\000\000"
}
type: TENSOR
}
doc_string: ""
}
node {
input: "_val_6"
input: "_val_5"
output: "_val_7"
name: "Expand_4"
op_type: "Expand"
doc_string: ""
}
node {
input: "_val_7"
input: "input_0"
output: "_val_8"
name: "CastLike_5"
op_type: "CastLike"
doc_string: ""
}
node {
input: "input_0"
input: "input_1"
input: "offsets"
input: "_val_8"
output: "_val_9"
name: "_aten_embedding_bag_onnx_6"
op_type: "_aten_embedding_bag_onnx"
attribute {
name: "include_last_offset"
i: 1
type: INT
}
attribute {
name: "mode"
i: 0
type: INT
}
doc_string: ""
domain: "pkg.onnxscript.torch_lib"
}
node {
input: "input_1"
output: "_val_10"
name: "Shape_7"
op_type: "Shape"
attribute {
name: "end"
i: 0
type: INT
}
attribute {
name: "start"
i: 0
type: INT
}
doc_string: ""
}
node {
input: "offsets"
output: "_val_11"
name: "Shape_8"
op_type: "Shape"
attribute {
name: "start"
i: 0
type: INT
}
doc_string: ""
}
node {
output: "_val_12"
name: "Constant_9"
op_type: "Constant"
attribute {
name: "value"
t {
data_type: 7
raw_data: "\000\000\000\000\000\000\000\000"
}
type: TENSOR
}
doc_string: ""
}
node {
input: "_val_12"
input: "_val_11"
output: "_val_13"
name: "Expand_10"
op_type: "Expand"
doc_string: ""
}
node {
input: "offsets"
output: "_val_14"
name: "Shape_11"
op_type: "Shape"
attribute {
name: "start"
i: 0
type: INT
}
doc_string: ""
}
node {
output: "_val_15"
name: "Constant_12"
op_type: "Constant"
attribute {
name: "value"
t {
data_type: 7
raw_data: "\000\000\000\000\000\000\000\000"
}
type: TENSOR
}
doc_string: ""
}
node {
input: "_val_15"
input: "_val_14"
output: "_val_16"
name: "Expand_13"
op_type: "Expand"
doc_string: ""
}
name: "torch_jit"
input {
name: "input_0"
type {
tensor_type {
elem_type: 10
shape {
dim {
dim_value: 10
}
dim {
dim_value: 5
}
}
}
}
}
input {
name: "input_1"
type {
tensor_type {
elem_type: 7
shape {
dim {
dim_value: 5
}
}
}
}
}
input {
name: "offsets"
type {
tensor_type {
elem_type: 7
shape {
dim {
dim_value: 3
}
}
}
}
}
output {
name: "_val_9"
type {
tensor_type {
elem_type: 10
shape {
dim {
dim_value: 2
}
dim {
dim_value: 5
}
}
}
}
}
output {
name: "_val_10"
type {
tensor_type {
elem_type: 7
shape {
dim {
dim_value: 0
}
}
}
}
}
output {
name: "_val_13"
type {
tensor_type {
elem_type: 7
shape {
dim {
dim_value: 3
}
}
}
}
}
output {
name: "_val_16"
type {
tensor_type {
elem_type: 7
shape {
dim {
dim_value: 3
}
}
}
}
}
value_info {
name: "_val_3"
type {
tensor_type {
elem_type: 7
shape {
dim {
dim_value: 1
}
}
}
}
}
value_info {
name: "_val_4"
type {
tensor_type {
elem_type: 7
shape {
dim {
dim_value: 5
}
}
}
}
}
value_info {
name: "_val_5"
type {
tensor_type {
elem_type: 7
shape {
dim {
dim_value: 1
}
}
}
}
}
value_info {
name: "_val_6"
type {
tensor_type {
elem_type: 7
shape {
}
}
}
}
value_info {
name: "_val_7"
type {
tensor_type {
elem_type: 7
shape {
dim {
dim_value: 5
}
}
}
}
}
value_info {
name: "_val_8"
type {
tensor_type {
elem_type: 10
shape {
dim {
dim_value: 5
}
}
}
}
}
value_info {
name: "_val_11"
type {
tensor_type {
elem_type: 7
shape {
dim {
dim_value: 1
}
}
}
}
}
value_info {
name: "_val_12"
type {
tensor_type {
elem_type: 7
shape {
}
}
}
}
value_info {
name: "_val_14"
type {
tensor_type {
elem_type: 7
shape {
dim {
dim_value: 1
}
}
}
}
}
value_info {
name: "_val_15"
type {
tensor_type {
elem_type: 7
shape {
}
}
}
}
}
opset_import {
domain: "pkg.onnxscript.torch_lib"
version: 1
}
opset_import {
domain: ""
version: 18
}
functions {
name: "_aten_embedding_bag_onnx"
input: "weight"
input: "indices"
input: "offsets"
input: "per_sample_weights"
output: "return_val"
attribute: "mode"
attribute: "include_last_offset"
node {
output: "neg_1"
name: "n0"
op_type: "Constant"
attribute {
name: "value_ints"
ints: -1
type: INTS
}
domain: ""
}
node {
input: "indices"
input: "neg_1"
output: "indices_1d"
name: "n1"
op_type: "Reshape"
domain: ""
}
node {
input: "weight"
input: "indices_1d"
output: "new_weight"
name: "n2"
op_type: "Gather"
domain: ""
}
node {
output: "int64_1"
name: "n3"
op_type: "Constant"
attribute {
name: "value"
t {
data_type: 7
int64_data: 1
name: "int64_1"
}
type: TENSOR
}
domain: ""
}
node {
input: "per_sample_weights"
input: "int64_1"
output: "tmp"
name: "n4"
op_type: "Unsqueeze"
domain: ""
}
node {
input: "new_weight"
input: "tmp"
output: "new_weight_0"
name: "n5"
op_type: "Mul"
domain: ""
}
node {
input: "weight"
output: "tmp_1"
name: "n6"
op_type: "Shape"
attribute {
name: "start"
i: 1
type: INT
}
domain: ""
}
node {
input: "tmp_1"
input: "neg_1"
output: "weight_dim_1"
name: "n7"
op_type: "Reshape"
domain: ""
}
node {
input: "indices_1d"
output: "indices_size"
name: "n8"
op_type: "Shape"
domain: ""
}
node {
input: "offsets"
output: "tmp_2"
name: "n9"
op_type: "Size"
domain: ""
}
node {
input: "tmp_2"
input: "neg_1"
output: "num_bag"
name: "n10"
op_type: "Reshape"
domain: ""
}
node {
output: "include_last_offset"
name: "n11"
op_type: "Constant"
attribute {
name: "value_int"
type: INT
ref_attr_name: "include_last_offset"
}
domain: ""
}
node {
input: "include_last_offset"
output: "include_last_offset_as_bool"
name: "n12"
op_type: "Cast"
attribute {
name: "to"
i: 9
type: INT
}
domain: ""
}
node {
output: "int64_True"
name: "n13"
op_type: "Constant"
attribute {
name: "value"
t {
data_type: 9
int32_data: 1
name: "int64_True"
}
type: TENSOR
}
domain: ""
}
node {
input: "include_last_offset_as_bool"
input: "int64_True"
output: "cond"
name: "n14"
op_type: "Equal"
domain: ""
}
node {
input: "cond"
output: "offsets_8"
output: "num_bag_9"
name: "n15"
op_type: "If"
attribute {
name: "then_branch"
g {
node {
output: "int64_1_3"
name: "n0"
op_type: "Constant"
attribute {
name: "value"
t {
data_type: 7
int64_data: 1
name: "int64_1_3"
}
type: TENSOR
}
domain: ""
}
node {
input: "int64_1_3"
input: "num_bag"
output: "int64_1_3_cast"
name: "n1"
op_type: "CastLike"
domain: ""
}
node {
input: "num_bag"
input: "int64_1_3_cast"
output: "num_bag_4"
name: "n2"
op_type: "Sub"
domain: ""
}
node {
input: "offsets"
output: "offsets_5"
name: "n3"
op_type: "Identity"
domain: ""
}
name: "thenGraph_23"
output {
name: "offsets_5"
type {
}
}
output {
name: "num_bag_4"
type {
}
}
}
type: GRAPH
}
attribute {
name: "else_branch"
g {
node {
input: "offsets"
input: "indices_size"
output: "offsets_6"
name: "n0"
op_type: "Concat"
attribute {
name: "axis"
i: 0
type: INT
}
domain: ""
}
node {
input: "num_bag"
output: "num_bag_7"
name: "n1"
op_type: "Identity"
domain: ""
}
name: "elseGraph_23"
output {
name: "offsets_6"
type {
}
}
output {
name: "num_bag_7"
type {
}
}
}
type: GRAPH
}
domain: ""
}
node {
output: "result"
name: "n16"
op_type: "SequenceEmpty"
domain: ""
}
node {
output: "tmp_10"
name: "n17"
op_type: "Constant"
attribute {
name: "value_int"
i: 0
type: INT
}
domain: ""
}
node {
input: "tmp_10"
input: "neg_1"
output: "index_tensor"
name: "n18"
op_type: "Reshape"
domain: ""
}
node {
input: "index_tensor"
input: "num_bag_9"
output: "cond_11"
name: "n19"
op_type: "Less"
domain: ""
}
node {
output: "true"
name: "n20"
op_type: "Constant"
attribute {
name: "value"
t {
data_type: 9
int32_data: 1
name: "true"
}
type: TENSOR
}
domain: ""
}
node {
input: ""
input: "true"
input: "result"
input: "index_tensor"
output: "result_59"
output: "index_tensor_60"
name: "n21"
op_type: "Loop"
attribute {
name: "body"
g {
node {
output: "int64_1_14"
name: "n0"
op_type: "Constant"
attribute {
name: "value"
t {
data_type: 7
int64_data: 1
name: "int64_1_14"
}
type: TENSOR
}
domain: ""
}
node {
input: "int64_1_14"
input: "index_tensor_13"
output: "int64_1_14_cast"
name: "n1"
op_type: "CastLike"
domain: ""
}
node {
input: "index_tensor_13"
input: "int64_1_14_cast"
output: "tmp_15"
name: "n2"
op_type: "Add"
domain: ""
}
node {
input: "offsets_8"
input: "index_tensor_13"
input: "tmp_15"
output: "start"
name: "n3"
op_type: "Slice"
domain: ""
}
node {
output: "int64_1_16"
name: "n4"
op_type: "Constant"
attribute {
name: "value"
t {
data_type: 7
int64_data: 1
name: "int64_1_16"
}
type: TENSOR
}
domain: ""
}
node {
input: "int64_1_16"
input: "index_tensor_13"
output: "int64_1_16_cast"
name: "n5"
op_type: "CastLike"
domain: ""
}
node {
input: "index_tensor_13"
input: "int64_1_16_cast"
output: "tmp_17"
name: "n6"
op_type: "Add"
domain: ""
}
node {
output: "int64_2"
name: "n7"
op_type: "Constant"
attribute {
name: "value"
t {
data_type: 7
int64_data: 2
name: "int64_2"
}
type: TENSOR
}
domain: ""
}
node {
input: "int64_2"
input: "index_tensor_13"
output: "int64_2_cast"
name: "n8"
op_type: "CastLike"
domain: ""
}
node {
input: "index_tensor_13"
input: "int64_2_cast"
output: "tmp_18"
name: "n9"
op_type: "Add"
domain: ""
}
node {
input: "offsets_8"
input: "tmp_17"
input: "tmp_18"
output: "end"
name: "n10"
op_type: "Slice"
domain: ""
}
node {
input: "start"
input: "end"
output: "cond_19"
name: "n11"
op_type: "Equal"
domain: ""
}
node {
input: "cond_19"
output: "row_result_54"
name: "n12"
op_type: "If"
attribute {
name: "then_branch"
g {
node {
output: "tmp_20"
name: "n0"
op_type: "Constant"
attribute {
name: "value_floats"
floats: 0.0
type: FLOATS
}
domain: ""
}
node {
output: "tmp_21"
name: "n1"
op_type: "Constant"
attribute {
name: "value_ints"
ints: 1
type: INTS
}
domain: ""
}
node {
input: "tmp_21"
input: "weight_dim_1"
output: "tmp_22"
name: "n2"
op_type: "Concat"
attribute {
name: "axis"
i: 0
type: INT
}
domain: ""
}
node {
input: "tmp_20"
input: "tmp_22"
output: "row_result"
name: "n3"
op_type: "Expand"
domain: ""
}
name: "thenGraph_40"
output {
name: "row_result"
type {
}
}
}
type: GRAPH
}
attribute {
name: "else_branch"
g {
node {
output: "mode"
name: "n0"
op_type: "Constant"
attribute {
name: "value_int"
type: INT
ref_attr_name: "mode"
}
domain: ""
}
node {
output: "int64_0"
name: "n1"
op_type: "Constant"
attribute {
name: "value"
t {
data_type: 7
int64_data: 0
name: "int64_0"
}
type: TENSOR
}
domain: ""
}
node {
input: "mode"
input: "int64_0"
output: "cond_23"
name: "n2"
op_type: "Equal"
domain: ""
}
node {
input: "cond_23"
output: "row_result_53"
name: "n3"
op_type: "If"
attribute {
name: "then_branch"
g {
node {
input: "new_weight_0"
input: "start"
input: "end"
output: "weight_rows"
name: "n0"
op_type: "Slice"
domain: ""
}
node {
output: "int64_0_1d"
name: "n1"
op_type: "Constant"
attribute {
name: "value"
t {
dims: 1
data_type: 7
int64_data: 0
name: "int64_0_1d"
}
type: TENSOR
}
domain: ""
}
node {
input: "weight_rows"
input: "int64_0_1d"
output: "row_result_24"
name: "n2"
op_type: "ReduceSum"
domain: ""
}
name: "thenGraph_46"
output {
name: "row_result_24"
type {
}
}
}
type: GRAPH
}
attribute {
name: "else_branch"
g {
node {
output: "mode_25"
name: "n0"
op_type: "Constant"
attribute {
name: "value_int"
type: INT
ref_attr_name: "mode"
}
domain: ""
}
node {
output: "int64_1_26"
name: "n1"
op_type: "Constant"
attribute {
name: "value"
t {
data_type: 7
int64_data: 1
name: "int64_1_26"
}
type: TENSOR
}
domain: ""
}
node {
input: "mode_25"
input: "int64_1_26"
output: "cond_27"
name: "n2"
op_type: "Equal"
domain: ""
}
node {
input: "cond_27"
output: "row_result_52"
name: "n3"
op_type: "If"
attribute {
name: "then_branch"
g {
node {
input: "new_weight_0"
input: "start"
input: "end"
output: "weight_rows_28"
name: "n0"
op_type: "Slice"
domain: ""
}
node {
output: "int64_1_29"
name: "n1"
op_type: "Constant"
attribute {
name: "value"
t {
data_type: 7
int64_data: 1
name: "int64_1_29"
}
type: TENSOR
}
domain: ""
}
node {
input: "int64_1_29"
input: "num_bag_9"
output: "int64_1_29_cast"
name: "n2"
op_type: "CastLike"
domain: ""
}
node {
input: "num_bag_9"
input: "int64_1_29_cast"
output: "tmp_30"
name: "n3"
op_type: "Sub"
domain: ""
}
node {
input: "index_tensor_13"
input: "tmp_30"
output: "cond_31"
name: "n4"
op_type: "Equal"
domain: ""
}
node {
input: "cond_31"
output: "row_result_43"
name: "n5"
op_type: "If"
attribute {
name: "then_branch"
g {
node {
output: "int64_0_1d_32"
name: "n0"
op_type: "Constant"
attribute {
name: "value"
t {
dims: 1
data_type: 7
int64_data: 0
name: "int64_0_1d_32"
}
type: TENSOR
}
domain: ""
}
node {
input: "weight_rows_28"
input: "int64_0_1d_32"
output: "row_result_33"
name: "n1"
op_type: "ReduceSum"
domain: ""
}
node {
input: "indices"
output: "tmp_34"
name: "n2"
op_type: "Shape"
attribute {
name: "end"
i: 1
type: INT
}
attribute {
name: "start"
i: 0
type: INT
}
domain: ""
}
node {
input: "tmp_34"
input: "start"
output: "denominator"
name: "n3"
op_type: "Sub"
domain: ""
}
node {
output: "int64_0_35"
name: "n4"
op_type: "Constant"
attribute {
name: "value"
t {
data_type: 7
int64_data: 0
name: "int64_0_35"
}
type: TENSOR
}
domain: ""
}
node {
input: "int64_0_35"
input: "denominator"
output: "int64_0_35_cast"
name: "n5"
op_type: "CastLike"
domain: ""
}
node {
input: "denominator"
input: "int64_0_35_cast"
output: "cond_36"
name: "n6"
op_type: "Greater"
domain: ""
}
node {
input: "cond_36"
output: "row_result_40"
name: "n7"
op_type: "If"
attribute {
name: "then_branch"
g {
node {
input: "denominator"
input: "new_weight_0"
output: "tmp_37"
name: "n0"
op_type: "CastLike"
domain: ""
}
node {
input: "row_result_33"
input: "tmp_37"
output: "row_result_38"
name: "n1"
op_type: "Div"
domain: ""
}
name: "thenGraph_56"
output {
name: "row_result_38"
type {
}
}
}
type: GRAPH
}
attribute {
name: "else_branch"
g {
node {
input: "row_result_33"
output: "row_result_39"
name: "n0"
op_type: "Identity"
domain: ""
}
name: "elseGraph_56"
output {
name: "row_result_39"
type {
}
}
}
type: GRAPH
}
domain: ""
}
name: "thenGraph_51"
output {
name: "row_result_40"
type {
}
}
}
type: GRAPH
}
attribute {
name: "else_branch"
g {
node {
output: "int64_0_1d_41"
name: "n0"
op_type: "Constant"
attribute {
name: "value"
t {
dims: 1
data_type: 7
int64_data: 0
name: "int64_0_1d_41"
}
type: TENSOR
}
domain: ""
}
node {
input: "weight_rows_28"
input: "int64_0_1d_41"
output: "row_result_42"
name: "n1"
op_type: "ReduceMean"
domain: ""
}
name: "elseGraph_51"
output {
name: "row_result_42"
type {
}
}
}
type: GRAPH
}
domain: ""
}
name: "thenGraph_49"
output {
name: "row_result_43"
type {
}
}
}
type: GRAPH
}
attribute {
name: "else_branch"
g {
node {
output: "int64_1_44"
name: "n0"
op_type: "Constant"
attribute {
name: "value"
t {
data_type: 7
int64_data: 1
name: "int64_1_44"
}
type: TENSOR
}
domain: ""
}
node {
input: "int64_1_44"
input: "num_bag_9"
output: "int64_1_44_cast"
name: "n1"
op_type: "CastLike"
domain: ""
}
node {
input: "num_bag_9"
input: "int64_1_44_cast"
output: "tmp_45"
name: "n2"
op_type: "Sub"
domain: ""
}
node {
input: "index_tensor_13"
input: "tmp_45"
output: "cond_46"
name: "n3"
op_type: "Equal"
domain: ""
}
node {
input: "cond_46"
output: "weight_rows_49"
name: "n4"
op_type: "If"
attribute {
name: "then_branch"
g {
node {
input: "new_weight_0"
input: "start"
input: "indices_size"
output: "weight_rows_47"
name: "n0"
op_type: "Slice"
domain: ""
}
name: "thenGraph_61"
output {
name: "weight_rows_47"
type {
}
}
}
type: GRAPH
}
attribute {
name: "else_branch"
g {
node {
input: "new_weight_0"
input: "start"
input: "end"
output: "weight_rows_48"
name: "n0"
op_type: "Slice"
domain: ""
}
name: "elseGraph_61"
output {
name: "weight_rows_48"
type {
}
}
}
type: GRAPH
}
domain: ""
}
node {
output: "int64_0_1d_50"
name: "n5"
op_type: "Constant"
attribute {
name: "value"
t {
dims: 1
data_type: 7
int64_data: 0
name: "int64_0_1d_50"
}
type: TENSOR
}
domain: ""
}
node {
input: "weight_rows_49"
input: "int64_0_1d_50"
output: "row_result_51"
name: "n6"
op_type: "ReduceMax"
domain: ""
}
name: "elseGraph_49"
output {
name: "row_result_51"
type {
}
}
}
type: GRAPH
}
domain: ""
}
name: "elseGraph_46"
output {
name: "row_result_52"
type {
}
}
}
type: GRAPH
}
domain: ""
}
name: "elseGraph_40"
output {
name: "row_result_53"
type {
}
}
}
type: GRAPH
}
domain: ""
}
node {
input: "result_12"
input: "row_result_54"
output: "result_55"
name: "n13"
op_type: "SequenceInsert"
domain: ""
}
node {
output: "int64_1_56"
name: "n14"
op_type: "Constant"
attribute {
name: "value"
t {
data_type: 7
int64_data: 1
name: "int64_1_56"
}
type: TENSOR
}
domain: ""
}
node {
input: "int64_1_56"
input: "index_tensor_13"
output: "int64_1_56_cast"
name: "n15"
op_type: "CastLike"
domain: ""
}
node {
input: "index_tensor_13"
input: "int64_1_56_cast"
output: "index_tensor_57"
name: "n16"
op_type: "Add"
domain: ""
}
node {
input: "index_tensor_57"
input: "num_bag_9"
output: "cond_58"
name: "n17"
op_type: "Less"
domain: ""
}
node {
input: "cond_58"
output: "cond_out"
name: "n18"
op_type: "Identity"
domain: ""
}
name: "loop_body"
input {
name: "infinite_loop"
type {
tensor_type {
elem_type: 7
shape {
}
}
}
}
input {
name: "cond"
type {
tensor_type {
elem_type: 9
shape {
}
}
}
}
input {
name: "result_12"
type {
}
}
input {
name: "index_tensor_13"
type {
}
}
output {
name: "cond_out"
type {
tensor_type {
elem_type: 9
shape {
}
}
}
}
output {
name: "result_55"
type {
}
}
output {
name: "index_tensor_57"
type {
}
}
}
type: GRAPH
}
domain: ""
}
node {
input: "result_59"
output: "result_61"
name: "n22"
op_type: "ConcatFromSequence"
attribute {
name: "axis"
i: 0
type: INT
}
domain: ""
}
node {
input: "result_61"
input: "weight"
output: "return_val"
name: "n23"
op_type: "CastLike"
domain: ""
}
opset_import {
domain: ""
version: 18
}
domain: "pkg.onnxscript.torch_lib"
}
"""
ort_inputs = {'input_0': array([[ 2.664 , -6.777 , 5.723 , -5.027 , 0.3164],
[ 1.397 , 8.945 , -0.0879, -8.93 , 2.04 ],
[ 0.8613, 5.78 , -1.837 , -7.18 , 1.143 ],
[-5.457 , -7.67 , -8.53 , 1.046 , 5.906 ],
[ 8.08 , -3.488 , -3.85 , 5.484 , 8.5 ],
[ 3.684 , 3.7 , -8.09 , 7.48 , -6.777 ],
[ 1.986 , 0.9756, -4.527 , 4.15 , -8.14 ],
[-7.664 , 7.56 , -0.9756, 7.867 , -1.292 ],
[-6.04 , 6.195 , -1.784 , 1.424 , 6.32 ],
[-0.5977, 5.977 , 8.74 , -7.594 , 5.09 ]], dtype=float16), 'input_1': array([9, 1, 4, 9, 6]), 'offsets': array([0, 2, 3])}
# Set up the inference session
session_options = ort.SessionOptions()
session_options.graph_optimization_level = ort.GraphOptimizationLevel.ORT_DISABLE_ALL
onnx_model = onnx.ModelProto()
google.protobuf.text_format.Parse(onnx_model_text, onnx_model)
# Uncomment this line to save the model to a file for examination
# onnx.save_model(onnx_model, "test_output_match_opinfo__ops_aten_embedding_bag_cpu_float16.onnx")
onnx.checker.check_model(onnx_model)
session = ort.InferenceSession(onnx_model.SerializeToString(), session_options, providers=("CPUExecutionProvider",))
# Run the model
for _ in range(N):
ort_outputs = session.run(None, ort_inputs)
[ONNXRuntimeError] : 1 : FAIL : Node (_0x7b6efe0_n21) Op (Loop) [TypeInferenceError] Graph attribute inferencing failed: Node (_0x7b6efe0_n12) Op (If) [TypeInferenceError] Mismatched tensor element type: source=10 target=1
File "/home/justinchu/dev/onnx-script/onnxscript/tests/function_libs/torch_lib/ops_test_common.py", line 533, in _capture_graph_and_evaluate_torch_script_evaluator
return _safe_ort_session_run(onnx_model.SerializeToString(), ort_inputs)
File "/home/justinchu/dev/onnx-script/onnxscript/tests/function_libs/torch_lib/ops_test_common.py", line 349, in _safe_ort_session_run
raise return_dict["error"]
OS: Linux-5.15.0-1042-azure-x86_64-with-glibc2.35
Python version: 3.10.9 (main, Jan 11 2023, 15:21:40) [GCC 11.2.0]
onnx==1.15.0.dev20230731
onnxruntime==1.15.1
numpy==1.25.1
torch==2.1.0.dev20230622+cpu
Describe the issue
op.SequenceEmpty(dtype=xxx) cannot be set to float16, it will ignore the dtype and result as float32.
To reproduce
Error
Urgency
No response
Platform
Windows
OS Version
11
ONNX Runtime Installation
Released Package
ONNX Runtime Version or Commit ID
1.14
ONNX Runtime API
Python
Architecture
X64
Execution Provider
Default CPU
Execution Provider Library Version
No response