Adding matmul_integer_to_float16 onnx models (#16978)

onnxruntime/test/testdata/matmul_integer_to_float16_int8.onnx

@@ -0,0 +1,51 @@
+	:Ì
+U
+
A
+
B
+a_zero_point
+b_zero_pointmatmul_output_int32MatMulInteger"MatMulInteger
+.
+a_scale
+b_scale
+multiplier	mul_right"Mul
+A
+matmul_output_int32matmul_output_floatcast"Cast*	
+to
+ 

+5
+matmul_output_float
+
+multiplier
Y
+mul_bottom"MulDynamicQuantizeMatMul_fusionZ
+
A
+
+
+
M
+
KZ
+
B
+
+
+
K
+
NZ
+a_scale
+
+
+
+
Z
+b_scale
+	
+
+
CZ
+a_zero_point
+
+
+
Z
+b_zero_point
+	
+
Cb
+
Y
+
+
+
+
M
+
NB

onnxruntime/test/testdata/matmul_integer_to_float16_int8_bias.onnx

@@ -0,0 +1,49 @@
+	:Ä
+9
+
A
+
Bmatmul_output_int32MatMulInteger"MatMulInteger
+.
+a_scale
+b_scale
+multiplier	mul_right"Mul
+A
+matmul_output_int32matmul_output_floatcast"Cast*	
+to
+ 

+E
+matmul_output_float
+
+multipliermul_bottom_output
+mul_bottom"Mul
+&
+mul_bottom_output
+bias
Yadd"AddDynamicQuantizeMatMul_fusionZ
+
A
+
+
+
M
+
KZ
+
B
+
+
+
K
+
NZ
+a_scale
+
+
+
+
Z
+b_scale
+	
+
+
CZ
+bias
+	
+
+
Nb
+
Y
+
+
+
+
M
+
NB

onnxruntime/test/testdata/matmul_integer_to_float16_int8_int8.onnx

@@ -0,0 +1,51 @@
+	:Ì
+U
+
A
+
B
+a_zero_point
+b_zero_pointmatmul_output_int32MatMulInteger"MatMulInteger
+.
+a_scale
+b_scale
+multiplier	mul_right"Mul
+A
+matmul_output_int32matmul_output_floatcast"Cast*	
+to
+ 

+5
+matmul_output_float
+
+multiplier
Y
+mul_bottom"MulDynamicQuantizeMatMul_fusionZ
+
A
+
+
+
M
+
KZ
+
B
+
+
+
K
+
NZ
+a_scale
+
+
+
+
Z
+b_scale
+	
+
+
CZ
+a_zero_point
+
+
+
Z
+b_zero_point
+	
+
Cb
+
Y
+
+
+
+
M
+
NB

onnxruntime/test/testdata/matmul_integer_to_float16_int8_int8_bias.onnx

@@ -0,0 +1,49 @@
+	:Ä
+9
+
A
+
Bmatmul_output_int32MatMulInteger"MatMulInteger
+.
+a_scale
+b_scale
+multiplier	mul_right"Mul
+A
+matmul_output_int32matmul_output_floatcast"Cast*	
+to
+ 

+E
+matmul_output_float
+
+multipliermul_bottom_output
+mul_bottom"Mul
+&
+mul_bottom_output
+bias
Yadd"AddDynamicQuantizeMatMul_fusionZ
+
A
+
+
+
M
+
KZ
+
B
+
+
+
K
+
NZ
+a_scale
+
+
+
+
Z
+b_scale
+	
+
+
CZ
+bias
+	
+
+
Nb
+
Y
+
+
+
+
M
+
NB

onnxruntime/test/testdata/transform/fusion/matmul_integer_to_float16.onnx

@@ -0,0 +1,90 @@
+	:Ö
+Q
+inputa_quantizeda_scalea_zpDynamicQuantizeLinear"DynamicQuantizeLinear
+a
+a_quantized
+b_quantized_1
+a_zp
+b_zp_1matmul_output_int32_1MatMulInteger_1"MatMulInteger
+4
+a_scale
+	b_scale_1multiplier_1mul_right_1"Mul
+G
+matmul_output_int32_1matmul_output_float_1cast_1"Cast*	
+to
 

+F
+matmul_output_float_1
+multiplier_1mul_output_1mul_bottom_1"Mul
+1
+mul_output_1
+bias_1output_1
+bias_add_1"Add
+a
+a_quantized
+b_quantized_2
+a_zp
+b_zp_2matmul_output_int32_2MatMulInteger_2"MatMulInteger
+4
+a_scale
+	b_scale_2multiplier_2mul_right_2"Mul
+G
+matmul_output_int32_2matmul_output_float_2cast_2"Cast*	
+to
 

+F
+matmul_output_float_2
+multiplier_2mul_output_2mul_bottom_2"Mul
+1
+mul_output_2
+bias_2output_2
+bias_add_2"Add
+a
+a_quantized
+b_quantized_3
+a_zp
+b_zp_3matmul_output_int32_3MatMulInteger_3"MatMulInteger
+4
+a_scale
+	b_scale_3multiplier_3mul_right_3"Mul
+G
+matmul_output_int32_3matmul_output_float_3cast_3"Cast*	
+to
 

+B
+matmul_output_float_3
+multiplier_3output_3mul_bottom_3"MulMatMulIntegerToFloat_fusion**Bb_quantized_1**€
Bb_zp_1*
+*³~B	b_scale_1**Bb_quantized_3**€
Bb_zp_3*
+*³~B	b_scale_3*
+*	€€
€ˆ
€Š
Bbias_1**
+*€x€€
€„
€ˆ
€Š
€Œ
€Ž
€
€‘
Bbias_2Z
+input
+
+
+
+Z
+b_quantized_2
+
+
+Z
+b_zp_2
+
+
+
Z
+	b_scale_2
+
+
+
+
b
+output_1
+
+
+
+b
+output_2
+
+
+
+b
+output_3
+
+
+
+B