File tree Expand file tree Collapse file tree 4 files changed +30
-0
lines changed Expand file tree Collapse file tree 4 files changed +30
-0
lines changed Original file line number Diff line number Diff line change @@ -121,6 +121,14 @@ xnn_datatype getDataType(const DataType& data_type) {
121
121
return xnn_datatype::xnn_datatype_qdint8;
122
122
case DataType::xnn_datatype_qbint4:
123
123
return xnn_datatype::xnn_datatype_qbint4;
124
+ case DataType::xnn_datatype_qpint8:
125
+ return xnn_datatype::xnn_datatype_qpint8;
126
+ case DataType::xnn_datatype_int32:
127
+ return xnn_datatype::xnn_datatype_int32;
128
+ case DataType::xnn_datatype_pfp32:
129
+ return xnn_datatype::xnn_datatype_pfp32;
130
+ case DataType::xnn_datatype_bf16:
131
+ return xnn_datatype::xnn_datatype_bf16;
124
132
default :
125
133
return xnn_datatype::xnn_datatype_invalid;
126
134
}
Original file line number Diff line number Diff line change @@ -29,6 +29,15 @@ enum XNNDatatype : short {
29
29
xnn_datatype_qdint8 = 9,
30
30
/// Quantized 4-bit signed integer with shared blockwise quantization parameters.
31
31
xnn_datatype_qbint4 = 10,
32
+ /// Dynamically quantized 8-bit signed integers packed with their per-row
33
+ /// quantization parameters.
34
+ xnn_datatype_qpint8 = 11,
35
+ /// 32-bit signed integers.
36
+ xnn_datatype_int32 = 12,
37
+ /// IEEE754 single-precision packed floating-point.
38
+ xnn_datatype_pfp32 = 13,
39
+ /// BFloat16, i.e. the upper 16 bits of a float32.
40
+ xnn_datatype_bf16 = 14,
32
41
}
33
42
34
43
// type of quantization
Original file line number Diff line number Diff line change @@ -29,6 +29,15 @@ enum XNNDatatype : short {
29
29
xnn_datatype_qdint8 = 9,
30
30
/// Quantized 4-bit signed integer with shared blockwise quantization parameters.
31
31
xnn_datatype_qbint4 = 10,
32
+ /// Dynamically quantized 8-bit signed integers packed with their per-row
33
+ /// quantization parameters.
34
+ xnn_datatype_qpint8 = 11,
35
+ /// 32-bit signed integers.
36
+ xnn_datatype_int32 = 12,
37
+ /// IEEE754 single-precision packed floating-point.
38
+ xnn_datatype_pfp32 = 13,
39
+ /// BFloat16, i.e. the upper 16 bits of a float32.
40
+ xnn_datatype_bf16 = 14,
32
41
}
33
42
34
43
// type of quantization
Original file line number Diff line number Diff line change @@ -430,6 +430,10 @@ class XNNDatatype(IntEnum):
430
430
xnn_datatype_qcint4 = 8
431
431
xnn_datatype_qdint8 = 9
432
432
xnn_datatype_qbint4 = 10
433
+ xnn_datatype_qpint8 = 11
434
+ xnn_datatype_int32 = 12
435
+ xnn_datatype_pfp32 = 13
436
+ xnn_datatype_bf16 = 14
433
437
434
438
435
439
@dataclass
You can’t perform that action at this time.
0 commit comments