File size: 15,703 Bytes
dc2106c
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
/*

 * SPDX-License-Identifier: Apache-2.0

 */

#include "onnx/defs/function.h"
#include "onnx/defs/schema.h"

namespace ONNX_NAMESPACE {

static const char* QuantizeLinear_ver19_doc = R"DOC(

The linear quantization operator. It consumes a high precision tensor, a scale, and a zero point to compute the low precision / quantized tensor.

The scale factor and zero point must have same shape, and can be either a scalar for per-tensor / per layer quantization, or a 1-D tensor for per-axis quantization.

The quantization formula is `y = saturate ((x / y_scale) + y_zero_point)`.

For saturation, it saturates to [0, 255] if it's uint8, or [-128, 127] if it's int8.

For (x / y_scale), it's rounding to the nearest even. Refer to https://en.wikipedia.org/wiki/Rounding for details.

'y_zero_point' and 'y' must have same type.

'y_zero_point' is usually not used for quantization to float8e4m3fn, float8e4m3fnuz, float8e5m2, float8e5m2fnuz,

but the quantization formula remains the same for consistency and

the type of the attribute 'y_zero_point' still determines the quantization type.

)DOC";

ONNX_OPERATOR_SET_SCHEMA(
    QuantizeLinear,
    19,
    OpSchema()
        .Input(0, "x", "N-D full precision Input tensor to be quantized.", "T1")
        .Input(
            1,
            "y_scale",
            "Scale for doing quantization to get 'y'. It can be a scalar, which means per-tensor/layer quantization, "
            "or a 1-D Tensor for per-axis quantization.",
            "T1")
        .Input(
            2,
            "y_zero_point",
            "Zero point for doing quantization to get 'y'. Shape must match y_scale. "
            "Default is uint8 with zero point of 0 if it's not specified.",
            "T2",
            OpSchema::Optional)
        .Output(0, "y", "N-D quantized output tensor. It has same shape as input 'x'.", "T2")
        .Attr(
            "axis",
            "(Optional) The axis of the quantization dimension of the input tensor. Ignored for per-tensor quantization. Negative value means counting dimensions from the back. Accepted range is [-r, r-1] where r = rank(input).",
            AttributeProto::INT,
            static_cast<int64_t>(1))
        .Attr(
            "saturate",
            "The parameter defines how the conversion behaves if an input value is out of "
            "range of the destination type. It only applies for float 8 quantization "
            "(float8e4m3fn, float8e4m3fnuz, float8e5m2, float8e5m2fnuz). It is true by default. "
            "All cases are fully described in two tables inserted in the operator description.",
            AttributeProto::INT,
            static_cast<int64_t>(1))
        .TypeConstraint(
            "T1",
            {"tensor(float)", "tensor(float16)", "tensor(bfloat16)", "tensor(int32)"},
            "Constrain 'x' to float, float16, bfloat16 or int32 tensor.")
        .TypeConstraint(
            "T2",
            {"tensor(int8)",
             "tensor(uint8)",
             "tensor(float8e4m3fn)",
             "tensor(float8e4m3fnuz)",
             "tensor(float8e5m2)",
             "tensor(float8e5m2fnuz)"},
            "Constrain 'y_zero_point' and 'y' to 8-bit integer/float tensor.")
        .SetDoc(QuantizeLinear_ver19_doc)
        .TypeAndShapeInferenceFunction([](ONNX_NAMESPACE::InferenceContext& ctx) {
          if (ctx.hasInput(2)) {
            propagateElemTypeFromInputToOutput(ctx, 2, 0);
          } else {
            updateOutputElemType(ctx, 0, TensorProto::UINT8);
          }
          if (!hasInputShape(ctx, 0)) {
            return;
          }

          auto& input_shape = getInputShape(ctx, 0);
          updateOutputShape(ctx, 0, input_shape);
        }));

static const char* DequantizeLinear_ver19_doc = R"DOC(

The linear dequantization operator. It consumes a quantized tensor, a scale, and a zero point to compute the full precision tensor.

The dequantization formula is `y = (x - x_zero_point) * x_scale`. `x_scale` and `x_zero_point` must have same shape, and can be either a scalar

for per-tensor / per layer quantization, or a 1-D tensor for per-axis quantization.

`x_zero_point` and `x` must have same type. `x` and `y` must have same shape. In the case of dequantizing int32,

there's no zero point (zero point is supposed to be 0).

`zero-point` is usually not used in the case of float8e4m3fn, float8e4m3fnuz, float8e5m2, float8e5m2fnuz quantization,

but the dequantization formula remains the same for consistency and 'x_scale' still determines the output type.

)DOC";

ONNX_OPERATOR_SET_SCHEMA(
    DequantizeLinear,
    19,
    OpSchema()
        .Input(0, "x", "N-D quantized input tensor to be de-quantized.", "T1")
        .Input(
            1,
            "x_scale",
            "Scale for input 'x'. It can be a scalar, which means a per-tensor/layer dequantization, "
            "or a 1-D tensor for per-axis dequantization.",
            "T2")
        .Input(
            2,
            "x_zero_point",
            "Zero point for input 'x'. Shape must match x_scale. "
            "It's optional. Zero point is 0 when it's not specified.",
            "T1",
            OpSchema::Optional)
        .Output(0, "y", "N-D full precision output tensor. It has same shape as input 'x'.", "T2")
        .Attr(
            "axis",
            "(Optional) The axis of the dequantizing dimension of the input tensor. Ignored for per-tensor quantization. Negative value means counting dimensions from the back. Accepted range is [-r, r-1] where r = rank(input).",
            AttributeProto::INT,
            static_cast<int64_t>(1))
        .TypeConstraint(
            "T1",
            {"tensor(int8)",
             "tensor(uint8)",
             "tensor(int32)",
             "tensor(float8e4m3fn)",
             "tensor(float8e4m3fnuz)",
             "tensor(float8e5m2)",
             "tensor(float8e5m2fnuz)"},
            "Constrain 'x_zero_point' and 'x' to 8-bit integer or float, or /32-bit integer tensor.")
        .TypeConstraint(
            "T2",
            {"tensor(float)", "tensor(float16)", "tensor(bfloat16)"},
            "'x_scale' determines the output type.")
        .SetDoc(DequantizeLinear_ver19_doc)
        .TypeAndShapeInferenceFunction([](ONNX_NAMESPACE::InferenceContext& ctx) {
          propagateElemTypeFromInputToOutput(ctx, 1, 0);
          if (!hasInputShape(ctx, 0)) {
            return;
          }
          auto& input_shape = getInputShape(ctx, 0);
          updateOutputShape(ctx, 0, input_shape);
        }));

static const char* QuantizeLinear_ver13_doc = R"DOC(

The linear quantization operator. It consumes a high precision tensor, a scale, and a zero point to compute the low precision / quantized tensor.

The scale factor and zero point must have same shape, and can be either a scalar for per-tensor / per layer quantization, or a 1-D tensor for per-axis quantization.

The quantization formula is y = saturate ((x / y_scale) + y_zero_point).

For saturation, it saturates to [0, 255] if it's uint8, or [-128, 127] if it's int8.

For (x / y_scale), it's rounding to the nearest even. Refer to https://en.wikipedia.org/wiki/Rounding for details. 'y_zero_point' and 'y' must have same type.

)DOC";

ONNX_OPERATOR_SET_SCHEMA(
    QuantizeLinear,
    13,
    OpSchema()
        .Input(0, "x", "N-D full precision Input tensor to be quantized.", "T1")
        .Input(
            1,
            "y_scale",
            "Scale for doing quantization to get 'y'. It can be a scalar, which means per-tensor/layer quantization, "
            "or a 1-D Tensor for per-axis quantization.",
            "tensor(float)")
        .Input(
            2,
            "y_zero_point",
            "Zero point for doing quantization to get 'y'. Shape must match y_scale. "
            "Default is uint8 with zero point of 0 if it's not specified.",
            "T2",
            OpSchema::Optional)
        .Output(0, "y", "N-D quantized output tensor. It has same shape as input 'x'.", "T2")
        .Attr(
            "axis",
            "(Optional) The axis of the quantization dimension of the input tensor. Ignored for per-tensor quantization. Negative value means counting dimensions from the back. Accepted range is [-r, r-1] where r = rank(input).",
            AttributeProto::INT,
            static_cast<int64_t>(1))
        .TypeConstraint("T1", {"tensor(float)", "tensor(int32)"}, "Constrain 'x' to float or int32 tensor.")
        .TypeConstraint(
            "T2",
            {"tensor(int8)", "tensor(uint8)"},
            "Constrain 'y_zero_point' and 'y' to 8-bit integer tensor.")
        .SetDoc(QuantizeLinear_ver13_doc)
        .TypeAndShapeInferenceFunction([](ONNX_NAMESPACE::InferenceContext& ctx) {
          if (ctx.hasInput(2)) {
            propagateElemTypeFromInputToOutput(ctx, 2, 0);
          } else {
            updateOutputElemType(ctx, 0, TensorProto::UINT8);
          }
          if (!hasInputShape(ctx, 0)) {
            return;
          }
          auto& input_shape = getInputShape(ctx, 0);
          updateOutputShape(ctx, 0, input_shape);
        }));

static const char* DequantizeLinear_ver13_doc = R"DOC(

The linear dequantization operator. It consumes a quantized tensor, a scale, and a zero point to compute the full precision tensor.

The dequantization formula is `y = (x - x_zero_point) * x_scale`. `x_scale` and `x_zero_point` must have same shape, and can be either a scalar

for per-tensor / per layer quantization, or a 1-D tensor for per-axis quantization.

`x_zero_point` and `x` must have same type. `x` and `y` must have same shape. In the case of dequantizing int32,

there's no zero point (zero point is supposed to be 0).

)DOC";

ONNX_OPERATOR_SET_SCHEMA(
    DequantizeLinear,
    13,
    OpSchema()
        .Input(0, "x", "N-D quantized input tensor to be de-quantized.", "T")
        .Input(
            1,
            "x_scale",
            "Scale for input 'x'. It can be a scalar, which means a per-tensor/layer dequantization, "
            "or a 1-D tensor for per-axis dequantization.",
            "tensor(float)")
        .Input(
            2,
            "x_zero_point",
            "Zero point for input 'x'. Shape must match x_scale. "
            "It's optional. Zero point is 0 when it's not specified.",
            "T",
            OpSchema::Optional)
        .Output(0, "y", "N-D full precision output tensor. It has same shape as input 'x'.", "tensor(float)")
        .Attr(
            "axis",
            "(Optional) The axis of the dequantizing dimension of the input tensor. Ignored for per-tensor quantization. Negative value means counting dimensions from the back. Accepted range is [-r, r-1] where r = rank(input).",
            AttributeProto::INT,
            static_cast<int64_t>(1))
        .TypeConstraint(
            "T",
            {"tensor(int8)", "tensor(uint8)", "tensor(int32)"},
            "Constrain 'x_zero_point' and 'x' to 8-bit/32-bit integer tensor.")
        .SetDoc(DequantizeLinear_ver13_doc)
        .TypeAndShapeInferenceFunction([](ONNX_NAMESPACE::InferenceContext& ctx) {
          auto y_type = ctx.getOutputType(0);
          // only float is supported
          y_type->mutable_tensor_type()->set_elem_type(ONNX_NAMESPACE::TensorProto::FLOAT);

          if (!hasInputShape(ctx, 0))
            return;

          auto& input_shape = getInputShape(ctx, 0);
          updateOutputShape(ctx, 0, input_shape);
        }));

static const char* QuantizeLinear_ver10_doc = R"DOC(

The linear per-tensor/layer quantization operator. It consumes a high precision tensor, a scale, a zero point to compute the low precision / quantized tensor.

The quantization formula is y = saturate ((x / y_scale) + y_zero_point). For saturation, it saturates to [0, 255] if it's uint8, or [-128, 127] if it's int8.

For (x / y_scale), it's rounding to the nearest even. Refer to https://en.wikipedia.org/wiki/Rounding for details. 'y_zero_point' and 'y' must have same type.

)DOC";

ONNX_OPERATOR_SET_SCHEMA(
    QuantizeLinear,
    10,
    OpSchema()
        .Input(0, "x", "N-D full precision Input tensor to be quantized.", "T1")
        .Input(
            1,
            "y_scale",
            "Scale for doing quantization to get 'y'. It's a scalar, which means a per-tensor/layer quantization.",
            "tensor(float)")
        .Input(
            2,
            "y_zero_point",
            "Zero point for doing quantization to get 'y'. It's a scalar, which means a per-tensor/layer quantization. "
            "Default value is uint8 typed 0 if it's not specified.",
            "T2",
            OpSchema::Optional)
        .Output(0, "y", "N-D quantized output tensor. It has same shape as input 'x'.", "T2")
        .TypeConstraint("T1", {"tensor(float)", "tensor(int32)"}, "Constrain 'x' to float or int32 tensor.")
        .TypeConstraint(
            "T2",
            {"tensor(int8)", "tensor(uint8)"},
            "Constrain 'y_zero_point' and 'y' to 8-bit integer tensor.")
        .SetDoc(QuantizeLinear_ver10_doc)
        .TypeAndShapeInferenceFunction([](ONNX_NAMESPACE::InferenceContext& ctx) {
          if (ctx.hasInput(2)) {
            propagateElemTypeFromInputToOutput(ctx, 2, 0);
          } else {
            updateOutputElemType(ctx, 0, TensorProto::UINT8);
          }
          if (!hasInputShape(ctx, 0)) {
            return;
          }

          auto& input_shape = getInputShape(ctx, 0);
          updateOutputShape(ctx, 0, input_shape);
        }));

static const char* DequantizeLinear_ver10_doc = R"DOC(

The linear dequantization operator. It consumes a quantized tensor, a scale, a zero point to compute the full precision tensor.

The dequantization formula is y = (x - x_zero_point) * x_scale. 'x_scale' and 'x_zero_point' are both scalars.

'x_zero_point' and 'x' must have same type. 'x' and 'y' must have same shape. In the case of dequantizing int32,

there's no zero point (zero point is supposed to be 0).

)DOC";

ONNX_OPERATOR_SET_SCHEMA(
    DequantizeLinear,
    10,
    OpSchema()
        .Input(0, "x", "N-D quantized input tensor to be de-quantized.", "T")
        .Input(
            1,
            "x_scale",
            "Scale for input 'x'. It's a scalar, which means a per-tensor/layer quantization.",
            "tensor(float)")
        .Input(
            2,
            "x_zero_point",
            "Zero point for input 'x'. It's a scalar, which means a per-tensor/layer quantization. "
            "It's optional. 0 is the default value when it's not specified.",
            "T",
            OpSchema::Optional)
        .Output(0, "y", "N-D full precision output tensor. It has same shape as input 'x'.", "tensor(float)")
        .TypeConstraint(
            "T",
            {"tensor(int8)", "tensor(uint8)", "tensor(int32)"},
            "Constrain 'x_zero_point' and 'x' to 8-bit/32-bit integer tensor.")
        .SetDoc(DequantizeLinear_ver10_doc)
        .TypeAndShapeInferenceFunction([](ONNX_NAMESPACE::InferenceContext& ctx) {
          auto y_type = ctx.getOutputType(0);
          // only float is supported
          y_type->mutable_tensor_type()->set_elem_type(ONNX_NAMESPACE::TensorProto::FLOAT);

          if (!hasInputShape(ctx, 0))
            return;

          auto& input_shape = getInputShape(ctx, 0);
          updateOutputShape(ctx, 0, input_shape);
        }));

} // namespace ONNX_NAMESPACE