xref: /aosp_15_r20/external/tensorflow/tensorflow/compiler/mlir/lite/quantization/lite/quantize_model.h (revision b6fb3261f9314811a0f4371741dbb8839866f948)
1 /* Copyright 2019 The TensorFlow Authors. All Rights Reserved.
2 
3 Licensed under the Apache License, Version 2.0 (the "License");
4 you may not use this file except in compliance with the License.
5 You may obtain a copy of the License at
6 
7     http://www.apache.org/licenses/LICENSE-2.0
8 
9 Unless required by applicable law or agreed to in writing, software
10 distributed under the License is distributed on an "AS IS" BASIS,
11 WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12 See the License for the specific language governing permissions and
13 limitations under the License.
14 ==============================================================================*/
15 #ifndef TENSORFLOW_COMPILER_MLIR_LITE_QUANTIZATION_LITE_QUANTIZE_MODEL_H_
16 #define TENSORFLOW_COMPILER_MLIR_LITE_QUANTIZATION_LITE_QUANTIZE_MODEL_H_
17 
18 #include <memory>
19 #include <string>
20 #include <unordered_set>
21 
22 #include "absl/container/flat_hash_set.h"
23 #include "tensorflow/lite/core/api/error_reporter.h"
24 #include "tensorflow/lite/model.h"
25 #include "tensorflow/lite/schema/schema_generated.h"
26 
27 namespace mlir {
28 namespace lite {
29 
30 using StringSet = absl::flat_hash_set<std::string>;
31 
32 // Quantize the `input_model` and write the result to a flatbuffer `builder`.
33 // The `input_type`, `output_type` and `inference_type` can be
34 // float32/qint8/int8/int16.
35 // Return partially quantized model if `fully_quantize` is false.
36 // When `verify_numeric` is true, the model will have it's original float ops
37 // and NumericVerify ops to compare output values from the quantized and float
38 // ops. When `legacy_float_scale` is true, the quantizer will use float scale
39 // instead of double, and call TOCO's quantization routines to maintain
40 // bit-exactness of the values with the TOCO quantizer.
41 TfLiteStatus QuantizeModel(
42     const tflite::ModelT& input_model, const tflite::TensorType& input_type,
43     const tflite::TensorType& output_type,
44     const tflite::TensorType& inference_type,
45     const std::unordered_set<std::string>& operator_names,
46     bool disable_per_channel, bool fully_quantize,
47     flatbuffers::FlatBufferBuilder* builder,
48     tflite::ErrorReporter* error_reporter, bool verify_numeric = false,
49     bool whole_model_verify = false, bool legacy_float_scale = true,
50     const StringSet& denylisted_ops = {},
51     const StringSet& denylisted_nodes = {});
52 }  // namespace lite
53 }  // namespace mlir
54 
55 #endif  // TENSORFLOW_COMPILER_MLIR_LITE_QUANTIZATION_LITE_QUANTIZE_MODEL_H_
56