offline-telespeech-ctc-model.cc
4.3 KB
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
// sherpa-onnx/csrc/offline-telespeech-ctc-model.cc
//
// Copyright (c) 2023-2024 Xiaomi Corporation
#include "sherpa-onnx/csrc/offline-telespeech-ctc-model.h"
#include "sherpa-onnx/csrc/macros.h"
#include "sherpa-onnx/csrc/onnx-utils.h"
#include "sherpa-onnx/csrc/session.h"
#include "sherpa-onnx/csrc/text-utils.h"
#include "sherpa-onnx/csrc/transpose.h"
namespace sherpa_onnx {
class OfflineTeleSpeechCtcModel::Impl {
public:
explicit Impl(const OfflineModelConfig &config)
: config_(config),
env_(ORT_LOGGING_LEVEL_ERROR),
sess_opts_(GetSessionOptions(config)),
allocator_{} {
auto buf = ReadFile(config_.telespeech_ctc);
Init(buf.data(), buf.size());
}
#if __ANDROID_API__ >= 9
Impl(AAssetManager *mgr, const OfflineModelConfig &config)
: config_(config),
env_(ORT_LOGGING_LEVEL_ERROR),
sess_opts_(GetSessionOptions(config)),
allocator_{} {
auto buf = ReadFile(mgr, config_.telespeech_ctc);
Init(buf.data(), buf.size());
}
#endif
std::vector<Ort::Value> Forward(Ort::Value features,
Ort::Value /*features_length*/) {
std::vector<int64_t> shape =
features.GetTensorTypeAndShapeInfo().GetShape();
if (static_cast<int32_t>(shape[0]) != 1) {
SHERPA_ONNX_LOGE("This model supports only batch size 1. Given %d",
static_cast<int32_t>(shape[0]));
}
auto out = sess_->Run({}, input_names_ptr_.data(), &features, 1,
output_names_ptr_.data(), output_names_ptr_.size());
std::vector<int64_t> logits_shape = {1};
Ort::Value logits_length = Ort::Value::CreateTensor<int64_t>(
allocator_, logits_shape.data(), logits_shape.size());
int64_t *dst = logits_length.GetTensorMutableData<int64_t>();
dst[0] = out[0].GetTensorTypeAndShapeInfo().GetShape()[0];
// (T, B, C) -> (B, T, C)
Ort::Value logits = Transpose01(allocator_, &out[0]);
std::vector<Ort::Value> ans;
ans.reserve(2);
ans.push_back(std::move(logits));
ans.push_back(std::move(logits_length));
return ans;
}
int32_t VocabSize() const { return vocab_size_; }
int32_t SubsamplingFactor() const { return subsampling_factor_; }
OrtAllocator *Allocator() { return allocator_; }
private:
void Init(void *model_data, size_t model_data_length) {
sess_ = std::make_unique<Ort::Session>(env_, model_data, model_data_length,
sess_opts_);
GetInputNames(sess_.get(), &input_names_, &input_names_ptr_);
GetOutputNames(sess_.get(), &output_names_, &output_names_ptr_);
// get meta data
Ort::ModelMetadata meta_data = sess_->GetModelMetadata();
if (config_.debug) {
std::ostringstream os;
PrintModelMetadata(os, meta_data);
SHERPA_ONNX_LOGE("%s\n", os.str().c_str());
}
{
auto shape =
sess_->GetOutputTypeInfo(0).GetTensorTypeAndShapeInfo().GetShape();
vocab_size_ = shape[2];
}
}
private:
OfflineModelConfig config_;
Ort::Env env_;
Ort::SessionOptions sess_opts_;
Ort::AllocatorWithDefaultOptions allocator_;
std::unique_ptr<Ort::Session> sess_;
std::vector<std::string> input_names_;
std::vector<const char *> input_names_ptr_;
std::vector<std::string> output_names_;
std::vector<const char *> output_names_ptr_;
int32_t vocab_size_ = 0;
int32_t subsampling_factor_ = 4;
};
OfflineTeleSpeechCtcModel::OfflineTeleSpeechCtcModel(
const OfflineModelConfig &config)
: impl_(std::make_unique<Impl>(config)) {}
#if __ANDROID_API__ >= 9
OfflineTeleSpeechCtcModel::OfflineTeleSpeechCtcModel(
AAssetManager *mgr, const OfflineModelConfig &config)
: impl_(std::make_unique<Impl>(mgr, config)) {}
#endif
OfflineTeleSpeechCtcModel::~OfflineTeleSpeechCtcModel() = default;
std::vector<Ort::Value> OfflineTeleSpeechCtcModel::Forward(
Ort::Value features, Ort::Value features_length) {
return impl_->Forward(std::move(features), std::move(features_length));
}
int32_t OfflineTeleSpeechCtcModel::VocabSize() const {
return impl_->VocabSize();
}
int32_t OfflineTeleSpeechCtcModel::SubsamplingFactor() const {
return impl_->SubsamplingFactor();
}
OrtAllocator *OfflineTeleSpeechCtcModel::Allocator() const {
return impl_->Allocator();
}
} // namespace sherpa_onnx