online-wenet-ctc-model.cc
8.0 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
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
// sherpa-onnx/csrc/online-wenet-ctc-model.cc
//
// Copyright (c) 2023 Xiaomi Corporation
#include "sherpa-onnx/csrc/online-wenet-ctc-model.h"
#include <algorithm>
#include <cmath>
#include <string>
#if __ANDROID_API__ >= 9
#include "android/asset_manager.h"
#include "android/asset_manager_jni.h"
#endif
#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"
namespace sherpa_onnx {
class OnlineWenetCtcModel::Impl {
public:
explicit Impl(const OnlineModelConfig &config)
: config_(config),
env_(ORT_LOGGING_LEVEL_ERROR),
sess_opts_(GetSessionOptions(config)),
allocator_{} {
{
auto buf = ReadFile(config.wenet_ctc.model);
Init(buf.data(), buf.size());
}
}
#if __ANDROID_API__ >= 9
Impl(AAssetManager *mgr, const OnlineModelConfig &config)
: config_(config),
env_(ORT_LOGGING_LEVEL_WARNING),
sess_opts_(GetSessionOptions(config)),
allocator_{} {
{
auto buf = ReadFile(mgr, config.wenet_ctc.model);
Init(buf.data(), buf.size());
}
}
#endif
std::vector<Ort::Value> Forward(Ort::Value x,
std::vector<Ort::Value> states) {
Ort::Value &attn_cache = states[0];
Ort::Value &conv_cache = states[1];
Ort::Value &offset = states[2];
int32_t chunk_size = config_.wenet_ctc.chunk_size;
int32_t left_chunks = config_.wenet_ctc.num_left_chunks;
// build attn_mask
std::array<int64_t, 3> attn_mask_shape{1, 1,
required_cache_size_ + chunk_size};
Ort::Value attn_mask = Ort::Value::CreateTensor<bool>(
allocator_, attn_mask_shape.data(), attn_mask_shape.size());
bool *p = attn_mask.GetTensorMutableData<bool>();
int32_t chunk_idx =
offset.GetTensorData<int64_t>()[0] / chunk_size - left_chunks;
if (chunk_idx < left_chunks) {
std::fill(p, p + required_cache_size_ - chunk_idx * chunk_size, 0);
std::fill(p + required_cache_size_ - chunk_idx * chunk_size,
p + attn_mask_shape[2], 1);
} else {
std::fill(p, p + attn_mask_shape[2], 1);
}
std::array<Ort::Value, 6> inputs = {std::move(x),
View(&offset),
View(&required_cache_size_tensor_),
std::move(attn_cache),
std::move(conv_cache),
std::move(attn_mask)};
auto out =
sess_->Run({}, input_names_ptr_.data(), inputs.data(), inputs.size(),
output_names_ptr_.data(), output_names_ptr_.size());
offset.GetTensorMutableData<int64_t>()[0] +=
out[0].GetTensorTypeAndShapeInfo().GetShape()[1];
out.push_back(std::move(offset));
return out;
}
int32_t VocabSize() const { return vocab_size_; }
int32_t ChunkLength() const {
// When chunk_size is 16, subsampling_factor_ is 4, right_context_ is 6,
// the returned value is (16 - 1)*4 + 6 + 1 = 67
return (config_.wenet_ctc.chunk_size - 1) * subsampling_factor_ +
right_context_ + 1;
}
int32_t ChunkShift() const { return required_cache_size_; }
OrtAllocator *Allocator() const { return allocator_; }
// Return a vector containing 3 tensors
// - attn_cache
// - conv_cache
// - offset
std::vector<Ort::Value> GetInitStates() {
std::vector<Ort::Value> ans;
ans.reserve(3);
ans.push_back(View(&attn_cache_));
ans.push_back(View(&conv_cache_));
int64_t offset_shape = 1;
Ort::Value offset =
Ort::Value::CreateTensor<int64_t>(allocator_, &offset_shape, 1);
offset.GetTensorMutableData<int64_t>()[0] = required_cache_size_;
ans.push_back(std::move(offset));
return ans;
}
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());
}
Ort::AllocatorWithDefaultOptions allocator; // used in the macro below
SHERPA_ONNX_READ_META_DATA(head_, "head");
SHERPA_ONNX_READ_META_DATA(num_blocks_, "num_blocks");
SHERPA_ONNX_READ_META_DATA(output_size_, "output_size");
SHERPA_ONNX_READ_META_DATA(cnn_module_kernel_, "cnn_module_kernel");
SHERPA_ONNX_READ_META_DATA(right_context_, "right_context");
SHERPA_ONNX_READ_META_DATA(subsampling_factor_, "subsampling_factor");
SHERPA_ONNX_READ_META_DATA(vocab_size_, "vocab_size");
required_cache_size_ =
config_.wenet_ctc.chunk_size * config_.wenet_ctc.num_left_chunks;
InitStates();
}
void InitStates() {
std::array<int64_t, 4> attn_cache_shape{
num_blocks_, head_, required_cache_size_, output_size_ / head_ * 2};
attn_cache_ = Ort::Value::CreateTensor<float>(
allocator_, attn_cache_shape.data(), attn_cache_shape.size());
Fill<float>(&attn_cache_, 0);
std::array<int64_t, 4> conv_cache_shape{num_blocks_, 1, output_size_,
cnn_module_kernel_ - 1};
conv_cache_ = Ort::Value::CreateTensor<float>(
allocator_, conv_cache_shape.data(), conv_cache_shape.size());
Fill<float>(&conv_cache_, 0);
int64_t shape = 1;
required_cache_size_tensor_ =
Ort::Value::CreateTensor<int64_t>(allocator_, &shape, 1);
required_cache_size_tensor_.GetTensorMutableData<int64_t>()[0] =
required_cache_size_;
}
private:
OnlineModelConfig 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 head_ = 0;
int32_t num_blocks_ = 0;
int32_t output_size_ = 0;
int32_t cnn_module_kernel_ = 0;
int32_t right_context_ = 0;
int32_t subsampling_factor_ = 0;
int32_t vocab_size_ = 0;
int32_t required_cache_size_ = 0;
Ort::Value attn_cache_{nullptr};
Ort::Value conv_cache_{nullptr};
Ort::Value required_cache_size_tensor_{nullptr};
};
OnlineWenetCtcModel::OnlineWenetCtcModel(const OnlineModelConfig &config)
: impl_(std::make_unique<Impl>(config)) {}
#if __ANDROID_API__ >= 9
OnlineWenetCtcModel::OnlineWenetCtcModel(AAssetManager *mgr,
const OnlineModelConfig &config)
: impl_(std::make_unique<Impl>(mgr, config)) {}
#endif
OnlineWenetCtcModel::~OnlineWenetCtcModel() = default;
std::vector<Ort::Value> OnlineWenetCtcModel::Forward(
Ort::Value x, std::vector<Ort::Value> states) const {
return impl_->Forward(std::move(x), std::move(states));
}
int32_t OnlineWenetCtcModel::VocabSize() const { return impl_->VocabSize(); }
int32_t OnlineWenetCtcModel::ChunkLength() const {
return impl_->ChunkLength();
}
int32_t OnlineWenetCtcModel::ChunkShift() const { return impl_->ChunkShift(); }
OrtAllocator *OnlineWenetCtcModel::Allocator() const {
return impl_->Allocator();
}
std::vector<Ort::Value> OnlineWenetCtcModel::GetInitStates() const {
return impl_->GetInitStates();
}
std::vector<Ort::Value> OnlineWenetCtcModel::StackStates(
std::vector<std::vector<Ort::Value>> states) const {
if (states.size() != 1) {
SHERPA_ONNX_LOGE("wenet CTC model supports only batch_size==1. Given: %d",
static_cast<int32_t>(states.size()));
}
return std::move(states[0]);
}
std::vector<std::vector<Ort::Value>> OnlineWenetCtcModel::UnStackStates(
std::vector<Ort::Value> states) const {
std::vector<std::vector<Ort::Value>> ans(1);
ans[0] = std::move(states);
return ans;
}
} // namespace sherpa_onnx