online-nemo-ctc-model.h
2.2 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
// sherpa-onnx/csrc/online-nemo-ctc-model.h
//
// Copyright (c) 2024 Xiaomi Corporation
#ifndef SHERPA_ONNX_CSRC_ONLINE_NEMO_CTC_MODEL_H_
#define SHERPA_ONNX_CSRC_ONLINE_NEMO_CTC_MODEL_H_
#include <memory>
#include <utility>
#include <vector>
#if __ANDROID_API__ >= 9
#include "android/asset_manager.h"
#include "android/asset_manager_jni.h"
#endif
#include "onnxruntime_cxx_api.h" // NOLINT
#include "sherpa-onnx/csrc/online-ctc-model.h"
#include "sherpa-onnx/csrc/online-model-config.h"
namespace sherpa_onnx {
class OnlineNeMoCtcModel : public OnlineCtcModel {
public:
explicit OnlineNeMoCtcModel(const OnlineModelConfig &config);
#if __ANDROID_API__ >= 9
OnlineNeMoCtcModel(AAssetManager *mgr, const OnlineModelConfig &config);
#endif
~OnlineNeMoCtcModel() override;
// A list of 3 tensors:
// - cache_last_channel
// - cache_last_time
// - cache_last_channel_len
std::vector<Ort::Value> GetInitStates() const override;
std::vector<Ort::Value> StackStates(
std::vector<std::vector<Ort::Value>> states) const override;
std::vector<std::vector<Ort::Value>> UnStackStates(
std::vector<Ort::Value> states) const override;
/**
*
* @param x A 3-D tensor of shape (N, T, C). N has to be 1.
* @param states It is from GetInitStates() or returned from this method.
*
* @return Return a list of tensors
* - ans[0] contains log_probs, of shape (N, T, C)
* - ans[1:] contains next_states
*/
std::vector<Ort::Value> Forward(
Ort::Value x, std::vector<Ort::Value> states) const override;
/** Return the vocabulary size of the model
*/
int32_t VocabSize() const override;
/** Return an allocator for allocating memory
*/
OrtAllocator *Allocator() const override;
// The model accepts this number of frames before subsampling as input
int32_t ChunkLength() const override;
// Similar to frame_shift in feature extractor, after processing
// ChunkLength() frames, we advance by ChunkShift() frames
// before we process the next chunk.
int32_t ChunkShift() const override;
bool SupportBatchProcessing() const override { return true; }
private:
class Impl;
std::unique_ptr<Impl> impl_;
};
} // namespace sherpa_onnx
#endif // SHERPA_ONNX_CSRC_ONLINE_NEMO_CTC_MODEL_H_