offline-whisper-model-config.cc
2.6 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
// sherpa-onnx/csrc/offline-whisper-model-config.cc
//
// Copyright (c) 2023 Xiaomi Corporation
#include "sherpa-onnx/csrc/offline-whisper-model-config.h"
#include "sherpa-onnx/csrc/file-utils.h"
#include "sherpa-onnx/csrc/macros.h"
namespace sherpa_onnx {
void OfflineWhisperModelConfig::Register(ParseOptions *po) {
po->Register("whisper-encoder", &encoder,
"Path to onnx encoder of whisper, e.g., tiny-encoder.onnx, "
"medium.en-encoder.onnx.");
po->Register("whisper-decoder", &decoder,
"Path to onnx decoder of whisper, e.g., tiny-decoder.onnx, "
"medium.en-decoder.onnx.");
po->Register(
"whisper-language", &language,
"The spoke language in the input audio file. Example values: "
"en, de, fr, zh, jp. If it is not given for a multilingual model, we will"
" infer the language from the input audio file. "
"Please refer to "
"https://github.com/openai/whisper/blob/main/whisper/tokenizer.py#L10"
" for valid values. Note that for non-multilingual models, it supports "
"only 'en'");
po->Register("whisper-task", &task,
"Valid values: transcribe, translate. "
"Note that for non-multilingual models, it supports "
"only 'transcribe'");
po->Register(
"whisper-tail-paddings", &tail_paddings,
"Suggest value: 50 for English models. 300 for multilingual models. "
"Since we have removed the 30-second constraint, we need to add some "
"tail padding frames "
"so that whisper can detect the eot token. Leave it to -1 to use 50 for "
"English models and 300 for multilingual models.");
}
bool OfflineWhisperModelConfig::Validate() const {
if (!FileExists(encoder)) {
SHERPA_ONNX_LOGE("whisper encoder file %s does not exist", encoder.c_str());
return false;
}
if (!FileExists(decoder)) {
SHERPA_ONNX_LOGE("whisper decoder file %s does not exist", decoder.c_str());
return false;
}
if (task != "translate" && task != "transcribe") {
SHERPA_ONNX_LOGE(
"--whisper-task supports only translate and transcribe. Given: %s",
task.c_str());
return false;
}
return true;
}
std::string OfflineWhisperModelConfig::ToString() const {
std::ostringstream os;
os << "OfflineWhisperModelConfig(";
os << "encoder=\"" << encoder << "\", ";
os << "decoder=\"" << decoder << "\", ";
os << "language=\"" << language << "\", ";
os << "task=\"" << task << "\", ";
os << "tail_paddings=" << tail_paddings << ")";
return os.str();
}
} // namespace sherpa_onnx