79c2ce5dd4
* Refactor online recognizer. Make it easier to support other streaming models. Note that it is a breaking change for the Python API. `sherpa_onnx.OnlineRecognizer()` used before should be replaced by `sherpa_onnx.OnlineRecognizer.from_transducer()`.
135 lines
3.9 KiB
C++
135 lines
3.9 KiB
C++
// sherpa-onnx/csrc/online-recognizer.cc
|
|
//
|
|
// Copyright (c) 2023 Xiaomi Corporation
|
|
// Copyright (c) 2023 Pingfeng Luo
|
|
|
|
#include "sherpa-onnx/csrc/online-recognizer.h"
|
|
|
|
#include <assert.h>
|
|
|
|
#include <algorithm>
|
|
#include <iomanip>
|
|
#include <memory>
|
|
#include <sstream>
|
|
#include <utility>
|
|
#include <vector>
|
|
|
|
#include "nlohmann/json.hpp"
|
|
#include "sherpa-onnx/csrc/online-recognizer-impl.h"
|
|
|
|
namespace sherpa_onnx {
|
|
|
|
std::string OnlineRecognizerResult::AsJsonString() const {
|
|
using json = nlohmann::json;
|
|
json j;
|
|
j["text"] = text;
|
|
j["tokens"] = tokens;
|
|
j["start_time"] = start_time;
|
|
#if 1
|
|
// This branch chooses number of decimal points to keep in
|
|
// the return json string
|
|
std::ostringstream os;
|
|
os << "[";
|
|
std::string sep = "";
|
|
for (auto t : timestamps) {
|
|
os << sep << std::fixed << std::setprecision(2) << t;
|
|
sep = ", ";
|
|
}
|
|
os << "]";
|
|
j["timestamps"] = os.str();
|
|
#else
|
|
j["timestamps"] = timestamps;
|
|
#endif
|
|
|
|
j["segment"] = segment;
|
|
j["is_final"] = is_final;
|
|
|
|
return j.dump();
|
|
}
|
|
|
|
void OnlineRecognizerConfig::Register(ParseOptions *po) {
|
|
feat_config.Register(po);
|
|
model_config.Register(po);
|
|
endpoint_config.Register(po);
|
|
lm_config.Register(po);
|
|
|
|
po->Register("enable-endpoint", &enable_endpoint,
|
|
"True to enable endpoint detection. False to disable it.");
|
|
po->Register("max-active-paths", &max_active_paths,
|
|
"beam size used in modified beam search.");
|
|
po->Register("context-score", &context_score,
|
|
"The bonus score for each token in context word/phrase. "
|
|
"Used only when decoding_method is modified_beam_search");
|
|
po->Register("decoding-method", &decoding_method,
|
|
"decoding method,"
|
|
"now support greedy_search and modified_beam_search.");
|
|
}
|
|
|
|
bool OnlineRecognizerConfig::Validate() const {
|
|
if (decoding_method == "modified_beam_search" && !lm_config.model.empty()) {
|
|
if (max_active_paths <= 0) {
|
|
SHERPA_ONNX_LOGE("max_active_paths is less than 0! Given: %d",
|
|
max_active_paths);
|
|
return false;
|
|
}
|
|
if (!lm_config.Validate()) return false;
|
|
}
|
|
return model_config.Validate();
|
|
}
|
|
|
|
std::string OnlineRecognizerConfig::ToString() const {
|
|
std::ostringstream os;
|
|
|
|
os << "OnlineRecognizerConfig(";
|
|
os << "feat_config=" << feat_config.ToString() << ", ";
|
|
os << "model_config=" << model_config.ToString() << ", ";
|
|
os << "lm_config=" << lm_config.ToString() << ", ";
|
|
os << "endpoint_config=" << endpoint_config.ToString() << ", ";
|
|
os << "enable_endpoint=" << (enable_endpoint ? "True" : "False") << ", ";
|
|
os << "max_active_paths=" << max_active_paths << ", ";
|
|
os << "context_score=" << context_score << ", ";
|
|
os << "decoding_method=\"" << decoding_method << "\")";
|
|
|
|
return os.str();
|
|
}
|
|
|
|
OnlineRecognizer::OnlineRecognizer(const OnlineRecognizerConfig &config)
|
|
: impl_(OnlineRecognizerImpl::Create(config)) {}
|
|
|
|
#if __ANDROID_API__ >= 9
|
|
OnlineRecognizer::OnlineRecognizer(AAssetManager *mgr,
|
|
const OnlineRecognizerConfig &config)
|
|
: impl_(OnlineRecognizerImpl::Create(mgr, config)) {}
|
|
#endif
|
|
|
|
OnlineRecognizer::~OnlineRecognizer() = default;
|
|
|
|
std::unique_ptr<OnlineStream> OnlineRecognizer::CreateStream() const {
|
|
return impl_->CreateStream();
|
|
}
|
|
|
|
std::unique_ptr<OnlineStream> OnlineRecognizer::CreateStream(
|
|
const std::vector<std::vector<int32_t>> &context_list) const {
|
|
return impl_->CreateStream(context_list);
|
|
}
|
|
|
|
bool OnlineRecognizer::IsReady(OnlineStream *s) const {
|
|
return impl_->IsReady(s);
|
|
}
|
|
|
|
void OnlineRecognizer::DecodeStreams(OnlineStream **ss, int32_t n) const {
|
|
impl_->DecodeStreams(ss, n);
|
|
}
|
|
|
|
OnlineRecognizerResult OnlineRecognizer::GetResult(OnlineStream *s) const {
|
|
return impl_->GetResult(s);
|
|
}
|
|
|
|
bool OnlineRecognizer::IsEndpoint(OnlineStream *s) const {
|
|
return impl_->IsEndpoint(s);
|
|
}
|
|
|
|
void OnlineRecognizer::Reset(OnlineStream *s) const { impl_->Reset(s); }
|
|
|
|
} // namespace sherpa_onnx
|