EngineX-Iluvatar/enginex-mr_series-sherpa-onnx
Archived
16
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Wiki Activity

Preserve more context after endpointing in transducer (#2061)

Browse Source
This commit is contained in:
Askars Salimbajevs
2025-04-02 18:33:47 +03:00
committed by GitHub
parent da4aad1189
commit 18a6ed5ddc
1 changed files with 10 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files

16
sherpa-onnx/csrc/online-recognizer-transducer-impl.h
View File

@@ -388,16 +388,20 @@ class OnlineRecognizerTransducerImpl : public OnlineRecognizerImpl {
auto r = decoder_->GetEmptyResult(); auto r = decoder_->GetEmptyResult();
auto last_result = s->GetResult(); auto last_result = s->GetResult();
// if last result is not empty, then // if last result is not empty, then
// preserve last tokens as the context for next result // truncate all last hyps and save as the context for next result
if (static_cast<int32_t>(last_result.tokens.size()) > context_size) { if (static_cast<int32_t>(last_result.tokens.size()) > context_size) {
std::vector<int64_t> context(last_result.tokens.end() - context_size, for (const auto &it : last_result.hyps) {
last_result.tokens.end()); auto h = it.second;
r.hyps.Add({std::vector<int64_t>(h.ys.end() - context_size,
h.ys.end()),
h.log_prob});
}
Hypotheses context_hyp({{context, 0}}); r.tokens = std::vector<int64_t> (last_result.tokens.end() - context_size,
r.hyps = std::move(context_hyp); last_result.tokens.end());
r.tokens = std::move(context);
} }
// but reset all contextual biasing graph states to root
if (config_.decoding_method == "modified_beam_search" && if (config_.decoding_method == "modified_beam_search" &&
nullptr != s->GetContextGraph()) { nullptr != s->GetContextGraph()) {
for (auto it = r.hyps.begin(); it != r.hyps.end(); ++it) { for (auto it = r.hyps.begin(); it != r.hyps.end(); ++it) {