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

Fix lm fusion (#157)

Browse Source 
* share GetHypsRowSplits interface and fix getting Topk not taking logprob

* fix lm score of lm fusion and make padding len same with 'icefall/egs/librispeech/ASR/pruned_transducer_stateless7_streaming/decode.py'
This commit is contained in:
PF Luo
2023-05-15 10:48:45 +08:00
committed by GitHub
parent 959f13eac8
commit 655c619bf3
3 changed files with 5 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
sherpa-onnx/csrc/sherpa-onnx.cc
View File

@@ -94,7 +94,7 @@ for a list of pre-trained models to download.
auto s = recognizer.CreateStream();
s->AcceptWaveform(sampling_rate, samples.data(), samples.size());
std::vector<float> tail_paddings(static_cast<int>(0.5 * sampling_rate));
std::vector<float> tail_paddings(static_cast<int>(0.3 * sampling_rate));
// Note: We can call AcceptWaveform() multiple times.
s->AcceptWaveform(sampling_rate, tail_paddings.data(), tail_paddings.size());