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

Pascal API for non-streaming ASR (#1247)

Browse Source
This commit is contained in:
Fangjun Kuang
2024-08-12 23:33:35 +08:00
committed by GitHub
parent 5791b695ea
commit a7dc6c2c16
26 changed files with 1616 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files

77
pascal-api-examples/non-streaming-asr/nemo_transducer.pas Normal file
View File

@@ -0,0 +1,77 @@
{ Copyright (c) 2024 Xiaomi Corporation }
{
This file shows how to use a non-streaming NeMo transducer
to decode files.
You can download the model files from
https://github.com/k2-fsa/sherpa-onnx/releases/tag/asr-models
}
program nemo_transducer;
{$mode objfpc}
uses
sherpa_onnx,
DateUtils,
SysUtils;
var
Wave: TSherpaOnnxWave;
WaveFilename: AnsiString;
Config: TSherpaOnnxOfflineRecognizerConfig;
Recognizer: TSherpaOnnxOfflineRecognizer;
Stream: TSherpaOnnxOfflineStream;
RecognitionResult: TSherpaOnnxOfflineRecognizerResult;
Start: TDateTime;
Stop: TDateTime;
Elapsed: Single;
Duration: Single;
RealTimeFactor: Single;
begin
Config.ModelConfig.Transducer.Encoder := './sherpa-onnx-nemo-fast-conformer-transducer-be-de-en-es-fr-hr-it-pl-ru-uk-20k/encoder.onnx';
Config.ModelConfig.Transducer.Decoder := './sherpa-onnx-nemo-fast-conformer-transducer-be-de-en-es-fr-hr-it-pl-ru-uk-20k/decoder.onnx';
Config.ModelConfig.Transducer.Joiner := './sherpa-onnx-nemo-fast-conformer-transducer-be-de-en-es-fr-hr-it-pl-ru-uk-20k/joiner.onnx';
Config.ModelConfig.ModelType := 'nemo_transducer';
Config.ModelConfig.Tokens := './sherpa-onnx-nemo-fast-conformer-transducer-be-de-en-es-fr-hr-it-pl-ru-uk-20k/tokens.txt';
Config.ModelConfig.Provider := 'cpu';
Config.ModelConfig.NumThreads := 1;
Config.ModelConfig.Debug := False;
WaveFilename := './sherpa-onnx-nemo-fast-conformer-transducer-be-de-en-es-fr-hr-it-pl-ru-uk-20k/test_wavs/de-german.wav';
Wave := SherpaOnnxReadWave(WaveFilename);
Recognizer := TSherpaOnnxOfflineRecognizer.Create(Config);
Stream := Recognizer.CreateStream();
Start := Now;
Stream.AcceptWaveform(Wave.Samples, Wave.SampleRate);
Recognizer.Decode(Stream);
RecognitionResult := Recognizer.GetResult(Stream);
Stop := Now;
Elapsed := MilliSecondsBetween(Stop, Start) / 1000;
Duration := Length(Wave.Samples) / Wave.SampleRate;
RealTimeFactor := Elapsed / Duration;
WriteLn(RecognitionResult.ToString);
WriteLn(Format('NumThreads %d', [Config.ModelConfig.NumThreads]));
WriteLn(Format('Elapsed %.3f s', [Elapsed]));
WriteLn(Format('Wave duration %.3f s', [Duration]));
WriteLn(Format('RTF = %.3f/%.3f = %.3f', [Elapsed, Duration, RealTimeFactor]));
{Free resources to avoid memory leak.
Note: You don't need to invoke them for this simple script.
However, you have to invoke them in your own large/complex project.
}
FreeAndNil(Stream);
FreeAndNil(Recognizer);
end.