Fix VAD+ASR C++ example. (#2335)
It was not able to handle short audios., e.g., 2.1 seconds.
This commit is contained in:
@@ -300,6 +300,66 @@ def get_models():
|
||||
|
||||
ls -lh
|
||||
|
||||
popd
|
||||
""",
|
||||
),
|
||||
Model(
|
||||
model_name="sherpa-onnx-streaming-zipformer-ctc-zh-int8-2025-06-30",
|
||||
idx=17,
|
||||
lang="zh",
|
||||
short_name="large_zipformer_int8",
|
||||
rule_fsts="itn_zh_number.fst",
|
||||
cmd="""
|
||||
if [ ! -f itn_zh_number.fst ]; then
|
||||
curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/itn_zh_number.fst
|
||||
fi
|
||||
pushd $model_name
|
||||
rm -fv bpe.model
|
||||
|
||||
rm -rf test_wavs
|
||||
|
||||
ls -lh
|
||||
|
||||
popd
|
||||
""",
|
||||
),
|
||||
Model(
|
||||
model_name="sherpa-onnx-streaming-zipformer-ctc-zh-2025-06-30",
|
||||
idx=18,
|
||||
lang="zh",
|
||||
short_name="large_zipformer",
|
||||
rule_fsts="itn_zh_number.fst",
|
||||
cmd="""
|
||||
if [ ! -f itn_zh_number.fst ]; then
|
||||
curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/itn_zh_number.fst
|
||||
fi
|
||||
pushd $model_name
|
||||
rm -fv bpe.model
|
||||
|
||||
rm -rf test_wavs
|
||||
|
||||
ls -lh
|
||||
|
||||
popd
|
||||
""",
|
||||
),
|
||||
Model(
|
||||
model_name="sherpa-onnx-streaming-zipformer-ctc-fp16-zh-2025-06-30",
|
||||
idx=19,
|
||||
lang="zh",
|
||||
short_name="large_zipformer_fp16",
|
||||
rule_fsts="itn_zh_number.fst",
|
||||
cmd="""
|
||||
if [ ! -f itn_zh_number.fst ]; then
|
||||
curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/itn_zh_number.fst
|
||||
fi
|
||||
pushd $model_name
|
||||
rm -fv bpe.model
|
||||
|
||||
rm -rf test_wavs
|
||||
|
||||
ls -lh
|
||||
|
||||
popd
|
||||
""",
|
||||
),
|
||||
|
||||
@@ -186,13 +186,15 @@ for a list of pre-trained models to download.
|
||||
fprintf(stderr, "Started!\n");
|
||||
int32_t window_size = vad_config.silero_vad.window_size;
|
||||
int32_t i = 0;
|
||||
while (i + window_size < samples.size()) {
|
||||
vad->AcceptWaveform(samples.data() + i, window_size);
|
||||
i += window_size;
|
||||
if (i >= samples.size()) {
|
||||
while (i < samples.size()) {
|
||||
if (i + window_size <= samples.size()) {
|
||||
vad->AcceptWaveform(samples.data() + i, window_size);
|
||||
} else {
|
||||
vad->Flush();
|
||||
}
|
||||
|
||||
i += window_size;
|
||||
|
||||
while (!vad->Empty()) {
|
||||
const auto &segment = vad->Front();
|
||||
float duration = segment.samples.size() / 16000.;
|
||||
|
||||
@@ -395,6 +395,36 @@ fun getModelConfig(type: Int): OnlineModelConfig? {
|
||||
)
|
||||
}
|
||||
|
||||
17 -> {
|
||||
val modelDir = "sherpa-onnx-streaming-zipformer-ctc-zh-int8-2025-06-30"
|
||||
return OnlineModelConfig(
|
||||
zipformer2Ctc = OnlineZipformer2CtcModelConfig(
|
||||
model = "$modelDir/model.int8.onnx",
|
||||
),
|
||||
tokens = "$modelDir/tokens.txt",
|
||||
)
|
||||
}
|
||||
|
||||
18 -> {
|
||||
val modelDir = "sherpa-onnx-streaming-zipformer-ctc-zh-2025-06-30"
|
||||
return OnlineModelConfig(
|
||||
zipformer2Ctc = OnlineZipformer2CtcModelConfig(
|
||||
model = "$modelDir/model.onnx",
|
||||
),
|
||||
tokens = "$modelDir/tokens.txt",
|
||||
)
|
||||
}
|
||||
|
||||
19 -> {
|
||||
val modelDir = "sherpa-onnx-streaming-zipformer-ctc-zh-fp16-2025-06-30"
|
||||
return OnlineModelConfig(
|
||||
zipformer2Ctc = OnlineZipformer2CtcModelConfig(
|
||||
model = "$modelDir/model.fp16.onnx",
|
||||
),
|
||||
tokens = "$modelDir/tokens.txt",
|
||||
)
|
||||
}
|
||||
|
||||
1000 -> {
|
||||
val modelDir = "sherpa-onnx-rk3588-streaming-zipformer-bilingual-zh-en-2023-02-20"
|
||||
return OnlineModelConfig(
|
||||
|
||||
Reference in New Issue
Block a user