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

Limit number of tokens per second for whisper. (#1958)

Browse Source 
Otherwise, it spends lots of time in the loop if the EOT token
is not predicted.
This commit is contained in:
Fangjun Kuang
2025-03-04 15:45:28 +08:00
committed by GitHub
parent 49177530ff
commit 209eaaae1d
4 changed files with 14 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
sherpa-onnx/csrc/offline-whisper-greedy-search-decoder.h
View File

@@ -18,8 +18,9 @@ class OfflineWhisperGreedySearchDecoder : public OfflineWhisperDecoder {
OfflineWhisperModel *model)
: config_(config), model_(model) {}
std::vector<OfflineWhisperDecoderResult> Decode(Ort::Value cross_k,
Ort::Value cross_v) override;
std::vector<OfflineWhisperDecoderResult> Decode(
Ort::Value cross_k, Ort::Value cross_v,
int32_t num_feature_frames) override;
void SetConfig(const OfflineWhisperModelConfig &config) override;