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

Export non-streaming NeMo faster conformer hybrid transducer and ctc to sherpa-onnx (#847)

Browse Source
This commit is contained in:
Fangjun Kuang
2024-05-09 13:59:47 +08:00
committed by GitHub
parent 68b25abf27
commit 5ed3ec1c04
16 changed files with 1055 additions and 20 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
scripts/nemo/fast-conformer-hybrid-transducer-ctc/export-onnx-transducer.py
View File

@@ -96,11 +96,15 @@ def main():
# encoder-model.onnx
# decoder_joint-model.onnx
normalize_type = asr_model.cfg.preprocessor.normalize
if normalize_type == "NA":
normalize_type = ""
meta_data = {
"vocab_size": asr_model.decoder.vocab_size,
"window_size": window_size,
"chunk_shift": chunk_shift,
"normalize_type": "None",
"normalize_type": normalize_type,
"cache_last_channel_dim1": cache_last_channel_dim1,
"cache_last_channel_dim2": cache_last_channel_dim2,
"cache_last_channel_dim3": cache_last_channel_dim3,