From 8136ad775b6cd02bf2ecc60d65e8641b709c2d41 Mon Sep 17 00:00:00 2001 From: Fangjun Kuang Date: Thu, 4 Jan 2024 13:59:32 +0800 Subject: [PATCH] Use high_freq -400 in computing fbank features. (#1447) See also https://github.com/k2-fsa/sherpa-onnx/issues/514 --- .../ASR/pruned_transducer_stateless2/pretrained.py | 1 + egs/aishell/ASR/conformer_ctc/pretrained.py | 1 + egs/aishell/ASR/pruned_transducer_stateless2/pretrained.py | 1 + egs/aishell/ASR/pruned_transducer_stateless3/pretrained.py | 1 + egs/aishell/ASR/pruned_transducer_stateless7/jit_pretrained.py | 1 + egs/aishell/ASR/pruned_transducer_stateless7/onnx_pretrained.py | 1 + .../ASR/pruned_transducer_stateless7_bbpe/jit_pretrained.py | 1 + egs/aishell/ASR/pruned_transducer_stateless7_bbpe/pretrained.py | 1 + .../pruned_transducer_stateless7_streaming/streaming_decode.py | 1 + egs/aishell/ASR/tdnn_lstm_ctc/pretrained.py | 1 + egs/aishell/ASR/transducer_stateless/pretrained.py | 1 + egs/aishell/ASR/transducer_stateless_modified-2/pretrained.py | 1 + egs/aishell/ASR/transducer_stateless_modified/pretrained.py | 1 + egs/aishell/ASR/zipformer/streaming_decode.py | 1 + egs/aishell2/ASR/pruned_transducer_stateless5/pretrained.py | 1 + egs/aishell4/ASR/pruned_transducer_stateless5/pretrained.py | 1 + egs/alimeeting/ASR/pruned_transducer_stateless2/pretrained.py | 1 + .../ASR/pruned_transducer_stateless7/onnx_pretrained.py | 1 + egs/commonvoice/ASR/pruned_transducer_stateless7/pretrained.py | 1 + .../pruned_transducer_stateless7_streaming/streaming_decode.py | 1 + .../jit_trace_pretrained.py | 1 + egs/csj/ASR/pruned_transducer_stateless7_streaming/pretrained.py | 1 + .../pruned_transducer_stateless7_streaming/streaming_decode.py | 1 + egs/gigaspeech/ASR/zipformer/streaming_decode.py | 1 + egs/libriheavy/ASR/zipformer_prompt_asr/pretrained.py | 1 + .../ASR/conformer_ctc/jit_pretrained_decode_with_H.py | 1 + .../ASR/conformer_ctc/jit_pretrained_decode_with_HL.py | 1 + .../ASR/conformer_ctc/jit_pretrained_decode_with_HLG.py | 1 + egs/librispeech/ASR/conformer_ctc/pretrained.py | 1 + egs/librispeech/ASR/conformer_ctc3/jit_pretrained.py | 1 + egs/librispeech/ASR/conformer_ctc3/pretrained.py | 1 + .../ASR/conv_emformer_transducer_stateless/streaming_decode.py | 1 + .../ASR/conv_emformer_transducer_stateless2/jit_pretrained.py | 1 + .../ASR/conv_emformer_transducer_stateless2/onnx_pretrained.py | 1 + .../conv_emformer_transducer_stateless2/streaming-ncnn-decode.py | 1 + .../ASR/conv_emformer_transducer_stateless2/streaming_decode.py | 1 + egs/librispeech/ASR/lstm_transducer_stateless/jit_pretrained.py | 1 + egs/librispeech/ASR/lstm_transducer_stateless/pretrained.py | 1 + .../ASR/lstm_transducer_stateless/streaming_decode.py | 1 + egs/librispeech/ASR/lstm_transducer_stateless2/jit_pretrained.py | 1 + egs/librispeech/ASR/lstm_transducer_stateless2/ncnn-decode.py | 1 + .../ASR/lstm_transducer_stateless2/onnx_pretrained.py | 1 + egs/librispeech/ASR/lstm_transducer_stateless2/pretrained.py | 1 + .../ASR/lstm_transducer_stateless2/streaming-ncnn-decode.py | 1 + .../ASR/lstm_transducer_stateless2/streaming-onnx-decode.py | 1 + egs/librispeech/ASR/lstm_transducer_stateless3/jit_pretrained.py | 1 + egs/librispeech/ASR/lstm_transducer_stateless3/pretrained.py | 1 + .../ASR/lstm_transducer_stateless3/streaming_decode.py | 1 + egs/librispeech/ASR/pruned_transducer_stateless/pretrained.py | 1 + .../ASR/pruned_transducer_stateless/streaming_decode.py | 1 + egs/librispeech/ASR/pruned_transducer_stateless2/pretrained.py | 1 + .../ASR/pruned_transducer_stateless2/streaming_decode.py | 1 + .../ASR/pruned_transducer_stateless3/jit_pretrained.py | 1 + .../ASR/pruned_transducer_stateless3/onnx_pretrained.py | 1 + egs/librispeech/ASR/pruned_transducer_stateless3/pretrained.py | 1 + .../ASR/pruned_transducer_stateless3/streaming_decode.py | 1 + .../ASR/pruned_transducer_stateless4/streaming_decode.py | 1 + .../pruned_transducer_stateless5/onnx_pretrained-streaming.py | 1 + egs/librispeech/ASR/pruned_transducer_stateless5/pretrained.py | 1 + .../ASR/pruned_transducer_stateless5/streaming_decode.py | 1 + .../ASR/pruned_transducer_stateless7/jit_pretrained.py | 1 + egs/librispeech/ASR/pruned_transducer_stateless7/pretrained.py | 1 + .../ASR/pruned_transducer_stateless7_ctc/jit_pretrained.py | 1 + .../ASR/pruned_transducer_stateless7_ctc/jit_pretrained_ctc.py | 1 + .../ASR/pruned_transducer_stateless7_ctc/pretrained.py | 1 + .../ASR/pruned_transducer_stateless7_ctc/pretrained_ctc.py | 1 + .../ASR/pruned_transducer_stateless7_ctc_bs/jit_pretrained.py | 1 + .../pruned_transducer_stateless7_ctc_bs/jit_pretrained_ctc.py | 1 + .../ASR/pruned_transducer_stateless7_ctc_bs/onnx_pretrained.py | 1 + .../ASR/pruned_transducer_stateless7_ctc_bs/pretrained.py | 1 + .../ASR/pruned_transducer_stateless7_ctc_bs/pretrained_ctc.py | 1 + .../ASR/pruned_transducer_stateless7_streaming/jit_pretrained.py | 1 + .../jit_trace_pretrained.py | 1 + .../pruned_transducer_stateless7_streaming/onnx_pretrained.py | 1 + .../ASR/pruned_transducer_stateless7_streaming/pretrained.py | 1 + .../streaming-ncnn-decode.py | 1 + .../pruned_transducer_stateless7_streaming/streaming_decode.py | 1 + .../ASR/pruned_transducer_stateless8/jit_pretrained.py | 1 + egs/librispeech/ASR/pruned_transducer_stateless8/pretrained.py | 1 + egs/librispeech/ASR/tdnn_lstm_ctc/pretrained.py | 1 + egs/librispeech/ASR/tiny_transducer_ctc/jit_pretrained.py | 1 + egs/librispeech/ASR/tiny_transducer_ctc/jit_pretrained_ctc.py | 1 + egs/librispeech/ASR/tiny_transducer_ctc/pretrained.py | 1 + egs/librispeech/ASR/tiny_transducer_ctc/pretrained_ctc.py | 1 + egs/librispeech/ASR/transducer/pretrained.py | 1 + egs/librispeech/ASR/transducer_stateless/pretrained.py | 1 + egs/librispeech/ASR/transducer_stateless2/pretrained.py | 1 + .../ASR/transducer_stateless_multi_datasets/pretrained.py | 1 + egs/librispeech/ASR/zipformer/jit_pretrained.py | 1 + egs/librispeech/ASR/zipformer/jit_pretrained_ctc.py | 1 + egs/librispeech/ASR/zipformer/jit_pretrained_streaming.py | 1 + egs/librispeech/ASR/zipformer/onnx_pretrained-streaming-ctc.py | 1 + egs/librispeech/ASR/zipformer/onnx_pretrained-streaming.py | 1 + egs/librispeech/ASR/zipformer/onnx_pretrained.py | 1 + egs/librispeech/ASR/zipformer/onnx_pretrained_ctc.py | 1 + egs/librispeech/ASR/zipformer/onnx_pretrained_ctc_H.py | 1 + egs/librispeech/ASR/zipformer/onnx_pretrained_ctc_HL.py | 1 + egs/librispeech/ASR/zipformer/onnx_pretrained_ctc_HLG.py | 1 + egs/librispeech/ASR/zipformer/pretrained.py | 1 + egs/librispeech/ASR/zipformer/pretrained_ctc.py | 1 + egs/librispeech/ASR/zipformer_mmi/jit_pretrained.py | 1 + egs/librispeech/ASR/zipformer_mmi/pretrained.py | 1 + egs/mgb2/ASR/conformer_ctc/pretrained.py | 1 + egs/mgb2/ASR/pruned_transducer_stateless5/pretrained.py | 1 + egs/multi_zh-hans/ASR/zipformer/pretrained.py | 1 + egs/multi_zh_en/ASR/zipformer/pretrained.py | 1 + egs/tal_csasr/ASR/pruned_transducer_stateless5/pretrained.py | 1 + .../ASR/pruned_transducer_stateless7_bbpe/jit_pretrained.py | 1 + .../ASR/pruned_transducer_stateless7_bbpe/pretrained.py | 1 + egs/tedlium3/ASR/pruned_transducer_stateless/pretrained.py | 1 + egs/tedlium3/ASR/transducer_stateless/pretrained.py | 1 + egs/timit/ASR/tdnn_ligru_ctc/pretrained.py | 1 + egs/timit/ASR/tdnn_lstm_ctc/pretrained.py | 1 + .../ASR/pruned_transducer_stateless2/jit_pretrained.py | 1 + egs/wenetspeech/ASR/pruned_transducer_stateless2/pretrained.py | 1 + .../pruned_transducer_stateless5/onnx_pretrained-streaming.py | 1 + .../ASR/pruned_transducer_stateless5/onnx_pretrained.py | 1 + egs/wenetspeech/ASR/pruned_transducer_stateless5/pretrained.py | 1 + .../ASR/pruned_transducer_stateless5/streaming_decode.py | 1 + egs/wenetspeech/ASR/zipformer/streaming_decode.py | 1 + egs/xbmu_amdo31/ASR/pruned_transducer_stateless5/pretrained.py | 1 + egs/xbmu_amdo31/ASR/pruned_transducer_stateless7/pretrained.py | 1 + egs/yesno/ASR/tdnn/jit_pretrained.py | 1 + egs/yesno/ASR/tdnn/jit_pretrained_decode_with_H.py | 1 + egs/yesno/ASR/tdnn/jit_pretrained_decode_with_HL.py | 1 + egs/yesno/ASR/tdnn/onnx_pretrained.py | 1 + egs/yesno/ASR/tdnn/pretrained.py | 1 + 127 files changed, 127 insertions(+) diff --git a/egs/aidatatang_200zh/ASR/pruned_transducer_stateless2/pretrained.py b/egs/aidatatang_200zh/ASR/pruned_transducer_stateless2/pretrained.py index 75c316eaf..17729e02e 100644 --- a/egs/aidatatang_200zh/ASR/pruned_transducer_stateless2/pretrained.py +++ b/egs/aidatatang_200zh/ASR/pruned_transducer_stateless2/pretrained.py @@ -242,6 +242,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/aishell/ASR/conformer_ctc/pretrained.py b/egs/aishell/ASR/conformer_ctc/pretrained.py index 66d583396..af1171a6f 100755 --- a/egs/aishell/ASR/conformer_ctc/pretrained.py +++ b/egs/aishell/ASR/conformer_ctc/pretrained.py @@ -261,6 +261,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/aishell/ASR/pruned_transducer_stateless2/pretrained.py b/egs/aishell/ASR/pruned_transducer_stateless2/pretrained.py index 82c10f129..c4aa98358 100755 --- a/egs/aishell/ASR/pruned_transducer_stateless2/pretrained.py +++ b/egs/aishell/ASR/pruned_transducer_stateless2/pretrained.py @@ -240,6 +240,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/aishell/ASR/pruned_transducer_stateless3/pretrained.py b/egs/aishell/ASR/pruned_transducer_stateless3/pretrained.py index ead393e6e..69fe3a40b 100755 --- a/egs/aishell/ASR/pruned_transducer_stateless3/pretrained.py +++ b/egs/aishell/ASR/pruned_transducer_stateless3/pretrained.py @@ -241,6 +241,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/aishell/ASR/pruned_transducer_stateless7/jit_pretrained.py b/egs/aishell/ASR/pruned_transducer_stateless7/jit_pretrained.py index e61190649..5143f2cae 100755 --- a/egs/aishell/ASR/pruned_transducer_stateless7/jit_pretrained.py +++ b/egs/aishell/ASR/pruned_transducer_stateless7/jit_pretrained.py @@ -230,6 +230,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = 16000 opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/aishell/ASR/pruned_transducer_stateless7/onnx_pretrained.py b/egs/aishell/ASR/pruned_transducer_stateless7/onnx_pretrained.py index a92182e8d..8e8e971eb 100755 --- a/egs/aishell/ASR/pruned_transducer_stateless7/onnx_pretrained.py +++ b/egs/aishell/ASR/pruned_transducer_stateless7/onnx_pretrained.py @@ -369,6 +369,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = args.sample_rate opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/aishell/ASR/pruned_transducer_stateless7_bbpe/jit_pretrained.py b/egs/aishell/ASR/pruned_transducer_stateless7_bbpe/jit_pretrained.py index 0c43bf74b..8fb7ac278 100755 --- a/egs/aishell/ASR/pruned_transducer_stateless7_bbpe/jit_pretrained.py +++ b/egs/aishell/ASR/pruned_transducer_stateless7_bbpe/jit_pretrained.py @@ -227,6 +227,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = 16000 opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/aishell/ASR/pruned_transducer_stateless7_bbpe/pretrained.py b/egs/aishell/ASR/pruned_transducer_stateless7_bbpe/pretrained.py index ea5bda4db..12004315b 100755 --- a/egs/aishell/ASR/pruned_transducer_stateless7_bbpe/pretrained.py +++ b/egs/aishell/ASR/pruned_transducer_stateless7_bbpe/pretrained.py @@ -250,6 +250,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/aishell/ASR/pruned_transducer_stateless7_streaming/streaming_decode.py b/egs/aishell/ASR/pruned_transducer_stateless7_streaming/streaming_decode.py index 6b4f183cf..aa0e07c83 100755 --- a/egs/aishell/ASR/pruned_transducer_stateless7_streaming/streaming_decode.py +++ b/egs/aishell/ASR/pruned_transducer_stateless7_streaming/streaming_decode.py @@ -317,6 +317,7 @@ def decode_dataset( opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = 16000 opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 log_interval = 50 diff --git a/egs/aishell/ASR/tdnn_lstm_ctc/pretrained.py b/egs/aishell/ASR/tdnn_lstm_ctc/pretrained.py index 7e7213501..9754b4939 100644 --- a/egs/aishell/ASR/tdnn_lstm_ctc/pretrained.py +++ b/egs/aishell/ASR/tdnn_lstm_ctc/pretrained.py @@ -158,6 +158,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/aishell/ASR/transducer_stateless/pretrained.py b/egs/aishell/ASR/transducer_stateless/pretrained.py index 40f430e13..540e7b61b 100755 --- a/egs/aishell/ASR/transducer_stateless/pretrained.py +++ b/egs/aishell/ASR/transducer_stateless/pretrained.py @@ -258,6 +258,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/aishell/ASR/transducer_stateless_modified-2/pretrained.py b/egs/aishell/ASR/transducer_stateless_modified-2/pretrained.py index 5d8ca2e11..4a4e9237c 100755 --- a/egs/aishell/ASR/transducer_stateless_modified-2/pretrained.py +++ b/egs/aishell/ASR/transducer_stateless_modified-2/pretrained.py @@ -238,6 +238,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/aishell/ASR/transducer_stateless_modified/pretrained.py b/egs/aishell/ASR/transducer_stateless_modified/pretrained.py index 9e4459247..66a91709e 100755 --- a/egs/aishell/ASR/transducer_stateless_modified/pretrained.py +++ b/egs/aishell/ASR/transducer_stateless_modified/pretrained.py @@ -238,6 +238,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/aishell/ASR/zipformer/streaming_decode.py b/egs/aishell/ASR/zipformer/streaming_decode.py index c3820447a..f54ffbd3c 100755 --- a/egs/aishell/ASR/zipformer/streaming_decode.py +++ b/egs/aishell/ASR/zipformer/streaming_decode.py @@ -572,6 +572,7 @@ def decode_dataset( opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = 16000 opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 log_interval = 100 diff --git a/egs/aishell2/ASR/pruned_transducer_stateless5/pretrained.py b/egs/aishell2/ASR/pruned_transducer_stateless5/pretrained.py index bc3ae7abf..f04632388 100755 --- a/egs/aishell2/ASR/pruned_transducer_stateless5/pretrained.py +++ b/egs/aishell2/ASR/pruned_transducer_stateless5/pretrained.py @@ -239,6 +239,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/aishell4/ASR/pruned_transducer_stateless5/pretrained.py b/egs/aishell4/ASR/pruned_transducer_stateless5/pretrained.py index ee898c303..e8b7f71b7 100755 --- a/egs/aishell4/ASR/pruned_transducer_stateless5/pretrained.py +++ b/egs/aishell4/ASR/pruned_transducer_stateless5/pretrained.py @@ -251,6 +251,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/alimeeting/ASR/pruned_transducer_stateless2/pretrained.py b/egs/alimeeting/ASR/pruned_transducer_stateless2/pretrained.py index f5a0dd8c8..a738bb3fb 100644 --- a/egs/alimeeting/ASR/pruned_transducer_stateless2/pretrained.py +++ b/egs/alimeeting/ASR/pruned_transducer_stateless2/pretrained.py @@ -242,6 +242,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/commonvoice/ASR/pruned_transducer_stateless7/onnx_pretrained.py b/egs/commonvoice/ASR/pruned_transducer_stateless7/onnx_pretrained.py index cf6ddfa36..52fed7331 100755 --- a/egs/commonvoice/ASR/pruned_transducer_stateless7/onnx_pretrained.py +++ b/egs/commonvoice/ASR/pruned_transducer_stateless7/onnx_pretrained.py @@ -370,6 +370,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = args.sample_rate opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/commonvoice/ASR/pruned_transducer_stateless7/pretrained.py b/egs/commonvoice/ASR/pruned_transducer_stateless7/pretrained.py index a22d1b4ba..b6e2451e8 100755 --- a/egs/commonvoice/ASR/pruned_transducer_stateless7/pretrained.py +++ b/egs/commonvoice/ASR/pruned_transducer_stateless7/pretrained.py @@ -260,6 +260,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/commonvoice/ASR/pruned_transducer_stateless7_streaming/streaming_decode.py b/egs/commonvoice/ASR/pruned_transducer_stateless7_streaming/streaming_decode.py index dbe65d0a7..018736d26 100755 --- a/egs/commonvoice/ASR/pruned_transducer_stateless7_streaming/streaming_decode.py +++ b/egs/commonvoice/ASR/pruned_transducer_stateless7_streaming/streaming_decode.py @@ -320,6 +320,7 @@ def decode_dataset( opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = 16000 opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 log_interval = 50 diff --git a/egs/csj/ASR/pruned_transducer_stateless7_streaming/jit_trace_pretrained.py b/egs/csj/ASR/pruned_transducer_stateless7_streaming/jit_trace_pretrained.py index d84cf04a3..58ee99e6a 100644 --- a/egs/csj/ASR/pruned_transducer_stateless7_streaming/jit_trace_pretrained.py +++ b/egs/csj/ASR/pruned_transducer_stateless7_streaming/jit_trace_pretrained.py @@ -177,6 +177,7 @@ def create_streaming_feature_extractor(sample_rate) -> OnlineFeature: opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = sample_rate opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 return OnlineFbank(opts) diff --git a/egs/csj/ASR/pruned_transducer_stateless7_streaming/pretrained.py b/egs/csj/ASR/pruned_transducer_stateless7_streaming/pretrained.py index 932026868..66fbae378 100644 --- a/egs/csj/ASR/pruned_transducer_stateless7_streaming/pretrained.py +++ b/egs/csj/ASR/pruned_transducer_stateless7_streaming/pretrained.py @@ -252,6 +252,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/csj/ASR/pruned_transducer_stateless7_streaming/streaming_decode.py b/egs/csj/ASR/pruned_transducer_stateless7_streaming/streaming_decode.py index 9700dd89e..7252665a7 100755 --- a/egs/csj/ASR/pruned_transducer_stateless7_streaming/streaming_decode.py +++ b/egs/csj/ASR/pruned_transducer_stateless7_streaming/streaming_decode.py @@ -337,6 +337,7 @@ def decode_dataset( opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = 16000 opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 log_interval = 50 diff --git a/egs/gigaspeech/ASR/zipformer/streaming_decode.py b/egs/gigaspeech/ASR/zipformer/streaming_decode.py index a76788859..09df2935c 100755 --- a/egs/gigaspeech/ASR/zipformer/streaming_decode.py +++ b/egs/gigaspeech/ASR/zipformer/streaming_decode.py @@ -553,6 +553,7 @@ def decode_dataset( opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = 16000 opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 log_interval = 100 diff --git a/egs/libriheavy/ASR/zipformer_prompt_asr/pretrained.py b/egs/libriheavy/ASR/zipformer_prompt_asr/pretrained.py index 48fd2612a..458109a3f 100644 --- a/egs/libriheavy/ASR/zipformer_prompt_asr/pretrained.py +++ b/egs/libriheavy/ASR/zipformer_prompt_asr/pretrained.py @@ -264,6 +264,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/librispeech/ASR/conformer_ctc/jit_pretrained_decode_with_H.py b/egs/librispeech/ASR/conformer_ctc/jit_pretrained_decode_with_H.py index 4bdec9e11..e9acf7e0b 100755 --- a/egs/librispeech/ASR/conformer_ctc/jit_pretrained_decode_with_H.py +++ b/egs/librispeech/ASR/conformer_ctc/jit_pretrained_decode_with_H.py @@ -195,6 +195,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = sample_rate opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/librispeech/ASR/conformer_ctc/jit_pretrained_decode_with_HL.py b/egs/librispeech/ASR/conformer_ctc/jit_pretrained_decode_with_HL.py index d5a1dba3c..5753aa5d3 100755 --- a/egs/librispeech/ASR/conformer_ctc/jit_pretrained_decode_with_HL.py +++ b/egs/librispeech/ASR/conformer_ctc/jit_pretrained_decode_with_HL.py @@ -192,6 +192,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = sample_rate opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/librispeech/ASR/conformer_ctc/jit_pretrained_decode_with_HLG.py b/egs/librispeech/ASR/conformer_ctc/jit_pretrained_decode_with_HLG.py index 216677a23..b6e3333ce 100755 --- a/egs/librispeech/ASR/conformer_ctc/jit_pretrained_decode_with_HLG.py +++ b/egs/librispeech/ASR/conformer_ctc/jit_pretrained_decode_with_HLG.py @@ -191,6 +191,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = sample_rate opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/librispeech/ASR/conformer_ctc/pretrained.py b/egs/librispeech/ASR/conformer_ctc/pretrained.py index df3e4d819..38b60fcb9 100755 --- a/egs/librispeech/ASR/conformer_ctc/pretrained.py +++ b/egs/librispeech/ASR/conformer_ctc/pretrained.py @@ -283,6 +283,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/librispeech/ASR/conformer_ctc3/jit_pretrained.py b/egs/librispeech/ASR/conformer_ctc3/jit_pretrained.py index 76db46cc8..19b26361e 100755 --- a/egs/librispeech/ASR/conformer_ctc3/jit_pretrained.py +++ b/egs/librispeech/ASR/conformer_ctc3/jit_pretrained.py @@ -271,6 +271,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/librispeech/ASR/conformer_ctc3/pretrained.py b/egs/librispeech/ASR/conformer_ctc3/pretrained.py index c37b99cce..a0cdfcf03 100755 --- a/egs/librispeech/ASR/conformer_ctc3/pretrained.py +++ b/egs/librispeech/ASR/conformer_ctc3/pretrained.py @@ -302,6 +302,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/librispeech/ASR/conv_emformer_transducer_stateless/streaming_decode.py b/egs/librispeech/ASR/conv_emformer_transducer_stateless/streaming_decode.py index e5a7c7116..9b8b4cce2 100755 --- a/egs/librispeech/ASR/conv_emformer_transducer_stateless/streaming_decode.py +++ b/egs/librispeech/ASR/conv_emformer_transducer_stateless/streaming_decode.py @@ -623,6 +623,7 @@ def create_streaming_feature_extractor() -> Fbank: opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = 16000 opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 return Fbank(opts) diff --git a/egs/librispeech/ASR/conv_emformer_transducer_stateless2/jit_pretrained.py b/egs/librispeech/ASR/conv_emformer_transducer_stateless2/jit_pretrained.py index 1fe358c79..58f587c91 100755 --- a/egs/librispeech/ASR/conv_emformer_transducer_stateless2/jit_pretrained.py +++ b/egs/librispeech/ASR/conv_emformer_transducer_stateless2/jit_pretrained.py @@ -184,6 +184,7 @@ def create_streaming_feature_extractor(sample_rate) -> OnlineFeature: opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = sample_rate opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 return OnlineFbank(opts) diff --git a/egs/librispeech/ASR/conv_emformer_transducer_stateless2/onnx_pretrained.py b/egs/librispeech/ASR/conv_emformer_transducer_stateless2/onnx_pretrained.py index a6c69d54f..c8aae04e8 100755 --- a/egs/librispeech/ASR/conv_emformer_transducer_stateless2/onnx_pretrained.py +++ b/egs/librispeech/ASR/conv_emformer_transducer_stateless2/onnx_pretrained.py @@ -326,6 +326,7 @@ def create_streaming_feature_extractor() -> OnlineFeature: opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = 16000 opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 return OnlineFbank(opts) diff --git a/egs/librispeech/ASR/conv_emformer_transducer_stateless2/streaming-ncnn-decode.py b/egs/librispeech/ASR/conv_emformer_transducer_stateless2/streaming-ncnn-decode.py index 74da9e6c8..1047100fc 100755 --- a/egs/librispeech/ASR/conv_emformer_transducer_stateless2/streaming-ncnn-decode.py +++ b/egs/librispeech/ASR/conv_emformer_transducer_stateless2/streaming-ncnn-decode.py @@ -276,6 +276,7 @@ def create_streaming_feature_extractor() -> OnlineFeature: opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = 16000 opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 return OnlineFbank(opts) diff --git a/egs/librispeech/ASR/conv_emformer_transducer_stateless2/streaming_decode.py b/egs/librispeech/ASR/conv_emformer_transducer_stateless2/streaming_decode.py index f5d894a7b..aaed7d31f 100755 --- a/egs/librispeech/ASR/conv_emformer_transducer_stateless2/streaming_decode.py +++ b/egs/librispeech/ASR/conv_emformer_transducer_stateless2/streaming_decode.py @@ -623,6 +623,7 @@ def create_streaming_feature_extractor() -> Fbank: opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = 16000 opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 return Fbank(opts) diff --git a/egs/librispeech/ASR/lstm_transducer_stateless/jit_pretrained.py b/egs/librispeech/ASR/lstm_transducer_stateless/jit_pretrained.py index c07956243..5350a54da 100755 --- a/egs/librispeech/ASR/lstm_transducer_stateless/jit_pretrained.py +++ b/egs/librispeech/ASR/lstm_transducer_stateless/jit_pretrained.py @@ -266,6 +266,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = args.sample_rate opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/librispeech/ASR/lstm_transducer_stateless/pretrained.py b/egs/librispeech/ASR/lstm_transducer_stateless/pretrained.py index 119fcf1fd..42c3a5d7f 100755 --- a/egs/librispeech/ASR/lstm_transducer_stateless/pretrained.py +++ b/egs/librispeech/ASR/lstm_transducer_stateless/pretrained.py @@ -251,6 +251,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/librispeech/ASR/lstm_transducer_stateless/streaming_decode.py b/egs/librispeech/ASR/lstm_transducer_stateless/streaming_decode.py index f989d9bc0..03472e2c3 100755 --- a/egs/librispeech/ASR/lstm_transducer_stateless/streaming_decode.py +++ b/egs/librispeech/ASR/lstm_transducer_stateless/streaming_decode.py @@ -615,6 +615,7 @@ def create_streaming_feature_extractor() -> Fbank: opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = 16000 opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 return Fbank(opts) diff --git a/egs/librispeech/ASR/lstm_transducer_stateless2/jit_pretrained.py b/egs/librispeech/ASR/lstm_transducer_stateless2/jit_pretrained.py index 728b09104..f4ec17221 100755 --- a/egs/librispeech/ASR/lstm_transducer_stateless2/jit_pretrained.py +++ b/egs/librispeech/ASR/lstm_transducer_stateless2/jit_pretrained.py @@ -267,6 +267,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = args.sample_rate opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/librispeech/ASR/lstm_transducer_stateless2/ncnn-decode.py b/egs/librispeech/ASR/lstm_transducer_stateless2/ncnn-decode.py index 3eeaa5397..5bab70fb0 100755 --- a/egs/librispeech/ASR/lstm_transducer_stateless2/ncnn-decode.py +++ b/egs/librispeech/ASR/lstm_transducer_stateless2/ncnn-decode.py @@ -255,6 +255,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = sample_rate opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/librispeech/ASR/lstm_transducer_stateless2/onnx_pretrained.py b/egs/librispeech/ASR/lstm_transducer_stateless2/onnx_pretrained.py index 06159e56a..06397965d 100755 --- a/egs/librispeech/ASR/lstm_transducer_stateless2/onnx_pretrained.py +++ b/egs/librispeech/ASR/lstm_transducer_stateless2/onnx_pretrained.py @@ -298,6 +298,7 @@ def create_streaming_feature_extractor() -> OnlineFeature: opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = 16000 opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 return OnlineFbank(opts) diff --git a/egs/librispeech/ASR/lstm_transducer_stateless2/pretrained.py b/egs/librispeech/ASR/lstm_transducer_stateless2/pretrained.py index 5d6d97320..dcff088e2 100755 --- a/egs/librispeech/ASR/lstm_transducer_stateless2/pretrained.py +++ b/egs/librispeech/ASR/lstm_transducer_stateless2/pretrained.py @@ -254,6 +254,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/librispeech/ASR/lstm_transducer_stateless2/streaming-ncnn-decode.py b/egs/librispeech/ASR/lstm_transducer_stateless2/streaming-ncnn-decode.py index cbbc77928..6166049ae 100755 --- a/egs/librispeech/ASR/lstm_transducer_stateless2/streaming-ncnn-decode.py +++ b/egs/librispeech/ASR/lstm_transducer_stateless2/streaming-ncnn-decode.py @@ -217,6 +217,7 @@ def create_streaming_feature_extractor() -> OnlineFeature: opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = 16000 opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 return OnlineFbank(opts) diff --git a/egs/librispeech/ASR/lstm_transducer_stateless2/streaming-onnx-decode.py b/egs/librispeech/ASR/lstm_transducer_stateless2/streaming-onnx-decode.py index 487fc2114..df9f6cf3f 100755 --- a/egs/librispeech/ASR/lstm_transducer_stateless2/streaming-onnx-decode.py +++ b/egs/librispeech/ASR/lstm_transducer_stateless2/streaming-onnx-decode.py @@ -344,6 +344,7 @@ def create_streaming_feature_extractor() -> OnlineFeature: opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = 16000 opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 return OnlineFbank(opts) diff --git a/egs/librispeech/ASR/lstm_transducer_stateless3/jit_pretrained.py b/egs/librispeech/ASR/lstm_transducer_stateless3/jit_pretrained.py index 237591a36..d9e7f3578 100755 --- a/egs/librispeech/ASR/lstm_transducer_stateless3/jit_pretrained.py +++ b/egs/librispeech/ASR/lstm_transducer_stateless3/jit_pretrained.py @@ -266,6 +266,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = args.sample_rate opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/librispeech/ASR/lstm_transducer_stateless3/pretrained.py b/egs/librispeech/ASR/lstm_transducer_stateless3/pretrained.py index 29a0d4d1a..e39637bd8 100755 --- a/egs/librispeech/ASR/lstm_transducer_stateless3/pretrained.py +++ b/egs/librispeech/ASR/lstm_transducer_stateless3/pretrained.py @@ -252,6 +252,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/librispeech/ASR/lstm_transducer_stateless3/streaming_decode.py b/egs/librispeech/ASR/lstm_transducer_stateless3/streaming_decode.py index c737e3611..c425b1f46 100755 --- a/egs/librispeech/ASR/lstm_transducer_stateless3/streaming_decode.py +++ b/egs/librispeech/ASR/lstm_transducer_stateless3/streaming_decode.py @@ -615,6 +615,7 @@ def create_streaming_feature_extractor() -> Fbank: opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = 16000 opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 return Fbank(opts) diff --git a/egs/librispeech/ASR/pruned_transducer_stateless/pretrained.py b/egs/librispeech/ASR/pruned_transducer_stateless/pretrained.py index 02f9f1b03..e06404619 100755 --- a/egs/librispeech/ASR/pruned_transducer_stateless/pretrained.py +++ b/egs/librispeech/ASR/pruned_transducer_stateless/pretrained.py @@ -277,6 +277,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/librispeech/ASR/pruned_transducer_stateless/streaming_decode.py b/egs/librispeech/ASR/pruned_transducer_stateless/streaming_decode.py index f4b01fd06..8586c66d6 100755 --- a/egs/librispeech/ASR/pruned_transducer_stateless/streaming_decode.py +++ b/egs/librispeech/ASR/pruned_transducer_stateless/streaming_decode.py @@ -334,6 +334,7 @@ def decode_dataset( opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = 16000 opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 log_interval = 100 diff --git a/egs/librispeech/ASR/pruned_transducer_stateless2/pretrained.py b/egs/librispeech/ASR/pruned_transducer_stateless2/pretrained.py index 029f55ba0..6923f4d40 100755 --- a/egs/librispeech/ASR/pruned_transducer_stateless2/pretrained.py +++ b/egs/librispeech/ASR/pruned_transducer_stateless2/pretrained.py @@ -278,6 +278,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/librispeech/ASR/pruned_transducer_stateless2/streaming_decode.py b/egs/librispeech/ASR/pruned_transducer_stateless2/streaming_decode.py index 9c4a13606..d17c3467a 100755 --- a/egs/librispeech/ASR/pruned_transducer_stateless2/streaming_decode.py +++ b/egs/librispeech/ASR/pruned_transducer_stateless2/streaming_decode.py @@ -336,6 +336,7 @@ def decode_dataset( opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = 16000 opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 log_interval = 50 diff --git a/egs/librispeech/ASR/pruned_transducer_stateless3/jit_pretrained.py b/egs/librispeech/ASR/pruned_transducer_stateless3/jit_pretrained.py index 0669284b3..6d09de6bd 100755 --- a/egs/librispeech/ASR/pruned_transducer_stateless3/jit_pretrained.py +++ b/egs/librispeech/ASR/pruned_transducer_stateless3/jit_pretrained.py @@ -285,6 +285,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = args.sample_rate opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/librispeech/ASR/pruned_transducer_stateless3/onnx_pretrained.py b/egs/librispeech/ASR/pruned_transducer_stateless3/onnx_pretrained.py index de3e03da6..8d12eae28 100755 --- a/egs/librispeech/ASR/pruned_transducer_stateless3/onnx_pretrained.py +++ b/egs/librispeech/ASR/pruned_transducer_stateless3/onnx_pretrained.py @@ -368,6 +368,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = args.sample_rate opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/librispeech/ASR/pruned_transducer_stateless3/pretrained.py b/egs/librispeech/ASR/pruned_transducer_stateless3/pretrained.py index abda4e2d4..05e6a6fba 100755 --- a/egs/librispeech/ASR/pruned_transducer_stateless3/pretrained.py +++ b/egs/librispeech/ASR/pruned_transducer_stateless3/pretrained.py @@ -287,6 +287,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/librispeech/ASR/pruned_transducer_stateless3/streaming_decode.py b/egs/librispeech/ASR/pruned_transducer_stateless3/streaming_decode.py index e7c1affc2..5e1acd735 100755 --- a/egs/librispeech/ASR/pruned_transducer_stateless3/streaming_decode.py +++ b/egs/librispeech/ASR/pruned_transducer_stateless3/streaming_decode.py @@ -337,6 +337,7 @@ def decode_dataset( opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = 16000 opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 log_interval = 50 diff --git a/egs/librispeech/ASR/pruned_transducer_stateless4/streaming_decode.py b/egs/librispeech/ASR/pruned_transducer_stateless4/streaming_decode.py index e966aa4b1..229b52e5b 100755 --- a/egs/librispeech/ASR/pruned_transducer_stateless4/streaming_decode.py +++ b/egs/librispeech/ASR/pruned_transducer_stateless4/streaming_decode.py @@ -353,6 +353,7 @@ def decode_dataset( opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = 16000 opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 log_interval = 50 diff --git a/egs/librispeech/ASR/pruned_transducer_stateless5/onnx_pretrained-streaming.py b/egs/librispeech/ASR/pruned_transducer_stateless5/onnx_pretrained-streaming.py index 6e290e799..2432c6010 100755 --- a/egs/librispeech/ASR/pruned_transducer_stateless5/onnx_pretrained-streaming.py +++ b/egs/librispeech/ASR/pruned_transducer_stateless5/onnx_pretrained-streaming.py @@ -326,6 +326,7 @@ def create_streaming_feature_extractor() -> OnlineFeature: opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = 16000 opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 return OnlineFbank(opts) diff --git a/egs/librispeech/ASR/pruned_transducer_stateless5/pretrained.py b/egs/librispeech/ASR/pruned_transducer_stateless5/pretrained.py index 304fa8693..a9ce75a7b 100755 --- a/egs/librispeech/ASR/pruned_transducer_stateless5/pretrained.py +++ b/egs/librispeech/ASR/pruned_transducer_stateless5/pretrained.py @@ -251,6 +251,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/librispeech/ASR/pruned_transducer_stateless5/streaming_decode.py b/egs/librispeech/ASR/pruned_transducer_stateless5/streaming_decode.py index f65f47fc2..8478a65fb 100755 --- a/egs/librispeech/ASR/pruned_transducer_stateless5/streaming_decode.py +++ b/egs/librispeech/ASR/pruned_transducer_stateless5/streaming_decode.py @@ -353,6 +353,7 @@ def decode_dataset( opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = 16000 opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 log_interval = 50 diff --git a/egs/librispeech/ASR/pruned_transducer_stateless7/jit_pretrained.py b/egs/librispeech/ASR/pruned_transducer_stateless7/jit_pretrained.py index 5af6dae25..88a05e09d 100755 --- a/egs/librispeech/ASR/pruned_transducer_stateless7/jit_pretrained.py +++ b/egs/librispeech/ASR/pruned_transducer_stateless7/jit_pretrained.py @@ -225,6 +225,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = 16000 opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/librispeech/ASR/pruned_transducer_stateless7/pretrained.py b/egs/librispeech/ASR/pruned_transducer_stateless7/pretrained.py index 86c922cda..4bf11ac24 100755 --- a/egs/librispeech/ASR/pruned_transducer_stateless7/pretrained.py +++ b/egs/librispeech/ASR/pruned_transducer_stateless7/pretrained.py @@ -260,6 +260,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/librispeech/ASR/pruned_transducer_stateless7_ctc/jit_pretrained.py b/egs/librispeech/ASR/pruned_transducer_stateless7_ctc/jit_pretrained.py index 280b95984..83dc29324 100755 --- a/egs/librispeech/ASR/pruned_transducer_stateless7_ctc/jit_pretrained.py +++ b/egs/librispeech/ASR/pruned_transducer_stateless7_ctc/jit_pretrained.py @@ -224,6 +224,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = 16000 opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/librispeech/ASR/pruned_transducer_stateless7_ctc/jit_pretrained_ctc.py b/egs/librispeech/ASR/pruned_transducer_stateless7_ctc/jit_pretrained_ctc.py index d50d231d5..d1b7eec65 100755 --- a/egs/librispeech/ASR/pruned_transducer_stateless7_ctc/jit_pretrained_ctc.py +++ b/egs/librispeech/ASR/pruned_transducer_stateless7_ctc/jit_pretrained_ctc.py @@ -280,6 +280,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/librispeech/ASR/pruned_transducer_stateless7_ctc/pretrained.py b/egs/librispeech/ASR/pruned_transducer_stateless7_ctc/pretrained.py index 78e0fa778..323ba2642 100755 --- a/egs/librispeech/ASR/pruned_transducer_stateless7_ctc/pretrained.py +++ b/egs/librispeech/ASR/pruned_transducer_stateless7_ctc/pretrained.py @@ -260,6 +260,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/librispeech/ASR/pruned_transducer_stateless7_ctc/pretrained_ctc.py b/egs/librispeech/ASR/pruned_transducer_stateless7_ctc/pretrained_ctc.py index 904c1deae..1e638aa7d 100755 --- a/egs/librispeech/ASR/pruned_transducer_stateless7_ctc/pretrained_ctc.py +++ b/egs/librispeech/ASR/pruned_transducer_stateless7_ctc/pretrained_ctc.py @@ -298,6 +298,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/librispeech/ASR/pruned_transducer_stateless7_ctc_bs/jit_pretrained.py b/egs/librispeech/ASR/pruned_transducer_stateless7_ctc_bs/jit_pretrained.py index da2c6a39a..a39fdee54 100755 --- a/egs/librispeech/ASR/pruned_transducer_stateless7_ctc_bs/jit_pretrained.py +++ b/egs/librispeech/ASR/pruned_transducer_stateless7_ctc_bs/jit_pretrained.py @@ -224,6 +224,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = 16000 opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/librispeech/ASR/pruned_transducer_stateless7_ctc_bs/jit_pretrained_ctc.py b/egs/librispeech/ASR/pruned_transducer_stateless7_ctc_bs/jit_pretrained_ctc.py index 653c25e06..80604ef4a 100755 --- a/egs/librispeech/ASR/pruned_transducer_stateless7_ctc_bs/jit_pretrained_ctc.py +++ b/egs/librispeech/ASR/pruned_transducer_stateless7_ctc_bs/jit_pretrained_ctc.py @@ -280,6 +280,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/librispeech/ASR/pruned_transducer_stateless7_ctc_bs/onnx_pretrained.py b/egs/librispeech/ASR/pruned_transducer_stateless7_ctc_bs/onnx_pretrained.py index 494a34d97..0ff110370 100755 --- a/egs/librispeech/ASR/pruned_transducer_stateless7_ctc_bs/onnx_pretrained.py +++ b/egs/librispeech/ASR/pruned_transducer_stateless7_ctc_bs/onnx_pretrained.py @@ -381,6 +381,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = args.sample_rate opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/librispeech/ASR/pruned_transducer_stateless7_ctc_bs/pretrained.py b/egs/librispeech/ASR/pruned_transducer_stateless7_ctc_bs/pretrained.py index 5d240cf30..a82f3562b 100755 --- a/egs/librispeech/ASR/pruned_transducer_stateless7_ctc_bs/pretrained.py +++ b/egs/librispeech/ASR/pruned_transducer_stateless7_ctc_bs/pretrained.py @@ -260,6 +260,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/librispeech/ASR/pruned_transducer_stateless7_ctc_bs/pretrained_ctc.py b/egs/librispeech/ASR/pruned_transducer_stateless7_ctc_bs/pretrained_ctc.py index 914107526..b98756a54 100755 --- a/egs/librispeech/ASR/pruned_transducer_stateless7_ctc_bs/pretrained_ctc.py +++ b/egs/librispeech/ASR/pruned_transducer_stateless7_ctc_bs/pretrained_ctc.py @@ -298,6 +298,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/librispeech/ASR/pruned_transducer_stateless7_streaming/jit_pretrained.py b/egs/librispeech/ASR/pruned_transducer_stateless7_streaming/jit_pretrained.py index c8301b2da..7116b10fb 100755 --- a/egs/librispeech/ASR/pruned_transducer_stateless7_streaming/jit_pretrained.py +++ b/egs/librispeech/ASR/pruned_transducer_stateless7_streaming/jit_pretrained.py @@ -231,6 +231,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = 16000 opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/librispeech/ASR/pruned_transducer_stateless7_streaming/jit_trace_pretrained.py b/egs/librispeech/ASR/pruned_transducer_stateless7_streaming/jit_trace_pretrained.py index f2ac1914d..d714670cf 100755 --- a/egs/librispeech/ASR/pruned_transducer_stateless7_streaming/jit_trace_pretrained.py +++ b/egs/librispeech/ASR/pruned_transducer_stateless7_streaming/jit_trace_pretrained.py @@ -186,6 +186,7 @@ def create_streaming_feature_extractor(sample_rate) -> OnlineFeature: opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = sample_rate opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 return OnlineFbank(opts) diff --git a/egs/librispeech/ASR/pruned_transducer_stateless7_streaming/onnx_pretrained.py b/egs/librispeech/ASR/pruned_transducer_stateless7_streaming/onnx_pretrained.py index 04861ea37..298d1889b 100755 --- a/egs/librispeech/ASR/pruned_transducer_stateless7_streaming/onnx_pretrained.py +++ b/egs/librispeech/ASR/pruned_transducer_stateless7_streaming/onnx_pretrained.py @@ -382,6 +382,7 @@ def create_streaming_feature_extractor() -> OnlineFeature: opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = 16000 opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 return OnlineFbank(opts) diff --git a/egs/librispeech/ASR/pruned_transducer_stateless7_streaming/pretrained.py b/egs/librispeech/ASR/pruned_transducer_stateless7_streaming/pretrained.py index bc42e8d05..aa2dd17fb 100755 --- a/egs/librispeech/ASR/pruned_transducer_stateless7_streaming/pretrained.py +++ b/egs/librispeech/ASR/pruned_transducer_stateless7_streaming/pretrained.py @@ -260,6 +260,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/librispeech/ASR/pruned_transducer_stateless7_streaming/streaming-ncnn-decode.py b/egs/librispeech/ASR/pruned_transducer_stateless7_streaming/streaming-ncnn-decode.py index 883fdcbdd..999f7e0b4 100755 --- a/egs/librispeech/ASR/pruned_transducer_stateless7_streaming/streaming-ncnn-decode.py +++ b/egs/librispeech/ASR/pruned_transducer_stateless7_streaming/streaming-ncnn-decode.py @@ -335,6 +335,7 @@ def create_streaming_feature_extractor() -> OnlineFeature: opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = 16000 opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 return OnlineFbank(opts) diff --git a/egs/librispeech/ASR/pruned_transducer_stateless7_streaming/streaming_decode.py b/egs/librispeech/ASR/pruned_transducer_stateless7_streaming/streaming_decode.py index a0f54b6e1..e27fb4e63 100755 --- a/egs/librispeech/ASR/pruned_transducer_stateless7_streaming/streaming_decode.py +++ b/egs/librispeech/ASR/pruned_transducer_stateless7_streaming/streaming_decode.py @@ -320,6 +320,7 @@ def decode_dataset( opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = 16000 opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 log_interval = 50 diff --git a/egs/librispeech/ASR/pruned_transducer_stateless8/jit_pretrained.py b/egs/librispeech/ASR/pruned_transducer_stateless8/jit_pretrained.py index 129497d5a..3ce2953c3 100755 --- a/egs/librispeech/ASR/pruned_transducer_stateless8/jit_pretrained.py +++ b/egs/librispeech/ASR/pruned_transducer_stateless8/jit_pretrained.py @@ -225,6 +225,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = 16000 opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/librispeech/ASR/pruned_transducer_stateless8/pretrained.py b/egs/librispeech/ASR/pruned_transducer_stateless8/pretrained.py index 64b38c9d5..c29b8d8c9 100755 --- a/egs/librispeech/ASR/pruned_transducer_stateless8/pretrained.py +++ b/egs/librispeech/ASR/pruned_transducer_stateless8/pretrained.py @@ -260,6 +260,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/librispeech/ASR/tdnn_lstm_ctc/pretrained.py b/egs/librispeech/ASR/tdnn_lstm_ctc/pretrained.py index fde724866..b3dfab64a 100755 --- a/egs/librispeech/ASR/tdnn_lstm_ctc/pretrained.py +++ b/egs/librispeech/ASR/tdnn_lstm_ctc/pretrained.py @@ -196,6 +196,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/librispeech/ASR/tiny_transducer_ctc/jit_pretrained.py b/egs/librispeech/ASR/tiny_transducer_ctc/jit_pretrained.py index 3888d3544..0cd876551 100755 --- a/egs/librispeech/ASR/tiny_transducer_ctc/jit_pretrained.py +++ b/egs/librispeech/ASR/tiny_transducer_ctc/jit_pretrained.py @@ -224,6 +224,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = 16000 opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/librispeech/ASR/tiny_transducer_ctc/jit_pretrained_ctc.py b/egs/librispeech/ASR/tiny_transducer_ctc/jit_pretrained_ctc.py index 6f2cbaabd..92dea3aa1 100755 --- a/egs/librispeech/ASR/tiny_transducer_ctc/jit_pretrained_ctc.py +++ b/egs/librispeech/ASR/tiny_transducer_ctc/jit_pretrained_ctc.py @@ -280,6 +280,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/librispeech/ASR/tiny_transducer_ctc/pretrained.py b/egs/librispeech/ASR/tiny_transducer_ctc/pretrained.py index 981039b8f..5c6956324 100755 --- a/egs/librispeech/ASR/tiny_transducer_ctc/pretrained.py +++ b/egs/librispeech/ASR/tiny_transducer_ctc/pretrained.py @@ -262,6 +262,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/librispeech/ASR/tiny_transducer_ctc/pretrained_ctc.py b/egs/librispeech/ASR/tiny_transducer_ctc/pretrained_ctc.py index a06d6d684..7698ada79 100755 --- a/egs/librispeech/ASR/tiny_transducer_ctc/pretrained_ctc.py +++ b/egs/librispeech/ASR/tiny_transducer_ctc/pretrained_ctc.py @@ -298,6 +298,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/librispeech/ASR/transducer/pretrained.py b/egs/librispeech/ASR/transducer/pretrained.py index c2413f5de..4d9bbf4b1 100755 --- a/egs/librispeech/ASR/transducer/pretrained.py +++ b/egs/librispeech/ASR/transducer/pretrained.py @@ -235,6 +235,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/librispeech/ASR/transducer_stateless/pretrained.py b/egs/librispeech/ASR/transducer_stateless/pretrained.py index 5898dd0f5..3b86e319e 100755 --- a/egs/librispeech/ASR/transducer_stateless/pretrained.py +++ b/egs/librispeech/ASR/transducer_stateless/pretrained.py @@ -247,6 +247,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/librispeech/ASR/transducer_stateless2/pretrained.py b/egs/librispeech/ASR/transducer_stateless2/pretrained.py index b69b347ef..2de4182f1 100755 --- a/egs/librispeech/ASR/transducer_stateless2/pretrained.py +++ b/egs/librispeech/ASR/transducer_stateless2/pretrained.py @@ -247,6 +247,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/librispeech/ASR/transducer_stateless_multi_datasets/pretrained.py b/egs/librispeech/ASR/transducer_stateless_multi_datasets/pretrained.py index 4f29d6f1f..83094ea51 100755 --- a/egs/librispeech/ASR/transducer_stateless_multi_datasets/pretrained.py +++ b/egs/librispeech/ASR/transducer_stateless_multi_datasets/pretrained.py @@ -247,6 +247,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/librispeech/ASR/zipformer/jit_pretrained.py b/egs/librispeech/ASR/zipformer/jit_pretrained.py index a41fbc1c9..52dfd3fb6 100755 --- a/egs/librispeech/ASR/zipformer/jit_pretrained.py +++ b/egs/librispeech/ASR/zipformer/jit_pretrained.py @@ -222,6 +222,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = 16000 opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/librispeech/ASR/zipformer/jit_pretrained_ctc.py b/egs/librispeech/ASR/zipformer/jit_pretrained_ctc.py index 660a4bfc6..fcd07ae34 100755 --- a/egs/librispeech/ASR/zipformer/jit_pretrained_ctc.py +++ b/egs/librispeech/ASR/zipformer/jit_pretrained_ctc.py @@ -285,6 +285,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/librispeech/ASR/zipformer/jit_pretrained_streaming.py b/egs/librispeech/ASR/zipformer/jit_pretrained_streaming.py index d4ceacefd..eade5a854 100755 --- a/egs/librispeech/ASR/zipformer/jit_pretrained_streaming.py +++ b/egs/librispeech/ASR/zipformer/jit_pretrained_streaming.py @@ -167,6 +167,7 @@ def create_streaming_feature_extractor(sample_rate) -> OnlineFeature: opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = sample_rate opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 return OnlineFbank(opts) diff --git a/egs/librispeech/ASR/zipformer/onnx_pretrained-streaming-ctc.py b/egs/librispeech/ASR/zipformer/onnx_pretrained-streaming-ctc.py index 44546cae5..dd47c0eb6 100755 --- a/egs/librispeech/ASR/zipformer/onnx_pretrained-streaming-ctc.py +++ b/egs/librispeech/ASR/zipformer/onnx_pretrained-streaming-ctc.py @@ -318,6 +318,7 @@ def create_streaming_feature_extractor() -> OnlineFeature: opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = 16000 opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 return OnlineFbank(opts) diff --git a/egs/librispeech/ASR/zipformer/onnx_pretrained-streaming.py b/egs/librispeech/ASR/zipformer/onnx_pretrained-streaming.py index e7c4f40ee..e011c4b24 100755 --- a/egs/librispeech/ASR/zipformer/onnx_pretrained-streaming.py +++ b/egs/librispeech/ASR/zipformer/onnx_pretrained-streaming.py @@ -413,6 +413,7 @@ def create_streaming_feature_extractor() -> OnlineFeature: opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = 16000 opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 return OnlineFbank(opts) diff --git a/egs/librispeech/ASR/zipformer/onnx_pretrained.py b/egs/librispeech/ASR/zipformer/onnx_pretrained.py index 334376093..662392b5f 100755 --- a/egs/librispeech/ASR/zipformer/onnx_pretrained.py +++ b/egs/librispeech/ASR/zipformer/onnx_pretrained.py @@ -369,6 +369,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = args.sample_rate opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/librispeech/ASR/zipformer/onnx_pretrained_ctc.py b/egs/librispeech/ASR/zipformer/onnx_pretrained_ctc.py index eb5cee9cd..ecca758f2 100755 --- a/egs/librispeech/ASR/zipformer/onnx_pretrained_ctc.py +++ b/egs/librispeech/ASR/zipformer/onnx_pretrained_ctc.py @@ -161,6 +161,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = args.sample_rate opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/librispeech/ASR/zipformer/onnx_pretrained_ctc_H.py b/egs/librispeech/ASR/zipformer/onnx_pretrained_ctc_H.py index 683a7dc20..a77c3bf2a 100755 --- a/egs/librispeech/ASR/zipformer/onnx_pretrained_ctc_H.py +++ b/egs/librispeech/ASR/zipformer/onnx_pretrained_ctc_H.py @@ -225,6 +225,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = args.sample_rate opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 logging.info(f"Loading H from {args.H}") H = kaldifst.StdVectorFst.read(args.H) diff --git a/egs/librispeech/ASR/zipformer/onnx_pretrained_ctc_HL.py b/egs/librispeech/ASR/zipformer/onnx_pretrained_ctc_HL.py index 0b94bfa65..6ef944514 100755 --- a/egs/librispeech/ASR/zipformer/onnx_pretrained_ctc_HL.py +++ b/egs/librispeech/ASR/zipformer/onnx_pretrained_ctc_HL.py @@ -223,6 +223,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = args.sample_rate opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 logging.info(f"Loading HL from {args.HL}") HL = kaldifst.StdVectorFst.read(args.HL) diff --git a/egs/librispeech/ASR/zipformer/onnx_pretrained_ctc_HLG.py b/egs/librispeech/ASR/zipformer/onnx_pretrained_ctc_HLG.py index 93569142a..ccb3107ea 100755 --- a/egs/librispeech/ASR/zipformer/onnx_pretrained_ctc_HLG.py +++ b/egs/librispeech/ASR/zipformer/onnx_pretrained_ctc_HLG.py @@ -223,6 +223,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = args.sample_rate opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 logging.info(f"Loading HLG from {args.HLG}") HLG = kaldifst.StdVectorFst.read(args.HLG) diff --git a/egs/librispeech/ASR/zipformer/pretrained.py b/egs/librispeech/ASR/zipformer/pretrained.py index 3104b6084..de0652893 100755 --- a/egs/librispeech/ASR/zipformer/pretrained.py +++ b/egs/librispeech/ASR/zipformer/pretrained.py @@ -303,6 +303,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/librispeech/ASR/zipformer/pretrained_ctc.py b/egs/librispeech/ASR/zipformer/pretrained_ctc.py index 9dff2e6fc..408d13576 100755 --- a/egs/librispeech/ASR/zipformer/pretrained_ctc.py +++ b/egs/librispeech/ASR/zipformer/pretrained_ctc.py @@ -304,6 +304,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/librispeech/ASR/zipformer_mmi/jit_pretrained.py b/egs/librispeech/ASR/zipformer_mmi/jit_pretrained.py index c9ef16ffa..6990c90a0 100755 --- a/egs/librispeech/ASR/zipformer_mmi/jit_pretrained.py +++ b/egs/librispeech/ASR/zipformer_mmi/jit_pretrained.py @@ -259,6 +259,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = 16000 opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/librispeech/ASR/zipformer_mmi/pretrained.py b/egs/librispeech/ASR/zipformer_mmi/pretrained.py index 3ba4da5dd..1e7afc777 100755 --- a/egs/librispeech/ASR/zipformer_mmi/pretrained.py +++ b/egs/librispeech/ASR/zipformer_mmi/pretrained.py @@ -282,6 +282,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/mgb2/ASR/conformer_ctc/pretrained.py b/egs/mgb2/ASR/conformer_ctc/pretrained.py index d30ca98d8..0ab2af527 100755 --- a/egs/mgb2/ASR/conformer_ctc/pretrained.py +++ b/egs/mgb2/ASR/conformer_ctc/pretrained.py @@ -287,6 +287,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/mgb2/ASR/pruned_transducer_stateless5/pretrained.py b/egs/mgb2/ASR/pruned_transducer_stateless5/pretrained.py index 77ba0873b..81a16f0ff 100755 --- a/egs/mgb2/ASR/pruned_transducer_stateless5/pretrained.py +++ b/egs/mgb2/ASR/pruned_transducer_stateless5/pretrained.py @@ -249,6 +249,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/multi_zh-hans/ASR/zipformer/pretrained.py b/egs/multi_zh-hans/ASR/zipformer/pretrained.py index 69ff382da..c15db11f7 100755 --- a/egs/multi_zh-hans/ASR/zipformer/pretrained.py +++ b/egs/multi_zh-hans/ASR/zipformer/pretrained.py @@ -303,6 +303,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/multi_zh_en/ASR/zipformer/pretrained.py b/egs/multi_zh_en/ASR/zipformer/pretrained.py index 676272e1f..2fcde550b 100755 --- a/egs/multi_zh_en/ASR/zipformer/pretrained.py +++ b/egs/multi_zh_en/ASR/zipformer/pretrained.py @@ -306,6 +306,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/tal_csasr/ASR/pruned_transducer_stateless5/pretrained.py b/egs/tal_csasr/ASR/pruned_transducer_stateless5/pretrained.py index 3305f5bd3..8a74ee745 100755 --- a/egs/tal_csasr/ASR/pruned_transducer_stateless5/pretrained.py +++ b/egs/tal_csasr/ASR/pruned_transducer_stateless5/pretrained.py @@ -248,6 +248,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/tal_csasr/ASR/pruned_transducer_stateless7_bbpe/jit_pretrained.py b/egs/tal_csasr/ASR/pruned_transducer_stateless7_bbpe/jit_pretrained.py index a23e2a04f..8c966a2f6 100755 --- a/egs/tal_csasr/ASR/pruned_transducer_stateless7_bbpe/jit_pretrained.py +++ b/egs/tal_csasr/ASR/pruned_transducer_stateless7_bbpe/jit_pretrained.py @@ -226,6 +226,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = 16000 opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/tal_csasr/ASR/pruned_transducer_stateless7_bbpe/pretrained.py b/egs/tal_csasr/ASR/pruned_transducer_stateless7_bbpe/pretrained.py index f365986f6..6e07b5949 100755 --- a/egs/tal_csasr/ASR/pruned_transducer_stateless7_bbpe/pretrained.py +++ b/egs/tal_csasr/ASR/pruned_transducer_stateless7_bbpe/pretrained.py @@ -261,6 +261,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/tedlium3/ASR/pruned_transducer_stateless/pretrained.py b/egs/tedlium3/ASR/pruned_transducer_stateless/pretrained.py index 8a89c3578..9e58fed00 100644 --- a/egs/tedlium3/ASR/pruned_transducer_stateless/pretrained.py +++ b/egs/tedlium3/ASR/pruned_transducer_stateless/pretrained.py @@ -256,6 +256,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/tedlium3/ASR/transducer_stateless/pretrained.py b/egs/tedlium3/ASR/transducer_stateless/pretrained.py index 81afd6a4e..5300fe764 100644 --- a/egs/tedlium3/ASR/transducer_stateless/pretrained.py +++ b/egs/tedlium3/ASR/transducer_stateless/pretrained.py @@ -270,6 +270,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/timit/ASR/tdnn_ligru_ctc/pretrained.py b/egs/timit/ASR/tdnn_ligru_ctc/pretrained.py index 3fdf3b855..0d77bc512 100644 --- a/egs/timit/ASR/tdnn_ligru_ctc/pretrained.py +++ b/egs/timit/ASR/tdnn_ligru_ctc/pretrained.py @@ -196,6 +196,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/timit/ASR/tdnn_lstm_ctc/pretrained.py b/egs/timit/ASR/tdnn_lstm_ctc/pretrained.py index 98c746ce5..f06c8c211 100644 --- a/egs/timit/ASR/tdnn_lstm_ctc/pretrained.py +++ b/egs/timit/ASR/tdnn_lstm_ctc/pretrained.py @@ -196,6 +196,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/wenetspeech/ASR/pruned_transducer_stateless2/jit_pretrained.py b/egs/wenetspeech/ASR/pruned_transducer_stateless2/jit_pretrained.py index f90dd2b43..aee1a2175 100755 --- a/egs/wenetspeech/ASR/pruned_transducer_stateless2/jit_pretrained.py +++ b/egs/wenetspeech/ASR/pruned_transducer_stateless2/jit_pretrained.py @@ -285,6 +285,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = args.sample_rate opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/wenetspeech/ASR/pruned_transducer_stateless2/pretrained.py b/egs/wenetspeech/ASR/pruned_transducer_stateless2/pretrained.py index c3d67ad92..642de72d7 100755 --- a/egs/wenetspeech/ASR/pruned_transducer_stateless2/pretrained.py +++ b/egs/wenetspeech/ASR/pruned_transducer_stateless2/pretrained.py @@ -238,6 +238,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/wenetspeech/ASR/pruned_transducer_stateless5/onnx_pretrained-streaming.py b/egs/wenetspeech/ASR/pruned_transducer_stateless5/onnx_pretrained-streaming.py index c31db6859..cca26feb0 100755 --- a/egs/wenetspeech/ASR/pruned_transducer_stateless5/onnx_pretrained-streaming.py +++ b/egs/wenetspeech/ASR/pruned_transducer_stateless5/onnx_pretrained-streaming.py @@ -327,6 +327,7 @@ def create_streaming_feature_extractor() -> OnlineFeature: opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = 16000 opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 return OnlineFbank(opts) diff --git a/egs/wenetspeech/ASR/pruned_transducer_stateless5/onnx_pretrained.py b/egs/wenetspeech/ASR/pruned_transducer_stateless5/onnx_pretrained.py index c784853ee..4b4ddd332 100755 --- a/egs/wenetspeech/ASR/pruned_transducer_stateless5/onnx_pretrained.py +++ b/egs/wenetspeech/ASR/pruned_transducer_stateless5/onnx_pretrained.py @@ -376,6 +376,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = args.sample_rate opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/wenetspeech/ASR/pruned_transducer_stateless5/pretrained.py b/egs/wenetspeech/ASR/pruned_transducer_stateless5/pretrained.py index 1cac20435..17428e19d 100644 --- a/egs/wenetspeech/ASR/pruned_transducer_stateless5/pretrained.py +++ b/egs/wenetspeech/ASR/pruned_transducer_stateless5/pretrained.py @@ -238,6 +238,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/wenetspeech/ASR/pruned_transducer_stateless5/streaming_decode.py b/egs/wenetspeech/ASR/pruned_transducer_stateless5/streaming_decode.py index 3a4dc3cb8..27a9b1714 100644 --- a/egs/wenetspeech/ASR/pruned_transducer_stateless5/streaming_decode.py +++ b/egs/wenetspeech/ASR/pruned_transducer_stateless5/streaming_decode.py @@ -378,6 +378,7 @@ def decode_dataset( opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = 16000 opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 log_interval = 100 diff --git a/egs/wenetspeech/ASR/zipformer/streaming_decode.py b/egs/wenetspeech/ASR/zipformer/streaming_decode.py index 94c5fae5f..96f339b07 100755 --- a/egs/wenetspeech/ASR/zipformer/streaming_decode.py +++ b/egs/wenetspeech/ASR/zipformer/streaming_decode.py @@ -572,6 +572,7 @@ def decode_dataset( opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = 16000 opts.mel_opts.num_bins = 80 + opts.mel_opts.high_freq = -400 log_interval = 100 diff --git a/egs/xbmu_amdo31/ASR/pruned_transducer_stateless5/pretrained.py b/egs/xbmu_amdo31/ASR/pruned_transducer_stateless5/pretrained.py index 74a2210c3..2c106c4cb 100755 --- a/egs/xbmu_amdo31/ASR/pruned_transducer_stateless5/pretrained.py +++ b/egs/xbmu_amdo31/ASR/pruned_transducer_stateless5/pretrained.py @@ -249,6 +249,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/xbmu_amdo31/ASR/pruned_transducer_stateless7/pretrained.py b/egs/xbmu_amdo31/ASR/pruned_transducer_stateless7/pretrained.py index d05bafcfb..6995ff2ff 100755 --- a/egs/xbmu_amdo31/ASR/pruned_transducer_stateless7/pretrained.py +++ b/egs/xbmu_amdo31/ASR/pruned_transducer_stateless7/pretrained.py @@ -260,6 +260,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/yesno/ASR/tdnn/jit_pretrained.py b/egs/yesno/ASR/tdnn/jit_pretrained.py index 7581ecb83..e29415ffb 100755 --- a/egs/yesno/ASR/tdnn/jit_pretrained.py +++ b/egs/yesno/ASR/tdnn/jit_pretrained.py @@ -142,6 +142,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/yesno/ASR/tdnn/jit_pretrained_decode_with_H.py b/egs/yesno/ASR/tdnn/jit_pretrained_decode_with_H.py index ff8c742af..72127aebd 100755 --- a/egs/yesno/ASR/tdnn/jit_pretrained_decode_with_H.py +++ b/egs/yesno/ASR/tdnn/jit_pretrained_decode_with_H.py @@ -164,6 +164,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = sample_rate opts.mel_opts.num_bins = 23 + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/yesno/ASR/tdnn/jit_pretrained_decode_with_HL.py b/egs/yesno/ASR/tdnn/jit_pretrained_decode_with_HL.py index 05ba74f9a..f8a057336 100755 --- a/egs/yesno/ASR/tdnn/jit_pretrained_decode_with_HL.py +++ b/egs/yesno/ASR/tdnn/jit_pretrained_decode_with_HL.py @@ -163,6 +163,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = sample_rate opts.mel_opts.num_bins = 23 + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/yesno/ASR/tdnn/onnx_pretrained.py b/egs/yesno/ASR/tdnn/onnx_pretrained.py index 72a1d69c8..968a9e9a8 100755 --- a/egs/yesno/ASR/tdnn/onnx_pretrained.py +++ b/egs/yesno/ASR/tdnn/onnx_pretrained.py @@ -186,6 +186,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts) diff --git a/egs/yesno/ASR/tdnn/pretrained.py b/egs/yesno/ASR/tdnn/pretrained.py index 987c49de6..bea520998 100755 --- a/egs/yesno/ASR/tdnn/pretrained.py +++ b/egs/yesno/ASR/tdnn/pretrained.py @@ -164,6 +164,7 @@ def main(): opts.frame_opts.snip_edges = False opts.frame_opts.samp_freq = params.sample_rate opts.mel_opts.num_bins = params.feature_dim + opts.mel_opts.high_freq = -400 fbank = kaldifeat.Fbank(opts)