Support CTC decoding on CPU using OpenFst and kaldi decoders.

[csukuangfj]

No description provided.

[csukuangfj]

Please see the CI logs for usages.

https://github.com/k2-fsa/icefall/actions/runs/6110904739/job/16584973570?pr=1244#step:10:21

python3 ./tdnn/jit_pretrained_decode_with_H.py \
      --nn-model ./tdnn/exp/cpu_jit.pt \
      --H ./data/lang_phone/H.fst \
      --tokens ./data/lang_phone/tokens.txt \
      ./download/waves_yesno/0_0_0_1_0_0_0_1.wav \
      ./download/waves_yesno/0_0_1_0_0_0_1_0.wav \
      ./download/waves_yesno/0_0_1_0_0_1_1_1.wav
  shell: /usr/bin/bash --noprofile --norc -e -o pipefail {0}
  env:
    pythonLocation: /opt/hostedtoolcache/Python/3.8.18/x64
    LD_LIBRARY_PATH: /opt/hostedtoolcache/Python/3.8.18/x64/lib
/home/runner/work/icefall/icefall:
2023-0[9](https://github.com/k2-fsa/icefall/actions/runs/6110904739/job/16584973570?pr=1244#step:10:9)-07 14:17:41,615 INFO [export.py:76] {'exp_dir': PosixPath('tdnn/exp'), 'lang_dir': PosixPath('data/lang_phone'), 'lr': 0.01, 'feature_dim': 23, 'weight_decay': 1e-06, 'start_epoch': 0, 'best_train_loss': inf, 'best_valid_loss': inf, 'best_train_epoch': -1, 'best_valid_epoch': -1, 'batch_idx_train': 0, 'log_interval': [10](https://github.com/k2-fsa/icefall/actions/runs/6110904739/job/16584973570?pr=1244#step:10:10), 'reset_interval': 20, 'valid_interval': 10, 'beam_size': 10, 'reduction': 'sum', 'use_double_scores': True, 'epoch': 14, 'avg': 2, 'jit': True}
2023-09-07 14:17:41,616 INFO [lexicon.py:168] Loading pre-compiled data/lang_phone/Linv.pt
2023-09-07 14:17:41,618 INFO [export.py:93] averaging ['tdnn/exp/epoch-[13](https://github.com/k2-fsa/icefall/actions/runs/6110904739/job/16584973570?pr=1244#step:10:13).pt', 'tdnn/exp/epoch-[14](https://github.com/k2-fsa/icefall/actions/runs/6110904739/job/16584973570?pr=1244#step:10:14).pt']
2023-09-07 14:17:41,622 INFO [export.py:100] Using torch.jit.script
2023-09-07 14:17:41,685 INFO [export.py:104] Saved to tdnn/exp/cpu_jit.pt
2023-09-07 14:17:43,340 INFO [jit_pretrained_decode_with_H.py:[15](https://github.com/k2-fsa/icefall/actions/runs/6110904739/job/16584973570?pr=1244#step:10:15)0] device: cpu
2023-09-07 14:17:43,340 INFO [jit_pretrained_decode_with_H.py:152] Loading torchscript model
2023-09-07 14:17:43,363 INFO [jit_pretrained_decode_with_H.py:157] Loading H from ./data/lang_phone/H.fst
2023-09-07 14:17:43,363 INFO [jit_pretrained_decode_with_H.py:[16](https://github.com/k2-fsa/icefall/actions/runs/6110904739/job/16584973570?pr=1244#step:10:16)2] Constructing Fbank computer
2023-09-07 14:[17](https://github.com/k2-fsa/icefall/actions/runs/6110904739/job/16584973570?pr=1244#step:10:17):43,363 INFO [jit_pretrained_decode_with_H.py:172] Reading sound files: ['./download/waves_yesno/0_0_0_1_0_0_0_1.wav', './download/waves_yesno/0_0_1_0_0_0_1_0.wav', './download/waves_yesno/0_0_1_0_0_1_1_1.wav']
[20](https://github.com/k2-fsa/icefall/actions/runs/6110904739/job/16584973570?pr=1244#step:10:21)[23](https://github.com/k2-fsa/icefall/actions/runs/6110904739/job/16584973570?pr=1244#step:10:24)-09-07 14:17:43,365 INFO [jit_pretrained_decode_with_H.py:178] Decoding started
2023-09-07 14:17:43,613 INFO [jit_pretrained_decode_with_H.py:201] 
./download/waves_yesno/0_0_0_1_0_0_0_1.wav:
N N N Y N N N Y

./download/waves_yesno/0_0_1_0_0_0_1_0.wav:
N N Y N N N Y N

./download/waves_yesno/0_0_1_0_0_1_1_1.wav:
N N Y N N Y Y Y


2023-09-07 14:17:43,613 INFO [jit_pretrained_decode_with_H.py:203] Decoding Done