2 Commits

Author	SHA1	Message	Date
Yifan Yang	070c77e724	Add Blankskip to Zipformer+CTC (#730) ... * init files * add ctc as auxiliary loss and ctc_decode.py * tuning the scalar of HLG score for 1best, nbest and nbest-oracle * rename to pruned_transducer_stateless7_ctc * fix doc * fix bug, recover the hlg scores * modify ctc_decode.py, move out the hlg scale * fix hlg_scale * add export.py and pretrained.py, and so on * upload files, update README.md and RESULTS.md * add CI test * update .gitignore * create symlinks * Add Blank Skip to Zipformer+CTC * Add warmup to blank skip * Add warmup to blank skip * Add __init__.py * Add parameters_names to Adam * Add warmup to blank skip * Modify frame_reducer * Modify frame_reducer * Add Blank Skip to decode. * Add ctc_decode.py * Add blank skip to Zipformer+CTC * process conflict * process conflict * modify ctc_guild_decode_bk.py * modify Lconv * produce the conflict * Add export.py * finish export * fix for running black * Add ci test * Add ci-test * chmod * chmod * fix bug for ci-test * fix bug for ci-test * fix bug for ci-test * rename the dirname * rename the dirname * change dirname * change dirname * fix notes * add pretrained.py * add pretrained.py * add pretrained.py * add pretrained.py * add pretrained.py * add pretrained.py * fix * fix * fix * finished * add the Copyright info and notes Co-authored-by: Zengwei Yao <yaozengwei@outlook.com> Co-authored-by: yifanyang <yifanyeung@yifanyangs-MacBook-Pro.local>	2022-12-21 17:41:31 +08:00
Wang, Guanbo	5fe58de43c	GigaSpeech recipe (#120) ... * initial commit * support download, data prep, and fbank * on-the-fly feature extraction by default * support BPE based lang * support HLG for BPE * small fix * small fix * chunked feature extraction by default * Compute features for GigaSpeech by splitting the manifest. * Fixes after review. * Split manifests into 2000 pieces. * set audio duration mismatch tolerance to 0.01 * small fix * add conformer training recipe * Add conformer.py without pre-commit checking * lazy loading and use SingleCutSampler * DynamicBucketingSampler * use KaldifeatFbank to compute fbank for musan * use pretrained language model and lexicon * use 3gram to decode, 4gram to rescore * Add decode.py * Update .flake8 * Delete compute_fbank_gigaspeech.py * Use BucketingSampler for valid and test dataloader * Update params in train.py * Use bpe_500 * update params in decode.py * Decrease num_paths while CUDA OOM * Added README * Update RESULTS * black * Decrease num_paths while CUDA OOM * Decode with post-processing * Update results * Remove lazy_load option * Use default `storage_type` * Keep the original tolerance * Use split-lazy * black * Update pretrained model Co-authored-by: Fangjun Kuang <csukuangfj@gmail.com>	2022-04-14 16:07:22 +08:00