You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
PaddleSpeech/examples/librispeech/s1
Hui Zhang eef8847a82
compute cmvn before build vocab
4 years ago
..
conf test refactor collator 4 years ago
local compute cmvn before build vocab 4 years ago
.gitignore E2E/Streaming Transformer/Conformer ASR (#578) 5 years ago
README.md compute cmvn before build vocab 4 years ago
cmd.sh librispeech s1 support sclite and multi process decode 4 years ago
path.sh disable __pycache__ 4 years ago
run.sh librispeech s1 support sclite and multi process decode 4 years ago
utils add dur utils 4 years ago

README.md

LibriSpeech

Conformer

Model Params Config Augmentation Test set Decode method Loss WER
conformer 47.63 M conf/conformer.yaml spec_aug + shift test-clean attention 6.35 0.030162
conformer 47.63 M conf/conformer.yaml spec_aug + shift test-clean ctc_greedy_search 6.35 0.037910
conformer 47.63 M conf/conformer.yaml spec_aug + shift test-clean ctc_prefix_beam_search 6.35 0.037761
conformer 47.63 M conf/conformer.yaml spec_aug + shift test-clean attention_rescoring 6.35 0.032115

Test w/o length filter

Model Params Config Augmentation Test set Decode method Loss WER
conformer 47.63 M conf/conformer.yaml spec_aug + shift test-clean-all attention 6.35 0.057117

Chunk Conformer

Model Params Config Augmentation Test set Decode method Chunk Size & Left Chunks Loss WER
conformer 47.63 M conf/chunk_conformer.yaml spec_aug + shift test-clean attention 16, -1 7.11 0.063193
conformer 47.63 M conf/chunk_conformer.yaml spec_aug + shift test-clean ctc_greedy_search 16, -1 7.11 0.082394
conformer 47.63 M conf/chunk_conformer.yaml spec_aug + shift test-clean ctc_prefix_beam_search 16, -1 7.11 0.082156
conformer 47.63 M conf/chunk_conformer.yaml spec_aug + shift test-clean attention_rescoring 16, -1 7.11 0.071000

Transformer

Model Params Config Augmentation Test set Decode method Loss WER
transformer 32.52 M conf/transformer.yaml spec_aug + shift test-clean attention 6.98 0.036

Test w/o length filter

Model Params Config Augmentation Test set Decode method Loss WER
transformer 32.52 M conf/transformer.yaml spec_aug + shift test-clean-all attention 7.63 0.056832
transformer 32.52 M conf/transformer.yaml spec_aug + shift test-clean-all ctc_greedy_search 7.63 0.059742
transformer 32.52 M conf/transformer.yaml spec_aug + shift test-clean-all ctc_prefix_beam_search 7.63 0.059057
transformer 32.52 M conf/transformer.yaml spec_aug + shift test-clean-all attention_rescoring 7.63 0.047417