since loss reduce and model more confidence, then less lm alpha

pull/567/head
Hui Zhang 5 years ago
parent f8719971b5
commit 1573675888

@ -3,5 +3,5 @@
## CTC ## CTC
| Model | Config | Test Set | WER | Valid Loss | | Model | Config | Test Set | WER | Valid Loss |
| --- | --- | --- | --- | --- | | --- | --- | --- | --- | --- |
| DeepSpeech2 | conf/deepspeech2.yaml | test-clean | 0.071391 | 15.078561 | | DeepSpeech2 | conf/deepspeech2.yaml | test-clean | 0.069357 | 15.078561 |
| DeepSpeech2 | release 1.8.5 | test-clean | 0.074939 | 15.351633 | | DeepSpeech2 | release 1.8.5 | test-clean | 0.074939 | 15.351633 |

@ -39,7 +39,7 @@ decoding:
error_rate_type: wer error_rate_type: wer
decoding_method: ctc_beam_search decoding_method: ctc_beam_search
lang_model_path: data/lm/common_crawl_00.prune01111.trie.klm lang_model_path: data/lm/common_crawl_00.prune01111.trie.klm
alpha: 2.5 alpha: 1.9
beta: 0.3 beta: 0.3
beam_size: 500 beam_size: 500
cutoff_prob: 1.0 cutoff_prob: 1.0

Loading…
Cancel
Save