nvidia
/

stt_uk_citrinet_1024_gamma_0_25

Automatic Speech Recognition

hf-asr-leaderboard

Model card Files Files and versions Community

dpykhtar commited on Jul 29, 2022

Commit

501c692

·

1 Parent(s): 1bf0f25

Update README.md

Files changed (1) hide show

README.md +7 -1

README.md CHANGED Viewed

@@ -182,7 +182,13 @@ The tokenizer for this models was built using the text transcripts of the train
 ### Datasets
-Model is trained on validated Mozilla Common Voice Corpus 10.0 dataset(excluding dev and test data) comprising of 69 hours of Ukrainian speech.
 ## Limitations

 ### Datasets
+Model is trained on validated Mozilla Common Voice Corpus 10.0 dataset (excluding dev and test data) comprising of 69 hours of Ukrainian speech.
+## Performance
+| Version       | Tokenizer             | Vocabulary Size  | MCV-8 test | MCV-8 dev | MCV-9 test | MCV-9 dev | MCV-10 test | MCV-10 dev |
+| :-----------: |:---------------------:| :--------------: | :--------: | :-------: | :--------: | :-------: | :---------: | :--------: |
+| 1.0.0         | SentencePiece Unigram | 1024             | 4.27       | 5.66      | 4.45       | 5.57      | 5.53        | 5.30       |
 ## Limitations