add kw asr models, add itn ru checkpoint (tagger-based) (#4595)

* add kw asr models, add itn ru checkpoint (tagger-based) Signed-off-by: Alexandra Antonova <aleksandraa@nvidia.com> * add rw results to docs Signed-off-by: Alexandra Antonova <aleksandraa@nvidia.com> Co-authored-by: Alexandra Antonova <aleksandraa@nvidia.com>
NVIDIA · Jul 25, 2022 · 468a3f3 · 468a3f3
1 parent 6b9617d
commit 468a3f3
Show file tree

Hide file tree

Showing 6 changed files with 38 additions and 2 deletions.
diff --git a/docs/source/asr/data/benchmark_rw.csv b/docs/source/asr/data/benchmark_rw.csv
@@ -0,0 +1,3 @@
+Model,Model Base Class,Model Card
+stt_rw_conformer_ctc_large,EncDecCTCModel,"https://ngc.nvidia.com/catalog/models/nvidia:nemo:stt_rw_conformer_ctc_large"
+stt_rw_conformer_transducer_large,EncDecRNNTBPEModel,"https://ngc.nvidia.com/catalog/models/nvidia:nemo:stt_rw_conformer_transducer_large"
diff --git a/docs/source/asr/data/scores/rw/conformer_rw.csv b/docs/source/asr/data/scores/rw/conformer_rw.csv
@@ -0,0 +1,3 @@
+Model Name,Language,MCV Test-Set v9.0 (rw)
+stt_rw_conformer_ctc_large,rw,18.22
+stt_rw_conformer_transducer_large,rw,16.19
diff --git a/docs/source/asr/results.rst b/docs/source/asr/results.rst
@@ -209,3 +209,13 @@ Marathi
    :widths: 40, 10, 50
    :header-rows: 1
 
+-----------------------------
+
+Kinyarwanda
+^^^^^^^
+.. csv-table::
+   :file: data/benchmark_rw.csv
+   :align: left
+   :widths: 40, 10, 50
+   :header-rows: 1
+
diff --git a/nemo/collections/asr/models/ctc_bpe_models.py b/nemo/collections/asr/models/ctc_bpe_models.py
@@ -527,4 +527,11 @@ def list_available_models(cls) -> Optional[PretrainedModelInfo]:
         )
         results.append(model)
 
+        model = PretrainedModelInfo(
+            pretrained_model_name="stt_rw_conformer_ctc_large",
+            description="For details about this model, please visit https://ngc.nvidia.com/catalog/models/nvidia:nemo:stt_rw_conformer_ctc_large",
+            location="https://api.ngc.nvidia.com/v2/models/nvidia/nemo/stt_rw_conformer_ctc_large/versions/1.11.0/files/stt_rw_conformer_ctc_large.nemo",
+        )
+        results.append(model)
+
         return results
diff --git a/nemo/collections/asr/models/rnnt_bpe_models.py b/nemo/collections/asr/models/rnnt_bpe_models.py
@@ -190,6 +190,13 @@ def list_available_models(cls) -> List[PretrainedModelInfo]:
         )
         results.append(model)
 
+        model = PretrainedModelInfo(
+            pretrained_model_name="stt_rw_conformer_transducer_large",
+            description="For details about this model, please visit https://ngc.nvidia.com/catalog/models/nvidia:nemo:stt_rw_conformer_transducer_large",
+            location="https://api.ngc.nvidia.com/v2/models/nvidia/nemo/stt_rw_conformer_transducer_large/versions/1.11.0/files/stt_rw_conformer_transducer_large.nemo",
+        )
+        results.append(model)
+
         return results
 
     def __init__(self, cfg: DictConfig, trainer: Trainer = None):

diff --git a/nemo/collections/nlp/models/text_normalization_as_tagging/thutmose_tagger.py b/nemo/collections/nlp/models/text_normalization_as_tagging/thutmose_tagger.py
@@ -411,8 +411,14 @@ def list_available_models(cls) -> Optional[PretrainedModelInfo]:
             PretrainedModelInfo(
                 pretrained_model_name="itn_en_thutmose_bert",
                 location="https://api.ngc.nvidia.com/v2/models/nvidia/nemo/itn_en_thutmose_bert/versions/1.9.0/files/itn_en_thutmose_bert.nemo",
-                description="A single-pass tagger-based model for inverse text normalization based"
-                "on bert-base-uncased, trained on 2 mln sentences from Google Text Normalization Dataset",
+                description="A single-pass tagger-based English model for inverse text normalization based"
+                "on BERT, trained on 2 mln sentences from Google Text Normalization Dataset",
+            ),
+            PretrainedModelInfo(
+                pretrained_model_name="itn_ru_thutmose_bert",
+                location="https://api.ngc.nvidia.com/v2/models/nvidia/nemo/itn_ru_thutmose_bert/versions/1.11.0/files/itn_ru_thutmose_bert.nemo",
+                description="A single-pass tagger-based Russian model for inverse text normalization based"
+                "on BERT, trained on 2 mln sentences from Google Text Normalization Dataset",
             ),
         ]
         return result