Import reazonspeech-nemo-v2.nemo

Signed-off-by: Fujimoto Seiji <[email protected]>

.gitattributes

@@ -32,4 +32,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.xz filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
+*.nemo filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text

README.md

@@ -1,3 +1,53 @@
 ---
 license: apache-2.0
+language:
+- ja
+library_name: nemo
+tags:
+  - automatic-speech-recognition
+  - NeMo
 ---
+
+# reazonspeech-nemo-v2
+
+`reazonspeech-nemo-v2` is an automatic speech recognition model trained
+on [ReazonSpeech v2.0 corpus](https://huggingface.co/datasets/reazon-research/reazonspeech).
+
+This model supports inference of long-form Japanese audio clips up to
+several hours.
+
+## Model Architecture
+
+The model features an improved Conformer architecture from
+[Fast Conformer with Linearly Scalable Attention for Efficient
+Speech Recognition](https://arxiv.org/abs/2305.05084).
+
+* Subword-based RNN-T model. The total parameter count is 619M.
+
+* Encoder uses [Longformer](https://arxiv.org/abs/2004.05150) attention
+  with local context size of 256, and has a single global token.
+
+* Decoder has a vocabulary space of 3000 tokens constructed by
+  [SentencePiece](https://github.com/google/sentencepiece)
+  unigram tokenizer.
+
+We trained this model for 1 million steps using AdamW optimizer
+following Noam annealing schedule.
+
+## Usage
+
+We recommend to use this model through our
+[reazonspeech](https://github.com/reazon-research/reazonspeech)
+library.
+
+```
+from reazonspeech.nemo.asr import load_model, transcribe
+
+model = load_model()
+ret = transcribe("speech.wav", model)
+print(ret.text)
+```
+
+## License
+
+[Apaceh Licence 2.0](https://choosealicense.com/licenses/apache-2.0/)

reazonspeech-nemo-v2.nemo

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d196d43ad03466ca88beeda4bf5fafb07bab7202d4b663b8e4f12cb0a4381fae
+size 2477946880