Updated readme with FastConformer Hybrid De relevant information and results

README.md

@@ -1,150 +1,116 @@
 ---
 language:
-- en
+- de
 library_name: nemo
 datasets:
-- librispeech_asr
-- fisher_corpus
-- Switchboard-1
-- WSJ-0
-- WSJ-1
-- National-Singapore-Corpus-Part-1
-- National-Singapore-Corpus-Part-6
-- vctk
-- VoxPopuli-(EN)
-- Europarl-ASR-(EN)
-- Multilingual-LibriSpeech-(2000-hours)
-- mozilla-foundation/common_voice_8_0
-- MLCommons/peoples_speech
+- multilingual_librispeech
+- mozilla-foundation/common_voice_12_0
+- VoxPopuli-(DE)
 thumbnail: null
 tags:
 - automatic-speech-recognition
 - speech
 - audio
 - Transducer
-- Conformer
+- FastConformer
+- CTC
 - Transformer
 - pytorch
 - NeMo
 - hf-asr-leaderboard
 license: cc-by-4.0
-widget:
-- example_title: Librispeech sample 1
-  src: https://cdn-media.huggingface.co/speech_samples/sample1.flac
-- example_title: Librispeech sample 2
-  src: https://cdn-media.huggingface.co/speech_samples/sample2.flac
 model-index:
-- name: stt_en_conformer_transducer_xlarge
+- name: stt_de_fastconformer_hybrid_large_pc
   results:
   - task:
       name: Automatic Speech Recognition
       type: automatic-speech-recognition
     dataset:
-      name: LibriSpeech (clean)
-      type: librispeech_asr
-      config: clean
+      name: common-voice-12-0
+      type: mozilla-foundation/common_voice_12_0
+      config: de
       split: test
       args:
-        language: en
+        language: de
     metrics:
     - name: Test WER
       type: wer
-      value: 1.62
-  - task:
-      type: Automatic Speech Recognition
-      name: automatic-speech-recognition
-    dataset:
-      name: LibriSpeech (other)
-      type: librispeech_asr
-      config: other
-      split: test
-      args:
-        language: en
-    metrics:
-    - name: Test WER
-      type: wer
-      value: 3.01
+      value: 4.93
   - task:
       type: Automatic Speech Recognition
       name: automatic-speech-recognition
     dataset:
       name: Multilingual LibriSpeech
       type: facebook/multilingual_librispeech
-      config: english
+      config: german
       split: test
       args:
-        language: en
+        language: de
     metrics:
     - name: Test WER
       type: wer
-      value: 5.32
+      value: 3.8
   - task:
       type: Automatic Speech Recognition
       name: automatic-speech-recognition
     dataset:
-      name: Mozilla Common Voice 7.0
-      type: mozilla-foundation/common_voice_7_0
-      config: en
+      name: Vox Populi
+      type: polinaeterna/voxpopuli
+      config: german
       split: test
       args:
-        language: en
+        language: de
     metrics:
     - name: Test WER
       type: wer
-      value: 5.13
+      value: 8.6
   - task:
-      type: Automatic Speech Recognition
-      name: automatic-speech-recognition
+      name: Automatic Speech Recognition
+      type: automatic-speech-recognition
     dataset:
-      name: Mozilla Common Voice 8.0
-      type: mozilla-foundation/common_voice_8_0
-      config: en
+      name: common-voice-12-0
+      type: mozilla-foundation/common_voice_12_0
+      config: German P&C
       split: test
       args:
-        language: en
+        language: de
     metrics:
-    - name: Test WER
+    - name: Test WER P&C
       type: wer
-      value: 6.46
+      value: 5.39
   - task:
       type: Automatic Speech Recognition
       name: automatic-speech-recognition
     dataset:
-      name: Wall Street Journal 92
-      type: wsj_0
-      args:
-        language: en
-    metrics:
-    - name: Test WER
-      type: wer
-      value: 1.17
-  - task:
-      type: Automatic Speech Recognition
-      name: automatic-speech-recognition
-    dataset:
-      name: Wall Street Journal 93
-      type: wsj_1
+      name: Multilingual LibriSpeech
+      type: facebook/multilingual_librispeech
+      config: German P&C
+      split: test
       args:
-        language: en
+        language: de
     metrics:
-    - name: Test WER
+    - name: Test WER P&C
       type: wer
-      value: 2.05
+      value: 11.1
   - task:
       type: Automatic Speech Recognition
       name: automatic-speech-recognition
     dataset:
-      name: National Singapore Corpus
-      type: nsc_part_1
+      name: Vox Populi
+      type: polinaeterna/voxpopuli
+      config: German P&C
+      split: test
       args:
-        language: en
+        language: de
     metrics:
-    - name: Test WER
+    - name: Test WER P&C
       type: wer
-      value: 5.7
+      value: 10.41
+metrics:
+- wer
 ---
 
-# NVIDIA Conformer-Transducer X-Large (en-US)
+# NVIDIA FastConformer-Hybrid Large (de)
 
 <style>
 img {
@@ -152,24 +118,20 @@ img {
 }
 </style>
 
-| [![Model architecture](https://img.shields.io/badge/Model_Arch-Conformer--Transducer-lightgrey#model-badge)](#model-architecture)
-| [![Model size](https://img.shields.io/badge/Params-600M-lightgrey#model-badge)](#model-architecture)
-| [![Language](https://img.shields.io/badge/Language-en--US-lightgrey#model-badge)](#datasets)
+| [![Model architecture](https://img.shields.io/badge/Model_Arch-FastConformer--Transducer_CTC-lightgrey#model-badge)](#model-architecture)
+| [![Model size](https://img.shields.io/badge/Params-115M-lightgrey#model-badge)](#model-architecture)
+| [![Language](https://img.shields.io/badge/Language-de-lightgrey#model-badge)](#datasets)
 
 
-This model transcribes speech in lower case English alphabet along with spaces and apostrophes.
-It is an "extra-large" versions of Conformer-Transducer (around 600M parameters) model.  
-See the [model architecture](#model-architecture) section and [NeMo documentation](https://docs.nvidia.com/deeplearning/nemo/user-guide/docs/en/main/asr/models.html#conformer-transducer) for complete architecture details.
+This model transcribes speech in upper and lower case German alphabet along with spaces, periods, commas, and question marks.
+It is a "large" version of FastConformer Transducer-CTC (around 115M parameters) model.  
+See the [model architecture](#model-architecture) section and [NeMo documentation](https://docs.nvidia.com/deeplearning/nemo/user-guide/docs/en/main/asr/models.html#fast-conformer) for complete architecture details.
 
 ## NVIDIA NeMo: Training
 
 To train, fine-tune or play with the model you will need to install [NVIDIA NeMo](https://github.com/NVIDIA/NeMo). We recommend you install it after you've installed latest Pytorch version.
 ```
 pip install nemo_toolkit['all']
-'''
-'''
-(if it causes an error): 
-pip install nemo_toolkit[all]
 ``` 
 
 ## How to Use this Model
@@ -180,7 +142,7 @@ The model is available for use in the NeMo toolkit [3], and can be used as a pre
 
 ```python
 import nemo.collections.asr as nemo_asr
-asr_model = nemo_asr.models.EncDecRNNTBPEModel.from_pretrained("nvidia/stt_en_conformer_transducer_xlarge")
+asr_model = nemo_asr.models.EncDecHybridRNNTCTCBPEModel.from_pretrained(model_name="nvidia/stt_de_fastconformer_hybrid_large_pc")
 ```
 
 ### Transcribing using Python
@@ -195,12 +157,21 @@ asr_model.transcribe(['2086-149220-0033.wav'])
 
 ### Transcribing many audio files
 
+Using Transducer mode inference:
 ```shell
 python [NEMO_GIT_FOLDER]/examples/asr/transcribe_speech.py 
- pretrained_name="nvidia/stt_en_conformer_transducer_xlarge" 
+ pretrained_name="nvidia/stt_de_fastconformer_hybrid_large_pc" 
  audio_dir="<DIRECTORY CONTAINING AUDIO FILES>"
 ```
 
+Using CTC mode inference:
+```shell
+python [NEMO_GIT_FOLDER]/examples/asr/transcribe_speech.py 
+ pretrained_name="nvidia/stt_de_fastconformer_hybrid_large_pc" 
+ audio_dir="<DIRECTORY CONTAINING AUDIO FILES>"
+ decoder_type="ctc"
+```
+
 ### Input
 
 This model accepts 16000 KHz Mono-channel Audio (wav files) as input.
@@ -211,42 +182,43 @@ This model provides transcribed speech as a string for a given audio sample.
 
 ## Model Architecture
 
-Conformer-Transducer model is an autoregressive variant of Conformer model [1] for Automatic Speech Recognition which uses Transducer loss/decoding instead of CTC Loss. You may find more info on the detail of this model here: [Conformer-Transducer Model](https://docs.nvidia.com/deeplearning/nemo/user-guide/docs/en/main/asr/models.html). 
+FastConformer is an optimized version of the Conformer model [1] with 8x depthwise-separable convolutional downsampling. The model is trained in a multitask setup with joint Transducer and CTC decoder loss. You may find more information on the details of FastConformer here: [Fast-Conformer Model](https://docs.nvidia.com/deeplearning/nemo/user-guide/docs/en/main/asr/models.html#fast-conformer) and about Hybrid Transducer-CTC training here: [Hybrid Transducer-CTC](https://docs.nvidia.com/deeplearning/nemo/user-guide/docs/en/main/asr/models.html#hybrid-transducer-ctc).
 
 ## Training
 
-The NeMo toolkit [3] was used for training the models for over several hundred epochs. These model are trained with this [example script](https://github.com/NVIDIA/NeMo/blob/main/examples/asr/asr_transducer/speech_to_text_rnnt_bpe.py) and this [base config](https://github.com/NVIDIA/NeMo/blob/main/examples/asr/conf/conformer/conformer_transducer_bpe.yaml).
+The NeMo toolkit [3] was used for training the models for over several hundred epochs. These model are trained with this [example script](https://github.com/NVIDIA/NeMo/blob/main/examples/asr/asr_hybrid_transducer_ctc/speech_to_text_hybrid_rnnt_ctc_bpe.py) and this [base config](https://github.com/NVIDIA/NeMo/blob/main/examples/asr/conf/fastconformer/hybrid_transducer_ctc/fastconformer_hybrid_transducer_ctc_bpe.yaml).
 
 The tokenizers for these models were built using the text transcripts of the train set with this [script](https://github.com/NVIDIA/NeMo/blob/main/scripts/tokenizers/process_asr_text_tokenizer.py).
 
 ### Datasets
 
-All the models in this collection are trained on a composite dataset (NeMo ASRSET) comprising of several thousand hours of English speech:
+All the models in this collection are trained on a composite dataset (NeMo PnC ASRSET) comprising of 2500 hours of German speech:
 
-- Librispeech 960 hours of English speech
-- Fisher Corpus
-- Switchboard-1 Dataset
-- WSJ-0 and WSJ-1
-- National Speech Corpus (Part 1, Part 6)
-- VCTK
-- VoxPopuli (EN)
-- Europarl-ASR (EN)
-- Multilingual Librispeech (MLS EN) - 2,000 hrs subset
-- Mozilla Common Voice (v8.0)
-- People's Speech  - 12,000 hrs subset
-
-Note: older versions of the model may have trained on smaller set of datasets.
+- MCV12 (800 hrs)
+- MLS (1500 hrs)
+- Voxpopuli (200 hrs)
 
 ## Performance
 
-The list of the available models in this collection is shown in the following table. Performances of the ASR models are reported in terms of Word Error Rate (WER%) with greedy decoding.
+The performance of Automatic Speech Recognition models is measuring using Word Error Rate. Since this dataset is trained on multiple domains and a much larger corpus, it will generally perform better at transcribing audio in general.
+
+The following tables summarizes the performance of the available models in this collection with the Transducer decoder. Performances of the ASR models are reported in terms of Word Error Rate (WER%) with greedy decoding. 
+
+
+a) On data without Punctuation and Capitalization with Transducer decoder
+| **Version** |     **Tokenizer**     | **Vocabulary Size** | **MCV12 DEV** | **MCV12 TEST** | **MLS DEV** | **MLS TEST** | **VOXPOPULI DEV** | **VOXPOPULI TEST** |
+|:-----------:|:---------------------:|:-------------------:|:-------------:|:--------------:|:-----------:|:------------:|:-----------------:|:------------------:|
+|    1.18.0    | SentencePiece Unigram |         1024        |      4.18     |      4.93      |     3.3     |      3.8     |        10.8       |         8.6        |
+
+
+b) On data with Punctuation and Capitalization with Transducer decoder
+| **Version** |     **Tokenizer**     | **Vocabulary Size** | **MCV12 DEV** | **MCV12 TEST** | **MLS DEV** | **MLS TEST** | **VOXPOPULI DEV** | **VOXPOPULI TEST** |
+|:-----------:|:---------------------:|:-------------------:|:-------------:|:--------------:|:-----------:|:------------:|:-----------------:|:------------------:|
+|    1.18.0    | SentencePiece Unigram |         1024        |      4.66     |      5.39      |    10.12    |     11.1     |       12.96       |        10.41       |
 
-| Version | Tokenizer | Vocabulary Size | LS test-other | LS test-clean | WSJ Eval92 | WSJ Dev93 | NSC Part 1 |  MLS Test | MLS Dev | MCV Test 8.0 | Train Dataset |
-|---------|-----------------------|-----------------|---------------|---------------|------------|-----------|-----|-------|------|----|------|
-| 1.10.0 | SentencePiece Unigram | 1024 | 3.01 | 1.62 | 1.17 | 2.05 | 5.70 | 5.32 | 4.59 | 6.46 | NeMo ASRSET 3.0 |
 
 ## Limitations
-Since this model was trained on publicly available speech datasets, the performance of this model might degrade for speech which includes technical terms, or vernacular that the model has not been trained on. The model might also perform worse for accented speech.
+Since this model was trained on publically available speech datasets, the performance of this model might degrade for speech which includes technical terms, or vernacular that the model has not been trained on. The model might also perform worse for accented speech. The model only outputs the punctuations: ```'.', ',', '?' ``` and hence might not do well in scenarios where other punctuations are also expected.
 
 ## NVIDIA Riva: Deployment
 
@@ -267,4 +239,4 @@ Check out [Riva live demo](https://developer.nvidia.com/riva#demos).
 
 ## Licence
 
-License to use this model is covered by the [CC-BY-4.0](https://creativecommons.org/licenses/by/4.0/). By downloading the public and release version of the model, you accept the terms and conditions of the [CC-BY-4.0](https://creativecommons.org/licenses/by/4.0/) license.
+License to use this model is covered by the [CC-BY-4.0](https://creativecommons.org/licenses/by/4.0/). By downloading the public and release version of the model, you accept the terms and conditions of the [CC-BY-4.0](https://creativecommons.org/licenses/by/4.0/) license.