NVIDIA NeMo

Introduction

NVIDIA NeMo is a conversational AI toolkit built for researchers working on automatic speech recognition (ASR), text-to-speech synthesis (TTS), large language models (LLMs), and natural language processing (NLP). The primary objective of NeMo is to help researchers from industry and academia to reuse prior work (code and pretrained models) and make it easier to create new conversational AI models.

All NeMo models are trained with Lightning and training is automatically scalable to 1000s of GPUs. Additionally, NeMo Megatron LLM models can be trained up to 1 trillion parameters using tensor and pipeline model parallelism. NeMo models can be optimized for inference and deployed for production use-cases with NVIDIA Riva.

Getting started with NeMo is simple. State of the Art pretrained NeMo models are freely available on HuggingFace Hub and NVIDIA NGC. These models can be used to transcribe audio, synthesize speech, or translate text in a just a few lines of code.

We have have extensive tutorials that can all be run on Google Colab.

For advanced users that want to train NeMo models from scratch or finetune existing NeMo models we have a full suite of example scripts that support multi-GPU/multi-node training.

Also see our introductory video for a high level overview of NeMo.

Key Features

• Speech processing

  • HuggingFace Space for Audio Transcription (File, Micriphone and YouTube)

  • Automatic Speech Recognition (ASR)

    • Supported models: Jasper, QuartzNet, CitriNet, Conformer-CTC, Conformer-Transducer, Squeezeformer-CTC, Squeezeformer-Transducer, ContextNet, LSTM-Transducer (RNNT), LSTM-CTC, ...

    • Supports CTC and Transducer/RNNT losses/decoders

      • NeMo Original Multi-blank Transducers

    • Beam Search decoding
    • Language Modelling for ASR: N-gram LM in fusion with Beam Search decoding, Neural Rescoring with Transformer
    • Streaming and Buffered ASR (CTC/Transducer) - Chunked Inference Examples
    • Support of long audios for Conformer with memory efficient local attention

  • Speech Classification and Speech Command Recognition: MatchboxNet (Command Recognition)

  • Voice activity Detection (VAD): MarbleNet

    • ASR with VAD Inference - Example

  • Speaker Recognition: TitaNet, ECAPA_TDNN, SpeakerNet

  • Speaker Diarization

    • Clustering Diarizer: TitaNet, ECAPA_TDNN, SpeakerNet
    • Neural Diarizer: MSDD (Multi-scale Diarization Decoder)

  • Speech Intent Detection and Slot Filling: Conformer-Transformer
  • Pretrained models on different languages.: English, Spanish, German, Russian, Chinese, French, Italian, Polish, ...
  • NGC collection of pre-trained speech processing models.

• Natural Language Processing

  • NeMo Megatron pre-training of Large Language Models
  • Neural Machine Translation (NMT)
  • Punctuation and Capitalization
  • Token classification (named entity recognition)
  • Text classification
  • Joint Intent and Slot Classification
  • Question answering
  • GLUE benchmark
  • Information retrieval
  • Entity Linking
  • Dialogue State Tracking
  • Prompt Learning
  • NGC collection of pre-trained NLP models.
  • Synthetic Tabular Data Generation

• Speech synthesis (TTS)

  • Spectrogram generation: Tacotron2, GlowTTS, TalkNet, FastPitch, FastSpeech2, Mixer-TTS, Mixer-TTS-X
  • Vocoders: WaveGlow, SqueezeWave, UniGlow, MelGAN, HiFiGAN, UnivNet
  • End-to-end speech generation: FastPitch_HifiGan_E2E, FastSpeech2_HifiGan_E2E
  • NGC collection of pre-trained TTS models.

• Tools

  • Text Processing (text normalization and inverse text normalization)
  • CTC-Segmentation tool
  • Speech Data Explorer: a dash-based tool for interactive exploration of ASR/TTS datasets

Built for speed, NeMo can utilize NVIDIA's Tensor Cores and scale out training to multiple GPUs and multiple nodes.

Requirements

1. Python 3.8 or above
2. Pytorch 1.10.0 or above
3. NVIDIA GPU for training

Documentation

Version	Status	Description
Latest		Documentation of the latest (i.e. main) branch.
Stable		Documentation of the stable (i.e. most recent release) branch.

Tutorials

A great way to start with NeMo is by checking one of our tutorials.

Getting help with NeMo

FAQ can be found on NeMo's Discussions board. You are welcome to ask questions or start discussions there.

Installation

Conda

We recommend installing NeMo in a fresh Conda environment.

conda create --name nemo python==3.8
conda activate nemo

Install PyTorch using their configurator.

conda install pytorch torchvision torchaudio cudatoolkit=11.3 -c pytorch

Note

The command used to install PyTorch may depend on your system.

Pip

Use this installation mode if you want the latest released version.

apt-get update && apt-get install -y libsndfile1 ffmpeg
pip install Cython
pip install nemo_toolkit['all']

Note

Depending on the shell used, you may need to use "nemo_toolkit[all]" instead in the above command.

Pip from source

Use this installation mode if you want the a version from particular GitHub branch (e.g main).

apt-get update && apt-get install -y libsndfile1 ffmpeg
pip install Cython
python -m pip install git+https://github.com/NVIDIA/NeMo.git@{BRANCH}#egg=nemo_toolkit[all]

From source

Use this installation mode if you are contributing to NeMo.

apt-get update && apt-get install -y libsndfile1 ffmpeg
git clone https://github.com/NVIDIA/NeMo
cd NeMo
./reinstall.sh

Note

If you only want the toolkit without additional conda-based dependencies, you may replace reinstall.sh with pip install -e . when your PWD is the root of the NeMo repository.

RNNT

Note that RNNT requires numba to be installed from conda.

conda remove numba
pip uninstall numba
conda install -c conda-forge numba

NeMo Megatron

NeMo Megatron training requires NVIDIA Apex to be installed. Install it manually if not using the NVIDIA PyTorch container.

git clone https://github.com/ericharper/apex.git
cd apex
git checkout nm_v1.14.0
pip install -v --disable-pip-version-check --no-cache-dir --global-option="--cpp_ext" --global-option="--cuda_ext" --global-option="--fast_layer_norm" --global-option="--distributed_adam" --global-option="--deprecated_fused_adam" ./

Transformer Engine

NeMo Megatron GPT has been integrated with NVIDIA Transformer Engine Transformer Engine enables FP8 training on NVIDIA Hopper GPUs. Install it manually if not using the NVIDIA PyTorch container.

Note

Transformer Engine requires PyTorch to be built with CUDA 11.8.

NeMo Text Processing

NeMo Text Processing, specifically (Inverse) Text Normalization, requires Pynini to be installed.

bash NeMo/nemo_text_processing/install_pynini.sh

Docker containers:

To build a nemo container with Dockerfile from a branch, please run

DOCKER_BUILDKIT=1 docker build -f Dockerfile -t nemo:latest .

If you chose to work with main branch, we recommend using NVIDIA's PyTorch container version 22.12-py3 and then installing from GitHub.

docker run --gpus all -it --rm -v <nemo_github_folder>:/NeMo --shm-size=8g \
-p 8888:8888 -p 6006:6006 --ulimit memlock=-1 --ulimit \
stack=67108864 --device=/dev/snd nvcr.io/nvidia/pytorch:22.12-py3

Examples

Many examples can be found under "Examples" folder.

Contributing

We welcome community contributions! Please refer to the CONTRIBUTING.md CONTRIBUTING.md for the process.

Publications

We provide an ever growing list of publications that utilize the NeMo framework. Please refer to PUBLICATIONS.md. We welcome the addition of your own articles to this list !

Citation

@article{kuchaiev2019nemo,
  title={Nemo: a toolkit for building ai applications using neural modules},
  author={Kuchaiev, Oleksii and Li, Jason and Nguyen, Huyen and Hrinchuk, Oleksii and Leary, Ryan and Ginsburg, Boris and Kriman, Samuel and Beliaev, Stanislav and Lavrukhin, Vitaly and Cook, Jack and others},
  journal={arXiv preprint arXiv:1909.09577},
  year={2019}
}

License

NeMo is under Apache 2.0 license.