NVIDIA NeMo

Introduction

NeMo is a toolkit for creating Conversational AI applications.

NeMo toolkit makes it possible for researchers to easily compose complex neural network architectures for conversational AI using reusable components - Neural Modules. Neural Modules are conceptual blocks of neural networks that take typed inputs and produce typed outputs. Such modules typically represent data layers, encoders, decoders, language models, loss functions, or methods of combining activations.

The toolkit comes with extendable collections of pre-built modules and ready-to-use models for automatic speech recognition (ASR), natural language processing (NLP) and text synthesis (TTS). Built for speed, NeMo can utilize NVIDIA's Tensor Cores and scale out training to multiple GPUs and multiple nodes.

Requirements

NeMo's works with:

1. Python 3.6 or 3.7
2. Pytorch 1.6 or above

Docker container:

We recommend using NVIDIA's PyTorch container version 20.06-py3 with NeMo's main branch.

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 nvcr.io/nvidia/pytorch:20.06-py3

Installation

Once requirements are satisfied (or you are inside NVIDIA docker container), simply install using pip:

• pip install nemo_toolkit[all]==version
• pip install nemo_toolkit[all] - latest released version (currently 0.11.0)

Or if you want the latest (or particular) version from GitHub:

• python -m pip install git+https://github.com/NVIDIA/NeMo.git@{BRANCH}#egg=nemo_toolkit[nlp] - where {BRANCH} should be replaced with the branch you want. This is recommended route if you are testing out the latest WIP version of NeMo.
• ./reinstall.sh - from NeMo's git root. This will install the version from current branch.

Examples

<nemo_github_folder>/examples/ folder contains various example scripts. Many of them look very similar and have the same arguments because we used Facebook's Hydra for configuration.

Here is an example command which trains ASR model (QuartzNet15x5) on LibriSpeech, using 4 GPUs and mixed precision training. (It assumes you are inside the container with NeMo installed)

root@987b39669a7e:/NeMo# python examples/asr/speech_to_text.py --config-name=quartznet_15x5 \
model.train_ds.manifest_filepath=<PATH_TO_DATA>/librispeech-train-all.json \
model.validation_ds.manifest_filepath=<PATH_TO_DATA>/librispeech-dev-other.json \
trainer.gpus=4 trainer.max_epochs=128 model.train_ds.batch_size=64 \
+trainer.precision=16 +trainer.amp_level=O1  \
+model.validation_ds.num_workers=16  +model.train_ds.num_workers=16

#(Optional) Tensorboard:
tensorboard --bind_all --logdir nemo_experiments

Documentation

Version	Status	Description
Latest		Documentation of the latest (i.e. main) branch
Stable		Documentation of the stable (i.e. 0.11.1) branch
Main		Documentation of the main branch
v0.11.1		Documentation of the v0.11.1 release
v0.11.0		Documentation of the v0.11.0 release
v0.10.1		Documentation of the v0.10.1 release

Tutorials

The best way to get started with NeMo is to checkout one of our tutorials.

Most NeMo tutorials can be run on Google's Colab.

To run tutorials:

1. Click on Colab link (see table below)
2. Connect to an instance with a GPU (Runtime -> Change runtime type -> select "GPU" for hardware accelerator)

NeMo tutorials

Domain	Title	GitHub URL
NeMo	Exploring NeMo Fundamentals	00_NeMo_Primer.ipynb
ASR	ASR with NeMo	01_ASR_with_NeMo.ipynb
ASR	Speech Commands	02_Speech_Commands.ipynb
ASR	Online Noise Augmentation	05_Online_Noise_Augmentation.ipynb
NLP	Token Classification (Named Entity Recognition)	Token_Classification_Named_Entity_Recognition.ipynb
NLP	GLUE Benchmark	GLUE_Benchmark.ipynb
NLP	Punctuation and Capitialization	Punctuation_and_Capitalization.ipynb
NLP	Question answering with SQuAD	Question_Answering_Squad.ipynb
TTS	Speech Synthesis	TTS_inference.ipynb

Contributing

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

License

NeMo is under Apache 2.0 license.