mlcommons-speech

Simple test code for speech dataset

to install

  python3 -m venv environment
  source environment/bin/activate
  
  pip install -r requirements.txt

to run

  python train.py -m config/sota.json 

helpful arguments

  AWS_REGION=us-east-2 TF_CPP_MIN_LOG_LEVEL=3 python train.py -m config/sota.json 

• Sets the correct AWS region
• Supressed tensorflow S3 logs

Principles

• Push button: It should be possible to clone this repo and reproduce the SOTA result on multiple machines with one short command.
• A good model: The SOTA model should have a short and clean implementation and achieve high accuracy.