Wattson is a Python library for estimating cloud compute carbon emissions.
It currently supports estimating emissions for a range of Amazon EC2 instances in a variety of regions.
You can install this package with pip by running pip install valohai-wattson
.
If you are currently using AWS instances, you can calculate the carbon emissions for your instances using the following code:
from wattson import estimate_carbon_emissions
training_emissions = estimate_carbon_emissions(
instance_type='c4.2xlarge',
region='us-east-1',
hours=1,
load_percentage=0.5,
)
The returned value will be of the type wattson.EmissionsEstimation
and have the following information:
data_license
: The license of the data used to estimate the emissions.region
: The region the original computation was performed in.instance_type
: The instance type of the original computation.avg_load
: The average CPU load during the original computation (defaults to 50% if not specified).scope_2_co2eq
: The estimated CO2 emissions for the electricity used in the original computation.scope_3_co2eq
: The estimated CO2 emissions for manufacturing of the device used in the original computation, assuming a 4-year usage.compensated
: Were any of the emissions of the original computation compensated using e.g. carbon emissions compensation or renewable energy credits.details
: Any additional details about e.g. the compensation methodology.
This project uses the EC2 Carbon Emissions Dataset by Teads Engineering.
- The library code is Copyright (c) 2022 Valohai, licensed under the MIT License.
- The Teads Engineering EC2 Carbon Emissions Dataset is licensed under the Creative Commons Attribution 4.0 International License.