v3.0: AI view, csv import-export, more help, some new data

Major updates of this version:

• The AI page combining a training and inference form, both relying on the base Green Algorithms calculator
• CSV import-export system, allowing to download the calculator state in a CSV file and then to upload it back to the platform

Minor updates

• Addition of some CPUs and GPUs
• Update of PUE default value to 2024 estimate
• Small design changes
• Bug fixes

⚠️ The carbon intensity values (of electricity consumption) have not been updated yet and should be soon. This is due to a change of the T&C of carbonfootprint.com, we are working at solving this as quickly as possible.

More info about this release

AI page:

• A new page dedicated to AI-related computations, intended to facilitate the estimation and reporting of AI systems's environmental footprint
• Includes a training and inference form with specific fields for retrainings, R&D experiments and two inference modes
• Aggregated and "per form" results are now displayed

CSV import-export

• The link sharing results is now disabled and replaced by two boxes: one for CSV export, one for CSV import
• This system allows to share results among users but also to manually compile different results on a single CSV, in order to compare results
• The CSV exported from the classic view contains all the fields to replicate the main form, whereas the ones exported from the AI view allow to replicate the state of both training and inference forms

Informative content

• The app is now made clearer, with tips for each form entry and help tabs associated to the training and inference form

Incoming features:

• Translation in multiple language

Fixed some bugs that appeared with the Dash upgrade.

Minor updates

• Fixed a bug where PUE wouldn't update properly when switching between platforms
• Added some CPUs ([issue])
  • AMD EPYC 7343
  • AMD EPYC 7513

Major changes:

• Updates carbon intensities with the 2022 electricity factors from carbonfootprint.com (mostly based from 2020 data).
• You can now load past versions of the app to replicate past results (and past versions’ data is available on github).

Minor changes:

• Better handles very large and very small carbon footprints.
• Improved the URL fill-in + reset options.
• Minor bug fixes.

Minor improvements, thanks to ongcp97:

• speed optimisation
• display for very large carbon footprints
• UX improvements
• links now open in a new window

v2.0 is here!

Biggest update yet for the Green Algorithms app, with the v2.0 out today!

Some long-awaited features, such as:

• Sharing results via a link
• Mixing GPUs and CPUs in the same estimation
• Better integration of cloud computing

Release log

Cloud computing

• Added the option to select a server rather than a location list for the major data centre providers who make this information available.

CPU and GPU

• Possibility to input both CPU and GPU time in the same estimation
• The pie chart shows the contribution of both CPU and GPUs

Permalink

• Possibility to share your results by sharing a permalink that leads to the calculator with a specific set of inputs.

General

• Adjusted layout and other design changes
• Added CPUs and GPUs to the list

Updated carbon intensities with 2020 values.
Minor changes ahead of v2.0

The v1.0.0 is out!

It brings together most of the functionalities needed to estimate the carbon footprint of an algorithm.

Most significant improvements:

• processors' usage factor can be specified
• a Pragmatic Scaling Factor (PSF) can be included
• Energy need (on top of carbon emissions) is displayed

Feel free to suggest improvements for the future versions!