ACCESS-NRI Release Stages
We develop and share our model configurations1 and tools in stages. Each stage has a different purpose and level of readiness. These terms are used to give a general indication on the state of the product you want to use and things to be aware of:
-
Prototype (not shown in table): Early versions shared with a small group of experts to explore new ideas. These are not polished or complete, and may change quickly.
-
Alpha: Early testing phase. The core features are in place, but things may change. No long-term support or maintenance. Feedback at this stage helps shape the product.
-
Beta: More stable versions open to a wider group. Most features and documentation are ready, and we welcome testing by anyone interested. Changes may still occur. No long-term support or maintenance.
-
Full Release: Fully tested, verified and documented. Supported versions with maintenance. Bitwise reproducible.2
Info
These release definitions will not be applied retroactively to any model configuration or tool release.
Release stages for model configurations1
The table below has some details for each release stage of ACCESS-NRI supported model configurations. Note that this table is valid only for model configurations, and not for tools.
ALPHA | BETA | FULL RELEASE | |
---|---|---|---|
Overall state of product | Unstable | Somewhat stable | Stable |
How to access and run the model configuration | Pre-release module | Pre-release module | Model/version module |
Input files for configurations | Not yet managed3 | Managed3 | Managed3, FAIR |
Availability & maintenance of the deployed model configuration | Short-term, for alpha testers only | Until next beta or full release | Long-term4 |
User support (via Hive Forum) | No guarantee, for alpha testers only | Until next beta or full release | Fully supported4 |
Spin-up and control experiment(s) | N/A | Under validation (if available) | Validated and published when available |
Documentation on Hive Docs | Unpublished | Published publicly with “beta” caveat | Published publicly |
Optimisation & scaling | No guarantee | In progress | Complete |
DOI | None | None | DOI issued; any model updates receive new DOIs |
Announcements to the community | Direct contact with alpha testers. No formal announcement. | Public post on Hive Forum. | Public post on Hive Forum. On ACCESS-NRI website, newsletter, and social media. May include a media release. |
-
A model configuration is the entire setup for a model experiment run. It is the compiled model code, the parameters, and input files. It is essentially an experiment waiting to happen. ↩↩
-
By bitwise reproducible, we mean that a model configuration will produce the same answers if re-run on Gadi (with the same layout, number of processors, etc., all of which are specified in a model configuration). ↩
-
By managed, we mean that the input files are collected in a specific place and are incorporated into the model configuration. ↩↩↩
-
Each release will have its own timeframe for availability, maintenance, and support. The timeframe will be based on factors that include: what our users need; availability, and suitability, of new models; machine changes (older models may not be ported to new machines); overall support burden (if we support too many models our ability to update current models or develop new ones is impaired); collaborators (if we use a model component that is deprecated it may make it harder for us to support it.) ↩↩