animals
:v0
. Every new version of an artifact bumps the index by one. You can imagine that once you have hundreds of versions, referring to a specific version by its index would be confusing and error prone. This is where aliases come in handy. An alias allows you to apply a human-readable name to given version.latest
image. Here's our new directory structure:log_artifact.py
to produce animals:v1
. W&B will automatically assign the newest version the alias latest
, so instead of using the version index we could also refer to it using animals:latest
. You can customize the aliases to apply to a version by passing in aliases=['my-cool-alias']
to log_artifact
.log_artifact.py
again and W&B will take care of the rest. The training scripts will then automatically pull in the latest
version.animals:v0
and animals:v1
, which track the following contents:v1
tracks a total of 6MB worth of files, it only takes up 3MB of space because it shares the remaining 3MB in common with v0
. If you were to delete v1
, you would reclaim the 3MB associated with rat.png
. On the other hand, if you were to delete v0
then v1
would need to inherit the storage costs of cat.png
and dog.png
bringing its size to 6MB.