Within an organization, people can create several projects. Typically, a project represents a single model they are
trying to fine-tune, but it doesn’t have to be limited to this use case. Projects provide a way to organize and manage
different initiatives or goals within the organization. Each project can have its own set of datasets, allowing for
multiple strategies of curating datasets for a single model. This structure enables teams to compartmentalize their work
and manage multiple fine-tuning efforts simultaneously.
Within a project, there can be several datasets. Each dataset is a subset of an organization’s samples that a person
wants to generate a fine-tuning dataset from. Datasets allow users to experiments with their curation process and organize
specific collections of samples for different model variants.A key feature of datasets in Alchemist is version control. When you complete a dataset, you can still edit it by adding
and removing samples from the set. Version control enables users to track changes, revert to previous versions if needed,
and maintain a history of dataset iterations.