arrow_backRetour aux issues
goldener-data/goldener
#229
Débutant
Ouvrirarrow_forward
Débutant
Ouvrirarrow_forward
Débutant
Ouvrirarrow_forward
Add vector transformation in the different `GoldDoer`
ecoDébutant
good first issue
help wanted
ready
descriptionDescription
**Context & Problem**
Depending on the use case and embeddings, we might want to transform the features before some actions. For instance, it can be a vector normalization (having channel vectors with norm of 1), or compute residual from prototypes, ... etc
There are multiple locations in the code where the transformation can be done:
- `GoldDescriptor`: before storing the embeddings in the local table
- `GoldVectorizer`: before storing the vectors in the local table
- `GoldSelector`: before making the selection among the vectors
- `GoldClusterizer`: before clustering the vectors
**Proposed Solution**
Like for the feature reduction, we create a base class called GoldTransformationTool. Then in all GoldDoers listed above we add it as an attribute. Then inside the computation functions, we transform the embeddings before the proper action.
As a first usable tool, we can add the normalization tool.
Issues similaires
calkit/calkit
star53
Poids du dépôt moyen
VS Code extension should be robust to YAML parser errors
Seeing this error: ``` Failed to read calkit.yaml: YAMLParseError: A block sequence may not be used as an implicit map…
Python
bug
good first issue
fu351/Doberman-Core
star211
Poids du dépôt léger
dash: a manual Refresh control
The dashboard polls: `refreshStats()` (`src/doberman/dash/app.py:408`) every 5 s and `refreshPending()` (`:546`) every …
Python
enhancement
good first issue
fu351/Doberman-Core
star211
Poids du dépôt léger
dash: "Copy details" button on each pending-approval card
Each pending-approval card in the dashboard (`renderPending`, `src/doberman/dash/app.py:448-544`) shows the risk badge,…
Python
enhancement
good first issue