steps.clip.model.trainer

trainer

Functions:

Name	Description
train	Training step for CLIP using the Picsellia training engine.

train(model, dataset_collection)

Training step for CLIP using the Picsellia training engine.

This step uses BLIP to generate captions for the dataset and runs CLIP fine-tuning using a CLI script.

Parameters:

Name	Type	Description	Default
model	CLIPModel	CLIP model instance to be trained.	required
dataset_collection	DatasetCollection[CocoDataset]	Dataset collection containing train/val/test splits.	required

Returns:

Type	Description
CLIPModel	The trained CLIP model with updated weights.