SimCLRConfig#
- class stable_ssl.joint_embedding.SimCLRConfig(name: str = 'Supervised', backbone_model: str = 'resnet18', sync_batchnorm: bool = False, memory_format: str = 'channels_last', pretrained: bool = False, with_classifier: bool = True, projector: list[int] = <factory>, temperature: float = 0.15)[source]#
Bases:
JEConfig
Configuration for the SimCLR model parameters.
- Parameters:
temperature (float) – Temperature parameter for the contrastive loss. Default is 0.15.