MoTune mmBERT with Within-Exit Partitioned InfoNCE (WEP-InfoNCE)

Trained checkpoints, metrics, and configuration artifacts for Two-Stream mmBERT with WEP-InfoNCE.

  • Temperature (tau): 0.10
  • Loss Weight: 0.08
  • Base Config: config/target_bottleneck_probing.json
Downloads last month
501
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support