qing-yao/ppt-pythia-160m-uniform250-previous_mse_delta_shuffle2-seed1024-stage2

Full fine-tune vs qing-yao/ppt-pythia-160m-uniform250-previous_mse-seed1024-stage1. Fine-tune: 100% of the weights changed, across 12 of 12 layers.

Tensors changed: Norms 47/50, Attention 48/48, MLP 48/48

On the Hub · Downloads on Model Pulse