bencyc1129/mitre-bert-base-cased
Full fine-tune vs google-bert/bert-base-cased. Fine-tune: 100% of the weights changed, across 12 of 12 layers.
Tensors changed: Embeddings 1/2, Attention 96/96, other 50/50
1 derivatives checked
- sarahwei/MITRE-v16-tactic-bert-case-based: Full fine-tune