deepvk/USER2-base

Full fine-tune vs deepvk/RuModernBERT-base. Fine-tune: 100% of the weights changed, across 22 of 22 layers.

Tensors changed: Embeddings 2/2, Norms 44/44, Attention 44/44, MLP 44/44

2 derivatives checked

• PruhaNLP/USER2-1C-code: Partial fine-tune
• deepvk/GeRaCl-USER2-base: Full fine-tune

On the Hub · Downloads on Model Pulse