dragonkue/colbert-ko-0.1b

Full fine-tune vs skt/A.X-Encoder-base. Fine-tune: 100% of the weights changed, across 22 of 22 layers.

Tensors changed: Embeddings 1/1, Norms 6/44, Attention 44/44, MLP 44/44

On the Hub · Downloads on Model Pulse