jcbtc/Ling-3.0-Flash-CIRU-IU4

Partial fine-tune vs inclusionAI/Ling-3.0-flash-int4. Partial change: 10% of the weights changed (mlp), in 40 of 43 layers.

Tensors changed: MLP 1514/2404, Attention 0/460, lm_head 0/1, Norms 0/133, other 0/1, Embeddings 0/1

On the Hub · Downloads on Model Pulse