KKHYA/llavaqwen3-1.7b-finetune

Full fine-tune (fewer layers) vs Qwen/Qwen3-1.7B. Qwen/Qwen3-1.7B matches part of KKHYA/llavaqwen3-1.7b-finetune (311 of its 706 tensors; the rest of KKHYA/llavaqwen3-1.7b-finetune has no counterpart). On the shared part: Fine-tune: 100% of the weights changed, across 28 of 28 layers.

Tensors changed: Embeddings 1/1, Norms 108/113, MLP 84/84, Attention 112/112, lm_head 1/1

1 derivatives checked

• KKHYA/llavaqwen3-1.7b-finetune-nm-mask-moe-sparse-4e-2k-4of8-imp-anchor_20260812_071057: Partial fine-tune

On the Hub · Downloads on Model Pulse