mradermacher/SabaVL1-2B-GGUF

Same weights, converted (fewer layers) vs Sakalti/SabaVL1-2B. mradermacher/SabaVL1-2B-GGUF matches part of Sakalti/SabaVL1-2B (338 of its 729 tensors; the rest of Sakalti/SabaVL1-2B has no counterpart). On the shared part: Same weights stored in another dtype (F16 → F16/F32).

Tensors changed: Embeddings 0/1, Norms 0/57, MLP 0/84, Attention 0/196

On the Hub · Downloads on Model Pulse