SandLogicTechnologies/mage-vl-GGUF

Quantization (fewer layers) vs microsoft/Mage-VL. SandLogicTechnologies/mage-vl-GGUF matches part of microsoft/Mage-VL (399 of its 696 tensors; the rest of microsoft/Mage-VL has no counterpart). On the shared part: SandLogicTechnologies/mage-vl-GGUF is a GGUF quantization (Q6_K) of the same weights: 145 tensors match exactly and 254 within quantization error.

Tensors changed: Embeddings 0/1, Norms 0/145, MLP 0/108, Attention 0/144, lm_head 0/1

On the Hub · Downloads on Model Pulse