tardellirs/embeddinggemma-pt-br

Same weights, repacked vs google/embeddinggemma-300m. Same weights in different files (re-saved, re-sharded or new metadata).

Tensors changed: Norms 0/145, MLP 0/72, Attention 0/96