EmbeddingGemma 2 for the Apple Neural Engine (ANEEmbed model package) This package contains a Core ML conversion of google/embeddinggemma-2 (https://huggingface.co/google/embeddinggemma-2), Copyright Google LLC / Google DeepMind, licensed under the Apache License, Version 2.0. Use of the model must also follow the Gemma Prohibited Use Policy (https://ai.google.dev/gemma/prohibited_use_policy). Files and their origin: EmbeddingGemma2Text.mlpackage / .mlmodelc converted by tools/convert/build.py from the original model.safetensors (weights stored as fp16) embeddings.bf16 token embedding table extracted unchanged (bf16) from the original model.safetensors tokenizer.json the original tokenizer, unchanged EmbeddingGemma2Vision.mlpackage / .mlmodelc the vision encoder conversion by FluidInference position_embeddings.f16 (https://huggingface.co/FluidInference/embeddinggemma-2-coreml, Apache-2.0), copied unchanged Modifications relative to the original model: re-expression of the text encoder as Core ML functions (fixed-length buckets, per-layer long-context functions), fp16 weights and activations, numerically neutral rewrites listed in the model card. No weights were retrained or altered beyond fp16 rounding.