File size: 1,453 Bytes
3c9b6f2 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 | 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.
|