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.