neich-cereales/sems-embeddinggemma-2-onnx

EmbeddingGemma 2 for sems (ONNX)

google/embeddinggemma-2 exported to ONNX for sems, a local semantic search CLI. sems downloads these files on first use; you do not need to fetch them yourself.

File	Contents
token_embedder.onnx	token ids -> input embeddings
text_encoder.onnx	input embeddings -> pooled, normalized embedding
vision_encoder.onnx	image and video-frame patches -> soft tokens
audio_encoder.onnx	30 s log-mel features -> soft tokens
config.json, tokenizer.json	from the original model

Weights are stored in float16 and cast to float32 when the graph loads, so inference runs in float32. Every graph reproduces the official sentence-transformers embeddings with cosine similarity of at least 0.9999 (measured 1.000000 at six decimals). The export scripts are in tools/export.

These files are a derivative of google/embeddinggemma-2 by Google DeepMind and are distributed under the same Apache 2.0 license.