| --- |
| license: gemma |
| base_model: google/embeddinggemma-300m |
| pipeline_tag: sentence-similarity |
| tags: [aether-kit, core-ai, text-embeddings, embeddinggemma] |
| --- |
| |
| # EmbeddingGemma 300M for Aether-Kit (Core AI) |
|
|
| [google/embeddinggemma-300m](https://huggingface.co/google/embeddinggemma-300m) at revision `57c266a7`, exported by |
| [Aether-Kit](https://github.com/bubstackdev/aether-kit)'s Forge exporter (`Forge/embeddinggemma/`) as one |
| static Core AI graph: `input_ids` and `attention_mask` (int32, 256 tokens, right-padded with 0) to |
| an L2-normalized 768-dimensional embedding. Mean pooling, the two dense projections and the |
| normalization are in the graph; it computes in float32 (EmbeddingGemma does not support float16 activations) |
| from float16-stored weights. Prefix queries with `task: search result | query: ` and documents with `title: none | text: `. |
| Load it with Aether-Kit's `AetherEmbeddings` (`text-embedder@1`, macOS); the |
| manifest is `.aether/aether.json`. Gated against the upstream implementation on CPU in float32 (see |
| `export-report.json`). |
|
|
| Gemma is provided under and subject to the Gemma Terms of Use found at |
| [ai.google.dev/gemma/terms](https://ai.google.dev/gemma/terms) (`LICENSE`). Use is restricted by the Gemma |
| Prohibited Use Policy (`GEMMA-PROHIBITED-USE-POLICY.txt`). A modified version, not an official Google |
| release (see `NOTICE`). |
|
|
