Model reference · open weights
ru-en-RoSBERTa is an open-weight embedding model from ai-forever, listed in the AxForge catalogue. AxForge can bring it up on EU-owned hardware for you on request — with the licence handled where one is required.
About
Model Card for ru-en-RoSBERTa The ru-en-RoSBERTa is a general text embedding model for Russian. The model is based on ruRoBERTa and fine-tuned with ~4M pairs of supervised, synthetic and unsupervised data in Russian and English. Tokenizer supports some English tokens from RoBERTa tokenizer. For more model details please refer to our article. Usage The model can be used as is with prefixes. It is recommended to use CLS pooling. The choice of prefix and pooling depends on the task. We use the following basic rules to choose a prefix: - "searchquery: " and "searchdocument: " prefixes are for answer or relevant paragraph retrieval - "classification: " prefix is for symmetric paraphrasing related tasks (STS, NLI, Bitext Mining) - "clustering: " prefix is for any tasks that rely on thematic features (topic classification, title-body retrieval) To better tailor the model to your needs, you can fine-tune it with relevant high-quality Russian and English datasets. Below are examples of texts encoding using the Transformers and SentenceTransformers libraries. Transformers SentenceTransformers or using prompts (sentence-transformers=2.4.0): Citation Limitations The model is designed to process texts in Russian, the quality in English is unknown. Maximum input text length is limited to 512 tokens.
Summarised from the published model card. Read the full card on the HuggingFace links below.
Specifications
| Maker | ai-forever |
|---|---|
| Type | Embedding models |
| Parameters (lead) | 404M |
| Context | 514 tokens |
| Variants | 1 |
| Runs with | sentence-transformers |
| Based on | ai-forever/ruRoberta-large |
| Released | 2024-07-29 |
| Popularity | 143k downloads / month |
| Likes | 89 |
| Licence | Open weights |
How it works
Variants
Open weights ship in several sizes and precisions. One page, all the variants — pick the one that fits your GPU. VRAM figures are estimates from model size.
| Variant | Params | Precision | VRAM | Fits 16 GB | Weights |
|---|---|---|---|---|---|
| ru-en-RoSBERTa | 404M | BF16 | ~0.9 GB | ✓ | Weights ↗ |
Benchmarks
As published on the model card — the maker's own numbers, not measured by AxForge.
| Task | Dataset | Metric | Score |
|---|---|---|---|
| MultilabelClassification | MTEB CEDRClassification (default) | accuracy | 44.687 |
| MultilabelClassification | MTEB CEDRClassification (default) | f1 | 40.76 |
| MultilabelClassification | MTEB CEDRClassification (default) | lrap | 70.696 |
| MultilabelClassification | MTEB CEDRClassification (default) | main_score | 44.687 |
| Classification | MTEB GeoreviewClassification (default) | accuracy | 49.697 |
| Classification | MTEB GeoreviewClassification (default) | f1 | 47.793 |
| Classification | MTEB GeoreviewClassification (default) | f1_weighted | 47.791 |
| Classification | MTEB GeoreviewClassification (default) | main_score | 49.697 |
| Clustering | MTEB GeoreviewClusteringP2P (default) | main_score | 65.422 |
| Clustering | MTEB GeoreviewClusteringP2P (default) | v_measure | 65.422 |
| Clustering | MTEB GeoreviewClusteringP2P (default) | v_measure_std | 0.852 |
| Classification | MTEB HeadlineClassification (default) | accuracy | 78.003 |
| Classification | MTEB HeadlineClassification (default) | f1 | 77.952 |
| Classification | MTEB HeadlineClassification (default) | f1_weighted | 77.951 |
| Classification | MTEB HeadlineClassification (default) | main_score | 78.003 |
| Classification | MTEB InappropriatenessClassification (default) | accuracy | 61.323 |
| Classification | MTEB InappropriatenessClassification (default) | ap | 57.11 |
| Classification | MTEB InappropriatenessClassification (default) | ap_weighted | 57.11 |
| Classification | MTEB InappropriatenessClassification (default) | f1 | 60.971 |
| Classification | MTEB InappropriatenessClassification (default) | f1_weighted | 60.971 |
| Classification | MTEB InappropriatenessClassification (default) | main_score | 61.323 |
| Classification | MTEB KinopoiskClassification (default) | accuracy | 63.273 |
| Classification | MTEB KinopoiskClassification (default) | f1 | 61.007 |
| Classification | MTEB KinopoiskClassification (default) | f1_weighted | 61.007 |
Using it via the API
Once AxForge deploys ru-en-rosberta for you, it answers on the OpenAI-compatible API — the same base URL and keys as every other model. (ru-en-rosberta below is illustrative; you get the exact model name on deployment.)
$ curl -sS https://api.axforge.ai/v1/embeddings \
-H "Authorization: Bearer $AXFORGE_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model":"ru-en-rosberta","input":"text to embed"}'
Licence
Open weights under mit — commercial use is permitted. Deploy it on AxForge EU hardware on request. Read the licence ↗
Explore