XLM-R
E435876
XLM-R is a multilingual transformer-based language model (XLM-RoBERTa) designed for cross-lingual understanding and natural language processing across many languages.
All labels observed (1)
| Label | Occurrences |
|---|---|
| XLM-R canonical | 1 |
Statements (48)
| Predicate | Object |
|---|---|
| instanceOf |
XLM-RoBERTa architecture
ⓘ
masked language model ⓘ multilingual language model ⓘ transformer-based model ⓘ |
| architecture | Transformer NERFINISHED ⓘ |
| basedOn | RoBERTa NERFINISHED ⓘ |
| compatibleWith | Hugging Face Transformers NERFINISHED ⓘ |
| designedFor |
cross-lingual generalization
ⓘ
multilingual NLP ⓘ |
| developedBy |
Facebook AI
NERFINISHED
ⓘ
Meta AI NERFINISHED ⓘ |
| family |
RoBERTa family
NERFINISHED
ⓘ
XLM family NERFINISHED ⓘ |
| handlesScriptTypes |
Arabic
ⓘ
CJK ⓘ Cyrillic ⓘ Devanagari NERFINISHED ⓘ Latin ⓘ |
| hasEncoderLayersApprox |
12 (base variant)
ⓘ
24 (large variant) ⓘ |
| inputType | text ⓘ |
| languageModelType | encoder-only transformer ⓘ |
| pretrainingDataType | CommonCrawl NERFINISHED ⓘ |
| pretrainingObjective | masked language modeling ⓘ |
| primaryUseContext |
production NLP systems
ⓘ
research ⓘ |
| releasedAs | open-source model ⓘ |
| releasedByOrganizationType | industry research lab ⓘ |
| supportsLanguages | multilingual ⓘ |
| supportsLanguagesCountApprox |
100+
ⓘ
over 100 languages ⓘ |
| supportsTask |
cross-lingual transfer learning
ⓘ
cross-lingual understanding ⓘ multilingual representation learning ⓘ named entity recognition ⓘ natural language processing ⓘ question answering ⓘ sentence embedding ⓘ sequence labeling ⓘ text classification ⓘ token-level classification ⓘ zero-shot cross-lingual transfer ⓘ |
| tokenizationMethod | SentencePiece NERFINISHED ⓘ |
| usesPositionalEncoding | true ⓘ |
| usesSelfAttention | true ⓘ |
| usesSubwordTokenization | true ⓘ |
| variant |
XLM-R-base
NERFINISHED
ⓘ
XLM-R-large NERFINISHED ⓘ |
Referenced by (1)
Full triples — surface form annotated when it differs from this entity's canonical label.