From the model card
Fields below are copied from the tags and counters on the HuggingFace repository google/electra-base-discriminator at our last fetch. They are set by the uploader, not verified by us; rows with no tag are omitted. How this page is made.
- Publisher (HF namespace)
- Library
- Transformers
- Framework tags
- PyTorch, TensorFlow, JAX, Rust (candle)
- License tag
apache-2.0— read the license file in the repo before relying on it- Language tags
- English (en)
- Papers cited
- arXiv:1406.2661
- Downloads (HF counter at last fetch)
- 56,993,793
- Likes (HF counter at last fetch)
- 158
- Model card
- https://huggingface.co/google/electra-base-discriminator
Use cases
- Fine-tuning for binary or multi-class text classification
- Natural language inference and textual entailment tasks
- NER when combined with a token classification head
- Extractive QA reading comprehension pipelines
- Feature extraction for downstream NLP classification
Pros
- More sample-efficient pre-training yields better performance per parameter vs. BERT
- English language representations from BookCorpus and Wikipedia
- Multi-framework support (PyTorch, TF, JAX, Rust), Apache 2.0 license
- Discriminator head provides richer training signal than masked LM
Cons
- No HuggingFace pipeline_tag means fewer automatic integrations
- Discriminator is not directly usable for text generation tasks
- Smaller community adoption than BERT/RoBERTa, fewer published fine-tuned checkpoints
- English-only; no multilingual pre-training variant at this model ID
- Surpassed by more recent efficient encoders on standard NLU benchmarks
Tags
transformerspytorchtfjaxrustelectrapretrainingenarxiv:1406.2661license:apache-2.0endpoints_compatibleregion:us