AI Tools.

Search

fill mask by FacebookAI

roberta-base

RoBERTa base from Facebook AI, trained with the same architecture as BERT base but significantly more data, longer training schedules, larger batch sizes, and dynamic masking. Pre-trained on BookCorpus, Wikipedia, CC-News, OpenWebText, and Stories — substantially more data than the original BERT. MIT licensed with multi-framework support.

Summary text generated by an automated pipeline from the model card · Not individually reviewed or run by us · How this page is made

From the model card

Fields below are copied from the tags and counters on the HuggingFace repository FacebookAI/roberta-base 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)
FacebookAI
Pipeline tag
fill-mask
Library
Transformers
Framework tags
PyTorch, TensorFlow, JAX, Rust (candle)
Weight formats
safetensors
License tag
mit — read the license file in the repo before relying on it
Language tags
English (en)
Papers cited
arXiv:1907.11692, arXiv:1806.02847
Datasets declared
bookcorpus, wikipedia
Downloads (HF counter at last fetch)
9,585,589
Likes (HF counter at last fetch)
640
Model card
https://huggingface.co/FacebookAI/roberta-base

Use cases

  • Fine-tuning for text classification (sentiment analysis, topic detection, intent recognition)
  • Named entity recognition with a token classification head
  • Natural language inference and textual entailment
  • Extractive question answering with span prediction
  • Sentence encoding as a higher-quality alternative to original BERT

Pros

  • More rigorous pre-training than BERT yields better NLU task performance
  • Multi-framework support (PyTorch, TF, JAX, Rust, safetensors)
  • MIT license; large ecosystem of fine-tuned domain-specific variants
  • Well-understood behavior from extensive published NLP research

Cons

  • English-only; no multilingual variant in this checkpoint
  • 512-token context limit requires chunking for long documents
  • Encoder-only architecture cannot generate free-form text
  • Surpassed on most benchmarks by DeBERTa variants and more recent efficient encoders
  • Heavier than distilled alternatives for limited accuracy gains on easy tasks

Tags

transformerspytorchtfjaxrustsafetensorsrobertafill-maskexbertendataset:bookcorpusdataset:wikipediaarxiv:1907.11692arxiv:1806.02847license:mitendpoints_compatibleregion:usdeploy:sagemakerdeploy:azure