AI Tools.

Search

fill mask by google-bert

bert-base-uncased

Google's original BERT base model in uncased form, pre-trained on BookCorpus and English Wikipedia via masked language modeling. Tokens are lowercased before processing, making it insensitive to capitalization. It remains a standard fine-tuning base for classification, NER, and extractive QA, though newer encoders outperform it on most benchmarks.

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 google-bert/bert-base-uncased 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)
google-bert
Pipeline tag
fill-mask
Library
Transformers
Framework tags
PyTorch, TensorFlow, JAX, Rust (candle)
Weight formats
Core ML, ONNX, safetensors
License tag
apache-2.0 — read the license file in the repo before relying on it
Language tags
English (en)
Papers cited
arXiv:1810.04805
Datasets declared
bookcorpus, wikipedia
Downloads (HF counter at last fetch)
58,556,227
Likes (HF counter at last fetch)
2,910
Model card
https://huggingface.co/google-bert/bert-base-uncased

Use cases

  • Fine-tuning for text classification (sentiment, topic, intent)
  • Named entity recognition with a token classification head
  • Extractive question answering on short passages
  • Sentence embedding via mean pooling of hidden states
  • Transfer learning starting point for domain-specific NLP tasks

Pros

  • Extensively benchmarked — failure modes and quirks well documented
  • Multi-framework support: PyTorch, TensorFlow, JAX, CoreML, ONNX, Rust
  • Apache 2.0 license; large ecosystem of domain-specific fine-tuned checkpoints
  • Low barrier for integration in HuggingFace-based pipelines

Cons

  • Lowercase tokenization breaks case-sensitive tasks like proper noun NER
  • 512-token context window insufficient for long documents without chunking
  • Encoder-only architecture cannot generate free-form text
  • Outperformed by DeBERTa and more recent encoders on most NLU benchmarks
  • No multilingual capability in the base checkpoint

Tags

transformerspytorchtfjaxrustcoremlonnxsafetensorsbertfill-maskexbertendataset:bookcorpusdataset:wikipediaarxiv:1810.04805license:apache-2.0endpoints_compatibledeploy:sagemakerdeploy:azureregion:us