EleutherAIPile-T5

Pile-T5 XXL

Pile-T5 XXL is an encoder-decoder model trained from scratch. Downloadable weights and intermediate checkpoints support research into text-to-text learning.

Research release

Encoder and decoder structure

XXL uses 24 encoder and 24 decoder layers with a hidden width of 4,096. It is the largest Pile-T5 size. The architecture keeps an explicit encoder rather than turning every input and output into one decoder stream.pile-t5-xxl model configurationpile-t5-xxl model card

An English base for supervised tasks

The model was trained on the deduplicated Pile with span corruption. Its card describes English-language research and later task adaptation. It does not establish general multilingual performance or an instruction-following interface.pile-t5-xxl model card

Training steps and formats

The repository includes 200 checkpoints spaced 10,000 training steps apart. T5x and Transformers files are distributions of this model. Task-specific fine-tunes are linked by the project and should be identified separately when reporting results.pile-t5-xxl model card

Specifications & access

Pile-T5 XXL specifications
DetailValue and conditions
StatusResearch release
Launch date
AccessDownloadable
Model identifiersEleutherAI/pile-t5-xxl
Context window512 tokens

Sequence length used during training; not a documented hard inference limit.

pile-t5-xxl model card
Parameters11,135,426,560pile-t5-xxl model card
Training exposureAbout 2 trillion tokens; 2 million stepspile-t5-xxl model card
LicenseApache 2.0pile-t5-xxl model card

Related models

  • A compact encoder-decoder model trained on two trillion Pile tokens.

    Launch date Research releaseDownloadable
  • A 783-million-parameter encoder-decoder model trained from scratch for text-to-text research.

    Launch date Research releaseDownloadable
  • A 2.85-billion-parameter text-to-text model with an open training history.

    Launch date Research releaseDownloadable