Issues

Issues

• Fix scaling in attention computation

  Bug

  PAIR-code/tiny-transformers#71 · opened on Apr 29, 2025

• Support running scripts with Nodejs or Deno, including webgpu support

  PAIR-code/tiny-transformers#62 · opened on Feb 19, 2025

• Add support for cross-entropy loss with integer labels

  PAIR-code/tiny-transformers#60 · opened on Feb 17, 2025

• Refactor transformer_gtensor.ts and gpt2.ts to share functions in a common file

  PAIR-code/tiny-transformers#54 · opened on Jan 16, 2025

• Add GPT2 tokenizer

  PAIR-code/tiny-transformers#52 · opened on Jan 16, 2025

• Fix typos in the codebase

  PAIR-code/tiny-transformers#51 · opened on Jan 16, 2025

• Support loading/saving of models from h5 files

  PAIR-code/tiny-transformers#41 · opened on Nov 5, 2024

• Create patchscopes module

  PAIR-code/tiny-transformers#40 · opened on Nov 2, 2024

• Pass dropout once for whole neural-network

  PAIR-code/tiny-transformers#38 · opened on Oct 14, 2024

• Create a Gemma compatible transformer implementation

  PAIR-code/tiny-transformers#30 · opened on Sep 18, 2024

• Provide a visualization for the norms of different parts of a transformer model

  PAIR-code/tiny-transformers#25 · opened on Aug 28, 2024