Hacker GitHub

← All repos

joennlae/tensorli

View on GitHub ↗ ★ 251 Python

Absolute minimalistic implementation of a GPT-like transformer using only numpy (<650 lines).

4 mentions on Hacker News by 1 person: 3 stories, 1 comment · first 2023-11-13 · latest 2024-05-16

Mentions per year 4 mentions · busiest year 2023 (3)
20082026
AllStoriesComments NewestTopOldest
  1. Trainable Llama-like transformer (with backpropagation) in numpy only (~600 lines) https://github.com/joennlae/tensorli
    comment by joennlae · 2024-05-16 · in Llama 3 implemented in pure NumPy
  2. ▲ 90
    Show HN: less than 650 LOC trainable GPT only using NumPy
    story by joennlae · 2023-11-17 · 18 comments
  3. ▲ 2
    Show HN: Tensorli – NumPy Only Transformer Training (<650 lines)
    story by joennlae · 2023-11-16 · 0 comments
  4. ▲ 3
    Show HN: Tensorli - GPT-Like Transformers in Pure NumPy
    story by joennlae · 2023-11-13 · 0 comments