Nikolai Varankine - "Multiplying Table LLM" Research Project
 Introduction   Projects   Contacts   About Me   News   LinkedIn™   Quora™ 

Application Code and Model Data

  • Python project (PyCharm™, PyTorch™) to learn multiplication and test the model

  • Playground


    Listed below are models exported into XLSX files. Calculator is located on the first tab. Type your numbers and verify result. You may want to open any other tab and examine every matrix in place.
    • 🧮 Standard Transformer XLSX (15.6 MB)
    • 🧮 Transformer ended by Binary Decoder XLSX (781 KB)
    • 🧮 B-Dec Transformer with TANH XLSX (16.1 MB)
  • Project notes:

    1. This is NOT a general purpose LLM! It knows nothing about Shakespeare. It knows only table of multiplication learned at school.
    2. The model doesn't know about zero. If you try to nullify something, the model will report wrong answer.
    3. Microsoft® Excel is required to play with exported model.
    4. CUDA™ and Excel™ crunch numbers differently. You may find noticeable mismatches at every tab versus binary model.
    5. Positional encoder isn't implemented.
    6. Analytical tabs in the XLSX use 3rd party plugin. These tabs are optional.
    7. All provided materials are covered by MIT license.
    8. This is a work in progress. No warranties provided, of any sort.
Sections