Dumb / README.md
56m's picture
Update README.md
d12b02d verified
metadata
license: apache-2.0
datasets:
  - trl-internal-testing/dolly-chatml-sft
  - HuggingFaceTB/smoltalk
  - diwank/expertllama-chatml
  - HuggingFaceTB/smollm-corpus
language:
  - en
pipeline_tag: text-generation
tags:
  - dumb
  - Asperger
  - LLM
  - SLM

THANKS 200+ download!!! Please Share, Please share this so that it can be used to record further downloads!!!!

  • 2026/05/06 132 DL!
  • 2026/05/07→182 DL!
  • 2026/05/09→240 DL!!!

dumb-7M: world's smallest LLM?

safetensors size: 31.5 MB, FP32

benchmarks(0-shot)

  • MMLU: 22.94%
  • GSM8K(500 problem): 1.2%

benchmark comparison

the goal

I'm creating an DUMBER Large Language Model. In other words, I made idiot Intelligence.

training

benchmark comparison


model family

  • Dumb-7M: Minimal Fucking models

  • Dumb-25M: Mini Fucking models → Dumb 1.1(19M)!

auuthor: 56m