smol_llama: 220M GQA

A small 220M param (total) decoder model. This is the first version of the model.

  • 1024 hidden size, 10 layers
  • GQA (32 heads, 8 key-value), context length 2048
  • train-from-scratch on one GPU :)

Links

Here are some fine-tunes we did, but there are many more possibilities out there!

  • instruct
    • openhermes - link
    • open-instruct - link
  • code
    • python (pypi) - link
  • zephyr DPO tune

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric Value
Avg. 29.44
AI2 Reasoning Challenge (25-Shot) 24.83
HellaSwag (10-Shot) 29.76
MMLU (5-Shot) 25.85
TruthfulQA (0-shot) 44.55
Winogrande (5-shot) 50.99
GSM8k (5-shot) 0.68

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric Value
Avg. 6.62
IFEval (0-Shot) 23.86
BBH (3-Shot) 3.04
MATH Lvl 5 (4-Shot) 0.00
GPQA (0-shot) 0.78
MuSR (0-shot) 9.07
MMLU-PRO (5-shot) 1.66
Downloads last month
2,614
Safetensors
Model size
218M params
Tensor type
BF16
·
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.

Model tree for BEE-spoke-data/smol_llama-220M-GQA

Finetunes
12 models
Merges
1 model
Quantizations
3 models

Datasets used to train BEE-spoke-data/smol_llama-220M-GQA

Collection including BEE-spoke-data/smol_llama-220M-GQA

Evaluation results