Hogleto
Hogleto is a 1.4M param model finetuned from BananaMind 2.1 Pico Preview. This model is an artifact from testing out the BananaAll app
Setup
LoRA was used to train this model, everything was kept as default unless the setting is listed below.
LoRA rank: 8 LoRA alpha: 16 Dataset: SupraThink Training steps: 5
Benchmarks
| Benchmark | Accuracy | Random Baseline |
|---|---|---|
| PIQA | 53.86% | 50.00% |
| ARC-Easy | 29.55% | 25.00% |
| ARC-Challenge | 22.61% | 25.00% |
| HellaSwag | 27.03% | 25.00% |
| ArithMark-3.0 | 32.50% | 25.00% |
- Downloads last month
- 279
Model tree for Hoglet-33/Hogleto
Base model
BananaMind/BananaMind-2.1-Pico-Preview