Downloads · 30 days
18
13% of all-time downloads
RexTRO111/Alisio-25M-Base-v1
Alisio-25M-Base-v1 is a machine learning model from RexTRO111. Use it for the machine learning task on the model card, and read the license before you ship it in a product. The card lists the license as apache-2.0.
WARNING: This model is very biased towards educational content and web articles. It has only ~24M parameters and was trained on only ~550M tokens, which if 550,109,184 tokens ÷ 24,833,280 parameters, = 22.15 tokens pe…
Downloads · 30 days
18
13% of all-time downloads
All-time downloads
134
Public
Parameters
24.8M
99.3 MB on disk
Likes
0
Public
Click a slice to open those files.
.safetensors99.3 MB · 99%
From the Hugging Face model README
WARNING: This model is very biased towards educational content and web articles. It has only ~24M parameters and was trained on only ~550M tokens, which if 550,109,184 tokens ÷ 24,833,280 parameters, = 22.15 tokens per parameter. We have released Alisio-25M, a 24M param SLM trained on 550M tokens of edu. web and HQ articles. It was trained on seven, or 7 FineWeb-Edu snapshots, ranging from December 2024 to June 2025. Note: This is the first version, and the base model, so isn't very good at following instructions unless you use few-shot samples, which isn't also very good for a 24M param model. Training data includes from December 2024 to June 2025, so knowledge cutoff = June 2025.