yacdev/SomaliLM-150M-Pretrained

🧠 SomaliLM-150M (Pretrained Foundation Model)

SomaliLM-150M is a compact, high-efficiency 147M parameter LLaMA-based language model pretrained on ~370 Million tokens of high-quality educational data (FineWeb-Edu).

Model Architecture

• —Parameters: 147.28 Million
• —Layers: 12
• —Hidden Dim: 768
• —Heads: 12
• —Vocab Size: 32,000
• —Context Length: 2048 tokens

Try It Live

Use the interactive Inference API widget on the right sidebar!