An edge-native, 162M-parameter progressive language model architecture engineered from scratch on consumer compute.
machine-learning deep-learning models transformers project language-model from-scratch model-training chinchilla pretraining instruction-tuning supervised-finetuning local-llm ollama chatml gguf continual-pre-training small-llms rope-embeddings ownllm
-
Updated
Sep 24, 2026 - Jupyter Notebook