A harness optimized to smaller LLMs
Active Learning for Text Classification in Python
A small proof-of-concept language model (not an LLM) that aims to incorporate latent-space prediction, internal state using recurrent trace units, and byte-by-byte output, built with MLX.
Local LLM Powered Recursive Search & Smart Knowledge Explorer
Survey of Small Language Models from Penn State, ...
[ICML 2026] effGen: Enabling Small Language Models as Capable Autonomous Agents
Doge Family of Small Language Models
"Generative AI in Action" book's code repository
[𝗜𝗖𝗠𝗟 𝟮𝟬𝟮𝟲] Dispersion loss counteracts embedding condensation and improves generalization in small language models
Small Language Model Inference, Fine-Tuning and Observability.
Repository for the companion Colab notebook of the Domain-Specific Small Language Models book.
Local decision model with calibrated probabilities: send a state and yes/no, choice or score questions, get a probability for every option. 0.8B GGUF on CPU, Jev-style API.
This Repository provides a Jupyter Notebook for building a small language model from scratch using 'TinyStories' dataset. Covers data preprocessing, BPE tokenization, binary storage, GPU memory management, and training a Transformer in PyTorch. Generate sample stories to test your model. Ideal for learning NLP and PyTorch.
Wonderful Matrices to Build Small Language Models
Remma-O1: An open-source Language Model with 1.17B Params, built on pytorch from scratch. Work in Progress!!! Open for collaboration.
xLM is a modular, research-friendly framework for developing and comparing non-autoregressive language models. Built on PyTorch and PyTorch Lightning, with Hydra for configuration management, XLM makes it effortless to experiment with cutting-edge NAR LM architectures.
A governed local AI build-and-memory system that trains small brains, compares them, protects the better one, archives the worse one, and preserves the evidence of why. v1.0.0/governed-v2.2.0+
A tiny jev-like model that answers Choice, Score and Noul questions in one forward pass and returns calibrated probabilities. MLX or PyTorch, fully offline, System One compatible.
Readable, composable PyTorch library for language models: build Llama, Qwen, Gemma, DeepSeek, Kimi and 20+ other architectures from plain nn.Modules, then train them on CPU, one GPU, or multi-GPU.
To associate your repository with the small-language-models topic, visit your repo's landing page and select "manage topics."