Agent frameworks & infrastructure
Models, fine-tuning & training
2,504 agents · most starred first
Models, fine-tuning & training, page 25 of 26
Thai large language model(LLM)
andromeda - Pytorch
Python tic tac toe state generator and GPT fine tuning
TILEARN for LLM
Intelligent agent for Anthropic's Tinker fine-tuning platform
Model classes and pre-training utilities for a tiny version of Llama in PyTorch.
Token Shift GPT - Pytorch
Easy Deep Reinforcement Learning for PyTorch
GPT-2 implemented in PyTorch
Object-oriented RL agents based on torchrl.
Toy GPT next-token prediction using a unigram model.
Toy GPT next-token prediction using a bigram model.
Toy GPT next-token prediction using a 2-token context window.
Toy GPT next-token prediction using a 3-token context window.
Toy LLM
Fine-tuning LLMs for instruction using QLoRA.
A modular PyTorch framework for fine-tuning Hugging Face models.
transnormerllm
Granite TSFM toolkit, model wrappers, and demos for time-series forecasting workflows.
TimesFM fork distribution for tsagentkit.
Unified Training of Universal Time Series Forecasting Transformers
its an implimentation of Perceptron
Type annotations for boto3 Bedrock 1.44.0 service generated with mypy-boto3-builder 8.12.0
ULLME: A Unified Framework for Large Language Model Embeddings with Generation-Augmented Learning
An automatic gradient engine for training your all in one deep learning model
Unit averaging in Python: efficient ensemble estimation of unit-specific parameters
Library to mitigate verbatim or near-verbatim memorization in LLMs
Short description of your package
语言模型中文识字率分析
Weather Swarm - Pytorch
Explore to building the intelligence easily! It's based on Torch, Transformers, etc.
Implementations of deep reinforcement learning algorithms in tensorflow 2.5
Simple & Cutting Edge LLM Finetuning
Automation for TTNN model bringup.
Create AI models instantly with plain English, no code, no fuss.
Ark7 model used for both backend and frontend
Build, fine-tune, and deploy custom AI models for better performance and lower costs.
Turn your LangGraph agent into a small fine-tuned model that runs with no orchestrator - near-frontier quality at a fraction of the inference cost.
AgentCamp SDK — where agents train their own experts. Observe, train, shadow, and activate cheaper expert models behind your existing agents.
TypeScript/JavaScript utilities for the GPT‑OSS Harmony format: renderers, parsers, tokenizers, and streaming helpers
Advanced LLM architecture augmentation library - surgical model expansion with minimal compute
Seren: A Self-Refining Decoder-Only LLM with GQA, RoPE, Flash Attention, and a novel two-pass self-refinement mechanism.
Streaming JSONL cleaner for LLM fine-tuning datasets.
A local toolkit for training tiny GPT models on your own data.
Building Language Models from Scratch — tokenizer, Transformer, training, inference, and fine-tuning in clean, tested PyTorch.
ViLLM Fast Tokenizer — Rust backend with embedded SentencePiece for Vietnamese-English code-switching
Fast C-backed trigram language model for word prediction and sentence completion
Curriculum learning for agent training — difficulty-scored stages
A lightweight, trainable language model SDK
ICONOCLAST — Discriminative representation editing for open-weight LLMs. Beats HERETIC baseline across all tested models.
Sample responses and log-probabilities from HuggingFace causal LMs
Nous Research is an AI research team focused on advancing open-source AI.
A clean, efficient framework for pretraining language models from scratch
Fine-tune LLMs (LoRA/QLoRA) and image classifiers locally — browser UI + optional TUI
Pull open-weight AI models from the ossllms torrent network. An HF-cache-compatible download shim for common huggingface_hub public model pulls.
Estimators for skill formation models
Mathieu Group M23 Weight Initialization for PyTorch Transformers and LLMs
Fine-tune cheap open-source LLMs (GLM, Qwen, DeepSeek) on your AI coding agent's successful runs with LoRA (SFT + DPO) so your model cascade escalates to expensive frontier models (GPT, Claude) less often — cutting cost-per-resolved. Turns run history int
aihu compiler native addon (napi) — linux-x64-gnu platform binary.
aihu compiler native addon (napi) — win32-x64-msvc platform binary.
aihu compiler native addon (napi) — darwin-arm64 platform binary.
Open-source, local chat with unrestricted small language models (ablate chat), plus a directional ablation (abliteration) toolkit to remove refusal from open LLMs.
Gradient descent, but the parameters are agents — a parallel, asynchronous framework for self-evolving agents (skills, prompts, harnesses)
Canonical PyTorch implementation of Gemma 4
LLM Surgery
AI-Native LLM Training & Infrastructure Abstraction Library
Fine-tune and serve small LLMs on your laptop
Ternary Additive Contrastive LLM — CPU-friendly integer neural network
A pure-Python (+numpy) byte-graph that is a 1-bit (ternary) LLM: dunder API, autograd, ultra-edges, visualization.
Modular Transformer & Multi-Corpus (English, Arabic, Code) LLM Library
Offline BPE tokenizer for OpenAI, Anthropic, and Gemini — zero dependencies
Native, in-process reinforcement learning SDK for AI agents — in-memory and local-file storage by default.
Default model selection shared by Agent entry points
Exact tokenizer and chat-template frontend for TuiskoLLM
A drop-in NSGA-II replacement whose authored machinery installs only when measurement says it beats the classical path
Easy data distillation of LLMs for text classification, information extraction, and open-ended tasks
RL-Split decision plane: masked PPO/TD3/bandit agents over the decision-parameter contract
Capture LLM traffic through a drop-in proxy, curate it into a dataset, and distill a small model that replaces the frontier one.
Self-improving agents through continual learning: in-process agentic RL that turns your own agent's work into training data. Observed tool outcomes, outbound-only, one runtime dependency.
Structured probabilistic decision inference for local causal language models using direct candidate scoring.
Use TypeSafe classification and scoring models with LLM
A simple high-level AI toolkit with scratch-trained LLM, image, video, and TTS models plus easy pretrained wrappers.
Highly optimized distributed training for diffusion language models.
AgentTune: build agentic workflows, then train, evaluate, distill, and self-heal them through one trajectory schema. A TRL-backed alignment + fine-tuning library (full RL coverage — DPO, PPO, GRPO) with an agentic integration spine on top: agent-design / memory-design / harness-design pillars, a Project lifecycle, and a FastAPI service.
A developer-first PyTorch engine for explicit, modular LLM training. Zero magic.
A small Transformer written from scratch in NumPy, built to be looked at.
Name reserved for jepa-agent. jepa-agent: training and planning with action-conditioned JEPA world models.
Unified Efficient Fine-Tuning of 100+ LLMs
Advanced LLM Fine-Tuner — SFT, DPO, ORPO, KTO, GRPO, reward modelling and evaluation
General-purpose, multi-objective hyperparameter optimizer for LLM training, with a pluggable model adapter interface.
Open-source framework for cheaper, faster, energy-efficient LLMs. Know before you compress.
LoRA fine-tuning, but more aggressively
RUMI-8M small decoder-only Transformer language model
Distill an LLM agent's closed-set decisions into cheap tabular classifiers: TabSchema, TabSynth, TabHead.
Bedrock-TriCore: An Attention-Free, Constant-State O(1) Sequence Operator (17.00 KiB) with Causal Delta-Rule Fast Weights and Latent Multi-Hop Chaining
Training-free amplitude outlier suppression for Transformer Q/O projections, with a measured cost.
Recursive Language Models for efficient long-context processing