Together ORG

No description.

OpenChatKit is a toolkit for running and fine-tuning open-source conversational AI models on your own hardware, including a 20-billion-parameter chat model and support for retrieval-augmented responses from custom documents.

The RedPajama-Data repository contains code for preparing large datasets for training large language models.

Together Mixture-Of-Agents (MoA) – 65.1% on AlpacaEval with OSS models

A collection of notebooks/recipes showcasing usecases of open-source models with Together AI.

Repository for StripedHyena, a state-of-the-art beyond Transformer architecture

Together Open Deep Research

Open AI data scientist agent that automates complex data analysis tasks using the ReAct framework. Execute Python code locally or in the cloud, upload datasets, and generate detailed analytical reports with minimal setup.

No description.

New state-of-the-art bounds for open problems

Finetune Llama-3-8b on the MathInstruct dataset

Extend the original llama.cpp repo to support redpajama model.

No description.

The Official Python Client for Together's API

No description.

A RAG Chatbot with Next.js, Together.ai and Llama Index

No description.

The official Together AI TypeScript library.

Fast and memory-efficient exact attention

Skills to help your coding agents use Together AI products.

Official implementation of Paper "System-Aware 4-Bit KV-Cache Quantization for Real-World LLM Serving"

No description.

XoRL

No description.

CREST is a training-free test-time steering framework that discovers cognitive heads via simple offline calibration and then rotates activations during decoding to guide the model’s reasoning—preserving norms to avoid per-model hyperparameter tuning. This improves accuracy and reduces tokens across models and datasets.

🤗 Transformers: State-of-the-art Machine Learning for Pytorch, TensorFlow, and JAX.

No description.

🤗 Diffusers: State-of-the-art diffusion models for image and audio generation in PyTorch

No description.

synthetic data pipeline for multi-image reasoning

Universal Stable Diffusion Pipeline(s) with Flash Attention

Transformer related optimization, including BERT, GPT

SDKs and CLIs for working with Together Sandboxes

A lightweight inference engine supporting speculative speculative decoding (SSD).

Streamlit Component, for a Chatbot UI

A high-throughput and memory-efficient inference and serving engine for LLMs

An implementation of model parallel autoregressive transformers on GPUs, based on the DeepSpeed library.

Official Together AI Go library

Elixir Client for the Stytch B2B SaaS authentication API

Sprocket SDK for building inference workers on Together Dedicated Containers

Autoscaling components for Kubernetes

AWQ: Activation-aware Weight Quantization for LLM Compression and Acceleration

⚡ Building applications with LLMs through composability ⚡

A high-throughput and memory-efficient inference and serving engine for LLMs

Fast and memory-efficient exact attention

Transformer related optimization, including BERT, GPT

Together port to run H3

Transformer related optimization, including BERT, GPT

Transformer related optimization, including BERT, GPT

No description.

No description.

Open Source Continuous Inference Benchmark Research Platform Kimi K2.6, DeepSeekv4, GLM5 - GB200 NVL72 vs MI355X vs B200 vs GB300 NVL72 & soon™ TPUv6e/v7/Trainium2/3

SGLang is a high-performance serving framework for large language models and multimodal models.

Shared modules and utilities for Elixir services at Together

Post-training with Tinker

Archipelago: eval framework for AI agents on professional services tasks

No description.

No description.

No description.

This project provides a framework that runs Slurm in Kubernetes.

No description.

fast bpe tokenizer

Private fork of https://github.com/All-Hands-AI/OpenHands.git

Running Networking Performance Tests against K8s

DeepGEMM: clean and efficient FP8 GEMM kernels with fine-grained scaling

Model components of the Llama Stack APIs

SGLang is a high-performance serving framework for large language models and multimodal models.

Pre-compiled wheels

Mooncake is the serving platform for Kimi, a leading LLM service provided by Moonshot AI.

No description.

XoRL Client

VeOmni: Scaling Any Modality Model Training with Model-Centric Distributed Recipe Zoo

NCCL Tests

DGXC Benchmarking provides recipes in ready-to-use templates for evaluating performance of specific AI use cases across hardware and software combinations.

Slurm debian packages

Democratizing Reinforcement Learning for LLMs

Genai-bench is a powerful benchmark tool designed for comprehensive token-level performance evaluation of large language model (LLM) serving systems.

PyTorch native post-training library

Save core dumps from a Kubernetes Service or RedHat OpenShift to an S3 protocol compatible object store

Modeling, training, eval, and inference code for OLMo

Gorilla: Training and Evaluating LLMs for Function Calls (Tool Calls)

Dockerized FastAPI wrapper for Kokoro-82M text-to-speech model w/CPU ONNX and NVIDIA GPU PyTorch support, handling, and auto-stitching

A PyTorch native platform for training generative AI models

A CoreDNS plugin to resolve all types of external Kubernetes resources

A CoreDNS plugin to resolve all types of external Kubernetes resources

Kubernetes Virtualization API and runtime in order to define and manage virtual machines.

Kubernetes Virtualization API and runtime in order to define and manage virtual machines.

Operator for provisioning and configuring SR-IOV CNI plugin and device plugin

Scalable and flexible workflow orchestration platform that seamlessly unifies data, ML and analytics stacks.

Mirror of FreeIPA, an integrated security information management solution

Cluster API Provider for KubeVirt

An Ansible playbook for installing and configuring MAAS

🚀 A simple way to launch, train, and use PyTorch models on almost any device and distributed configuration, automatic mixed precision (including fp8), and easy-to-configure FSDP and DeepSpeed support

EventSource client for Node.js and Browser (polyfill)

scalable and robust tree-based speculative decoding algorithm

TensorRT-LLM provides users with an easy-to-use Python API to define Large Language Models (LLMs) and build TensorRT engines that contain state-of-the-art optimizations to perform inference efficiently on NVIDIA GPUs. TensorRT-LLM also contains components to create Python and C++ runtimes that execute those TensorRT engines.

Holistic Evaluation of Language Models (HELM), a framework to increase the transparency of language models (https://arxiv.org/abs/2211.09110).

A framework for few-shot evaluation of autoregressive language models.

Transformer related optimization, including BERT, GPT