The AI developer platform. Use Weights & Biases to train and fine-tune models, and manage models from experimentation to production.

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, 17+ clouds, or on-prem).

Accessible large language models via k-bit quantization for PyTorch.

Utilities intended for use with Llama models.

[ICLR 2024] Efficient Streaming Language Models with Attention Sinks

An interactive NVIDIA-GPU process viewer and beyond, the one-stop solution for GPU process management.

Simple and efficient pytorch-native transformer text generation in <1000 LOC of python.

Efficient Triton Kernels for LLM Training

Depth Pro: Sharp Monocular Metric Depth in Less Than a Second.

A PyTorch native platform for training generative AI models

Perf monitoring CLI tool for Apple Silicon

🚀 Efficient implementations of state-of-the-art linear attention models

[MLSys 2024 Best Paper Award] AWQ: Activation-aware Weight Quantization for LLM Compression and Acceleration

Open source process design kit for usage with SkyWater Technology Foundry's 130nm node.

Sparsity-aware deep learning inference runtime for CPUs

Implementation for MatMul-free LM.

Run LLMs with MLX

SOTA low-bit LLM quantization (INT8/FP8/MXFP8/INT4/MXFP4/NVFP4) & sparsity; leading model compression techniques on TensorFlow, PyTorch, and ONNX Runtime

PyTorch native quantization and sparsity for training and inference

Minimalistic large language model 3D-parallelism training

Code for the ICLR 2023 paper "GPTQ: Accurate Post-training Quantization of Generative Pretrained Transformers".

Libraries for applying sparsification recipes to neural networks with a few lines of code, enabling faster and smaller models

dstack is an open-source control plane for running development, training, and inference jobs on GPUs—across hyperscalers, neoclouds, or on-prem.

Implementation of "BitNet: Scaling 1-bit Transformers for Large Language Models" in pytorch

MLX-VLM is a package for inference and fine-tuning of Vision Language Models (VLMs) on your Mac using MLX.

Machine learning on FPGAs using HLS

Implementing DeepSeek R1's GRPO algorithm from scratch

[ICML 2023] SmoothQuant: Accurate and Efficient Post-Training Quantization for Large Language Models

Modular hardware build system