NVIDIA / TransformerEngineLinks
A library for accelerating Transformer models on NVIDIA GPUs, including using 8-bit and 4-bit floating point (FP8 and FP4) precision on Hopper, Ada and Blackwell GPUs, to provide better performance with lower memory utilization in both training and inference.
☆2,993Updated this week
Alternatives and similar repositories for TransformerEngine
Users that are interested in TransformerEngine are comparing it to the libraries listed below
Sorting:
- FlashInfer: Kernel Library for LLM Serving☆4,195Updated this week
- PyTorch native quantization and sparsity for training and inference☆2,560Updated this week
- [ICML 2023] SmoothQuant: Accurate and Efficient Post-Training Quantization for Large Language Models☆1,570Updated last year
- A unified library of SOTA model optimization techniques like quantization, pruning, distillation, speculative decoding, etc. It compresse…☆1,637Updated this week
- [MLSys 2024 Best Paper Award] AWQ: Activation-aware Weight Quantization for LLM Compression and Acceleration☆3,386Updated 4 months ago
- Transformer related optimization, including BERT, GPT☆6,365Updated last year
- Ongoing research training transformer language models at scale, including: BERT & GPT-2☆2,203Updated 3 months ago
- Tile primitives for speedy kernels☆2,994Updated this week
- Automatically Discovering Fast Parallelization Strategies for Distributed Deep Neural Network Training☆1,849Updated this week
- Pipeline Parallelism for PyTorch☆784Updated last year
- Code for the ICLR 2023 paper "GPTQ: Accurate Post-training Quantization of Generative Pretrained Transformers".☆2,234Updated last year
- MII makes low-latency and high-throughput inference possible, powered by DeepSpeed.☆2,081Updated 5 months ago
- PyTorch extensions for high performance and large scale training.☆3,388Updated 7 months ago
- The Triton TensorRT-LLM Backend☆910Updated 2 weeks ago
- Mirage Persistent Kernel: Compiling LLMs into a MegaKernel☆1,981Updated this week
- Kernl lets you run PyTorch transformer models several times faster on GPU with a single line of code, and is designed to be easily hackab…☆1,585Updated last year
- Transformers-compatible library for applying various compression algorithms to LLMs for optimized deployment with vLLM☆2,348Updated last week
- SOTA low-bit LLM quantization (INT8/FP8/MXFP8/INT4/MXFP4/NVFP4) & sparsity; leading model compression techniques on PyTorch, TensorFlow, …☆2,543Updated this week
- Distributed Compiler based on Triton for Parallel Systems☆1,269Updated this week
- Minimalistic large language model 3D-parallelism training☆2,362Updated 3 weeks ago
- AutoAWQ implements the AWQ algorithm for 4-bit quantization with a 2x speedup during inference. Documentation:☆2,289Updated 7 months ago
- A pytorch quantization backend for optimum☆1,012Updated 3 weeks ago
- Flash Attention in ~100 lines of CUDA (forward pass only)☆1,015Updated 11 months ago
- FP16xINT4 LLM inference kernel that can achieve near-ideal ~4x speedups up to medium batchsizes of 16-32 tokens.☆960Updated last year
- NCCL Tests☆1,363Updated 3 weeks ago
- Puzzles for learning Triton☆2,153Updated last year
- Tutel MoE: Optimized Mixture-of-Experts Library, Support GptOss/DeepSeek/Kimi-K2/Qwen3 using FP8/NVFP4/MXFP4☆948Updated last month
- A fast communication-overlapping library for tensor/expert parallelism on GPUs.☆1,188Updated 3 months ago
- 🚀 Efficient implementations of state-of-the-art linear attention models☆3,996Updated last week
- Byted PyTorch Distributed for Hyperscale Training of LLMs and RLs☆896Updated 2 weeks ago