aredden / torch-bnb-fp4Links
Faster Pytorch bitsandbytes 4bit fp4 nn.Linear ops
☆28Updated last year
Alternatives and similar repositories for torch-bnb-fp4
Users that are interested in torch-bnb-fp4 are comparing it to the libraries listed below
Sorting:
- Repository for Sparse Finetuning of LLMs via modified version of the MosaicML llmfoundry☆42Updated last year
- This repository contains the experimental PyTorch native float8 training UX☆223Updated 10 months ago
- Demo of the unit_scaling library, showing how a model can be easily adapted to train in FP8.☆44Updated 10 months ago
- ☆108Updated last year
- QuIP quantization☆52Updated last year
- ☆157Updated last year
- ☆46Updated last week
- Work in progress.☆67Updated this week
- Flash-Muon: An Efficient Implementation of Muon Optimizer☆121Updated last week
- A place to store reusable transformer components of my own creation or found on the interwebs☆56Updated 3 weeks ago
- A library for unit scaling in PyTorch☆125Updated 6 months ago
- Experiment of using Tangent to autodiff triton☆79Updated last year
- Repository for CPU Kernel Generation for LLM Inference☆26Updated last year
- ☆129Updated 3 months ago
- research impl of Native Sparse Attention (2502.11089)☆54Updated 3 months ago
- 🎬 3.7× faster video generation E2E 🖼️ 1.6× faster image generation E2E ⚡ ColumnSparseAttn 9.3× vs FlashAttn‑3 💨 ColumnSparseGEMM 2.5× …☆65Updated this week
- Boosting 4-bit inference kernels with 2:4 Sparsity☆76Updated 9 months ago
- This repository contains code for the MicroAdam paper.☆19Updated 5 months ago
- Triton Implementation of HyperAttention Algorithm☆48Updated last year
- ☆130Updated 2 months ago
- 32 times longer context window than vanilla Transformers and up to 4 times longer than memory efficient Transformers.☆48Updated last year
- ☆73Updated 4 months ago
- Tritonbench is a collection of PyTorch custom operators with example inputs to measure their performance.☆127Updated this week
- Low-Rank Llama Custom Training☆23Updated last year
- Fast low-bit matmul kernels in Triton☆311Updated this week
- Load compute kernels from the Hub☆139Updated this week
- ☆144Updated 2 years ago
- [WIP] Better (FP8) attention for Hopper☆30Updated 3 months ago
- ☆28Updated 10 months ago
- Odysseus: Playground of LLM Sequence Parallelism☆69Updated 11 months ago