graphcore-research / out-of-the-box-fp8-training
Demo of the unit_scaling library, showing how a model can be easily adapted to train in FP8.
☆45Updated 9 months ago
Alternatives and similar repositories for out-of-the-box-fp8-training:
Users that are interested in out-of-the-box-fp8-training are comparing it to the libraries listed below
- Experiment of using Tangent to autodiff triton☆78Updated last year
- ☆78Updated 10 months ago
- A library for unit scaling in PyTorch☆125Updated 5 months ago
- CUDA implementation of autoregressive linear attention, with all the latest research findings☆44Updated last year
- JAX bindings for Flash Attention v2☆89Updated 9 months ago
- ☆103Updated 11 months ago
- Triton Implementation of HyperAttention Algorithm☆47Updated last year
- Personal solutions to the Triton Puzzles☆18Updated 9 months ago
- A place to store reusable transformer components of my own creation or found on the interwebs☆55Updated this week
- Make triton easier☆47Updated 10 months ago
- Tritonbench is a collection of PyTorch custom operators with example inputs to measure their performance.☆120Updated this week
- Using FlexAttention to compute attention with different masking patterns☆43Updated 7 months ago
- ☆104Updated 8 months ago
- This repository contains the experimental PyTorch native float8 training UX☆224Updated 9 months ago
- Automatically take good care of your preemptible TPUs☆36Updated last year
- Flash-Muon: An Efficient Implementation of Muon Optimzer☆91Updated this week
- Fast and memory-efficient exact attention☆68Updated 2 months ago
- PyTorch centric eager mode debugger☆47Updated 4 months ago
- ☆21Updated 2 months ago
- supporting pytorch FSDP for optimizers☆80Updated 4 months ago
- The simplest implementation of recent Sparse Attention patterns for efficient LLM inference.☆60Updated 3 months ago
- extensible collectives library in triton☆85Updated last month
- A bunch of kernels that might make stuff slower 😉☆40Updated this week
- Utilities for Training Very Large Models☆58Updated 7 months ago
- Load compute kernels from the Hub☆115Updated last week
- Here we will test various linear attention designs.☆60Updated last year
- ☆29Updated 2 years ago
- FlexAttention w/ FlashAttention3 Support☆26Updated 7 months ago
- ☆48Updated last year
- ☆81Updated last year