facebookresearch / fairring
Fairring (FAIR + Herring) is a plug-in for PyTorch that provides a process group for distributed training that outperforms NCCL at large scales
☆65Updated 3 years ago
Alternatives and similar repositories for fairring:
Users that are interested in fairring are comparing it to the libraries listed below
- ☆43Updated last year
- System for automated integration of deep learning backends.☆48Updated 2 years ago
- (NeurIPS 2022) Automatically finding good model-parallel strategies, especially for complex models and clusters.☆38Updated 2 years ago
- FTPipe and related pipeline model parallelism research.☆41Updated last year
- Training neural networks in TensorFlow 2.0 with 5x less memory☆129Updated 3 years ago
- ☆140Updated 2 months ago
- extensible collectives library in triton☆84Updated this week
- pytorch-profiler☆51Updated last year
- A Python library transfers PyTorch tensors between CPU and NVMe☆111Updated 4 months ago
- PipeTransformer: Automated Elastic Pipelining for Distributed Training of Large-scale Models. ICML 2021☆56Updated 3 years ago
- Memory Optimizations for Deep Learning (ICML 2023)☆62Updated last year
- An external memory allocator example for PyTorch.☆14Updated 3 years ago
- ☆49Updated last year
- TileFusion is an experimental C++ macro kernel template library that elevates the abstraction level in CUDA C for tile processing.☆74Updated this week
- ☆60Updated 3 months ago
- Research and development for optimizing transformers☆125Updated 4 years ago
- A performant, memory-efficient checkpointing library for PyTorch applications, designed with large, complex distributed workloads in mind…☆155Updated 3 months ago
- ☆72Updated 3 years ago
- Tritonbench is a collection of PyTorch custom operators with example inputs to measure their performance.☆107Updated this week
- Automated Parallelization System and Infrastructure for Multiple Ecosystems