fairscale

We have hosted the application fairscale in order to run this application in our online workstations with Wine or directly.

Quick description about fairscale:

FairScale is a collection of PyTorch performance and scaling primitives that pioneered many of the ideas now used for large-model training. It introduced Fully Sharded Data Parallel (FSDP) style techniques that shard model parameters, gradients, and optimizer states across ranks to fit bigger models into the same memory budget. The library also provides pipeline parallelism, activation checkpointing, mixed precision, optimizer state sharding (OSS), and auto-wrapping policies that reduce boilerplate in complex distributed setups. Its components are modular, so teams can adopt just the sharding optimizer or the pipeline engine without rewriting their training loop. FairScale puts emphasis on correctness and debuggability, offering hook points, logging, and reference examples for common trainer patterns. Although many ideas have since landed in core PyTorch, FairScale remains a valuable reference and a practical toolbox for squeezing more performance out of multi-GPU and multi-node jobs.

Features:

• Fully Sharded Data Parallel style parameter, grad, and optimizer sharding
• Pipeline parallelism utilities with schedule control
• Activation checkpointing to trade compute for memory
• Optimizer State Sharding (OSS) drop-in optimizers
• Mixed precision and auto-wrap policies for easy adoption
• Examples and hooks for production-grade distributed training

Programming Language: Python.
Categories: