Parallelformers: An Efficient Model Parallelization Toolkit for Deployment
Parallelformers: An Efficient Model Parallelization Toolkit for Deployment
Cross-node inference
Use this library for CNN networks like Unet
Support for LLaMA
Title: RuntimeError: Timed out initializing process group in store based barrier
Still in development?
RuntimeError: CUDA error: peer access is not supported between these two devices
freeze_support()
Support for Falcon-7B and Falcon-40B models
Bus error in parallelformers 1.2.7 for OPT model
[Feature Request] Add Bloom to the Auto Policy