tldr/pages/common/accelerate.md

Accelerate

A library that enables the same PyTorch code to be run across any distributed configuration. More information: https://huggingface.co/docs/accelerate/index.

• Print environment information:

accelerate env

• Interactively create a configuration file:

accelerate config

• Print the estimated GPU memory cost of running a huggingface model with different data types:

accelerate estimate-memory {{name/model}}

• Test an Accelerate configuration file:

accelerate test --config_file {{path/to/config.yaml}}

• Run a model on CPU with Accelerate:

accelerate launch {{path/to/script.py}} {{--cpu}}

• Run a model on multi-GPU with Accelerate, with 2 machines:

accelerate launch {{path/to/script.py}} --multi_gpu --num_machines {{2}}