lilith/tldr
1
0
Fork 0
mirror of https://github.com/tldr-pages/tldr.git synced 2025-04-29 23:24:55 +02:00
Code Activity
tldr/pages/common/accelerate.md
WU Zhenglong f994b81ad5
accelerate: add page (#11340) 
* accelerate: add page

* accelerate: modify command description

* accelerate: update template format

Co-authored-by: Sebastiaan Speck <12570668+sebastiaanspeck@users.noreply.github.com>

* accelerate: remove inappropriate description

Co-authored-by: Juri Dispan <juri.dispan@posteo.net>

* accelerate: update command description

Co-authored-by: Juri Dispan <juri.dispan@posteo.net>

* accelerate: fix typo and update appropriate description

Co-authored-by: Juri Dispan <juri.dispan@posteo.net>

---------

Co-authored-by: Sebastiaan Speck <12570668+sebastiaanspeck@users.noreply.github.com>
Co-authored-by: Juri Dispan <juri.dispan@posteo.net>
2023-10-31 18:17:44 +01:00

781 B
Raw Blame History

Accelerate

Accelerate is 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}}