"How to set 'num_workers' during training to speed up CPU data loading?" #1783
Answered
by
ancestor-mithril
KevinWuang
asked this question in
Q&A
-
|
"My dataset is quite large, and the training speed is slow. How can I speed up the training by adjusting 'num_workers'? Are there any other methods? Can you provide more details?" |
Beta Was this translation helpful? Give feedback.
Answered by
ancestor-mithril
Nov 6, 2023
Replies: 1 comment 1 reply
-
|
Use the environment variables |
Beta Was this translation helpful? Give feedback.
1 reply
-
|
Thanks |
Beta Was this translation helpful? Give feedback.
Answer selected by
KevinWuang
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Use the environment variables
nnUNet_n_proc_DA(default: 8) to change the number of workers.It is not recommended to increase it if you have less than 64GB RAM.
If you have less, the training might be faster if you decrease the number of workers.