Open rabeeqasem opened 7 months ago
If you are training on CPU, I am not sure if multi threading is supported by detectron2. I am using their launch function for splitting the work over multiple GPUs. Perhaps you can find more in their documentation. To set the number of workers for the dataloader use:
DATALOADER:
NUM_WORKERS: 16
how can we define in the YAML file the multi treathreading CPU and the number of workers to train ?