DDP `WORLD_SIZE`-safe dataloader workers (#5631)

train.py

@@ -266,7 +266,7 @@ def train(hyp,  # path/to/hyp.yaml or hyp dictionary
     stopper = EarlyStopping(patience=opt.patience)
     compute_loss = ComputeLoss(model)  # init loss class
     LOGGER.info(f'Image sizes {imgsz} train, {imgsz} val\n'
-                f'Using {train_loader.num_workers} dataloader workers\n'
+                f'Using {train_loader.num_workers * WORLD_SIZE} dataloader workers\n'
                 f"Logging results to {colorstr('bold', save_dir)}\n"
                 f'Starting training for {epochs} epochs...')
     for epoch in range(start_epoch, epochs):  # epoch ------------------------------------------------------------------
@@ -460,7 +460,7 @@ def parse_opt(known=False):
     parser.add_argument('--single-cls', action='store_true', help='train multi-class data as single-class')
     parser.add_argument('--adam', action='store_true', help='use torch.optim.Adam() optimizer')
     parser.add_argument('--sync-bn', action='store_true', help='use SyncBatchNorm, only available in DDP mode')
-    parser.add_argument('--workers', type=int, default=8, help='maximum number of dataloader workers')
+    parser.add_argument('--workers', type=int, default=8, help='max dataloader workers (per RANK in DDP mode)')
     parser.add_argument('--project', default=ROOT / 'runs/train', help='save to project/name')
     parser.add_argument('--name', default='exp', help='save to project/name')
     parser.add_argument('--exist-ok', action='store_true', help='existing project/name ok, do not increment')

utils/datasets.py

@@ -34,6 +34,7 @@ from utils.torch_utils import torch_distributed_zero_first
 HELP_URL = 'https://github.com/ultralytics/yolov5/wiki/Train-Custom-Data'
 IMG_FORMATS = ['bmp', 'jpg', 'jpeg', 'png', 'tif', 'tiff', 'dng', 'webp', 'mpo']  # acceptable image suffixes
 VID_FORMATS = ['mov', 'avi', 'mp4', 'mpg', 'mpeg', 'm4v', 'wmv', 'mkv']  # acceptable video suffixes
+WORLD_SIZE = int(os.getenv('WORLD_SIZE', 1))  # DPP
 NUM_THREADS = min(8, os.cpu_count())  # number of multiprocessing threads
 
 # Get orientation exif tag
@@ -107,7 +108,7 @@ def create_dataloader(path, imgsz, batch_size, stride, single_cls=False, hyp=Non
                                       prefix=prefix)
 
     batch_size = min(batch_size, len(dataset))
-    nw = min([os.cpu_count(), batch_size if batch_size > 1 else 0, workers])  # number of workers
+    nw = min([os.cpu_count() // WORLD_SIZE, batch_size if batch_size > 1 else 0, workers])  # number of workers
     sampler = torch.utils.data.distributed.DistributedSampler(dataset) if rank != -1 else None
     loader = torch.utils.data.DataLoader if image_weights else InfiniteDataLoader
     # Use torch.utils.data.DataLoader() if dataset.properties will update during training else InfiniteDataLoader()