Fix2 select_device() for Multi-GPU (#6461)

* Fix2 select_device() for Multi-GPU

* Cleanup

* Cleanup

* Simplify error message

* Improve assert

* Update torch_utils.py

utils/datasets.py

@@ -29,13 +29,12 @@ from tqdm import tqdm
 from utils.augmentations import Albumentations, augment_hsv, copy_paste, letterbox, mixup, random_perspective
 from utils.general import (LOGGER, NUM_THREADS, check_dataset, check_requirements, check_yaml, clean_str,
                            segments2boxes, xyn2xy, xywh2xyxy, xywhn2xyxy, xyxy2xywhn)
-from utils.torch_utils import device_count, torch_distributed_zero_first
+from utils.torch_utils import torch_distributed_zero_first
 
 # Parameters
 HELP_URL = 'https://github.com/ultralytics/yolov5/wiki/Train-Custom-Data'
 IMG_FORMATS = ['bmp', 'dng', 'jpeg', 'jpg', 'mpo', 'png', 'tif', 'tiff', 'webp']  # include image suffixes
 VID_FORMATS = ['asf', 'avi', 'gif', 'm4v', 'mkv', 'mov', 'mp4', 'mpeg', 'mpg', 'wmv']  # include video suffixes
-DEVICE_COUNT = max(device_count(), 1)  # number of CUDA devices
 
 # Get orientation exif tag
 for orientation in ExifTags.TAGS.keys():
@@ -110,7 +109,8 @@ 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() // DEVICE_COUNT, batch_size if batch_size > 1 else 0, workers])  # number of workers
+    nd = torch.cuda.device_count()  # number of CUDA devices
+    nw = min([os.cpu_count() // max(nd, 1), batch_size if batch_size > 1 else 0, workers])  # number of workers
     sampler = None if rank == -1 else distributed.DistributedSampler(dataset, shuffle=shuffle)
     loader = DataLoader if image_weights else InfiniteDataLoader  # only DataLoader allows for attribute updates
     return loader(dataset,