Merge pull request #2181 from huggingface/Delaunay-dist-backend

Delaunay dist backend flag

timm/utils/distributed.py

@@ -108,9 +108,16 @@ def init_distributed_device_so(
     world_size = 1
     global_rank = 0
     local_rank = 0
+    device_type, *device_idx = device.split(':', maxsplit=1)
+
     if dist_backend is None:
-        # FIXME sane defaults for other device backends?
-        dist_backend = 'nccl' if 'cuda' in device else 'gloo'
+        # FIXME: verify that ROCm transform nccl to rccl
+        dist_backends = {
+            "xpu": "ccl",
+            "hpu": "hccl",
+            "cuda": "nccl",
+        }
+        dist_backend = dist_backends.get(device_type, 'gloo')
     dist_url = dist_url or 'env://'
 
     # TBD, support horovod?
@@ -150,18 +157,15 @@ def init_distributed_device_so(
             global_rank = torch.distributed.get_rank()
         distributed = True
 
-    if 'cuda' in device:
+    if device_type == 'cuda':
         assert torch.cuda.is_available(), f'CUDA is not available but {device} was specified.'
 
     if distributed and device != 'cpu':
-        device, *device_idx = device.split(':', maxsplit=1)
-
         # Ignore manually specified device index in distributed mode and
         # override with resolved local rank, fewer headaches in most setups.
         if device_idx:
             _logger.warning(f'device index {device_idx[0]} removed from specified ({device}).')
-
-        device = f'{device}:{local_rank}'
+        device = f'{device_type}:{local_rank}'
 
     if device.startswith('cuda:'):
         torch.cuda.set_device(device)