bug for distributed training (#1985)

fix a bug for distributed training in windows platform

Author: CapMocha

mmdet/apis/train.py

@@ -73,7 +73,7 @@ def parse_losses(losses):
     log_vars['loss'] = loss
     for loss_name, loss_value in log_vars.items():
         # reduce loss when distributed training
-        if dist.is_initialized():
+        if dist.is_available() and dist.is_initialized():
             loss_value = loss_value.data.clone()
             dist.all_reduce(loss_value.div_(dist.get_world_size()))
         log_vars[loss_name] = loss_value.item()