Fix bug in examples: double wrap into DataParallel during eval

examples/hans/test_hans.py

@@ -255,7 +255,7 @@ def evaluate(args, model, tokenizer, prefix=""):
         eval_dataloader = DataLoader(eval_dataset, sampler=eval_sampler, batch_size=args.eval_batch_size)
 
         # multi-gpu eval
-        if args.n_gpu > 1:
+        if args.n_gpu > 1 and not isinstance(model, torch.nn.DataParallel):
             model = torch.nn.DataParallel(model)
 
         # Eval!

examples/mm-imdb/run_mmimdb.py

@@ -278,7 +278,7 @@ def evaluate(args, model, tokenizer, criterion, prefix=""):
     )
 
     # multi-gpu eval
-    if args.n_gpu > 1:
+    if args.n_gpu > 1 and not isinstance(model, torch.nn.DataParallel):
         model = torch.nn.DataParallel(model)
 
     # Eval!

examples/ner/run_ner.py

@@ -253,7 +253,7 @@ def evaluate(args, model, tokenizer, labels, pad_token_label_id, mode, prefix=""
     eval_dataloader = DataLoader(eval_dataset, sampler=eval_sampler, batch_size=args.eval_batch_size)
 
     # multi-gpu evaluate
-    if args.n_gpu > 1:
+    if args.n_gpu > 1 and not isinstance(model, torch.nn.DataParallel):
         model = torch.nn.DataParallel(model)
 
     # Eval!

examples/run_language_modeling.py

@@ -427,7 +427,7 @@ def evaluate(args, model: PreTrainedModel, tokenizer: PreTrainedTokenizer, prefi
     )
 
     # multi-gpu evaluate
-    if args.n_gpu > 1:
+    if args.n_gpu > 1 and not isinstance(model, torch.nn.DataParallel):
         model = torch.nn.DataParallel(model)
 
     # Eval!

examples/run_multiple_choice.py

@@ -256,7 +256,7 @@ def evaluate(args, model, tokenizer, prefix="", test=False):
         eval_dataloader = DataLoader(eval_dataset, sampler=eval_sampler, batch_size=args.eval_batch_size)
 
         # multi-gpu evaluate
-        if args.n_gpu > 1:
+        if args.n_gpu > 1 and not isinstance(model, torch.nn.DataParallel):
             model = torch.nn.DataParallel(model)
 
         # Eval!

examples/run_xnli.py

@@ -266,7 +266,7 @@ def evaluate(args, model, tokenizer, prefix=""):
         eval_dataloader = DataLoader(eval_dataset, sampler=eval_sampler, batch_size=args.eval_batch_size)
 
         # multi-gpu eval
-        if args.n_gpu > 1:
+        if args.n_gpu > 1 and not isinstance(model, torch.nn.DataParallel):
             model = torch.nn.DataParallel(model)
 
         # Eval!