Add --model-dtype (pure bfloat16/float16) support to inference.py

Author: rwightman

inference.py

@@ -105,6 +105,8 @@
                     help='use Native AMP for mixed precision training')
 parser.add_argument('--amp-dtype', default='float16', type=str,
                     help='lower precision AMP dtype (default: float16)')
+parser.add_argument('--model-dtype', default=None, type=str,
+                   help='Model dtype override (non-AMP) (default: float32)')
 parser.add_argument('--fuser', default='', type=str,
                     help="Select jit fuser. One of ('', 'te', 'old', 'nvfuser')")
 parser.add_argument('--model-kwargs', nargs='*', default={}, action=ParseKwargs)
@@ -161,9 +163,15 @@ def main():
 
     device = torch.device(args.device)
 
+    model_dtype = None
+    if args.model_dtype:
+        assert args.model_dtype in ('float32', 'float16', 'bfloat16')
+        model_dtype = getattr(torch, args.model_dtype)
+
     # resolve AMP arguments based on PyTorch / Apex availability
     amp_autocast = suppress
     if args.amp:
+        assert model_dtype is None or model_dtype == torch.float32, 'float32 model dtype must be used with AMP'
         assert args.amp_dtype in ('float16', 'bfloat16')
         amp_dtype = torch.bfloat16 if args.amp_dtype == 'bfloat16' else torch.float16
         amp_autocast = partial(torch.autocast, device_type=device.type, dtype=amp_dtype)
@@ -201,7 +209,7 @@ def main():
     if args.test_pool:
         model, test_time_pool = apply_test_time_pool(model, data_config)
 
-    model = model.to(device)
+    model = model.to(device=device, dtype=model_dtype)
     model.eval()
     if args.channels_last:
         model = model.to(memory_format=torch.channels_last)
@@ -237,6 +245,7 @@ def main():
         use_prefetcher=True,
         num_workers=workers,
         device=device,
+        img_dtype=model_dtype or torch.float32,
         **data_config,
     )
 
@@ -280,7 +289,7 @@ def main():
                     np_labels = to_label(np_indices)
                     all_labels.append(np_labels)
 
-            all_outputs.append(output.cpu().numpy())
+            all_outputs.append(output.float().cpu().numpy())
 
             # measure elapsed time
             batch_time.update(time.time() - end)