Add support for FP16 (half) to export.py (#3010)

* Added support for fp16 (half) to export.py

* minimize code additions

Co-authored-by: Glenn Jocher <glenn.jocher@ultralytics.com>

models/export.py

@@ -28,6 +28,7 @@ if __name__ == '__main__':
     parser.add_argument('--batch-size', type=int, default=1, help='batch size')
     parser.add_argument('--grid', action='store_true', help='export Detect() layer grid')
     parser.add_argument('--device', default='cpu', help='cuda device, i.e. 0 or 0,1,2,3 or cpu')
+    parser.add_argument('--half', action='store_true', help='FP16 half-precision export')
     parser.add_argument('--dynamic', action='store_true', help='dynamic ONNX axes')  # ONNX-only
     parser.add_argument('--simplify', action='store_true', help='simplify ONNX model')  # ONNX-only
     opt = parser.parse_args()
@@ -44,11 +45,14 @@ if __name__ == '__main__':
     # Checks
     gs = int(max(model.stride))  # grid size (max stride)
     opt.img_size = [check_img_size(x, gs) for x in opt.img_size]  # verify img_size are gs-multiples
+    assert not (opt.device.lower() == "cpu" and opt.half), '--half only compatible with GPU export, i.e. use --device 0'
 
     # Input
     img = torch.zeros(opt.batch_size, 3, *opt.img_size).to(device)  # image size(1,3,320,192) iDetection
 
     # Update model
+    if opt.half:
+        img, model = img.half(), model.half()  # to FP16
     for k, m in model.named_modules():
         m._non_persistent_buffers_set = set()  # pytorch 1.6.0 compatibility
         if isinstance(m, models.common.Conv):  # assign export-friendly activations