Fix for `python models/yolo.py --profile` (#4541)

Profiling fix copies input to Detect layer to circumvent inplace changes to the feature maps.

models/yolo.py

@@ -48,7 +48,6 @@ class Detect(nn.Module):
         self.inplace = inplace  # use in-place ops (e.g. slice assignment)
 
     def forward(self, x):
-        # x = x.copy()  # for profiling
         z = []  # inference output
         for i in range(self.nl):
             x[i] = self.m[i](x[i])  # conv
@@ -143,10 +142,11 @@ class Model(nn.Module):
                 x = y[m.f] if isinstance(m.f, int) else [x if j == -1 else y[j] for j in m.f]  # from earlier layers
 
             if profile:
-                o = thop.profile(m, inputs=(x,), verbose=False)[0] / 1E9 * 2 if thop else 0  # FLOPs
+                c = isinstance(m, Detect)  # copy input as inplace fix
+                o = thop.profile(m, inputs=(x.copy() if c else x,), verbose=False)[0] / 1E9 * 2 if thop else 0  # FLOPs
                 t = time_sync()
                 for _ in range(10):
-                    _ = m(x)
+                    m(x.copy() if c else x)
                 dt.append((time_sync() - t) * 100)
                 if m == self.model[0]:
                     LOGGER.info(f"{'time (ms)':>10s} {'GFLOPs':>10s} {'params':>10s}  {'module'}")