Fix grad checkpointing in focalnet

timm/models/focalnet.py

@@ -266,6 +266,10 @@ class BasicLayer(nn.Module):
             )
             for i in range(depth)])
 
+    @torch.jit.ignore
+    def set_grad_checkpointing(self, enable=True):
+        self.grad_checkpointing = enable
+
     def forward(self, x):
         x = self.downsample(x)
         for blk in self.blocks: