Merge pull request #428 from P2Oileen:fix-pad

GitOrigin-RevId: f33ea46ad62025f6531d98311df3d39462e7d862

imperative/python/megengine/functional/nn.py

@@ -1708,7 +1708,7 @@ def sliding_window_transpose(
 
 def pad(
     src: Tensor,
-    pad_witdth: Tuple[Tuple[int, int], ...],
+    pad_width: Tuple[Tuple[int, int], ...],
     mode: str = "constant",
     constant_value: float = 0.0,
 ) -> Tensor:
@@ -1723,9 +1723,9 @@ def pad(
     if mode.lower() == "edge":
         mode = "replicate"
 
-    for i in range(0, len(pad_witdth)):
-        p_offsets[i * 2] = pad_witdth[i][0]
-        p_offsets[i * 2 + 1] = pad_witdth[i][1]
+    for i in range(0, len(pad_width)):
+        p_offsets[i * 2] = pad_width[i][0]
+        p_offsets[i * 2 + 1] = pad_width[i][1]
 
     op = builtin.Padding(
         front_offset_dim0=p_offsets[0],

imperative/python/megengine/module/padding.py

@@ -12,16 +12,16 @@ class Pad(Module):
 
     def __init__(
         self,
-        pad_witdth: Tuple[Tuple[int, int], ...],
+        pad_width: Tuple[Tuple[int, int], ...],
         mode: str = "constant",
         constant_val: float = 0.0,
     ):
         super().__init__()
-        self.pad_width = pad_witdth
+        self.pad_width = pad_width
         self.mode = mode
         self.pad_val = constant_val
 
     def forward(self, src):
         return nn.pad(
-            src, pad_witdth=self.pad_width, mode=self.mode, constant_value=self.pad_val
+            src, pad_width=self.pad_width, mode=self.mode, constant_value=self.pad_val
         )

imperative/python/megengine/traced_module/compat.py

@@ -162,3 +162,11 @@ def tensor_gen_func_loader(expr):
         else:
             device = None
         expr.set_args_kwargs(shape, dtype=dtype, device=device)
+
+
+@register_functional_loader(("megengine.functional.nn", "pad"))
+def pad_func_loader(expr):
+    if "pad_witdth" in expr.kwargs:
+        kwargs = expr.kwargs
+        kwargs["pad_width"] = kwargs.pop("pad_witdth")
+        expr.set_args_kwargs(*expr.args, **kwargs)