Make pad_to_patch_size function work on multi dim.

Author: comfyanonymous

comfy/ldm/common_dit.py

@@ -4,9 +4,12 @@
 def pad_to_patch_size(img, patch_size=(2, 2), padding_mode="circular"):
     if padding_mode == "circular" and (torch.jit.is_tracing() or torch.jit.is_scripting()):
         padding_mode = "reflect"
-    pad_h = (patch_size[0] - img.shape[-2] % patch_size[0]) % patch_size[0]
-    pad_w = (patch_size[1] - img.shape[-1] % patch_size[1]) % patch_size[1]
-    return torch.nn.functional.pad(img, (0, pad_w, 0, pad_h), mode=padding_mode)
+
+    pad = ()
+    for i in range(img.ndim - 2):
+        pad = (0, (patch_size[i] - img.shape[i + 2] % patch_size[i]) % patch_size[i]) + pad
+
+    return torch.nn.functional.pad(img, pad, mode=padding_mode)
 
 try:
     rms_norm_torch = torch.nn.functional.rms_norm