Minor adjustment, mutable default arg, extra check of valid len...

timm/models/vision_transformer.py

@@ -352,7 +352,7 @@ def _init_vit_weights(m, n: str = '', head_bias: float = 0., jax_impl: bool = Fa
         nn.init.ones_(m.weight)
 
 
-def resize_pos_embed(posemb, posemb_new, num_tokens=1, gs_new=[]):
+def resize_pos_embed(posemb, posemb_new, num_tokens=1, gs_new=()):
     # Rescale the grid of position embeddings when loading from state_dict. Adapted from
     # https://github.com/google-research/vision_transformer/blob/00883dd691c63a6830751563748663526e811cee/vit_jax/checkpoint.py#L224
     _logger.info('Resized position embedding: %s to %s', posemb.shape, posemb_new.shape)
@@ -365,6 +365,7 @@ def resize_pos_embed(posemb, posemb_new, num_tokens=1, gs_new=[]):
     gs_old = int(math.sqrt(len(posemb_grid)))
     if not len(gs_new):  # backwards compatibility
         gs_new = [int(math.sqrt(ntok_new))] * 2
+    assert len(gs_new) >= 2
     _logger.info('Position embedding grid-size from %s to %s', [gs_old, gs_old], gs_new)
     posemb_grid = posemb_grid.reshape(1, gs_old, gs_old, -1).permute(0, 3, 1, 2)
     posemb_grid = F.interpolate(posemb_grid, size=gs_new, mode='bilinear')