Confusion over the naming in pos_embed
Author: SweeRotyCreated Jul 10, 2025Updated Jul 10, 2025
emb_h = get_1d_sincos_pos_embed_from_grid(embed_dim // 2, grid[0]) # (H*W, D/2)
emb_w = get_1d_sincos_pos_embed_from_grid(embed_dim // 2, grid[1]) # (H*W, D/2)In get_2d_sincos_pos_embed_from_grid , since w goes first, doesn't grid[0] strand for emb_w and grid[1] stand for emb_h? Is this a typo here?
Source: facebookresearch/mae