RuntimeError: Mask shape should match input shape

[MINXIANWEN]

hello，author，i have a question want to ask you. when i use clip.model.text_encode , the text.tensor.shape must is [77,77]?,meaning that the text must is consist of 77 sequences? the attn_mask could be changed? the attn_mask's tensor shape is [77,77],looks like a upper triangular matrix, but when i changed my text_tensor_shape to be [77,77],it also have the same error,mask shape should match input shape, so i don't where is the problem, need your help. thanks author