size mismatch for encoder.deit.pos_embed

[dreamlychina]

Describe the bug Model I am using (trocr):

The problem arises when using:

• [ ] the official example scripts: (give details below) python pic_inference.py

A clear and concise description of what the bug is. RuntimeError: Error(s) in loading state_dict for TrOCRModel: size mismatch for encoder.deit.pos_embed: copying a param with shape torch.Size([1, 578, 384]) from checkpoint, the shape in current model is torch.Size([1, 577, 384]). the model is trocr and the checkpoints is the offical trocr-small-handwritten.pt

To Reproduce Steps to reproduce the behavior:

Expected behavior A clear and concise description of what you expected to happen.

• Platform:
• Python version: 3.8.3
• PyTorch version (GPU?): 1.13.1

[domitix]

Hi, I have this problem too. Have you solved it?