search

mlii0117 / DCL

Official code for "Dynamic Graph Enhanced Contrastive Learning for Chest X-ray Report Generation" (CVPR 2023)
MIT License
90 stars 10 forks source link

Regarding the mismatch in vocab size of the SCI model loaded for training #19

Open Ammexm opened 9 months ago

Ammexm commented 9 months ago

size mismatch for cls.predictions.decoder.bias: copying a param with shape torch.Size([31090]) from checkpoint, the shape in current model is torch.Size([31092]).

Hello, when I run the training command, the program reports an error as shown above. How to solve this problem?

Voolion commented 2 weeks ago

i have the same question, and i adjust the config file (maybe med_config_sci.json ,depend on what config you use), there is property 'vocab_size' and set it 31090 (your vocal_size)