search

czczup / ViT-Adapter

[ICLR 2023 Spotlight] Vision Transformer Adapter for Dense Predictions
https://arxiv.org/abs/2205.08534
Apache License 2.0
1.27k stars 140 forks source link

Download Pretrained Model, Finetune on separate dataset #58

Closed dhosk22 closed 1 year ago

dhosk22 commented 1 year ago

Hi, for a project in my deep learning course, I am attempting to download your pretrained model for semantic segmentation tested on cityscapes dataset. I want to finetune it on a separate dataset that I found on the web. I am having quite a difficult time getting my colab notebook set up. I'm hoping I could get some advice in order to move forward with this project.

czczup commented 1 year ago

Hi, this is an example of colab notebook, provided by dudifrid.

dhosk22 commented 1 year ago

Thanks for the quick response. Is the purpose of the checkpoint to save on computation? Also, why is the check using a Cocostuff model, while the model used for testing is the Ade20k?

czczup commented 1 year ago

Thanks for the quick response. Is the purpose of the checkpoint to save on computation? Also, why is the check using a Cocostuff model, while the model used for testing is the Ade20k?

with_cp=True is used to save GPU memory. Besides, there might be some mistakes in this example. You should use a ade20k model for the testing of ade20k dataset.