search

SHI-Labs / Neighborhood-Attention-Transformer

Neighborhood Attention Transformer, arxiv 2022 / CVPR 2023. Dilated Neighborhood Attention Transformer, arxiv 2022
MIT License
1.05k stars 86 forks source link

Issue with Neighborhood Attention Model (NAT) Pretrained Weights #101

Closed Miiitiii closed 6 months ago

Miiitiii commented 6 months ago

Dear Mr. Hassani,

I hope this message finds you well.

I am a user of your powerful and highly effective Neighborhood Attention Model (NAT). Firstly, I want to express my gratitude for creating such a helpful model and for your outstanding contributions to the field.

I have been utilizing NAT following the guidelines provided in your GitHub repository, and it has been instrumental in my work. However, after several days of successful usage, I encountered an error regarding the matching of pretrained weights in your model. I have attached a screenshot of this error for your reference. it happens in your latest NATTEN version(natten==0.17.0).

I would greatly appreciate it if you could advise me on how to resolve this issue or if there might be a problem with the model that you could address.

Thank you for your dedication and for all the invaluable work you do. Your assistance in this matter is highly appreciated.

Best regards

error

alihassanijr commented 6 months ago

Thank you for your interest and apologies for the inconvenience. This was caused by NATTEN renaming the bias argument to rel_pos_bias since version 0.17.0.

I'll start a PR fixing this.