search

GXYM / TextBPN-Plus-Plus

Arbitrary Shape Text Detection via Boundary Transformer；The paper at: https://arxiv.org/abs/2205.05320, which has been accepted by IEEE Transactions on Multimedia (T-MM 2023).
172 stars 37 forks source link

TITAN V inference ERROR！ #5

Open yazheng0307 opened 2 years ago

yazheng0307 commented 2 years ago

thank for you code contribute，we met a unclear error. when we inference a image by using a trained checkpoint on GPU named TITAN V,we got nothing result. But on another GPU card（NVIDIA 3080TI）, the reasoning will have right results. Can you tell us why this phenomenon？ Is it related to the compiling environment of deformable convolution?

yazheng0307 commented 2 years ago

cdea97f8dffeea608745831e95e62f8 3c647f3af52b7409e37c250cccc3ac9

GXYM commented 2 years ago

thank for you code contribute，we met a unclear error. when we inference a image by using a trained checkpoint on GPU named TITAN V,we got nothing result. But on another GPU card（NVIDIA 3080TI）, the reasoning will have right results. Can you tell us why this phenomenon？ Is it related to the compiling environment of deformable convolution?

Yes, you may need to compile some CUDA files and set the GPU environment to TITAN V. such as

image

Release the marked code and set the GPU ID you need (TITAN V). Otherwise, GPU 0 will be used by default.