search

fudan-zvg / SETR

[CVPR 2021] Rethinking Semantic Segmentation from a Sequence-to-Sequence Perspective with Transformers
MIT License
1.05k stars 150 forks source link

always CUDA out of memory #24

Closed daixiaolei623 closed 3 years ago

daixiaolei623 commented 3 years ago

@lzrobots @VictorLlu @sixiaozheng Hi, thank you for your sharing. however, when i run "./tools/dist_test.sh configs/SETR/SETR_PUP_512x512_160k_ade20k_bs_16_MS.py", i got the error: CUDA out of memory. I have 2 NVIDIA Tesla P100 about 16GB per GPU. Could you please tell me what is wrong. Thank you.

daixiaolei623 commented 3 years ago

could you please tell which GPU do you use and how big they are.

lzrobots commented 3 years ago

Hey have you ever run any segmentation model using 2*16G GPUs, eg, PSPNet, DeepLab V3 with ResNet101?

Please refer to #13

GusRoth commented 3 years ago

@lzrobots @VictorLlu @sixiaozheng Hi, thank you for your sharing. however, when i run "./tools/dist_test.sh configs/SETR/SETR_PUP_512x512_160k_ade20k_bs_16_MS.py", i got the error: CUDA out of memory. I have 2 NVIDIA Tesla P100 about 16GB per GPU. Could you please tell me what is wrong. Thank you.

the authors have said that they used the 8x32g GPUs to run the SETR at zhihu