search

jiasenlu / vilbert_beta

473 stars 96 forks source link

Error when extract feature from faster rcnn #11

Closed neverland7D closed 5 years ago

neverland7D commented 5 years ago

when I use generate_tsv to extract feature from pretrained caffe model, I got error "Check failed: error == cudaSuccess (8 vs. 0) invalid device function" which happend in this layer: caffe::PoolingLayer<>::Forward_gpu(). I've checked my caffe version is able to run with gpu, I have no idea what's wrong with it.

jiasenlu commented 5 years ago

I think you didn't install caffe correctly. please follow the original bottom-up repo to install caffe first. The caffe requires python2 and opencv.

neverland7D commented 5 years ago

I think you didn't install caffe correctly. please follow the original bottom-up repo to install caffe first. The caffe requires python2 and opencv.

Thanks a lot for your reply!