Andres

It is the same model as the one the repo uses to extract but rather than dealing with caffe (can be painful) it is on pytorch. Check that repo and...

@LetsGoFir Yes I have tried. The training seems to converge but the results are not as good as the ones with the original caffe feats. Performance on Flickr30K (Paper) (R@1,...

Well atm I don't have the npy file with the features. Nonetheless, in literature the most common split used in Flickr30K is the Karpathy split (https://cs.stanford.edu/people/karpathy/deepimagesent/). You can donwload the...

@jasscia18 FYI, issue #17 addresses this. I think they use the img_ids from the original Karpahty split.

@tonghe90 I have the same problem. It was not resolved by deleting the reverse_layer.cu file...

Here is the log: CXX src/caffe/layers/crop_layer.cpp In file included from ./include/caffe/common.hpp:19:0, from ./include/caffe/blob.hpp:8, from ./include/caffe/layers/reverse_axis_layer.hpp:6, from src/caffe/layers/bkp/reverse_axis_layer.cpp:1: ./include/caffe/util/device_alternate.hpp:14:15: error: expected initializer before ‘

@tonghe90 Well the problem appears with GPU and CPU, so I guess it is not the CUDNN version.

Hello, I already tried with your config file. The problem persist CXX src/caffe/solvers/rmsprop_solver.cpp CXX src/caffe/solvers/sgd_solver.cpp CXX src/caffe/data_transformer.cpp CXX src/caffe/layer_factory.cpp NVCC src/caffe/util/math_functions.cu NVCC src/caffe/util/im2col.cu NVCC src/caffe/layers/cudnn_conv_layer.cu NVCC src/caffe/layers/pooling_layer.cu NVCC src/caffe/layers/prelu_layer.cu NVCC...

@tonghe90 I lowered gcc to version 4.8 but the problem persists. The same data type errors on the same files. I have no idea or clue what could be wrong...

Apparently this code only works with Cuda 8.0 and CUDNN 6.0. Maybe you should try to include this info as well.