Thank you for your implementation of Densenet Caffe and I can already train Cifar 10 dataset. However, when it comes to my own dataset, it will be stuck for a long time at here:
After a long time (about 1hour), the process will be killed without any error information.
My input size of images is 512*512. I use "train_test_BCBN_C10plusNoBias.prototxt" for my network architecture and I only modified the input and the "num_output" in InnerProduct1 layer. Do you think it is too large? Or I have made some other mistakes? Can you give me some advice? Thanks a lot!
Hi Tongcheng,
Thank you for your implementation of Densenet Caffe and I can already train Cifar 10 dataset. However, when it comes to my own dataset, it will be stuck for a long time at here:
I0111 00:30:41.302551 8065 layer_factory.hpp:77] Creating layer Data1 I0111 00:30:41.337936 8065 db_lmdb.cpp:35] Opened lmdb /home/gaia/Dev/caffe/examples/Apollo/lmdb_test/apollo_general18_train_lmdb I0111 00:30:41.356104 8065 net.cpp:86] Creating Layer Data1 I0111 00:30:41.356127 8065 net.cpp:382] Data1 -> Data1 I0111 00:30:41.356150 8065 net.cpp:382] Data1 -> Data2 I0111 00:30:41.357967 8065 data_layer.cpp:48] output data size: 1,3,512,512
After a long time (about 1hour), the process will be killed without any error information. My input size of images is 512*512. I use "train_test_BCBN_C10plusNoBias.prototxt" for my network architecture and I only modified the input and the "num_output" in InnerProduct1 layer. Do you think it is too large? Or I have made some other mistakes? Can you give me some advice? Thanks a lot!