Skip to first unread message
林俊凡
May 25, 2016, 6:54:37 AM5/25/16
to Caffe Users
Hi,
Nice to meet you, my name is Jack.
I'm a freshman to Caffe and I had already followed several tutorials. And recently, I want to re-implement the logreg tutorial by using MemoryData layer in Python instead of Data layer with hdf5. However, it didn't work at all, and when I ran solver.solve(), the loss finally turned into NAN(It's nothing to do with the lr_base).... I found that when I used solver.step(1), the 'data' blob's data was very strange and it's different from the data I input.
Two files attached:
the html file of ipynb
the ipynb
I pointed out the strange behavior in the last several cell in the notebook.
Could you tell me what's wrong with my implement? Thanks in advanced!
Nice to meet you, my name is Jack.
I'm a freshman to Caffe and I had already followed several tutorials. And recently, I want to re-implement the logreg tutorial by using MemoryData layer in Python instead of Data layer with hdf5. However, it didn't work at all, and when I ran solver.solve(), the loss finally turned into NAN(It's nothing to do with the lr_base).... I found that when I used solver.step(1), the 'data' blob's data was very strange and it's different from the data I input.
Two files attached:
the html file of ipynb
the ipynb
I pointed out the strange behavior in the last several cell in the notebook.
Could you tell me what's wrong with my implement? Thanks in advanced!
Reply all
Reply to author
Forward
0 new messages