Two questions about CAFFE.

1. I think the CAFFE has been revised for python interface. But the speed extraction speed is slower than previous version. about 6 to 7s per image. instead of 300ms as it used to be.
   I checked the wrapper " out = self.forward_all(**{self.inputs[0]: caffe_in})" this line took about 7s for each image.
   Could anyone help me this question? Thanks in advance.
2. Another question is about the pycaffe.py. I think the python 2.6 cannot read this file. There is some error on the screen.

File "../python/caffe/pycaffe.py", line 69
outs = {out: self.blobs[out].data for out in set(self.outputs + blobs)}

Thanks.