don't squeeze blob arrays for python

Preserve the non-batch dimensions of blob arrays, even for singletons.

The forward() and backward() helpers take lists of ndarrays instead of a
single ndarray per blob, and lists of ndarrays are likewise returned.

Note that for output the blob array could actually be returned as a
single ndarray instead of a list.

diff --git a/python/caffe/pycaffe.py b/python/caffe/pycaffe.py
index 5275a07..d965227 100644 (file)
--- a/python/caffe/pycaffe.py
+++ b/python/caffe/pycaffe.py
@@ -77,7 +77,7 @@ def _Net_forward(self, blobs=None, **kwargs):
   out_blobs.extend([self.blobs[blob].data for blob in blobs])
   out_blob_names = self.outputs + blobs
   for out, out_blob in zip(out_blob_names, out_blobs):
-    outs[out] = [out_blob[ix, :, :, :].squeeze()
+    outs[out] = [out_blob[ix, :, :, :]
                   for ix in range(out_blob.shape[0])]
   return outs
 
@@ -118,7 +118,7 @@ def _Net_backward(self, diffs=None, **kwargs):
   out_diffs.extend([self.blobs[diff].diff for diff in diffs])
   out_diff_names = self.inputs + diffs
   for out, out_diff in zip(out_diff_names, out_diffs):
-    outs[out] = [out_diff[ix, :, :, :].squeeze()
+    outs[out] = [out_diff[ix, :, :, :]
                            for ix in range(out_diff.shape[0])]
   return outs