Do not force float64 in the graph when it isn't needed. Do the same dtype requirement as on the CPU.

Remove a useless check. It can be a cached constant. This is safe in the make_node. inplace opt will disallow that later, but it should be safe.

Clean up

Add conv2d transpose2doc

#5742 Removed theano/sandbox/linalg/ops.py and theano/sandbox/linalg/tests/test_linalg.py from flake8 whitelist

Open tempfile in mode 'w' to prevent TypeError

.. autofunction:: theano.tensor.nnet.conv2d_transpose added in order to generate docuentation for conv2d transpose.

For printing compilation errors, a NamedTemporaryFile object is opened in the default mode ('w+b'), and strings are written to it. This is not an issue in Python 2, but in Python 3, strings cannot be implicitly converted to bytes so a TypeError occurs on line 2310 (previously 2309). This commit explicitly opens the NamedTemporaryFile in mode 'w'.

Update Windows installation documentation for Anaconda.

Fix opt crash fix and  more information when verify_grad detect an error.

print_test_value in debugging tutorial

Updated numpy as np #4218

2. Fixed indentation

2. Fixed indentation

2. Fixed indentation

2. Fixed indentation

2. Fixed indentation

2. Fixed indentation

2. Fixed indentation

2. Fixed indentation

2. Fixed indentation

2. Fixed indentation

2. Fixed indentation

2. Fixed indentation

2. Fixed indentation

2. Fixed indentation

2. Fixed indentation