Moved consider_constant from theano.tensor.extra_ops to theano.gradient, moved the optimization that removes it to theano.tensor.opt. Modified theano.compile.builders so it doesn't import theano.gradient, because theano.gradient now imports ViewOp from theano.compile and this would lead to a circular dependency.

Add fast_compile tag so the optimization to remove the Op always runs. Use g_out.zeros_like instead of tensor.zeros_like so the op also works for sparse variables.

added consider_constant op, which truncates the gradient of an expression to 0. Also added an optimization to remove it from the graph, so other optimizations can be applied properly.

test whether the gradient is computed correctly, and test whether the consider_constant op is removed from the graph.

Update to the blas ops following API change in compyte.

theano.tensor.batched_tensordot(x, y, axes)

Fix test in py26 (travis-ci)

Lift subtensor constant indexes as constants in the graph rather than keep them op-internal.

Fix test in python 3.

[MRG]Doc

This is related to the unpickle problem of ndarray.

Fix numerous grammar, syntax and stylistic problems.

Fix Variable.eval() (issue #1724 on github)

Canopy