IndexError in test_gradients

[fonnesbeck]

Two identical failures occur during calls to check_gradients in the test_gradients unit tests:

Traceback (most recent call last):
  File "/Users/fonnescj/Code/pymc/pymc/tests/test_gradients.py", line 286, in test_gradients
    check_gradients(norm2)
  File "/Users/fonnescj/Code/pymc/pymc/tests/test_gradients.py", line 107, in check_gradients
    numeric_gradient = get_numeric_gradient(stochastics, s)
  File "/Users/fonnescj/Code/pymc/pymc/tests/test_gradients.py", line 134, in get_numeric_gradient
    delta[unravel_index(i,i_shape or (1,))] += e
IndexError: 0-d arrays can only use a single () or a list of newaxes (and a single ...) as an index

Also occurs on line 360 of test_gradients.

[fonnesbeck]

I dealt with the IndexError (commit 9988ebb), but the gradient assertion does not appear to fare well:

======================================================================
FAIL: test_gradients (pymc.tests.test_gradients.test_gradients)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/Users/fonnescj/Code/pymc/pymc/tests/test_gradients.py", line 291, in test_gradients
    check_gradients(gamma)
  File "/Users/fonnescj/Code/pymc/pymc/tests/test_gradients.py", line 112, in check_gradients
    " is not correct.")
  File "/Library/Python/2.7/site-packages/numpy-2.0.0.dev_730b861_20110815-py2.7-macosx-10.7-x86_64.egg/numpy/testing/utils.py", line 799, in assert_array_almost_equal
    header=('Arrays are not almost equal to %d decimals' % decimal))
  File "/Library/Python/2.7/site-packages/numpy-2.0.0.dev_730b861_20110815-py2.7-macosx-10.7-x86_64.egg/numpy/testing/utils.py", line 635, in assert_array_compare
    raise AssertionError(msg)
AssertionError: 
Arrays are not almost equal to 4 decimals
analytic gradient for gamma with respect to parameter a is not correct.
(mismatch 100.0%)
 x: array([-0.96673602, -1.41302081, -2.3154314 , -2.84824608, -1.73328374])
 y: array([-0.54395252, -0.99024297, -1.89264995, -2.42546566, -1.31050626])

[fonnesbeck]

This error seems to have gone away.