Unnecessary call to `fit()` before `cross_validate`

[Rob Crowell]

ml-class/scikit/pipeline-bigrams-cross-validate.py

Line 19 in e50faa3

 p.fit(fixed_text, fixed_target) 

this confused me for a while, re: my question about whether when we do cross-validation do we start with a new model each iteration or do we improve an existing model.