Mercurial > ift6266
diff deep/convolutional_dae/scdae.py @ 292:8108d271c30c
Fix stuff (imports, ...) so that it can run under jobman properly.
| author | Arnaud Bergeron <abergeron@gmail.com> |
|---|---|
| date | Fri, 26 Mar 2010 18:49:27 -0400 |
| parents | 518589bfee55 |
| children | a222af1d0598 |
line wrap: on
line diff
--- a/deep/convolutional_dae/scdae.py Fri Mar 26 18:35:23 2010 -0400 +++ b/deep/convolutional_dae/scdae.py Fri Mar 26 18:49:27 2010 -0400 @@ -174,25 +174,27 @@ title="Reconstruction error (mse)"), reduce_every=100) - series['training_err'] = AccumulatorSeriesWrapper( + series['train_error'] = AccumulatorSeriesWrapper( base_series=ErrorSeries(error_name='training_error', table_name='training_error', hdf5_file=h5f, index_names=('iter',), - titles='Training error (nll)'), + title='Training error (nll)'), reduce_every=100) - - series['valid_err'] = ErrorSeries(error_name='valid_error', - table_name='valid_error', - hdf5_file=h5f, - index_names=('iter',), - titles='Validation error (class)') - - series['test_err'] = ErrorSeries(error_name='test_error', - table_name='test_error', - hdf5_file=h5f, - index_names=('iter',), - titles='Test error (class)') + + series['valid_error'] = ErrorSeries(error_name='valid_error', + table_name='valid_error', + hdf5_file=h5f, + index_names=('iter',), + title='Validation error (class)') + + series['test_error'] = ErrorSeries(error_name='test_error', + table_name='test_error', + hdf5_file=h5f, + index_names=('iter',), + title='Test error (class)') + + return series def run_exp(state, channel): from ift6266 import datasets @@ -206,7 +208,7 @@ # TODO: maybe record pynnet version? channel.save() - dset = dataset.nist_all() + dset = dataset.nist_all(1000) nfilts = [] if state.nfilts1 != 0: