diff --git a/mergekit/scripts/ABM/extract_activations.py b/mergekit/scripts/ABM/extract_activations.py index 96b7695b..7cb5961b 100644 --- a/mergekit/scripts/ABM/extract_activations.py +++ b/mergekit/scripts/ABM/extract_activations.py @@ -132,11 +132,6 @@ def main( model_config = model.config() model_arch_info = get_architecture_info(model_config) - # things to do: find the residual space - # rest difference the ignore_spaces - # residual space is the one to for hidden states - # the rest we will attach hooks based on the module name - _json = model_arch_info.definition residual_space = None diff --git a/mergekit/scripts/ABM/extract_permutation_matrices.py b/mergekit/scripts/ABM/extract_permutation_matrices.py index 84a98e4b..75c58692 100644 --- a/mergekit/scripts/ABM/extract_permutation_matrices.py +++ b/mergekit/scripts/ABM/extract_permutation_matrices.py @@ -91,10 +91,8 @@ def match_tensors_permute_MHA( corr_submatrix, maximize=True ) - # store cost (cost is maximized here) costs[j, k] = corr_submatrix[row_ind, col_ind].sum() - # store perm so we don't have to recompute it later col_inds_storage[j][k] = col_ind outer_row_ind, outer_col_ind = scipy.optimize.linear_sum_assignment(