remove repeated open_ended_activations processing (#58)

This caused errors when attempting to run
feature_intervention_generate() with and without
return_activations=True.

Author: jakobhansen-blai

circuit_tracer/replacement_model.py

@@ -881,10 +881,6 @@ def clear_and_add_hooks(tensor, hook):
         self.reset_hooks()
 
         logits = torch.cat((logit_cache[0], *open_ended_logits), dim=1)  # type:ignore
-        open_ended_activations = torch.stack(
-            [torch.cat(acts, dim=0) for acts in open_ended_activations],  # type:ignore
-            dim=0,
-        )
         if return_activations:
             activation_cache = torch.stack(activation_cache)
             if open_ended_activations and any(acts for acts in open_ended_activations):