Update generate.py

generate.py

@@ -130,8 +130,8 @@ def generate(
     forced_eos_token_id= False,
     remove_invalid_values=None,
     synced_gpus=None,
-    n_ahead=1,
-    n_ahead_talk=1,
+    n_ahead=12,
+    n_ahead_talk=4,
     merged_talk_heads=True,
     merged_lm_and_talk_heads=False,
     merged_lm_and_think_heads=True,
@@ -161,7 +161,7 @@ def generate(
     self.use_end_thought_token = True
     self.use_start_thought_token = True
     self.n_ahead = n_ahead
-    self.n_passes = 1
+    self.n_passes = 2
     self.eval_mode = True
     self.first_run = False
     self.rm_initialized = True