diff --git a/configuration_moss_audio_tokenizer.py b/configuration_moss_audio_tokenizer.py
index d82791b..67c3fd0 100644
--- a/configuration_moss_audio_tokenizer.py
+++ b/configuration_moss_audio_tokenizer.py
@@ -250,26 +250,6 @@ def __init__(
                     "module_type": "PatchedPretransform",
                     "patch_size": 2,
                 },
-                {
-                    "module_type": "Transformer",
-                    "input_dimension": 384,
-                    "output_dimension": 768,
-                    "d_model": 768,
-                    "num_heads": 12,
-                    "num_layers": 12,
-                    "dim_feedforward": 3072,
-                    "causal": True,
-                    "norm": "layer_norm",
-                    "positional_embedding": "rope",
-                    "max_period": 10000,
-                    "gating": "none",
-                    "layer_scale": 0.01,
-                    "conv_layout": True,
-                },
-                {
-                    "module_type": "PatchedPretransform",
-                    "patch_size": 2,
-                },
                 {
                     "module_type": "Transformer",
                     "input_dimension": 384,