diff --git a/part_segmentation/models/__init__.py b/part_segmentation/models/__init__.py
index 75a221d..83a619e 100755
--- a/part_segmentation/models/__init__.py
+++ b/part_segmentation/models/__init__.py
@@ -1,5 +1,2 @@
 from __future__ import absolute_import
-from .curvenet_seg import CurveNet
-from .pointMLP31 import model31G, model31J, model31K, model31D, model31A
-from .pointMLP32 import model32A, model32B, model32C, model32D, model32E, model32F, model32G
-from .pointMLP33 import model33A, model33B, model33C, model33D, model33E, model33F, model33G
+from .pointMLP import pointMLP
diff --git a/part_segmentation/models/pointMLP31.py b/part_segmentation/models/pointMLP.py
similarity index 70%
rename from part_segmentation/models/pointMLP31.py
rename to part_segmentation/models/pointMLP.py
index f3a373c..5ce90fb 100644
--- a/part_segmentation/models/pointMLP31.py
+++ b/part_segmentation/models/pointMLP.py
@@ -1,13 +1,3 @@
-"""
-Based on PointMLP9, change to 128, 256, 512, 512
-Based on PointMLP4, use fps replace random sample.
-Based on PointMLP3, change dropout to 0.1
-Based on PointMLP2, add blocks in decode, change channel numbers.
-Based on PointMLP1, using poseExtraction to replace MLP in decode.
-PointsformerE2, 1)change relu to GELU, 2) change backbone to model24.
-"""
-
-
 
 import torch
 import torch.nn as nn
@@ -337,14 +327,14 @@ class PointNetFeaturePropagation(nn.Module):
 
 
 
-class PointMLP31(nn.Module):
+class PointMLP(nn.Module):
     def __init__(self, num_classes=50,points=2048, embed_dim=64, groups=1, res_expansion=1.0,
                  activation="relu", bias=True, use_xyz=True, normalize="anchor",
                  dim_expansion=[2, 2, 2, 2], pre_blocks=[2, 2, 2, 2], pos_blocks=[2, 2, 2, 2],
                  k_neighbors=[32, 32, 32, 32], reducers=[4, 4, 4, 4],
                  de_dims=[512, 256, 128, 128], de_blocks=[2,2,2,2],
                  gmp_dim=64,cls_dim=64, **kwargs):
-        super(PointMLP31, self).__init__()
+        super(PointMLP, self).__init__()
         self.stages = len(pre_blocks)
         self.class_num = num_classes
         self.points = points
@@ -455,143 +445,20 @@ class PointMLP31(nn.Module):
         return x
 
 
-def model31A(num_classes=50, **kwargs) -> PointMLP31:
-    return PointMLP31(num_classes=num_classes, points=2048, embed_dim=64, groups=1, res_expansion=1.0,
-                 activation="relu", bias=True, use_xyz=False, normalize="anchor",
-                 dim_expansion=[2, 2, 2, 2], pre_blocks=[2, 2, 2, 2], pos_blocks=[2, 2, 2, 2],
-                 k_neighbors=[32, 32, 32, 32], reducers=[4, 4, 4, 4],
-                 de_dims=[512, 256, 128, 128], de_blocks=[4,4,4,4],
-                 gmp_dim=64,cls_dim=64, **kwargs)
-
-
-def model31B(num_classes=50, **kwargs) -> PointMLP31:
-    return PointMLP31(num_classes=num_classes,points=2048, embed_dim=64, groups=1, res_expansion=1.0,
-                 activation="relu", bias=True, use_xyz=False, normalize="anchor",
-                 dim_expansion=[2, 2, 2, 2], pre_blocks=[3, 3, 3, 3], pos_blocks=[3, 3, 3, 3],
-                 k_neighbors=[32, 32, 32, 32], reducers=[4, 4, 4, 4],
-                 de_dims=[512, 256, 128, 128], de_blocks=[3, 3, 3, 3],
-                 gmp_dim=64,cls_dim=64, **kwargs)
-
-def model31C(num_classes=50, **kwargs) -> PointMLP31:
-    return PointMLP31(num_classes=num_classes,points=2048, embed_dim=64, groups=1, res_expansion=1.0,
-                 activation="relu", bias=True, use_xyz=False, normalize="anchor",
-                 dim_expansion=[2, 2, 2, 2], pre_blocks=[2, 2, 2, 2], pos_blocks=[2, 2, 2, 2],
-                 k_neighbors=[24, 24, 24, 24], reducers=[4, 4, 4, 4],
-                 de_dims=[512, 256, 128, 128], de_blocks=[3, 3, 3, 3],
-                 gmp_dim=64,cls_dim=64, **kwargs)
-
-
-
-def model31D(num_classes=50, **kwargs) -> PointMLP31:
-    return PointMLP31(num_classes=num_classes, points=2048, embed_dim=32, groups=1, res_expansion=1.0,
-                 activation="relu", bias=True, use_xyz=False, normalize="anchor",
-                 dim_expansion=[2, 2, 2, 2], pre_blocks=[2, 2, 2, 2], pos_blocks=[2, 2, 2, 2],
-                 k_neighbors=[32, 32, 32, 32], reducers=[4, 4, 4, 4],
-                 de_dims=[512, 256, 128, 128], de_blocks=[4,4,4,4],
-                 gmp_dim=64,cls_dim=64, **kwargs)
-
-
-def model31E(num_classes=50, **kwargs) -> PointMLP31:
-    return PointMLP31(num_classes=num_classes, points=2048, embed_dim=64, groups=1, res_expansion=1.0,
-                 activation="relu", bias=True, use_xyz=False, normalize="anchor",
-                 dim_expansion=[2, 2, 2, 2], pre_blocks=[4, 4, 4, 4], pos_blocks=[4, 4, 4, 4],
-                 k_neighbors=[32, 32, 32, 32], reducers=[4, 4, 4, 4],
-                 de_dims=[512, 256, 128, 128], de_blocks=[6,6,6,6],
-                 gmp_dim=64,cls_dim=64, **kwargs)
-
-def model31F(num_classes=50, **kwargs) -> PointMLP31:
-    return PointMLP31(num_classes=num_classes, points=2048, embed_dim=64, groups=1, res_expansion=1.0,
-                 activation="relu", bias=True, use_xyz=False, normalize="anchor",
-                 dim_expansion=[2, 2, 2, 2], pre_blocks=[2, 2, 2, 2], pos_blocks=[2, 2, 2, 2],
-                 k_neighbors=[32, 32, 32, 32], reducers=[4, 2, 4, 2],
-                 de_dims=[512, 256, 128, 128], de_blocks=[4,4,4,4],
-                 gmp_dim=64,cls_dim=64, **kwargs)
-
-def model31G(num_classes=50, **kwargs) -> PointMLP31:
-    return PointMLP31(num_classes=num_classes, points=2048, embed_dim=64, groups=1, res_expansion=1.0,
+def pointMLP(num_classes=50, **kwargs) -> PointMLP:
+    return PointMLP(num_classes=num_classes, points=2048, embed_dim=64, groups=1, res_expansion=1.0,
                  activation="relu", bias=True, use_xyz=True, normalize="anchor",
                  dim_expansion=[2, 2, 2, 2], pre_blocks=[2, 2, 2, 2], pos_blocks=[2, 2, 2, 2],
                  k_neighbors=[32, 32, 32, 32], reducers=[4, 4, 4, 4],
                  de_dims=[512, 256, 128, 128], de_blocks=[4,4,4,4],
                  gmp_dim=64,cls_dim=64, **kwargs)
 
-def model31G1(num_classes=50, **kwargs) -> PointMLP31:
-    return PointMLP31(num_classes=num_classes, points=2048, embed_dim=64, groups=1, res_expansion=1.0,
-                 activation="leakyrelu0.2", bias=True, use_xyz=True, normalize="anchor",
-                 dim_expansion=[2, 2, 2, 2], pre_blocks=[2, 2, 2, 2], pos_blocks=[2, 2, 2, 2],
-                 k_neighbors=[32, 32, 32, 32], reducers=[4, 4, 4, 4],
-                 de_dims=[512, 256, 128, 128], de_blocks=[4,4,4,4],
-                 gmp_dim=64,cls_dim=64, **kwargs)
-
-def model31G2(num_classes=50, **kwargs) -> PointMLP31:
-    return PointMLP31(num_classes=num_classes, points=2048, embed_dim=64, groups=1, res_expansion=1.0,
-                 activation="leakyrelu", bias=True, use_xyz=True, normalize="anchor",
-                 dim_expansion=[2, 2, 2, 2], pre_blocks=[2, 2, 2, 2], pos_blocks=[2, 2, 2, 2],
-                 k_neighbors=[32, 32, 32, 32], reducers=[4, 4, 4, 4],
-                 de_dims=[512, 256, 128, 128], de_blocks=[4,4,4,4],
-                 gmp_dim=64,cls_dim=64, **kwargs)
-
-def model31G3(num_classes=50, **kwargs) -> PointMLP31:
-    return PointMLP31(num_classes=num_classes, points=2048, embed_dim=64, groups=1, res_expansion=1.0,
-                 activation="gelu", bias=True, use_xyz=True, normalize="anchor",
-                 dim_expansion=[2, 2, 2, 2], pre_blocks=[2, 2, 2, 2], pos_blocks=[2, 2, 2, 2],
-                 k_neighbors=[32, 32, 32, 32], reducers=[4, 4, 4, 4],
-                 de_dims=[512, 256, 128, 128], de_blocks=[4,4,4,4],
-                 gmp_dim=64,cls_dim=64, **kwargs)
-
-def model31H(num_classes=50, **kwargs) -> PointMLP31:
-    return PointMLP31(num_classes=num_classes, points=2048, embed_dim=64, groups=1, res_expansion=1.0,
-                 activation="gelu", bias=True, use_xyz=True, normalize="anchor",
-                 dim_expansion=[2, 2, 2, 2], pre_blocks=[2, 2, 2, 2], pos_blocks=[2, 2, 2, 2],
-                 k_neighbors=[32, 32, 32, 32], reducers=[4, 4, 4, 4],
-                 de_dims=[512, 256, 128, 128], de_blocks=[4,4,4,4],
-                 gmp_dim=64,cls_dim=64, **kwargs)
-
-def model31I(num_classes=50, **kwargs) -> PointMLP31:
-    return PointMLP31(num_classes=num_classes, points=2048, embed_dim=64, groups=1, res_expansion=1.0,
-                 activation="relu", bias=True, use_xyz=True, normalize="anchor",
-                 dim_expansion=[2, 2, 2, 1], pre_blocks=[3, 3, 3, 3], pos_blocks=[3, 3, 3, 3],
-                 k_neighbors=[32, 32, 32, 32], reducers=[4, 4, 4, 4],
-                 de_dims=[512, 256, 128, 128], de_blocks=[4, 4, 4, 4],
-                 gmp_dim=64,cls_dim=64, **kwargs)
-
-def model31J(num_classes=50, **kwargs) -> PointMLP31:
-    return PointMLP31(num_classes=num_classes, points=2048, embed_dim=64, groups=1, res_expansion=1.0,
-                 activation="leakyrelu", bias=True, use_xyz=True, normalize="anchor",
-                 dim_expansion=[2, 2, 2, 2], pre_blocks=[2, 2, 2, 2], pos_blocks=[2, 2, 2, 2],
-                 k_neighbors=[32, 32, 32, 32], reducers=[4, 4, 4, 4],
-                 de_dims=[512, 256, 128, 128], de_blocks=[4,4,4,4],
-                 gmp_dim=64,cls_dim=64, **kwargs)
-
-
-def model31K(num_classes=50, **kwargs) -> PointMLP31:
-    return PointMLP31(num_classes=num_classes, points=2048, embed_dim=64, groups=1, res_expansion=1.0,
-                 activation="leakyrelu0.2", bias=True, use_xyz=True, normalize="anchor",
-                 dim_expansion=[2, 2, 2, 2], pre_blocks=[2, 2, 2, 2], pos_blocks=[2, 2, 2, 2],
-                 k_neighbors=[32, 32, 32, 32], reducers=[4, 4, 4, 4],
-                 de_dims=[512, 256, 128, 128], de_blocks=[4,4,4,4],
-                 gmp_dim=64,cls_dim=64, **kwargs)
-
-
-
 
 if __name__ == '__main__':
     data = torch.rand(2, 3, 2048)
     norm = torch.rand(2, 3, 2048)
     cls_label = torch.rand([2, 16])
     print("===> testing modelD ...")
-    model = model31D(50)
+    model = model31G(50)
     out = model(data, cls_label)  # [2,2048,50]
     print(out.shape)
-
-    model = model31A(50)
-    out = model(data, cls_label)  # [2,2048,50]
-    print(out.shape)
-
-    model = model31B(50)
-    out = model(data, cls_label)  # [2,2048,50]
-    print(out.shape)
-
-    model = model31C(50)
-    out = model(data, cls_label)  # [2,50, 2048]
-    print(out.shape)