train psdf回退

brep2sdf/train.py

@@ -307,7 +307,7 @@ class Trainer:
         self.model.train()
         total_loss = 0.0
         step = 0 # 如果你的训练是分批次的，这里应该用批次索引
-        batch_size = 8192  # 设置合适的batch大小
+        batch_size = 8192*16  # 设置合适的batch大小
 
         # 数据处理
         # manfld
@@ -319,19 +319,16 @@ class Trainer:
         if epoch % 10 == 1 or self.cached_train_data is None:
             # 计算流形点的掩码和操作符
             # 生成非流形点
-            _psdf_pnts, _psdf = self.sampler.get_norm_points(_mnfld_pnts, _normals,local_sigma=0.001)
+            _nonmnfld_pnts, _psdf = self.sampler.get_norm_points(_mnfld_pnts, _normals)
-            _nonmnfld_pnts = self.sampler.get_points(_mnfld_pnts, local_sigma=0.01):
             
             # 更新缓存
             self.cached_train_data = {
                 "nonmnfld_pnts": _nonmnfld_pnts,
-                "psdf_pnts": _psdf_pnts,
                 "psdf": _psdf,
             }
         else:
             # 从缓存中读取数据
             _nonmnfld_pnts = self.cached_train_data["nonmnfld_pnts"]
-            _psdf_pnts =  self.cached_train_data["psdf_pnts"]
             _psdf = self.cached_train_data["psdf"]
         
         logger.debug((_mnfld_pnts, _nonmnfld_pnts, _psdf))
@@ -353,13 +350,11 @@ class Trainer:
             
             # 非流形点使用缓存数据（整个batch共享）
             nonmnfld_pnts = _nonmnfld_pnts[start_idx:end_idx]
-            psdf_pnts = _psdf_pnts[start_idx:end_idx]
             psdf = _psdf[start_idx:end_idx]
 
             # --- 准备模型输入，启用梯度 ---
             mnfld_pnts.requires_grad_(True) # 在检查之后启用梯度
             nonmnfld_pnts.requires_grad_(True) # 在检查之后启用梯度
-            psdf_pnts.requires_grad_(True) # 在检查之后启用梯度
 
             # --- 前向传播 ---
             mnfld_pred = self.model.forward_background(
@@ -368,9 +363,6 @@ class Trainer:
             nonmnfld_pred = self.model.forward_background(
                 nonmnfld_pnts
             )
-            psdf_pred = self.model.forward_background(
-                psdf_pnts
-            )
 
 
 
@@ -390,7 +382,6 @@ class Trainer:
                     loss, loss_details = self.loss_manager.compute_loss(
                         mnfld_pnts,
                         nonmnfld_pnts,
-                        psdf_pnts,
                         normals, # 传递检查过的 normals
                         gt_sdf,
                         mnfld_pred,