ncut-pytorch 3.0.0.dev8__tar.gz → 3.0.0.dev9__tar.gz

{ncut_pytorch-3.0.0.dev8 → ncut_pytorch-3.0.0.dev9}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: ncut_pytorch
-Version: 3.0.0.dev8
+Version: 3.0.0.dev9
 Summary: Normalized Cut and Spectral Embedding
 Author-email: Huzheng Yang <huze.yann@gmail.com>
 License-Expression: MIT

{ncut_pytorch-3.0.0.dev8 → ncut_pytorch-3.0.0.dev9}/ncut_pytorch/color/mspace.py

@@ -232,7 +232,7 @@ class TrainEncoder(pl.LightningModule):
                  lr=1e-3, progress_bar=True, training_steps=1000,
                  encoder_activation='gelu', decoder_activation='gelu', 
                  final_activation='identity',
-                 log_grad_norm=True, 
+                 log_grad_norm=False, 
                  **kwargs):
         super().__init__()
         
@@ -396,7 +396,7 @@ def suppress_lightning_logs(func):
 
 
 @suppress_lightning_logs
-def train_mspace_model(compress_feats, uncompress_feats, training_steps=1000, decoder_training_steps=1000,
+def train_mspace_model(compress_feats, uncompress_feats, encoder_training_steps=1000, decoder_training_steps=1000,
                     batch_size=1000, return_trainer=False, progress_bar=False,
                     logger=False, use_wandb=False, **model_kwargs):
     # check args
@@ -410,7 +410,7 @@ def train_mspace_model(compress_feats, uncompress_feats, training_steps=1000, de
     l, c_in = compress_feats.shape
     c_out = uncompress_feats.shape[1]
 
-    model = TrainEncoder(c_in, c_out, training_steps=training_steps, progress_bar=progress_bar, **model_kwargs)
+    model = TrainEncoder(c_in, c_out, training_steps=encoder_training_steps, progress_bar=progress_bar, **model_kwargs)
     
     dataset = TensorDataset(compress_feats, uncompress_feats)
     dataloader = torch.utils.data.DataLoader(dataset, batch_size=batch_size, shuffle=True, num_workers=0)
@@ -427,7 +427,7 @@ def train_mspace_model(compress_feats, uncompress_feats, training_steps=1000, de
         logger = pl.loggers.WandbLogger(project='mspace', name='mspace')
 
     # train the autoencoder jointly
-    trainer = pl.Trainer(max_steps=training_steps, logger=logger, **trainer_args)
+    trainer = pl.Trainer(max_steps=encoder_training_steps, logger=logger, **trainer_args)
     trainer.fit(model, dataloader)
     
     mspace_ae = model.mspace_ae

{ncut_pytorch-3.0.0.dev8 → ncut_pytorch-3.0.0.dev9}/ncut_pytorch.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: ncut_pytorch
-Version: 3.0.0.dev8
+Version: 3.0.0.dev9
 Summary: Normalized Cut and Spectral Embedding
 Author-email: Huzheng Yang <huze.yann@gmail.com>
 License-Expression: MIT

{ncut_pytorch-3.0.0.dev8 → ncut_pytorch-3.0.0.dev9}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "ncut_pytorch"
-version = "3.0.0dev8"
+version = "3.0.0dev9"
 authors = [
     { name = "Huzheng Yang", email = "huze.yann@gmail.com" },
 ]