Misc cleanup

vjoki · Jan 13, 2021 · 13fbf07 · 13fbf07
1 parent 06e775b
commit 13fbf07
Show file tree

Hide file tree

Showing 3 changed files with 5 additions and 16 deletions.
diff --git a/snn/librispeech/datamodule.py b/snn/librispeech/datamodule.py
@@ -109,7 +109,8 @@ def setup(self, stage: Optional[str] = None):
                                                      max_sample_length=self.max_sample_length)
 
             if self.num_train != 0:
-                self.training_sampler = torch.utils.data.RandomSampler(self.training_set, replacement=True,
+                self.training_sampler = torch.utils.data.RandomSampler(self.training_set,  # type: ignore
+                                                                       replacement=True,
                                                                        num_samples=self.num_train)
 
             self.validation_set = PairDataset(val_dataset, max_sample_length=self.max_sample_length)

diff --git a/snn/librispeech/loss/angularproto.py b/snn/librispeech/loss/angularproto.py
@@ -6,7 +6,7 @@
 
 
 class AngularPrototypicalLoss(nn.Module):
-    def __init__(self, init_scale=10.0, init_bias=-5.0):
+    def __init__(self, init_scale=10.0, init_bias=-5.0, **kwargs):
         super().__init__()
         self.w = nn.Parameter(torch.as_tensor(init_scale))
         self.b = nn.Parameter(torch.as_tensor(init_bias))

diff --git a/snn/librispeech/model/snn_angularproto.py b/snn/librispeech/model/snn_angularproto.py
@@ -1,11 +1,10 @@
-from typing import Tuple, List
+from typing import Tuple
 import torch
 import torch.nn.functional as F
 import pytorch_lightning as pl
 
 from .base import BaseNet
 from snn.librispeech.loss.angularproto import AngularPrototypicalLoss
-from resnet.utils import accuracy
 
 
 class SNNAngularProto(BaseNet):
@@ -22,7 +21,7 @@ def forward(self, x: torch.Tensor) -> torch.Tensor:  # type: ignore[override]
     def training_step(self,  # type: ignore[override]
                       batch: Tuple[torch.Tensor, torch.Tensor],
                       batch_idx: int) -> torch.Tensor:
-        support_sets, labels = batch
+        support_sets, _ = batch
         #print('support_sets', support_sets.shape)
         #print('labels', labels.shape)
 
@@ -50,17 +49,6 @@ def training_step(self,  # type: ignore[override]
                                   support.size(-1))
         assert support.size(1) == self.num_ways
 
-        # supports = []
-        # for shots in support_sets:
-        #     s = []
-        #     for waveform in shots:
-        #         x = self.spectogram_transform(waveform, augment=self.specaugment)
-        #         print('ResNet (spectro)', x.shape)
-        #         s.append(self.cnn(x))
-        #     supports.append(torch.cat(s, dim=0))
-
-        # support = torch.stack(supports, dim=0)
-
         loss, acc = self.training_loss_fn(support)
 
         self.log('train_acc_step', acc, on_step=True, on_epoch=False)