update docstrings (#20)

Author: Atticus1806

i6_models/parts/conformer/convolution.py

@@ -12,16 +12,20 @@
 
 @dataclass
 class ConformerConvolutionV1Config(ModelConfiguration):
+    """
+    Attributes:
+        channels: number of channels for conv layers
+        kernel_size: kernel size of conv layers
+        dropout: dropout probability
+        activation: activation function applied after normalization
+        norm: normalization layer with input of shape [N,C,T]
+    """
+
     channels: int
-    """number of channels for conv layers"""
     kernel_size: int
-    """kernel size of conv layers"""
     dropout: float
-    """dropout probability"""
     activation: Union[nn.Module, Callable[[torch.Tensor], torch.Tensor]]
-    """activation function applied after norm"""
     norm: Union[nn.Module, Callable[[torch.Tensor], torch.Tensor]]
-    """normalization layer with input of shape [N,C,T]"""
 
     def check_valid(self):
         assert self.kernel_size % 2 == 1, "ConformerConvolutionV1 only supports odd kernel sizes"

i6_models/parts/conformer/feedforward.py

@@ -13,14 +13,18 @@
 
 @dataclass
 class ConformerPositionwiseFeedForwardV1Config(ModelConfiguration):
+    """
+    Attributes:
+        input_dim: input dimension
+        hidden_dim: hidden dimension (normally set to 4*input_dim as suggested by the paper)
+        dropout: dropout probability
+        activation: activation function
+    """
+
     input_dim: int
-    """input dimension"""
     hidden_dim: int
-    """hidden dimension (normally set to 4*input_dim as suggested by the paper)"""
     dropout: float
-    """dropout probability"""
     activation: Callable[[torch.Tensor], torch.Tensor] = nn.functional.silu
-    """activation function"""
 
 
 class ConformerPositionwiseFeedForwardV1(nn.Module):

i6_models/parts/conformer/mhsa.py

@@ -10,14 +10,18 @@
 
 @dataclass
 class ConformerMHSAV1Config(ModelConfiguration):
+    """
+    Attributes:
+        input_dim: input dim and total dimension for query/key and value projections, should be divisible by `num_att_heads`
+        num_att_heads: number of attention heads
+        att_weights_dropout: attention weights dropout
+        dropout: multi-headed self attention output dropout
+    """
+
     input_dim: int
-    """input dim and total dimension for query/key and value projections, should be divisible by `num_att_heads`"""
     num_att_heads: int
-    """number of attention heads"""
     att_weights_dropout: float
-    """attention weights dropout"""
     dropout: float
-    """multi-headed self attention output dropout"""
 
     def __post_init__(self) -> None:
         super().__post_init__()