add doc

i6_models/assemblies/lstm/lstm_v1.py

@@ -14,6 +14,16 @@
 
 @dataclass
 class LstmEncoderV1Config(ModelConfiguration):
+    """
+    :param init_args: used to initialize parameters of modules, example:
+                      ```
+                        {
+                            "init_args_w": {"func": "normal", "arg": {"mean": 0.0, "std": 0.1}},
+                            "init_args_b": {"func": "normal", "arg": {"mean": 0.0, "std": 0.1}},
+                        }
+                      ```
+    """
+
     input_dim: int
     embed_dim: int
     embed_dropout: float
@@ -30,6 +40,13 @@ def from_dict(cls, model_cfg_dict: Dict[str, Any]):
 
 class LstmEncoderV1(nn.Module):
     def __init__(self, model_cfg: Union[LstmEncoderV1Config, Dict[str, Any]], **kwargs):
+        """
+        Model definition of LSTM encoder. Contains embedding layer followed by single lstm stack, dropout after both.
+        Padding sequence in forward call.
+
+        :param model_cfg: holds model configuration as dataclass or dict instance.
+        :param kwargs:
+        """
         super().__init__()
 
         self.cfg = LstmEncoderV1Config.from_dict(model_cfg) if isinstance(model_cfg, dict) else model_cfg