[#41] add model init

sgnlp/models/sentic_asgcn/config.py

@@ -11,6 +11,7 @@ class SenticASGCNConfig(PreTrainedConfig):
     Args:
         embed_dim (:obj:`int`, defaults to 300): Embedding dimension size.
         hidden_dim (:obj:`int`, defaults to 300): Size of hidden dimension.
+        dropout (:obj:`float`, defaults to 0.3): Droput percentage.
         polarities_dim (:obj:`int`, defaults to 3): Size of output dimension represeting available polarities (e.g. Positive, Negative, Neutral).
         device (:obj:`torch.device`, defaults to torch.device('cuda`)): Type of torch device.
 
@@ -27,11 +28,13 @@ def __init__(
         embed_dim=300,
         hidden_dim=300,
         polarities_dim=3,
+        dropout=0.3,
         device=torch.device("cuda"),
         **kwargs
     ):
         super().__init__(**kwargs)
         self.embed_dim = embed_dim
         self.hidden_dim = hidden_dim
+        self.dropout = dropout
         self.polarities_dim = polarities_dim
         self.device = device

sgnlp/models/sentic_asgcn/config/sentic_asgcn_config.json

@@ -9,6 +9,7 @@
     "embed_dim": 300,
     "hidden_dim": 300,
     "polarities_dim": 3,
+    "dropout": 0.3,
     "save": true,
     "seed": 776,
     "device": "cuda"

sgnlp/models/sentic_asgcn/data_class.py

@@ -22,6 +22,9 @@ class SenticASGCNTrainArgs:
     hidden_dim: int = field(
         default=300, metadata={"help": "Number of neurons for hidden layer."}
     )
+    dropout: float = field(
+        default=0.3, metadata={"help": "Default value for dropout percentages."}
+    )
     polarities_dim: int = field(
         default=3, metadata={"help": "Default dimension for polarities."}
     )

sgnlp/models/sentic_asgcn/modeling.py

@@ -5,20 +5,39 @@
 from transformers import PreTrainedModel
 from transformers.file_utils import ModelOutput
 
+from .modules.dynamic_rnn import DynamicLSTM
+from .modules.gcn import GraphConvolution
+from .config import SenticASGCNConfig
+
 
 @dataclass
 class SenticASGCNModelOutput(ModelOutput):
     pass
 
 
 class SenticASGCNPreTrainedModel(PreTrainedModel):
-    # config_class =
+    """
+    An abstract class to handle weights initialization and a simple interface for download and loading pretrained models.
+    """
+
+    config_class = SenticASGCNConfig
     base_model_prefix = "sentic_asgcn"
 
     def _init_weights(self, module):
         pass
 
 
 class SenticASGCNModel(SenticASGCNPreTrainedModel):
-    def __init__(self, config):
-        pass
+    def __init__(self, config: SenticASGCNConfig):
+        super().__init__(config)
+        self.text_lstm = DynamicLSTM(
+            config.embed_dim,
+            config.hidden_dim,
+            num_layers=1,
+            batch_first=True,
+            bidirectional=True,
+        )
+        self.gc1 = GraphConvolution(2 * config.hidden_dim, 2 * config.hidden_dim)
+        self.gc2 = GraphConvolution(2 * config.hidden_dim, 2 * config.hidden_dim)
+        self.fc = nn.Linear(2 * config.hidden_dim, config.polarities_dim)
+        self.text_embed_dropout = nn.Dropout(config.dropout)

sgnlp/models/sentic_asgcn/train.py

@@ -1,5 +1,5 @@
 from .data_class import SenticASGCNTrainArgs
-from .utils import parse_args_and_load_config
+from .utils import parse_args_and_load_config, set_random_seed
 
 
 def train_model(cfg: SenticASGCNTrainArgs):
@@ -8,4 +8,6 @@ def train_model(cfg: SenticASGCNTrainArgs):
 
 if __name__ == "__main__":
     cfg = parse_args_and_load_config()
+    if cfg.seed is not None:
+        set_random_seed(cfg.seed)
     train_model(cfg)