Upload SimpleMLPForClassification

config.json

@@ -2,6 +2,10 @@
   "architectures": [
     "SimpleMLPForClassification"
   ],
+  "auto_map": {
+    "AutoConfig": "simple_mlp_configuration.SimpleMLPConfig",
+    "AutoModelForSequenceClassification": "modeling_simple_mlp.SimpleMLPForClassification"
+  },
   "dropout_rate": 0.1,
   "dtype": "float32",
   "hidden_dim": 64,

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:67f8eb4584db50f83ce345511ce70e85dbdbf76eb56cf27d8ad7057880425680
+oid sha256:85bf988b2688164019bb0a0766f9cba665d10e18d2dc816439d1cfc26c90b534
 size 34124

modeling_simple_mlp.py

@@ -0,0 +1,43 @@
+import torch
+import torch.nn as nn
+from transformers import PreTrainedModel
+from simple_mlp_configuration import SimpleMLPConfig
+from transformers.modeling_outputs import SequenceClassifierOutput
+
+class SimpleMLPForClassification(PreTrainedModel):
+    config_class = SimpleMLPConfig
+
+    def __init__(self, config):
+        super().__init__(config)
+        self.config = config
+        self.num_labels = config.num_classes
+
+        self.fc1 = nn.Linear(config.input_dim, config.hidden_dim)
+        self.activation = nn.ReLU()
+        self.dropout = nn.Dropout(config.dropout_rate)
+        self.fc2 = nn.Linear(config.hidden_dim, config.num_classes)
+
+        self.post_init()
+    
+    def forward(self, inputs_embeds, labels=None, return_dict=None):
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        x = self.fc1(inputs_embeds)
+        x = self.activation(x)
+        x = self.dropout(x)
+        logits = self.fc2(x)
+
+        loss = None
+        if labels is not None:
+            loss_fct = nn.CrossEntropyLoss()
+            loss = loss_fct(logits.view(-1, self.num_labels), labels.view(-1))
+        
+        if not return_dict:
+            output = (logits,)
+            return ((loss,) + output) if loss is not None else output
+        
+        return SequenceClassifierOutput(
+            loss=loss,
+            logits=logits,
+        )
+

simple_mlp_configuration.py

@@ -0,0 +1,18 @@
+from transformers import PretrainedConfig
+
+class SimpleMLPConfig(PretrainedConfig):
+    model_type = "simple_mlp"
+
+    def __init__(
+            self,
+            input_dim=768,
+            hidden_dim=256,
+            num_classes=2,
+            dropout_rate=0.1,
+            **kwargs
+    ):
+        self.input_dim = input_dim
+        self.hidden_dim = hidden_dim
+        self.num_classes = num_classes
+        self.dropout_rate = dropout_rate
+        super().__init__(**kwargs)