Upload RobertaEmotion

config.json

@@ -2,12 +2,10 @@
   "architectures": [
     "RobertaEmotion"
   ],
-  "attention_probs_dropout_prob": 0.1,
-  "bos_token_id": 0,
-  "classifier_dropout": null,
-  "eos_token_id": 2,
-  "hidden_act": "gelu",
-  "hidden_dropout_prob": 0.1,
+  "auto_map": {
+    "AutoConfig": "configuration_roberta_emotion.RobertaEmotionConfig",
+    "AutoModel": "modeling_roberta_emotion.RobertaEmotion"
+  },
   "hidden_size": 768,
   "id2label": {
     "0": "sadness",
@@ -17,8 +15,6 @@
     "4": "fear",
     "5": "surprise"
   },
-  "initializer_range": 0.02,
-  "intermediate_size": 3072,
   "label2id": {
     "anger": 3,
     "fear": 4,
@@ -27,16 +23,7 @@
     "sadness": 0,
     "surprise": 5
   },
-  "layer_norm_eps": 1e-05,
-  "max_position_embeddings": 514,
-  "model_type": "roberta",
-  "num_attention_heads": 12,
-  "num_hidden_layers": 12,
-  "pad_token_id": 1,
-  "position_embedding_type": "absolute",
+  "model_type": "ma2za/roberta-emotion",
   "torch_dtype": "float32",
-  "transformers_version": "4.28.1",
-  "type_vocab_size": 1,
-  "use_cache": true,
-  "vocab_size": 50265
+  "transformers_version": "4.28.1"
 }

configuration_roberta_emotion.py

@@ -0,0 +1,7 @@
+from transformers import PretrainedConfig
+
+class RobertaEmotionConfig(PretrainedConfig):
+    model_type = "ma2za/roberta-emotion"
+
+    def __init__(self, **kwargs):
+        super().__init__(**kwargs)

modeling_roberta_emotion.py

@@ -0,0 +1,32 @@
+from torch.nn import CrossEntropyLoss
+
+import torch 
+
+from transformers import AutoModel, PreTrainedModel
+
+from transformers.modeling_outputs import SequenceClassifierOutput
+
+from .configuration_roberta_emotion import RobertaEmotionConfig
+
+class RobertaEmotion(PreTrainedModel):
+    config_class = RobertaEmotionConfig
+
+    def __init__(self, config):
+        super().__init__(config)
+        self.num_labels = config.num_labels
+        self.backbone = AutoModel.from_pretrained("roberta-base", config)
+        self.dropout = torch.nn.Dropout(p=0.1)
+        self.output = torch.nn.Linear(config.hidden_size, config.num_labels)
+
+    def forward(self, input_ids, labels=None, attention_mask=None):
+        model_output = self.backbone(input_ids)
+        hidden = model_output.last_hidden_state
+        logits = self.output(self.dropout(hidden[:, 0, :]))
+
+        loss = None
+        if labels is not None:
+            labels = labels.to(logits.device)
+            loss_fct = CrossEntropyLoss()
+            loss = loss_fct(logits.view(-1, self.num_labels), labels.view(-1))
+
+        return SequenceClassifierOutput(loss=loss, logits=logits)

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d37bec16044b9e2af2580094e0e2952d869d7db188b9b6a6b7c80eb2ffd3b663
+oid sha256:680e4142e5ffd42a300aa54b976193bfc499640ad8161ce83044dc273d128d80
 size 498674549