Upload model

config.json

@@ -0,0 +1,56 @@
+{
+  "_name_or_path": "./",
+  "_num_labels": 3,
+  "activation_dropout": 0.0,
+  "activation_function": "gelu",
+  "architectures": [
+    "CustomLEDForQAModel"
+  ],
+  "attention_dropout": 0.0,
+  "attention_window": [
+    1024,
+    1024,
+    1024,
+    1024,
+    1024,
+    1024,
+    1024,
+    1024,
+    1024,
+    1024,
+    1024,
+    1024
+  ],
+  "auto_map": {
+    "AutoModel": "modeling_CustomLEDForQA.CustomLEDForQAModel"
+  },
+  "bos_token_id": 0,
+  "classif_dropout": 0.0,
+  "classifier_dropout": 0.0,
+  "d_model": 1024,
+  "decoder_attention_heads": 16,
+  "decoder_ffn_dim": 4096,
+  "decoder_layerdrop": 0.0,
+  "decoder_layers": 12,
+  "decoder_start_token_id": 2,
+  "dropout": 0.1,
+  "encoder_attention_heads": 16,
+  "encoder_ffn_dim": 4096,
+  "encoder_layerdrop": 0.0,
+  "encoder_layers": 12,
+  "eos_token_id": 2,
+  "gradient_checkpointing": false,
+  "init_std": 0.02,
+  "is_encoder_decoder": true,
+  "max_decoder_position_embeddings": 1024,
+  "max_encoder_position_embeddings": 16384,
+  "model_type": "led",
+  "num_hidden_layers": 12,
+  "output_past": false,
+  "pad_token_id": 1,
+  "prefix": " ",
+  "torch_dtype": "float32",
+  "transformers_version": "4.35.0",
+  "use_cache": true,
+  "vocab_size": 50265
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f93ab471be2eed93a99b14a9d4168a2fa1e2dc351d36c38f5fe2216d27323fd2
+size 1028803816

modeling_CustomLEDForQA.py

@@ -0,0 +1,40 @@
+# from transformers.models.led.modeling_led import LEDEncoder
+from transformers import LEDConfig, LEDModel, LEDPreTrainedModel
+import torch.nn as nn
+
+# NEED TO REPLACE nn.Module with PreTrainedModel
+class CustomLEDForQAModel(LEDPreTrainedModel):
+    config_class = LEDConfig
+    
+    def __init__(self, config: LEDConfig): 
+
+        super().__init__(config)
+        config.num_labels = 2
+        self.num_labels = config.num_labels
+
+        self.led = LEDModel(config).get_encoder()
+        self.qa_outputs = nn.Linear(config.hidden_size, config.num_labels)
+
+    def forward(self, input_ids=None, attention_mask=None, global_attention_mask=None, start_positions=None, end_positions=None):
+
+        outputs = self.led(input_ids=input_ids, attention_mask=attention_mask, global_attention_mask=global_attention_mask)
+        
+        logits = self.qa_outputs(outputs.last_hidden_state)
+        start_logits, end_logits = logits.split(1, dim=-1)
+        start_logits = start_logits.squeeze(-1).contiguous()
+        end_logits = end_logits.squeeze(-1).contiguous()
+
+        total_loss = None
+        if start_positions is not None and end_positions is not None:
+            
+            loss_fct = nn.CrossEntropyLoss()
+            start_loss = loss_fct(start_logits, start_positions[0])
+            end_loss = loss_fct(end_logits, end_positions[0])
+            total_loss = (start_loss + end_loss) / 2
+
+    
+        return {
+            'loss': total_loss,
+            'start_logits': start_logits,
+            'end_logits': end_logits,
+        }