Upload model

config.json

@@ -1,6 +1,5 @@
 {
-  "_name_or_path": "./",
-  "_num_labels": 3,
+  "_name_or_path": "allenai/led-base-16384",
   "activation_dropout": 0.0,
   "activation_function": "gelu",
   "architectures": [
@@ -8,12 +7,6 @@
   ],
   "attention_dropout": 0.0,
   "attention_window": [
-    1024,
-    1024,
-    1024,
-    1024,
-    1024,
-    1024,
     1024,
     1024,
     1024,
@@ -27,17 +20,17 @@
   "bos_token_id": 0,
   "classif_dropout": 0.0,
   "classifier_dropout": 0.0,
-  "d_model": 1024,
-  "decoder_attention_heads": 16,
-  "decoder_ffn_dim": 4096,
+  "d_model": 768,
+  "decoder_attention_heads": 12,
+  "decoder_ffn_dim": 3072,
   "decoder_layerdrop": 0.0,
-  "decoder_layers": 12,
+  "decoder_layers": 6,
   "decoder_start_token_id": 2,
   "dropout": 0.1,
-  "encoder_attention_heads": 16,
-  "encoder_ffn_dim": 4096,
+  "encoder_attention_heads": 12,
+  "encoder_ffn_dim": 3072,
   "encoder_layerdrop": 0.0,
-  "encoder_layers": 12,
+  "encoder_layers": 6,
   "eos_token_id": 2,
   "gradient_checkpointing": false,
   "init_std": 0.02,
@@ -45,10 +38,8 @@
   "max_decoder_position_embeddings": 1024,
   "max_encoder_position_embeddings": 16384,
   "model_type": "led",
-  "num_hidden_layers": 12,
-  "output_past": false,
+  "num_hidden_layers": 6,
   "pad_token_id": 1,
-  "prefix": " ",
   "torch_dtype": "float32",
   "transformers_version": "4.35.0",
   "use_cache": true,

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f93ab471be2eed93a99b14a9d4168a2fa1e2dc351d36c38f5fe2216d27323fd2
-size 1028803816
+oid sha256:fd556d7d74d025da749bb5e36c3c3163516ab32d3be851ae5c52ee2f6daf5f3e
+size 417405656

modeling_CustomLEDForQA.py

@@ -6,13 +6,17 @@ import torch.nn as nn
 class CustomLEDForQAModel(LEDPreTrainedModel):
     config_class = LEDConfig
     
-    def __init__(self, config: LEDConfig): 
+    def __init__(self, config: LEDConfig, checkpoint): 
 
         super().__init__(config)
         config.num_labels = 2
         self.num_labels = config.num_labels
 
-        self.led = LEDModel(config).get_encoder()
+        if (checkpoint):
+            self.led = LEDModel.from_pretrained(checkpoint, config=config).get_encoder()
+        else:
+            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):