Adds config and model file for pipeline.

config.json

@@ -0,0 +1,19 @@
+{
+  "model_type": "llama",
+  "model_path": "meta-llama/Llama-3.2-1B",
+  "num_labels": 4,
+  "head_dim": 64,
+  "hidden_size": 2048,
+  "id2label": {
+    "0": "None",
+    "1": "Keyhole",
+    "2": "Lack of Fusion",
+    "3": "Balling"
+  },
+  "label2id": {
+    "None": 0,
+    "Keyhole": 1,
+    "Lack of Fusion": 2,
+    "Balling": 3
+  }
+}

model.py

@@ -0,0 +1,19 @@
+import torch.nn as nn
+from transformers import AutoModel, PreTrainedModel
+
+class LlamaClassificationModel(PreTrainedModel):
+    def __init__(self, config):
+        super().__init__(config)
+        self.base_model = AutoModel.from_pretrained(config.model_path, config=config)
+        self.classifier = nn.Linear(config.hidden_size, config.num_labels)
+        self.config = config
+
+    def forward(self, input_ids, attention_mask, labels=None):
+        outputs = self.base_model(input_ids=input_ids, attention_mask=attention_mask)
+        summed_representation = outputs.last_hidden_state.sum(dim=1)
+        logits = self.classifier(summed_representation)
+        loss = None
+        if labels is not None:
+            loss_fn = nn.BCEWithLogitsLoss()
+            loss = loss_fn(logits, labels.float())
+        return {"loss": loss, "logits": logits}