Upload 3 files

config (3).py

@@ -0,0 +1,61 @@
+import torch
+import os
+
+# --- Paths ---
+DATA_PATH = '/kaggle/input/co-deberta/synthetic_transactions_samples_5000.csv'
+TOKENIZER_PATH = './tokenizer/'
+LABEL_ENCODERS_PATH = './label_encoders.pkl'
+MODEL_SAVE_DIR = './saved_models/'
+PREDICTIONS_SAVE_DIR = './predictions/'
+
+# --- Data Columns ---
+TEXT_COLUMN = "Sanction_Context"
+LABEL_COLUMNS = [
+    "Red_Flag_Reason",
+    "Maker_Action",
+    "Escalation_Level",
+    "Risk_Category",
+    "Risk_Drivers",
+    "Investigation_Outcome"
+]
+METADATA_COLUMNS = []
+
+# --- Model Hyperparameters ---
+MAX_LEN = 128
+BATCH_SIZE = 16
+LEARNING_RATE = 2e-5
+NUM_EPOCHS = 3
+DROPOUT_RATE = 0.3
+
+# --- Device Configuration ---
+DEVICE = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+
+# --- Model Names ---
+BERT_MODEL_NAME = 'bert-base-uncased'
+ROBERTA_MODEL_NAME = 'roberta-base'
+DEBERTA_MODEL_NAME = 'microsoft/deberta-base'
+
+# --- TF-IDF ---
+TFIDF_MAX_FEATURES = 5000
+
+# --- Optional Strategy Definitions ---
+FIELD_STRATEGIES = {
+    "Maker_Action": {
+        "loss": "focal_loss",
+        "enhancements": ["action_templates", "context_prompt_tuning"]
+    },
+    "Risk_Category": {
+        "enhancements": ["numerical_metadata", "transaction_patterns"]
+    },
+    "Escalation_Level": {
+        "enhancements": ["class_balancing", "policy_keyword_patterns"]
+    },
+    "Investigation_Outcome": {
+        "type": "classification_or_generation"
+    }
+}
+
+# --- Ensure Save Directories Exist ---
+os.makedirs(MODEL_SAVE_DIR, exist_ok=True)
+os.makedirs(PREDICTIONS_SAVE_DIR, exist_ok=True)
+os.makedirs(TOKENIZER_PATH, exist_ok=True)

dataset_utils (3).py

@@ -0,0 +1,93 @@
+import pandas as pd
+import torch
+from torch.utils.data import Dataset
+from sklearn.preprocessing import LabelEncoder
+from transformers import BertTokenizer, RobertaTokenizer, DebertaTokenizer
+import pickle
+import os
+
+from config import TEXT_COLUMN, LABEL_COLUMNS, MAX_LEN, TOKENIZER_PATH, LABEL_ENCODERS_PATH, METADATA_COLUMNS
+
+class ComplianceDataset(Dataset):
+    def __init__(self, texts, labels, tokenizer, max_len):
+        self.texts = texts
+        self.labels = labels
+        self.tokenizer = tokenizer
+        self.max_len = max_len
+
+    def __len__(self):
+        return len(self.texts)
+
+    def __getitem__(self, idx):
+        text = str(self.texts[idx])
+        inputs = self.tokenizer(
+            text,
+            padding='max_length',
+            truncation=True,
+            max_length=self.max_len,
+            return_tensors="pt"
+        )
+        inputs = {key: val.squeeze(0) for key, val in inputs.items()}
+        labels = torch.tensor(self.labels[idx], dtype=torch.long)
+        return inputs, labels
+
+class ComplianceDatasetWithMetadata(Dataset):
+    def __init__(self, texts, metadata, labels, tokenizer, max_len):
+        self.texts = texts
+        self.metadata = metadata
+        self.labels = labels
+        self.tokenizer = tokenizer
+        self.max_len = max_len
+
+    def __len__(self):
+        return len(self.texts)
+
+    def __getitem__(self, idx):
+        text = str(self.texts[idx])
+        inputs = self.tokenizer(
+            text,
+            padding='max_length',
+            truncation=True,
+            max_length=self.max_len,
+            return_tensors="pt"
+        )
+        inputs = {key: val.squeeze(0) for key, val in inputs.items()}
+        metadata = torch.tensor(self.metadata[idx], dtype=torch.float)
+        labels = torch.tensor(self.labels[idx], dtype=torch.long)
+        return inputs, metadata, labels
+
+def load_and_preprocess_data(data_path):
+    data = pd.read_csv(data_path)
+    data.fillna("Unknown", inplace=True)
+
+    for col in METADATA_COLUMNS:
+        if col in data.columns:
+            data[col] = pd.to_numeric(data[col], errors='coerce').fillna(0)
+
+    label_encoders = {col: LabelEncoder() for col in LABEL_COLUMNS}
+    for col in LABEL_COLUMNS:
+        data[col] = label_encoders[col].fit_transform(data[col])
+    return data, label_encoders
+
+def get_tokenizer(model_name):
+    model_name = model_name.lower()
+    if "bert" in model_name and "deberta" not in model_name:
+        return BertTokenizer.from_pretrained(model_name)
+    elif "roberta" in model_name:
+        return RobertaTokenizer.from_pretrained(model_name)
+    elif "deberta" in model_name:
+        return DebertaTokenizer.from_pretrained(model_name)
+    else:
+        raise ValueError(f"Unsupported tokenizer for model: {model_name}")
+
+def save_label_encoders(label_encoders):
+    with open(LABEL_ENCODERS_PATH, "wb") as f:
+        pickle.dump(label_encoders, f)
+    print(f"✅ Label encoders saved to {LABEL_ENCODERS_PATH}")
+
+def load_label_encoders():
+    with open(LABEL_ENCODERS_PATH, "rb") as f:
+        return pickle.load(f)
+
+def get_num_labels(label_encoders):
+    return [len(label_encoders[col].classes_) for col in LABEL_COLUMNS]

label_encoders (1).pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:be834abbaaa80f915d0a0015f541a17ae6fda5c75d9485cb23c6a7b7bb7b7c97
+size 2047