Create train.py

train.py

@@ -0,0 +1,54 @@
+from datasets import load_dataset
+from transformers import AutoTokenizer, AutoModelForSequenceClassification, Trainer, TrainingArguments
+import evaluate
+
+# 1. Load BiScope dataset
+dataset = load_dataset("HanxiGuo/BiScope_Data")
+
+# 2. Tokenizer
+MODEL = "microsoft/deberta-v3-small"
+tokenizer = AutoTokenizer.from_pretrained(MODEL)
+
+def preprocess(examples):
+    return tokenizer(examples["text"], truncation=True, padding="max_length", max_length=256)
+
+encoded_dataset = dataset.map(preprocess, batched=True)
+
+# 3. Load model
+model = AutoModelForSequenceClassification.from_pretrained(MODEL, num_labels=2)
+
+# 4. Metrics
+accuracy = evaluate.load("accuracy")
+
+def compute_metrics(eval_pred):
+    logits, labels = eval_pred
+    predictions = logits.argmax(axis=-1)
+    return accuracy.compute(predictions=predictions, references=labels)
+
+# 5. Training args
+training_args = TrainingArguments(
+    output_dir="./results",
+    evaluation_strategy="epoch",
+    save_strategy="epoch",
+    learning_rate=2e-5,
+    per_device_train_batch_size=16,
+    per_device_eval_batch_size=16,
+    num_train_epochs=2,
+    weight_decay=0.01,
+    push_to_hub=True,  # ✅ Upload to HF
+    hub_model_id="your-username/biscope-detector"
+)
+
+# 6. Trainer
+trainer = Trainer(
+    model=model,
+    args=training_args,
+    train_dataset=encoded_dataset["train"],
+    eval_dataset=encoded_dataset["validation"],
+    tokenizer=tokenizer,
+    compute_metrics=compute_metrics,
+)
+
+# 7. Train & Push
+trainer.train()
+trainer.push_to_hub()