Create t5_urdu_translation

t5_urdu_translation

@@ -0,0 +1,53 @@
+import torch
+from datasets import load_dataset
+from transformers import T5Tokenizer, T5ForConditionalGeneration, Trainer, TrainingArguments
+
+# Load dataset (replace 'your_dataset' with your actual dataset path or Hugging Face dataset name)
+dataset = load_dataset('csv', data_files={'train': 'train.csv', 'validation': 'validation.csv'})
+
+# Preprocess dataset
+def preprocess_function(examples):
+    inputs = ["translate English to Urdu: " + ex for ex in examples["English"]]
+    targets = examples["Urdu"]
+    model_inputs = tokenizer(inputs, max_length=512, truncation=True)
+    labels = tokenizer(targets, max_length=512, truncation=True).input_ids
+    model_inputs["labels"] = labels
+    return model_inputs
+
+# Load T5 tokenizer and model
+model_name = "t5-small"
+tokenizer = T5Tokenizer.from_pretrained(model_name)
+model = T5ForConditionalGeneration.from_pretrained(model_name)
+
+# Tokenize datasets
+tokenized_datasets = dataset.map(preprocess_function, batched=True)
+
+# Define training arguments
+training_args = TrainingArguments(
+    output_dir="./t5_urdu_translation",
+    evaluation_strategy="epoch",
+    learning_rate=5e-5,
+    per_device_train_batch_size=16,
+    per_device_eval_batch_size=16,
+    num_train_epochs=3,
+    weight_decay=0.01,
+    save_total_limit=2,
+    predict_with_generate=True,
+    logging_dir="./logs",
+)
+
+# Define Trainer
+trainer = Trainer(
+    model=model,
+    args=training_args,
+    train_dataset=tokenized_datasets["train"],
+    eval_dataset=tokenized_datasets["validation"],
+    tokenizer=tokenizer,
+)
+
+# Train model
+trainer.train()
+
+# Save model
+trainer.save_model("./t5_urdu_translation")
+tokenizer.save_pretrained("./t5_urdu_translation")