Update main.py

main.py

@@ -2,3 +2,68 @@ from datasets import load_dataset
 
 dataset = load_dataset("mteb/tweet_sentiment_extraction")
 df = pd.DataFrame(dataset['train'])
+
+
+from transformers import GPT2Tokenizer
+
+# Loading the dataset to train our model
+dataset = load_dataset("mteb/tweet_sentiment_extraction")
+
+tokenizer = GPT2Tokenizer.from_pretrained("gpt2")
+tokenizer.pad_token = tokenizer.eos_token
+def tokenize_function(examples):
+   return tokenizer(examples["text"], padding="max_length", truncation=True)
+
+tokenized_datasets = dataset.map(tokenize_function, batched=True)
+
+
+small_train_dataset = tokenized_datasets["train"].shuffle(seed=42).select(range(1000))
+small_eval_dataset = tokenized_datasets["test"].shuffle(seed=42).select(range(1000))
+
+
+from transformers import GPT2ForSequenceClassification
+
+model = GPT2ForSequenceClassification.from_pretrained("gpt2", num_labels=3)
+
+
+
+import evaluate
+
+metric = evaluate.load("accuracy")
+
+def compute_metrics(eval_pred):
+   logits, labels = eval_pred
+   predictions = np.argmax(logits, axis=-1)
+   return metric.compute(predictions=predictions, references=labels)
+
+
+
+
+from transformers import TrainingArguments, Trainer
+
+training_args = TrainingArguments(
+   output_dir="test_trainer",
+   #evaluation_strategy="epoch",
+   per_device_train_batch_size=1,  # Reduce batch size here
+   per_device_eval_batch_size=1,    # Optionally, reduce for evaluation as well
+   gradient_accumulation_steps=4
+   )
+
+
+trainer = Trainer(
+   model=model,
+   args=training_args,
+   train_dataset=small_train_dataset,
+   eval_dataset=small_eval_dataset,
+   compute_metrics=compute_metrics,
+
+)
+
+trainer.train()
+
+
+
+
+import evaluate
+
+trainer.evaluate()