train_sentiment_model.py

import pandas as pd
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import LabelEncoder
from transformers import DistilBertTokenizerFast, DistilBertForSequenceClassification, Trainer, TrainingArguments
import torch
from datasets import Dataset

# 1. Load and clean data
df = pd.read_csv("/Users/milanradovanovich/Downloads/twitter_training.csv/twitter_training.csv", header=None)
df.columns = ['tweet_id', 'topic', 'sentiment', 'text']
df.dropna(subset=['text', 'sentiment'], inplace=True)
df['sentiment'] = df['sentiment'].str.strip().str.lower()

# 2. Encode sentiment labels
label_map = {'positive': 0, 'negative': 1, 'neutral': 2, 'irrelevant': 3}
df['label'] = df['sentiment'].map(label_map)

# 3. Split into train and validation
train_texts, val_texts, train_labels, val_labels = train_test_split(
    df['text'].tolist(),
    df['label'].tolist(),
    test_size=0.1,
    stratify=df['label'],
    random_state=42
)

# 4. Tokenize
tokenizer = DistilBertTokenizerFast.from_pretrained('distilbert-base-uncased')
train_encodings = tokenizer(train_texts, truncation=True, padding=True, max_length=128)
val_encodings = tokenizer(val_texts, truncation=True, padding=True, max_length=128)

# 5. Build Dataset class
class SentimentDataset(torch.utils.data.Dataset):
    def __init__(self, encodings, labels):
        self.encodings = encodings
        self.labels = labels

    def __getitem__(self, idx):
        item = {key: torch.tensor(val[idx]) for key, val in self.encodings.items()}
        item["labels"] = torch.tensor(self.labels[idx])
        return item

    def __len__(self):
        return len(self.labels)

train_dataset = SentimentDataset(train_encodings, train_labels)
val_dataset = SentimentDataset(val_encodings, val_labels)

# 6. Load model
model = DistilBertForSequenceClassification.from_pretrained(
    "distilbert-base-uncased", num_labels=4
)

# 7. Training config
training_args = TrainingArguments(
    output_dir="./model_output",
    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=3,
    weight_decay=0.01,
    logging_dir='./logs',
    logging_steps=50,
)
# 8. Train
trainer = Trainer(
    model=model,
    args=training_args,
    train_dataset=train_dataset,
    eval_dataset=val_dataset,
)

trainer.train()

# 9. Save model and tokenizer
model.save_pretrained("./sentiment_model")
tokenizer.save_pretrained("./sentiment_model")

print(" Training complete and model saved to ./sentiment_model")