import torch import torch.nn as nn from torch.optim import AdamW from torch.cuda.amp import autocast, GradScaler from tqdm import tqdm from dataset import get_data_loaders from model import FinancialSentimentClassifier def train_epoch(model, data_loader, loss_fn, optimizer, device, scaler): model = model.train() losses = [] correct_predictions = 0 for d in tqdm(data_loader, desc="Training"): input_ids = d["input_ids"].to(device) attention_mask = d["attention_mask"].to(device) labels = d["labels"].to(device) optimizer.zero_grad() # Mixed Precision for hardware optimization with autocast(): outputs = model(input_ids=input_ids, attention_mask=attention_mask) _, preds = torch.max(outputs, dim=1) loss = loss_fn(outputs, labels) correct_predictions += torch.sum(preds == labels) losses.append(loss.item()) # Scale gradients and step scaler.scale(loss).backward() scaler.step(optimizer) scaler.update() return correct_predictions.double() / len(data_loader.dataset), sum(losses) / len(losses) if __name__ == "__main__": device = torch.device("cuda" if torch.cuda.is_available() else "cpu") print(f"Executing training on: {device}") train_loader, val_loader = get_data_loaders(batch_size=16) model = FinancialSentimentClassifier(n_classes=3).to(device) optimizer = AdamW(model.parameters(), lr=2e-5, weight_decay=0.01) loss_fn = nn.CrossEntropyLoss().to(device) scaler = GradScaler() EPOCHS = 3 for epoch in range(EPOCHS): print(f"Epoch {epoch + 1}/{EPOCHS}") train_acc, train_loss = train_epoch(model, train_loader, loss_fn, optimizer, device, scaler) print(f"Train Loss: {train_loss:.4f} | Train Acc: {train_acc:.4f}") # Save the professional weights torch.save(model.state_dict(), "../models/finbert_sentiment.pt") print("Model weights secured.")