AnchorQuant / src /train.py
abdurrehmansaeed's picture
Upload 9 files
025391e verified
Raw History Blame Contribute Delete
2.05 kB
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.")