# predict.py — REFERENCE SOLUTION (PyTorch) import torch import torch.nn.functional as F import numpy as np def load_trained_model(path, model_class, **kwargs): """Load saved model weights.""" model = model_class(**kwargs) model.load_state_dict(torch.load(path, weights_only=True)) model.eval() return model def predict_single(model, x): """Predict class probabilities for a single input vector.""" model.eval() x_tensor = torch.tensor(x, dtype=torch.float32) x_tensor = x_tensor.unsqueeze(0) with torch.no_grad(): predictions = model(x_tensor) return predictions.numpy()[0] def predict_batch(model, x_batch): """Predict on a batch of inputs.""" model.eval() x_tensor = torch.tensor(x_batch, dtype=torch.float32) with torch.no_grad(): predictions = model(x_tensor) probs = predictions.numpy() classes = torch.argmax(predictions, dim=1).numpy() return {"probabilities": probs, "classes": classes} def get_top_k(model, x, k=2): """Get top-k predictions for an input.""" model.eval() x_tensor = torch.tensor(x, dtype=torch.float32) if x_tensor.dim() == 1: x_tensor = x_tensor.unsqueeze(0) with torch.no_grad(): predictions = model(x_tensor) top_k = torch.topk(predictions, k=k) return {"values": top_k.values.numpy(), "indices": top_k.indices.numpy()}