import os import torch from PIL import Image from huggingface_hub import hf_hub_download from models.biohcl_sssm import BioHCLSSSM from dataset import get_transforms def load_pretrained_model(repo_id="unknownBrain/BioHCL-SSSM-Pediatric-Pneumonia", device="cpu"): """ Downloads and instantiates the best BioHCL-SSSM model checkpoint from Hugging Face. """ print(f"Downloading best model from {repo_id}...") ckpt_path = hf_hub_download(repo_id=repo_id, filename="checkpoints/best_student_cv.pt") model = BioHCLSSSM(num_classes=2, embed_dim=512, chain_mode="dual", num_sssm_blocks=3) state_dict = torch.load(ckpt_path, map_location=device) if "model_state_dict" in state_dict: state_dict = state_dict["model_state_dict"] model.load_state_dict(state_dict) model.to(device) model.eval() print("Model loaded successfully!") return model def predict_chest_xray(model, image_path, device="cpu"): """ Runs diagnostic prediction on a single chest X-ray image. """ _, eval_transform = get_transforms(img_size=224) image = Image.open(image_path).convert("RGB") tensor = eval_transform(image=np.array(image))["image"].unsqueeze(0).to(device) with torch.no_grad(): logits, _ = model(tensor) probs = torch.softmax(logits, dim=1)[0] pred_idx = torch.argmax(probs).item() classes = ["NORMAL", "PNEUMONIA"] return { "prediction": classes[pred_idx], "confidence": float(probs[pred_idx]), "probabilities": {"NORMAL": float(probs[0]), "PNEUMONIA": float(probs[1])} } if __name__ == "__main__": device = "cuda" if torch.cuda.is_available() else "cpu" print(f"Running inference on {device}...") # model = load_pretrained_model(device=device) # result = predict_chest_xray(model, "sample_xray.jpeg", device=device) # print("Result:", result)