submission_task1.py

"""MediaEval Medico 2026 Task 1 submission entry point.

This is a repository-ready skeleton for the HuggingFace model repo:
    minhquang47/medico2026-task1-cata-qwen3b

The official `medvqa` validator imports/runs this script from the repo root.
Keep all paths relative to this file.
"""

from __future__ import annotations

import json
import os
import sys
from pathlib import Path
from typing import Any, Dict, Iterable, List, Optional

import numpy as np
import torch
from PIL import Image

REPO_DIR = Path(__file__).resolve().parent
HF_REPO_ID = "minhquang47/medico2026-task1-cata-qwen3b"

if str(REPO_DIR) not in sys.path:
    sys.path.insert(0, str(REPO_DIR))


def _ensure_full_repo_available() -> Path:
    """Return a repo directory that contains the full submission assets.

    The official medvqa validator first runs a lightweight snapshot that may
    contain only `submission_task1.py` and `requirements.txt`. In that case,
    top-level imports from `src` would fail even though the full Hugging Face
    repository contains `src/` and `checkpoints/`. Download the full snapshot
    lazily and use it as the runtime root.
    """

    if (REPO_DIR / "src").exists() and (REPO_DIR / "checkpoints").exists():
        return REPO_DIR

    try:
        from huggingface_hub import snapshot_download

        full_repo = Path(
            snapshot_download(
                repo_id=HF_REPO_ID,
                repo_type="model",
                local_files_only=False,
            )
        )
        if str(full_repo) not in sys.path:
            sys.path.insert(0, str(full_repo))
        return full_repo
    except Exception as exc:  # pragma: no cover - validator/network dependent
        print("Warning: could not download full repo snapshot:", repr(exc))
        return REPO_DIR


RUNTIME_DIR = _ensure_full_repo_available()
if str(RUNTIME_DIR) not in sys.path:
    sys.path.insert(0, str(RUNTIME_DIR))


try:
    from src.models.structural_vqa_generative import build_structural_generative_vqa
except Exception as exc:  # pragma: no cover - validator environment dependent
    print("Warning: model import failed lazily:", repr(exc))
    build_structural_generative_vqa = None


CHECKPOINT_CANDIDATES = [
    RUNTIME_DIR / "checkpoints" / "final.pt",
    RUNTIME_DIR / "checkpoints" / "epoch_3.pt",
    RUNTIME_DIR / "checkpoints" / "epoch_4.pt",
    RUNTIME_DIR / "checkpoints" / "last.pt",
    RUNTIME_DIR / "model.pt",
    REPO_DIR / "checkpoints" / "final.pt",
    REPO_DIR / "model.pt",
]

DEFAULT_IMAGE_SIZE = (224, 224)
DEFAULT_GRID_SIZE = (14, 14)
DEFAULT_TOPO_FEATURE_DIM = 12
DEFAULT_GLOBAL_FEATURE_DIM = 8

_MODEL = None
_DEVICE = None


# --- Deep TopoAdapter runtime injection (mirrors training scripts) ---
class TopoAdapter(torch.nn.Module):
    def __init__(self, hidden_dim: int, topo_dim: int, bottleneck_dim: int = 32):
        super().__init__()
        self.topo_projector = torch.nn.Sequential(
            torch.nn.Linear(topo_dim, bottleneck_dim),
            torch.nn.SiLU(),
            torch.nn.Linear(bottleneck_dim, bottleneck_dim),
        )
        self.down_proj = torch.nn.Linear(hidden_dim, bottleneck_dim, bias=False)
        self.gating = torch.nn.Linear(bottleneck_dim, bottleneck_dim)
        self.up_proj = torch.nn.Linear(bottleneck_dim, hidden_dim, bias=False)
        torch.nn.init.zeros_(self.up_proj.weight)  # exact no-op at init

    def forward(self, hidden_states: torch.Tensor, topo_condition: torch.Tensor) -> torch.Tensor:
        if topo_condition.ndim == 2:
            topo_condition = topo_condition.unsqueeze(1).expand(-1, hidden_states.shape[1], -1)
        topo_condition = topo_condition.to(
            device=hidden_states.device,
            dtype=self.topo_projector[0].weight.dtype,
        )
        topo_condition = torch.nan_to_num(topo_condition, nan=0.0, posinf=0.0, neginf=0.0)
        hs = hidden_states.to(dtype=self.down_proj.weight.dtype)
        t_emb = self.topo_projector(topo_condition)
        gate = torch.sigmoid(self.gating(t_emb))
        delta = self.up_proj(self.down_proj(hs) * gate).to(dtype=hidden_states.dtype)
        return hidden_states + delta


class DecoderLayerWithTopoAdapter(torch.nn.Module):
    def __init__(self, base_layer: torch.nn.Module, adapter: TopoAdapter):
        super().__init__()
        self.base_layer = base_layer
        self.topo_adapter = adapter
        self.topo_condition: Optional[torch.Tensor] = None
        # Newer Transformers/Qwen generation code reads decoder-layer metadata
        # directly from each layer object before calling forward(). Preserve
        # those attributes when wrapping the layer with TopoAdapter.
        if hasattr(base_layer, "attention_type"):
            self.attention_type = base_layer.attention_type

    def set_topo_condition(self, condition: Optional[torch.Tensor]):
        self.topo_condition = condition

    def forward(self, *args, **kwargs):
        out = self.base_layer(*args, **kwargs)
        if self.topo_condition is None:
            return out
        if isinstance(out, tuple):
            hs = self.topo_adapter(out[0], self.topo_condition)
            return (hs,) + out[1:]
        return self.topo_adapter(out, self.topo_condition)


def _find_decoder_layers(llm) -> torch.nn.ModuleList:
    candidates = [
        "model.layers",
        "base_model.model.model.layers",
        "base_model.model.layers",
        "model.model.layers",
    ]
    for path in candidates:
        obj = llm
        ok = True
        for part in path.split("."):
            if not hasattr(obj, part):
                ok = False
                break
            obj = getattr(obj, part)
        if ok and isinstance(obj, torch.nn.ModuleList):
            return obj
    raise RuntimeError("Cannot locate Qwen decoder layers for TopoAdapter injection")


def _install_topo_adapters(
    model,
    topo_dim: int = 47,
    bottleneck_dim: int = 32,
    last_n_layers: int = 8,
    every_n_layers: int = 0,
) -> List[DecoderLayerWithTopoAdapter]:
    layers = _find_decoder_layers(model.llm)
    hidden_dim = int(model.llm.get_input_embeddings().embedding_dim)
    n_layers = len(layers)
    selected = set(range(max(0, n_layers - last_n_layers), n_layers)) if last_n_layers > 0 else set()
    if every_n_layers > 0:
        selected.update(range(0, n_layers, every_n_layers))
    wrappers: List[DecoderLayerWithTopoAdapter] = []
    for idx in sorted(selected):
        if isinstance(layers[idx], DecoderLayerWithTopoAdapter):
            wrappers.append(layers[idx])
            continue
        wrapper = DecoderLayerWithTopoAdapter(
            layers[idx],
            TopoAdapter(hidden_dim, topo_dim, bottleneck_dim),
        )
        layers[idx] = wrapper
        wrappers.append(wrapper)
    print(
        f"Installed {len(wrappers)} TopoAdapters / {n_layers} decoder layers "
        f"| hidden={hidden_dim} topo_dim={topo_dim}"
    )
    return wrappers


def _topo_condition(
    prior_mask: torch.Tensor,
    topo_features: torch.Tensor,
    global_features: torch.Tensor,
    mode: str = "all",
) -> torch.Tensor:
    tf = topo_features.float()
    flat = tf.flatten(1, 2)
    parts = [flat.mean(1), flat.std(1), flat.amax(1)]
    if mode == "all":
        pm = prior_mask.float().flatten(1)
        parts.extend([
            pm.mean(1, keepdim=True),
            pm.std(1, keepdim=True),
            pm.amax(1, keepdim=True),
        ])
        parts.append(global_features.float())
    cond = torch.cat(parts, dim=-1)
    return torch.nan_to_num(cond, nan=0.0, posinf=0.0, neginf=0.0)


def _set_topo_condition(wrappers: List[DecoderLayerWithTopoAdapter], condition: Optional[torch.Tensor]):
    for wrapper in wrappers:
        wrapper.set_topo_condition(condition)


def _patch_model_forward(model, wrappers: List[DecoderLayerWithTopoAdapter], mode: str = "all"):
    original_forward = model.forward
    original_generate = model.generate

    def forward_with_topo(
        self,
        image,
        prior_mask,
        topo_features,
        global_features,
        question_text,
        answer_text=None,
        return_diagnostics=True,
    ):
        condition = _topo_condition(prior_mask, topo_features, global_features, mode).to(image.device)
        _set_topo_condition(wrappers, condition)
        try:
            return original_forward(
                image,
                prior_mask,
                topo_features,
                global_features,
                question_text,
                answer_text,
                return_diagnostics,
            )
        finally:
            _set_topo_condition(wrappers, None)

    @torch.no_grad()
    def generate_with_topo(
        self,
        image,
        prior_mask,
        topo_features,
        global_features,
        question_text,
        max_new_tokens=64,
    ):
        condition = _topo_condition(prior_mask, topo_features, global_features, mode).to(image.device)
        _set_topo_condition(wrappers, condition)
        try:
            return original_generate(
                image,
                prior_mask,
                topo_features,
                global_features,
                question_text,
                max_new_tokens,
            )
        finally:
            _set_topo_condition(wrappers, None)

    import types

    model.forward = types.MethodType(forward_with_topo, model)
    model.generate = types.MethodType(generate_with_topo, model)


def _find_checkpoint() -> Path:
    for path in CHECKPOINT_CANDIDATES:
        if path.exists():
            return path
    raise FileNotFoundError(
        "No checkpoint found. Expected one of: "
        + ", ".join(str(path.relative_to(REPO_DIR)) for path in CHECKPOINT_CANDIDATES)
    )


def _select_device() -> torch.device:
    return torch.device("cuda" if torch.cuda.is_available() else "cpu")


def _safe_load_checkpoint(checkpoint_path: Path) -> Dict[str, Any]:
    """Load a checkpoint while tolerating optimizer/RNG pickle incompatibilities."""

    try:
        return torch.load(checkpoint_path, map_location="cpu", weights_only=False)
    except TypeError:
        loaded = torch.load(checkpoint_path, map_location="cpu", weights_only=True)
        if isinstance(loaded, dict) and "model_state_dict" in loaded:
            return loaded
        return {
            "model_state_dict": loaded,
            "args": {
                "llm_name_or_path": "Qwen/Qwen2.5-3B-Instruct",
                "vision_backend": "timm",
                "freeze_llm": True,
                "use_lora": True,
                "lora_r": 16,
                "lora_alpha": 32,
                "lora_dropout": 0.05,
                "lora_target_modules": "q_proj,v_proj",
                "use_ot": True,
                "use_ot_fusion": True,
                "ot_fusion_mode": "prefix",
                "ot_fusion_dropout": 0.10,
                "use_prior_as_ot_target": True,
                "use_topological_loss": True,
                "use_prior_align_loss": True,
                "use_global_topo_loss": True,
                "use_patch_topo_loss": False,
            },
        }


def _load_model_from_checkpoint(checkpoint_path: Path, device: torch.device):
    if build_structural_generative_vqa is None:
        raise ImportError(
            "Could not import src.models.structural_vqa_generative. "
            "Copy the project src/ directory into the HF repo root."
        )

    checkpoint = _safe_load_checkpoint(checkpoint_path)
    train_args = checkpoint.get("args", {})
    config = {**train_args}
    config.update(
        {
            "llm_name_or_path": train_args.get("llm_name_or_path", "Qwen/Qwen2.5-3B-Instruct"),
            "vision_pretrained": False,
            "vision_backend": train_args.get("vision_backend", "timm"),
            "freeze_vision_backbone": True,
            "freeze_llm": train_args.get("freeze_llm", True),
            "use_lora": train_args.get("use_lora", True),
            "lora_r": train_args.get("lora_r", 16),
            "lora_alpha": train_args.get("lora_alpha", 32),
            "lora_dropout": train_args.get("lora_dropout", 0.05),
            "lora_target_modules": train_args.get(
                "lora_target_modules",
                "q_proj,k_proj,v_proj,o_proj,gate_proj,up_proj,down_proj",
            ),
        }
    )
    allowed = {
        "llm_name_or_path",
        "vision_pretrained",
        "vision_backend",
        "freeze_vision_backbone",
        "freeze_llm",
        "max_question_length",
        "max_answer_length",
        "use_lora",
        "lora_r",
        "lora_alpha",
        "lora_dropout",
        "lora_target_modules",
        "ot_loss_weight",
        "use_ot",
        "use_ot_fusion",
        "ot_fusion_mode",
        "ot_fusion_dropout",
        "use_prior_as_ot_target",
        "prior_ot_global_mass",
        "use_topological_loss",
        "use_prior_align_loss",
        "use_global_topo_loss",
        "use_patch_topo_loss",
    }
    build_kwargs = {key: value for key, value in config.items() if key in allowed}
    model = build_structural_generative_vqa(**build_kwargs).to(device)

    # Install TopoAdapters and patch the model forward/generate methods
    # Using mode="all" matching "train_qwen3b_curriculum_topo_adapter" full configuration
    wrappers = _install_topo_adapters(model, topo_dim=47, bottleneck_dim=32, last_n_layers=8, every_n_layers=0)
    for wrapper in wrappers:
        wrapper.to(device)
    _patch_model_forward(model, wrappers, mode="all")

    state = checkpoint.get("model_state_dict", checkpoint)
    missing, unexpected = model.load_state_dict(state, strict=False)
    critical_missing = [key for key in missing if "topo_adapter" in key]
    unexpected_topo = [key for key in unexpected if "topo_adapter" in key]
    if critical_missing and unexpected_topo:
        remapped_state = dict(state)
        prefix_pairs = [
            ("llm.base_model.model.model.layers.", "llm.model.layers."),
            ("llm.model.layers.", "llm.base_model.model.model.layers."),
        ]
        for old_prefix, new_prefix in prefix_pairs:
            for key, value in state.items():
                if key.startswith(old_prefix):
                    remapped_state[new_prefix + key[len(old_prefix):]] = value
        missing, unexpected = model.load_state_dict(remapped_state, strict=False)
        critical_missing = [key for key in missing if "topo_adapter" in key]

    if critical_missing:
        raise RuntimeError(f"Missing TopoAdapter checkpoint keys: {critical_missing[:20]}")
    if unexpected:
        print(f"Warning: unexpected checkpoint keys: {unexpected[:20]}")
    print(f"Loaded checkpoint successfully. Status: OK")
    model.eval()
    return model


def load_model():
    """Load and cache the VQA model."""

    global _MODEL, _DEVICE
    if _MODEL is None:
        _DEVICE = _select_device()
        _MODEL = _load_model_from_checkpoint(_find_checkpoint(), _DEVICE)
    return _MODEL


def _pil_to_tensor(image: Image.Image) -> torch.Tensor:
    import cv2

    image_rgb = np.asarray(image.convert("RGB")).astype(np.uint8)
    image_rgb = cv2.resize(image_rgb, DEFAULT_IMAGE_SIZE[::-1], interpolation=cv2.INTER_AREA)
    arr = image_rgb.astype(np.float32) / 255.0
    arr = np.transpose(arr, (2, 0, 1))
    return torch.from_numpy(arr).float().unsqueeze(0)


_STRUCTURAL_EXTRACTORS = None


def _get_structural_extractors():
    """Create the same structural extractors used by offline precompute."""

    global _STRUCTURAL_EXTRACTORS
    if _STRUCTURAL_EXTRACTORS is not None:
        return _STRUCTURAL_EXTRACTORS

    from src.topology.lesion_prior import LesionPriorExtractor
    from src.topology.tda_morphology import TopologicalExtractor as MorphologyTopologicalExtractor

    prior_extractor = LesionPriorExtractor(
        image_size=DEFAULT_IMAGE_SIZE,
        grid_size=DEFAULT_GRID_SIZE,
        use_morphology=False,
    )
    morpho_extractor = MorphologyTopologicalExtractor(
        grid_size=DEFAULT_GRID_SIZE,
        image_size=DEFAULT_IMAGE_SIZE,
    )
    _STRUCTURAL_EXTRACTORS = (prior_extractor, morpho_extractor)
    return _STRUCTURAL_EXTRACTORS


def _extract_online_structural_tensors(image: Image.Image) -> Dict[str, torch.Tensor]:
    """Compute structural features online with the original precompute logic.

    This mirrors `scripts/precompute_structural_features.py`:
      prior_output = LesionPriorExtractor.extract_prior(image_bgr)
      morpho_output = MorphologyTopologicalExtractor.extract_features(image_bgr)

    Shapes:
      - prior_mask: [1, 14, 14]
      - topo_features: [1, 14, 14, 12]
      - global_features: [1, 8]
    """

    import cv2

    prior_extractor, morpho_extractor = _get_structural_extractors()

    image_rgb = np.asarray(image.convert("RGB")).astype(np.uint8)
    image_bgr = cv2.cvtColor(image_rgb, cv2.COLOR_RGB2BGR)

    prior_output = prior_extractor.extract_prior(image_bgr)
    morpho_output = morpho_extractor.extract_features(image_bgr)

    prior_mask = prior_output["prior_mask"].astype(np.float32)
    topo_features = morpho_output["topo_features"].astype(np.float32)
    global_features = morpho_output["global_features"].astype(np.float32)

    if prior_mask.shape != DEFAULT_GRID_SIZE:
        raise ValueError(f"prior_mask shape {prior_mask.shape} != {DEFAULT_GRID_SIZE}")
    expected_topo_shape = (*DEFAULT_GRID_SIZE, DEFAULT_TOPO_FEATURE_DIM)
    if topo_features.shape != expected_topo_shape:
        raise ValueError(f"topo_features shape {topo_features.shape} != {expected_topo_shape}")
    if global_features.shape != (DEFAULT_GLOBAL_FEATURE_DIM,):
        raise ValueError(
            f"global_features shape {global_features.shape} != {(DEFAULT_GLOBAL_FEATURE_DIM,)}"
        )

    return {
        "prior_mask": torch.from_numpy(prior_mask).float().unsqueeze(0),
        "topo_features": torch.from_numpy(topo_features).float().unsqueeze(0),
        "global_features": torch.from_numpy(global_features).float().unsqueeze(0),
    }


def predict_one(image: Image.Image, question: str, max_new_tokens: int = 48) -> str:
    model = load_model()
    device = _DEVICE or _select_device()
    image_tensor = _pil_to_tensor(image).to(device)
    structural = {key: value.to(device) for key, value in _extract_online_structural_tensors(image).items()}

    with torch.inference_mode():
        prediction = model.generate(
            image=image_tensor,
            prior_mask=structural["prior_mask"],
            topo_features=structural["topo_features"],
            global_features=structural["global_features"],
            question_text=[question],
            max_new_tokens=max_new_tokens,
        )[0]
    return prediction.strip()


def predict(batch: Iterable[Dict[str, Any]]) -> List[Dict[str, Any]]:
    """Generic batch interface for validators that pass dict records.

    Each record is expected to contain an `image` PIL object/path and `question`.
    The return keeps IDs if present and always includes `answer`.
    """

    outputs = []
    for record in batch:
        image = record.get("image")
        if isinstance(image, (str, os.PathLike)):
            image = Image.open(image)
        if not isinstance(image, Image.Image):
            raise TypeError("record['image'] must be a PIL.Image.Image or image path")
        question = str(record.get("question", ""))
        answer = predict_one(image, question)
        out = {key: record[key] for key in ("id", "img_id", "question") if key in record}
        out["answer"] = answer
        outputs.append(out)
    return outputs


class MedicoTask1Model:
    """Small callable wrapper for template-style validators."""

    def __call__(self, image: Image.Image, question: str) -> str:
        return predict_one(image, question)

    def predict(self, image: Image.Image, question: str) -> str:
        return predict_one(image, question)


def load():
    """Factory hook used by some submission templates."""

    load_model()
    return MedicoTask1Model()


SUBMISSION_INFO = {
    "Participant_Names": "Minh Quang Nguyen",
    "Affiliations": "Independent",
    "Contact_emails": ["nmquang04072005@gmail.com"],
    "Team_Name": "Sweet&Sour",
    "Country": "Vietnam",
    "Notes_to_organizers": (
        "Task 1 submission using Qwen2.5-3B-Instruct + QLoRA r16 + Curriculum-Gated full structural Deep TopoAdapter. "
        "Trained for 2 epochs on 30k with warmup/gate regularization then 1 epoch full-data continuation. "
        "Clean single-sample inference layout."
    ),
}


def _run_medvqa_validation(output_name: str = "predictions_1.json") -> Path:
    """Generate `predictions_1.json` in the official Task 1 format."""

    import platform
    import subprocess
    import time

    from datasets import Image as HfImage
    from datasets import load_dataset
    from evaluate import load as load_metric
    from tqdm import tqdm

    output_path = REPO_DIR / output_name
    gpu_name = torch.cuda.get_device_name(0) if torch.cuda.is_available() else "cpu"
    device_name = "cuda" if torch.cuda.is_available() else "cpu"

    def get_mem() -> float:
        return torch.cuda.memory_allocated(device_name) / (1024**2) if torch.cuda.is_available() else 0.0

    initial_mem = get_mem()
    start_time = time.time()

    ds = load_dataset("SimulaMet/Kvasir-VQA-x1")["test"]
    val_dataset = ds.shuffle(seed=42).select(range(1500)).cast_column("image", HfImage())

    # Force model load before measuring generation-time memory, matching the template.
    load_model()
    post_model_mem = get_mem()
    model_mem_used = round(post_model_mem - initial_mem, 2)

    predictions = []
    for idx, sample in enumerate(tqdm(val_dataset, desc="Validating", unit="samples")):
        question = str(sample.get("question", ""))
        image = sample["image"]
        answer = predict_one(image, question).strip()
        predictions.append(
            {
                "index": int(idx),
                "img_id": str(sample.get("img_id", "")),
                "question": question,
                "answer": answer,
            }
        )

        # Print periodic generation samples
        if (idx + 1) % 50 == 0 or idx == 0:
            print(f"\n[{idx + 1}/1500] Q: {question}")
            print(f"Pred: '{answer}'")
            print(f"Ref:  '{sample.get('answer', '')}'")
            print("-" * 30)

    assert len(predictions) == len(val_dataset), "Mismatch between predictions and dataset length"

    references = [[answer] for answer in val_dataset["answer"]]
    preds = [pred["answer"] for pred in predictions]

    bleu = load_metric("bleu")
    rouge = load_metric("rouge")
    meteor = load_metric("meteor")

    bleu_result = bleu.compute(predictions=preds, references=references)
    rouge_result = rouge.compute(predictions=preds, references=references)
    meteor_result = meteor.compute(predictions=preds, references=references)
    public_scores = {
        "bleu": round(float(bleu_result["bleu"]), 4),
        "rouge1": round(float(rouge_result["rouge1"]), 4),
        "rouge2": round(float(rouge_result["rouge2"]), 4),
        "rougeL": round(float(rouge_result["rougeL"]), 4),
        "meteor": round(float(meteor_result["meteor"]), 4),
    }
    print("✨Public scores:", public_scores)

    total_time = round(time.time() - start_time, 4)
    final_mem = round(get_mem() - post_model_mem, 2)

    output_data = {
        "submission_info": SUBMISSION_INFO,
        "public_scores": public_scores,
        "predictions": predictions,
        "total_time": total_time,
        "time_per_item": total_time / len(val_dataset),
        "memory_used_mb": final_mem,
        "model_memory_mb": model_mem_used,
        "gpu_name": gpu_name,
        "debug": {
            "packages": json.loads(
                subprocess.check_output([sys.executable, "-m", "pip", "list", "--format=json"])
            ),
            "system": {
                "python": platform.python_version(),
                "os": platform.system(),
                "platform": platform.platform(),
                "arch": platform.machine(),
            },
        },
    }

    with output_path.open("w", encoding="utf-8") as f:
        json.dump(output_data, f, ensure_ascii=False, indent=4)

    print(
        f"Time: {total_time}s | Mem: {final_mem}MB | "
        f"Model Load Mem: {model_mem_used}MB | GPU: {gpu_name}"
    )
    print("✅ Results saved to predictions_1.json")
    return output_path


def main():
    # medvqa runs this file directly and expects predictions_1.json afterwards.
    print("Medico 2026 Task 1 submission script is importable.")
    print("Repo:", REPO_DIR)
    print("Runtime repo:", RUNTIME_DIR)
    print("Checkpoint exists:", any(path.exists() for path in CHECKPOINT_CANDIDATES))
    _run_medvqa_validation()


if __name__ == "__main__":
    main()