gnn_data/gnn_world_model_loader.py

from __future__ import annotations

import json
from dataclasses import dataclass
from functools import lru_cache
from pathlib import Path
from typing import Dict, List, Optional, Tuple

import numpy as np
import torch
import yaml
from torch_geometric.data import Data


TYPE_ENCODING_DIM = 10
SAM2_EMB_DIM = 256
POS_DIM = 3
VIS_DIM = 1
NODE_DIM = SAM2_EMB_DIM + POS_DIM + TYPE_ENCODING_DIM + VIS_DIM
ROBOT_STATE_DIM = 13

DATASET_ROOT = Path(__file__).resolve().parent
TYPE_ENCODING_ROOT = DATASET_ROOT / "config"


@lru_cache(maxsize=4)
def load_type_encoding(encoding_method: str = "random") -> Dict[str, np.ndarray]:
    path = TYPE_ENCODING_ROOT / f"type_encoding_{encoding_method}.yaml"
    with path.open("r", encoding="utf-8") as f:
        raw = yaml.safe_load(f)
    return {k: np.asarray(v, dtype=np.float32) for k, v in raw.items()}


def type_encode(comp_type: str, encoding_method: str = "random") -> np.ndarray:
    table = load_type_encoding(encoding_method)
    vec = table.get(comp_type)
    if vec is None:
        return np.zeros(TYPE_ENCODING_DIM, dtype=np.float32)
    return vec.astype(np.float32)


def list_labeled_frames(episode_dir: str | Path) -> List[int]:
    mask_dir = Path(episode_dir) / "annotations" / "side_masks"
    if not mask_dir.exists():
        return []
    frames = []
    for p in mask_dir.glob("frame_*.npz"):
        try:
            frames.append(int(p.stem.split("_")[1]))
        except (ValueError, IndexError):
            continue
    return sorted(frames)


def resolve_frame_state(graph_json: dict, frame_idx: int) -> Tuple[Dict[str, bool], Dict[str, bool], Dict[str, bool]]:
    constraints: Dict[str, bool] = {}
    visibility: Dict[str, bool] = {}
    held: Dict[str, bool] = {}

    for c in graph_json["components"]:
        cid = c["id"]
        visibility[cid] = True
        held[cid] = False
    for e in graph_json["edges"]:
        constraints[f"{e['src']}->{e['dst']}"] = True

    fs_dict = graph_json.get("frame_states", {})
    for f in sorted(int(k) for k in fs_dict):
        if f > frame_idx:
            break
        fs = fs_dict[str(f)]
        constraints.update(fs.get("constraints", {}))
        visibility.update(fs.get("visibility", {}))
        held.update(fs.get("held", {}))

    return constraints, visibility, held


@dataclass
class FrameData:
    graph: dict
    masks: dict
    embeddings: dict
    depth_info: dict
    robot: Optional[dict]
    constraints: Dict[str, bool]
    visibility: Dict[str, bool]
    held: Dict[str, bool]


def _npz_dict(path: Path) -> Dict[str, np.ndarray]:
    if not path.exists():
        return {}
    data = np.load(path)
    return {k: data[k] for k in data.files}


def load_frame_data(episode_dir: str | Path, frame_idx: int) -> FrameData:
    episode_dir = Path(episode_dir)
    anno = episode_dir / "annotations"
    with (anno / "side_graph.json").open("r", encoding="utf-8") as f:
        graph = json.load(f)

    masks = _npz_dict(anno / "side_masks" / f"frame_{frame_idx:06d}.npz")
    embeddings = _npz_dict(anno / "side_embeddings" / f"frame_{frame_idx:06d}.npz")
    depth_info = _npz_dict(anno / "side_depth_info" / f"frame_{frame_idx:06d}.npz")

    robot = None
    robot_path = anno / "side_robot" / f"frame_{frame_idx:06d}.npz"
    if robot_path.exists():
        r = np.load(robot_path)
        if "visible" in r.files and int(r["visible"][0]) == 1:
            robot = {k: r[k] for k in r.files}

    constraints, visibility, held = resolve_frame_state(graph, frame_idx)
    return FrameData(graph, masks, embeddings, depth_info, robot, constraints, visibility, held)


def _build_product_node_features(nodes: List[dict], fd: FrameData, encoding_method: str) -> torch.Tensor:
    feats = []
    for node in nodes:
        cid = node["id"]
        emb = fd.embeddings.get(cid, np.zeros(SAM2_EMB_DIM, dtype=np.float32))
        depth_valid_key = f"{cid}_depth_valid"
        centroid_key = f"{cid}_centroid"
        if depth_valid_key in fd.depth_info and int(fd.depth_info[depth_valid_key][0]) == 1:
            pos = fd.depth_info[centroid_key].astype(np.float32)
        else:
            pos = np.zeros(POS_DIM, dtype=np.float32)

        visible = 1.0 if fd.visibility.get(cid, True) else 0.0
        if visible == 0.0:
            emb = np.zeros(SAM2_EMB_DIM, dtype=np.float32)
            pos = np.zeros(POS_DIM, dtype=np.float32)

        feats.append(
            np.concatenate(
                [
                    emb.astype(np.float32),
                    pos,
                    type_encode(node["type"], encoding_method),
                    np.array([visible], dtype=np.float32),
                ]
            )
        )

    if not feats:
        return torch.empty((0, NODE_DIM), dtype=torch.float32)
    return torch.tensor(np.stack(feats), dtype=torch.float32)


def _build_product_edges(nodes: List[dict], graph: dict, fd: FrameData):
    constraint_set = {(edge["src"], edge["dst"]) for edge in graph["edges"]}
    pair_forward = {frozenset([src, dst]): (src, dst) for src, dst in constraint_set}

    src_idx, dst_idx, edge_attr = [], [], []
    for i, src_node in enumerate(nodes):
        for j, dst_node in enumerate(nodes):
            if i == j:
                continue
            src_id = src_node["id"]
            dst_id = dst_node["id"]
            src_idx.append(i)
            dst_idx.append(j)

            pair_key = frozenset([src_id, dst_id])
            if pair_key in pair_forward:
                forward = pair_forward[pair_key]
                constraint_key = f"{forward[0]}->{forward[1]}"
                is_locked = bool(fd.constraints.get(constraint_key, True))
                if fd.held.get(src_id, False) or fd.held.get(dst_id, False):
                    is_locked = False
                src_blocks_dst = 1.0 if src_id == forward[0] else 0.0
                edge_attr.append([1.0, 1.0 if is_locked else 0.0, src_blocks_dst])
            else:
                edge_attr.append([0.0, 0.0, 0.0])

    return src_idx, dst_idx, edge_attr


def load_pyg_frame_products_only(
    episode_dir: str | Path,
    frame_idx: int,
    encoding_method: str = "random",
) -> Data:
    fd = load_frame_data(episode_dir, frame_idx)
    nodes = fd.graph["components"]
    x = _build_product_node_features(nodes, fd, encoding_method)
    src, dst, edge_attr = _build_product_edges(nodes, fd.graph, fd)
    return Data(
        x=x,
        edge_index=torch.tensor([src, dst], dtype=torch.long),
        edge_attr=torch.tensor(edge_attr, dtype=torch.float32),
        y=torch.tensor([frame_idx], dtype=torch.long),
        num_nodes=len(nodes),
    )


def load_pyg_frame_with_robot(
    episode_dir: str | Path,
    frame_idx: int,
    encoding_method: str = "random",
) -> Data:
    fd = load_frame_data(episode_dir, frame_idx)
    if fd.robot is None:
        return load_pyg_frame_products_only(episode_dir, frame_idx, encoding_method)

    products = fd.graph["components"]
    product_count = len(products)
    x_prod = _build_product_node_features(products, fd, encoding_method)
    robot_emb = fd.robot["embedding"].astype(np.float32)
    robot_pos = (
        fd.robot["centroid"].astype(np.float32)
        if int(fd.robot["depth_valid"][0]) == 1
        else np.zeros(POS_DIM, dtype=np.float32)
    )
    robot_feat = np.concatenate(
        [
            robot_emb,
            robot_pos,
            type_encode("robot", encoding_method),
            np.array([1.0], dtype=np.float32),
        ]
    )
    x = torch.cat([x_prod, torch.tensor(robot_feat, dtype=torch.float32).unsqueeze(0)], dim=0)

    src, dst, edge_attr = _build_product_edges(products, fd.graph, fd)
    robot_idx = product_count
    for i in range(product_count):
        src.append(robot_idx)
        dst.append(i)
        edge_attr.append([0.0, 0.0, 0.0])
        src.append(i)
        dst.append(robot_idx)
        edge_attr.append([0.0, 0.0, 0.0])

    data = Data(
        x=x,
        edge_index=torch.tensor([src, dst], dtype=torch.long),
        edge_attr=torch.tensor(edge_attr, dtype=torch.float32),
        y=torch.tensor([frame_idx], dtype=torch.long),
        num_nodes=product_count + 1,
    )
    data.robot_point_cloud = torch.tensor(fd.robot["point_cloud"], dtype=torch.float32)
    data.robot_pixel_coords = torch.tensor(fd.robot["pixel_coords"], dtype=torch.int32)
    data.robot_mask = torch.tensor(fd.robot["mask"], dtype=torch.uint8)
    return data


def load_pyg_frame_with_robot_state(
    episode_dir: str | Path,
    frame_idx: int,
    encoding_method: str = "random",
) -> Data:
    episode_dir = Path(episode_dir)
    data = load_pyg_frame_products_only(episode_dir, frame_idx, encoding_method)
    robot_states = np.load(episode_dir / "robot_states.npy")
    data.robot_state = torch.tensor(robot_states[frame_idx].astype(np.float32), dtype=torch.float32)
    return data


def load_pyg_frame_with_robot_action(
    episode_dir: str | Path,
    frame_idx: int,
    encoding_method: str = "random",
) -> Data:
    episode_dir = Path(episode_dir)
    data = load_pyg_frame_with_robot_state(episode_dir, frame_idx, encoding_method)
    robot_states = np.load(episode_dir / "robot_states.npy")
    if frame_idx + 1 < robot_states.shape[0]:
        action = robot_states[frame_idx + 1] - robot_states[frame_idx]
    else:
        action = np.zeros(ROBOT_STATE_DIM, dtype=np.float32)
    data.robot_action = torch.tensor(action.astype(np.float32), dtype=torch.float32)
    return data


_VARIANTS = {
    "products_only": load_pyg_frame_products_only,
    "with_robot": load_pyg_frame_with_robot,
    "with_robot_state": load_pyg_frame_with_robot_state,
    "with_robot_action": load_pyg_frame_with_robot_action,
}


def list_all_frame_graphs(
    episode_dir: str | Path,
    variant: str = "with_robot_state",
    encoding_method: str = "random",
):
    if variant not in _VARIANTS:
        raise ValueError(f"variant must be one of {list(_VARIANTS)}, got {variant!r}")
    loader = _VARIANTS[variant]
    for frame_idx in list_labeled_frames(episode_dir):
        yield frame_idx, loader(episode_dir, frame_idx, encoding_method=encoding_method)