svae_johanna_noise_trainer.py

"""
Johanna-128 Omega — Continue Gaussian-pretrained model on 16 noise types
==========================================================================
Loads the Gaussian-trained checkpoint (ep200, MSE=0.059) and expands
the signal vocabulary to all 16 noise distributions at 128×128.

The Gaussian knowledge is the foundation — the MLP already knows how to
invert the geometric projection for one distribution. Now we teach it
the other 15 without destroying what it learned.

Strategy:
  - Moderate lr (1e-4): fast enough to learn new distributions,
    slow enough to preserve Gaussian knowledge
  - Gaussian is 1 of 16 types, so it stays in the training mix
  - Same architecture: V=256, D=16, hidden=768, depth=4, 17M params
  - Batch=128 to stay under cusolver limit (128 × 64 patches = 8192 calls)
"""

import os
import torch
import torch.nn as nn
import torch.nn.functional as F
import torchvision
import torchvision.transforms as T
import math
import time
import numpy as np
from tqdm import tqdm

# ── HuggingFace auth from Colab secrets ──
try:
    from google.colab import userdata
    os.environ["HF_TOKEN"] = userdata.get('HF_TOKEN')
    from huggingface_hub import login
    login(token=os.environ["HF_TOKEN"])
except Exception:
    pass

# ── SVD Backend ──────────────────────────────────────────────────

try:
    from geolip_core.linalg.eigh import FLEigh, _FL_MAX_N
    HAS_FL = True
except ImportError:
    HAS_FL = False


def gram_eigh_svd_fp64(A):
    orig_dtype = A.dtype
    with torch.amp.autocast('cuda', enabled=False):
        A_d = A.double()
        G = torch.bmm(A_d.transpose(1, 2), A_d)
        G.diagonal(dim1=-2, dim2=-1).add_(1e-12)
        eigenvalues, V = torch.linalg.eigh(G)
        eigenvalues = eigenvalues.flip(-1)
        V = V.flip(-1)
        S = torch.sqrt(eigenvalues.clamp(min=1e-24))
        U = torch.bmm(A_d, V) / S.unsqueeze(1).clamp(min=1e-16)
        Vh = V.transpose(-2, -1).contiguous()
    return U.to(orig_dtype), S.to(orig_dtype), Vh.to(orig_dtype)


def svd_fp64(A):
    B, M, N = A.shape
    if HAS_FL and N <= _FL_MAX_N and A.is_cuda:
        orig_dtype = A.dtype
        with torch.amp.autocast('cuda', enabled=False):
            A_d = A.double()
            G = torch.bmm(A_d.transpose(1, 2), A_d)
            eigenvalues, V = FLEigh()(G.float())
            eigenvalues = eigenvalues.double().flip(-1)
            V = V.double().flip(-1)
            S = torch.sqrt(eigenvalues.clamp(min=1e-24))
            U = torch.bmm(A_d, V) / S.unsqueeze(1).clamp(min=1e-16)
            Vh = V.transpose(-2, -1).contiguous()
        return U.to(orig_dtype), S.to(orig_dtype), Vh.to(orig_dtype)
    else:
        return gram_eigh_svd_fp64(A)


# ── CV Monitoring ────────────────────────────────────────────────

def cayley_menger_vol2(points):
    B, N, D = points.shape
    pts = points.double()
    gram = torch.bmm(pts, pts.transpose(1, 2))
    norms = torch.diagonal(gram, dim1=1, dim2=2)
    d2 = F.relu(norms.unsqueeze(2) + norms.unsqueeze(1) - 2 * gram)
    cm = torch.zeros(B, N + 1, N + 1, device=points.device, dtype=torch.float64)
    cm[:, 0, 1:] = 1.0
    cm[:, 1:, 0] = 1.0
    cm[:, 1:, 1:] = d2
    k = N - 1
    sign = (-1.0) ** (k + 1)
    fact = math.factorial(k)
    return sign * torch.linalg.det(cm) / ((2 ** k) * (fact ** 2))


def cv_of(emb, n_samples=200):
    if emb.dim() != 2 or emb.shape[0] < 5:
        return 0.0
    N, D = emb.shape
    pool = min(N, 512)
    indices = torch.stack([torch.randperm(pool, device=emb.device)[:5] for _ in range(n_samples)])
    vol2 = cayley_menger_vol2(emb[:pool][indices])
    valid = vol2 > 1e-20
    if valid.sum() < 10:
        return 0.0
    vols = vol2[valid].sqrt()
    return (vols.std() / (vols.mean() + 1e-8)).item()


# ── Comprehensive Noise Dataset (128×128) ────────────────────────

class OmegaNoiseDataset(torch.utils.data.Dataset):
    """16 noise types at arbitrary resolution. Seed rotation."""

    N_TYPES = 16

    def __init__(self, size=1000000, img_size=128, seed_rotate_every=1000):
        self.size = size
        self.img_size = img_size
        self.seed_rotate_every = seed_rotate_every
        self._rng = np.random.RandomState(42)
        self._call_count = 0

    def __len__(self):
        return self.size

    def _rotate_seed(self):
        self._call_count += 1
        if self._call_count % self.seed_rotate_every == 0:
            new_seed = int.from_bytes(os.urandom(4), 'big')
            self._rng = np.random.RandomState(new_seed)
            torch.manual_seed(new_seed)

    def _pink_noise(self, shape):
        white = torch.randn(shape)
        S = torch.fft.rfft2(white)
        h, w = shape[-2], shape[-1]
        fy = torch.fft.fftfreq(h).unsqueeze(-1).expand(-1, w // 2 + 1)
        fx = torch.fft.rfftfreq(w).unsqueeze(0).expand(h, -1)
        f = torch.sqrt(fx**2 + fy**2).clamp(min=1e-8)
        S = S / f
        return torch.fft.irfft2(S, s=(h, w))

    def _brown_noise(self, shape):
        white = torch.randn(shape)
        S = torch.fft.rfft2(white)
        h, w = shape[-2], shape[-1]
        fy = torch.fft.fftfreq(h).unsqueeze(-1).expand(-1, w // 2 + 1)
        fx = torch.fft.rfftfreq(w).unsqueeze(0).expand(h, -1)
        f = (fx**2 + fy**2).clamp(min=1e-8)
        S = S / f
        return torch.fft.irfft2(S, s=(h, w))

    def __getitem__(self, idx):
        self._rotate_seed()
        s = self.img_size
        noise_type = idx % self.N_TYPES

        if noise_type == 0:
            img = torch.randn(3, s, s)
        elif noise_type == 1:
            img = torch.rand(3, s, s) * 2 - 1
        elif noise_type == 2:
            img = (torch.rand(3, s, s) - 0.5) * 4
        elif noise_type == 3:
            lam = self._rng.uniform(0.5, 20.0)
            img = torch.poisson(torch.full((3, s, s), lam)) / lam - 1.0
        elif noise_type == 4:
            img = self._pink_noise((3, s, s))
            img = img / (img.std() + 1e-8)
        elif noise_type == 5:
            img = self._brown_noise((3, s, s))
            img = img / (img.std() + 1e-8)
        elif noise_type == 6:
            img = torch.where(torch.rand(3, s, s) > 0.5,
                              torch.ones(3, s, s) * 2, torch.ones(3, s, s) * -2)
            img = img + torch.randn(3, s, s) * 0.1
        elif noise_type == 7:
            mask = torch.rand(3, s, s) > 0.9
            img = torch.randn(3, s, s) * mask.float() * 3
        elif noise_type == 8:
            block = self._rng.randint(2, 16)
            small = torch.randn(3, s // block + 1, s // block + 1)
            img = F.interpolate(small.unsqueeze(0), size=s, mode='nearest').squeeze(0)
        elif noise_type == 9:
            gy = torch.linspace(-2, 2, s).unsqueeze(1).expand(s, s)
            gx = torch.linspace(-2, 2, s).unsqueeze(0).expand(s, s)
            angle = self._rng.uniform(0, 2 * math.pi)
            grad = math.cos(angle) * gx + math.sin(angle) * gy
            img = grad.unsqueeze(0).expand(3, -1, -1) + torch.randn(3, s, s) * 0.5
        elif noise_type == 10:
            check_size = self._rng.randint(2, 16)
            coords_y = torch.arange(s) // check_size
            coords_x = torch.arange(s) // check_size
            checker = ((coords_y.unsqueeze(1) + coords_x.unsqueeze(0)) % 2).float() * 2 - 1
            img = checker.unsqueeze(0).expand(3, -1, -1) + torch.randn(3, s, s) * 0.3
        elif noise_type == 11:
            a = torch.randn(3, s, s)
            b = torch.rand(3, s, s) * 2 - 1
            alpha = self._rng.uniform(0.2, 0.8)
            img = alpha * a + (1 - alpha) * b
        elif noise_type == 12:
            img = torch.zeros(3, s, s)
            h2, w2 = s // 2, s // 2
            img[:, :h2, :w2] = torch.randn(3, h2, w2)
            img[:, :h2, w2:] = torch.rand(3, h2, w2) * 2 - 1
            img[:, h2:, :w2] = self._pink_noise((3, h2, w2)) / 2
            sp = torch.where(torch.rand(3, h2, w2) > 0.5,
                             torch.ones(3, h2, w2), -torch.ones(3, h2, w2))
            img[:, h2:, w2:] = sp
        elif noise_type == 13:
            u = torch.rand(3, s, s)
            img = torch.tan(math.pi * (u - 0.5))
            img = img.clamp(-3, 3)
        elif noise_type == 14:
            img = torch.empty(3, s, s).exponential_(1.0) - 1.0
        elif noise_type == 15:
            u = torch.rand(3, s, s) - 0.5
            img = -torch.sign(u) * torch.log1p(-2 * u.abs())

        img = img.clamp(-4, 4)
        return img.float(), noise_type


# ── Patch Utilities ──────────────────────────────────────────────

def extract_patches(images, patch_size=16):
    B, C, H, W = images.shape
    gh, gw = H // patch_size, W // patch_size
    patches = images.reshape(B, C, gh, patch_size, gw, patch_size)
    patches = patches.permute(0, 2, 4, 1, 3, 5)
    return patches.reshape(B, gh * gw, C * patch_size * patch_size), gh, gw


def stitch_patches(patches, gh, gw, patch_size=16):
    B = patches.shape[0]
    patches = patches.reshape(B, gh, gw, 3, patch_size, patch_size)
    patches = patches.permute(0, 3, 1, 4, 2, 5)
    return patches.reshape(B, 3, gh * patch_size, gw * patch_size)


class BoundarySmooth(nn.Module):
    def __init__(self, channels=3, mid=16):
        super().__init__()
        self.net = nn.Sequential(
            nn.Conv2d(channels, mid, 3, padding=1), nn.GELU(),
            nn.Conv2d(mid, channels, 3, padding=1))
        nn.init.zeros_(self.net[-1].weight)
        nn.init.zeros_(self.net[-1].bias)
    def forward(self, x):
        return x + self.net(x)


class SpectralCrossAttention(nn.Module):
    def __init__(self, D, n_heads=4, max_alpha=0.2, alpha_init=-2.0):
        super().__init__()
        self.n_heads = n_heads
        self.head_dim = D // n_heads
        self.max_alpha = max_alpha
        assert D % n_heads == 0
        self.qkv = nn.Linear(D, 3 * D)
        self.out_proj = nn.Linear(D, D)
        self.norm = nn.LayerNorm(D)
        self.scale = self.head_dim ** -0.5
        self.alpha_logits = nn.Parameter(torch.full((D,), alpha_init))

    @property
    def alpha(self):
        return self.max_alpha * torch.sigmoid(self.alpha_logits)

    def forward(self, S):
        B, N, D = S.shape
        S_normed = self.norm(S)
        qkv = self.qkv(S_normed).reshape(B, N, 3, self.n_heads, self.head_dim)
        qkv = qkv.permute(2, 0, 3, 1, 4)
        q, k, v = qkv[0], qkv[1], qkv[2]
        attn = (q @ k.transpose(-2, -1)) * self.scale
        attn = attn.softmax(dim=-1)
        out = (attn @ v).transpose(1, 2).reshape(B, N, D)
        gate = torch.tanh(self.out_proj(out))
        return S * (1.0 + self.alpha.unsqueeze(0).unsqueeze(0) * gate)


class PatchSVAE(nn.Module):
    def __init__(self, matrix_v=256, D=16, patch_size=16, hidden=768,
                 depth=4, n_cross_layers=2):
        super().__init__()
        self.matrix_v = matrix_v
        self.D = D
        self.patch_size = patch_size
        self.patch_dim = 3 * patch_size * patch_size
        self.mat_dim = matrix_v * D

        self.enc_in = nn.Linear(self.patch_dim, hidden)
        self.enc_blocks = nn.ModuleList([
            nn.Sequential(nn.LayerNorm(hidden), nn.Linear(hidden, hidden),
                          nn.GELU(), nn.Linear(hidden, hidden))
            for _ in range(depth)])
        self.enc_out = nn.Linear(hidden, self.mat_dim)

        self.dec_in = nn.Linear(self.mat_dim, hidden)
        self.dec_blocks = nn.ModuleList([
            nn.Sequential(nn.LayerNorm(hidden), nn.Linear(hidden, hidden),
                          nn.GELU(), nn.Linear(hidden, hidden))
            for _ in range(depth)])
        self.dec_out = nn.Linear(hidden, self.patch_dim)

        nn.init.orthogonal_(self.enc_out.weight)

        self.cross_attn = nn.ModuleList([
            SpectralCrossAttention(D, n_heads=min(4, D))
            for _ in range(n_cross_layers)])
        self.boundary_smooth = BoundarySmooth(channels=3, mid=16)

    def encode_patches(self, patches):
        B, N, _ = patches.shape
        flat = patches.reshape(B * N, -1)
        h = F.gelu(self.enc_in(flat))
        for block in self.enc_blocks:
            h = h + block(h)
        M = self.enc_out(h).reshape(B * N, self.matrix_v, self.D)
        M = F.normalize(M, dim=-1)
        U, S, Vt = svd_fp64(M)
        U = U.reshape(B, N, self.matrix_v, self.D)
        S = S.reshape(B, N, self.D)
        Vt = Vt.reshape(B, N, self.D, self.D)
        M = M.reshape(B, N, self.matrix_v, self.D)
        S_coord = S
        for layer in self.cross_attn:
            S_coord = layer(S_coord)
        return {'U': U, 'S_orig': S, 'S': S_coord, 'Vt': Vt, 'M': M}

    def decode_patches(self, U, S, Vt):
        B, N, V, D = U.shape
        U_flat = U.reshape(B * N, V, D)
        S_flat = S.reshape(B * N, D)
        Vt_flat = Vt.reshape(B * N, D, D)
        M_hat = torch.bmm(U_flat * S_flat.unsqueeze(1), Vt_flat)
        h = F.gelu(self.dec_in(M_hat.reshape(B * N, -1)))
        for block in self.dec_blocks:
            h = h + block(h)
        return self.dec_out(h).reshape(B, N, -1)

    def forward(self, images):
        patches, gh, gw = extract_patches(images, self.patch_size)
        svd = self.encode_patches(patches)
        decoded = self.decode_patches(svd['U'], svd['S'], svd['Vt'])
        recon = stitch_patches(decoded, gh, gw, self.patch_size)
        recon = self.boundary_smooth(recon)
        return {'recon': recon, 'svd': svd, 'gh': gh, 'gw': gw}

    @staticmethod
    def effective_rank(S):
        p = S / (S.sum(-1, keepdim=True) + 1e-8)
        p = p.clamp(min=1e-8)
        return (-(p * p.log()).sum(-1)).exp()


# ── Training ─────────────────────────────────────────────────────

def train():
    # ── Config ──
    V, D, patch_size = 256, 16, 16
    hidden, depth = 768, 4
    n_cross_layers = 2
    batch_size = 128  # 128 × 64 patches = 8192 eigh calls (safe)
    lr = 1e-4
    epochs = 200
    target_cv = 0.125
    cv_weight, boost, sigma = 0.3, 0.5, 0.15
    img_size = 128

    save_dir = '/content/checkpoints'
    save_every = 10
    report_every = 5000
    hf_repo = 'AbstractPhil/geolip-SVAE'
    hf_version = 'v16_johanna_omega'
    tb_dir = '/content/runs'

    # ── Pretrained checkpoint ──
    pretrained_repo = 'AbstractPhil/geolip-SVAE'
    pretrained_file = 'v14_noise/checkpoints/epoch_0200.pt'

    os.makedirs(save_dir, exist_ok=True)
    device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')

    # ── TensorBoard ──
    from torch.utils.tensorboard import SummaryWriter
    run_name = f"johanna_omega_V{V}_D{D}_h{hidden}_d{depth}"
    tb_path = os.path.join(tb_dir, run_name)
    writer = SummaryWriter(tb_path)
    print(f"  TensorBoard: {tb_path}")

    # ── HuggingFace ──
    hf_enabled = False
    try:
        from huggingface_hub import HfApi, hf_hub_download
        api = HfApi()
        api.whoami()
        hf_enabled = True
        hf_prefix = f"{hf_version}/checkpoints"
        print(f"  HuggingFace: {hf_repo}/{hf_prefix}")
    except Exception as e:
        print(f"  HuggingFace: disabled ({e})")

    def upload_to_hf(local_path, remote_name):
        if not hf_enabled:
            return
        try:
            api.upload_file(path_or_fileobj=local_path,
                            path_in_repo=f"{hf_prefix}/{remote_name}",
                            repo_id=hf_repo, repo_type="model")
            print(f"  ☁️  Uploaded: {hf_repo}/{hf_prefix}/{remote_name}")
        except Exception as e:
            print(f"  ⚠️  HF upload failed: {e}")

    # ── Load pretrained Johanna-128 Gaussian ──
    print(f"\n  Loading pretrained: {pretrained_repo}/{pretrained_file}")
    ckpt_path = hf_hub_download(repo_id=pretrained_repo, filename=pretrained_file,
                                 repo_type="model")
    ckpt = torch.load(ckpt_path, map_location='cpu', weights_only=False)
    print(f"  Pretrained epoch: {ckpt['epoch']}, MSE: {ckpt['test_mse']:.6f}")
    print(f"  Pretrained config: {ckpt['config']}")

    # ── Model ──
    model = PatchSVAE(matrix_v=V, D=D, patch_size=patch_size,
                      hidden=hidden, depth=depth,
                      n_cross_layers=n_cross_layers).to(device)
    model.load_state_dict(ckpt['model_state_dict'], strict=True)
    print(f"  Loaded {sum(p.numel() for p in model.parameters()):,} parameters")

    # Fresh optimizer — don't carry Gaussian momentum into omega training
    opt = torch.optim.Adam(model.parameters(), lr=lr)
    sched = torch.optim.lr_scheduler.CosineAnnealingLR(opt, T_max=epochs)

    # ── Data: 16-type omega noise at 128×128 ──
    train_ds = OmegaNoiseDataset(size=1280000, img_size=img_size)
    val_ds = OmegaNoiseDataset(size=10000, img_size=img_size)
    train_loader = torch.utils.data.DataLoader(
        train_ds, batch_size=batch_size, shuffle=True,
        num_workers=4, pin_memory=True, drop_last=True)
    test_loader = torch.utils.data.DataLoader(
        val_ds, batch_size=batch_size, shuffle=False,
        num_workers=4, pin_memory=True)

    n_patches = (img_size // patch_size) ** 2
    batches_per_epoch = len(train_loader)
    total_params = sum(p.numel() for p in model.parameters())

    print(f"\n  JOHANNA-128 OMEGA CONTINUATION")
    print(f"  Pretrained on: Gaussian N(0,1), 200 epochs, MSE=0.059")
    print(f"  Now training on: 16 noise types at {img_size}×{img_size}")
    print(f"  {n_patches} patches, ({V},{D}), hidden={hidden}, depth={depth}")
    print(f"  Params: {total_params:,}, batch={batch_size}")
    print(f"  Batches/epoch: {batches_per_epoch}, lr={lr}")
    print(f"  Report every {report_every} batches")
    print("=" * 100)
    print(f" {'ep':>3} {'batch':>8} | {'loss':>7} {'recon':>7} | "
          f"{'S0':>6} {'SD':>6} {'ratio':>5} {'erank':>5} | "
          f"{'row_cv':>7} {'prox':>5} {'rw':>5} | "
          f"{'S_delta':>7}")
    print("-" * 100)

    best_recon = float('inf')
    global_batch = 0

    def save_checkpoint(path, epoch, test_mse, extra=None, upload=True):
        ckpt_out = {
            'epoch': epoch, 'test_mse': test_mse,
            'global_batch': global_batch,
            'model_state_dict': model.state_dict(),
            'optimizer_state_dict': opt.state_dict(),
            'scheduler_state_dict': sched.state_dict(),
            'config': {
                'V': V, 'D': D, 'patch_size': patch_size,
                'hidden': hidden, 'depth': depth,
                'n_cross_layers': n_cross_layers,
                'target_cv': target_cv,
                'dataset': 'omega_noise_16types_128',
                'pretrained_from': 'v14_noise/epoch_0200.pt',
                'img_size': img_size, 'lr': lr,
            },
        }
        if extra:
            ckpt_out.update(extra)
        torch.save(ckpt_out, path)
        size_mb = os.path.getsize(path) / (1024 * 1024)
        print(f"  💾 Saved: {path} ({size_mb:.1f}MB, ep{epoch}, MSE={test_mse:.6f})")
        if upload:
            upload_to_hf(path, os.path.basename(path))

    # ── Training Loop ──
    for epoch in range(1, epochs + 1):
        model.train()
        total_loss, total_recon, n = 0, 0, 0
        last_cv, last_prox, recon_w = target_cv, 1.0, 1.0 + boost
        t0 = time.time()

        pbar = tqdm(train_loader, desc=f"Ep {epoch}/{epochs}",
                    bar_format='{l_bar}{bar:20}{r_bar}')
        for batch_idx, (images, noise_types) in enumerate(pbar):
            images = images.to(device)
            opt.zero_grad()
            out = model(images)
            recon_loss = F.mse_loss(out['recon'], images)

            with torch.no_grad():
                if batch_idx % 50 == 0:
                    current_cv = cv_of(out['svd']['M'][0, 0])
                    if current_cv > 0:
                        last_cv = current_cv
                    delta = last_cv - target_cv
                    last_prox = math.exp(-delta**2 / (2 * sigma**2))

            recon_w = 1.0 + boost * last_prox
            cv_pen = cv_weight * (1.0 - last_prox)
            cv_l = (last_cv - target_cv) ** 2
            loss = recon_w * recon_loss + cv_pen * cv_l
            loss.backward()

            torch.nn.utils.clip_grad_norm_(model.cross_attn.parameters(), max_norm=0.5)
            opt.step()

            total_loss += loss.item() * len(images)
            total_recon += recon_loss.item() * len(images)
            n += len(images)
            global_batch += 1

            pbar.set_postfix_str(
                f"loss={recon_loss.item():.4f} cv={last_cv:.3f} prox={last_prox:.2f}",
                refresh=False)

            # ── Readout ──
            if global_batch % report_every == 0:
                model.eval()
                with torch.no_grad():
                    test_imgs, _ = next(iter(test_loader))
                    test_imgs = test_imgs.to(device)
                    test_out = model(test_imgs)
                    test_mse = F.mse_loss(test_out['recon'], test_imgs).item()
                    S_mean = test_out['svd']['S'].mean(dim=(0, 1))
                    S_orig = test_out['svd']['S_orig'].mean(dim=(0, 1))
                    erank = model.effective_rank(
                        test_out['svd']['S'].reshape(-1, D)).mean().item()
                    s_delta = (S_mean - S_orig).abs().mean().item()
                    ratio = (S_mean[0] / (S_mean[-1] + 1e-8)).item()

                writer.add_scalar('train/recon', total_recon / n, global_batch)
                writer.add_scalar('test/recon_mse', test_mse, global_batch)
                writer.add_scalar('geo/row_cv', last_cv, global_batch)
                writer.add_scalar('geo/ratio', ratio, global_batch)
                writer.add_scalar('geo/erank', erank, global_batch)
                writer.add_scalar('geo/S0', S_mean[0].item(), global_batch)
                writer.add_scalar('cross_attn/s_delta', s_delta, global_batch)

                print(f"\n {epoch:3d} {global_batch:8d} | "
                      f"{total_loss/n:7.4f} {total_recon/n:7.4f} | "
                      f"{S_mean[0]:6.3f} {S_mean[-1]:6.3f} {ratio:5.2f} {erank:5.2f} | "
                      f"{last_cv:7.4f} {last_prox:5.3f} {recon_w:5.2f} | "
                      f"{s_delta:7.5f}")

                if test_mse < best_recon:
                    best_recon = test_mse
                    save_checkpoint(os.path.join(save_dir, 'best.pt'),
                                    epoch, test_mse, upload=False)
                model.train()

        pbar.close()
        sched.step()
        epoch_time = time.time() - t0

        writer.add_scalar('train/epoch_time', epoch_time, epoch)

        # ── Epoch eval ──
        model.eval()
        test_recon_total, test_n = 0, 0
        with torch.no_grad():
            for test_imgs, _ in test_loader:
                test_imgs = test_imgs.to(device)
                out = model(test_imgs)
                test_recon_total += F.mse_loss(out['recon'], test_imgs).item() * len(test_imgs)
                test_n += len(test_imgs)
        epoch_test_mse = test_recon_total / test_n

        print(f"  Epoch {epoch} done: {epoch_time:.1f}s, test_mse={epoch_test_mse:.6f}, "
              f"best={best_recon:.6f}")

        if epoch_test_mse < best_recon:
            best_recon = epoch_test_mse
            save_checkpoint(os.path.join(save_dir, 'best.pt'),
                            epoch, epoch_test_mse, upload=False)

        if epoch % save_every == 0:
            save_checkpoint(os.path.join(save_dir, f'epoch_{epoch:04d}.pt'),
                            epoch, epoch_test_mse)
            best_path = os.path.join(save_dir, 'best.pt')
            if os.path.exists(best_path):
                upload_to_hf(best_path, 'best.pt')
            writer.flush()
            if hf_enabled:
                try:
                    api.upload_folder(folder_path=tb_path,
                                      path_in_repo=f"{hf_version}/tensorboard/{run_name}",
                                      repo_id=hf_repo, repo_type="model")
                    print(f"  ☁️  TB synced")
                except:
                    pass

    writer.close()
    print(f"\n  JOHANNA-128 OMEGA TRAINING COMPLETE")
    print(f"  Best MSE: {best_recon:.6f}")
    print(f"  Checkpoints: {save_dir}/")


if __name__ == "__main__":
    torch.set_float32_matmul_precision('high')
    train()