path: root/experiments/synth_nonlinearity_ladder.py

"""
Synthetic Nonlinearity Ladder: teacher-student classification with controllable nonlinearity.

Teacher dynamics:
  h_{l+1}^* = h_l^* + W_l^* phi_alpha(h_l^*)
  phi_alpha(z) = (1-alpha)*z + alpha*tanh(z)

alpha=0 -> linear, alpha=1 -> fully nonlinear.
Sweep (alpha, L) to find where state bridge fails and credit bridge degrades.

Methods: BP, DFA, State Bridge, Credit Bridge (all reuse project conventions).
"""
import os
import sys
import json
import argparse
import time
import numpy as np
import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.optim as optim
from torch.utils.data import DataLoader, TensorDataset
import copy

sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))

from models.residual_mlp import ResidualMLP, ResidualBlock
from models.value_net import ValueNet, create_ema_model, update_ema
from models.state_bridge import StateBridgeNet
from metrics.credit_metrics import (
    cosine_similarity_batch, perturbation_correlation, nudging_test,
    offline_bp_cosine, feature_drift
)


# =============================================================================
# Teacher network and data generation
# =============================================================================
class TeacherNet:
    """Fixed teacher network with controllable nonlinearity."""

    def __init__(self, d_hidden, num_blocks, num_classes, alpha, seed=0):
        rng = np.random.RandomState(seed)
        self.d_hidden = d_hidden
        self.num_blocks = num_blocks
        self.num_classes = num_classes
        self.alpha = alpha

        # Teacher block weights: scaled for stability
        self.Ws = []
        for l in range(num_blocks):
            W = rng.randn(d_hidden, d_hidden).astype(np.float32)
            # Scale so spectral norm of residual part is small
            W = W / (np.linalg.norm(W, ord=2) + 1e-8) * 0.3
            self.Ws.append(torch.from_numpy(W))

        # Output projection
        U = rng.randn(num_classes, d_hidden).astype(np.float32)
        U = U / (np.linalg.norm(U, ord=2) + 1e-8)
        self.U = torch.from_numpy(U)

    def to(self, device):
        self.Ws = [W.to(device) for W in self.Ws]
        self.U = self.U.to(device)
        return self

    def phi(self, z):
        """Controllable activation: (1-alpha)*z + alpha*tanh(z)."""
        return (1 - self.alpha) * z + self.alpha * torch.tanh(z)

    def forward(self, h0):
        """Forward pass through teacher, returns logits and hidden states."""
        h = h0
        hiddens = [h]
        for l in range(self.num_blocks):
            f = F.linear(self.phi(h), self.Ws[l])
            h = h + f
            hiddens.append(h)
        logits = F.linear(h, self.U)
        return logits, hiddens


def generate_dataset(teacher, num_samples, d_hidden, device, seed=0):
    """Generate classification dataset from teacher network."""
    torch.manual_seed(seed)
    X = torch.randn(num_samples, d_hidden, device=device)
    with torch.no_grad():
        logits, _ = teacher.forward(X)
        Y = logits.argmax(dim=-1)
    return X, Y


# =============================================================================
# Student network (same architecture family as teacher)
# =============================================================================
class StudentBlock(nn.Module):
    """Residual block with pre-LayerNorm + phi_alpha."""

    def __init__(self, d_hidden, alpha):
        super().__init__()
        self.ln = nn.LayerNorm(d_hidden)
        self.w = nn.Linear(d_hidden, d_hidden, bias=False)
        self.alpha = alpha
        # Small init for stability
        nn.init.normal_(self.w.weight, std=0.01)

    def phi(self, z):
        return (1 - self.alpha) * z + self.alpha * torch.tanh(z)

    def forward(self, h):
        """Returns residual F_l(h), NOT h + F_l(h)."""
        return self.w(self.phi(self.ln(h)))


class StudentNet(nn.Module):
    """Student network: L residual blocks + linear output head."""

    def __init__(self, d_hidden, num_classes, num_blocks, alpha):
        super().__init__()
        self.blocks = nn.ModuleList([StudentBlock(d_hidden, alpha) for _ in range(num_blocks)])
        self.out_head = nn.Linear(d_hidden, num_classes)
        self.num_blocks = num_blocks
        self.d_hidden = d_hidden

    def forward(self, x, return_hidden=False):
        h = x  # No embedding needed, input is already d_hidden
        hiddens = [h] if return_hidden else None
        for block in self.blocks:
            f = block(h)
            h = h + f
            if return_hidden:
                hiddens.append(h)
        logits = self.out_head(h)
        if return_hidden:
            return logits, hiddens
        return logits

    def forward_from_layer(self, h, start_layer):
        """Run forward from a given layer to output."""
        for i in range(start_layer, self.num_blocks):
            f = self.blocks[i](h)
            h = h + f
        return self.out_head(h)


# =============================================================================
# Training methods
# =============================================================================
def evaluate(model, test_loader, device):
    model.eval()
    correct, total = 0, 0
    with torch.no_grad():
        for x, y in test_loader:
            x, y = x.to(device), y.to(device)
            logits = model(x)
            correct += (logits.argmax(1) == y).sum().item()
            total += x.size(0)
    return correct / total


def train_bp(model, train_loader, test_loader, device, args):
    """Standard BP training."""
    optimizer = optim.AdamW(model.parameters(), lr=args.lr, weight_decay=args.wd)
    scheduler = optim.lr_scheduler.CosineAnnealingLR(optimizer, T_max=args.epochs)
    log = {'train_loss': [], 'train_acc': [], 'test_acc': []}

    for epoch in range(1, args.epochs + 1):
        model.train()
        total_loss, correct, total = 0, 0, 0
        for x, y in train_loader:
            x, y = x.to(device), y.to(device)
            logits = model(x)
            loss = F.cross_entropy(logits, y)
            optimizer.zero_grad()
            loss.backward()
            optimizer.step()
            total_loss += loss.item() * x.size(0)
            correct += (logits.argmax(1) == y).sum().item()
            total += x.size(0)
        scheduler.step()
        log['train_loss'].append(total_loss / total)
        log['train_acc'].append(correct / total)
        log['test_acc'].append(evaluate(model, test_loader, device))
        if epoch % 20 == 0 or epoch == 1:
            print(f"    [BP] Ep {epoch}: loss={log['train_loss'][-1]:.4f} "
                  f"train={log['train_acc'][-1]:.4f} test={log['test_acc'][-1]:.4f}")
    return log


def train_dfa(model, train_loader, test_loader, device, args):
    """DFA training with fixed random feedback matrices."""
    d = model.d_hidden
    L = model.num_blocks
    C = args.num_classes

    Bs = [torch.randn(d, C, device=device) / np.sqrt(C) for _ in range(L)]
    block_opts = [optim.AdamW(block.parameters(), lr=args.lr, weight_decay=args.wd)
                  for block in model.blocks]
    head_opt = optim.AdamW(model.out_head.parameters(), lr=args.lr, weight_decay=args.wd)

    all_schedulers = ([optim.lr_scheduler.CosineAnnealingLR(o, T_max=args.epochs) for o in block_opts]
                      + [optim.lr_scheduler.CosineAnnealingLR(head_opt, T_max=args.epochs)])

    log = {'train_loss': [], 'train_acc': [], 'test_acc': []}

    for epoch in range(1, args.epochs + 1):
        model.train()
        total_loss, correct, total = 0, 0, 0
        for x, y in train_loader:
            x, y = x.to(device), y.to(device)
            batch = x.size(0)

            with torch.no_grad():
                logits, hiddens = model(x, return_hidden=True)
                loss_val = F.cross_entropy(logits, y)
                e_T = logits.softmax(dim=-1)
                e_T[torch.arange(batch), y] -= 1

            # Update output head (h_L detached)
            hL_det = hiddens[-1].detach()
            logits_out = model.out_head(hL_det)
            loss_out = F.cross_entropy(logits_out, y)
            head_opt.zero_grad()
            loss_out.backward()
            head_opt.step()

            # Update each block with DFA local surrogate
            for l in range(L):
                h_l = hiddens[l].detach()
                a_dfa = (e_T @ Bs[l].T).detach()
                rms = (a_dfa ** 2).mean(dim=-1, keepdim=True).sqrt() + 1e-6
                a_norm = a_dfa / rms
                f_l = model.blocks[l](h_l)
                local_loss = (f_l * a_norm).sum(dim=-1).mean()
                block_opts[l].zero_grad()
                local_loss.backward()
                torch.nn.utils.clip_grad_norm_(model.blocks[l].parameters(), 1.0)
                block_opts[l].step()

            total_loss += loss_val.item() * batch
            correct += (logits.argmax(1) == y).sum().item()
            total += batch

        for s in all_schedulers:
            s.step()
        log['train_loss'].append(total_loss / total)
        log['train_acc'].append(correct / total)
        log['test_acc'].append(evaluate(model, test_loader, device))
        if epoch % 20 == 0 or epoch == 1:
            print(f"    [DFA] Ep {epoch}: loss={log['train_loss'][-1]:.4f} "
                  f"train={log['train_acc'][-1]:.4f} test={log['test_acc'][-1]:.4f}")
    return log, Bs


def train_state_bridge(model, train_loader, test_loader, device, args):
    """State Bridge training."""
    d = model.d_hidden
    L = model.num_blocks
    C = args.num_classes

    state_pred = StateBridgeNet(d_hidden=d, s_dim=C, time_embed_dim=32,
                                hidden_dim=256, num_layers=3).to(device)

    block_opts = [optim.AdamW(block.parameters(), lr=args.lr, weight_decay=args.wd)
                  for block in model.blocks]
    head_opt = optim.AdamW(model.out_head.parameters(), lr=args.lr, weight_decay=args.wd)
    state_opt = optim.Adam(state_pred.parameters(), lr=args.lr_fb)

    all_schedulers = ([optim.lr_scheduler.CosineAnnealingLR(o, T_max=args.epochs) for o in block_opts]
                      + [optim.lr_scheduler.CosineAnnealingLR(head_opt, T_max=args.epochs)])

    log = {'train_loss': [], 'train_acc': [], 'test_acc': [], 'state_pred_error': []}

    for epoch in range(1, args.epochs + 1):
        model.train()
        state_pred.train()
        total_loss, correct, total = 0, 0, 0
        total_se = 0

        for x, y in train_loader:
            x, y = x.to(device), y.to(device)
            batch = x.size(0)

            with torch.no_grad():
                logits, hiddens = model(x, return_hidden=True)
                loss_val = F.cross_entropy(logits, y)
                e_T = logits.softmax(dim=-1)
                e_T[torch.arange(batch), y] -= 1
                s = e_T.detach()

            hL_det = hiddens[-1].detach()

            # Train state predictor
            state_loss = 0.0
            for l in range(L):
                h_l_det = hiddens[l].detach()
                t_l = torch.full((batch,), l / L, device=device)
                pred_hL = state_pred(h_l_det, t_l, s)
                target = hL_det
                target_norm = target.norm(dim=-1, keepdim=True).clamp(min=1.0)
                state_loss = state_loss + (((pred_hL - target) / target_norm) ** 2).sum(dim=-1).mean()
            state_loss = state_loss / L
            state_opt.zero_grad()
            state_loss.backward()
            state_opt.step()
            total_se += state_loss.item() * batch

            # Compute credits
            credits = []
            for l in range(L):
                h_l_det = hiddens[l].detach().requires_grad_(True)
                t_l = torch.full((batch,), l / L, device=device)
                pred_hL = state_pred(h_l_det, t_l, s)
                pred_logits = model.out_head(pred_hL)
                pred_loss = F.cross_entropy(pred_logits, y, reduction='sum')
                a_l = torch.autograd.grad(pred_loss, h_l_det, create_graph=False)[0]
                credits.append(a_l.detach())

            # Update output head
            logits_out = model.out_head(hL_det)
            loss_out = F.cross_entropy(logits_out, y)
            head_opt.zero_grad()
            loss_out.backward()
            head_opt.step()

            # Update blocks
            for l in range(L):
                h_l = hiddens[l].detach()
                a = credits[l]
                rms = (a ** 2).mean(dim=-1, keepdim=True).sqrt() + 1e-6
                a_norm = a / rms
                f_l = model.blocks[l](h_l)
                local_loss = (f_l * a_norm).sum(dim=-1).mean()
                block_opts[l].zero_grad()
                local_loss.backward()
                torch.nn.utils.clip_grad_norm_(model.blocks[l].parameters(), 1.0)
                block_opts[l].step()

            total_loss += loss_val.item() * batch
            correct += (logits.argmax(1) == y).sum().item()
            total += batch

        for sch in all_schedulers:
            sch.step()

        log['train_loss'].append(total_loss / total)
        log['train_acc'].append(correct / total)
        log['test_acc'].append(evaluate(model, test_loader, device))
        log['state_pred_error'].append(total_se / total)
        if epoch % 20 == 0 or epoch == 1:
            print(f"    [SB] Ep {epoch}: loss={log['train_loss'][-1]:.4f} "
                  f"train={log['train_acc'][-1]:.4f} test={log['test_acc'][-1]:.4f} "
                  f"se={log['state_pred_error'][-1]:.6f}")
    return log, state_pred


def train_credit_bridge(model, train_loader, test_loader, device, args):
    """Credit Bridge training with terminal gradient matching + bridge consistency."""
    d = model.d_hidden
    L = model.num_blocks
    C = args.num_classes
    warmup_epochs = max(1, args.epochs // 5)

    value_net = ValueNet(d_hidden=d, s_dim=C, time_embed_dim=32,
                         hidden_dim=256, num_layers=3).to(device)
    value_net_ema = create_ema_model(value_net)

    Bs_fallback = [torch.randn(d, C, device=device) / np.sqrt(C) for _ in range(L)]

    block_opts = [optim.AdamW(block.parameters(), lr=args.lr, weight_decay=args.wd)
                  for block in model.blocks]
    head_opt = optim.AdamW(model.out_head.parameters(), lr=args.lr, weight_decay=args.wd)
    value_opt = optim.Adam(value_net.parameters(), lr=args.lr_fb)

    all_schedulers = ([optim.lr_scheduler.CosineAnnealingLR(o, T_max=args.epochs) for o in block_opts]
                      + [optim.lr_scheduler.CosineAnnealingLR(head_opt, T_max=args.epochs)])

    lam = args.lam
    K_samples = args.K
    sigma_bridge = args.sigma_bridge
    ema_momentum = args.ema_momentum
    term_grad_weight = args.term_grad_weight

    log = {'train_loss': [], 'train_acc': [], 'test_acc': [],
           'value_loss': [], 'term_loss': [], 'bridge_loss': [], 'tgrad_loss': []}

    for epoch in range(1, args.epochs + 1):
        model.train()
        value_net.train()
        total_loss, correct, total = 0, 0, 0
        total_vloss, total_term, total_bridge, total_tgrad = 0, 0, 0, 0

        if epoch <= warmup_epochs:
            credit_blend = 0.0
        else:
            credit_blend = min(1.0, (epoch - warmup_epochs) / max(1, warmup_epochs))

        for x, y in train_loader:
            x, y = x.to(device), y.to(device)
            batch = x.size(0)

            with torch.no_grad():
                logits, hiddens = model(x, return_hidden=True)
                loss_val = F.cross_entropy(logits, y)
                e_T = logits.softmax(dim=-1)
                e_T[torch.arange(batch), y] -= 1
                s = e_T.detach()
                true_loss = F.cross_entropy(logits, y, reduction='none').detach()

            hL_det = hiddens[-1].detach()

            # ---- Train value net ----
            t_L = torch.ones(batch, device=device)
            V_terminal = value_net(hL_det, t_L, s)
            loss_term = ((V_terminal - true_loss) ** 2).mean()

            loss_tgrad = torch.tensor(0.0, device=device)
            if term_grad_weight > 0:
                hL_req = hL_det.clone().requires_grad_(True)
                V_at_L = value_net(hL_req, t_L, s)
                grad_V_L = torch.autograd.grad(V_at_L.sum(), hL_req, create_graph=True)[0]
                hL_req2 = hL_det.clone().requires_grad_(True)
                logits_tgt = model.out_head(hL_req2)
                ce_loss = F.cross_entropy(logits_tgt, y, reduction='sum')
                a_L_exact = torch.autograd.grad(ce_loss, hL_req2, create_graph=False)[0].detach()
                loss_tgrad = ((grad_V_L - a_L_exact) ** 2).sum(dim=-1).mean()

            loss_bridge = 0.0
            for l in range(L):
                h_l_det = hiddens[l].detach()
                t_l = torch.full((batch,), l / L, device=device)
                t_l_next = torch.full((batch,), (l + 1) / L, device=device)
                V_l = value_net(h_l_det, t_l, s)

                with torch.no_grad():
                    h_next_det = hiddens[l + 1].detach()
                    log_terms = []
                    for k in range(K_samples):
                        noise = sigma_bridge * torch.randn_like(h_next_det)
                        V_next = value_net_ema(h_next_det + noise, t_l_next, s)
                        log_terms.append(-V_next / lam)
                    log_stack = torch.stack(log_terms, dim=-1)
                    V_target = -lam * (torch.logsumexp(log_stack, dim=-1) - np.log(K_samples))

                loss_bridge = loss_bridge + ((V_l - V_target.detach()) ** 2).mean()
            loss_bridge = loss_bridge / L

            value_loss = loss_term + loss_bridge + term_grad_weight * loss_tgrad
            value_opt.zero_grad()
            value_loss.backward()
            torch.nn.utils.clip_grad_norm_(value_net.parameters(), 1.0)
            value_opt.step()
            update_ema(value_net, value_net_ema, ema_momentum)

            total_vloss += value_loss.item() * batch
            total_term += loss_term.item() * batch
            total_bridge += (loss_bridge.item() if isinstance(loss_bridge, torch.Tensor) else loss_bridge) * batch
            total_tgrad += loss_tgrad.item() * batch

            # ---- Compute credits ----
            cb_credits = []
            for l in range(L):
                h_l_det = hiddens[l].detach().requires_grad_(True)
                t_l = torch.full((batch,), l / L, device=device)
                V_l = value_net(h_l_det, t_l, s)
                a_l = torch.autograd.grad(V_l.sum(), h_l_det, create_graph=False)[0]
                cb_credits.append(a_l.detach())

            dfa_credits = [(e_T @ Bs_fallback[l].T).detach() for l in range(L)]

            credits = []
            for l in range(L):
                if credit_blend >= 1.0:
                    a = cb_credits[l]
                elif credit_blend <= 0.0:
                    a = dfa_credits[l]
                else:
                    cb_rms = (cb_credits[l] ** 2).mean(dim=-1, keepdim=True).sqrt() + 1e-6
                    dfa_rms = (dfa_credits[l] ** 2).mean(dim=-1, keepdim=True).sqrt() + 1e-6
                    a = credit_blend * (cb_credits[l] / cb_rms) + (1 - credit_blend) * (dfa_credits[l] / dfa_rms)
                credits.append(a)

            # ---- Update output head ----
            logits_out = model.out_head(hL_det)
            loss_out = F.cross_entropy(logits_out, y)
            head_opt.zero_grad()
            loss_out.backward()
            head_opt.step()

            # ---- Update blocks ----
            for l in range(L):
                h_l = hiddens[l].detach()
                a = credits[l]
                rms = (a ** 2).mean(dim=-1, keepdim=True).sqrt() + 1e-6
                a_norm = a / rms
                f_l = model.blocks[l](h_l)
                local_loss = (f_l * a_norm).sum(dim=-1).mean()
                block_opts[l].zero_grad()
                local_loss.backward()
                torch.nn.utils.clip_grad_norm_(model.blocks[l].parameters(), 1.0)
                block_opts[l].step()

            total_loss += loss_val.item() * batch
            correct += (logits.argmax(1) == y).sum().item()
            total += batch

        for sch in all_schedulers:
            sch.step()

        log['train_loss'].append(total_loss / total)
        log['train_acc'].append(correct / total)
        log['test_acc'].append(evaluate(model, test_loader, device))
        log['value_loss'].append(total_vloss / total)
        log['term_loss'].append(total_term / total)
        log['bridge_loss'].append(total_bridge / total)
        log['tgrad_loss'].append(total_tgrad / total)
        if epoch % 20 == 0 or epoch == 1:
            phase = "warmup" if epoch <= warmup_epochs else f"blend={credit_blend:.2f}"
            print(f"    [CB] Ep {epoch} ({phase}): loss={log['train_loss'][-1]:.4f} "
                  f"train={log['train_acc'][-1]:.4f} test={log['test_acc'][-1]:.4f} "
                  f"vloss={log['value_loss'][-1]:.6f}")
    return log, value_net, value_net_ema


# =============================================================================
# Diagnostics (per-layer)
# =============================================================================
def compute_diagnostics(model, method_name, test_loader, device, args,
                        value_net=None, state_predictor=None, dfa_Bs=None):
    """Compute per-layer diagnostic metrics."""
    model.eval()
    if value_net is not None:
        value_net.eval()
    if state_predictor is not None:
        state_predictor.eval()

    d = model.d_hidden
    L = model.num_blocks
    C = args.num_classes

    # Get one batch
    for x, y in test_loader:
        x, y = x.to(device), y.to(device)
        break

    batch = x.size(0)

    # BP gradients (for offline BP cosine)
    # Manual forward to get gradable hidden states
    h = x.detach().requires_grad_(True)
    hiddens_bp = [h]
    for block in model.blocks:
        f = block(hiddens_bp[-1])
        h_next = hiddens_bp[-1] + f
        hiddens_bp.append(h_next)
    logits_bp = model.out_head(hiddens_bp[-1])
    loss_bp = F.cross_entropy(logits_bp, y)
    bp_grads = {}
    grads = torch.autograd.grad(loss_bp, hiddens_bp, retain_graph=False)
    for l in range(L + 1):
        bp_grads[l] = grads[l].detach().clone()

    # Clean forward
    with torch.no_grad():
        logits, hiddens = model(x, return_hidden=True)
        e_T = logits.softmax(dim=-1)
        e_T[torch.arange(batch), y] -= 1
        s = e_T.detach()

    results = {
        'bp_cosine': [],
        'perturbation_rho': [],
        'nudging': {'0.001': [], '0.003': [], '0.01': []},
    }

    # State bridge prediction error (if applicable)
    if method_name == 'state_bridge' and state_predictor is not None:
        state_pred_errors = []
        for l in range(L):
            h_l_det = hiddens[l].detach()
            t_l = torch.full((batch,), l / L, device=device)
            with torch.no_grad():
                pred_hL = state_predictor(h_l_det, t_l, s)
                hL_det = hiddens[-1].detach()
                err = ((pred_hL - hL_det) ** 2).sum(dim=-1).mean().item()
                state_pred_errors.append(err)
        results['state_pred_error_per_layer'] = state_pred_errors

    for l in range(L):
        h_l = hiddens[l].detach()
        t_l = torch.full((batch,), l / L, device=device)

        if method_name == 'bp':
            a_l = bp_grads[l]
        elif method_name == 'dfa':
            a_l = (e_T @ dfa_Bs[l].T).detach()
        elif method_name == 'state_bridge':
            h_l_req = h_l.clone().requires_grad_(True)
            pred_hL = state_predictor(h_l_req, t_l, s)
            pred_logits = model.out_head(pred_hL)
            pred_loss = F.cross_entropy(pred_logits, y, reduction='sum')
            a_l = torch.autograd.grad(pred_loss, h_l_req, create_graph=False)[0].detach()
        elif method_name == 'credit_bridge':
            h_l_req = h_l.clone().requires_grad_(True)
            V_l = value_net(h_l_req, t_l, s)
            a_l = torch.autograd.grad(V_l.sum(), h_l_req, create_graph=False)[0].detach()
        else:
            raise ValueError(f"Unknown method: {method_name}")

        # BP cosine
        bp_cos = cosine_similarity_batch(a_l, bp_grads[l])
        results['bp_cosine'].append(bp_cos)

        # Forward function for perturbation and nudging
        def make_fwd_fn(start_l):
            def fwd_fn(h):
                with torch.no_grad():
                    curr = h
                    for i in range(start_l, L):
                        curr = curr + model.blocks[i](curr)
                    out = model.out_head(curr)
                    return F.cross_entropy(out, y, reduction='none')
            return fwd_fn

        fwd_fn = make_fwd_fn(l)
        rho = perturbation_correlation(h_l, a_l, fwd_fn, epsilon=1e-3, M=16)
        results['perturbation_rho'].append(rho)

        for eta in [0.001, 0.003, 0.01]:
            nud = nudging_test(h_l, a_l, fwd_fn, eta=eta)
            results['nudging'][str(eta)].append(nud)

    return results


# =============================================================================
# Main experiment runner
# =============================================================================
def run_single(alpha, L, seed, args, device):
    """Run all methods for a single (alpha, L, seed) configuration."""
    d = args.d_hidden
    C = args.num_classes

    print(f"\n  === alpha={alpha}, L={L}, seed={seed} ===")
    t0 = time.time()

    # Generate data from teacher
    teacher = TeacherNet(d, L, C, alpha, seed=0).to(device)  # Fixed teacher seed=0
    X_train, Y_train = generate_dataset(teacher, args.n_train, d, device, seed=seed)
    X_test, Y_test = generate_dataset(teacher, args.n_test, d, device, seed=seed + 10000)

    train_ds = TensorDataset(X_train, Y_train)
    test_ds = TensorDataset(X_test, Y_test)
    train_loader = DataLoader(train_ds, batch_size=args.batch_size, shuffle=True)
    test_loader = DataLoader(test_ds, batch_size=args.batch_size, shuffle=False)

    results = {}

    # ---- BP ----
    print("    --- BP ---")
    torch.manual_seed(seed)
    model_bp = StudentNet(d, C, L, alpha).to(device)
    bp_log = train_bp(model_bp, train_loader, test_loader, device, args)
    bp_diag = compute_diagnostics(model_bp, 'bp', test_loader, device, args)
    results['bp'] = {'log': bp_log, 'diagnostics': bp_diag}

    # ---- DFA ----
    print("    --- DFA ---")
    torch.manual_seed(seed)
    model_dfa = StudentNet(d, C, L, alpha).to(device)
    dfa_log, dfa_Bs = train_dfa(model_dfa, train_loader, test_loader, device, args)
    dfa_diag = compute_diagnostics(model_dfa, 'dfa', test_loader, device, args, dfa_Bs=dfa_Bs)
    results['dfa'] = {'log': dfa_log, 'diagnostics': dfa_diag}

    # ---- State Bridge ----
    print("    --- State Bridge ---")
    torch.manual_seed(seed)
    model_sb = StudentNet(d, C, L, alpha).to(device)
    sb_log, state_pred = train_state_bridge(model_sb, train_loader, test_loader, device, args)
    sb_diag = compute_diagnostics(model_sb, 'state_bridge', test_loader, device, args,
                                   state_predictor=state_pred)
    results['state_bridge'] = {'log': sb_log, 'diagnostics': sb_diag}

    # ---- Credit Bridge ----
    print("    --- Credit Bridge ---")
    torch.manual_seed(seed)
    model_cb = StudentNet(d, C, L, alpha).to(device)
    cb_log, vnet, vnet_ema = train_credit_bridge(model_cb, train_loader, test_loader, device, args)
    cb_diag = compute_diagnostics(model_cb, 'credit_bridge', test_loader, device, args,
                                   value_net=vnet)
    results['credit_bridge'] = {'log': cb_log, 'diagnostics': cb_diag}

    elapsed = time.time() - t0
    print(f"  === Done alpha={alpha}, L={L}, seed={seed} in {elapsed:.1f}s ===")

    # Summary
    for m in ['bp', 'dfa', 'state_bridge', 'credit_bridge']:
        test_acc = results[m]['log']['test_acc'][-1]
        mean_gamma = np.mean(results[m]['diagnostics']['bp_cosine'])
        mean_rho = np.mean(results[m]['diagnostics']['perturbation_rho'])
        mean_nudge = np.mean(results[m]['diagnostics']['nudging']['0.01'])
        print(f"    {m:20s}: acc={test_acc:.4f} Gamma={mean_gamma:.4f} "
              f"rho={mean_rho:.4f} nudge={mean_nudge:.6f}")

    return results


def serialize(obj):
    if isinstance(obj, dict):
        return {str(k): serialize(v) for k, v in obj.items()}
    elif isinstance(obj, list):
        return [serialize(v) for v in obj]
    elif isinstance(obj, (np.floating, np.integer)):
        return float(obj)
    elif isinstance(obj, np.ndarray):
        return obj.tolist()
    elif isinstance(obj, torch.Tensor):
        return obj.cpu().numpy().tolist()
    return obj


def main():
    parser = argparse.ArgumentParser(description='Synthetic Nonlinearity Ladder')
    parser.add_argument('--alphas', type=float, nargs='+', default=[0.0, 0.5, 1.0])
    parser.add_argument('--depths', type=int, nargs='+', default=[2, 8])
    parser.add_argument('--seeds', type=int, nargs='+', default=[42])
    parser.add_argument('--d_hidden', type=int, default=128)
    parser.add_argument('--num_classes', type=int, default=10)
    parser.add_argument('--n_train', type=int, default=10000)
    parser.add_argument('--n_test', type=int, default=2000)
    parser.add_argument('--batch_size', type=int, default=256)
    parser.add_argument('--epochs', type=int, default=60)
    parser.add_argument('--lr', type=float, default=1e-3)
    parser.add_argument('--lr_fb', type=float, default=1e-3)
    parser.add_argument('--wd', type=float, default=0.01)
    parser.add_argument('--lam', type=float, default=0.1)
    parser.add_argument('--K', type=int, default=4)
    parser.add_argument('--sigma_bridge', type=float, default=0.05)
    parser.add_argument('--ema_momentum', type=float, default=0.995)
    parser.add_argument('--term_grad_weight', type=float, default=1.0)
    parser.add_argument('--gpu', type=int, default=1)
    parser.add_argument('--output_dir', type=str, default='results/synth_ladder')
    args = parser.parse_args()

    device = torch.device(f'cuda:{args.gpu}' if torch.cuda.is_available() else 'cpu')
    print(f"Device: {device}")
    print(f"Alphas: {args.alphas}")
    print(f"Depths: {args.depths}")
    print(f"Seeds: {args.seeds}")

    os.makedirs(args.output_dir, exist_ok=True)

    all_results = {}

    for alpha in args.alphas:
        for L in args.depths:
            for seed in args.seeds:
                key = f"a{alpha}_L{L}_s{seed}"
                result = run_single(alpha, L, seed, args, device)
                all_results[key] = result

                # Save incrementally
                out_path = os.path.join(args.output_dir, f'synth_{key}.json')
                with open(out_path, 'w') as f:
                    json.dump(serialize(result), f, indent=2)

    # Save combined summary
    summary = {}
    for key, result in all_results.items():
        s = {}
        for method in ['bp', 'dfa', 'state_bridge', 'credit_bridge']:
            r = result[method]
            diag = r['diagnostics']
            s[method] = {
                'test_acc': r['log']['test_acc'][-1],
                'mean_bp_cosine': float(np.mean(diag['bp_cosine'])),
                'mean_rho': float(np.mean(diag['perturbation_rho'])),
                'mean_nudge_001': float(np.mean(diag['nudging']['0.001'])),
                'mean_nudge_003': float(np.mean(diag['nudging']['0.003'])),
                'mean_nudge_01': float(np.mean(diag['nudging']['0.01'])),
                'bp_cosine_per_layer': [float(x) for x in diag['bp_cosine']],
                'rho_per_layer': [float(x) for x in diag['perturbation_rho']],
                'nudge_per_layer': [float(x) for x in diag['nudging']['0.01']],
            }
            if method == 'state_bridge' and 'state_pred_error_per_layer' in diag:
                s[method]['state_pred_error_per_layer'] = [float(x) for x in diag['state_pred_error_per_layer']]
                s[method]['mean_state_pred_error'] = float(np.mean(diag['state_pred_error_per_layer']))
            if method == 'credit_bridge':
                s[method]['final_value_loss'] = r['log']['value_loss'][-1]
                s[method]['final_term_loss'] = r['log']['term_loss'][-1]
                s[method]['final_bridge_loss'] = r['log']['bridge_loss'][-1]
                s[method]['final_tgrad_loss'] = r['log']['tgrad_loss'][-1]
        summary[key] = s

    summary_path = os.path.join(args.output_dir, 'summary.json')
    with open(summary_path, 'w') as f:
        json.dump(summary, f, indent=2)
    print(f"\nSummary saved to {summary_path}")

    # Save config
    config = serialize(vars(args))
    config_path = os.path.join(args.output_dir, 'config.json')
    with open(config_path, 'w') as f:
        json.dump(config, f, indent=2)

    # Print final summary table
    print("\n" + "=" * 100)
    print("SYNTHETIC NONLINEARITY LADDER - SUMMARY")
    print("=" * 100)
    print(f"{'Config':<20} {'Method':<20} {'Acc':>8} {'Gamma':>8} {'rho':>8} {'nudge':>10}")
    print("-" * 100)
    for key in sorted(summary.keys()):
        for method in ['bp', 'dfa', 'state_bridge', 'credit_bridge']:
            s = summary[key][method]
            print(f"{key:<20} {method:<20} {s['test_acc']:>8.4f} {s['mean_bp_cosine']:>8.4f} "
                  f"{s['mean_rho']:>8.4f} {s['mean_nudge_01']:>10.6f}")
        print("-" * 100)


if __name__ == '__main__':
    main()