Upload 4 files

5a1.py

@@ -0,0 +1,886 @@
+#!/usr/bin/env python3
+# 5L.py — joint AR+NAT+SAT trainer/decoder (Qwen3 tokenizer)
+# Robust fresh-start, ignores *.pt.tmp, AMP dtype auto, OOM backoff, progressive block growth.
+
+from __future__ import annotations
+import argparse, json, math, pathlib, random, time, os
+from contextlib import nullcontext
+from typing import Dict, Any, List, Optional, Tuple
+
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+from datasets import load_dataset
+from transformers import AutoTokenizer, logging as hf_log
+from tqdm.auto import tqdm
+
+# ───────────────────────── Globals ─────────────────────────
+hf_log.set_verbosity_error()
+DEV = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+torch.backends.cuda.matmul.allow_tf32 = True
+try:
+    torch.set_float32_matmul_precision("high")
+except Exception:
+    pass
+
+# Use the Qwen3 tokenizer (can override with env TOKENIZER_ID if needed)
+TOKENIZER_ID = os.environ.get(
+    "TOKENIZER_ID",
+    "Qwen/Qwen3-235B-A22B-Thinking-2507"
+)
+
+# Some Qwen tokenizers require trust_remote_code
+tok = AutoTokenizer.from_pretrained(TOKENIZER_ID, use_fast=True, trust_remote_code=True)
+if tok.pad_token is None:
+    tok.add_special_tokens({"pad_token": "[PAD]"})
+VOCAB, BLANK, EOS = (
+    max(tok.get_vocab().values()) + 1,   # allow new [PAD] if appended
+    tok.pad_token_id,
+    tok.eos_token_id if tok.eos_token_id is not None else tok.sep_token_id
+)
+
+PRESETS: Dict[str, Dict[str, int]] = {
+    "small":   dict(d=512, layers=8,  heads=16, rank=64),
+    "smallx2": dict(d=512, layers=16, heads=16, rank=64),  # NEW preset: small ×2
+    "base":    dict(d=768, layers=12, heads=24, rank=96),
+}
+
+# Safe default for 1× Tesla P40; override with --block
+DEFAULT_BLOCK = 576
+SAT_BLOCK = 2
+LR_CORE, LR_HEAD = 5e-5, 2e-4
+EMIT_LAMBDA = 0.1
+DEFAULT_SAVE_SEC = 8 * 24 * 3600  # 8 days
+CKDIR = pathlib.Path("ckpts_joint")
+
+
+# ───────────────────────── Utilities ─────────────────────────
+def rng_state():
+    if DEV.type == "cuda":
+        try:
+            return torch.cuda.get_rng_state(DEV)
+        except TypeError:
+            return torch.cuda.get_rng_state()
+    return torch.get_rng_state()
+
+
+def _is_probably_ckpt(path: pathlib.Path) -> bool:
+    try:
+        return path.is_file() and path.suffix == ".pt" and not path.name.endswith(".pt.tmp") and path.stat().st_size > (1<<20)
+    except Exception:
+        return False
+
+
+def _resolve_ckpt(path: pathlib.Path) -> pathlib.Path | None:
+    """
+    Return a solid .pt (never .tmp). If 'path' is dir, pick newest *.pt.
+    If not usable, return None.
+    """
+    try:
+        if path.is_dir():
+            cands = sorted([p for p in path.glob("*.pt") if _is_probably_ckpt(p)],
+                           key=lambda p: p.stat().st_mtime, reverse=True)
+            return cands[0] if cands else None
+        if path.suffix == ".tmp":
+            solid = path.with_suffix("")
+            return solid if _is_probably_ckpt(solid) else _resolve_ckpt(path.parent)
+        return path if _is_probably_ckpt(path) else _resolve_ckpt(path.parent)
+    except Exception:
+        return None
+
+
+def _try_load(path: pathlib.Path, map_location="cpu"):
+    try:
+        # NOTE: keep default weights_only behavior for compatibility with older checkpoints
+        return torch.load(path, map_location=map_location)
+    except Exception as e:
+        print(f"[ckpt-skip] {path} not usable: {e}")
+        return None
+
+
+# ───────────────────────── Repetition penalty utility ─────────────────────────
+def apply_repetition_penalty_(logits: torch.Tensor, generated_tokens, penalty: float):
+    """
+    In-place apply HuggingFace-style repetition penalty to logits.
+
+    - logits: Tensor of shape (V,), (B, V), or (B, L, V).
+      If (B, L, V) the function will operate on the last time dimension (i.e., logits[:, :L, :]).
+      Returns the modified logits (same object or a view).
+    - generated_tokens: for batch_size=1: list[int] (tokens already present in sequence).
+      for batch_size>1: list[list[int]] of length B.
+      We assume tokens are token ids between 0 and V-1.
+    - penalty: float; >1.0 applies penalty, 1.0 disables.
+
+    Rule:
+      if score < 0: score *= penalty
+      else: score /= penalty
+    """
+    if penalty is None or penalty <= 1.0:
+        return logits
+
+    # normalize shapes
+    orig_dim = logits.dim()
+    if orig_dim == 1:
+        logits_view = logits.unsqueeze(0)  # (1, V)
+        wrote_back = True
+    elif orig_dim == 2:
+        logits_view = logits  # (B, V)
+        wrote_back = False
+    elif orig_dim == 3:
+        # operate on last time-steps block(s). Caller can pass logits[:, :stride, :].
+        # Here we operate on logits_view which will be shape (B, L, V) -> we'll flatten batch/time for per-token ops.
+        logits_view = logits.view(-1, logits.size(-1))  # (B*L, V)
+        wrote_back = True
+    else:
+        raise ValueError(f"Unsupported logits dimensionality: {orig_dim}")
+
+    # At this point logits_view is (M, V) where M = B or B*L
+    M, V = logits_view.shape
+
+    # Normalize generated_tokens to a list-of-lists of length B (assuming caller knows batch size)
+    # If orig_dim == 1 or 2, treat M==B; if orig_dim==3 we can't reconstruct B,L from caller here (so caller should pass slice)
+    if orig_dim == 1:
+        batch_generated = [list(generated_tokens) if isinstance(generated_tokens, (list, tuple)) else [int(generated_tokens)]]
+    elif orig_dim == 2:
+        # Expect generated_tokens as list-of-lists or a single list for all batches
+        if isinstance(generated_tokens, (list, tuple)) and len(generated_tokens) > 0 and isinstance(generated_tokens[0], (list, tuple)):
+            batch_generated = [list(g) for g in generated_tokens]
+        else:
+            # single list applied to all batch entries
+            batch_generated = [list(generated_tokens) if isinstance(generated_tokens, (list, tuple)) else [int(generated_tokens)]] * M
+    else:
+        # orig_dim == 3 -> caller should pass appropriate generated per (B, L) slice; fallback: use first list across M rows
+        if isinstance(generated_tokens, (list, tuple)) and len(generated_tokens) > 0 and isinstance(generated_tokens[0], (list, tuple)):
+            # flatten lists across time if any; then reuse for each row
+            flat0 = [int(t) for t in generated_tokens[0]]
+            batch_generated = [flat0] * M
+        else:
+            flat = list(generated_tokens) if isinstance(generated_tokens, (list, tuple)) else [int(generated_tokens)]
+            batch_generated = [flat] * M
+
+    # apply penalty row-wise
+    # To reduce Python loop overhead we still iterate token-wise per row (vocab may be big but unique tokens are few)
+    for row_idx in range(M):
+        gen = batch_generated[row_idx] if row_idx < len(batch_generated) else batch_generated[0]
+        if not gen:
+            continue
+        unique_tokens = set(int(t) for t in gen if 0 <= int(t) < V)
+        if not unique_tokens:
+            continue
+        # vectorized-ish: apply per token
+        for tokid in unique_tokens:
+            col = logits_view[row_idx, tokid]
+            if col < 0:
+                logits_view[row_idx, tokid] = col * penalty
+            else:
+                logits_view[row_idx, tokid] = col / penalty
+
+    # write back if needed
+    if wrote_back and orig_dim == 1:
+        return logits_view.squeeze(0)
+    elif wrote_back and orig_dim == 3:
+        # caller provided the (B,L,V) as a contiguous tensor originally; we flattened so we need user to pass slice or handle writing
+        return logits.view_as(logits)  # unchanged reference; the in-place ops already modified original memory if contiguous
+    else:
+        return logits_view
+
+
+# ───────────────────────── AMP helper ─────────────────────────
+try:
+    from torch.amp import autocast as _ac, GradScaler
+except ImportError:
+    from torch.cuda.amp import autocast as _ac, GradScaler
+
+def _auto_amp_dtype():
+    if DEV.type == "cuda":
+        try:
+            # prefer bf16 only when actually supported; otherwise fp16
+            if torch.cuda.is_bf16_supported():
+                return torch.bfloat16
+            return torch.float16
+        except Exception:
+            return torch.float16
+    return torch.float32
+
+def amp(enabled):
+    return nullcontext() if not enabled else _ac(device_type="cuda", dtype=_auto_amp_dtype())
+
+
+# ───────────────────────── Data stream ─────────────────────────
+def token_stream(ds_name: str, target: int, seed: int = 42):
+    ds = load_dataset(ds_name, split="train", streaming=True)
+    ds = ds.shuffle(buffer_size=10_000, seed=seed)
+    emitted = 0
+    for ex in ds:
+        # ensure EOS between docs
+        enc = tok.encode(ex["text"])
+        if EOS is not None and (len(enc) == 0 or enc[-1] != EOS):
+            enc = enc + [EOS]
+        for t in enc:
+            yield t
+            emitted += 1
+            if emitted >= target:
+                return
+
+
+# ───────────────────────── Relative positional bias (ALiBi) ─────────────────────────
+def _alibi_slopes(n_heads: int):
+    import math
+    def pow2slopes(n):
+        start = 2 ** (-2 ** -(math.log2(n) - 3))
+        ratio = start
+        return [start * (ratio ** i) for i in range(n)]
+    if math.log2(n_heads).is_integer():
+        vals = pow2slopes(n_heads)
+    else:
+        closest = 2 ** math.floor(math.log2(n_heads))
+        vals = pow2slopes(closest)
+        extra = pow2slopes(2 * closest)
+        vals += extra[0::2][: n_heads - closest]
+    return torch.tensor(vals, device=DEV).view(1, n_heads, 1, 1)
+
+def alibi_bias(n_heads: int, n_tokens: int):
+    i = torch.arange(n_tokens, device=DEV).view(1, 1, n_tokens, 1)
+    j = torch.arange(n_tokens, device=DEV).view(1, 1, 1, n_tokens)
+    dist = (j - i).clamp_min(0)  # only penalize future
+    slopes = _alibi_slopes(n_heads)
+    return -slopes * dist
+
+
+# ───────────────────────── Model components ─────────────────────────
+class LowRankMHA(nn.Module):
+    """
+    Cache-aware MHA with low-rank projections; supports kv caching for decode.
+    """
+    def __init__(self, d: int, h: int, r: int, use_relpos: bool = True):
+        super().__init__()
+        assert d % h == 0, "d must be divisible by number of heads"
+        self.h, self.dk = h, d // h
+        self.use_relpos = use_relpos
+        self.q = nn.Linear(d, d, bias=False)
+        self.k = nn.Linear(d, d, bias=False)
+        self.v = nn.Linear(d, d, bias=False)
+        self.U = nn.Parameter(torch.randn(self.dk, r))
+        nn.init.orthogonal_(self.U)
+        self.proj = nn.Linear(h * r, d, bias=False)
+        self.drop = nn.Dropout(0.1)
+
+    def _proj(self, x):
+        B, N, _ = x.shape
+        return (x.view(B, N, self.h, self.dk).transpose(1, 2) @ self.U)
+
+    def forward(
+        self,
+        x: torch.Tensor,
+        mask: Optional[torch.Tensor] = None,
+        rel_bias_tokens: Optional[int] = None,
+        kv_cache: Optional[Tuple[torch.Tensor, torch.Tensor]] = None,
+        use_cache: bool = False,
+    ):
+        q = self._proj(self.q(x))
+        k_new = self._proj(self.k(x))
+        v_new = self._proj(self.v(x))
+
+        if kv_cache is None:
+            k, v = k_new, v_new
+        else:
+            k, v = kv_cache
+            if use_cache:
+                k = torch.cat([k, k_new], dim=2)
+                v = torch.cat([v, v_new], dim=2)
+
+        att = (q @ k.transpose(-1, -2)) / math.sqrt(self.dk)
+
+        if q.size(2) == k.size(2):
+            if self.use_relpos and rel_bias_tokens is not None:
+                att = att + alibi_bias(self.h, rel_bias_tokens)
+            if mask is not None:
+                att = att + mask
+
+        z = (att.softmax(-1) @ v).transpose(1, 2)  # (B,Nq,h,r)
+        z = z.reshape(x.size(0), x.size(1), -1)
+        out = self.drop(self.proj(z))
+        return (out, (k, v)) if use_cache else out
+
+
+class Block(nn.Module):
+    def __init__(self, d: int, h: int, r: int):
+        super().__init__()
+        self.ln1, self.ln2 = nn.LayerNorm(d), nn.LayerNorm(d)
+        self.mha = LowRankMHA(d, h, r, use_relpos=True)
+        self.ff = nn.Sequential(nn.Linear(d, 4 * d), nn.ReLU(), nn.Linear(4 * d, d))
+
+    def forward(
+        self,
+        x: torch.Tensor,
+        mask: Optional[torch.Tensor],
+        kv: Optional[Tuple[torch.Tensor, torch.Tensor]] = None,
+        use_cache: bool = False
+    ):
+        n = x.size(1)
+        if use_cache:
+            y, new_kv = self.mha(self.ln1(x), mask, rel_bias_tokens=n if mask is not None else None, kv_cache=kv, use_cache=True)
+            x = x + y
+            x = x + self.ff(self.ln2(x))
+            return x, new_kv
+        else:
+            x = x + self.mha(self.ln1(x), mask, rel_bias_tokens=n)
+            return x + self.ff(self.ln2(x))
+
+
+class Encoder(nn.Module):
+    """
+    Transformer encoder with optional kv caching (for AR/SAT decode).
+    """
+    def __init__(self, cfg: Dict[str, int]):
+        super().__init__()
+        d, l, h, r = cfg["d"], cfg["layers"], cfg["heads"], cfg["rank"]
+        self.emb = nn.Embedding(VOCAB, d)
+        self.blocks = nn.ModuleList(Block(d, h, r) for _ in range(l))
+        self.ln = nn.LayerNorm(d)
+
+    def forward(
+        self,
+        ids: torch.Tensor,
+        mask: Optional[torch.Tensor],
+        kv_caches: Optional[List[Optional[Tuple[torch.Tensor, torch.Tensor]]]] = None,
+        use_cache: bool = False
+    ):
+        x = self.emb(ids)
+        if not use_cache:
+            for blk in self.blocks:
+                x = blk(x, mask)
+            return self.ln(x)
+
+        new_kvs: List[Tuple[torch.Tensor, torch.Tensor]] = []
+        for i, blk in enumerate(self.blocks):
+            kv = kv_caches[i] if (kv_caches is not None) else None
+            x, kv_out = blk(x, mask, kv, use_cache=True)
+            new_kvs.append(kv_out)
+        return self.ln(x), new_kvs
+
+
+class ARHead(nn.Module):
+    def __init__(self, d):
+        super().__init__()
+        self.proj = nn.Linear(d, VOCAB)
+    def forward(self, h): return self.proj(h)
+
+
+class NATHead(nn.Module):
+    def __init__(self, d):
+        super().__init__()
+        self.proj = nn.Linear(d, VOCAB)
+    def forward(self, h): return self.proj(h)
+
+
+class SATHead(nn.Module):
+    def __init__(self, d, mode="var"):
+        super().__init__()
+        self.proj = nn.Linear(d, VOCAB)
+        self.mode = mode
+        self.gate = nn.Linear(d, 2) if mode == "var" else None
+    def forward(self, h_last):
+        logits = self.proj(h_last)
+        gate = self.gate(h_last[:, 0]) if self.gate is not None else None
+        return logits, gate
+
+
+# ───────────────────────── Masks ─────────────────────────
+def causal_mask(n):
+    m = torch.full((1, 1, n, n), float("-inf"), device=DEV)
+    return torch.triu(m, 1)
+
+def sat_mask(n, block=SAT_BLOCK):
+    idx = torch.arange(n, device=DEV)
+    grp = idx.unsqueeze(0) // block
+    allow = (grp.T == grp) | (grp.T > grp)
+    return torch.where(allow, 0.0, float("-inf")).unsqueeze(0).unsqueeze(0)
+
+
+# ───────────────────────── Checkpoint helpers ─────────────────────────
+def save_ckpt(
+    path: pathlib.Path,
+    core: nn.Module,
+    ar_h: nn.Module,
+    nat_h: nn.Module,
+    sat_h: nn.Module,
+    opt: torch.optim.Optimizer,
+    scaler: GradScaler,
+    meta: Dict[str, Any],
+):
+    path.parent.mkdir(exist_ok=True, parents=True)
+    tmp = path.with_suffix(path.suffix + ".tmp")
+    state = {
+        "core": core.state_dict(),
+        "ar": ar_h.state_dict(),
+        "nat": nat_h.state_dict(),
+        "sat": sat_h.state_dict(),
+        "opt": opt.state_dict(),
+        "scaler": scaler.state_dict(),
+        "cfg": meta.get("cfg"),
+        "tokenizer_id": TOKENIZER_ID,
+        **{k: v for k, v in meta.items() if k not in {"cfg"}},
+    }
+    torch.save(state, tmp, _use_new_zipfile_serialization=False)
+    tmp.replace(path)
+    (path.parent / "latest.json").write_text(json.dumps({"path": str(path), "step": meta["step"]}))
+    print(f"\n✓ saved checkpoint {path.name}")
+
+
+def load_ckpt(
+    path: pathlib.Path,
+    core: nn.Module,
+    ar_h: nn.Module,
+    nat_h: nn.Module,
+    sat_h: nn.Module,
+    opt: torch.optim.Optimizer,
+    scaler: GradScaler,
+):
+    p = _resolve_ckpt(path) or path
+    ck = _try_load(p, map_location=DEV)
+    if ck is None:
+        raise FileNotFoundError(f"No valid checkpoint at {p}")
+    core.load_state_dict(ck["core"])
+    ar_h.load_state_dict(ck["ar"])
+    nat_h.load_state_dict(ck["nat"])
+    sat_h.load_state_dict(ck["sat"])
+    opt.load_state_dict(ck["opt"])
+    scaler.load_state_dict(ck["scaler"])
+    return ck.get("step", 0), ck.get("seen_tok", 0), ck.get("wall_time", time.time())
+
+
+def _safe_load_any(path: pathlib.Path, tgt: nn.Module, key: str | None = None, rename: str | None = None):
+    p = _resolve_ckpt(path) or path
+    if not p.exists(): return 0
+    ck = _try_load(p, map_location=DEV)
+    if ck is None: return 0
+    sd = ck.get(key, ck) if key else ck
+    if isinstance(sd, dict) and "state_dict" in sd:
+        sd = sd["state_dict"]
+    if rename:
+        sd = {k.replace(rename, "proj."): v for k, v in sd.items() if rename in k}
+    tgt_sd = tgt.state_dict()
+    filt = {k: v for k, v in sd.items() if k in tgt_sd and v.shape == tgt_sd[k].shape}
+    if filt:
+        tgt.load_state_dict(filt, strict=False)
+    return len(filt)
+
+
+def infer_cfg_from_ckpt(path: pathlib.Path):
+    p = _resolve_ckpt(path) or path
+    if not p.exists(): return None
+    sd = _try_load(p, map_location="cpu")
+    if sd is None: return None
+    if isinstance(sd, dict) and "cfg" in sd and isinstance(sd["cfg"], dict):
+        return dict(sd["cfg"])
+    core = sd.get("core")
+    if core is None: return None
+    emb_w = core.get("emb.weight")
+    if emb_w is None: return None
+    d = emb_w.shape[1]
+    layer_ids = []
+    for k in core.keys():
+        if k.startswith("blocks."):
+            parts = k.split(".")
+            if len(parts) > 2 and parts[1].isdigit():
+                layer_ids.append(int(parts[1]))
+    layers = (max(layer_ids) + 1) if layer_ids else None
+    U = core.get("blocks.0.mha.U")
+    heads = rank = None
+    if U is not None:
+        dk, r = U.shape
+        rank = r
+        heads = d // dk if dk > 0 else None
+    out = {"d": d}
+    if layers is not None: out["layers"] = layers
+    if heads is not None:  out["heads"] = heads
+    if rank is not None:   out["rank"] = rank
+    return out
+
+
+# ───────────────────────── Train loop ─────────────────────────
+def _parse_grow_plan(s: str) -> List[int]:
+    steps = []
+    for part in s.split(","):
+        part = part.strip()
+        if part:
+            v = int(part)
+            if v >= 128:
+                steps.append(v)
+    return sorted(set(steps))
+
+
+def train(args):
+    cfg = PRESETS[args.preset].copy()
+
+    # Previous topology probe (unless --fresh)
+    if not args.fresh:
+        src_probe = pathlib.Path(args.warmstart_from) if args.warmstart_from else pathlib.Path(args.save_dir) / "final.pt"
+        prev_cfg = infer_cfg_from_ckpt(src_probe)
+    else:
+        prev_cfg = None
+
+    if prev_cfg:
+        cfg["d"] = prev_cfg.get("d", cfg["d"])
+        if prev_cfg.get("heads"):
+            cfg["heads"] = prev_cfg["heads"]
+        if args.rank is None and prev_cfg.get("rank"):
+            cfg["rank"] = prev_cfg["rank"]
+        # NEW: copy layers from ckpt even without --x2
+        if prev_cfg.get("layers"):
+            cfg["layers"] = prev_cfg["layers"]
+        # Optional doubling only when explicitly requested
+        if args.x2 and prev_cfg.get("layers"):
+            cfg["layers"] = max(cfg["layers"], prev_cfg["layers"] * 2)
+    if args.rank:
+        cfg["rank"] = args.rank
+    if args.x2 and not prev_cfg:
+        cfg["layers"] *= 2
+
+    BLOCK = args.block or DEFAULT_BLOCK
+
+    core = Encoder(cfg).to(DEV)
+    ar_h, nat_h = ARHead(cfg["d"]).to(DEV), NATHead(cfg["d"]).to(DEV)
+    sat_h = SATHead(cfg["d"], mode="var").to(DEV)
+
+    # Warm start unless --fresh
+    loaded = 0
+    if not args.fresh:
+        src = pathlib.Path(args.warmstart_from) if args.warmstart_from else pathlib.Path(args.save_dir) / "final.pt"
+        src = _resolve_ckpt(src)
+        if src:
+            loaded += _safe_load_any(src, core, key="core")
+            loaded += _safe_load_any(src, ar_h, key="ar")
+            loaded += _safe_load_any(src, nat_h, key="nat")
+            loaded += _safe_load_any(src, sat_h, key="sat")
+            if loaded:
+                print(f"Warm-start: loaded {loaded} matching tensors from {src}")
+
+    opt = torch.optim.AdamW(
+        [
+            {"params": core.parameters(), "lr": LR_CORE},
+            {"params": ar_h.parameters(), "lr": LR_HEAD},
+            {"params": nat_h.parameters(), "lr": LR_HEAD},
+            {"params": sat_h.parameters(), "lr": LR_HEAD},
+        ]
+    )
+    scaler = GradScaler(enabled=args.amp and DEV.type == "cuda")
+
+    ce_tok = nn.CrossEntropyLoss(label_smoothing=0.1)
+    ctc = nn.CTCLoss(blank=BLANK, zero_infinity=True)
+    ce_gate = nn.CrossEntropyLoss()
+
+    # ---------- resume bookkeeping ----------
+    start_step, seen_tok = 0, 0
+    last_save_time = time.time()
+    if args.resume and not args.fresh:
+        start_step, seen_tok, last_save_time = load_ckpt(
+            pathlib.Path(args.resume), core, ar_h, nat_h, sat_h, opt, scaler
+        )
+        print(f"✓ resumed from step {start_step:,}, seen_tokens={seen_tok:,}")
+
+    # Target tokens
+    if args.target_tokens:
+        target_tokens = args.target_tokens
+    else:
+        param_count = sum(p.numel() for p in core.parameters())
+        target_tokens = int(25 * param_count)
+
+    new_tokens_needed = target_tokens - seen_tok
+    if new_tokens_needed <= 0:
+        print("Target already reached – nothing to train.")
+        return
+    new_steps = new_tokens_needed // BLOCK
+    if args.steps:
+        new_steps = min(new_steps, args.steps)
+        new_tokens_needed = new_steps * BLOCK
+
+    total_tokens_needed = seen_tok + new_tokens_needed
+    print(f"[auto-steps] {new_steps:,} training steps (@ {BLOCK} tokens/step)")
+
+    # Progressive growth plan
+    grow_plan = _parse_grow_plan(args.grow_plan) if args.auto_grow else []
+    if args.auto_grow:
+        if BLOCK not in grow_plan:
+            grow_plan = sorted(set(grow_plan + [BLOCK]))
+        print(f"[auto-grow] plan: {grow_plan} every {args.grow_every_steps} steps")
+
+    stream = token_stream(args.source, target_tokens, seed=42)
+    buf: list[int] = []
+    pbar = tqdm(total=total_tokens_needed, initial=seen_tok, unit="tok")
+    step = start_step
+    steps_since_last_grow = 0
+
+    while seen_tok < total_tokens_needed:
+        # ------- assemble one batch -------
+        try:
+            while len(buf) < BLOCK:
+                buf.append(next(stream))
+        except StopIteration:
+            break
+        ids = torch.tensor(buf[:BLOCK], device=DEV).unsqueeze(0)  # (B=1, N)
+        buf = buf[BLOCK:]
+
+        tgt_ar = ids.clone()                           # (1, N)
+        ids_nat = torch.repeat_interleave(ids, 2, 1)   # (1, 2N) for NAT only
+
+        try:
+            with amp(args.amp):
+                # AR path
+                h_ar = core(ids, causal_mask(ids.size(1)))
+                logits_ar = ar_h(h_ar)[:, :-1]
+                loss_ar = ce_tok(logits_ar.reshape(-1, VOCAB), tgt_ar[:, 1:].reshape(-1))
+
+                # NAT path (uses doubled sequence)
+                h_nat = core(ids_nat, None)
+                log_nat = nat_h(h_nat).log_softmax(-1).transpose(0, 1)  # (T,B,V)
+                ilen = tlen = torch.tensor([ids_nat.size(1) // 2], device=DEV)
+                loss_nat = ctc(log_nat, tgt_ar, ilen, tlen)
+
+                # SAT path
+                h_sat = core(ids, sat_mask(ids.size(1)))
+                logits_sat, gate = sat_h(h_sat[:, -SAT_BLOCK:])
+                tgt_sat = ids[:, 1:SAT_BLOCK+1]
+                loss_sat = ce_tok(logits_sat.reshape(-1, VOCAB), tgt_sat.reshape(-1))
+                if gate is not None:
+                    loss_sat += EMIT_LAMBDA * ce_gate(gate, torch.ones(ids.size(0), device=DEV, dtype=torch.long))
+
+                loss = loss_ar + loss_nat + loss_sat
+
+            # optimisation
+            scaler.scale(loss).backward()
+            scaler.unscale_(opt)
+            nn.utils.clip_grad_norm_(core.parameters(), 1.0)
+            scaler.step(opt)
+            scaler.update()
+            opt.zero_grad(set_to_none=True)
+
+        except RuntimeError as e:
+            msg = str(e).lower()
+            if "out of memory" in msg or "cuda error" in msg:
+                new_block = max(128, BLOCK // 2)
+                if new_block < BLOCK:
+                    print(f"\n[OOM] reducing block from {BLOCK} -> {new_block}")
+                    BLOCK = new_block
+                    torch.cuda.empty_cache()
+                    buf = ids[0].tolist() + buf
+                    steps_since_last_grow = 0
+                    continue
+            raise
+
+        # progress
+        step += 1
+        seen_tok += BLOCK
+        pbar.update(BLOCK)
+        pbar.set_postfix(loss=f"{loss.item():.3f}", block=BLOCK)
+
+        # checkpoint cadence
+        now = time.time()
+        time_due = (now - last_save_time) >= args.save_every_sec > 0
+        step_due = args.save_every_steps > 0 and step % args.save_every_steps == 0
+        if time_due or step_due:
+            ck_name = f"step{step:08d}.pt"
+            save_ckpt(
+                pathlib.Path(args.save_dir) / ck_name,
+                core, ar_h, nat_h, sat_h, opt, scaler,
+                meta={
+                    "cfg": cfg,
+                    "step": step,
+                    "seen_tok": seen_tok,
+                    "wall_time": now,
+                    "py_state": random.getstate(),
+                    "torch_state": rng_state(),
+                },
+            )
+            last_save_time = now
+
+        # progressive growth
+        if args.auto_grow:
+            steps_since_last_grow += 1
+            if steps_since_last_grow >= args.grow_every_steps:
+                steps_since_last_grow = 0
+                try:
+                    idx = grow_plan.index(BLOCK)
+                    if idx + 1 < len(grow_plan):
+                        candidate = grow_plan[idx + 1]
+                        print(f"[auto-grow] attempting BLOCK {BLOCK} -> {candidate}")
+                        BLOCK = candidate
+                        torch.cuda.empty_cache()
+                    else:
+                        print("[auto-grow] at max planned block; no further growth.")
+                except ValueError:
+                    grow_plan = sorted(set(grow_plan + [BLOCK]))
+                    idx = grow_plan.index(BLOCK)
+                    if idx + 1 < len(grow_plan):
+                        candidate = grow_plan[idx + 1]
+                        print(f"[auto-grow] moving to planned BLOCK {candidate}")
+                        BLOCK = candidate
+                        torch.cuda.empty_cache()
+
+    pbar.close()
+
+    # final save
+    save_ckpt(
+        pathlib.Path(args.save_dir) / "final.pt",
+        core, ar_h, nat_h, sat_h, opt, scaler,
+        meta={
+            "cfg": cfg,
+            "step": step,
+            "seen_tok": seen_tok,
+            "wall_time": time.time(),
+            "py_state": random.getstate(),
+            "torch_state": rng_state(),
+        },
+    )
+    print("🎉 training complete")
+
+
+# ───────────────────────── Inference helpers ─────────────────────────
+def load_joint(ckpt: str, preset: str):
+    path = _resolve_ckpt(pathlib.Path(ckpt)) or pathlib.Path(ckpt)
+    sd = _try_load(path, map_location=DEV)
+    if sd is None:
+        raise FileNotFoundError(f"No valid checkpoint at {path}")
+    cfg = sd["cfg"] if "cfg" in sd and isinstance(sd["cfg"], dict) else PRESETS[preset]
+    core = Encoder(cfg).to(DEV)
+    ar_h, nat_h = ARHead(cfg["d"]).to(DEV), NATHead(cfg["d"]).to(DEV)
+    sat_h = SATHead(cfg["d"]).to(DEV)
+    core.load_state_dict(sd["core"])
+    ar_h.load_state_dict(sd["ar"])
+    nat_h.load_state_dict(sd["nat"])
+    sat_h.load_state_dict(sd["sat"])
+    return core, ar_h, nat_h, sat_h
+
+
+@torch.no_grad()
+def ar_decode(core, ar_h, prompt: str, max_new: int, T: float, repetition_penalty: float = 1.0):
+    ids = torch.tensor([tok.encode(prompt)], device=DEV)
+    if ids.size(1) == 0:
+        ids = torch.tensor([[EOS] if EOS is not None else [0]], device=DEV)
+    # prepare generated token list (include prompt tokens so we avoid repeating prompt if desired)
+    generated = ids[0].tolist()
+
+    # full warmup pass to populate kv caches
+    h_full, kvs = core(ids, causal_mask(ids.size(1)), use_cache=True)
+    logits_full = ar_h(h_full)  # (B, N, V)
+    last_logits = logits_full[:, -1, :]  # (B, V)
+    # apply repetition penalty with current generated tokens
+    apply_repetition_penalty_(last_logits, generated, repetition_penalty)
+    # sample first next token
+    probs = (last_logits / max(T, 1e-5)).softmax(-1)
+    nxt = probs.multinomial(1)
+    ids = torch.cat([ids, nxt], 1)
+    generated.append(int(nxt[0, 0].item()))
+
+    start = time.time()
+    for _ in range(max(0, max_new - 1)):
+        x = ids[:, -1:]
+        h_step, kvs = core(x, None, kv_caches=kvs, use_cache=True)
+        logits_step = ar_h(h_step)[:, -1]  # (B, V)
+        apply_repetition_penalty_(logits_step, generated, repetition_penalty)
+        probs = (logits_step / max(T, 1e-5)).softmax(-1)
+        nxt = probs.multinomial(1)
+        ids = torch.cat([ids, nxt], 1)
+        generated.append(int(nxt[0, 0].item()))
+    print(tok.decode(ids[0].tolist(), skip_special_tokens=True))
+    print(f"[{max_new} tok in {time.time() - start:.2f}s]")
+
+
+@torch.no_grad()
+def sat_decode(core, sat_h, prompt, max_new, T, var, repetition_penalty: float = 1.0):
+    ids = torch.tensor([tok.encode(prompt)], device=DEV)
+    added, t0 = 0, time.time()
+    # track generated tokens as list
+    generated = ids[0].tolist()
+    while added < max_new:
+        h = core(ids, sat_mask(ids.size(1)))
+        logits, gate = sat_h(h[:, -SAT_BLOCK:])  # logits shape (B, SAT_BLOCK, V)
+        stride = 2 if (not var or gate is None) else (gate.softmax(-1).multinomial(1) + 1).item()
+        # slice first `stride` positions (along SAT_BLOCK) and apply repetition penalty
+        logits_slice = logits[:, :stride, :]  # (B, stride, V)
+        # our penalty function expects (B, L, V) or (B, V) etc., so pass logits_slice directly
+        apply_repetition_penalty_(logits_slice, generated, repetition_penalty)
+        probs = torch.softmax(logits_slice / max(T, 1e-5), -1)  # (B, stride, V)
+        # sampling: reshape to (B*stride, V) then multinomial per row — but current code expects (1, stride, V)
+        # keep same shape-handling as original: assume B==1
+        nxt = probs.reshape(1, stride, VOCAB).multinomial(1).squeeze(-1)
+        # nxt shape (1, stride) -> add to ids
+        ids = torch.cat([ids, nxt], 1)
+        # update generated tokens list
+        generated.extend([int(x) for x in nxt[0].tolist()])
+        added += stride
+    print(tok.decode(ids[0].tolist(), skip_special_tokens=True))
+    print(f"[{added} tok in {time.time() - t0:.2f}s]")
+
+
+@torch.no_grad()
+def nat_decode(core, nat_h, prompt, max_new, passes, streams):
+    ids = torch.tensor([tok.encode(prompt) + [BLANK] * (max_new * 2)], device=DEV)
+    t0 = time.time()
+    for _ in range(passes):
+        h = core(ids, None)
+        logits = nat_h(h)
+        logits[..., BLANK] = -1e9
+        cand = logits.topk(streams, -1).indices.permute(2, 0, 1)
+        best = (cand != BLANK).float().mean(-1).argmax(0)
+        ids = cand[best, torch.arange(ids.size(0), device=DEV)][:, ::2]
+    out = [t for t in ids[0].tolist() if t != BLANK]
+    print(tok.decode(out, skip_special_tokens=True))
+    print(f"[{len(out)} output tokens in {time.time() - t0:.2f}s]")
+
+
+# ───────────────────────── CLI ─────────────────────────
+def main():
+    ap = argparse.ArgumentParser()
+    sub = ap.add_subparsers(dest="cmd", required=True)
+
+    tr = sub.add_parser("train")
+    tr.add_argument("--preset", choices=PRESETS, default="small")
+    tr.add_argument("--rank", type=int)
+    tr.add_argument("--block", type=int, default=DEFAULT_BLOCK)
+    tr.add_argument("--source", default="cerebras/SlimPajama-627B")
+    tr.add_argument("--target_tokens", type=int)
+    tr.add_argument("--steps", type=int)
+    tr.add_argument("--amp", action="store_true")
+    tr.add_argument("--save_every_sec", type=int, default=DEFAULT_SAVE_SEC)
+    tr.add_argument("--save_every_steps", type=int, default=0)
+    tr.add_argument("--save_dir", default=str(CKDIR))
+    tr.add_argument("--resume", type=str)
+    tr.add_argument("--x2", action="store_true", help="~2x params by doubling layers")
+    tr.add_argument("--warmstart_from", type=str, default=None, help="Path to previous final.pt for shape-safe warm start")
+    tr.add_argument("--fresh", action="store_true", help="Start from scratch: do not probe or load any checkpoints")
+
+    # Progressive block growth
+    tr.add_argument("--auto_grow", action="store_true", help="Automatically grow block size over time")
+    tr.add_argument("--grow_plan", type=str, default="576,640,768,896,1024", help="Comma list of block sizes to try in order")
+    tr.add_argument("--grow_every_steps", type=int, default=50000, help="Steps between growth attempts")
+
+    inf = sub.add_parser("infer")
+    inf.add_argument("--mode", choices=["ar", "nat", "sat"], required=True)
+    inf.add_argument("--ckpt", required=True)
+    inf.add_argument("--preset", default="small")
+    inf.add_argument("--prompt", required=True)
+    inf.add_argument("--max_new", type=int, default=120)
+    inf.add_argument("--temperature", type=float, default=1.0)
+    inf.add_argument("--var", action="store_true")
+    inf.add_argument("--passes", type=int, default=1)
+    inf.add_argument("--streams", type=int, default=5)
+    inf.add_argument("--repetition_penalty", type=float, default=1.0,
+                     help=">1.0 penalises previously generated tokens (HuggingFace style). 1.0 disables.")
+
+    args = ap.parse_args()
+    if args.cmd == "train":
+        train(args)
+    else:
+        core, ar_h, nat_h, sat_h = load_joint(args.ckpt, args.preset)
+        if args.mode == "ar":
+            ar_decode(core, ar_h, args.prompt, args.max_new, args.temperature, args.repetition_penalty)
+        elif args.mode == "sat":
+            sat_decode(core, sat_h, args.prompt, args.max_new, args.temperature, args.var, args.repetition_penalty)
+        else:
+            nat_decode(core, nat_h, args.prompt, args.max_new, args.passes, args.streams)
+
+
+if __name__ == "__main__":
+    main()

5a2.py

@@ -0,0 +1,868 @@
+#!/usr/bin/env python3
+# 5L.py — joint AR+NAT+SAT trainer/decoder (Qwen3 tokenizer)
+# Robust fresh-start, ignores *.pt.tmp, AMP dtype auto, OOM backoff, progressive block growth.
+# + repetition penalty for sampling (AR & SAT), tokenizer auto-sync from ckpt, SAT multinomial shape fix.
+
+from __future__ import annotations
+import argparse, json, math, pathlib, random, time, os
+from contextlib import nullcontext
+from typing import Dict, Any, List, Optional, Tuple
+from collections import deque
+
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+from datasets import load_dataset
+from transformers import AutoTokenizer, logging as hf_log
+from tqdm.auto import tqdm
+
+# ───────────────────────── Globals ─────────────────────────
+hf_log.set_verbosity_error()
+DEV = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+torch.backends.cuda.matmul.allow_tf32 = True
+try:
+    torch.set_float32_matmul_precision("high")
+except Exception:
+    pass
+
+# Use the Qwen3 tokenizer (can override with env TOKENIZER_ID if needed)
+TOKENIZER_ID = os.environ.get(
+    "TOKENIZER_ID",
+    "Qwen/Qwen3-235B-A22B-Thinking-2507"
+)
+
+# Some Qwen tokenizers require trust_remote_code
+tok = AutoTokenizer.from_pretrained(TOKENIZER_ID, use_fast=True, trust_remote_code=True)
+if tok.pad_token is None:
+    tok.add_special_tokens({"pad_token": "[PAD]"})
+VOCAB, BLANK, EOS = (
+    max(tok.get_vocab().values()) + 1,   # allow new [PAD] if appended
+    tok.pad_token_id,
+    tok.eos_token_id if tok.eos_token_id is not None else tok.sep_token_id
+)
+
+PRESETS: Dict[str, Dict[str, int]] = {
+    "small":   dict(d=512, layers=8,  heads=16, rank=64),
+    "smallx2": dict(d=512, layers=16, heads=16, rank=64),  # NEW preset: small ×2
+    "base":    dict(d=768, layers=12, heads=24, rank=96),
+}
+
+# Safe default for 1× Tesla P40; override with --block
+DEFAULT_BLOCK = 576
+SAT_BLOCK = 2
+LR_CORE, LR_HEAD = 5e-5, 2e-4
+EMIT_LAMBDA = 0.1
+DEFAULT_SAVE_SEC = 8 * 24 * 3600  # 8 days
+CKDIR = pathlib.Path("ckpts_joint")
+
+
+# ───────────────────────── Utilities ─────────────────────────
+def rng_state():
+    if DEV.type == "cuda":
+        try:
+            return torch.cuda.get_rng_state(DEV)
+        except TypeError:
+            return torch.cuda.get_rng_state()
+    return torch.get_rng_state()
+
+
+def _is_probably_ckpt(path: pathlib.Path) -> bool:
+    try:
+        return path.is_file() and path.suffix == ".pt" and not path.name.endswith(".pt.tmp") and path.stat().st_size > (1<<20)
+    except Exception:
+        return False
+
+
+def _resolve_ckpt(path: pathlib.Path) -> pathlib.Path | None:
+    """
+    Return a solid .pt (never .tmp). If 'path' is dir, pick newest *.pt.
+    If not usable, return None.
+    """
+    try:
+        if path.is_dir():
+            cands = sorted([p for p in path.glob("*.pt") if _is_probably_ckpt(p)],
+                           key=lambda p: p.stat().st_mtime, reverse=True)
+            return cands[0] if cands else None
+        if path.suffix == ".tmp":
+            solid = path.with_suffix("")
+            return solid if _is_probably_ckpt(solid) else _resolve_ckpt(path.parent)
+        return path if _is_probably_ckpt(path) else _resolve_ckpt(path.parent)
+    except Exception:
+        return None
+
+
+def _try_load(path: pathlib.Path, map_location="cpu"):
+    try:
+        # NOTE: keep default weights_only behavior for compatibility with older checkpoints
+        return torch.load(path, map_location=map_location)
+    except Exception as e:
+        print(f"[ckpt-skip] {path} not usable: {e}")
+        return None
+
+
+# ───────────────────────── AMP helper ─────────────────────────
+try:
+    from torch.amp import autocast as _ac, GradScaler
+except ImportError:
+    from torch.cuda.amp import autocast as _ac, GradScaler
+
+def _auto_amp_dtype():
+    if DEV.type == "cuda":
+        try:
+            # prefer bf16 only when actually supported; otherwise fp16
+            if torch.cuda.is_bf16_supported():
+                return torch.bfloat16
+            return torch.float16
+        except Exception:
+            return torch.float16
+    return torch.float32
+
+def amp(enabled):
+    return nullcontext() if not enabled else _ac(device_type="cuda", dtype=_auto_amp_dtype())
+
+
+# ───────────────────────── Data stream ─────────────────────────
+def token_stream(ds_name: str, target: int, seed: int = 42):
+    ds = load_dataset(ds_name, split="train", streaming=True)
+    ds = ds.shuffle(buffer_size=10_000, seed=seed)
+    emitted = 0
+    for ex in ds:
+        # ensure EOS between docs
+        enc = tok.encode(ex["text"])
+        if EOS is not None and (len(enc) == 0 or enc[-1] != EOS):
+            enc = enc + [EOS]
+        for t in enc:
+            yield t
+            emitted += 1
+            if emitted >= target:
+                return
+
+
+# ───────────────────────── Relative positional bias (ALiBi) ─────────────────────────
+def _alibi_slopes(n_heads: int):
+    import math
+    def pow2slopes(n):
+        start = 2 ** (-2 ** -(math.log2(n) - 3))
+        ratio = start
+        return [start * (ratio ** i) for i in range(n)]
+    if math.log2(n_heads).is_integer():
+        vals = pow2slopes(n_heads)
+    else:
+        closest = 2 ** math.floor(math.log2(n_heads))
+        vals = pow2slopes(closest)
+        extra = pow2slopes(2 * closest)
+        vals += extra[0::2][: n_heads - closest]
+    return torch.tensor(vals, device=DEV).view(1, n_heads, 1, 1)
+
+def alibi_bias(n_heads: int, n_tokens: int):
+    i = torch.arange(n_tokens, device=DEV).view(1, 1, n_tokens, 1)
+    j = torch.arange(n_tokens, device=DEV).view(1, 1, 1, n_tokens)
+    dist = (j - i).clamp_min(0)  # only penalize future
+    slopes = _alibi_slopes(n_heads)
+    return -slopes * dist
+
+
+# ───────────────────────── Model components ─────────────────────────
+class LowRankMHA(nn.Module):
+    """
+    Cache-aware MHA with low-rank projections; supports kv caching for decode.
+    """
+    def __init__(self, d: int, h: int, r: int, use_relpos: bool = True):
+        super().__init__()
+        assert d % h == 0, "d must be divisible by number of heads"
+        self.h, self.dk = h, d // h
+        self.use_relpos = use_relpos
+        self.q = nn.Linear(d, d, bias=False)
+        self.k = nn.Linear(d, d, bias=False)
+        self.v = nn.Linear(d, d, bias=False)
+        self.U = nn.Parameter(torch.randn(self.dk, r))
+        nn.init.orthogonal_(self.U)
+        self.proj = nn.Linear(h * r, d, bias=False)
+        self.drop = nn.Dropout(0.1)
+
+    def _proj(self, x):
+        B, N, _ = x.shape
+        return (x.view(B, N, self.h, self.dk).transpose(1, 2) @ self.U)
+
+    def forward(
+        self,
+        x: torch.Tensor,
+        mask: Optional[torch.Tensor] = None,
+        rel_bias_tokens: Optional[int] = None,
+        kv_cache: Optional[Tuple[torch.Tensor, torch.Tensor]] = None,
+        use_cache: bool = False,
+    ):
+        q = self._proj(self.q(x))
+        k_new = self._proj(self.k(x))
+        v_new = self._proj(self.v(x))
+
+        if kv_cache is None:
+            k, v = k_new, v_new
+        else:
+            k, v = kv_cache
+            if use_cache:
+                k = torch.cat([k, k_new], dim=2)
+                v = torch.cat([v, v_new], dim=2)
+
+        att = (q @ k.transpose(-1, -2)) / math.sqrt(self.dk)
+
+        if q.size(2) == k.size(2):
+            if self.use_relpos and rel_bias_tokens is not None:
+                att = att + alibi_bias(self.h, rel_bias_tokens)
+            if mask is not None:
+                att = att + mask
+
+        z = (att.softmax(-1) @ v).transpose(1, 2)  # (B,Nq,h,r)
+        z = z.reshape(x.size(0), x.size(1), -1)
+        out = self.drop(self.proj(z))
+        return (out, (k, v)) if use_cache else out
+
+
+class Block(nn.Module):
+    def __init__(self, d: int, h: int, r: int):
+        super().__init__()
+        self.ln1, self.ln2 = nn.LayerNorm(d), nn.LayerNorm(d)
+        self.mha = LowRankMHA(d, h, r, use_relpos=True)
+        self.ff = nn.Sequential(nn.Linear(d, 4 * d), nn.ReLU(), nn.Linear(4 * d, d))
+
+    def forward(
+        self,
+        x: torch.Tensor,
+        mask: Optional[torch.Tensor],
+        kv: Optional[Tuple[torch.Tensor, torch.Tensor]] = None,
+        use_cache: bool = False
+    ):
+        n = x.size(1)
+        if use_cache:
+            y, new_kv = self.mha(self.ln1(x), mask, rel_bias_tokens=n if mask is not None else None, kv_cache=kv, use_cache=True)
+            x = x + y
+            x = x + self.ff(self.ln2(x))
+            return x, new_kv
+        else:
+            x = x + self.mha(self.ln1(x), mask, rel_bias_tokens=n)
+            return x + self.ff(self.ln2(x))
+
+
+class Encoder(nn.Module):
+    """
+    Transformer encoder with optional kv caching (for AR/SAT decode).
+    """
+    def __init__(self, cfg: Dict[str, int]):
+        super().__init__()
+        d, l, h, r = cfg["d"], cfg["layers"], cfg["heads"], cfg["rank"]
+        self.emb = nn.Embedding(VOCAB, d)
+        self.blocks = nn.ModuleList(Block(d, h, r) for _ in range(l))
+        self.ln = nn.LayerNorm(d)
+
+    def forward(
+        self,
+        ids: torch.Tensor,
+        mask: Optional[torch.Tensor],
+        kv_caches: Optional[List[Optional[Tuple[torch.Tensor, torch.Tensor]]]] = None,
+        use_cache: bool = False
+    ):
+        x = self.emb(ids)
+        if not use_cache:
+            for blk in self.blocks:
+                x = blk(x, mask)
+            return self.ln(x)
+
+        new_kvs: List[Tuple[torch.Tensor, torch.Tensor]] = []
+        for i, blk in enumerate(self.blocks):
+            kv = kv_caches[i] if (kv_caches is not None) else None
+            x, kv_out = blk(x, mask, kv, use_cache=True)
+            new_kvs.append(kv_out)
+        return self.ln(x), new_kvs
+
+
+class ARHead(nn.Module):
+    def __init__(self, d):
+        super().__init__()
+        self.proj = nn.Linear(d, VOCAB)
+    def forward(self, h): return self.proj(h)
+
+
+class NATHead(nn.Module):
+    def __init__(self, d):
+        super().__init__()
+        self.proj = nn.Linear(d, VOCAB)
+    def forward(self, h): return self.proj(h)
+
+
+class SATHead(nn.Module):
+    def __init__(self, d, mode="var"):
+        super().__init__()
+        self.proj = nn.Linear(d, VOCAB)
+        self.mode = mode
+        self.gate = nn.Linear(d, 2) if mode == "var" else None
+    def forward(self, h_last):
+        logits = self.proj(h_last)
+        gate = self.gate(h_last[:, 0]) if self.gate is not None else None
+        return logits, gate
+
+
+# ───────────────────────── Masks ─────────────────────────
+def causal_mask(n):
+    m = torch.full((1, 1, n, n), float("-inf"), device=DEV)
+    return torch.triu(m, 1)
+
+def sat_mask(n, block=SAT_BLOCK):
+    idx = torch.arange(n, device=DEV)
+    grp = idx.unsqueeze(0) // block
+    allow = (grp.T == grp) | (grp.T > grp)
+    return torch.where(allow, 0.0, float("-inf")).unsqueeze(0).unsqueeze(0)
+
+
+# ───────────────────────── Checkpoint helpers ─────────────────────────
+def save_ckpt(
+    path: pathlib.Path,
+    core: nn.Module,
+    ar_h: nn.Module,
+    nat_h: nn.Module,
+    sat_h: nn.Module,
+    opt: torch.optim.Optimizer,
+    scaler: GradScaler,
+    meta: Dict[str, Any],
+):
+    path.parent.mkdir(exist_ok=True, parents=True)
+    tmp = path.with_suffix(path.suffix + ".tmp")
+    state = {
+        "core": core.state_dict(),
+        "ar": ar_h.state_dict(),
+        "nat": nat_h.state_dict(),
+        "sat": sat_h.state_dict(),
+        "opt": opt.state_dict(),
+        "scaler": scaler.state_dict(),
+        "cfg": meta.get("cfg"),
+        "tokenizer_id": TOKENIZER_ID,
+        **{k: v for k, v in meta.items() if k not in {"cfg"}},
+    }
+    torch.save(state, tmp, _use_new_zipfile_serialization=False)
+    tmp.replace(path)
+    (path.parent / "latest.json").write_text(json.dumps({"path": str(path), "step": meta["step"]}))
+    print(f"\n✓ saved checkpoint {path.name}")
+
+
+def load_ckpt(
+    path: pathlib.Path,
+    core: nn.Module,
+    ar_h: nn.Module,
+    nat_h: nn.Module,
+    sat_h: nn.Module,
+    opt: torch.optim.Optimizer,
+    scaler: GradScaler,
+):
+    p = _resolve_ckpt(path) or path
+    ck = _try_load(p, map_location=DEV)
+    if ck is None:
+        raise FileNotFoundError(f"No valid checkpoint at {p}")
+    core.load_state_dict(ck["core"])
+    ar_h.load_state_dict(ck["ar"])
+    nat_h.load_state_dict(ck["nat"])
+    sat_h.load_state_dict(ck["sat"])
+    opt.load_state_dict(ck["opt"])
+    scaler.load_state_dict(ck["scaler"])
+    return ck.get("step", 0), ck.get("seen_tok", 0), ck.get("wall_time", time.time())
+
+
+def _safe_load_any(path: pathlib.Path, tgt: nn.Module, key: str | None = None, rename: str | None = None):
+    p = _resolve_ckpt(path) or path
+    if not p.exists(): return 0
+    ck = _try_load(p, map_location=DEV)
+    if ck is None: return 0
+    sd = ck.get(key, ck) if key else ck
+    if isinstance(sd, dict) and "state_dict" in sd:
+        sd = sd["state_dict"]
+    if rename:
+        sd = {k.replace(rename, "proj."): v for k, v in sd.items() if rename in k}
+    tgt_sd = tgt.state_dict()
+    filt = {k: v for k, v in sd.items() if k in tgt_sd and v.shape == tgt_sd[k].shape}
+    if filt:
+        tgt.load_state_dict(filt, strict=False)
+    return len(filt)
+
+
+def infer_cfg_from_ckpt(path: pathlib.Path):
+    p = _resolve_ckpt(path) or path
+    if not p.exists(): return None
+    sd = _try_load(p, map_location="cpu")
+    if sd is None: return None
+    if isinstance(sd, dict) and "cfg" in sd and isinstance(sd["cfg"], dict):
+        return dict(sd["cfg"])
+    core = sd.get("core")
+    if core is None: return None
+    emb_w = core.get("emb.weight")
+    if emb_w is None: return None
+    d = emb_w.shape[1]
+    layer_ids = []
+    for k in core.keys():
+        if k.startswith("blocks."):
+            parts = k.split(".")
+            if len(parts) > 2 and parts[1].isdigit():
+                layer_ids.append(int(parts[1]))
+    layers = (max(layer_ids) + 1) if layer_ids else None
+    U = core.get("blocks.0.mha.U")
+    heads = rank = None
+    if U is not None:
+        dk, r = U.shape
+        rank = r
+        heads = d // dk if dk > 0 else None
+    out = {"d": d}
+    if layers is not None: out["layers"] = layers
+    if heads is not None:  out["heads"] = heads
+    if rank is not None:   out["rank"] = rank
+    return out
+
+
+# ───────────────────────── Train loop ─────────────────────────
+def _parse_grow_plan(s: str) -> List[int]:
+    steps = []
+    for part in s.split(","):
+        part = part.strip()
+        if part:
+            v = int(part)
+            if v >= 128:
+                steps.append(v)
+    return sorted(set(steps))
+
+
+def train(args):
+    cfg = PRESETS[args.preset].copy()
+
+    # Previous topology probe (unless --fresh)
+    if not args.fresh:
+        src_probe = pathlib.Path(args.warmstart_from) if args.warmstart_from else pathlib.Path(args.save_dir) / "final.pt"
+        prev_cfg = infer_cfg_from_ckpt(src_probe)
+    else:
+        prev_cfg = None
+
+    if prev_cfg:
+        cfg["d"] = prev_cfg.get("d", cfg["d"])
+        if prev_cfg.get("heads"):
+            cfg["heads"] = prev_cfg["heads"]
+        if args.rank is None and prev_cfg.get("rank"):
+            cfg["rank"] = prev_cfg["rank"]
+        # NEW: copy layers from ckpt even without --x2
+        if prev_cfg.get("layers"):
+            cfg["layers"] = prev_cfg["layers"]
+        # Optional doubling only when explicitly requested
+        if args.x2 and prev_cfg.get("layers"):
+            cfg["layers"] = max(cfg["layers"], prev_cfg["layers"] * 2)
+    if args.rank:
+        cfg["rank"] = args.rank
+    if args.x2 and not prev_cfg:
+        cfg["layers"] *= 2
+
+    BLOCK = args.block or DEFAULT_BLOCK
+
+    core = Encoder(cfg).to(DEV)
+    ar_h, nat_h = ARHead(cfg["d"]).to(DEV), NATHead(cfg["d"]).to(DEV)
+    sat_h = SATHead(cfg["d"], mode="var").to(DEV)
+
+    # Warm start unless --fresh
+    loaded = 0
+    if not args.fresh:
+        src = pathlib.Path(args.warmstart_from) if args.warmstart_from else pathlib.Path(args.save_dir) / "final.pt"
+        src = _resolve_ckpt(src)
+        if src:
+            loaded += _safe_load_any(src, core, key="core")
+            loaded += _safe_load_any(src, ar_h, key="ar")
+            loaded += _safe_load_any(src, nat_h, key="nat")
+            loaded += _safe_load_any(src, sat_h, key="sat")
+            if loaded:
+                print(f"Warm-start: loaded {loaded} matching tensors from {src}")
+
+    opt = torch.optim.AdamW(
+        [
+            {"params": core.parameters(), "lr": LR_CORE},
+            {"params": ar_h.parameters(), "lr": LR_HEAD},
+            {"params": nat_h.parameters(), "lr": LR_HEAD},
+            {"params": sat_h.parameters(), "lr": LR_HEAD},
+        ]
+    )
+    scaler = GradScaler(enabled=args.amp and DEV.type == "cuda")
+
+    ce_tok = nn.CrossEntropyLoss(label_smoothing=0.1)
+    ctc = nn.CTCLoss(blank=BLANK, zero_infinity=True)
+    ce_gate = nn.CrossEntropyLoss()
+
+    # ---------- resume bookkeeping ----------
+    start_step, seen_tok = 0, 0
+    last_save_time = time.time()
+    if args.resume and not args.fresh:
+        start_step, seen_tok, last_save_time = load_ckpt(
+            pathlib.Path(args.resume), core, ar_h, nat_h, sat_h, opt, scaler
+        )
+        print(f"✓ resumed from step {start_step:,}, seen_tokens={seen_tok:,}")
+
+    # Target tokens
+    if args.target_tokens:
+        target_tokens = args.target_tokens
+    else:
+        param_count = sum(p.numel() for p in core.parameters())
+        target_tokens = int(25 * param_count)
+
+    new_tokens_needed = target_tokens - seen_tok
+    if new_tokens_needed <= 0:
+        print("Target already reached – nothing to train.")
+        return
+    new_steps = new_tokens_needed // BLOCK
+    if args.steps:
+        new_steps = min(new_steps, args.steps)
+        new_tokens_needed = new_steps * BLOCK
+
+    total_tokens_needed = seen_tok + new_tokens_needed
+    print(f"[auto-steps] {new_steps:,} training steps (@ {BLOCK} tokens/step)")
+
+    # Progressive growth plan
+    grow_plan = _parse_grow_plan(args.grow_plan) if args.auto_grow else []
+    if args.auto_grow:
+        if BLOCK not in grow_plan:
+            grow_plan = sorted(set(grow_plan + [BLOCK]))
+        print(f"[auto-grow] plan: {grow_plan} every {args.grow_every_steps} steps")
+
+    stream = token_stream(args.source, target_tokens, seed=42)
+    buf: list[int] = []
+    pbar = tqdm(total=total_tokens_needed, initial=seen_tok, unit="tok")
+    step = start_step
+    steps_since_last_grow = 0
+
+    while seen_tok < total_tokens_needed:
+        # ------- assemble one batch -------
+        try:
+            while len(buf) < BLOCK:
+                buf.append(next(stream))
+        except StopIteration:
+            break
+        ids = torch.tensor(buf[:BLOCK], device=DEV).unsqueeze(0)  # (B=1, N)
+        buf = buf[BLOCK:]
+
+        tgt_ar = ids.clone()                           # (1, N)
+        ids_nat = torch.repeat_interleave(ids, 2, 1)   # (1, 2N) for NAT only
+
+        try:
+            with amp(args.amp):
+                # AR path
+                h_ar = core(ids, causal_mask(ids.size(1)))
+                logits_ar = ar_h(h_ar)[:, :-1]
+                loss_ar = ce_tok(logits_ar.reshape(-1, VOCAB), tgt_ar[:, 1:].reshape(-1))
+
+                # NAT path (uses doubled sequence)
+                h_nat = core(ids_nat, None)
+                log_nat = nat_h(h_nat).log_softmax(-1).transpose(0, 1)  # (T,B,V)
+                ilen = tlen = torch.tensor([ids_nat.size(1) // 2], device=DEV)
+                loss_nat = ctc(log_nat, tgt_ar, ilen, tlen)
+
+                # SAT path
+                h_sat = core(ids, sat_mask(ids.size(1)))
+                logits_sat, gate = sat_h(h_sat[:, -SAT_BLOCK:])
+                tgt_sat = ids[:, 1:SAT_BLOCK+1]
+                loss_sat = ce_tok(logits_sat.reshape(-1, VOCAB), tgt_sat.reshape(-1))
+                if gate is not None:
+                    loss_sat += EMIT_LAMBDA * ce_gate(gate, torch.ones(ids.size(0), device=DEV, dtype=torch.long))
+
+                loss = loss_ar + loss_nat + loss_sat
+
+            # optimisation
+            scaler.scale(loss).backward()
+            scaler.unscale_(opt)
+            nn.utils.clip_grad_norm_(core.parameters(), 1.0)
+            scaler.step(opt)
+            scaler.update()
+            opt.zero_grad(set_to_none=True)
+
+        except RuntimeError as e:
+            msg = str(e).lower()
+            if "out of memory" in msg or "cuda error" in msg:
+                new_block = max(128, BLOCK // 2)
+                if new_block < BLOCK:
+                    print(f"\n[OOM] reducing block from {BLOCK} -> {new_block}")
+                    BLOCK = new_block
+                    if DEV.type == "cuda":
+                        torch.cuda.empty_cache()
+                    buf = ids[0].tolist() + buf
+                    steps_since_last_grow = 0
+                    continue
+            raise
+
+        # progress
+        step += 1
+        seen_tok += BLOCK
+        pbar.update(BLOCK)
+        pbar.set_postfix(loss=f"{loss.item():.3f}", block=BLOCK)
+
+        # checkpoint cadence
+        now = time.time()
+        time_due = (now - last_save_time) >= args.save_every_sec > 0
+        step_due = args.save_every_steps > 0 and step % args.save_every_steps == 0
+        if time_due or step_due:
+            ck_name = f"step{step:08d}.pt"
+            save_ckpt(
+                pathlib.Path(args.save_dir) / ck_name,
+                core, ar_h, nat_h, sat_h, opt, scaler,
+                meta={
+                    "cfg": cfg,
+                    "step": step,
+                    "seen_tok": seen_tok,
+                    "wall_time": now,
+                    "py_state": random.getstate(),
+                    "torch_state": rng_state(),
+                },
+            )
+            last_save_time = now
+
+        # progressive growth
+        if args.auto_grow:
+            steps_since_last_grow += 1
+            if steps_since_last_grow >= args.grow_every_steps:
+                steps_since_last_grow = 0
+                try:
+                    idx = grow_plan.index(BLOCK)
+                    if idx + 1 < len(grow_plan):
+                        candidate = grow_plan[idx + 1]
+                        print(f"[auto-grow] attempting BLOCK {BLOCK} -> {candidate}")
+                        BLOCK = candidate
+                        if DEV.type == "cuda":
+                            torch.cuda.empty_cache()
+                    else:
+                        print("[auto-grow] at max planned block; no further growth.")
+                except ValueError:
+                    grow_plan = sorted(set(grow_plan + [BLOCK]))
+                    idx = grow_plan.index(BLOCK)
+                    if idx + 1 < len(grow_plan):
+                        candidate = grow_plan[idx + 1]
+                        print(f"[auto-grow] moving to planned BLOCK {candidate}")
+                        BLOCK = candidate
+                        if DEV.type == "cuda":
+                            torch.cuda.empty_cache()
+
+    pbar.close()
+
+    # final save
+    save_ckpt(
+        pathlib.Path(args.save_dir) / "final.pt",
+        core, ar_h, nat_h, sat_h, opt, scaler,
+        meta={
+            "cfg": cfg,
+            "step": step,
+            "seen_tok": seen_tok,
+            "wall_time": time.time(),
+            "py_state": random.getstate(),
+            "torch_state": rng_state(),
+        },
+    )
+    print("🎉 training complete")
+
+
+# ───────────────────────── Repetition penalty helper ─────────────────────────
+@torch.no_grad()
+def apply_repetition_penalty_(logits: torch.Tensor,
+                              recent_ids: deque | List[int],
+                              penalty: float,
+                              window: int) -> None:
+    """
+    In-place adjustment of logits using HF-style repetition penalty.
+    Penalizes tokens that appeared in the last `window` ids.
+    Works for shapes (V,) or (1,V). We assume batch=1 for decode.
+    """
+    if penalty is None or penalty <= 1.0 or window <= 0 or not recent_ids:
+        return
+    # View final-dim as a 1D vector
+    lview = logits
+    while lview.dim() > 1:
+        lview = lview[0]
+    tail = list(recent_ids)[-window:]
+    if not tail:
+        return
+    u, cnt = torch.unique(torch.tensor(tail, device=lview.device, dtype=torch.long), return_counts=True)
+    powv = (torch.ones_like(cnt, dtype=lview.dtype) * penalty).pow(cnt.to(lview.dtype))
+    sel = lview.index_select(0, u)
+    sel = torch.where(sel > 0, sel / powv, sel * powv)
+    lview.index_copy_(0, u, sel)
+
+
+# ───────────────────────── Inference helpers ─────────────────────────
+def _sync_tokenizer_for_checkpoint(sd: dict):
+    global tok, TOKENIZER_ID, VOCAB, BLANK, EOS
+    ck_tok = sd.get("tokenizer_id")
+    if isinstance(ck_tok, str) and ck_tok and ck_tok != TOKENIZER_ID:
+        TOKENIZER_ID = ck_tok
+        tok = AutoTokenizer.from_pretrained(TOKENIZER_ID, use_fast=True, trust_remote_code=True)
+        if tok.pad_token is None:
+            tok.add_special_tokens({"pad_token": "[PAD]"})
+        VOCAB = max(tok.get_vocab().values()) + 1
+        BLANK = tok.pad_token_id
+        EOS = tok.eos_token_id if tok.eos_token_id is not None else tok.sep_token_id
+
+
+def load_joint(ckpt: str, preset: str):
+    path = _resolve_ckpt(pathlib.Path(ckpt)) or pathlib.Path(ckpt)
+    # map to CPU to avoid accidental GPU OOM during load
+    sd = _try_load(path, map_location="cpu")
+    if sd is None:
+        raise FileNotFoundError(f"No valid checkpoint at {path}")
+    _sync_tokenizer_for_checkpoint(sd)  # update tokenizer & vocab if needed
+    cfg = sd["cfg"] if "cfg" in sd and isinstance(sd["cfg"], dict) else PRESETS[preset]
+    core = Encoder(cfg).to(DEV)
+    ar_h, nat_h = ARHead(cfg["d"]).to(DEV), NATHead(cfg["d"]).to(DEV)
+    sat_h = SATHead(cfg["d"]).to(DEV)
+    core.load_state_dict(sd["core"])
+    ar_h.load_state_dict(sd["ar"])
+    nat_h.load_state_dict(sd["nat"])
+    sat_h.load_state_dict(sd["sat"])
+    return core, ar_h, nat_h, sat_h
+
+
+@torch.no_grad()
+def ar_decode(core, ar_h, prompt: str, max_new: int, T: float,
+              repetition_penalty: float = 1.0, rep_window: int = 256):
+    ids = torch.tensor([tok.encode(prompt)], device=DEV)
+    if ids.size(1) == 0:
+        ids = torch.tensor([[EOS] if EOS is not None else [0]], device=DEV)
+    # cache
+    h_full, kvs = core(ids, causal_mask(ids.size(1)), use_cache=True)
+    logits = ar_h(h_full)[:, -1]
+    recent = deque(ids[0].tolist(), maxlen=max(1, rep_window))
+    apply_repetition_penalty_(logits, recent, repetition_penalty, rep_window)
+
+    if T <= 1e-6:
+        nxt = torch.argmax(logits, dim=-1, keepdim=True)
+    else:
+        nxt = torch.softmax(logits / T, -1).multinomial(1)
+
+    ids = torch.cat([ids, nxt], 1)
+    recent.append(nxt.item())
+    start = time.time()
+    for _ in range(max(0, max_new - 1)):
+        x = ids[:, -1:]
+        h_step, kvs = core(x, None, kv_caches=kvs, use_cache=True)
+        logits = ar_h(h_step)[:, -1]
+        apply_repetition_penalty_(logits, recent, repetition_penalty, rep_window)
+
+        if T <= 1e-6:
+            nxt = torch.argmax(logits, dim=-1, keepdim=True)
+        else:
+            nxt = torch.softmax(logits / T, -1).multinomial(1)
+
+        ids = torch.cat([ids, nxt], 1)
+        recent.append(nxt.item())
+    print(tok.decode(ids[0].tolist(), skip_special_tokens=True))
+    print(f"[{max_new} tok in {time.time() - start:.2f}s]")
+
+
+@torch.no_grad()
+def sat_decode(core, sat_h, prompt, max_new, T, var,
+               repetition_penalty: float = 1.0, rep_window: int = 256):
+    ids = torch.tensor([tok.encode(prompt)], device=DEV)
+    recent = deque(ids[0].tolist(), maxlen=max(1, rep_window))
+    added, t0 = 0, time.time()
+    while added < max_new:
+        h = core(ids, sat_mask(ids.size(1)))
+        logits, gate = sat_h(h[:, -SAT_BLOCK:])
+        # stride selection
+        if not var or gate is None:
+            stride = SAT_BLOCK
+        else:
+            gprob = torch.softmax(gate, -1)
+            # sample 0 or 1, add 1 to get 1 or 2 tokens
+            stride = int(torch.multinomial(gprob, 1).item() + 1)
+
+        # logits shape (1, SAT_BLOCK, V); apply penalty per position
+        logits = logits[:, :stride, :]
+        for s in range(logits.size(1)):
+            apply_repetition_penalty_(logits[:, s, :], recent, repetition_penalty, rep_window)
+
+        if T <= 1e-6:
+            nxt = torch.argmax(logits, dim=-1)           # (1, stride)
+        else:
+            probs = torch.softmax(logits / T, -1)        # (1, stride, V)
+            flat = probs.view(-1, probs.size(-1))        # (stride, V) since B=1
+            picks = torch.multinomial(flat, 1).view(1, -1)
+            nxt = picks                                   # (1, stride)
+
+        tok_ids = nxt.squeeze(0).tolist()
+        for tid in tok_ids:
+            ids = torch.cat([ids, torch.tensor([[tid]], device=ids.device)], 1)
+            recent.append(tid)
+            added += 1
+            if added >= max_new:
+                break
+    print(tok.decode(ids[0].tolist(), skip_special_tokens=True))
+    print(f"[{added} tok in {time.time() - t0:.2f}s]")
+
+
+@torch.no_grad()
+def nat_decode(core, nat_h, prompt, max_new, passes, streams):
+    ids = torch.tensor([tok.encode(prompt) + [BLANK] * (max_new * 2)], device=DEV)
+    t0 = time.time()
+    for _ in range(passes):
+        h = core(ids, None)
+        logits = nat_h(h)
+        logits[..., BLANK] = -1e9
+        cand = logits.topk(streams, -1).indices.permute(2, 0, 1)
+        best = (cand != BLANK).float().mean(-1).argmax(0)
+        ids = cand[best, torch.arange(ids.size(0), device=DEV)][:, ::2]
+    out = [t for t in ids[0].tolist() if t != BLANK]
+    print(tok.decode(out, skip_special_tokens=True))
+    print(f"[{len(out)} output tokens in {time.time() - t0:.2f}s]")
+
+
+# ───────────────────────── CLI ─────────────────────────
+def main():
+    ap = argparse.ArgumentParser()
+    sub = ap.add_subparsers(dest="cmd", required=True)
+
+    tr = sub.add_parser("train")
+    tr.add_argument("--preset", choices=PRESETS, default="small")
+    tr.add_argument("--rank", type=int)
+    tr.add_argument("--block", type=int, default=DEFAULT_BLOCK)
+    tr.add_argument("--source", default="cerebras/SlimPajama-627B")
+    tr.add_argument("--target_tokens", type=int)
+    tr.add_argument("--steps", type=int)
+    tr.add_argument("--amp", action="store_true")
+    tr.add_argument("--save_every_sec", type=int, default=DEFAULT_SAVE_SEC)
+    tr.add_argument("--save_every_steps", type=int, default=0)
+    tr.add_argument("--save_dir", default=str(CKDIR))
+    tr.add_argument("--resume", type=str)
+    tr.add_argument("--x2", action="store_true", help="~2x params by doubling layers")
+    tr.add_argument("--warmstart_from", type=str, default=None, help="Path to previous final.pt for shape-safe warm start")
+    tr.add_argument("--fresh", action="store_true", help="Start from scratch: do not probe or load any checkpoints")
+
+    # Progressive block growth
+    tr.add_argument("--auto_grow", action="store_true", help="Automatically grow block size over time")
+    tr.add_argument("--grow_plan", type=str, default="576,640,768,896,1024", help="Comma list of block sizes to try in order")
+    tr.add_argument("--grow_every_steps", type=int, default=50000, help="Steps between growth attempts")
+
+    inf = sub.add_parser("infer")
+    inf.add_argument("--mode", choices=["ar", "nat", "sat"], required=True)
+    inf.add_argument("--ckpt", required=True)
+    inf.add_argument("--preset", default="small")
+    inf.add_argument("--prompt", required=True)
+    inf.add_argument("--max_new", type=int, default=120)
+    inf.add_argument("--temperature", type=float, default=1.0)
+    inf.add_argument("--var", action="store_true")
+    inf.add_argument("--passes", type=int, default=1)
+    inf.add_argument("--streams", type=int, default=5)
+    # repetition penalty knobs
+    inf.add_argument("--repetition_penalty", type=float, default=1.0,
+                     help=">1.0 discourages repeating recently emitted tokens (HF-style; default off)")
+    inf.add_argument("--rep_window", type=int, default=256,
+                     help="Number of most-recent tokens to penalize (default 256)")
+
+    args = ap.parse_args()
+    if args.cmd == "train":
+        train(args)
+    else:
+        core, ar_h, nat_h, sat_h = load_joint(args.ckpt, args.preset)
+        if args.mode == "ar":
+            ar_decode(core, ar_h, args.prompt, args.max_new, args.temperature,
+                      repetition_penalty=args.repetition_penalty, rep_window=args.rep_window)
+        elif args.mode == "sat":
+            sat_decode(core, sat_h, args.prompt, args.max_new, args.temperature, args.var,
+                       repetition_penalty=args.repetition_penalty, rep_window=args.rep_window)
+        else:
+            nat_decode(core, nat_h, args.prompt, args.max_new, args.passes, args.streams)
+
+
+if __name__ == "__main__":
+    main()

5ap.py

@@ -0,0 +1,949 @@
+#!/usr/bin/env python3
+# 5L.py — joint AR+NAT+SAT trainer/decoder (Qwen3 tokenizer)
+# Robust fresh-start, ignores *.pt.tmp, AMP dtype auto, OOM backoff, progressive block growth.
+# Added: repetition/presence/frequency penalties, top-k/top-p/min-p, greedy, no-repeat-ngrams.
+# Fixes: SAT multinomial shape; checkpoint loads on CPU; cfg fallback if ckpt missing cfg.
+
+from __future__ import annotations
+import argparse, json, math, pathlib, random, time, os
+from contextlib import nullcontext
+from typing import Dict, Any, List, Optional, Tuple
+
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+from datasets import load_dataset
+from transformers import AutoTokenizer, logging as hf_log
+from tqdm.auto import tqdm
+
+# ───────────────────────── Globals ─────────────────────────
+hf_log.set_verbosity_error()
+DEV = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+torch.backends.cuda.matmul.allow_tf32 = True
+try:
+    torch.set_float32_matmul_precision("high")
+except Exception:
+    pass
+
+# Use the Qwen3 tokenizer (can override with env TOKENIZER_ID if needed)
+TOKENIZER_ID = os.environ.get(
+    "TOKENIZER_ID",
+    "Qwen/Qwen3-235B-A22B-Thinking-2507"
+)
+
+# Some Qwen tokenizers require trust_remote_code
+tok = AutoTokenizer.from_pretrained(TOKENIZER_ID, use_fast=True, trust_remote_code=True)
+if tok.pad_token is None:
+    tok.add_special_tokens({"pad_token": "[PAD]"})
+VOCAB, BLANK, EOS = (
+    max(tok.get_vocab().values()) + 1,   # allow new [PAD] if appended
+    tok.pad_token_id,
+    tok.eos_token_id if tok.eos_token_id is not None else tok.sep_token_id
+)
+
+PRESETS: Dict[str, Dict[str, int]] = {
+    "small":   dict(d=512, layers=8,  heads=16, rank=64),
+    "smallx2": dict(d=512, layers=16, heads=16, rank=64),
+    "base":    dict(d=768, layers=12, heads=24, rank=96),
+}
+
+# Safe default for 1× Tesla P40; override with --block
+DEFAULT_BLOCK = 576
+SAT_BLOCK = 2
+LR_CORE, LR_HEAD = 5e-5, 2e-4
+EMIT_LAMBDA = 0.1
+DEFAULT_SAVE_SEC = 8 * 24 * 3600  # 8 days
+CKDIR = pathlib.Path("ckpts_joint")
+
+
+# ───────────────────────── Utilities ─────────────────────────
+def rng_state():
+    if DEV.type == "cuda":
+        try:
+            return torch.cuda.get_rng_state(DEV)
+        except TypeError:
+            return torch.cuda.get_rng_state()
+    return torch.get_rng_state()
+
+
+def _is_probably_ckpt(path: pathlib.Path) -> bool:
+    try:
+        return path.is_file() and path.suffix == ".pt" and not path.name.endswith(".pt.tmp") and path.stat().st_size > (1<<20)
+    except Exception:
+        return False
+
+
+def _resolve_ckpt(path: pathlib.Path) -> pathlib.Path | None:
+    """
+    Return a solid .pt (never .tmp). If 'path' is dir, pick newest *.pt.
+    If not usable, return None.
+    """
+    try:
+        if path.is_dir():
+            cands = sorted([p for p in path.glob("*.pt") if _is_probably_ckpt(p)],
+                           key=lambda p: p.stat().st_mtime, reverse=True)
+            return cands[0] if cands else None
+        if path.suffix == ".tmp":
+            solid = path.with_suffix("")
+            return solid if _is_probably_ckpt(solid) else _resolve_ckpt(path.parent)
+        return path if _is_probably_ckpt(path) else _resolve_ckpt(path.parent)
+    except Exception:
+        return None
+
+
+def _try_load(path: pathlib.Path, map_location="cpu"):
+    """
+    Always load on CPU to avoid CUDA fragmentation/OOM during torch.load.
+    """
+    try:
+        return torch.load(path, map_location="cpu")
+    except Exception as e:
+        print(f"[ckpt-skip] {path} not usable: {e}")
+        return None
+
+
+# ───────────────────────── AMP helper ─────────────────────────
+try:
+    from torch.amp import autocast as _ac, GradScaler
+except ImportError:
+    from torch.cuda.amp import autocast as _ac, GradScaler
+
+def _auto_amp_dtype():
+    if DEV.type == "cuda":
+        try:
+            if torch.cuda.is_bf16_supported():
+                return torch.bfloat16
+            return torch.float16
+        except Exception:
+            return torch.float16
+    return torch.float32
+
+def amp(enabled):
+    return nullcontext() if not enabled else _ac(device_type="cuda", dtype=_auto_amp_dtype())
+
+
+# ───────────────────────── Data stream ─────────────────────────
+def token_stream(ds_name: str, target: int, seed: int = 42):
+    ds = load_dataset(ds_name, split="train", streaming=True)
+    ds = ds.shuffle(buffer_size=10_000, seed=seed)
+    emitted = 0
+    for ex in ds:
+        # ensure EOS between docs
+        enc = tok.encode(ex["text"])
+        if EOS is not None and (len(enc) == 0 or enc[-1] != EOS):
+            enc = enc + [EOS]
+        for t in enc:
+            yield t
+            emitted += 1
+            if emitted >= target:
+                return
+
+
+# ───────────────────────── Relative positional bias (ALiBi) ─────────────────────────
+def _alibi_slopes(n_heads: int):
+    import math
+    def pow2slopes(n):
+        start = 2 ** (-2 ** -(math.log2(n) - 3))
+        ratio = start
+        return [start * (ratio ** i) for i in range(n)]
+    if math.log2(n_heads).is_integer():
+        vals = pow2slopes(n_heads)
+    else:
+        closest = 2 ** math.floor(math.log2(n_heads))
+        vals = pow2slopes(closest)
+        extra = pow2slopes(2 * closest)
+        vals += extra[0::2][: n_heads - closest]
+    return torch.tensor(vals, device=DEV).view(1, n_heads, 1, 1)
+
+def alibi_bias(n_heads: int, n_tokens: int):
+    i = torch.arange(n_tokens, device=DEV).view(1, 1, n_tokens, 1)
+    j = torch.arange(n_tokens, device=DEV).view(1, 1, 1, n_tokens)
+    dist = (j - i).clamp_min(0)  # only penalize future
+    slopes = _alibi_slopes(n_heads)
+    return -slopes * dist
+
+
+# ───────────────────────── Model components ─────────────────────────
+class LowRankMHA(nn.Module):
+    """
+    Cache-aware MHA with low-rank projections; supports kv caching for decode.
+    """
+    def __init__(self, d: int, h: int, r: int, use_relpos: bool = True):
+        super().__init__()
+        assert d % h == 0, "d must be divisible by number of heads"
+        self.h, self.dk = h, d // h
+        self.use_relpos = use_relpos
+        self.q = nn.Linear(d, d, bias=False)
+        self.k = nn.Linear(d, d, bias=False)
+        self.v = nn.Linear(d, d, bias=False)
+        self.U = nn.Parameter(torch.randn(self.dk, r))
+        nn.init.orthogonal_(self.U)
+        self.proj = nn.Linear(h * r, d, bias=False)
+        self.drop = nn.Dropout(0.1)
+
+    def _proj(self, x):
+        B, N, _ = x.shape
+        return (x.view(B, N, self.h, self.dk).transpose(1, 2) @ self.U)
+
+    def forward(
+        self,
+        x: torch.Tensor,
+        mask: Optional[torch.Tensor] = None,
+        rel_bias_tokens: Optional[int] = None,
+        kv_cache: Optional[Tuple[torch.Tensor, torch.Tensor]] = None,
+        use_cache: bool = False,
+    ):
+        q = self._proj(self.q(x))
+        k_new = self._proj(self.k(x))
+        v_new = self._proj(self.v(x))
+
+        if kv_cache is None:
+            k, v = k_new, v_new
+        else:
+            k, v = kv_cache
+            if use_cache:
+                k = torch.cat([k, k_new], dim=2)
+                v = torch.cat([v, v_new], dim=2)
+
+        att = (q @ k.transpose(-1, -2)) / math.sqrt(self.dk)
+
+        if q.size(2) == k.size(2):
+            if self.use_relpos and rel_bias_tokens is not None:
+                att = att + alibi_bias(self.h, rel_bias_tokens)
+            if mask is not None:
+                att = att + mask
+
+        z = (att.softmax(-1) @ v).transpose(1, 2)  # (B,Nq,h,r)
+        z = z.reshape(x.size(0), x.size(1), -1)
+        out = self.drop(self.proj(z))
+        return (out, (k, v)) if use_cache else out
+
+
+class Block(nn.Module):
+    def __init__(self, d: int, h: int, r: int):
+        super().__init__()
+        self.ln1, self.ln2 = nn.LayerNorm(d), nn.LayerNorm(d)
+        self.mha = LowRankMHA(d, h, r, use_relpos=True)
+        self.ff = nn.Sequential(nn.Linear(d, 4 * d), nn.ReLU(), nn.Linear(4 * d, d))
+
+    def forward(
+        self,
+        x: torch.Tensor,
+        mask: Optional[torch.Tensor],
+        kv: Optional[Tuple[torch.Tensor, torch.Tensor]] = None,
+        use_cache: bool = False
+    ):
+        n = x.size(1)
+        if use_cache:
+            y, new_kv = self.mha(self.ln1(x), mask, rel_bias_tokens=n if mask is not None else None, kv_cache=kv, use_cache=True)
+            x = x + y
+            x = x + self.ff(self.ln2(x))
+            return x, new_kv
+        else:
+            x = x + self.mha(self.ln1(x), mask, rel_bias_tokens=n)
+            return x + self.ff(self.ln2(x))
+
+
+class Encoder(nn.Module):
+    """
+    Transformer encoder with optional kv caching (for AR/SAT decode).
+    """
+    def __init__(self, cfg: Dict[str, int]):
+        super().__init__()
+        d, l, h, r = cfg["d"], cfg["layers"], cfg["heads"], cfg["rank"]
+        self.emb = nn.Embedding(VOCAB, d)
+        self.blocks = nn.ModuleList(Block(d, h, r) for _ in range(l))
+        self.ln = nn.LayerNorm(d)
+
+    def forward(
+        self,
+        ids: torch.Tensor,
+        mask: Optional[torch.Tensor],
+        kv_caches: Optional[List[Optional[Tuple[torch.Tensor, torch.Tensor]]]] = None,
+        use_cache: bool = False
+    ):
+        x = self.emb(ids)
+        if not use_cache:
+            for blk in self.blocks:
+                x = blk(x, mask)
+            return self.ln(x)
+
+        new_kvs: List[Tuple[torch.Tensor, torch.Tensor]] = []
+        for i, blk in enumerate(self.blocks):
+            kv = kv_caches[i] if (kv_caches is not None) else None
+            x, kv_out = blk(x, mask, kv, use_cache=True)
+            new_kvs.append(kv_out)
+        return self.ln(x), new_kvs
+
+
+class ARHead(nn.Module):
+    def __init__(self, d):
+        super().__init__()
+        self.proj = nn.Linear(d, VOCAB)
+    def forward(self, h): return self.proj(h)
+
+
+class NATHead(nn.Module):
+    def __init__(self, d):
+        super().__init__()
+        self.proj = nn.Linear(d, VOCAB)
+    def forward(self, h): return self.proj(h)
+
+
+class SATHead(nn.Module):
+    def __init__(self, d, mode="var"):
+        super().__init__()
+        self.proj = nn.Linear(d, VOCAB)
+        self.mode = mode
+        self.gate = nn.Linear(d, 2) if mode == "var" else None
+    def forward(self, h_last):
+        logits = self.proj(h_last)
+        gate = self.gate(h_last[:, 0]) if self.gate is not None else None
+        return logits, gate
+
+
+# ───────────────────────── Masks ─────────────────────────
+def causal_mask(n):
+    m = torch.full((1, 1, n, n), float("-inf"), device=DEV)
+    return torch.triu(m, 1)
+
+def sat_mask(n, block=SAT_BLOCK):
+    idx = torch.arange(n, device=DEV)
+    grp = idx.unsqueeze(0) // block
+    allow = (grp.T == grp) | (grp.T > grp)
+    return torch.where(allow, 0.0, float("-inf")).unsqueeze(0).unsqueeze(0)
+
+
+# ───────────────────────── Checkpoint helpers ─────────────────────────
+def save_ckpt(
+    path: pathlib.Path,
+    core: nn.Module,
+    ar_h: nn.Module,
+    nat_h: nn.Module,
+    sat_h: nn.Module,
+    opt: torch.optim.Optimizer,
+    scaler: GradScaler,
+    meta: Dict[str, Any],
+):
+    path.parent.mkdir(exist_ok=True, parents=True)
+    tmp = path.with_suffix(path.suffix + ".tmp")
+    state = {
+        "core": core.state_dict(),
+        "ar": ar_h.state_dict(),
+        "nat": nat_h.state_dict(),
+        "sat": sat_h.state_dict(),
+        "opt": opt.state_dict(),
+        "scaler": scaler.state_dict(),
+        "cfg": meta.get("cfg"),
+        "tokenizer_id": TOKENIZER_ID,
+        **{k: v for k, v in meta.items() if k not in {"cfg"}},
+    }
+    torch.save(state, tmp, _use_new_zipfile_serialization=False)
+    tmp.replace(path)
+    (path.parent / "latest.json").write_text(json.dumps({"path": str(path), "step": meta["step"]}))
+    print(f"\n✓ saved checkpoint {path.name}")
+
+
+def load_ckpt(
+    path: pathlib.Path,
+    core: nn.Module,
+    ar_h: nn.Module,
+    nat_h: nn.Module,
+    sat_h: nn.Module,
+    opt: torch.optim.Optimizer,
+    scaler: GradScaler,
+):
+    p = _resolve_ckpt(path) or path
+    ck = _try_load(p, map_location="cpu")
+    if ck is None:
+        raise FileNotFoundError(f"No valid checkpoint at {p}")
+    core.load_state_dict(ck["core"])
+    ar_h.load_state_dict(ck["ar"])
+    nat_h.load_state_dict(ck["nat"])
+    sat_h.load_state_dict(ck["sat"])
+    opt.load_state_dict(ck["opt"])
+    scaler.load_state_dict(ck["scaler"])
+    return ck.get("step", 0), ck.get("seen_tok", 0), ck.get("wall_time", time.time())
+
+
+def _safe_load_any(path: pathlib.Path, tgt: nn.Module, key: str | None = None, rename: str | None = None):
+    p = _resolve_ckpt(path) or path
+    if not p.exists(): return 0
+    ck = _try_load(p, map_location="cpu")
+    if ck is None: return 0
+    sd = ck.get(key, ck) if key else ck
+    if isinstance(sd, dict) and "state_dict" in sd:
+        sd = sd["state_dict"]
+    if rename:
+        sd = {k.replace(rename, "proj."): v for k, v in sd.items() if rename in k}
+    tgt_sd = tgt.state_dict()
+    filt = {k: v for k, v in sd.items() if k in tgt_sd and v.shape == tgt_sd[k].shape}
+    if filt:
+        tgt.load_state_dict(filt, strict=False)
+    return len(filt)
+
+
+def infer_cfg_from_ckpt(path: pathlib.Path):
+    p = _resolve_ckpt(path) or path
+    if not p.exists(): return None
+    sd = _try_load(p, map_location="cpu")
+    if sd is None: return None
+    if isinstance(sd, dict) and "cfg" in sd and isinstance(sd["cfg"], dict):
+        return dict(sd["cfg"])
+    core = sd.get("core")
+    if core is None: return None
+    emb_w = core.get("emb.weight")
+    if emb_w is None: return None
+    d = emb_w.shape[1]
+    layer_ids = []
+    for k in core.keys():
+        if k.startswith("blocks."):
+            parts = k.split(".")
+            if len(parts) > 2 and parts[1].isdigit():
+                layer_ids.append(int(parts[1]))
+    layers = (max(layer_ids) + 1) if layer_ids else None
+    U = core.get("blocks.0.mha.U")
+    heads = rank = None
+    if U is not None:
+        dk, r = U.shape
+        rank = r
+        heads = d // dk if dk > 0 else None
+    out = {"d": d}
+    if layers is not None: out["layers"] = layers
+    if heads is not None:  out["heads"] = heads
+    if rank is not None:   out["rank"] = rank
+    return out
+
+
+# ───────────────────────── Train loop ─────────────────────────
+def _parse_grow_plan(s: str) -> List[int]:
+    steps = []
+    for part in s.split(","):
+        part = part.strip()
+        if part:
+            v = int(part)
+            if v >= 128:
+                steps.append(v)
+    return sorted(set(steps))
+
+
+def train(args):
+    cfg = PRESETS[args.preset].copy()
+
+    # Previous topology probe (unless --fresh)
+    if not args.fresh:
+        src_probe = pathlib.Path(args.warmstart_from) if args.warmstart_from else pathlib.Path(args.save_dir) / "final.pt"
+        prev_cfg = infer_cfg_from_ckpt(src_probe)
+    else:
+        prev_cfg = None
+
+    if prev_cfg:
+        cfg["d"] = prev_cfg.get("d", cfg["d"])
+        if prev_cfg.get("heads"):
+            cfg["heads"] = prev_cfg["heads"]
+        if args.rank is None and prev_cfg.get("rank"):
+            cfg["rank"] = prev_cfg["rank"]
+        if prev_cfg.get("layers"):
+            cfg["layers"] = prev_cfg["layers"]
+        if args.x2 and prev_cfg.get("layers"):
+            cfg["layers"] = max(cfg["layers"], prev_cfg["layers"] * 2)
+    if args.rank:
+        cfg["rank"] = args.rank
+    if args.x2 and not prev_cfg:
+        cfg["layers"] *= 2
+
+    BLOCK = args.block or DEFAULT_BLOCK
+
+    core = Encoder(cfg).to(DEV)
+    ar_h, nat_h = ARHead(cfg["d"]).to(DEV), NATHead(cfg["d"]).to(DEV)
+    sat_h = SATHead(cfg["d"], mode="var").to(DEV)
+
+    # Warm start unless --fresh
+    loaded = 0
+    if not args.fresh:
+        src = pathlib.Path(args.warmstart_from) if args.warmstart_from else pathlib.Path(args.save_dir) / "final.pt"
+        src = _resolve_ckpt(src)
+        if src:
+            loaded += _safe_load_any(src, core, key="core")
+            loaded += _safe_load_any(src, ar_h, key="ar")
+            loaded += _safe_load_any(src, nat_h, key="nat")
+            loaded += _safe_load_any(src, sat_h, key="sat")
+            if loaded:
+                print(f"Warm-start: loaded {loaded} matching tensors from {src}")
+
+    opt = torch.optim.AdamW(
+        [
+            {"params": core.parameters(), "lr": LR_CORE},
+            {"params": ar_h.parameters(), "lr": LR_HEAD},
+            {"params": nat_h.parameters(), "lr": LR_HEAD},
+            {"params": sat_h.parameters(), "lr": LR_HEAD},
+        ]
+    )
+    scaler = GradScaler(enabled=args.amp and DEV.type == "cuda")
+
+    ce_tok = nn.CrossEntropyLoss(label_smoothing=0.1)
+    ctc = nn.CTCLoss(blank=BLANK, zero_infinity=True)
+    ce_gate = nn.CrossEntropyLoss()
+
+    # ---------- resume bookkeeping ----------
+    start_step, seen_tok = 0, 0
+    last_save_time = time.time()
+    if args.resume and not args.fresh:
+        start_step, seen_tok, last_save_time = load_ckpt(
+            pathlib.Path(args.resume), core, ar_h, nat_h, sat_h, opt, scaler
+        )
+        print(f"✓ resumed from step {start_step:,}, seen_tokens={seen_tok:,}")
+
+    # Target tokens
+    if args.target_tokens:
+        target_tokens = args.target_tokens
+    else:
+        param_count = sum(p.numel() for p in core.parameters())
+        target_tokens = int(25 * param_count)
+
+    new_tokens_needed = target_tokens - seen_tok
+    if new_tokens_needed <= 0:
+        print("Target already reached – nothing to train.")
+        return
+    new_steps = new_tokens_needed // BLOCK
+    if args.steps:
+        new_steps = min(new_steps, args.steps)
+        new_tokens_needed = new_steps * BLOCK
+
+    total_tokens_needed = seen_tok + new_tokens_needed
+    print(f"[auto-steps] {new_steps:,} training steps (@ {BLOCK} tokens/step)")
+
+    # Progressive growth plan
+    grow_plan = _parse_grow_plan(args.grow_plan) if args.auto_grow else []
+    if args.auto_grow:
+        if BLOCK not in grow_plan:
+            grow_plan = sorted(set(grow_plan + [BLOCK]))
+        print(f"[auto-grow] plan: {grow_plan} every {args.grow_every_steps} steps")
+
+    stream = token_stream(args.source, target_tokens, seed=42)
+    buf: list[int] = []
+    pbar = tqdm(total=total_tokens_needed, initial=seen_tok, unit="tok")
+    step = start_step
+    steps_since_last_grow = 0
+
+    while seen_tok < total_tokens_needed:
+        # ------- assemble one batch -------
+        try:
+            while len(buf) < BLOCK:
+                buf.append(next(stream))
+        except StopIteration:
+            break
+        ids = torch.tensor(buf[:BLOCK], device=DEV).unsqueeze(0)  # (B=1, N)
+        buf = buf[BLOCK:]
+
+        tgt_ar = ids.clone()                           # (1, N)
+        ids_nat = torch.repeat_interleave(ids, 2, 1)   # (1, 2N) for NAT only
+
+        try:
+            with amp(args.amp):
+                # AR path
+                h_ar = core(ids, causal_mask(ids.size(1)))
+                logits_ar = ar_h(h_ar)[:, :-1]
+                loss_ar = ce_tok(logits_ar.reshape(-1, VOCAB), tgt_ar[:, 1:].reshape(-1))
+
+                # NAT path (uses doubled sequence)
+                h_nat = core(ids_nat, None)
+                log_nat = nat_h(h_nat).log_softmax(-1).transpose(0, 1)  # (T,B,V)
+                ilen = tlen = torch.tensor([ids_nat.size(1) // 2], device=DEV)
+                loss_nat = ctc(log_nat, tgt_ar, ilen, tlen)
+
+                # SAT path
+                h_sat = core(ids, sat_mask(ids.size(1)))
+                logits_sat, gate = sat_h(h_sat[:, -SAT_BLOCK:])
+                tgt_sat = ids[:, 1:SAT_BLOCK+1]
+                loss_sat = ce_tok(logits_sat.reshape(-1, VOCAB), tgt_sat.reshape(-1))
+                if gate is not None:
+                    loss_sat += EMIT_LAMBDA * ce_gate(gate, torch.ones(ids.size(0), device=DEV, dtype=torch.long))
+
+                loss = loss_ar + loss_nat + loss_sat
+
+            # optimisation
+            scaler.scale(loss).backward()
+            scaler.unscale_(opt)
+            nn.utils.clip_grad_norm_(core.parameters(), 1.0)
+            scaler.step(opt)
+            scaler.update()
+            opt.zero_grad(set_to_none=True)
+
+        except RuntimeError as e:
+            msg = str(e).lower()
+            if "out of memory" in msg or "cuda error" in msg:
+                new_block = max(128, BLOCK // 2)
+                if new_block < BLOCK:
+                    print(f"\n[OOM] reducing block from {BLOCK} -> {new_block}")
+                    BLOCK = new_block
+                    if DEV.type == "cuda":
+                        torch.cuda.empty_cache()
+                    buf = ids[0].tolist() + buf
+                    steps_since_last_grow = 0
+                    continue
+            raise
+
+        # progress
+        step += 1
+        seen_tok += BLOCK
+        pbar.update(BLOCK)
+        pbar.set_postfix(loss=f"{loss.item():.3f}", block=BLOCK)
+
+        # checkpoint cadence
+        now = time.time()
+        time_due = (now - last_save_time) >= args.save_every_sec > 0
+        step_due = args.save_every_steps > 0 and step % args.save_every_steps == 0
+        if time_due or step_due:
+            ck_name = f"step{step:08d}.pt"
+            save_ckpt(
+                pathlib.Path(args.save_dir) / ck_name,
+                core, ar_h, nat_h, sat_h, opt, scaler,
+                meta={
+                    "cfg": cfg,
+                    "step": step,
+                    "seen_tok": seen_tok,
+                    "wall_time": now,
+                    "py_state": random.getstate(),
+                    "torch_state": rng_state(),
+                },
+            )
+            last_save_time = now
+
+        # progressive growth
+        if args.auto_grow:
+            steps_since_last_grow += 1
+            if steps_since_last_grow >= args.grow_every_steps:
+                steps_since_last_grow = 0
+                try:
+                    idx = grow_plan.index(BLOCK)
+                    if idx + 1 < len(grow_plan):
+                        candidate = grow_plan[idx + 1]
+                        print(f"[auto-grow] attempting BLOCK {BLOCK} -> {candidate}")
+                        BLOCK = candidate
+                        if DEV.type == "cuda":
+                            torch.cuda.empty_cache()
+                    else:
+                        print("[auto-grow] at max planned block; no further growth.")
+                except ValueError:
+                    grow_plan = sorted(set(grow_plan + [BLOCK]))
+                    idx = grow_plan.index(BLOCK)
+                    if idx + 1 < len(grow_plan):
+                        candidate = grow_plan[idx + 1]
+                        print(f"[auto-grow] moving to planned BLOCK {candidate}")
+                        BLOCK = candidate
+                        if DEV.type == "cuda":
+                            torch.cuda.empty_cache()
+
+    pbar.close()
+
+    # final save
+    save_ckpt(
+        pathlib.Path(args.save_dir) / "final.pt",
+        core, ar_h, nat_h, sat_h, opt, scaler,
+        meta={
+            "cfg": cfg,
+            "step": step,
+            "seen_tok": seen_tok,
+            "wall_time": time.time(),
+            "py_state": random.getstate(),
+            "torch_state": rng_state(),
+        },
+    )
+    print("🎉 training complete")
+
+
+# ───────────────────────── Sampling utils ─────────────────────────
+def _apply_no_repeat_ngram(logits: torch.Tensor, ids: torch.Tensor, n: int):
+    """
+    Block tokens that would complete any previously seen n-gram.
+    ids: (1, t)
+    logits: (..., V) where ... may be (1,) or (stride,)
+    """
+    if n <= 0 or ids.size(1) < n - 1:
+        return logits
+    prefix = ids[0, - (n - 1):].tolist()
+    # Build set of next tokens forbidden after this prefix.
+    banned = []
+    tokens = ids[0].tolist()
+    for i in range(len(tokens) - n + 1):
+        if tokens[i:i + n - 1] == prefix:
+            banned.append(tokens[i + n - 1])
+    if banned:
+        banned_idx = torch.tensor(banned, device=logits.device, dtype=torch.long)
+        logits[..., banned_idx] = float("-inf")
+    return logits
+
+
+def _apply_rep_presence_frequency(
+    logits: torch.Tensor, ids: torch.Tensor, last_n: int,
+    repetition_penalty: float, presence_penalty: float, frequency_penalty: float
+):
+    """
+    logits: (..., V) where ... may be (1,) or (stride,)
+    ids: (1, t) history
+    """
+    if ids.numel() == 0:
+        return logits
+    if last_n > 0:
+        hist = ids[0, -last_n:].to(torch.long)
+    else:
+        hist = ids[0].to(torch.long)
+
+    if hist.numel() == 0:
+        return logits
+
+    uniq, counts = torch.unique(hist, return_counts=True)
+
+    # presence/frequency penalties (OpenAI-like)
+    if presence_penalty != 0.0 or frequency_penalty != 0.0:
+        # subtract presence for seen tokens; subtract frequency * count
+        adjust = presence_penalty + frequency_penalty * counts.to(logits.dtype)
+        logits[..., uniq] = logits[..., uniq] - adjust
+
+    # repetition penalty (CTRL/GPT-NeoX style)
+    if repetition_penalty and abs(repetition_penalty - 1.0) > 1e-6:
+        sel = logits[..., uniq]
+        # if logit > 0: divide by penalty; else multiply by penalty
+        sel = torch.where(sel > 0, sel / repetition_penalty, sel * repetition_penalty)
+        logits[..., uniq] = sel
+
+    return logits
+
+
+def _filter_top_k_top_p_min_p(
+    logits: torch.Tensor, top_k: int, top_p: float, min_p: float, temperature: float
+) -> torch.Tensor:
+    """
+    Works on 1D or 2D logits (..., V). Applies temperature, then filtering.
+    Returns normalized probabilities ready for sampling.
+    """
+    logits = logits / max(temperature, 1e-8)
+
+    # shape handling
+    if logits.dim() == 1:
+        logits = logits.unsqueeze(0)
+
+    B, V = logits.size(0), logits.size(-1)
+
+    # Convert to probabilities for p-based filtering
+    probs = logits.softmax(-1)
+
+    # Top-k
+    if top_k and top_k < V:
+        vals, idx = torch.topk(probs, top_k, dim=-1)
+        mask = torch.full_like(probs, 0.0)
+        mask.scatter_(1, idx, 1.0)
+        probs = probs * mask
+
+    # Top-p (nucleus)
+    if top_p < 1.0:
+        sorted_probs, sorted_idx = torch.sort(probs, descending=True, dim=-1)
+        cumsum = torch.cumsum(sorted_probs, dim=-1)
+        keep = cumsum <= top_p
+        # Always keep at least one
+        keep[..., 0] = True
+        # Build mask
+        mask = torch.zeros_like(probs)
+        mask.scatter_(1, sorted_idx, keep.to(mask.dtype))
+        probs = probs * mask
+
+    # Min-p
+    if min_p > 0.0:
+        probs = torch.where(probs >= min_p, probs, torch.zeros_like(probs))
+
+    # If everything zeroed (can happen at extreme settings), fall back to the argmax token
+    sums = probs.sum(-1, keepdim=True)
+    empty = (sums == 0)
+    if empty.any():
+        fallback_idx = logits.argmax(-1, keepdim=True)
+        probs = torch.where(empty, torch.zeros_like(probs), probs)
+        probs.scatter_(-1, fallback_idx, torch.where(empty, torch.ones_like(sums), torch.zeros_like(sums)))
+
+    # Renormalize
+    probs = probs / probs.sum(-1, keepdim=True)
+    return probs
+
+
+# ───────────────────────── Inference helpers ─────────────────────────
+def load_joint(ckpt: str, preset: str):
+    path = _resolve_ckpt(pathlib.Path(ckpt)) or pathlib.Path(ckpt)
+    sd = _try_load(path, map_location="cpu")
+    if sd is None:
+        raise FileNotFoundError(f"No valid checkpoint at {path}")
+    cfg = sd["cfg"] if "cfg" in sd and isinstance(sd["cfg"], dict) else (infer_cfg_from_ckpt(path) or PRESETS[preset])
+    core = Encoder(cfg).to(DEV)
+    ar_h, nat_h = ARHead(cfg["d"]).to(DEV), NATHead(cfg["d"]).to(DEV)
+    sat_h = SATHead(cfg["d"]).to(DEV)
+    core.load_state_dict(sd["core"])
+    ar_h.load_state_dict(sd["ar"])
+    nat_h.load_state_dict(sd["nat"])
+    sat_h.load_state_dict(sd["sat"])
+    return core, ar_h, nat_h, sat_h
+
+
+@torch.no_grad()
+def ar_decode(core, ar_h, prompt: str, max_new: int, T: float,
+              greedy: bool, top_k: int, top_p: float, min_p: float,
+              repetition_penalty: float, presence_penalty: float,
+              frequency_penalty: float, penalty_last_n: int,
+              no_repeat_ngram_size: int):
+    ids = torch.tensor([tok.encode(prompt)], device=DEV)
+    if ids.size(1) == 0:
+        ids = torch.tensor([[EOS] if EOS is not None else [0]], device=DEV)
+    h_full, kvs = core(ids, causal_mask(ids.size(1)), use_cache=True)
+
+    start = time.time()
+    for _ in range(max_new):
+        logits = ar_h(h_full)[:, -1]  # (1, V)
+
+        # penalties
+        logits = _apply_no_repeat_ngram(logits, ids, no_repeat_ngram_size)
+        logits = _apply_rep_presence_frequency(
+            logits, ids, penalty_last_n, repetition_penalty, presence_penalty, frequency_penalty
+        )
+
+        if greedy:
+            nxt = logits.argmax(-1, keepdim=True)
+        else:
+            probs = _filter_top_k_top_p_min_p(logits.squeeze(0), top_k, top_p, min_p, T)
+            nxt = probs.multinomial(1)
+
+        ids = torch.cat([ids, nxt.unsqueeze(0) if nxt.dim()==1 else nxt], 1)
+
+        # step with kv cache
+        x = ids[:, -1:]
+        h_full, kvs = core(x, None, kv_caches=kvs, use_cache=True)
+
+    print(tok.decode(ids[0].tolist(), skip_special_tokens=True))
+    print(f"[{max_new} tok in {time.time() - start:.2f}s]")
+
+
+@torch.no_grad()
+def sat_decode(core, sat_h, prompt, max_new, T, var,
+               greedy: bool, top_k: int, top_p: float, min_p: float,
+               repetition_penalty: float, presence_penalty: float,
+               frequency_penalty: float, penalty_last_n: int,
+               no_repeat_ngram_size: int):
+    ids = torch.tensor([tok.encode(prompt)], device=DEV)
+    added, t0 = 0, time.time()
+    while added < max_new:
+        h = core(ids, sat_mask(ids.size(1)))
+        logits_all, gate = sat_h(h[:, -SAT_BLOCK:])  # (1, SAT_BLOCK, V)
+        stride = 2 if (not var or gate is None) else (gate.softmax(-1).multinomial(1) + 1).item()
+        stride = int(stride)
+
+        # Sequentially sample within the stride so penalties apply cumulatively
+        for pos in range(stride):
+            row_logits = logits_all[:, pos, :]  # (1, V)
+
+            # penalties
+            row_logits = _apply_no_repeat_ngram(row_logits, ids, no_repeat_ngram_size)
+            row_logits = _apply_rep_presence_frequency(
+                row_logits, ids, penalty_last_n, repetition_penalty, presence_penalty, frequency_penalty
+            )
+
+            if greedy:
+                nxt = row_logits.argmax(-1, keepdim=True)  # (1,1)
+            else:
+                probs = _filter_top_k_top_p_min_p(row_logits.squeeze(0), top_k, top_p, min_p, T)
+                nxt = probs.multinomial(1)  # (1,1)
+
+            ids = torch.cat([ids, nxt], 1)
+            added += 1
+            if added >= max_new:
+                break
+
+    print(tok.decode(ids[0].tolist(), skip_special_tokens=True))
+    print(f"[{added} tok in {time.time() - t0:.2f}s]")
+
+
+@torch.no_grad()
+def nat_decode(core, nat_h, prompt, max_new, passes, streams):
+    ids = torch.tensor([tok.encode(prompt) + [BLANK] * (max_new * 2)], device=DEV)
+    t0 = time.time()
+    for _ in range(passes):
+        h = core(ids, None)
+        logits = nat_h(h)
+        logits[..., BLANK] = -1e9
+        cand = logits.topk(streams, -1).indices.permute(2, 0, 1)
+        best = (cand != BLANK).float().mean(-1).argmax(0)
+        ids = cand[best, torch.arange(ids.size(0), device=DEV)][:, ::2]
+    out = [t for t in ids[0].tolist() if t != BLANK]
+    print(tok.decode(out, skip_special_tokens=True))
+    print(f"[{len(out)} output tokens in {time.time() - t0:.2f}s]")
+
+
+# ───────────────────────── CLI ─────────────────────────
+def main():
+    ap = argparse.ArgumentParser()
+    sub = ap.add_subparsers(dest="cmd", required=True)
+
+    tr = sub.add_parser("train")
+    tr.add_argument("--preset", choices=PRESETS, default="small")
+    tr.add_argument("--rank", type=int)
+    tr.add_argument("--block", type=int, default=DEFAULT_BLOCK)
+    tr.add_argument("--source", default="cerebras/SlimPajama-627B")
+    tr.add_argument("--target_tokens", type=int)
+    tr.add_argument("--steps", type=int)
+    tr.add_argument("--amp", action="store_true")
+    tr.add_argument("--save_every_sec", type=int, default=DEFAULT_SAVE_SEC)
+    tr.add_argument("--save_every_steps", type=int, default=0)
+    tr.add_argument("--save_dir", default=str(CKDIR))
+    tr.add_argument("--resume", type=str)
+    tr.add_argument("--x2", action="store_true", help="~2x params by doubling layers")
+    tr.add_argument("--warmstart_from", type=str, default=None, help="Path to previous final.pt for shape-safe warm start")
+    tr.add_argument("--fresh", action="store_true", help="Start from scratch: do not probe or load any checkpoints")
+
+    # Progressive block growth
+    tr.add_argument("--auto_grow", action="store_true", help="Automatically grow block size over time")
+    tr.add_argument("--grow_plan", type=str, default="576,640,768,896,1024", help="Comma list of block sizes to try in order")
+    tr.add_argument("--grow_every_steps", type=int, default=50000, help="Steps between growth attempts")
+
+    inf = sub.add_parser("infer")
+    inf.add_argument("--mode", choices=["ar", "nat", "sat"], required=True)
+    inf.add_argument("--ckpt", required=True)
+    inf.add_argument("--preset", default="small")
+    inf.add_argument("--prompt", required=True)
+    inf.add_argument("--max_new", type=int, default=120)
+    inf.add_argument("--temperature", type=float, default=1.0)
+
+    # New decode controls
+    inf.add_argument("--greedy", action="store_true", help="Greedy decode (overrides sampling)")
+    inf.add_argument("--top_k", type=int, default=0)
+    inf.add_argument("--top_p", type=float, default=1.0)
+    inf.add_argument("--min_p", type=float, default=0.0)
+
+    inf.add_argument("--repetition_penalty", type=float, default=1.0)
+    inf.add_argument("--presence_penalty", type=float, default=0.0)
+    inf.add_argument("--frequency_penalty", type=float, default=0.0)
+    inf.add_argument("--penalty_last_n", type=int, default=64)
+    inf.add_argument("--no_repeat_ngram_size", type=int, default=0)
+
+    inf.add_argument("--var", action="store_true")
+    inf.add_argument("--passes", type=int, default=1)
+    inf.add_argument("--streams", type=int, default=5)
+
+    args = ap.parse_args()
+    if args.cmd == "train":
+        train(args)
+    else:
+        core, ar_h, nat_h, sat_h = load_joint(args.ckpt, args.preset)
+        if args.mode == "ar":
+            ar_decode(core, ar_h, args.prompt, args.max_new, args.temperature,
+                      args.greedy, args.top_k, args.top_p, args.min_p,
+                      args.repetition_penalty, args.presence_penalty,
+                      args.frequency_penalty, args.penalty_last_n,
+                      args.no_repeat_ngram_size)
+        elif args.mode == "sat":
+            sat_decode(core, sat_h, args.prompt, args.max_new, args.temperature, args.var,
+                       args.greedy, args.top_k, args.top_p, args.min_p,
+                       args.repetition_penalty, args.presence_penalty,
+                       args.frequency_penalty, args.penalty_last_n,
+                       args.no_repeat_ngram_size)
+        else:
+            nat_decode(core, nat_h, args.prompt, args.max_new, args.passes, args.streams)
+
+
+if __name__ == "__main__":
+    main()

step01015312.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8ea0ab49a2505d0d393265e1778c7a75ea7fff68cf9a500eed7d2d5799483b56
+size 4388630747