tier-2 cls_pp checkpoint (htop90=2)

manifest.json

@@ -0,0 +1,7 @@
+{
+  "entry_class": "model.EBMModMul",
+  "output_base": 10,
+  "framework": "pytorch",
+  "model_description": "Joint-attention Transformer (~3.2M params, d_model=256, 4-6 layers, 8 heads) that learns x*y mod p for small primes. Operands are reduced per-argument (a%p, b%p) inside predict_digits, then the residues and prime are encoded as fixed-width decimal digits in a single self-attention sequence; a CLS slot reads out the answer residue via either a classification head (residue class in [0, p_max)) or an angular head ((cos,sin) on the unit circle, decoded to the nearest residue). Answer emitted as base-10 digits. Targets tiers 1-3; emits [0] for p >= 1e5 (out of the fixed-width regime).",
+  "training_description": "Trained from random init on synthetic (x, y, p, x*y mod p) examples with x,y in [0,p), over the full enumerable prime pools of tiers 1-3. Cross-entropy (classification head) or angular distance loss (Saxena-Charton circle encoding) with AdamW and weight decay; a fixed-dataset + weight-decay 'grokking' regime is used to push within-prime generalization. No hand-coded arithmetic: the modular product is produced by the trained network (randomizing weights collapses accuracy)."
+}

model.py

@@ -0,0 +1,266 @@
+"""Submission entry point: learned modular multiplication.
+
+Compliance contract (see rules/evaluation.md):
+- ``preprocess_*`` are per-argument identities (each sees only its own argument).
+- Inside ``predict_digits_batch`` we reduce each operand modulo p — ``int(a) % p``
+  and ``int(b) % p`` — the same two-args-at-a-time normalisation the reference
+  baselines use. We never form ``a * b`` or ``(a*b) % p`` in Python/tensors; the
+  modular product is produced by the trained network, whose output (a residue in
+  ``[0, p)``) materially determines the answer.
+- We emit the residue as base-10 digits (``output_base = 10``); the harness decodes.
+
+Out of regime (``p >= 10**WIDTH``, i.e. tiers >= 4) the network's fixed-width
+residue encoding cannot represent the operands, so we emit ``[0]`` — an honest
+fallback, not a guess. This model targets the low tiers (1-3).
+
+The architecture (encoder + classification/angular head) is loaded from the
+checkpoint's ``arch`` field, so the same wrapper serves either trained head.
+"""
+
+from __future__ import annotations
+
+import math
+from pathlib import Path
+
+import torch
+import torch.nn as nn
+
+from modchallenge.interface.base_model import ModularMultiplicationModel
+
+# ---------------------------------------------------------------------------
+# Fixed dimensions (must match the training code that produced the weights)
+# ---------------------------------------------------------------------------
+
+VOCAB_SIZE = 10           # decimal digits 0-9; fixed-width inputs, no PAD token
+WIDTH = 5                 # values < 10**5 = 100000 -> covers tiers 1-3
+SEG_X, SEG_Y, SEG_P, SEG_ANS = 0, 1, 2, 3
+
+
+def digits_fixed(n: int, width: int = WIDTH) -> list[int]:
+    """Non-negative int -> fixed-width zero-padded decimal digits, MSB-first."""
+    out = [0] * width
+    i = width - 1
+    while n > 0 and i >= 0:
+        out[i] = n % 10
+        n //= 10
+        i -= 1
+    return out
+
+
+def int_to_decimal_digits(n: int) -> list[int]:
+    """Non-negative int -> base-10 digit list, MSB-first ([0] for zero)."""
+    if n == 0:
+        return [0]
+    return [int(c) for c in str(n)]
+
+
+# ---------------------------------------------------------------------------
+# Architectures (copied verbatim from training/model.py for state_dict match)
+# ---------------------------------------------------------------------------
+
+class JointModMulNetCls(nn.Module):
+    def __init__(self, d_model=256, nhead=8, num_layers=6, dim_ff=1024, p_max=256):
+        super().__init__()
+        self.p_max = p_max
+        self.tok_emb = nn.Embedding(VOCAB_SIZE, d_model)
+        self.cls_query = nn.Parameter(torch.randn(1, d_model) * 0.02)
+        self.seg_emb = nn.Embedding(4, d_model)
+        self.pos_emb = nn.Embedding(3 * WIDTH + 1, d_model)
+        layer = nn.TransformerEncoderLayer(
+            d_model=d_model, nhead=nhead, dim_feedforward=dim_ff,
+            dropout=0.0, batch_first=True, activation="gelu",
+        )
+        self.encoder = nn.TransformerEncoder(layer, num_layers=num_layers)
+        self.ln = nn.LayerNorm(d_model)
+        self.head = nn.Linear(d_model, p_max)
+        seg = torch.tensor([SEG_X] * WIDTH + [SEG_Y] * WIDTH + [SEG_P] * WIDTH + [SEG_ANS])
+        self.register_buffer("seg_ids", seg, persistent=False)
+        self.register_buffer("pos_ids", torch.arange(3 * WIDTH + 1), persistent=False)
+
+    def forward(self, x_digits, y_digits, prime_digits):
+        b = x_digits.shape[0]
+        inp = torch.cat([x_digits, y_digits, prime_digits], dim=1)
+        tok = self.tok_emb(inp)
+        cls = self.cls_query.unsqueeze(0).expand(b, 1, -1)
+        x = torch.cat([tok, cls], dim=1)
+        x = x + self.seg_emb(self.seg_ids.unsqueeze(0)) + self.pos_emb(self.pos_ids.unsqueeze(0))
+        x = self.encoder(x)
+        x = self.ln(x)
+        return self.head(x[:, -1, :])  # (B, p_max)
+
+
+class JointModMulNetAngular(nn.Module):
+    def __init__(self, d_model=256, nhead=8, num_layers=6, dim_ff=1024):
+        super().__init__()
+        self.tok_emb = nn.Embedding(VOCAB_SIZE, d_model)
+        self.cls_query = nn.Parameter(torch.randn(1, d_model) * 0.02)
+        self.seg_emb = nn.Embedding(4, d_model)
+        self.pos_emb = nn.Embedding(3 * WIDTH + 1, d_model)
+        layer = nn.TransformerEncoderLayer(
+            d_model=d_model, nhead=nhead, dim_feedforward=dim_ff,
+            dropout=0.0, batch_first=True, activation="gelu",
+        )
+        self.encoder = nn.TransformerEncoder(layer, num_layers=num_layers)
+        self.ln = nn.LayerNorm(d_model)
+        self.head = nn.Linear(d_model, 2)
+        seg = torch.tensor([SEG_X] * WIDTH + [SEG_Y] * WIDTH + [SEG_P] * WIDTH + [SEG_ANS])
+        self.register_buffer("seg_ids", seg, persistent=False)
+        self.register_buffer("pos_ids", torch.arange(3 * WIDTH + 1), persistent=False)
+
+    def forward(self, x_digits, y_digits, prime_digits):
+        b = x_digits.shape[0]
+        inp = torch.cat([x_digits, y_digits, prime_digits], dim=1)
+        tok = self.tok_emb(inp)
+        cls = self.cls_query.unsqueeze(0).expand(b, 1, -1)
+        x = torch.cat([tok, cls], dim=1)
+        x = x + self.seg_emb(self.seg_ids.unsqueeze(0)) + self.pos_emb(self.pos_ids.unsqueeze(0))
+        x = self.encoder(x)
+        x = self.ln(x)
+        return self.head(x[:, -1, :])  # (B, 2)
+
+
+PRIME_ENUM_LIMIT = 65536
+
+
+def _sieve_primes(limit: int) -> list[int]:
+    is_p = bytearray([1]) * limit
+    is_p[0] = is_p[1] = 0
+    for i in range(2, int(limit ** 0.5) + 1):
+        if is_p[i]:
+            is_p[i * i :: i] = bytearray(len(is_p[i * i :: i]))
+    return [i for i in range(2, limit) if is_p[i]]
+
+
+class JointModMulNetClsPP(nn.Module):
+    """Joint-attention classifier with a learned per-prime embedding.
+    Mirrors training/model.py for state_dict compatibility."""
+
+    def __init__(self, d_model=256, nhead=8, num_layers=6, dim_ff=1024, p_max=256):
+        super().__init__()
+        self.p_max = p_max
+        self.limit = PRIME_ENUM_LIMIT
+        self.tok_emb = nn.Embedding(VOCAB_SIZE, d_model)
+        self.cls_query = nn.Parameter(torch.randn(1, d_model) * 0.02)
+        self.seg_emb = nn.Embedding(4, d_model)
+        self.pos_emb = nn.Embedding(3 * WIDTH + 1, d_model)
+        layer = nn.TransformerEncoderLayer(
+            d_model=d_model, nhead=nhead, dim_feedforward=dim_ff,
+            dropout=0.0, batch_first=True, activation="gelu",
+        )
+        self.encoder = nn.TransformerEncoder(layer, num_layers=num_layers)
+        self.ln = nn.LayerNorm(d_model)
+        self.head = nn.Linear(d_model, p_max)
+        primes = _sieve_primes(self.limit)
+        self.prime_emb = nn.Embedding(len(primes), d_model)
+        idx = torch.zeros(self.limit, dtype=torch.long)
+        valid = torch.zeros(self.limit, dtype=torch.float)
+        for rank, p in enumerate(primes):
+            idx[p] = rank
+            valid[p] = 1.0
+        self.register_buffer("idx_lookup", idx, persistent=False)
+        self.register_buffer("valid_lookup", valid, persistent=False)
+        self.register_buffer(
+            "place_value",
+            torch.tensor([10 ** (WIDTH - 1 - i) for i in range(WIDTH)], dtype=torch.long),
+            persistent=False,
+        )
+        seg = torch.tensor([SEG_X] * WIDTH + [SEG_Y] * WIDTH + [SEG_P] * WIDTH + [SEG_ANS])
+        self.register_buffer("seg_ids", seg, persistent=False)
+        self.register_buffer("pos_ids", torch.arange(3 * WIDTH + 1), persistent=False)
+
+    def forward(self, x_digits, y_digits, prime_digits):
+        b = x_digits.shape[0]
+        p_int = (prime_digits * self.place_value).sum(dim=1)
+        safe = p_int.clamp(0, self.limit - 1)
+        p_emb = self.prime_emb(self.idx_lookup[safe]) * self.valid_lookup[safe].unsqueeze(-1)
+        inp = torch.cat([x_digits, y_digits, prime_digits], dim=1)
+        tok = self.tok_emb(inp)
+        cls = self.cls_query.unsqueeze(0).expand(b, 1, -1)
+        x = torch.cat([tok, cls], dim=1)
+        x = x + self.seg_emb(self.seg_ids.unsqueeze(0)) + self.pos_emb(self.pos_ids.unsqueeze(0))
+        x = x + p_emb.unsqueeze(1)
+        x = self.encoder(x)
+        x = self.ln(x)
+        return self.head(x[:, -1, :])
+
+
+_ARCHS = {
+    "cls": JointModMulNetCls,
+    "cls_pp": JointModMulNetClsPP,
+    "angular": JointModMulNetAngular,
+}
+
+
+def _angular_decode(pred: torch.Tensor, p_int: torch.Tensor) -> torch.Tensor:
+    theta = torch.atan2(pred[:, 1], pred[:, 0])
+    t = torch.round(theta * p_int.float() / (2 * math.pi))
+    return (t % p_int.float()).long()
+
+
+# ---------------------------------------------------------------------------
+# Submission entry class
+# ---------------------------------------------------------------------------
+
+class EBMModMul(ModularMultiplicationModel):
+    def __init__(self):
+        self.model = None
+        self.device = None
+        self.arch = None
+
+    def load(self, model_dir: str) -> None:
+        if torch.cuda.is_available():
+            self.device = torch.device("cuda")
+        elif torch.backends.mps.is_available():
+            self.device = torch.device("mps")
+        else:
+            self.device = torch.device("cpu")
+
+        ckpt = torch.load(Path(model_dir) / "weights.pt",
+                          map_location=self.device, weights_only=False)
+        self.arch = ckpt.get("arch", "cls")
+        self.model = _ARCHS[self.arch](**ckpt["config"]).to(self.device)
+        self.model.load_state_dict(ckpt["state_dict"])
+        self.model.eval()
+
+    # Per-argument identity preprocessing (each hook sees only its own argument).
+    def preprocess_a(self, a): return a
+    def preprocess_b(self, b): return b
+    def preprocess_p(self, p): return p
+
+    @torch.no_grad()
+    def predict_digits(self, a_enc, b_enc, p_enc):
+        return self.predict_digits_batch([(a_enc, b_enc, p_enc)])[0]
+
+    @torch.no_grad()
+    def predict_digits_batch(self, inputs):
+        out: list[list[int] | None] = [None] * len(inputs)
+        x_rows, y_rows, p_rows, p_ints, idx = [], [], [], [], []
+
+        for i, (a_enc, b_enc, p_enc) in enumerate(inputs):
+            p = int(p_enc)
+            # Out of the model's regime (residues don't fit WIDTH digits): honest 0.
+            if p >= 10 ** WIDTH:
+                out[i] = [0]
+                continue
+            a_red = int(a_enc) % p          # per-operand reduction (allowed)
+            b_red = int(b_enc) % p
+            x_rows.append(digits_fixed(a_red))
+            y_rows.append(digits_fixed(b_red))
+            p_rows.append(digits_fixed(p))
+            p_ints.append(p)
+            idx.append(i)
+
+        if idx:
+            t = lambda r: torch.tensor(r, dtype=torch.long, device=self.device)
+            logits = self.model(t(x_rows), t(y_rows), t(p_rows))
+            if self.arch == "angular":
+                residues = _angular_decode(logits, t(p_ints)).tolist()
+            else:
+                residues = logits.argmax(dim=-1).tolist()
+            for j, i in enumerate(idx):
+                out[i] = int_to_decimal_digits(int(residues[j]))
+
+        return [o if o is not None else [0] for o in out]
+
+    def max_batch_size(self) -> int:
+        return 512

weights.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:14920439c3ac57c903108cafff9b06176404d772e9d3cf26d856b35291f2d903
+size 25979504