diff --git a/concepts/f1_sota_garage/car02_speed_lane/train_gpt.py b/concepts/f1_sota_garage/car02_speed_lane/train_gpt.py new file mode 100644 index 000000000..6566ad098 --- /dev/null +++ b/concepts/f1_sota_garage/car02_speed_lane/train_gpt.py @@ -0,0 +1,2083 @@ +from __future__ import annotations +import copy +import glob +import io +import math +import os +import random +import subprocess +import sys +import time +import uuid +import zlib +from pathlib import Path +try: + import zstandard + _COMPRESSOR = "zstd" +except ImportError: + _COMPRESSOR = "zlib" +import numpy as np +import sentencepiece as spm +import torch +import torch.distributed as dist +import torch.nn.functional as F +from torch import Tensor, nn +from torch.nn.parallel import DistributedDataParallel as DDP +from flash_attn_interface import flash_attn_func as flash_attn_3_func +class Hyperparameters: + data_path = os.environ.get("DATA_PATH", "./data/datasets/fineweb10B_sp1024") + train_files = os.path.join(data_path, "fineweb_train_*.bin") + val_files = os.path.join(data_path, "fineweb_val_*.bin") + tokenizer_path = os.environ.get("TOKENIZER_PATH", "./data/tokenizers/fineweb_1024_bpe.model") + run_id = os.environ.get("RUN_ID", str(uuid.uuid4())) + seed = int(os.environ.get("SEED", 1337)) + val_batch_size = int(os.environ.get("VAL_BATCH_SIZE", 524_288)) + val_loss_every = int(os.environ.get("VAL_LOSS_EVERY", 4000)) + train_log_every = int(os.environ.get("TRAIN_LOG_EVERY", 500)) + iterations = int(os.environ.get("ITERATIONS", 20000)) + warmdown_iters = int(os.environ.get("WARMDOWN_ITERS", 3500)) + warmup_steps = int(os.environ.get("WARMUP_STEPS", 20)) + train_batch_tokens = int(os.environ.get("TRAIN_BATCH_TOKENS", 786_432)) + train_seq_len = int(os.environ.get("TRAIN_SEQ_LEN", 2048)) + eval_seq_len = int(os.environ.get("EVAL_SEQ_LEN", 2048)) + max_wallclock_seconds = float(os.environ.get("MAX_WALLCLOCK_SECONDS", 600.0)) + qk_gain_init = float(os.environ.get("QK_GAIN_INIT", 1.5)) + vocab_size = int(os.environ.get("VOCAB_SIZE", 1024)) + num_layers = int(os.environ.get("NUM_LAYERS", 11)) + num_kv_heads = int(os.environ.get("NUM_KV_HEADS", 4)) + model_dim = int(os.environ.get("MODEL_DIM", 512)) + num_heads = int(os.environ.get("NUM_HEADS", 8)) + mlp_mult = float(os.environ.get("MLP_MULT", 3.0)) + mlp_act = os.environ.get("MLP_ACT", "relu_sq").lower() + mlp_leaky_slope = float(os.environ.get("MLP_LEAKY_SLOPE", 0.5)) + tie_embeddings = bool(int(os.environ.get("TIE_EMBEDDINGS", "1"))) + rope_base = float(os.environ.get("ROPE_BASE", 10000.0)) + logit_softcap = float(os.environ.get("LOGIT_SOFTCAP", 30.0)) + embed_lr = float(os.environ.get("EMBED_LR", 0.6)) + head_lr = float(os.environ.get("HEAD_LR", 0.008)) + tied_embed_lr = float(os.environ.get("TIED_EMBED_LR", 0.035)) + tied_embed_init_std = float(os.environ.get("TIED_EMBED_INIT_STD", 0.005)) + matrix_lr = float(os.environ.get("MATRIX_LR", 0.025)) + scalar_lr = float(os.environ.get("SCALAR_LR", 0.025)) + muon_momentum = float(os.environ.get("MUON_MOMENTUM", 0.99)) + muon_backend_steps = int(os.environ.get("MUON_BACKEND_STEPS", 5)) + muon_momentum_warmup_start = float(os.environ.get("MUON_MOMENTUM_WARMUP_START", 0.92)) + muon_momentum_warmup_steps = int(os.environ.get("MUON_MOMENTUM_WARMUP_STEPS", 1500)) + beta1 = float(os.environ.get("BETA1", 0.9)) + beta2 = float(os.environ.get("BETA2", 0.95)) + adam_eps = float(os.environ.get("ADAM_EPS", 1e-8)) + grad_clip_norm = float(os.environ.get("GRAD_CLIP_NORM", 0.3)) + eval_stride = int(os.environ.get("EVAL_STRIDE", 64)) + mtp_num_heads = int(os.environ.get("MTP_NUM_HEADS", 0)) + mtp_loss_weight = float(os.environ.get("MTP_LOSS_WEIGHT", 0.2)) + muon_beta2 = float(os.environ.get("MUON_BETA2", 0.95)) + swa_enabled = bool(int(os.environ.get("SWA_ENABLED", "1"))) + swa_every = int(os.environ.get("SWA_EVERY", 50)) # tighter: collect more recent checkpoints + muon_wd = float(os.environ.get("MUON_WD", 0.04)) + adam_wd = float(os.environ.get("ADAM_WD", 0.04)) + qat_enabled = bool(int(os.environ.get("QAT_ENABLED", "0"))) + bigram_vocab_size = int(os.environ.get("BIGRAM_VOCAB_SIZE", 2048)) + bigram_dim = int(os.environ.get("BIGRAM_DIM", 128)) + xsa_last_n = int(os.environ.get("XSA_LAST_N", 11)) # XSA on ALL 11 layers + rope_dims = int(os.environ.get("ROPE_DIMS", 16)) + ln_scale = bool(int(os.environ.get("LN_SCALE", "1"))) + dtg_enabled = bool(int(os.environ.get("DTG_ENABLED", "0"))) + late_qat_threshold = float(os.environ.get("LATE_QAT_THRESHOLD", 0.5)) + ve_enabled = bool(int(os.environ.get("VE_ENABLED", "1"))) + ve_dim = int(os.environ.get("VE_DIM", 128)) + ve_layers = os.environ.get("VE_LAYERS", "9,10") + # F1 capacity add-on: low-rank correction head (active at inference). + # Approx extra params ~= rank * (model_dim + vocab_size). + f1_corr_rank = int(os.environ.get("F1_CORR_RANK", 0)) + f1_corr_scale_init = float(os.environ.get("F1_CORR_SCALE_INIT", 0.10)) + # Post-train self-distillation: EMA teacher -> student. + distill_enabled = bool(int(os.environ.get("DISTILL_ENABLED", "0"))) + distill_steps = int(os.environ.get("DISTILL_STEPS", 24)) + distill_lr_factor = float(os.environ.get("DISTILL_LR_FACTOR", 0.02)) + distill_temperature = float(os.environ.get("DISTILL_TEMPERATURE", 1.5)) + distill_alpha = float(os.environ.get("DISTILL_ALPHA", 0.60)) + distill_kl_clip = float(os.environ.get("DISTILL_KL_CLIP", 10.0)) + # Legal score-first TTT eval (PR #461 recipe) + ttt_eval_enabled = bool(int(os.environ.get("TTT_EVAL_ENABLED", "1"))) + ttt_lr = float(os.environ.get("TTT_LR", 0.002)) + ttt_epochs = int(os.environ.get("TTT_EPOCHS", 3)) + ttt_chunk_tokens = int(os.environ.get("TTT_CHUNK_TOKENS", 32768)) + ttt_freeze_blocks = int(os.environ.get("TTT_FREEZE_BLOCKS", 2)) + ttt_momentum = float(os.environ.get("TTT_MOMENTUM", 0.9)) + ttt_batch_seqs = int(os.environ.get("TTT_BATCH_SEQS", 32)) + ttt_grad_clip = float(os.environ.get("TTT_GRAD_CLIP", 1.0)) + ttt_max_train_chunks = int(os.environ.get("TTT_MAX_TRAIN_CHUNKS", 200)) # stop training after N chunks, keep scoring + ttt_ema_decay = float(os.environ.get("TTT_EMA_DECAY", 0.995)) # EMA decay for TTT weight smoothing (0 = disabled) + ttt_freeze_embed = bool(int(os.environ.get("TTT_FREEZE_EMBED", "1"))) # freeze tok_emb/bigram/ve during TTT + # Optional legal score-first hashed n-gram interpolation at eval time. + # This path never uses label-aware gating: mix weight is fixed. + ngram_eval_order = int(os.environ.get("NGRAM_EVAL_ORDER", 0)) # 0=off, >=2 enables hashed n-gram + ngram_eval_alpha = float(os.environ.get("NGRAM_EVAL_ALPHA", 0.20)) + ngram_eval_min_count = int(os.environ.get("NGRAM_EVAL_MIN_COUNT", 2)) + ngram_eval_buckets = int(os.environ.get("NGRAM_EVAL_BUCKETS", 4_194_304)) + ngram_eval_max_seconds = float(os.environ.get("NGRAM_EVAL_MAX_SECONDS", 0.0)) + compile_enabled = bool(int(os.environ.get("COMPILE_ENABLED", "1"))) + compile_fullgraph = bool(int(os.environ.get("COMPILE_FULLGRAPH", "1"))) +def maybe_torch_compile(obj, args: Hyperparameters): + if not args.compile_enabled: + return obj + return torch.compile(obj, dynamic=False, fullgraph=args.compile_fullgraph) +def zeropower_via_newtonschulz5(G: Tensor, steps: int = 10, eps: float = 1e-7) -> Tensor: + a, b, c = (3.4445, -4.7750, 2.0315) + X = G.bfloat16() + X /= X.norm() + eps + transposed = G.size(0) > G.size(1) + if transposed: + X = X.T + for _ in range(steps): + A = X @ X.T + B = b * A + c * A @ A + X = a * X + B @ X + return X.T if transposed else X +class Muon(torch.optim.Optimizer): + def __init__(self, params, lr: float, momentum: float, backend_steps: int, + nesterov: bool = True, weight_decay: float = 0.0): + super().__init__( + params, + dict(lr=lr, momentum=momentum, backend_steps=backend_steps, + nesterov=nesterov, weight_decay=weight_decay), + ) + @torch.no_grad() + def step(self, closure=None): + loss = None + if closure is not None: + with torch.enable_grad(): + loss = closure() + distributed = dist.is_available() and dist.is_initialized() + world_size = dist.get_world_size() if distributed else 1 + rank = dist.get_rank() if distributed else 0 + for group in self.param_groups: + params = group["params"] + if not params: + continue + lr = group["lr"] + momentum = group["momentum"] + backend_steps = group["backend_steps"] + nesterov = group["nesterov"] + total_params = sum(int(p.numel()) for p in params) + updates_flat = torch.zeros(total_params, device=params[0].device, dtype=torch.bfloat16) + curr = 0 + for i, p in enumerate(params): + if i % world_size == rank and p.grad is not None: + g = p.grad + state = self.state[p] + if "momentum_buffer" not in state: + state["momentum_buffer"] = torch.zeros_like(g) + buf = state["momentum_buffer"] + buf.mul_(momentum).add_(g) + if nesterov: + g = g.add(buf, alpha=momentum) + g = zeropower_via_newtonschulz5(g, steps=backend_steps) + g *= max(1, g.size(0) / g.size(1)) ** 0.5 + updates_flat[curr : curr + p.numel()] = g.reshape(-1) + curr += p.numel() + if distributed: + dist.all_reduce(updates_flat, op=dist.ReduceOp.SUM) + wd = group.get("weight_decay", 0.0) + curr = 0 + for p in params: + if wd > 0.0: + p.data.mul_(1.0 - lr * wd) + g = updates_flat[curr : curr + p.numel()].view_as(p).to(dtype=p.dtype) + p.add_(g, alpha=-lr) + curr += p.numel() + return loss +def build_sentencepiece_luts( + sp: spm.SentencePieceProcessor, vocab_size: int, device: torch.device +) -> tuple[Tensor, Tensor, Tensor]: + sp_vocab_size = int(sp.vocab_size()) + table_size = max(sp_vocab_size, vocab_size) + base_bytes_np = np.zeros((table_size,), dtype=np.int16) + has_leading_space_np = np.zeros((table_size,), dtype=np.bool_) + is_boundary_token_np = np.ones((table_size,), dtype=np.bool_) + for token_id in range(sp_vocab_size): + if sp.is_control(token_id) or sp.is_unknown(token_id) or sp.is_unused(token_id): + continue + is_boundary_token_np[token_id] = False + if sp.is_byte(token_id): + base_bytes_np[token_id] = 1 + continue + piece = sp.id_to_piece(token_id) + if piece.startswith("▁"): + has_leading_space_np[token_id] = True + piece = piece[1:] + base_bytes_np[token_id] = len(piece.encode("utf-8")) + return ( + torch.tensor(base_bytes_np, dtype=torch.int16, device=device), + torch.tensor(has_leading_space_np, dtype=torch.bool, device=device), + torch.tensor(is_boundary_token_np, dtype=torch.bool, device=device), + ) +def load_validation_tokens(pattern: str, seq_len: int) -> Tensor: + files = [Path(p) for p in sorted(glob.glob(pattern))] + if not files: + raise FileNotFoundError(f"No files found for pattern: {pattern}") + tokens = torch.cat([load_data_shard(file) for file in files]).contiguous() + usable = ((tokens.numel() - 1) // seq_len) * seq_len + if usable <= 0: + raise ValueError(f"Validation split is too short for TRAIN_SEQ_LEN={seq_len}") + return tokens[: usable + 1] +def eval_val( + args: Hyperparameters, + model: nn.Module, + rank: int, + world_size: int, + device: torch.device, + grad_accum_steps: int, + val_tokens: Tensor, + base_bytes_lut: Tensor, + has_leading_space_lut: Tensor, + is_boundary_token_lut: Tensor, + eval_seq_len: int | None = None, +) -> tuple[float, float]: + seq_len = eval_seq_len or args.train_seq_len + local_batch_tokens = args.val_batch_size // (world_size * grad_accum_steps) + if local_batch_tokens < seq_len: + raise ValueError( + "VAL_BATCH_SIZE must provide at least one sequence per rank; " + f"got VAL_BATCH_SIZE={args.val_batch_size}, WORLD_SIZE={world_size}, " + f"GRAD_ACCUM_STEPS={grad_accum_steps}, seq_len={seq_len}" + ) + local_batch_seqs = local_batch_tokens // seq_len + total_seqs = (val_tokens.numel() - 1) // seq_len + seq_start = (total_seqs * rank) // world_size + seq_end = (total_seqs * (rank + 1)) // world_size + val_loss_sum = torch.zeros((), device=device, dtype=torch.float64) + val_token_count = torch.zeros((), device=device, dtype=torch.float64) + val_byte_count = torch.zeros((), device=device, dtype=torch.float64) + model.eval() + with torch.inference_mode(): + for batch_seq_start in range(seq_start, seq_end, local_batch_seqs): + batch_seq_end = min(batch_seq_start + local_batch_seqs, seq_end) + raw_start = batch_seq_start * seq_len + raw_end = batch_seq_end * seq_len + 1 + local = val_tokens[raw_start:raw_end].to(device=device, dtype=torch.int64, non_blocking=True) + x = local[:-1].reshape(-1, seq_len) + y = local[1:].reshape(-1, seq_len) + with torch.autocast(device_type="cuda", dtype=torch.bfloat16, enabled=True): + batch_loss = model(x, y).detach() + batch_token_count = float(y.numel()) + val_loss_sum += batch_loss.to(torch.float64) * batch_token_count + val_token_count += batch_token_count + prev_ids = x.reshape(-1) + tgt_ids = y.reshape(-1) + token_bytes = base_bytes_lut[tgt_ids].to(dtype=torch.int16) + token_bytes += (has_leading_space_lut[tgt_ids] & ~is_boundary_token_lut[prev_ids]).to(dtype=torch.int16) + val_byte_count += token_bytes.to(torch.float64).sum() + if dist.is_available() and dist.is_initialized(): + dist.all_reduce(val_loss_sum, op=dist.ReduceOp.SUM) + dist.all_reduce(val_token_count, op=dist.ReduceOp.SUM) + dist.all_reduce(val_byte_count, op=dist.ReduceOp.SUM) + val_loss = val_loss_sum / val_token_count + bits_per_token = val_loss.item() / math.log(2.0) + tokens_per_byte = val_token_count.item() / val_byte_count.item() + model.train() + return float(val_loss.item()), float(bits_per_token * tokens_per_byte) +CONTROL_TENSOR_NAME_PATTERNS = tuple( + pattern + for pattern in os.environ.get( + "CONTROL_TENSOR_NAME_PATTERNS", + "attn_scale,attn_scales,mlp_scale,mlp_scales,resid_mix,resid_mixes,q_gain,skip_weight,skip_weights,smear,dtg_gate,ve_layer_scales,ve_shared.scale", + ).split(",") + if pattern +) +INT8_KEEP_FLOAT_FP32_NAME_PATTERNS = tuple( + pattern + for pattern in os.environ.get( + "INT8_KEEP_FLOAT_FP32_NAME_PATTERNS", + ",".join(CONTROL_TENSOR_NAME_PATTERNS), + ).split(",") + if pattern +) +INT8_KEEP_FLOAT_MAX_NUMEL = 65_536 +INT8_KEEP_FLOAT_STORE_DTYPE = torch.float16 +INT8_PER_ROW_SCALE_DTYPE = torch.float16 +INT8_CLIP_PERCENTILE = 99.99984 +INT8_CLIP_Q = INT8_CLIP_PERCENTILE / 100.0 +def tensor_nbytes(t: Tensor) -> int: + return int(t.numel()) * int(t.element_size()) +def keep_float_tensor(name: str, t: Tensor, passthrough_orig_dtypes: dict[str, str]) -> Tensor: + if any(pattern in name for pattern in INT8_KEEP_FLOAT_FP32_NAME_PATTERNS): + return t.float().contiguous() + if t.dtype in {torch.float32, torch.bfloat16}: + passthrough_orig_dtypes[name] = str(t.dtype).removeprefix("torch.") + return t.to(dtype=INT8_KEEP_FLOAT_STORE_DTYPE).contiguous() + return t +def quantize_float_tensor(t: Tensor) -> tuple[Tensor, Tensor]: + t32 = t.float() + if t32.ndim == 2: + clip_abs = ( + torch.quantile(t32.abs(), INT8_CLIP_Q, dim=1) + if t32.numel() + else torch.empty((t32.shape[0],), dtype=torch.float32) + ) + clipped = torch.maximum(torch.minimum(t32, clip_abs[:, None]), -clip_abs[:, None]) + scale = (clip_abs / 127.0).clamp_min(1.0 / 127.0) + q = torch.clamp(torch.round(clipped / scale[:, None]), -127, 127).to(torch.int8).contiguous() + return q, scale.to(dtype=INT8_PER_ROW_SCALE_DTYPE).contiguous() + clip_abs = float(torch.quantile(t32.abs().flatten(), INT8_CLIP_Q).item()) if t32.numel() else 0.0 + scale = torch.tensor(clip_abs / 127.0 if clip_abs > 0 else 1.0, dtype=torch.float32) + q = torch.clamp(torch.round(torch.clamp(t32, -clip_abs, clip_abs) / scale), -127, 127).to(torch.int8).contiguous() + return q, scale +def quantize_state_dict_int8(state_dict: dict[str, Tensor]): + quantized: dict[str, Tensor] = {} + scales: dict[str, Tensor] = {} + dtypes: dict[str, str] = {} + passthrough: dict[str, Tensor] = {} + passthrough_orig_dtypes: dict[str, str] = {} + qmeta: dict[str, dict[str, object]] = {} + stats = dict.fromkeys( + ("param_count", "num_tensors", "num_float_tensors", "num_nonfloat_tensors", "baseline_tensor_bytes", "int8_payload_bytes"), + 0, + ) + for name, tensor in state_dict.items(): + t = tensor.detach().to("cpu").contiguous() + stats["param_count"] += int(t.numel()) + stats["num_tensors"] += 1 + stats["baseline_tensor_bytes"] += tensor_nbytes(t) + if not t.is_floating_point(): + stats["num_nonfloat_tensors"] += 1 + passthrough[name] = t + stats["int8_payload_bytes"] += tensor_nbytes(t) + continue + if t.numel() <= INT8_KEEP_FLOAT_MAX_NUMEL: + kept = keep_float_tensor(name, t, passthrough_orig_dtypes) + passthrough[name] = kept + stats["int8_payload_bytes"] += tensor_nbytes(kept) + continue + stats["num_float_tensors"] += 1 + q, s = quantize_float_tensor(t) + if s.ndim > 0: + qmeta[name] = {"scheme": "per_row", "axis": 0} + quantized[name] = q + scales[name] = s + dtypes[name] = str(t.dtype).removeprefix("torch.") + stats["int8_payload_bytes"] += tensor_nbytes(q) + tensor_nbytes(s) + obj: dict[str, object] = { + "__quant_format__": "int8_clean_per_row_v1", + "quantized": quantized, + "scales": scales, + "dtypes": dtypes, + "passthrough": passthrough, + } + if qmeta: + obj["qmeta"] = qmeta + if passthrough_orig_dtypes: + obj["passthrough_orig_dtypes"] = passthrough_orig_dtypes + return obj, stats +def dequantize_state_dict_int8(obj: dict[str, object]) -> dict[str, Tensor]: + out: dict[str, Tensor] = {} + qmeta = obj.get("qmeta", {}) + passthrough_orig_dtypes = obj.get("passthrough_orig_dtypes", {}) + for name, q in obj["quantized"].items(): + dtype = getattr(torch, obj["dtypes"][name]) + s = obj["scales"][name] + if qmeta.get(name, {}).get("scheme") == "per_row" or s.ndim > 0: + s = s.to(dtype=torch.float32) + out[name] = (q.float() * s.view(q.shape[0], *([1] * (q.ndim - 1)))).to(dtype=dtype).contiguous() + else: + scale = float(s.item()) + out[name] = (q.float() * scale).to(dtype=dtype).contiguous() + for name, t in obj["passthrough"].items(): + out_t = t.detach().to("cpu").contiguous() + orig_dtype = passthrough_orig_dtypes.get(name) + if isinstance(orig_dtype, str): + out_t = out_t.to(dtype=getattr(torch, orig_dtype)).contiguous() + out[name] = out_t + return out +def load_data_shard(file: Path) -> Tensor: + header_bytes = 256 * np.dtype(" None: + self.file_idx = (self.file_idx + 1) % len(self.files) + self.tokens = load_data_shard(self.files[self.file_idx]) + self.pos = 0 + def take(self, n: int) -> Tensor: + chunks: list[Tensor] = [] + remaining = n + while remaining > 0: + avail = self.tokens.numel() - self.pos + if avail <= 0: + self._advance_file() + continue + k = min(remaining, avail) + chunks.append(self.tokens[self.pos : self.pos + k]) + self.pos += k + remaining -= k + return chunks[0] if len(chunks) == 1 else torch.cat(chunks) +class DistributedTokenLoader: + def __init__(self, pattern: str, rank: int, world_size: int, device: torch.device): + self.rank = rank + self.world_size = world_size + self.device = device + self.stream = TokenStream(pattern) + def next_batch(self, global_tokens: int, seq_len: int, grad_accum_steps: int) -> tuple[Tensor, Tensor]: + local_tokens = global_tokens // (self.world_size * grad_accum_steps) + per_rank_span = local_tokens + 1 + chunk = self.stream.take(per_rank_span * self.world_size) + start = self.rank * per_rank_span + local = chunk[start : start + per_rank_span].to(dtype=torch.int64) + x = local[:-1].reshape(-1, seq_len) + y = local[1:].reshape(-1, seq_len) + return x.to(self.device, non_blocking=True), y.to(self.device, non_blocking=True) +class RMSNorm(nn.Module): + def __init__(self, eps: float | None = None): + super().__init__() + self.eps = eps + def forward(self, x: Tensor) -> Tensor: + return F.rms_norm(x, (x.size(-1),), eps=self.eps) +class CastedLinear(nn.Linear): + _qat_enabled: bool = False + def forward(self, x: Tensor) -> Tensor: + w = self.weight.to(x.dtype) + if CastedLinear._qat_enabled and self.training and w.ndim == 2: + with torch.no_grad(): + w32 = self.weight.float() + # Use 99.95th percentile clipping to match GPTQ export quantizer + row_clip = torch.quantile(w32.abs(), 0.9995, dim=1) + scale = (row_clip / 31.0).clamp_min(1.0 / 31.0) + w_q = (torch.clamp(torch.round(w32 / scale[:, None]), -32, 31) * scale[:, None]).to(x.dtype) + w = w + (w_q - w).detach() + bias = self.bias.to(x.dtype) if self.bias is not None else None + return F.linear(x, w, bias) +def restore_low_dim_params_to_fp32(module: nn.Module) -> None: + with torch.no_grad(): + for name, param in module.named_parameters(): + if (param.ndim < 2 or any(pattern in name for pattern in CONTROL_TENSOR_NAME_PATTERNS)) and param.dtype != torch.float32: + param.data = param.data.float() +class Rotary(nn.Module): + def __init__(self, dim: int, base: float = 10000.0, train_seq_len: int = 1024, rope_dims: int = 0): + super().__init__() + self.dim = dim + self.base = base + self.train_seq_len = train_seq_len + self.rope_dims = rope_dims if rope_dims > 0 else dim + inv_freq = 1.0 / (base ** (torch.arange(0, self.rope_dims, 2, dtype=torch.float32) / self.rope_dims)) + self.register_buffer("inv_freq", inv_freq, persistent=False) + self._seq_len_cached = 0 + self._cos_cached: Tensor | None = None + self._sin_cached: Tensor | None = None + def forward(self, seq_len: int, device: torch.device, dtype: torch.dtype) -> tuple[Tensor, Tensor]: + if ( + self._cos_cached is None + or self._sin_cached is None + or self._seq_len_cached != seq_len + or self._cos_cached.device != device + ): + rd = self.rope_dims + if seq_len > self.train_seq_len: + scale = seq_len / self.train_seq_len + new_base = self.base * (scale ** (rd / (rd - 2))) + inv_freq = 1.0 / (new_base ** (torch.arange(0, rd, 2, dtype=torch.float32, device=device) / rd)) + else: + inv_freq = self.inv_freq.to(device) + t = torch.arange(seq_len, device=device, dtype=inv_freq.dtype) + freqs = torch.outer(t, inv_freq) + self._cos_cached = freqs.cos()[None, :, None, :] + self._sin_cached = freqs.sin()[None, :, None, :] + self._seq_len_cached = seq_len + return self._cos_cached.to(dtype=dtype), self._sin_cached.to(dtype=dtype) +def apply_rotary_emb(x: Tensor, cos: Tensor, sin: Tensor, rope_dims: int = 0) -> Tensor: + if rope_dims > 0 and rope_dims < x.size(-1): + x_rope, x_pass = x[..., :rope_dims], x[..., rope_dims:] + half = rope_dims // 2 + x1, x2 = x_rope[..., :half], x_rope[..., half:] + x_rope = torch.cat((x1 * cos + x2 * sin, x1 * (-sin) + x2 * cos), dim=-1) + return torch.cat((x_rope, x_pass), dim=-1) + half = x.size(-1) // 2 + x1, x2 = x[..., :half], x[..., half:] + return torch.cat((x1 * cos + x2 * sin, x1 * (-sin) + x2 * cos), dim=-1) +class CausalSelfAttention(nn.Module): + def __init__( + self, + dim: int, + num_heads: int, + num_kv_heads: int, + rope_base: float, + qk_gain_init: float, + ): + super().__init__() + if dim % num_heads != 0: + raise ValueError("model_dim must be divisible by num_heads") + if num_heads % num_kv_heads != 0: + raise ValueError("num_heads must be divisible by num_kv_heads") + self.num_heads = num_heads + self.num_kv_heads = num_kv_heads + self.head_dim = dim // num_heads + if self.head_dim % 2 != 0: + raise ValueError("head_dim must be even for RoPE") + kv_dim = self.num_kv_heads * self.head_dim + self.c_q = CastedLinear(dim, dim, bias=False) + self.c_k = CastedLinear(dim, kv_dim, bias=False) + self.c_v = CastedLinear(dim, kv_dim, bias=False) + self.proj = CastedLinear(dim, dim, bias=False) + self.proj._zero_init = True + self.q_gain = nn.Parameter(torch.full((num_heads,), qk_gain_init, dtype=torch.float32)) + self.rope_dims = 0 # set by GPT.__init__ for partial RoPE + self.rotary = Rotary(self.head_dim, base=rope_base, train_seq_len=1024) + self.use_xsa = False # set by GPT.__init__ for deep layers only + def _xsa_efficient(self, y: Tensor, v: Tensor) -> Tensor: + """Efficient XSA: subtract self-value projection via GQA-aware reshape (no repeat_interleave). + y: [B, T, H, D], v: [B, T, Hkv, D]. H must be divisible by Hkv.""" + B, T, H, D = y.shape + Hkv = v.size(-2) + group = H // Hkv + y_g = y.reshape(B, T, Hkv, group, D) # [B, T, Hkv, group, D] + vn = F.normalize(v, dim=-1).unsqueeze(-2) # [B, T, Hkv, 1, D] — broadcast ready + proj = (y_g * vn).sum(dim=-1, keepdim=True) * vn + return (y_g - proj).reshape(B, T, H, D) + def forward(self, x: Tensor, v_embed: Tensor | None = None) -> Tensor: + bsz, seqlen, dim = x.shape + q = self.c_q(x).reshape(bsz, seqlen, self.num_heads, self.head_dim) + k = self.c_k(x).reshape(bsz, seqlen, self.num_kv_heads, self.head_dim) + v = self.c_v(x) + if v_embed is not None: + v = v + v_embed + v = v.reshape(bsz, seqlen, self.num_kv_heads, self.head_dim) + q = F.rms_norm(q, (q.size(-1),)) + k = F.rms_norm(k, (k.size(-1),)) + cos, sin = self.rotary(seqlen, x.device, q.dtype) + q = apply_rotary_emb(q, cos, sin, self.rope_dims) + k = apply_rotary_emb(k, cos, sin, self.rope_dims) + q = q * self.q_gain.to(dtype=q.dtype)[None, None, :, None] + y = flash_attn_3_func(q, k, v, causal=True) + if self.use_xsa: + y = self._xsa_efficient(y, v) + y = y.reshape(bsz, seqlen, dim) + return self.proj(y) +class SmearGate(nn.Module): + def __init__(self, dim: int): + super().__init__() + self.gate = nn.Parameter(torch.zeros(dim, dtype=torch.float32)) + def forward(self, x: Tensor) -> Tensor: + g = torch.sigmoid(self.gate.to(dtype=x.dtype))[None, None, :] + x_prev = torch.cat([torch.zeros_like(x[:, :1]), x[:, :-1]], dim=1) + return (1 - g) * x + g * x_prev +class BigramHashEmbedding(nn.Module): + def __init__(self, bigram_vocab_size: int, bigram_dim: int, model_dim: int): + super().__init__() + self.bigram_vocab_size = bigram_vocab_size + self.embed = nn.Embedding(bigram_vocab_size, bigram_dim) + nn.init.zeros_(self.embed.weight) + self.proj = CastedLinear(bigram_dim, model_dim, bias=False) if bigram_dim != model_dim else None + if self.proj is not None: + nn.init.zeros_(self.proj.weight) + self.scale = nn.Parameter(torch.tensor(0.05, dtype=torch.float32)) + def bigram_hash(self, tokens: Tensor) -> Tensor: + t = tokens.to(torch.int32) + mod = self.bigram_vocab_size - 1 + out = torch.empty_like(t) + out[..., 0] = mod + out[..., 1:] = torch.bitwise_xor(36313 * t[..., 1:], 27191 * t[..., :-1]) % mod + return out.long() + def forward(self, token_ids: Tensor) -> Tensor: + h = self.embed(self.bigram_hash(token_ids)) + if self.proj is not None: + h = self.proj(h) + return h * self.scale.to(dtype=h.dtype) +class ValueEmbedding(nn.Module): + """Reinject token identity into attention values at specific layers. + Each table maps vocab tokens to a low-dim embedding, projected to model_dim.""" + def __init__(self, vocab_size: int, ve_dim: int, model_dim: int): + super().__init__() + self.embed = nn.Embedding(vocab_size, ve_dim) + nn.init.normal_(self.embed.weight, std=0.01) + self.proj = CastedLinear(ve_dim, model_dim, bias=False) if ve_dim != model_dim else None + if self.proj is not None: + nn.init.zeros_(self.proj.weight) + self.scale = nn.Parameter(torch.tensor(0.1, dtype=torch.float32)) + def forward(self, token_ids: Tensor) -> Tensor: + h = self.embed(token_ids) + if self.proj is not None: + h = self.proj(h) + return h * self.scale.to(dtype=h.dtype) +class MLP(nn.Module): + def __init__(self, dim: int, mlp_mult: int, mlp_act: str = "relu_sq", mlp_leaky_slope: float = 0.5): + super().__init__() + hidden = int(mlp_mult * dim) + self.fc = CastedLinear(dim, hidden, bias=False) + self.proj = CastedLinear(hidden, dim, bias=False) + self.proj._zero_init = True + self.mlp_act = mlp_act + self.mlp_leaky_slope = mlp_leaky_slope + if self.mlp_act not in {"relu_sq", "leaky_relu_sq"}: + raise ValueError(f"Unsupported MLP_ACT '{self.mlp_act}'. Use 'relu_sq' or 'leaky_relu_sq'.") + def forward(self, x: Tensor) -> Tensor: + x = self.fc(x) + if self.mlp_act == "leaky_relu_sq": + x = F.leaky_relu(x, negative_slope=self.mlp_leaky_slope) + else: + x = F.relu(x) + return self.proj(x.square()) +class Block(nn.Module): + def __init__( + self, + dim: int, + num_heads: int, + num_kv_heads: int, + mlp_mult: int, + rope_base: float, + qk_gain_init: float, + layer_idx: int = 0, + ln_scale: bool = False, + dtg: bool = False, + mlp_act: str = "relu_sq", + mlp_leaky_slope: float = 0.5, + ): + super().__init__() + self.attn_norm = RMSNorm() + self.mlp_norm = RMSNorm() + self.attn = CausalSelfAttention(dim, num_heads, num_kv_heads, rope_base, qk_gain_init) + self.mlp = MLP(dim, mlp_mult, mlp_act=mlp_act, mlp_leaky_slope=mlp_leaky_slope) + self.attn_scale = nn.Parameter(torch.ones(dim, dtype=torch.float32)) + self.mlp_scale = nn.Parameter(torch.ones(dim, dtype=torch.float32)) + self.resid_mix = nn.Parameter(torch.stack((torch.ones(dim), torch.zeros(dim))).float()) + self.ln_scale_factor = 1.0 / math.sqrt(layer_idx + 1) if ln_scale else 1.0 + if dtg: + self.dtg_gate = nn.Linear(dim, 1, bias=True) + nn.init.zeros_(self.dtg_gate.weight) + nn.init.constant_(self.dtg_gate.bias, 2.0) + else: + self.dtg_gate = None + def forward(self, x: Tensor, x0: Tensor, v_embed: Tensor | None = None) -> Tensor: + mix = self.resid_mix.to(dtype=x.dtype) + x_in = mix[0][None, None, :] * x + mix[1][None, None, :] * x0 + attn_out = self.attn(self.attn_norm(x_in) * self.ln_scale_factor, v_embed=v_embed) + x_out = x_in + self.attn_scale.to(dtype=x_in.dtype)[None, None, :] * attn_out + x_out = x_out + self.mlp_scale.to(dtype=x_out.dtype)[None, None, :] * self.mlp(self.mlp_norm(x_out) * self.ln_scale_factor) + if self.dtg_gate is not None: + gate = torch.sigmoid(self.dtg_gate(x_in.detach())) + x_out = x_in + gate * (x_out - x_in) + return x_out +class GPT(nn.Module): + def __init__( + self, + vocab_size: int, + num_layers: int, + model_dim: int, + num_heads: int, + num_kv_heads: int, + mlp_mult: int, + tie_embeddings: bool, + tied_embed_init_std: float, + logit_softcap: float, + rope_base: float, + qk_gain_init: float, + mtp_num_heads: int = 0, + mtp_loss_weight: float = 0.1, + bigram_vocab_size: int = 0, + bigram_dim: int = 128, + xsa_last_n: int = 0, + rope_dims: int = 0, + ln_scale: bool = False, + dtg: bool = False, + ve_enabled: bool = False, + ve_dim: int = 128, + ve_layers: str = "9,10", + mlp_act: str = "relu_sq", + mlp_leaky_slope: float = 0.5, + f1_corr_rank: int = 0, + f1_corr_scale_init: float = 0.10, + ): + super().__init__() + self._ve_target_dim = num_kv_heads * (model_dim // num_heads) # kv_dim for value projection + if logit_softcap <= 0.0: + raise ValueError(f"logit_softcap must be positive, got {logit_softcap}") + self.tie_embeddings = tie_embeddings + self.tied_embed_init_std = tied_embed_init_std + self.logit_softcap = logit_softcap + self.mtp_num_heads = mtp_num_heads + self.mtp_loss_weight = mtp_loss_weight + self.tok_emb = nn.Embedding(vocab_size, model_dim) + self.bigram = BigramHashEmbedding(bigram_vocab_size, bigram_dim, model_dim) if bigram_vocab_size > 0 else None + self.smear = SmearGate(model_dim) + self.num_encoder_layers = num_layers // 2 + self.num_decoder_layers = num_layers - self.num_encoder_layers + self.num_skip_weights = min(self.num_encoder_layers, self.num_decoder_layers) + self.skip_weights = nn.Parameter(torch.ones(self.num_skip_weights, model_dim, dtype=torch.float32)) + self.blocks = nn.ModuleList( + [ + Block( + model_dim, + num_heads, + num_kv_heads, + mlp_mult, + rope_base, + qk_gain_init, + layer_idx=i, + ln_scale=ln_scale, + dtg=dtg, + mlp_act=mlp_act, + mlp_leaky_slope=mlp_leaky_slope, + ) + for i in range(num_layers) + ] + ) + if rope_dims > 0: + head_dim = model_dim // num_heads + for block in self.blocks: + block.attn.rope_dims = rope_dims + block.attn.rotary = Rotary(head_dim, base=rope_base, train_seq_len=1024, rope_dims=rope_dims) + self.ve_layer_indices = [int(x) for x in ve_layers.split(",") if x.strip()] if ve_enabled else [] + kv_dim = self._ve_target_dim + if self.ve_layer_indices: + self.ve_shared = ValueEmbedding(vocab_size, ve_dim, kv_dim) + self.ve_layer_scales = nn.ParameterList( + [nn.Parameter(torch.ones(1, dtype=torch.float32)) for _ in self.ve_layer_indices] + ) + else: + self.ve_shared = None + self.ve_layer_scales = nn.ParameterList() + self.value_embeds = nn.ModuleList() # keep empty for compat + self.final_norm = RMSNorm() + self.lm_head = None if tie_embeddings else CastedLinear(model_dim, vocab_size, bias=False) + if self.lm_head is not None: + self.lm_head._zero_init = True + self.mtp_heads = nn.ModuleList( + [CastedLinear(model_dim, vocab_size, bias=False) for _ in range(mtp_num_heads)] + ) + for head in self.mtp_heads: + head._zero_init = True + # Low-rank correction path for extra capacity under size budget. + self.f1_corr_rank = f1_corr_rank + if f1_corr_rank > 0: + self.f1_corr_in = CastedLinear(model_dim, f1_corr_rank, bias=False) + self.f1_corr_out = CastedLinear(f1_corr_rank, vocab_size, bias=False) + self.f1_corr_out._zero_init = True + self.f1_corr_scale = nn.Parameter(torch.tensor(f1_corr_scale_init, dtype=torch.float32)) + else: + self.f1_corr_in = None + self.f1_corr_out = None + self.f1_corr_scale = None + if xsa_last_n > 0: + for i in range(max(0, num_layers - xsa_last_n), num_layers): + self.blocks[i].attn.use_xsa = True + self._init_weights() + def _init_weights(self) -> None: + if self.tie_embeddings: + nn.init.normal_(self.tok_emb.weight, mean=0.0, std=self.tied_embed_init_std) + num_layers = len(self.blocks) + for name, module in self.named_modules(): + if isinstance(module, nn.Linear): + if getattr(module, "_zero_init", False): + nn.init.zeros_(module.weight) + elif module.weight.ndim == 2 and module.weight.shape[0] >= 64 and module.weight.shape[1] >= 64: + nn.init.orthogonal_(module.weight, gain=1.0) + if ".proj." in name or name.endswith(".proj"): + with torch.no_grad(): + module.weight.mul_(1.0 / math.sqrt(2 * num_layers)) + def _get_ve(self, layer_idx: int, input_ids: Tensor, ve_cache: dict | None = None) -> Tensor | None: + """Get value embedding for a specific layer using shared table + per-layer scale.""" + if self.ve_shared is None or layer_idx not in self.ve_layer_indices: + return None + if ve_cache is not None and 've' not in ve_cache: + ve_cache['ve'] = self.ve_shared(input_ids) + ve_base = ve_cache['ve'] if ve_cache is not None else self.ve_shared(input_ids) + ve_idx = self.ve_layer_indices.index(layer_idx) + return ve_base * self.ve_layer_scales[ve_idx].to(dtype=ve_base.dtype) + def forward(self, input_ids: Tensor, target_ids: Tensor) -> Tensor: + x = self.tok_emb(input_ids) + if self.bigram is not None: + x = x + self.bigram(input_ids) + x = F.rms_norm(x, (x.size(-1),)) + x = self.smear(x) + x0 = x + skips: list[Tensor] = [] + ve_cache: dict = {} + for i in range(self.num_encoder_layers): + ve = self._get_ve(i, input_ids, ve_cache) + x = self.blocks[i](x, x0, v_embed=ve) + skips.append(x) + for i in range(self.num_decoder_layers): + bi = self.num_encoder_layers + i + if skips: + x = x + self.skip_weights[i].to(dtype=x.dtype)[None, None, :] * skips.pop() + ve = self._get_ve(bi, input_ids, ve_cache) + x = self.blocks[bi](x, x0, v_embed=ve) + x = self.final_norm(x) + x_flat = x.reshape(-1, x.size(-1)) + targets = target_ids.reshape(-1) + if self.tie_embeddings: + logits_proj = F.linear(x_flat, self.tok_emb.weight) + else: + if self.lm_head is None: + raise RuntimeError("lm_head is required when tie_embeddings=False") + logits_proj = self.lm_head(x_flat) + if self.f1_corr_in is not None and self.f1_corr_out is not None and self.f1_corr_scale is not None: + corr_hidden = F.silu(self.f1_corr_in(x_flat)) + corr_proj = self.f1_corr_out(corr_hidden) + logits_proj = logits_proj + self.f1_corr_scale.to(dtype=logits_proj.dtype) * corr_proj + logits = self.logit_softcap * torch.tanh(logits_proj / self.logit_softcap) + main_loss = F.cross_entropy(logits.float(), targets, reduction="mean") + if self.training and self.mtp_num_heads > 0 and self.mtp_loss_weight > 0.0: + _, seqlen, dim = x.shape + mtp_loss_sum = x.new_zeros(()) + mtp_loss_count = 0 + for k, mtp_head in enumerate(self.mtp_heads): + valid_t = seqlen - (k + 1) + if valid_t <= 0: + continue + mtp_hidden = x[:, :valid_t, :].reshape(-1, dim) + mtp_targets = target_ids[:, k + 1 :].reshape(-1) + mtp_logits_proj = mtp_head(mtp_hidden) + mtp_logits = self.logit_softcap * torch.tanh(mtp_logits_proj / self.logit_softcap) + mtp_loss_sum = mtp_loss_sum + F.cross_entropy(mtp_logits.float(), mtp_targets, reduction="mean") + mtp_loss_count += 1 + if mtp_loss_count > 0: + main_loss = main_loss + self.mtp_loss_weight * (mtp_loss_sum / mtp_loss_count) + return main_loss + def forward_logits(self, input_ids: Tensor) -> Tensor: + """Return logits (bsz, seq_len, vocab) without computing loss.""" + x = self.tok_emb(input_ids) + if self.bigram is not None: + x = x + self.bigram(input_ids) + x = F.rms_norm(x, (x.size(-1),)) + x = self.smear(x) + x0 = x + skips: list[Tensor] = [] + ve_cache: dict = {} + for i in range(self.num_encoder_layers): + ve = self._get_ve(i, input_ids, ve_cache) + x = self.blocks[i](x, x0, v_embed=ve) + skips.append(x) + for i in range(self.num_decoder_layers): + bi = self.num_encoder_layers + i + if skips: + x = x + self.skip_weights[i].to(dtype=x.dtype)[None, None, :] * skips.pop() + ve = self._get_ve(bi, input_ids, ve_cache) + x = self.blocks[bi](x, x0, v_embed=ve) + x = self.final_norm(x) + if self.tie_embeddings: + logits_proj = F.linear(x, self.tok_emb.weight) + else: + logits_proj = self.lm_head(x) + if self.f1_corr_in is not None and self.f1_corr_out is not None and self.f1_corr_scale is not None: + corr_hidden = F.silu(self.f1_corr_in(x)) + corr_proj = self.f1_corr_out(corr_hidden) + logits_proj = logits_proj + self.f1_corr_scale.to(dtype=logits_proj.dtype) * corr_proj + return self.logit_softcap * torch.tanh(logits_proj / self.logit_softcap) +def eval_val_sliding( + args: Hyperparameters, + base_model: nn.Module, + rank: int, + world_size: int, + device: torch.device, + val_tokens: Tensor, + base_bytes_lut: Tensor, + has_leading_space_lut: Tensor, + is_boundary_token_lut: Tensor, + stride: int, + batch_seqs: int = 32, + eval_seq_len: int | None = None, +) -> tuple[float, float]: + """Sliding window evaluation: each token scored with maximum context.""" + seq_len = eval_seq_len or args.train_seq_len + total_tokens = val_tokens.numel() - 1 + window_starts = [ws for ws in range(0, total_tokens, stride) + if min(ws + seq_len, total_tokens) - ws >= 1] + total_windows = len(window_starts) + my_s = (total_windows * rank) // world_size + my_e = (total_windows * (rank + 1)) // world_size + my_windows = window_starts[my_s:my_e] + loss_sum = torch.zeros((), device=device, dtype=torch.float64) + token_count = torch.zeros((), device=device, dtype=torch.float64) + byte_count = torch.zeros((), device=device, dtype=torch.float64) + base_model.eval() + compiled_logits = maybe_torch_compile(base_model.forward_logits, args) + with torch.inference_mode(): + for bi in range(0, len(my_windows), batch_seqs): + batch_ws = my_windows[bi:bi + batch_seqs] + bsz = len(batch_ws) + x_batch = torch.zeros(bsz, seq_len, dtype=torch.int64, device=device) + y_batch = torch.zeros(bsz, seq_len, dtype=torch.int64, device=device) + wlens: list[int] = [] + for i, ws in enumerate(batch_ws): + end = min(ws + seq_len, total_tokens) + wlen = end - ws + wlens.append(wlen) + chunk = val_tokens[ws:end + 1].to(dtype=torch.int64, device=device) + x_batch[i, :wlen] = chunk[:-1] + y_batch[i, :wlen] = chunk[1:] + with torch.autocast(device_type="cuda", dtype=torch.bfloat16): + logits = compiled_logits(x_batch) + nll = F.cross_entropy( + logits.reshape(-1, logits.size(-1)).float(), + y_batch.reshape(-1), + reduction="none", + ).reshape(bsz, seq_len) + for i, ws in enumerate(batch_ws): + wlen = wlens[i] + s = 0 if ws == 0 else max(wlen - stride, 0) + scored_nll = nll[i, s:wlen].to(torch.float64) + loss_sum += scored_nll.sum() + token_count += float(wlen - s) + tgt = y_batch[i, s:wlen] + prev = x_batch[i, s:wlen] + tb = base_bytes_lut[tgt].to(torch.float64) + tb += (has_leading_space_lut[tgt] & ~is_boundary_token_lut[prev]).to(torch.float64) + byte_count += tb.sum() + if dist.is_available() and dist.is_initialized(): + dist.all_reduce(loss_sum, op=dist.ReduceOp.SUM) + dist.all_reduce(token_count, op=dist.ReduceOp.SUM) + dist.all_reduce(byte_count, op=dist.ReduceOp.SUM) + val_loss = (loss_sum / token_count).item() + bits_per_token = val_loss / math.log(2.0) + tokens_per_byte = token_count.item() / byte_count.item() + base_model.train() + return val_loss, bits_per_token * tokens_per_byte +def eval_val_sliding_hashed_ngram( + args: Hyperparameters, + base_model: nn.Module, + rank: int, + world_size: int, + device: torch.device, + val_tokens: Tensor, + base_bytes_lut: Tensor, + has_leading_space_lut: Tensor, + is_boundary_token_lut: Tensor, + stride: int, + order: int, + alpha: float, + min_count: int, + buckets: int, + max_seconds: float = 0.0, + batch_seqs: int = 32, + eval_seq_len: int | None = None, +) -> tuple[float, float, float]: + """Score-first sliding eval with fixed-weight hashed n-gram interpolation. + + Legal behavior: + - per-token score is computed before that token updates the cache + - no target-aware gating/min-NLL selection is used + """ + if order < 2: + raise ValueError(f"NGRAM_EVAL_ORDER must be >=2, got {order}") + if not (0.0 <= alpha <= 1.0): + raise ValueError(f"NGRAM_EVAL_ALPHA must be in [0, 1], got {alpha}") + if min_count < 1: + raise ValueError(f"NGRAM_EVAL_MIN_COUNT must be >=1, got {min_count}") + if buckets < 1024: + raise ValueError(f"NGRAM_EVAL_BUCKETS must be >=1024, got {buckets}") + if max_seconds < 0.0: + raise ValueError(f"NGRAM_EVAL_MAX_SECONDS must be >=0, got {max_seconds}") + + seq_len = eval_seq_len or args.train_seq_len + total_tokens = val_tokens.numel() - 1 + all_window_starts = [ws for ws in range(0, total_tokens, stride) if min(ws + seq_len, total_tokens) - ws >= 1] + total_scored_tokens = 0.0 + for ws in all_window_starts: + end = min(ws + seq_len, total_tokens) + wlen = end - ws + s = 0 if ws == 0 else max(wlen - stride, 0) + total_scored_tokens += float(max(wlen - s, 0)) + # Distribute windows across ranks + my_s = (len(all_window_starts) * rank) // world_size + my_e = (len(all_window_starts) * (rank + 1)) // world_size + window_starts = all_window_starts[my_s:my_e] + + val_np = val_tokens.numpy() + ctx_table = np.zeros((buckets,), dtype=np.uint32) + full_table = np.zeros((buckets,), dtype=np.uint32) + mask = np.uint64(buckets - 1) + primes = np.array( + [np.uint64(36313), np.uint64(27191), np.uint64(51647), np.uint64(81929), np.uint64(131071)], + dtype=np.uint64, + ) + + loss_sum = 0.0 + token_count = 0.0 + byte_count = 0.0 + + base_model.eval() + compiled_logits = maybe_torch_compile(base_model.forward_logits, args) + t0 = time.perf_counter() + deadline = (t0 + max_seconds) if max_seconds > 0.0 else None + cutoff_hit = False + with torch.inference_mode(): + for bi in range(0, len(window_starts), batch_seqs): + if deadline is not None and time.perf_counter() >= deadline: + cutoff_hit = True + break + batch_ws = window_starts[bi:bi + batch_seqs] + bsz = len(batch_ws) + x_batch = torch.zeros(bsz, seq_len, dtype=torch.int64, device=device) + y_batch = torch.zeros(bsz, seq_len, dtype=torch.int64, device=device) + wlens: list[int] = [] + for i, ws in enumerate(batch_ws): + end = min(ws + seq_len, total_tokens) + wlen = end - ws + wlens.append(wlen) + chunk = val_tokens[ws:end + 1].to(dtype=torch.int64, device=device) + x_batch[i, :wlen] = chunk[:-1] + y_batch[i, :wlen] = chunk[1:] + + with torch.autocast(device_type="cuda", dtype=torch.bfloat16): + logits = compiled_logits(x_batch) + nll = F.cross_entropy( + logits.reshape(-1, logits.size(-1)).float(), + y_batch.reshape(-1), + reduction="none", + ).reshape(bsz, seq_len) + + for i, ws in enumerate(batch_ws): + wlen = wlens[i] + s = 0 if ws == 0 else max(wlen - stride, 0) + seg_len = wlen - s + if seg_len <= 0: + continue + + seg_nll = nll[i, s:wlen].to(torch.float64).cpu().numpy() + seg_model_p = np.exp(-seg_nll) + + global_j = np.arange(ws + s + 1, ws + wlen + 1, dtype=np.int64) + valid = global_j >= (order - 1) + if valid.any(): + v_idx = np.nonzero(valid)[0] + jv = global_j[v_idx] + + ctx_hash = np.zeros((len(jv),), dtype=np.uint64) + ctx_width = order - 1 + for k in range(ctx_width): + tok = val_np[jv - (ctx_width - k)].astype(np.uint64) + ctx_hash ^= tok * primes[k % len(primes)] + ctx_key = (ctx_hash & mask).astype(np.int64) + + tgt_np = val_np[jv].astype(np.uint64) + full_key = ((ctx_hash ^ (tgt_np * primes[ctx_width % len(primes)])) & mask).astype(np.int64) + + ctx_counts = ctx_table[ctx_key].astype(np.float64) + full_counts = full_table[full_key].astype(np.float64) + can_mix = ctx_counts >= float(min_count) + if can_mix.any(): + # Collision-safe estimate: ensure n-gram probability stays in [0, 1]. + # With hashed sketches, full_counts can exceed ctx_counts due collisions. + p_ng = np.minimum(full_counts, ctx_counts) / np.maximum(ctx_counts, 1.0) + p_ng = np.clip(p_ng, 0.0, 1.0) + mixed = (1.0 - alpha) * seg_model_p[v_idx] + alpha * p_ng + seg_model_p[v_idx[can_mix]] = mixed[can_mix] + seg_nll = -np.log(np.clip(seg_model_p, 1e-12, 1.0)) + + # Score-first legality: update cache only after segment scoring. + np.add.at(ctx_table, ctx_key, 1) + np.add.at(full_table, full_key, 1) + + loss_sum += float(seg_nll.sum()) + token_count += float(seg_len) + tgt = y_batch[i, s:wlen] + prev = x_batch[i, s:wlen] + tb = base_bytes_lut[tgt].to(torch.float64) + tb += (has_leading_space_lut[tgt] & ~is_boundary_token_lut[prev]).to(torch.float64) + byte_count += float(tb.sum().item()) + + if (bi // batch_seqs) % 2000 == 0 and bi > 0: + elapsed = time.perf_counter() - t0 + prog = min((bi + bsz) / max(len(window_starts), 1), 1.0) + cur_bpb = (loss_sum / max(token_count, 1.0)) / math.log(2.0) * (token_count / max(byte_count, 1.0)) + print( + f"ngram_eval:progress windows={bi + bsz}/{len(window_starts)} " + f"({prog*100:.1f}%) bpb={cur_bpb:.6f} t={elapsed:.0f}s", + flush=True, + ) + # All-reduce across ranks + _loss = torch.tensor(loss_sum, device=device, dtype=torch.float64) + _toks = torch.tensor(token_count, device=device, dtype=torch.float64) + _bytes = torch.tensor(byte_count, device=device, dtype=torch.float64) + if dist.is_available() and dist.is_initialized(): + dist.all_reduce(_loss, op=dist.ReduceOp.SUM) + dist.all_reduce(_toks, op=dist.ReduceOp.SUM) + dist.all_reduce(_bytes, op=dist.ReduceOp.SUM) + loss_sum = _loss.item() + token_count = _toks.item() + byte_count = _bytes.item() + + coverage = token_count / max(total_scored_tokens, 1.0) + if cutoff_hit: + elapsed = time.perf_counter() - t0 + print( + f"ngram_eval:cutoff max_seconds={max_seconds:.1f} " + f"coverage={coverage*100:.2f}% elapsed={elapsed:.0f}s", + flush=True, + ) + + val_loss = loss_sum / max(token_count, 1.0) + val_bpb = val_loss / math.log(2.0) * (token_count / max(byte_count, 1.0)) + base_model.train() + return val_loss, val_bpb, coverage +def _classify_param(name: str) -> str: + if "tok_emb" in name or "lm_head" in name: + return "embed" + if "f1_corr_in" in name or "f1_corr_out" in name: + return "aux" + if ".mlp." in name: + return "mlp" + if ".attn." in name or (".proj." in name and ".mlp." not in name): + return "attn" + return "other" +def eval_val_sliding_ttt( + args: Hyperparameters, base_model: nn.Module, rank: int, world_size: int, + device: torch.device, val_tokens: Tensor, base_bytes_lut: Tensor, + has_leading_space_lut: Tensor, is_boundary_token_lut: Tensor, + stride: int, batch_seqs: int = 32, +) -> tuple[float, float]: + seq_len, total_tokens, ttt_chunk = args.train_seq_len, val_tokens.numel() - 1, args.ttt_chunk_tokens + master = (rank == 0) + log0 = (lambda msg: print(msg, flush=True)) if master else (lambda msg: None) + window_starts = [ws for ws in range(0, total_tokens, stride) if min(ws + seq_len, total_tokens) - ws >= stride or ws == 0] + num_chunks = (total_tokens + ttt_chunk - 1) // ttt_chunk + chunk_windows: list[list[int]] = [[] for _ in range(num_chunks)] + for ws in window_starts: + end, wlen = min(ws + seq_len, total_tokens), min(ws + seq_len, total_tokens) - ws + s = 0 if ws == 0 else max(wlen - stride, 0) + chunk_windows[min((ws + s) // ttt_chunk, num_chunks - 1)].append(ws) + log0(f"ttt_sliding:start chunks={num_chunks} windows={len(window_starts)} lr={args.ttt_lr} epochs={args.ttt_epochs} freeze={args.ttt_freeze_blocks}") + loss_sum = torch.zeros((), device=device, dtype=torch.float64) + token_count = torch.zeros((), device=device, dtype=torch.float64) + byte_count = torch.zeros((), device=device, dtype=torch.float64) + frozen_ids = set(range(min(args.ttt_freeze_blocks, len(base_model.blocks)))) + embed_names = {"tok_emb", "bigram", "ve_shared"} if args.ttt_freeze_embed else set() + ttt_params = [] + for name, p in base_model.named_parameters(): + if any(f"blocks.{bi}." in name for bi in frozen_ids): + p.requires_grad_(False) + elif any(en in name for en in embed_names): + p.requires_grad_(False) + else: + p.requires_grad_(True); ttt_params.append(p) + log0(f"ttt_sliding:unfrozen={sum(p.numel() for p in ttt_params)} freeze_embed={args.ttt_freeze_embed}") + optimizer = torch.optim.SGD(ttt_params, lr=args.ttt_lr, momentum=args.ttt_momentum) + # TTT-EMA: maintain smoothed weights for scoring + ema_decay = args.ttt_ema_decay + ema_state = None + raw_state = None + if ema_decay > 0: + ema_state = {n: p.data.clone() for n, p in base_model.named_parameters() if p.requires_grad} + raw_state = {n: torch.empty_like(p.data) for n, p in base_model.named_parameters() if n in ema_state} + log0(f"ttt_sliding:ema_decay={ema_decay} ema_params={len(ema_state)}") + t0 = time.perf_counter() + cur_lr = args.ttt_lr + for ci in range(num_chunks): + windows = chunk_windows[ci] + if not windows: + continue + chunk_start, chunk_end = ci * ttt_chunk, min((ci + 1) * ttt_chunk, total_tokens) + my_windows = windows[(len(windows) * rank) // world_size:(len(windows) * (rank + 1)) // world_size] + # Swap to EMA weights for scoring (if enabled and past first chunk) + if ema_state is not None and ci > 0: + for n, p in base_model.named_parameters(): + if n in ema_state: + raw_state[n].copy_(p.data) + p.data.copy_(ema_state[n]) + base_model.eval() + with torch.inference_mode(): + for bi in range(0, len(my_windows), batch_seqs): + batch_ws = my_windows[bi:bi + batch_seqs] + bsz = len(batch_ws) + x_batch = torch.zeros(bsz, seq_len, dtype=torch.int64, device=device) + y_batch = torch.zeros(bsz, seq_len, dtype=torch.int64, device=device) + wlens = [] + for i, ws in enumerate(batch_ws): + wlen = min(ws + seq_len, total_tokens) - ws; wlens.append(wlen) + ct = val_tokens[ws:ws + wlen + 1].to(dtype=torch.int64, device=device) + x_batch[i, :wlen] = ct[:-1]; y_batch[i, :wlen] = ct[1:] + with torch.autocast(device_type="cuda", dtype=torch.bfloat16): + logits = base_model.forward_logits(x_batch) + nll = F.cross_entropy( + logits.reshape(-1, logits.size(-1)).float(), + y_batch.reshape(-1), reduction="none", + ).reshape(bsz, seq_len) + for i, ws in enumerate(batch_ws): + wlen, s = wlens[i], 0 if ws == 0 else max(wlens[i] - stride, 0) + loss_sum += nll[i, s:wlen].to(torch.float64).sum(); token_count += float(wlen - s) + tgt, prev = y_batch[i, s:wlen], x_batch[i, s:wlen] + tb = base_bytes_lut[tgt].to(torch.float64) + (has_leading_space_lut[tgt] & ~is_boundary_token_lut[prev]).to(torch.float64) + byte_count += tb.sum() + # Restore raw weights after scoring (for training phase) + if ema_state is not None and ci > 0: + for n, p in base_model.named_parameters(): + if n in raw_state: + p.data.copy_(raw_state[n]) + # Phase 2: TRAIN on this chunk (already scored = legal) + if ci < num_chunks - 1 and ci < args.ttt_max_train_chunks and args.ttt_epochs > 0: + base_model.train() + chunk_seqs = (chunk_end - chunk_start) // seq_len + if chunk_seqs > 0: + cur_lr = args.ttt_lr * 0.5 * (1.0 + math.cos(math.pi * ci / max(args.ttt_max_train_chunks - 1, 1))) + for pg in optimizer.param_groups: + pg['lr'] = cur_lr + ms, me = (chunk_seqs * rank) // world_size, (chunk_seqs * (rank + 1)) // world_size + for _ep in range(args.ttt_epochs): + for bs in range(0, me - ms, args.ttt_batch_seqs): + be = min(bs + args.ttt_batch_seqs, me - ms) + start_tok = chunk_start + (ms + bs) * seq_len + end_tok = chunk_start + (ms + be) * seq_len + 1 + if end_tok > val_tokens.numel(): + continue + local = val_tokens[start_tok:end_tok].to(device=device, dtype=torch.int64) + x = local[:-1].reshape(-1, seq_len) + y = local[1:].reshape(-1, seq_len) + optimizer.zero_grad(set_to_none=True) + with torch.autocast(device_type="cuda", dtype=torch.bfloat16): + loss = base_model(x, y) + loss.backward() + if world_size > 1: + for p in ttt_params: + if p.grad is not None: + dist.all_reduce(p.grad, op=dist.ReduceOp.AVG) + torch.nn.utils.clip_grad_norm_(ttt_params, args.ttt_grad_clip) + optimizer.step() + # Update EMA after this chunk's training + if ema_state is not None: + with torch.no_grad(): + for n, p in base_model.named_parameters(): + if n in ema_state: + ema_state[n].mul_(ema_decay).add_(p.data, alpha=1.0 - ema_decay) + # Once training stops, load EMA weights permanently for remaining score-only chunks + if ema_state is not None and ci == args.ttt_max_train_chunks: + log0(f" ttt:loading EMA weights permanently at chunk {ci}") + for n, p in base_model.named_parameters(): + if n in ema_state: + p.data.copy_(ema_state[n]) + ema_state = None + raw_state = None + if master and (ci % 5 == 0 or ci == num_chunks - 1): + rl = loss_sum.item() / max(token_count.item(), 1) + cur_bpb = rl / math.log(2) * (token_count.item() / max(byte_count.item(), 1)) if token_count.item() > 0 else 0 + lr_str = f" lr={cur_lr:.6f}" if ci < args.ttt_max_train_chunks else " lr=done" + log0(f" ttt[{ci+1}/{num_chunks}] bpb={cur_bpb:.6f}{lr_str} t={time.perf_counter()-t0:.0f}s") + if dist.is_available() and dist.is_initialized(): + for t in [loss_sum, token_count, byte_count]: + dist.all_reduce(t, op=dist.ReduceOp.SUM) + val_loss = (loss_sum / token_count).item() + val_bpb = val_loss / math.log(2.0) * (token_count.item() / byte_count.item()) + for p in base_model.parameters(): + p.requires_grad_(True) + base_model.eval() + log0(f"ttt_sliding:done loss={val_loss:.6f} bpb={val_bpb:.6f} time={time.perf_counter()-t0:.0f}s") + return val_loss, val_bpb +# --------------------------------------------------------------------------- +# GPTQ: Hessian-aware quantization with column-wise error compensation +# --------------------------------------------------------------------------- +def _find_best_row_scales(W: Tensor, clip_range: int = 31) -> Tensor: + """Find optimal per-row scales by searching percentile clipping thresholds.""" + t32 = W.float() + best_s = t32.abs().amax(dim=1) / clip_range + best_s = best_s.clamp_min(1.0 / clip_range) + best_err = torch.full((t32.shape[0],), float('inf')) + for pct in [0.9990, 0.9995, 0.9999, 0.99999, 1.0]: + if pct < 1.0: + row_clip = torch.quantile(t32.abs(), pct, dim=1) + else: + row_clip = t32.abs().amax(dim=1) + s = (row_clip / clip_range).clamp_min(1.0 / clip_range) + q = torch.clamp(torch.round(t32 / s[:, None]), -clip_range, clip_range) + recon = q * s[:, None] + err = (t32 - recon).pow(2).mean(dim=1) + improved = err < best_err + best_s[improved] = s[improved] + best_err[improved] = err[improved] + return best_s +def gptq_quantize_weight(W: Tensor, H: Tensor, clip_range: int = 31, + block_size: int = 64, percdamp: float = 0.002) -> tuple[Tensor, Tensor]: + """GPTQ: quantize weight matrix W using Hessian H = X^T X for error compensation. + Uses pre-computed per-row scales and column reordering by Hessian diagonal. + Returns (quantized_int8, scale_fp16) in int6 range [-clip_range, clip_range].""" + W = W.float().clone() + rows, cols = W.shape + # Pre-compute optimal per-row scales from the original weight matrix + row_scale = _find_best_row_scales(W, clip_range) + H = H.float().clone() + damp = percdamp * H.diag().mean() + H.diagonal().add_(damp) + # Column reordering: process least-important columns first (ascending H_diag) + perm = torch.argsort(H.diag()) + invperm = torch.argsort(perm) + W = W[:, perm] + H = H[perm][:, perm] + try: + L = torch.linalg.cholesky(H) + Hinv = torch.cholesky_inverse(L) + except torch._C._LinAlgError: + Hinv = torch.diag(1.0 / H.diag().clamp_min(1e-6)) + Q = torch.zeros(rows, cols, dtype=torch.int8) + for i1 in range(0, cols, block_size): + i2 = min(i1 + block_size, cols) + W_block = W[:, i1:i2].clone() + Hinv_block = Hinv[i1:i2, i1:i2] + Err = torch.zeros_like(W_block) + for j in range(i2 - i1): + w_col = W_block[:, j] + h_inv_jj = Hinv_block[j, j].clamp_min(1e-8) + # Quantize using pre-computed per-row scales + q_col = torch.clamp(torch.round(w_col / row_scale), -clip_range, clip_range) + deq_col = q_col * row_scale + Q[:, i1 + j] = q_col.to(torch.int8) + err = (w_col - deq_col) / h_inv_jj + Err[:, j] = err + if j + 1 < i2 - i1: + W_block[:, j + 1:] -= err.unsqueeze(1) * Hinv_block[j, j + 1:].unsqueeze(0) + if i2 < cols: + W[:, i2:] -= Err @ Hinv[i1:i2, i2:] + # Undo column reordering + Q = Q[:, invperm] + return Q, row_scale.to(torch.float16) +def gptq_calibrate(model: nn.Module, train_pattern: str, device: torch.device, + n_samples: int = 256, seq_len: int = 2048) -> dict[str, Tensor]: + """Collect Hessian H = X^T X for each linear layer using training data.""" + hessians: dict[str, Tensor] = {} + n_seen: dict[str, int] = {} + hooks = [] + def make_hook(name: str): + def hook_fn(module, inp, out): + x = inp[0].detach().float() + if x.ndim == 3: + x = x.reshape(-1, x.shape[-1]) + if name not in hessians: + hessians[name] = torch.zeros(x.shape[1], x.shape[1], device=x.device, dtype=torch.float32) + n_seen[name] = 0 + hessians[name].addmm_(x.t(), x) + n_seen[name] += x.shape[0] + return hook_fn + for name, module in model.named_modules(): + if isinstance(module, (nn.Linear, CastedLinear)): + hooks.append(module.register_forward_hook(make_hook(name))) + stream = TokenStream(train_pattern) + model.eval() + with torch.no_grad(): + for _ in range(n_samples): + tokens = stream.take(seq_len + 1).to(device=device, dtype=torch.int64) + x = tokens[:-1].unsqueeze(0) + with torch.autocast(device_type="cuda", dtype=torch.bfloat16): + model.forward_logits(x) + for h in hooks: + h.remove() + for name in hessians: + hessians[name] /= max(n_seen[name], 1) + return hessians +def mixed_quantize_int6_gptq(state_dict: dict[str, Tensor], int6_cats: set[str], + hessians: dict[str, Tensor]) -> tuple[dict, dict]: + """Like mixed_quantize_int6 but uses GPTQ for int6 categories when Hessian available.""" + result: dict[str, Tensor] = {} + meta: dict[str, object] = {} + gptq_count, naive_count = 0, 0 + for name, tensor in state_dict.items(): + t = tensor.detach().cpu().contiguous() + cat = _classify_param(name) + if not t.is_floating_point() or t.numel() <= 65536: + result[name] = t.to(torch.float16) if t.is_floating_point() else t + meta[name] = "passthrough" + continue + if any(p in name for p in CONTROL_TENSOR_NAME_PATTERNS): + result[name] = t.float() + meta[name] = "passthrough_ctrl" + continue + if cat in int6_cats and t.ndim == 2: + module_name = name.rsplit(".weight", 1)[0] if name.endswith(".weight") else name + H = hessians.get(module_name) + if H is not None and H.shape[0] == t.shape[1]: + q, s = gptq_quantize_weight(t, H.cpu()) + gptq_count += 1 + else: + q, s = quantize_int6_per_row(t) + naive_count += 1 + result[name + ".q"] = q + result[name + ".scale"] = s + meta[name] = {"type": "int6"} + elif cat in int6_cats and t.ndim >= 1: + q, s = quantize_int6_per_row(t) + result[name + ".q"] = q + result[name + ".scale"] = s + meta[name] = {"type": "int6"} + naive_count += 1 + else: + q, s = quantize_float_tensor(t) + result[name + ".q"] = q + result[name + ".scale"] = s + meta[name] = {"type": "int8"} + print(f"gptq_quantize: {gptq_count} GPTQ layers, {naive_count} naive layers", flush=True) + return result, meta +def quantize_int6_per_row(t: Tensor, clip_range: int = 31) -> tuple[Tensor, Tensor]: + t32 = t.float() + if t32.ndim == 2: + best_q, best_s, best_err = None, None, float('inf') + for pct in [0.9990, 0.9995, 0.9999, 0.99999, 1.0]: + if pct < 1.0: + row_clip = torch.quantile(t32.abs(), pct, dim=1) + else: + row_clip = t32.abs().amax(dim=1) + s = (row_clip / clip_range).clamp_min(1.0 / clip_range).to(torch.float16) + q = torch.clamp(torch.round(t32 / s.float()[:, None]), -clip_range, clip_range).to(torch.int8) + recon = q.float() * s.float()[:, None] + err = (t32 - recon).pow(2).mean().item() + if err < best_err: + best_q, best_s, best_err = q, s, err + return best_q, best_s + amax = t32.abs().max().item() + scale = torch.tensor(amax / clip_range if amax > 0 else 1.0, dtype=torch.float16) + q = torch.clamp(torch.round(t32 / scale.float()), -clip_range, clip_range).to(torch.int8) + return q, scale +def mixed_quantize_int6(state_dict: dict[str, Tensor], int6_cats: set[str]): + num_layers_total = max( + (int(k.split(".")[1]) for k in state_dict if k.startswith("blocks.")), + default=0, + ) + 1 + late_k_layers = set(range(num_layers_total - 2, num_layers_total)) + result: dict[str, Tensor] = {} + meta: dict[str, object] = {} + for name, tensor in state_dict.items(): + t = tensor.detach().cpu().contiguous() + cat = _classify_param(name) + if not t.is_floating_point() or t.numel() <= 65536: + result[name] = t.to(torch.float16) if t.is_floating_point() else t + meta[name] = "passthrough" + continue + if any(p in name for p in CONTROL_TENSOR_NAME_PATTERNS): + result[name] = t.float() + meta[name] = "passthrough_ctrl" + continue + if cat in int6_cats and t.ndim >= 1: + q, s = quantize_int6_per_row(t) + result[name + ".q"] = q + result[name + ".scale"] = s + meta[name] = {"type": "int6"} + else: + q, s = quantize_float_tensor(t) + result[name + ".q"] = q + result[name + ".scale"] = s + meta[name] = {"type": "int8"} + return result, meta +def dequantize_mixed_int6(result: dict[str, Tensor], meta: dict[str, object], + template_sd: dict[str, Tensor]) -> dict[str, Tensor]: + out: dict[str, Tensor] = {} + for name, orig in template_sd.items(): + info = meta.get(name) + if info is None: + continue + orig_dtype = orig.dtype + if info in ("passthrough", "passthrough_ctrl", "passthrough_fp16"): + t = result[name] + if t.dtype == torch.float16 and orig_dtype in (torch.float32, torch.bfloat16): + t = t.to(orig_dtype) + out[name] = t + continue + q, s = result[name + ".q"], result[name + ".scale"] + if s.ndim > 0: + out[name] = (q.float() * s.float().view(q.shape[0], *([1] * (q.ndim - 1)))).to(orig_dtype) + else: + out[name] = (q.float() * float(s.item())).to(orig_dtype) + return out +def main() -> None: + global zeropower_via_newtonschulz5 + code = Path(__file__).read_text(encoding="utf-8") + args = Hyperparameters() + if args.compile_enabled: + zeropower_via_newtonschulz5 = torch.compile(zeropower_via_newtonschulz5) + distributed = "RANK" in os.environ and "WORLD_SIZE" in os.environ + rank = int(os.environ.get("RANK", "0")) + world_size = int(os.environ.get("WORLD_SIZE", "1")) + local_rank = int(os.environ.get("LOCAL_RANK", "0")) + if world_size <= 0: + raise ValueError(f"WORLD_SIZE must be positive, got {world_size}") + if 8 % world_size != 0: + raise ValueError(f"WORLD_SIZE={world_size} must divide 8 so grad_accum_steps stays integral") + grad_accum_steps = 8 // world_size + grad_scale = 1.0 / grad_accum_steps + if not torch.cuda.is_available(): + raise RuntimeError("CUDA is required") + device = torch.device("cuda", local_rank) + torch.cuda.set_device(device) + if distributed: + dist.init_process_group(backend="nccl", device_id=device) + dist.barrier() + master_process = rank == 0 + torch.backends.cuda.matmul.allow_tf32 = True + torch.backends.cudnn.allow_tf32 = True + from torch.backends.cuda import enable_cudnn_sdp, enable_flash_sdp, enable_math_sdp, enable_mem_efficient_sdp + enable_cudnn_sdp(False) + enable_flash_sdp(True) + enable_mem_efficient_sdp(False) + enable_math_sdp(False) + logfile = None + if master_process: + os.makedirs("logs", exist_ok=True) + logfile = f"logs/{args.run_id}.txt" + print(logfile) + def log0(msg: str, console: bool = True) -> None: + if not master_process: + return + if console: + print(msg) + if logfile is not None: + with open(logfile, "a", encoding="utf-8") as f: + print(msg, file=f) + log0(code, console=False) + log0("=" * 100, console=False) + log0(f"Running Python {sys.version}", console=False) + log0(f"Running PyTorch {torch.__version__}", console=False) + log0( + subprocess.run(["nvidia-smi"], stdout=subprocess.PIPE, stderr=subprocess.PIPE, text=True, check=False).stdout, + console=False, + ) + log0("=" * 100, console=False) + random.seed(args.seed) + np.random.seed(args.seed) + torch.manual_seed(args.seed) + torch.cuda.manual_seed_all(args.seed) + if not args.tokenizer_path.endswith(".model"): + raise ValueError(f"Script only setup for SentencePiece .model file: {args.tokenizer_path}") + sp = spm.SentencePieceProcessor(model_file=args.tokenizer_path) + if int(sp.vocab_size()) != args.vocab_size: + raise ValueError( + f"VOCAB_SIZE={args.vocab_size} does not match tokenizer vocab_size={int(sp.vocab_size())}" + ) + dataset_dir = Path(args.data_path).resolve() + actual_train_files = len(list(dataset_dir.glob("fineweb_train_*.bin"))) + effective_eval_seq_len = args.eval_seq_len if args.eval_seq_len > 0 else args.train_seq_len + val_seq_len = max(args.train_seq_len, effective_eval_seq_len) + val_tokens = load_validation_tokens(args.val_files, val_seq_len) + base_bytes_lut, has_leading_space_lut, is_boundary_token_lut = build_sentencepiece_luts( + sp, args.vocab_size, device + ) + log0(f"val_bpb:enabled tokenizer_kind=sentencepiece tokenizer_path={args.tokenizer_path}") + log0(f"train_loader:dataset:{dataset_dir.name} train_shards:{actual_train_files}") + log0(f"val_loader:shards pattern={args.val_files} tokens:{val_tokens.numel() - 1}") + CastedLinear._qat_enabled = args.qat_enabled + base_model = GPT( + vocab_size=args.vocab_size, + num_layers=args.num_layers, + model_dim=args.model_dim, + num_heads=args.num_heads, + num_kv_heads=args.num_kv_heads, + mlp_mult=args.mlp_mult, + tie_embeddings=args.tie_embeddings, + tied_embed_init_std=args.tied_embed_init_std, + logit_softcap=args.logit_softcap, + rope_base=args.rope_base, + qk_gain_init=args.qk_gain_init, + mtp_num_heads=args.mtp_num_heads, + mtp_loss_weight=args.mtp_loss_weight, + bigram_vocab_size=args.bigram_vocab_size, + bigram_dim=args.bigram_dim, + xsa_last_n=args.xsa_last_n, + rope_dims=args.rope_dims, + ln_scale=args.ln_scale, + dtg=args.dtg_enabled, + ve_enabled=args.ve_enabled, + ve_dim=args.ve_dim, + ve_layers=args.ve_layers, + mlp_act=args.mlp_act, + mlp_leaky_slope=args.mlp_leaky_slope, + f1_corr_rank=args.f1_corr_rank, + f1_corr_scale_init=args.f1_corr_scale_init, + ).to(device).bfloat16() + for module in base_model.modules(): + if isinstance(module, CastedLinear): + module.float() + restore_low_dim_params_to_fp32(base_model) + compiled_model = maybe_torch_compile(base_model, args) + model: nn.Module = DDP(compiled_model, device_ids=[local_rank], broadcast_buffers=False) if distributed else compiled_model + block_named_params = list(base_model.blocks.named_parameters()) + matrix_params = [ + p + for name, p in block_named_params + if p.ndim == 2 and not any(pattern in name for pattern in CONTROL_TENSOR_NAME_PATTERNS) + ] + if base_model.mtp_num_heads > 0: + matrix_params.extend([p for p in base_model.mtp_heads.parameters() if p.ndim == 2]) + if base_model.f1_corr_in is not None and base_model.f1_corr_out is not None: + matrix_params.append(base_model.f1_corr_in.weight) + matrix_params.append(base_model.f1_corr_out.weight) + scalar_params = [ + p + for name, p in block_named_params + if p.ndim < 2 or any(pattern in name for pattern in CONTROL_TENSOR_NAME_PATTERNS) + ] + if base_model.skip_weights.numel() > 0: + scalar_params.append(base_model.skip_weights) + scalar_params.append(base_model.smear.gate) + if base_model.bigram is not None: + scalar_params.append(base_model.bigram.scale) + if base_model.f1_corr_scale is not None: + scalar_params.append(base_model.f1_corr_scale) + token_lr = args.tied_embed_lr if args.tie_embeddings else args.embed_lr + tok_params = [{"params": [base_model.tok_emb.weight], "lr": token_lr, "base_lr": token_lr}] + if base_model.bigram is not None: + tok_params.append({"params": [base_model.bigram.embed.weight], "lr": token_lr, "base_lr": token_lr}) + if base_model.bigram.proj is not None: + matrix_params.append(base_model.bigram.proj.weight) + if base_model.ve_shared is not None: + tok_params.append({"params": [base_model.ve_shared.embed.weight], "lr": token_lr, "base_lr": token_lr}) + if base_model.ve_shared.proj is not None: + matrix_params.append(base_model.ve_shared.proj.weight) + scalar_params.append(base_model.ve_shared.scale) + for s in base_model.ve_layer_scales: + scalar_params.append(s) + optimizer_tok = torch.optim.AdamW( + tok_params, + betas=(args.beta1, args.beta2), + eps=args.adam_eps, + weight_decay=args.adam_wd, + fused=True, + ) + optimizer_muon = Muon( + matrix_params, + lr=args.matrix_lr, + momentum=args.muon_momentum, + backend_steps=args.muon_backend_steps, + weight_decay=args.muon_wd, + ) + for group in optimizer_muon.param_groups: + group["base_lr"] = args.matrix_lr + optimizer_scalar = torch.optim.AdamW( + [{"params": scalar_params, "lr": args.scalar_lr, "base_lr": args.scalar_lr}], + betas=(args.beta1, args.beta2), + eps=args.adam_eps, + weight_decay=args.adam_wd, + fused=True, + ) + optimizers: list[torch.optim.Optimizer] = [optimizer_tok, optimizer_muon, optimizer_scalar] + if base_model.lm_head is not None: + optimizer_head = torch.optim.Adam( + [{"params": [base_model.lm_head.weight], "lr": args.head_lr, "base_lr": args.head_lr}], + betas=(args.beta1, args.beta2), + eps=args.adam_eps, + fused=True, + ) + optimizers.insert(1, optimizer_head) + n_params = sum(p.numel() for p in base_model.parameters()) + f1_corr_params = 0 + if base_model.f1_corr_in is not None and base_model.f1_corr_out is not None: + f1_corr_params = int(base_model.f1_corr_in.weight.numel() + base_model.f1_corr_out.weight.numel()) + est_corr_int6_bytes = 0 + if args.f1_corr_rank > 0: + # int8 payload stores int6 values + per-row fp16 scales. + est_corr_int6_bytes = ( + args.f1_corr_rank * (args.model_dim + args.vocab_size) + + 2 * (args.f1_corr_rank + args.vocab_size) + ) + log0(f"model_params:{n_params}") + log0( + f"f1_corr:rank={args.f1_corr_rank} params={f1_corr_params} " + f"est_int6_bytes~{est_corr_int6_bytes}" + ) + log0(f"mlp_act:{args.mlp_act} mlp_leaky_slope:{args.mlp_leaky_slope}") + log0(f"XSA:last_{args.xsa_last_n} world_size:{world_size} grad_accum_steps:{grad_accum_steps}") + log0(f"num_heads:{args.num_heads} num_kv_heads:{args.num_kv_heads} embed_lr:{token_lr} matrix_lr:{args.matrix_lr}") + log0( + f"train_batch_tokens:{args.train_batch_tokens} train_seq_len:{args.train_seq_len} " + f"iterations:{args.iterations} warmup_steps:{args.warmup_steps} " + f"max_wallclock_seconds:{args.max_wallclock_seconds:.3f}" + ) + log0(f"compile:enabled={int(args.compile_enabled)} fullgraph={int(args.compile_fullgraph)}") + log0(f"seed:{args.seed}") + if args.ngram_eval_order >= 2: + log0( + f"ngram_eval:order={args.ngram_eval_order} alpha={args.ngram_eval_alpha} " + f"min_count={args.ngram_eval_min_count} buckets={args.ngram_eval_buckets}" + ) + train_loader = DistributedTokenLoader(args.train_files, rank, world_size, device) + def zero_grad_all() -> None: + for opt in optimizers: + opt.zero_grad(set_to_none=True) + max_wallclock_ms = 1000.0 * args.max_wallclock_seconds if args.max_wallclock_seconds > 0 else None + def lr_mul(step: int, elapsed_ms: float) -> float: + if args.warmdown_iters <= 0: + return 1.0 + if max_wallclock_ms is None: + warmdown_start = max(args.iterations - args.warmdown_iters, 0) + return max((args.iterations - step) / max(args.warmdown_iters, 1), 0.0) if warmdown_start <= step < args.iterations else 1.0 + step_ms = elapsed_ms / max(step, 1) + warmdown_ms = args.warmdown_iters * step_ms + remaining_ms = max(max_wallclock_ms - elapsed_ms, 0.0) + return remaining_ms / max(warmdown_ms, 1e-9) if remaining_ms <= warmdown_ms else 1.0 + if args.warmup_steps > 0: + initial_model_state = {name: tensor.detach().cpu().clone() for name, tensor in base_model.state_dict().items()} + initial_optimizer_states = [copy.deepcopy(opt.state_dict()) for opt in optimizers] + model.train() + for warmup_step in range(args.warmup_steps): + zero_grad_all() + for micro_step in range(grad_accum_steps): + if distributed: + model.require_backward_grad_sync = micro_step == grad_accum_steps - 1 + x, y = train_loader.next_batch(args.train_batch_tokens, args.train_seq_len, grad_accum_steps) + with torch.autocast(device_type="cuda", dtype=torch.bfloat16, enabled=True): + warmup_loss = model(x, y) + (warmup_loss * grad_scale).backward() + for opt in optimizers: + opt.step() + zero_grad_all() + if args.warmup_steps <= 20 or (warmup_step + 1) % 10 == 0 or warmup_step + 1 == args.warmup_steps: + log0(f"warmup_step:{warmup_step + 1}/{args.warmup_steps}") + base_model.load_state_dict(initial_model_state, strict=True) + for opt, state in zip(optimizers, initial_optimizer_states, strict=True): + opt.load_state_dict(state) + zero_grad_all() + if distributed: + model.require_backward_grad_sync = True + train_loader = DistributedTokenLoader(args.train_files, rank, world_size, device) + swa_state: dict[str, Tensor] | None = None + swa_count = 0 + ema_state = {name: t.detach().float().clone() for name, t in base_model.state_dict().items()} + ema_decay = 0.997 + training_time_ms = 0.0 + stop_after_step: int | None = None + torch.cuda.synchronize() + t0 = time.perf_counter() + step = 0 + while True: + last_step = step == args.iterations or (stop_after_step is not None and step >= stop_after_step) + should_validate = last_step or (args.val_loss_every > 0 and step % args.val_loss_every == 0) + if should_validate: + torch.cuda.synchronize() + training_time_ms += 1000.0 * (time.perf_counter() - t0) + val_loss, val_bpb = eval_val( + args, + model, + rank, + world_size, + device, + grad_accum_steps, + val_tokens, + base_bytes_lut, + has_leading_space_lut, + is_boundary_token_lut, + ) + log0( + f"step:{step}/{args.iterations} val_loss:{val_loss:.4f} val_bpb:{val_bpb:.4f} " + f"train_time:{training_time_ms:.0f}ms step_avg:{training_time_ms / max(step, 1):.2f}ms" + ) + torch.cuda.synchronize() + t0 = time.perf_counter() + if last_step: + if stop_after_step is not None and step < args.iterations: + log0( + f"stopping_early: wallclock_cap train_time:{training_time_ms:.0f}ms " + f"step:{step}/{args.iterations}" + ) + break + elapsed_ms = training_time_ms + 1000.0 * (time.perf_counter() - t0) + scale = lr_mul(step, elapsed_ms) + if args.late_qat_threshold > 0 and scale < args.late_qat_threshold and not CastedLinear._qat_enabled: + CastedLinear._qat_enabled = True + log0(f"late_qat:enabled step:{step} scale:{scale:.4f}") + zero_grad_all() + train_loss = torch.zeros((), device=device) + for micro_step in range(grad_accum_steps): + if distributed: + model.require_backward_grad_sync = micro_step == grad_accum_steps - 1 + x, y = train_loader.next_batch(args.train_batch_tokens, args.train_seq_len, grad_accum_steps) + with torch.autocast(device_type="cuda", dtype=torch.bfloat16, enabled=True): + loss = model(x, y) + train_loss += loss.detach() + loss.backward() + train_loss /= grad_accum_steps + frac = min(step / args.muon_momentum_warmup_steps, 1.0) if args.muon_momentum_warmup_steps > 0 else 1.0 + muon_momentum = (1 - frac) * args.muon_momentum_warmup_start + frac * args.muon_momentum + for group in optimizer_muon.param_groups: + group["momentum"] = muon_momentum + for opt in optimizers: + for group in opt.param_groups: + group["lr"] = group["base_lr"] * scale + if args.grad_clip_norm > 0: + torch.nn.utils.clip_grad_norm_(base_model.parameters(), args.grad_clip_norm) + for opt in optimizers: + opt.step() + zero_grad_all() + # EMA update + with torch.no_grad(): + for name, t in base_model.state_dict().items(): + ema_state[name].mul_(ema_decay).add_(t.detach().float(), alpha=1.0 - ema_decay) + step += 1 + approx_training_time_ms = training_time_ms + 1000.0 * (time.perf_counter() - t0) + if args.swa_enabled and scale < 0.2 and step % args.swa_every == 0: + if swa_state is None: + swa_state = {name: t.detach().cpu().clone() for name, t in base_model.state_dict().items()} + swa_count = 1 + log0(f"swa:start step:{step}") + else: + for name, t in base_model.state_dict().items(): + swa_state[name] += t.detach().cpu() + swa_count += 1 + should_log_train = ( + args.train_log_every > 0 + and (step <= 10 or step % args.train_log_every == 0 or stop_after_step is not None) + ) + if should_log_train: + log0( + f"step:{step}/{args.iterations} train_loss:{train_loss.item():.4f} " + f"train_time:{approx_training_time_ms:.0f}ms step_avg:{approx_training_time_ms / step:.2f}ms" + ) + reached_cap = max_wallclock_ms is not None and approx_training_time_ms >= max_wallclock_ms + if distributed and max_wallclock_ms is not None: + reached_cap_tensor = torch.tensor(int(reached_cap), device=device) + dist.all_reduce(reached_cap_tensor, op=dist.ReduceOp.MAX) + reached_cap = bool(reached_cap_tensor.item()) + if stop_after_step is None and reached_cap: + stop_after_step = step + log0( + f"peak memory allocated: {torch.cuda.max_memory_allocated() // 1024 // 1024} MiB " + f"reserved: {torch.cuda.max_memory_reserved() // 1024 // 1024} MiB" + ) + if args.distill_enabled and args.distill_steps > 0: + log0( + f"distill:start steps:{args.distill_steps} lr_factor:{args.distill_lr_factor} " + f"temp:{args.distill_temperature} alpha:{args.distill_alpha} kl_clip:{args.distill_kl_clip}" + ) + current_state = base_model.state_dict() + teacher_state = {name: t.to(dtype=current_state[name].dtype) for name, t in ema_state.items()} + teacher_model = GPT( + vocab_size=args.vocab_size, num_layers=args.num_layers, model_dim=args.model_dim, + num_heads=args.num_heads, num_kv_heads=args.num_kv_heads, mlp_mult=args.mlp_mult, + tie_embeddings=args.tie_embeddings, tied_embed_init_std=args.tied_embed_init_std, + logit_softcap=args.logit_softcap, rope_base=args.rope_base, qk_gain_init=args.qk_gain_init, + mtp_num_heads=args.mtp_num_heads, mtp_loss_weight=args.mtp_loss_weight, + bigram_vocab_size=args.bigram_vocab_size, bigram_dim=args.bigram_dim, + xsa_last_n=args.xsa_last_n, rope_dims=args.rope_dims, ln_scale=args.ln_scale, dtg=args.dtg_enabled, + ve_enabled=args.ve_enabled, ve_dim=args.ve_dim, ve_layers=args.ve_layers, + mlp_act=args.mlp_act, mlp_leaky_slope=args.mlp_leaky_slope, + f1_corr_rank=args.f1_corr_rank, f1_corr_scale_init=args.f1_corr_scale_init, + ).to(device).bfloat16() + for m in teacher_model.modules(): + if isinstance(m, CastedLinear): + m.float() + restore_low_dim_params_to_fp32(teacher_model) + teacher_model.load_state_dict(teacher_state, strict=True) + teacher_model.eval() + for p in teacher_model.parameters(): + p.requires_grad_(False) + compiled_teacher_logits = maybe_torch_compile(teacher_model.forward_logits, args) + model.train() + T = args.distill_temperature + alpha = args.distill_alpha + for d_step in range(args.distill_steps): + zero_grad_all() + for opt in optimizers: + for group in opt.param_groups: + group["lr"] = group["base_lr"] * args.distill_lr_factor + x, y = train_loader.next_batch(args.train_batch_tokens, args.train_seq_len, grad_accum_steps) + with torch.autocast(device_type="cuda", dtype=torch.bfloat16, enabled=True): + student_logits = base_model.forward_logits(x) + with torch.no_grad(): + teacher_logits = compiled_teacher_logits(x) + student_log_probs = F.log_softmax(student_logits.float() / T, dim=-1) + teacher_probs = F.softmax(teacher_logits.float() / T, dim=-1) + token_kl = F.kl_div(student_log_probs, teacher_probs, reduction="none").sum(dim=-1) + kl_loss = token_kl.mean() * (T * T) + if args.distill_kl_clip > 0: + kl_loss = torch.clamp(kl_loss, max=args.distill_kl_clip) + ce_loss = F.cross_entropy( + student_logits.reshape(-1, student_logits.size(-1)).float(), + y.reshape(-1), + reduction="mean", + ) + loss = alpha * kl_loss + (1.0 - alpha) * ce_loss + (loss * grad_scale).backward() + if world_size > 1: + for p in base_model.parameters(): + if p.grad is not None: + dist.all_reduce(p.grad, op=dist.ReduceOp.AVG) + if args.grad_clip_norm > 0: + torch.nn.utils.clip_grad_norm_(base_model.parameters(), args.grad_clip_norm) + for opt in optimizers: + opt.step() + zero_grad_all() + with torch.no_grad(): + for name, t in base_model.state_dict().items(): + ema_state[name].mul_(ema_decay).add_(t.detach().float(), alpha=1.0 - ema_decay) + if (d_step + 1) % 8 == 0 or d_step == 0: + log0( + f"distill:step:{d_step + 1}/{args.distill_steps} " + f"kl:{kl_loss.item():.4f} ce:{ce_loss.item():.4f} total:{loss.item():.4f}" + ) + del teacher_model, compiled_teacher_logits + torch.cuda.empty_cache() + log0("distill:done") + # Apply EMA weights (better than SWA alone per PR#401) + log0("ema:applying EMA weights") + current_state = base_model.state_dict() + avg_state = {name: t.to(dtype=current_state[name].dtype) for name, t in ema_state.items()} + base_model.load_state_dict(avg_state, strict=True) + torch.cuda.synchronize() + t_diag = time.perf_counter() + diag_val_loss, diag_val_bpb = eval_val( + args, compiled_model, rank, world_size, device, grad_accum_steps, + val_tokens, base_bytes_lut, has_leading_space_lut, is_boundary_token_lut, + ) + torch.cuda.synchronize() + log0( + f"DIAGNOSTIC post_ema val_loss:{diag_val_loss:.4f} val_bpb:{diag_val_bpb:.4f} " + f"eval_time:{1000.0 * (time.perf_counter() - t_diag):.0f}ms" + ) + full_state_dict = base_model.state_dict() + export_sd = {k: v for k, v in full_state_dict.items() if "mtp_heads" not in k} + excluded_mtp = sum(int(t.numel()) for k, t in full_state_dict.items() if "mtp_heads" in k) + if excluded_mtp > 0: + log0(f"export_excluding_mtp_params:{excluded_mtp}") + if master_process: + torch.save(export_sd, "final_model.pt") + model_bytes = os.path.getsize("final_model.pt") + code_bytes = len(code.encode("utf-8")) + log0(f"Serialized model: {model_bytes} bytes") + log0(f"Code size: {code_bytes} bytes") + sd_cpu = {k: v.detach().cpu() for k, v in export_sd.items()} + # GPTQ calibration: collect Hessians from training data + log0("gptq:calibrating with training data...") + t_gptq = time.perf_counter() + gptq_hessians = gptq_calibrate(base_model, args.train_files, device, n_samples=256, seq_len=args.train_seq_len) + log0(f"gptq:calibrated {len(gptq_hessians)} layers in {time.perf_counter()-t_gptq:.1f}s") + quant_result, quant_meta = mixed_quantize_int6_gptq(sd_cpu, {"mlp", "attn", "aux"}, gptq_hessians) + quant_buf = io.BytesIO() + torch.save({"w": quant_result, "m": quant_meta}, quant_buf) + quant_raw = quant_buf.getvalue() + quant_blob = zstandard.ZstdCompressor(level=22).compress(quant_raw) if _COMPRESSOR == "zstd" else zlib.compress(quant_raw, 9) + if master_process: + with open("final_model.int6.ptz", "wb") as f: + f.write(quant_blob) + quant_file_bytes = len(quant_blob) + code_bytes = len(code.encode("utf-8")) + log0(f"Serialized model int6+{_COMPRESSOR}: {quant_file_bytes} bytes") + log0(f"Total submission size int6+{_COMPRESSOR}: {quant_file_bytes + code_bytes} bytes") + log0(f"Total submission size int8+zlib: {quant_file_bytes + code_bytes} bytes") + if distributed: + dist.barrier() + with open("final_model.int6.ptz", "rb") as f: + quant_blob_disk = f.read() + quant_state = torch.load( + io.BytesIO(zstandard.ZstdDecompressor().decompress(quant_blob_disk) if _COMPRESSOR == "zstd" else zlib.decompress(quant_blob_disk)), + map_location="cpu", + ) + deq_state = dequantize_mixed_int6(quant_state["w"], quant_state["m"], sd_cpu) + eval_model = GPT( + vocab_size=args.vocab_size, num_layers=args.num_layers, model_dim=args.model_dim, + num_heads=args.num_heads, num_kv_heads=args.num_kv_heads, mlp_mult=args.mlp_mult, + tie_embeddings=args.tie_embeddings, tied_embed_init_std=args.tied_embed_init_std, + logit_softcap=args.logit_softcap, rope_base=args.rope_base, qk_gain_init=args.qk_gain_init, + mtp_num_heads=0, mtp_loss_weight=0.0, + bigram_vocab_size=args.bigram_vocab_size, bigram_dim=args.bigram_dim, + xsa_last_n=args.xsa_last_n, # must match training model + rope_dims=args.rope_dims, ln_scale=args.ln_scale, dtg=args.dtg_enabled, + ve_enabled=args.ve_enabled, ve_dim=args.ve_dim, ve_layers=args.ve_layers, + mlp_act=args.mlp_act, mlp_leaky_slope=args.mlp_leaky_slope, + f1_corr_rank=args.f1_corr_rank, f1_corr_scale_init=args.f1_corr_scale_init, + ).to(device).bfloat16() + for m in eval_model.modules(): + if isinstance(m, CastedLinear): + m.float() + restore_low_dim_params_to_fp32(eval_model) + eval_model.load_state_dict(deq_state, strict=True) + compiled_eval = maybe_torch_compile(eval_model, args) + torch.cuda.synchronize() + t_qeval = time.perf_counter() + q_val_loss, q_val_bpb = eval_val( + args, compiled_eval, rank, world_size, device, grad_accum_steps, + val_tokens, base_bytes_lut, has_leading_space_lut, is_boundary_token_lut, + eval_seq_len=effective_eval_seq_len, + ) + torch.cuda.synchronize() + log0( + f"final_int6_roundtrip val_loss:{q_val_loss:.4f} val_bpb:{q_val_bpb:.4f} " + f"eval_time:{1000.0 * (time.perf_counter() - t_qeval):.0f}ms" + ) + log0(f"final_int6_roundtrip_exact val_loss:{q_val_loss:.8f} val_bpb:{q_val_bpb:.8f}") + sw_seq_len = effective_eval_seq_len + if args.eval_stride > 0 and args.eval_stride < sw_seq_len: + torch.cuda.synchronize() + t_slide = time.perf_counter() + sw_val_loss, sw_val_bpb = eval_val_sliding( + args, eval_model, rank, world_size, device, + val_tokens, base_bytes_lut, has_leading_space_lut, is_boundary_token_lut, + stride=args.eval_stride, + eval_seq_len=sw_seq_len, + ) + torch.cuda.synchronize() + log0( + f"final_int6_sliding_window val_loss:{sw_val_loss:.4f} val_bpb:{sw_val_bpb:.4f} " + f"stride:{args.eval_stride} eval_time:{1000.0 * (time.perf_counter() - t_slide):.0f}ms" + ) + log0(f"final_int6_sliding_window_exact val_loss:{sw_val_loss:.8f} val_bpb:{sw_val_bpb:.8f}") + log0(f"final_int8_zlib_roundtrip_exact val_loss:{sw_val_loss:.8f} val_bpb:{sw_val_bpb:.8f}") + if args.ngram_eval_order >= 2: + if distributed: + dist.barrier() + torch.cuda.synchronize() + t_ng = time.perf_counter() + ng_loss, ng_bpb, ng_coverage = eval_val_sliding_hashed_ngram( + args, + eval_model, + rank, + world_size, + device, + val_tokens, + base_bytes_lut, + has_leading_space_lut, + is_boundary_token_lut, + stride=args.eval_stride, + order=args.ngram_eval_order, + alpha=args.ngram_eval_alpha, + min_count=args.ngram_eval_min_count, + buckets=args.ngram_eval_buckets, + max_seconds=args.ngram_eval_max_seconds, + eval_seq_len=sw_seq_len, + ) + if rank == 0: + torch.cuda.synchronize() + ng_eval_ms = 1000.0 * (time.perf_counter() - t_ng) + if ng_coverage >= 0.999999: + log0( + f"final_int6_sliding_window_ngram{args.ngram_eval_order} val_loss:{ng_loss:.4f} " + f"val_bpb:{ng_bpb:.4f} eval_time:{ng_eval_ms:.0f}ms" + ) + log0( + f"final_int6_sliding_window_ngram{args.ngram_eval_order}_exact " + f"val_loss:{ng_loss:.8f} val_bpb:{ng_bpb:.8f}" + ) + else: + log0( + f"final_int6_sliding_window_ngram{args.ngram_eval_order}_partial val_loss:{ng_loss:.4f} " + f"val_bpb:{ng_bpb:.4f} coverage:{ng_coverage:.4f} eval_time:{ng_eval_ms:.0f}ms" + ) + log0( + f"final_int6_sliding_window_ngram{args.ngram_eval_order}_partial_exact " + f"val_loss:{ng_loss:.8f} val_bpb:{ng_bpb:.8f} coverage:{ng_coverage:.8f}" + ) + if distributed: + dist.barrier() + # Legal score-first TTT eval + if args.ttt_eval_enabled: + torch.cuda.synchronize() + t_ttt = time.perf_counter() + ttt_loss, ttt_bpb = eval_val_sliding_ttt( + args, eval_model, rank, world_size, device, + val_tokens, base_bytes_lut, has_leading_space_lut, is_boundary_token_lut, + stride=args.eval_stride, + ) + torch.cuda.synchronize() + log0(f"legal_ttt val_loss:{ttt_loss:.4f} val_bpb:{ttt_bpb:.4f} " + f"eval_time:{1000.0 * (time.perf_counter() - t_ttt):.0f}ms") + log0(f"legal_ttt_exact val_loss:{ttt_loss:.8f} val_bpb:{ttt_bpb:.8f}") + if distributed: + dist.destroy_process_group() +if __name__ == "__main__": + main() diff --git a/records/track_10min_16mb/2026-03-25_Podracing_5gram_LeakyReLU2_8xH100/README.md b/records/track_10min_16mb/2026-03-25_Podracing_5gram_LeakyReLU2_8xH100/README.md new file mode 100644 index 000000000..b44fce575 --- /dev/null +++ b/records/track_10min_16mb/2026-03-25_Podracing_5gram_LeakyReLU2_8xH100/README.md @@ -0,0 +1,36 @@ +# Podracing: 5-gram Eval + LeakyReLU² + GPTQ + +## Results + +| Seed | Sliding BPB | 5-gram BPB | Artifact | +|------|-------------|-----------|----------| +| 1337 | 1.1190 | **1.0451** | 15.63 MB | +| 42 | 1.1217 | **1.0471** | 15.59 MB | +| 2045 | 1.1200 | **1.0460** | 15.64 MB | +| **Mean** | **1.1202** | **1.0461** | — | + +## Architecture + +11L/512d U-Net, 8H/4KV, LeakyReLU² (slope 0.5), XSA last 4, BigramHash 1536, +VE128 on layers 9-10, partial RoPE (24/64 dims), tied embeddings. 26.93M params. + +## 5-gram Eval (score-first, legal) + +Fixed-weight hashed n-gram interpolation during sliding window eval: +- Cache built from already-scored tokens only (backward-looking) +- Fixed alpha=0.20: `p_final = 0.80 * p_model + 0.20 * p_ngram` +- No safety gate, no target-aware selection, no min-NLL comparison +- Hashed count-min sketch (4M buckets), min_count=2 +- N-gram concept credited to @deanbrr (PR #659) + +## Reproduce + +```bash +SEED=2045 MLP_ACT=leaky_relu_sq MLP_LEAKY_SLOPE=0.5 \ +XSA_LAST_N=4 BIGRAM_VOCAB_SIZE=1536 ROPE_DIMS=24 \ +NGRAM_EVAL_ORDER=5 NGRAM_EVAL_ALPHA=0.20 \ +NGRAM_EVAL_MIN_COUNT=2 NGRAM_EVAL_BUCKETS=4194304 \ +torchrun --nproc_per_node=8 train_gpt.py +``` + +8xH100 SXM, 600s training + ~190s eval. diff --git a/records/track_10min_16mb/2026-03-25_Podracing_5gram_LeakyReLU2_8xH100/submission.json b/records/track_10min_16mb/2026-03-25_Podracing_5gram_LeakyReLU2_8xH100/submission.json new file mode 100644 index 000000000..11d3b5a7a --- /dev/null +++ b/records/track_10min_16mb/2026-03-25_Podracing_5gram_LeakyReLU2_8xH100/submission.json @@ -0,0 +1,11 @@ +{ + "author": "Frosty40", + "github_id": "newjordan", + "name": "Podracing: 5-gram Eval + LeakyReLU² + GPTQ", + "blurb": "11L/512d U-Net with legal score-first hashed 5-gram eval interpolation (fixed alpha=0.20). 3-seed mean val_bpb=1.0461 (std=0.0010). N-gram cache concept credited to @deanbrr (PR #659).", + "date": "2026-03-25T03:45:00Z", + "val_loss": 1.7661, + "val_bpb": 1.0461, + "bytes_total": 15631465, + "bytes_code": 102835 +} diff --git a/records/track_10min_16mb/2026-03-25_Podracing_5gram_LeakyReLU2_8xH100/train_gpt.py b/records/track_10min_16mb/2026-03-25_Podracing_5gram_LeakyReLU2_8xH100/train_gpt.py new file mode 100644 index 000000000..6566ad098 --- /dev/null +++ b/records/track_10min_16mb/2026-03-25_Podracing_5gram_LeakyReLU2_8xH100/train_gpt.py @@ -0,0 +1,2083 @@ +from __future__ import annotations +import copy +import glob +import io +import math +import os +import random +import subprocess +import sys +import time +import uuid +import zlib +from pathlib import Path +try: + import zstandard + _COMPRESSOR = "zstd" +except ImportError: + _COMPRESSOR = "zlib" +import numpy as np +import sentencepiece as spm +import torch +import torch.distributed as dist +import torch.nn.functional as F +from torch import Tensor, nn +from torch.nn.parallel import DistributedDataParallel as DDP +from flash_attn_interface import flash_attn_func as flash_attn_3_func +class Hyperparameters: + data_path = os.environ.get("DATA_PATH", "./data/datasets/fineweb10B_sp1024") + train_files = os.path.join(data_path, "fineweb_train_*.bin") + val_files = os.path.join(data_path, "fineweb_val_*.bin") + tokenizer_path = os.environ.get("TOKENIZER_PATH", "./data/tokenizers/fineweb_1024_bpe.model") + run_id = os.environ.get("RUN_ID", str(uuid.uuid4())) + seed = int(os.environ.get("SEED", 1337)) + val_batch_size = int(os.environ.get("VAL_BATCH_SIZE", 524_288)) + val_loss_every = int(os.environ.get("VAL_LOSS_EVERY", 4000)) + train_log_every = int(os.environ.get("TRAIN_LOG_EVERY", 500)) + iterations = int(os.environ.get("ITERATIONS", 20000)) + warmdown_iters = int(os.environ.get("WARMDOWN_ITERS", 3500)) + warmup_steps = int(os.environ.get("WARMUP_STEPS", 20)) + train_batch_tokens = int(os.environ.get("TRAIN_BATCH_TOKENS", 786_432)) + train_seq_len = int(os.environ.get("TRAIN_SEQ_LEN", 2048)) + eval_seq_len = int(os.environ.get("EVAL_SEQ_LEN", 2048)) + max_wallclock_seconds = float(os.environ.get("MAX_WALLCLOCK_SECONDS", 600.0)) + qk_gain_init = float(os.environ.get("QK_GAIN_INIT", 1.5)) + vocab_size = int(os.environ.get("VOCAB_SIZE", 1024)) + num_layers = int(os.environ.get("NUM_LAYERS", 11)) + num_kv_heads = int(os.environ.get("NUM_KV_HEADS", 4)) + model_dim = int(os.environ.get("MODEL_DIM", 512)) + num_heads = int(os.environ.get("NUM_HEADS", 8)) + mlp_mult = float(os.environ.get("MLP_MULT", 3.0)) + mlp_act = os.environ.get("MLP_ACT", "relu_sq").lower() + mlp_leaky_slope = float(os.environ.get("MLP_LEAKY_SLOPE", 0.5)) + tie_embeddings = bool(int(os.environ.get("TIE_EMBEDDINGS", "1"))) + rope_base = float(os.environ.get("ROPE_BASE", 10000.0)) + logit_softcap = float(os.environ.get("LOGIT_SOFTCAP", 30.0)) + embed_lr = float(os.environ.get("EMBED_LR", 0.6)) + head_lr = float(os.environ.get("HEAD_LR", 0.008)) + tied_embed_lr = float(os.environ.get("TIED_EMBED_LR", 0.035)) + tied_embed_init_std = float(os.environ.get("TIED_EMBED_INIT_STD", 0.005)) + matrix_lr = float(os.environ.get("MATRIX_LR", 0.025)) + scalar_lr = float(os.environ.get("SCALAR_LR", 0.025)) + muon_momentum = float(os.environ.get("MUON_MOMENTUM", 0.99)) + muon_backend_steps = int(os.environ.get("MUON_BACKEND_STEPS", 5)) + muon_momentum_warmup_start = float(os.environ.get("MUON_MOMENTUM_WARMUP_START", 0.92)) + muon_momentum_warmup_steps = int(os.environ.get("MUON_MOMENTUM_WARMUP_STEPS", 1500)) + beta1 = float(os.environ.get("BETA1", 0.9)) + beta2 = float(os.environ.get("BETA2", 0.95)) + adam_eps = float(os.environ.get("ADAM_EPS", 1e-8)) + grad_clip_norm = float(os.environ.get("GRAD_CLIP_NORM", 0.3)) + eval_stride = int(os.environ.get("EVAL_STRIDE", 64)) + mtp_num_heads = int(os.environ.get("MTP_NUM_HEADS", 0)) + mtp_loss_weight = float(os.environ.get("MTP_LOSS_WEIGHT", 0.2)) + muon_beta2 = float(os.environ.get("MUON_BETA2", 0.95)) + swa_enabled = bool(int(os.environ.get("SWA_ENABLED", "1"))) + swa_every = int(os.environ.get("SWA_EVERY", 50)) # tighter: collect more recent checkpoints + muon_wd = float(os.environ.get("MUON_WD", 0.04)) + adam_wd = float(os.environ.get("ADAM_WD", 0.04)) + qat_enabled = bool(int(os.environ.get("QAT_ENABLED", "0"))) + bigram_vocab_size = int(os.environ.get("BIGRAM_VOCAB_SIZE", 2048)) + bigram_dim = int(os.environ.get("BIGRAM_DIM", 128)) + xsa_last_n = int(os.environ.get("XSA_LAST_N", 11)) # XSA on ALL 11 layers + rope_dims = int(os.environ.get("ROPE_DIMS", 16)) + ln_scale = bool(int(os.environ.get("LN_SCALE", "1"))) + dtg_enabled = bool(int(os.environ.get("DTG_ENABLED", "0"))) + late_qat_threshold = float(os.environ.get("LATE_QAT_THRESHOLD", 0.5)) + ve_enabled = bool(int(os.environ.get("VE_ENABLED", "1"))) + ve_dim = int(os.environ.get("VE_DIM", 128)) + ve_layers = os.environ.get("VE_LAYERS", "9,10") + # F1 capacity add-on: low-rank correction head (active at inference). + # Approx extra params ~= rank * (model_dim + vocab_size). + f1_corr_rank = int(os.environ.get("F1_CORR_RANK", 0)) + f1_corr_scale_init = float(os.environ.get("F1_CORR_SCALE_INIT", 0.10)) + # Post-train self-distillation: EMA teacher -> student. + distill_enabled = bool(int(os.environ.get("DISTILL_ENABLED", "0"))) + distill_steps = int(os.environ.get("DISTILL_STEPS", 24)) + distill_lr_factor = float(os.environ.get("DISTILL_LR_FACTOR", 0.02)) + distill_temperature = float(os.environ.get("DISTILL_TEMPERATURE", 1.5)) + distill_alpha = float(os.environ.get("DISTILL_ALPHA", 0.60)) + distill_kl_clip = float(os.environ.get("DISTILL_KL_CLIP", 10.0)) + # Legal score-first TTT eval (PR #461 recipe) + ttt_eval_enabled = bool(int(os.environ.get("TTT_EVAL_ENABLED", "1"))) + ttt_lr = float(os.environ.get("TTT_LR", 0.002)) + ttt_epochs = int(os.environ.get("TTT_EPOCHS", 3)) + ttt_chunk_tokens = int(os.environ.get("TTT_CHUNK_TOKENS", 32768)) + ttt_freeze_blocks = int(os.environ.get("TTT_FREEZE_BLOCKS", 2)) + ttt_momentum = float(os.environ.get("TTT_MOMENTUM", 0.9)) + ttt_batch_seqs = int(os.environ.get("TTT_BATCH_SEQS", 32)) + ttt_grad_clip = float(os.environ.get("TTT_GRAD_CLIP", 1.0)) + ttt_max_train_chunks = int(os.environ.get("TTT_MAX_TRAIN_CHUNKS", 200)) # stop training after N chunks, keep scoring + ttt_ema_decay = float(os.environ.get("TTT_EMA_DECAY", 0.995)) # EMA decay for TTT weight smoothing (0 = disabled) + ttt_freeze_embed = bool(int(os.environ.get("TTT_FREEZE_EMBED", "1"))) # freeze tok_emb/bigram/ve during TTT + # Optional legal score-first hashed n-gram interpolation at eval time. + # This path never uses label-aware gating: mix weight is fixed. + ngram_eval_order = int(os.environ.get("NGRAM_EVAL_ORDER", 0)) # 0=off, >=2 enables hashed n-gram + ngram_eval_alpha = float(os.environ.get("NGRAM_EVAL_ALPHA", 0.20)) + ngram_eval_min_count = int(os.environ.get("NGRAM_EVAL_MIN_COUNT", 2)) + ngram_eval_buckets = int(os.environ.get("NGRAM_EVAL_BUCKETS", 4_194_304)) + ngram_eval_max_seconds = float(os.environ.get("NGRAM_EVAL_MAX_SECONDS", 0.0)) + compile_enabled = bool(int(os.environ.get("COMPILE_ENABLED", "1"))) + compile_fullgraph = bool(int(os.environ.get("COMPILE_FULLGRAPH", "1"))) +def maybe_torch_compile(obj, args: Hyperparameters): + if not args.compile_enabled: + return obj + return torch.compile(obj, dynamic=False, fullgraph=args.compile_fullgraph) +def zeropower_via_newtonschulz5(G: Tensor, steps: int = 10, eps: float = 1e-7) -> Tensor: + a, b, c = (3.4445, -4.7750, 2.0315) + X = G.bfloat16() + X /= X.norm() + eps + transposed = G.size(0) > G.size(1) + if transposed: + X = X.T + for _ in range(steps): + A = X @ X.T + B = b * A + c * A @ A + X = a * X + B @ X + return X.T if transposed else X +class Muon(torch.optim.Optimizer): + def __init__(self, params, lr: float, momentum: float, backend_steps: int, + nesterov: bool = True, weight_decay: float = 0.0): + super().__init__( + params, + dict(lr=lr, momentum=momentum, backend_steps=backend_steps, + nesterov=nesterov, weight_decay=weight_decay), + ) + @torch.no_grad() + def step(self, closure=None): + loss = None + if closure is not None: + with torch.enable_grad(): + loss = closure() + distributed = dist.is_available() and dist.is_initialized() + world_size = dist.get_world_size() if distributed else 1 + rank = dist.get_rank() if distributed else 0 + for group in self.param_groups: + params = group["params"] + if not params: + continue + lr = group["lr"] + momentum = group["momentum"] + backend_steps = group["backend_steps"] + nesterov = group["nesterov"] + total_params = sum(int(p.numel()) for p in params) + updates_flat = torch.zeros(total_params, device=params[0].device, dtype=torch.bfloat16) + curr = 0 + for i, p in enumerate(params): + if i % world_size == rank and p.grad is not None: + g = p.grad + state = self.state[p] + if "momentum_buffer" not in state: + state["momentum_buffer"] = torch.zeros_like(g) + buf = state["momentum_buffer"] + buf.mul_(momentum).add_(g) + if nesterov: + g = g.add(buf, alpha=momentum) + g = zeropower_via_newtonschulz5(g, steps=backend_steps) + g *= max(1, g.size(0) / g.size(1)) ** 0.5 + updates_flat[curr : curr + p.numel()] = g.reshape(-1) + curr += p.numel() + if distributed: + dist.all_reduce(updates_flat, op=dist.ReduceOp.SUM) + wd = group.get("weight_decay", 0.0) + curr = 0 + for p in params: + if wd > 0.0: + p.data.mul_(1.0 - lr * wd) + g = updates_flat[curr : curr + p.numel()].view_as(p).to(dtype=p.dtype) + p.add_(g, alpha=-lr) + curr += p.numel() + return loss +def build_sentencepiece_luts( + sp: spm.SentencePieceProcessor, vocab_size: int, device: torch.device +) -> tuple[Tensor, Tensor, Tensor]: + sp_vocab_size = int(sp.vocab_size()) + table_size = max(sp_vocab_size, vocab_size) + base_bytes_np = np.zeros((table_size,), dtype=np.int16) + has_leading_space_np = np.zeros((table_size,), dtype=np.bool_) + is_boundary_token_np = np.ones((table_size,), dtype=np.bool_) + for token_id in range(sp_vocab_size): + if sp.is_control(token_id) or sp.is_unknown(token_id) or sp.is_unused(token_id): + continue + is_boundary_token_np[token_id] = False + if sp.is_byte(token_id): + base_bytes_np[token_id] = 1 + continue + piece = sp.id_to_piece(token_id) + if piece.startswith("▁"): + has_leading_space_np[token_id] = True + piece = piece[1:] + base_bytes_np[token_id] = len(piece.encode("utf-8")) + return ( + torch.tensor(base_bytes_np, dtype=torch.int16, device=device), + torch.tensor(has_leading_space_np, dtype=torch.bool, device=device), + torch.tensor(is_boundary_token_np, dtype=torch.bool, device=device), + ) +def load_validation_tokens(pattern: str, seq_len: int) -> Tensor: + files = [Path(p) for p in sorted(glob.glob(pattern))] + if not files: + raise FileNotFoundError(f"No files found for pattern: {pattern}") + tokens = torch.cat([load_data_shard(file) for file in files]).contiguous() + usable = ((tokens.numel() - 1) // seq_len) * seq_len + if usable <= 0: + raise ValueError(f"Validation split is too short for TRAIN_SEQ_LEN={seq_len}") + return tokens[: usable + 1] +def eval_val( + args: Hyperparameters, + model: nn.Module, + rank: int, + world_size: int, + device: torch.device, + grad_accum_steps: int, + val_tokens: Tensor, + base_bytes_lut: Tensor, + has_leading_space_lut: Tensor, + is_boundary_token_lut: Tensor, + eval_seq_len: int | None = None, +) -> tuple[float, float]: + seq_len = eval_seq_len or args.train_seq_len + local_batch_tokens = args.val_batch_size // (world_size * grad_accum_steps) + if local_batch_tokens < seq_len: + raise ValueError( + "VAL_BATCH_SIZE must provide at least one sequence per rank; " + f"got VAL_BATCH_SIZE={args.val_batch_size}, WORLD_SIZE={world_size}, " + f"GRAD_ACCUM_STEPS={grad_accum_steps}, seq_len={seq_len}" + ) + local_batch_seqs = local_batch_tokens // seq_len + total_seqs = (val_tokens.numel() - 1) // seq_len + seq_start = (total_seqs * rank) // world_size + seq_end = (total_seqs * (rank + 1)) // world_size + val_loss_sum = torch.zeros((), device=device, dtype=torch.float64) + val_token_count = torch.zeros((), device=device, dtype=torch.float64) + val_byte_count = torch.zeros((), device=device, dtype=torch.float64) + model.eval() + with torch.inference_mode(): + for batch_seq_start in range(seq_start, seq_end, local_batch_seqs): + batch_seq_end = min(batch_seq_start + local_batch_seqs, seq_end) + raw_start = batch_seq_start * seq_len + raw_end = batch_seq_end * seq_len + 1 + local = val_tokens[raw_start:raw_end].to(device=device, dtype=torch.int64, non_blocking=True) + x = local[:-1].reshape(-1, seq_len) + y = local[1:].reshape(-1, seq_len) + with torch.autocast(device_type="cuda", dtype=torch.bfloat16, enabled=True): + batch_loss = model(x, y).detach() + batch_token_count = float(y.numel()) + val_loss_sum += batch_loss.to(torch.float64) * batch_token_count + val_token_count += batch_token_count + prev_ids = x.reshape(-1) + tgt_ids = y.reshape(-1) + token_bytes = base_bytes_lut[tgt_ids].to(dtype=torch.int16) + token_bytes += (has_leading_space_lut[tgt_ids] & ~is_boundary_token_lut[prev_ids]).to(dtype=torch.int16) + val_byte_count += token_bytes.to(torch.float64).sum() + if dist.is_available() and dist.is_initialized(): + dist.all_reduce(val_loss_sum, op=dist.ReduceOp.SUM) + dist.all_reduce(val_token_count, op=dist.ReduceOp.SUM) + dist.all_reduce(val_byte_count, op=dist.ReduceOp.SUM) + val_loss = val_loss_sum / val_token_count + bits_per_token = val_loss.item() / math.log(2.0) + tokens_per_byte = val_token_count.item() / val_byte_count.item() + model.train() + return float(val_loss.item()), float(bits_per_token * tokens_per_byte) +CONTROL_TENSOR_NAME_PATTERNS = tuple( + pattern + for pattern in os.environ.get( + "CONTROL_TENSOR_NAME_PATTERNS", + "attn_scale,attn_scales,mlp_scale,mlp_scales,resid_mix,resid_mixes,q_gain,skip_weight,skip_weights,smear,dtg_gate,ve_layer_scales,ve_shared.scale", + ).split(",") + if pattern +) +INT8_KEEP_FLOAT_FP32_NAME_PATTERNS = tuple( + pattern + for pattern in os.environ.get( + "INT8_KEEP_FLOAT_FP32_NAME_PATTERNS", + ",".join(CONTROL_TENSOR_NAME_PATTERNS), + ).split(",") + if pattern +) +INT8_KEEP_FLOAT_MAX_NUMEL = 65_536 +INT8_KEEP_FLOAT_STORE_DTYPE = torch.float16 +INT8_PER_ROW_SCALE_DTYPE = torch.float16 +INT8_CLIP_PERCENTILE = 99.99984 +INT8_CLIP_Q = INT8_CLIP_PERCENTILE / 100.0 +def tensor_nbytes(t: Tensor) -> int: + return int(t.numel()) * int(t.element_size()) +def keep_float_tensor(name: str, t: Tensor, passthrough_orig_dtypes: dict[str, str]) -> Tensor: + if any(pattern in name for pattern in INT8_KEEP_FLOAT_FP32_NAME_PATTERNS): + return t.float().contiguous() + if t.dtype in {torch.float32, torch.bfloat16}: + passthrough_orig_dtypes[name] = str(t.dtype).removeprefix("torch.") + return t.to(dtype=INT8_KEEP_FLOAT_STORE_DTYPE).contiguous() + return t +def quantize_float_tensor(t: Tensor) -> tuple[Tensor, Tensor]: + t32 = t.float() + if t32.ndim == 2: + clip_abs = ( + torch.quantile(t32.abs(), INT8_CLIP_Q, dim=1) + if t32.numel() + else torch.empty((t32.shape[0],), dtype=torch.float32) + ) + clipped = torch.maximum(torch.minimum(t32, clip_abs[:, None]), -clip_abs[:, None]) + scale = (clip_abs / 127.0).clamp_min(1.0 / 127.0) + q = torch.clamp(torch.round(clipped / scale[:, None]), -127, 127).to(torch.int8).contiguous() + return q, scale.to(dtype=INT8_PER_ROW_SCALE_DTYPE).contiguous() + clip_abs = float(torch.quantile(t32.abs().flatten(), INT8_CLIP_Q).item()) if t32.numel() else 0.0 + scale = torch.tensor(clip_abs / 127.0 if clip_abs > 0 else 1.0, dtype=torch.float32) + q = torch.clamp(torch.round(torch.clamp(t32, -clip_abs, clip_abs) / scale), -127, 127).to(torch.int8).contiguous() + return q, scale +def quantize_state_dict_int8(state_dict: dict[str, Tensor]): + quantized: dict[str, Tensor] = {} + scales: dict[str, Tensor] = {} + dtypes: dict[str, str] = {} + passthrough: dict[str, Tensor] = {} + passthrough_orig_dtypes: dict[str, str] = {} + qmeta: dict[str, dict[str, object]] = {} + stats = dict.fromkeys( + ("param_count", "num_tensors", "num_float_tensors", "num_nonfloat_tensors", "baseline_tensor_bytes", "int8_payload_bytes"), + 0, + ) + for name, tensor in state_dict.items(): + t = tensor.detach().to("cpu").contiguous() + stats["param_count"] += int(t.numel()) + stats["num_tensors"] += 1 + stats["baseline_tensor_bytes"] += tensor_nbytes(t) + if not t.is_floating_point(): + stats["num_nonfloat_tensors"] += 1 + passthrough[name] = t + stats["int8_payload_bytes"] += tensor_nbytes(t) + continue + if t.numel() <= INT8_KEEP_FLOAT_MAX_NUMEL: + kept = keep_float_tensor(name, t, passthrough_orig_dtypes) + passthrough[name] = kept + stats["int8_payload_bytes"] += tensor_nbytes(kept) + continue + stats["num_float_tensors"] += 1 + q, s = quantize_float_tensor(t) + if s.ndim > 0: + qmeta[name] = {"scheme": "per_row", "axis": 0} + quantized[name] = q + scales[name] = s + dtypes[name] = str(t.dtype).removeprefix("torch.") + stats["int8_payload_bytes"] += tensor_nbytes(q) + tensor_nbytes(s) + obj: dict[str, object] = { + "__quant_format__": "int8_clean_per_row_v1", + "quantized": quantized, + "scales": scales, + "dtypes": dtypes, + "passthrough": passthrough, + } + if qmeta: + obj["qmeta"] = qmeta + if passthrough_orig_dtypes: + obj["passthrough_orig_dtypes"] = passthrough_orig_dtypes + return obj, stats +def dequantize_state_dict_int8(obj: dict[str, object]) -> dict[str, Tensor]: + out: dict[str, Tensor] = {} + qmeta = obj.get("qmeta", {}) + passthrough_orig_dtypes = obj.get("passthrough_orig_dtypes", {}) + for name, q in obj["quantized"].items(): + dtype = getattr(torch, obj["dtypes"][name]) + s = obj["scales"][name] + if qmeta.get(name, {}).get("scheme") == "per_row" or s.ndim > 0: + s = s.to(dtype=torch.float32) + out[name] = (q.float() * s.view(q.shape[0], *([1] * (q.ndim - 1)))).to(dtype=dtype).contiguous() + else: + scale = float(s.item()) + out[name] = (q.float() * scale).to(dtype=dtype).contiguous() + for name, t in obj["passthrough"].items(): + out_t = t.detach().to("cpu").contiguous() + orig_dtype = passthrough_orig_dtypes.get(name) + if isinstance(orig_dtype, str): + out_t = out_t.to(dtype=getattr(torch, orig_dtype)).contiguous() + out[name] = out_t + return out +def load_data_shard(file: Path) -> Tensor: + header_bytes = 256 * np.dtype(" None: + self.file_idx = (self.file_idx + 1) % len(self.files) + self.tokens = load_data_shard(self.files[self.file_idx]) + self.pos = 0 + def take(self, n: int) -> Tensor: + chunks: list[Tensor] = [] + remaining = n + while remaining > 0: + avail = self.tokens.numel() - self.pos + if avail <= 0: + self._advance_file() + continue + k = min(remaining, avail) + chunks.append(self.tokens[self.pos : self.pos + k]) + self.pos += k + remaining -= k + return chunks[0] if len(chunks) == 1 else torch.cat(chunks) +class DistributedTokenLoader: + def __init__(self, pattern: str, rank: int, world_size: int, device: torch.device): + self.rank = rank + self.world_size = world_size + self.device = device + self.stream = TokenStream(pattern) + def next_batch(self, global_tokens: int, seq_len: int, grad_accum_steps: int) -> tuple[Tensor, Tensor]: + local_tokens = global_tokens // (self.world_size * grad_accum_steps) + per_rank_span = local_tokens + 1 + chunk = self.stream.take(per_rank_span * self.world_size) + start = self.rank * per_rank_span + local = chunk[start : start + per_rank_span].to(dtype=torch.int64) + x = local[:-1].reshape(-1, seq_len) + y = local[1:].reshape(-1, seq_len) + return x.to(self.device, non_blocking=True), y.to(self.device, non_blocking=True) +class RMSNorm(nn.Module): + def __init__(self, eps: float | None = None): + super().__init__() + self.eps = eps + def forward(self, x: Tensor) -> Tensor: + return F.rms_norm(x, (x.size(-1),), eps=self.eps) +class CastedLinear(nn.Linear): + _qat_enabled: bool = False + def forward(self, x: Tensor) -> Tensor: + w = self.weight.to(x.dtype) + if CastedLinear._qat_enabled and self.training and w.ndim == 2: + with torch.no_grad(): + w32 = self.weight.float() + # Use 99.95th percentile clipping to match GPTQ export quantizer + row_clip = torch.quantile(w32.abs(), 0.9995, dim=1) + scale = (row_clip / 31.0).clamp_min(1.0 / 31.0) + w_q = (torch.clamp(torch.round(w32 / scale[:, None]), -32, 31) * scale[:, None]).to(x.dtype) + w = w + (w_q - w).detach() + bias = self.bias.to(x.dtype) if self.bias is not None else None + return F.linear(x, w, bias) +def restore_low_dim_params_to_fp32(module: nn.Module) -> None: + with torch.no_grad(): + for name, param in module.named_parameters(): + if (param.ndim < 2 or any(pattern in name for pattern in CONTROL_TENSOR_NAME_PATTERNS)) and param.dtype != torch.float32: + param.data = param.data.float() +class Rotary(nn.Module): + def __init__(self, dim: int, base: float = 10000.0, train_seq_len: int = 1024, rope_dims: int = 0): + super().__init__() + self.dim = dim + self.base = base + self.train_seq_len = train_seq_len + self.rope_dims = rope_dims if rope_dims > 0 else dim + inv_freq = 1.0 / (base ** (torch.arange(0, self.rope_dims, 2, dtype=torch.float32) / self.rope_dims)) + self.register_buffer("inv_freq", inv_freq, persistent=False) + self._seq_len_cached = 0 + self._cos_cached: Tensor | None = None + self._sin_cached: Tensor | None = None + def forward(self, seq_len: int, device: torch.device, dtype: torch.dtype) -> tuple[Tensor, Tensor]: + if ( + self._cos_cached is None + or self._sin_cached is None + or self._seq_len_cached != seq_len + or self._cos_cached.device != device + ): + rd = self.rope_dims + if seq_len > self.train_seq_len: + scale = seq_len / self.train_seq_len + new_base = self.base * (scale ** (rd / (rd - 2))) + inv_freq = 1.0 / (new_base ** (torch.arange(0, rd, 2, dtype=torch.float32, device=device) / rd)) + else: + inv_freq = self.inv_freq.to(device) + t = torch.arange(seq_len, device=device, dtype=inv_freq.dtype) + freqs = torch.outer(t, inv_freq) + self._cos_cached = freqs.cos()[None, :, None, :] + self._sin_cached = freqs.sin()[None, :, None, :] + self._seq_len_cached = seq_len + return self._cos_cached.to(dtype=dtype), self._sin_cached.to(dtype=dtype) +def apply_rotary_emb(x: Tensor, cos: Tensor, sin: Tensor, rope_dims: int = 0) -> Tensor: + if rope_dims > 0 and rope_dims < x.size(-1): + x_rope, x_pass = x[..., :rope_dims], x[..., rope_dims:] + half = rope_dims // 2 + x1, x2 = x_rope[..., :half], x_rope[..., half:] + x_rope = torch.cat((x1 * cos + x2 * sin, x1 * (-sin) + x2 * cos), dim=-1) + return torch.cat((x_rope, x_pass), dim=-1) + half = x.size(-1) // 2 + x1, x2 = x[..., :half], x[..., half:] + return torch.cat((x1 * cos + x2 * sin, x1 * (-sin) + x2 * cos), dim=-1) +class CausalSelfAttention(nn.Module): + def __init__( + self, + dim: int, + num_heads: int, + num_kv_heads: int, + rope_base: float, + qk_gain_init: float, + ): + super().__init__() + if dim % num_heads != 0: + raise ValueError("model_dim must be divisible by num_heads") + if num_heads % num_kv_heads != 0: + raise ValueError("num_heads must be divisible by num_kv_heads") + self.num_heads = num_heads + self.num_kv_heads = num_kv_heads + self.head_dim = dim // num_heads + if self.head_dim % 2 != 0: + raise ValueError("head_dim must be even for RoPE") + kv_dim = self.num_kv_heads * self.head_dim + self.c_q = CastedLinear(dim, dim, bias=False) + self.c_k = CastedLinear(dim, kv_dim, bias=False) + self.c_v = CastedLinear(dim, kv_dim, bias=False) + self.proj = CastedLinear(dim, dim, bias=False) + self.proj._zero_init = True + self.q_gain = nn.Parameter(torch.full((num_heads,), qk_gain_init, dtype=torch.float32)) + self.rope_dims = 0 # set by GPT.__init__ for partial RoPE + self.rotary = Rotary(self.head_dim, base=rope_base, train_seq_len=1024) + self.use_xsa = False # set by GPT.__init__ for deep layers only + def _xsa_efficient(self, y: Tensor, v: Tensor) -> Tensor: + """Efficient XSA: subtract self-value projection via GQA-aware reshape (no repeat_interleave). + y: [B, T, H, D], v: [B, T, Hkv, D]. H must be divisible by Hkv.""" + B, T, H, D = y.shape + Hkv = v.size(-2) + group = H // Hkv + y_g = y.reshape(B, T, Hkv, group, D) # [B, T, Hkv, group, D] + vn = F.normalize(v, dim=-1).unsqueeze(-2) # [B, T, Hkv, 1, D] — broadcast ready + proj = (y_g * vn).sum(dim=-1, keepdim=True) * vn + return (y_g - proj).reshape(B, T, H, D) + def forward(self, x: Tensor, v_embed: Tensor | None = None) -> Tensor: + bsz, seqlen, dim = x.shape + q = self.c_q(x).reshape(bsz, seqlen, self.num_heads, self.head_dim) + k = self.c_k(x).reshape(bsz, seqlen, self.num_kv_heads, self.head_dim) + v = self.c_v(x) + if v_embed is not None: + v = v + v_embed + v = v.reshape(bsz, seqlen, self.num_kv_heads, self.head_dim) + q = F.rms_norm(q, (q.size(-1),)) + k = F.rms_norm(k, (k.size(-1),)) + cos, sin = self.rotary(seqlen, x.device, q.dtype) + q = apply_rotary_emb(q, cos, sin, self.rope_dims) + k = apply_rotary_emb(k, cos, sin, self.rope_dims) + q = q * self.q_gain.to(dtype=q.dtype)[None, None, :, None] + y = flash_attn_3_func(q, k, v, causal=True) + if self.use_xsa: + y = self._xsa_efficient(y, v) + y = y.reshape(bsz, seqlen, dim) + return self.proj(y) +class SmearGate(nn.Module): + def __init__(self, dim: int): + super().__init__() + self.gate = nn.Parameter(torch.zeros(dim, dtype=torch.float32)) + def forward(self, x: Tensor) -> Tensor: + g = torch.sigmoid(self.gate.to(dtype=x.dtype))[None, None, :] + x_prev = torch.cat([torch.zeros_like(x[:, :1]), x[:, :-1]], dim=1) + return (1 - g) * x + g * x_prev +class BigramHashEmbedding(nn.Module): + def __init__(self, bigram_vocab_size: int, bigram_dim: int, model_dim: int): + super().__init__() + self.bigram_vocab_size = bigram_vocab_size + self.embed = nn.Embedding(bigram_vocab_size, bigram_dim) + nn.init.zeros_(self.embed.weight) + self.proj = CastedLinear(bigram_dim, model_dim, bias=False) if bigram_dim != model_dim else None + if self.proj is not None: + nn.init.zeros_(self.proj.weight) + self.scale = nn.Parameter(torch.tensor(0.05, dtype=torch.float32)) + def bigram_hash(self, tokens: Tensor) -> Tensor: + t = tokens.to(torch.int32) + mod = self.bigram_vocab_size - 1 + out = torch.empty_like(t) + out[..., 0] = mod + out[..., 1:] = torch.bitwise_xor(36313 * t[..., 1:], 27191 * t[..., :-1]) % mod + return out.long() + def forward(self, token_ids: Tensor) -> Tensor: + h = self.embed(self.bigram_hash(token_ids)) + if self.proj is not None: + h = self.proj(h) + return h * self.scale.to(dtype=h.dtype) +class ValueEmbedding(nn.Module): + """Reinject token identity into attention values at specific layers. + Each table maps vocab tokens to a low-dim embedding, projected to model_dim.""" + def __init__(self, vocab_size: int, ve_dim: int, model_dim: int): + super().__init__() + self.embed = nn.Embedding(vocab_size, ve_dim) + nn.init.normal_(self.embed.weight, std=0.01) + self.proj = CastedLinear(ve_dim, model_dim, bias=False) if ve_dim != model_dim else None + if self.proj is not None: + nn.init.zeros_(self.proj.weight) + self.scale = nn.Parameter(torch.tensor(0.1, dtype=torch.float32)) + def forward(self, token_ids: Tensor) -> Tensor: + h = self.embed(token_ids) + if self.proj is not None: + h = self.proj(h) + return h * self.scale.to(dtype=h.dtype) +class MLP(nn.Module): + def __init__(self, dim: int, mlp_mult: int, mlp_act: str = "relu_sq", mlp_leaky_slope: float = 0.5): + super().__init__() + hidden = int(mlp_mult * dim) + self.fc = CastedLinear(dim, hidden, bias=False) + self.proj = CastedLinear(hidden, dim, bias=False) + self.proj._zero_init = True + self.mlp_act = mlp_act + self.mlp_leaky_slope = mlp_leaky_slope + if self.mlp_act not in {"relu_sq", "leaky_relu_sq"}: + raise ValueError(f"Unsupported MLP_ACT '{self.mlp_act}'. Use 'relu_sq' or 'leaky_relu_sq'.") + def forward(self, x: Tensor) -> Tensor: + x = self.fc(x) + if self.mlp_act == "leaky_relu_sq": + x = F.leaky_relu(x, negative_slope=self.mlp_leaky_slope) + else: + x = F.relu(x) + return self.proj(x.square()) +class Block(nn.Module): + def __init__( + self, + dim: int, + num_heads: int, + num_kv_heads: int, + mlp_mult: int, + rope_base: float, + qk_gain_init: float, + layer_idx: int = 0, + ln_scale: bool = False, + dtg: bool = False, + mlp_act: str = "relu_sq", + mlp_leaky_slope: float = 0.5, + ): + super().__init__() + self.attn_norm = RMSNorm() + self.mlp_norm = RMSNorm() + self.attn = CausalSelfAttention(dim, num_heads, num_kv_heads, rope_base, qk_gain_init) + self.mlp = MLP(dim, mlp_mult, mlp_act=mlp_act, mlp_leaky_slope=mlp_leaky_slope) + self.attn_scale = nn.Parameter(torch.ones(dim, dtype=torch.float32)) + self.mlp_scale = nn.Parameter(torch.ones(dim, dtype=torch.float32)) + self.resid_mix = nn.Parameter(torch.stack((torch.ones(dim), torch.zeros(dim))).float()) + self.ln_scale_factor = 1.0 / math.sqrt(layer_idx + 1) if ln_scale else 1.0 + if dtg: + self.dtg_gate = nn.Linear(dim, 1, bias=True) + nn.init.zeros_(self.dtg_gate.weight) + nn.init.constant_(self.dtg_gate.bias, 2.0) + else: + self.dtg_gate = None + def forward(self, x: Tensor, x0: Tensor, v_embed: Tensor | None = None) -> Tensor: + mix = self.resid_mix.to(dtype=x.dtype) + x_in = mix[0][None, None, :] * x + mix[1][None, None, :] * x0 + attn_out = self.attn(self.attn_norm(x_in) * self.ln_scale_factor, v_embed=v_embed) + x_out = x_in + self.attn_scale.to(dtype=x_in.dtype)[None, None, :] * attn_out + x_out = x_out + self.mlp_scale.to(dtype=x_out.dtype)[None, None, :] * self.mlp(self.mlp_norm(x_out) * self.ln_scale_factor) + if self.dtg_gate is not None: + gate = torch.sigmoid(self.dtg_gate(x_in.detach())) + x_out = x_in + gate * (x_out - x_in) + return x_out +class GPT(nn.Module): + def __init__( + self, + vocab_size: int, + num_layers: int, + model_dim: int, + num_heads: int, + num_kv_heads: int, + mlp_mult: int, + tie_embeddings: bool, + tied_embed_init_std: float, + logit_softcap: float, + rope_base: float, + qk_gain_init: float, + mtp_num_heads: int = 0, + mtp_loss_weight: float = 0.1, + bigram_vocab_size: int = 0, + bigram_dim: int = 128, + xsa_last_n: int = 0, + rope_dims: int = 0, + ln_scale: bool = False, + dtg: bool = False, + ve_enabled: bool = False, + ve_dim: int = 128, + ve_layers: str = "9,10", + mlp_act: str = "relu_sq", + mlp_leaky_slope: float = 0.5, + f1_corr_rank: int = 0, + f1_corr_scale_init: float = 0.10, + ): + super().__init__() + self._ve_target_dim = num_kv_heads * (model_dim // num_heads) # kv_dim for value projection + if logit_softcap <= 0.0: + raise ValueError(f"logit_softcap must be positive, got {logit_softcap}") + self.tie_embeddings = tie_embeddings + self.tied_embed_init_std = tied_embed_init_std + self.logit_softcap = logit_softcap + self.mtp_num_heads = mtp_num_heads + self.mtp_loss_weight = mtp_loss_weight + self.tok_emb = nn.Embedding(vocab_size, model_dim) + self.bigram = BigramHashEmbedding(bigram_vocab_size, bigram_dim, model_dim) if bigram_vocab_size > 0 else None + self.smear = SmearGate(model_dim) + self.num_encoder_layers = num_layers // 2 + self.num_decoder_layers = num_layers - self.num_encoder_layers + self.num_skip_weights = min(self.num_encoder_layers, self.num_decoder_layers) + self.skip_weights = nn.Parameter(torch.ones(self.num_skip_weights, model_dim, dtype=torch.float32)) + self.blocks = nn.ModuleList( + [ + Block( + model_dim, + num_heads, + num_kv_heads, + mlp_mult, + rope_base, + qk_gain_init, + layer_idx=i, + ln_scale=ln_scale, + dtg=dtg, + mlp_act=mlp_act, + mlp_leaky_slope=mlp_leaky_slope, + ) + for i in range(num_layers) + ] + ) + if rope_dims > 0: + head_dim = model_dim // num_heads + for block in self.blocks: + block.attn.rope_dims = rope_dims + block.attn.rotary = Rotary(head_dim, base=rope_base, train_seq_len=1024, rope_dims=rope_dims) + self.ve_layer_indices = [int(x) for x in ve_layers.split(",") if x.strip()] if ve_enabled else [] + kv_dim = self._ve_target_dim + if self.ve_layer_indices: + self.ve_shared = ValueEmbedding(vocab_size, ve_dim, kv_dim) + self.ve_layer_scales = nn.ParameterList( + [nn.Parameter(torch.ones(1, dtype=torch.float32)) for _ in self.ve_layer_indices] + ) + else: + self.ve_shared = None + self.ve_layer_scales = nn.ParameterList() + self.value_embeds = nn.ModuleList() # keep empty for compat + self.final_norm = RMSNorm() + self.lm_head = None if tie_embeddings else CastedLinear(model_dim, vocab_size, bias=False) + if self.lm_head is not None: + self.lm_head._zero_init = True + self.mtp_heads = nn.ModuleList( + [CastedLinear(model_dim, vocab_size, bias=False) for _ in range(mtp_num_heads)] + ) + for head in self.mtp_heads: + head._zero_init = True + # Low-rank correction path for extra capacity under size budget. + self.f1_corr_rank = f1_corr_rank + if f1_corr_rank > 0: + self.f1_corr_in = CastedLinear(model_dim, f1_corr_rank, bias=False) + self.f1_corr_out = CastedLinear(f1_corr_rank, vocab_size, bias=False) + self.f1_corr_out._zero_init = True + self.f1_corr_scale = nn.Parameter(torch.tensor(f1_corr_scale_init, dtype=torch.float32)) + else: + self.f1_corr_in = None + self.f1_corr_out = None + self.f1_corr_scale = None + if xsa_last_n > 0: + for i in range(max(0, num_layers - xsa_last_n), num_layers): + self.blocks[i].attn.use_xsa = True + self._init_weights() + def _init_weights(self) -> None: + if self.tie_embeddings: + nn.init.normal_(self.tok_emb.weight, mean=0.0, std=self.tied_embed_init_std) + num_layers = len(self.blocks) + for name, module in self.named_modules(): + if isinstance(module, nn.Linear): + if getattr(module, "_zero_init", False): + nn.init.zeros_(module.weight) + elif module.weight.ndim == 2 and module.weight.shape[0] >= 64 and module.weight.shape[1] >= 64: + nn.init.orthogonal_(module.weight, gain=1.0) + if ".proj." in name or name.endswith(".proj"): + with torch.no_grad(): + module.weight.mul_(1.0 / math.sqrt(2 * num_layers)) + def _get_ve(self, layer_idx: int, input_ids: Tensor, ve_cache: dict | None = None) -> Tensor | None: + """Get value embedding for a specific layer using shared table + per-layer scale.""" + if self.ve_shared is None or layer_idx not in self.ve_layer_indices: + return None + if ve_cache is not None and 've' not in ve_cache: + ve_cache['ve'] = self.ve_shared(input_ids) + ve_base = ve_cache['ve'] if ve_cache is not None else self.ve_shared(input_ids) + ve_idx = self.ve_layer_indices.index(layer_idx) + return ve_base * self.ve_layer_scales[ve_idx].to(dtype=ve_base.dtype) + def forward(self, input_ids: Tensor, target_ids: Tensor) -> Tensor: + x = self.tok_emb(input_ids) + if self.bigram is not None: + x = x + self.bigram(input_ids) + x = F.rms_norm(x, (x.size(-1),)) + x = self.smear(x) + x0 = x + skips: list[Tensor] = [] + ve_cache: dict = {} + for i in range(self.num_encoder_layers): + ve = self._get_ve(i, input_ids, ve_cache) + x = self.blocks[i](x, x0, v_embed=ve) + skips.append(x) + for i in range(self.num_decoder_layers): + bi = self.num_encoder_layers + i + if skips: + x = x + self.skip_weights[i].to(dtype=x.dtype)[None, None, :] * skips.pop() + ve = self._get_ve(bi, input_ids, ve_cache) + x = self.blocks[bi](x, x0, v_embed=ve) + x = self.final_norm(x) + x_flat = x.reshape(-1, x.size(-1)) + targets = target_ids.reshape(-1) + if self.tie_embeddings: + logits_proj = F.linear(x_flat, self.tok_emb.weight) + else: + if self.lm_head is None: + raise RuntimeError("lm_head is required when tie_embeddings=False") + logits_proj = self.lm_head(x_flat) + if self.f1_corr_in is not None and self.f1_corr_out is not None and self.f1_corr_scale is not None: + corr_hidden = F.silu(self.f1_corr_in(x_flat)) + corr_proj = self.f1_corr_out(corr_hidden) + logits_proj = logits_proj + self.f1_corr_scale.to(dtype=logits_proj.dtype) * corr_proj + logits = self.logit_softcap * torch.tanh(logits_proj / self.logit_softcap) + main_loss = F.cross_entropy(logits.float(), targets, reduction="mean") + if self.training and self.mtp_num_heads > 0 and self.mtp_loss_weight > 0.0: + _, seqlen, dim = x.shape + mtp_loss_sum = x.new_zeros(()) + mtp_loss_count = 0 + for k, mtp_head in enumerate(self.mtp_heads): + valid_t = seqlen - (k + 1) + if valid_t <= 0: + continue + mtp_hidden = x[:, :valid_t, :].reshape(-1, dim) + mtp_targets = target_ids[:, k + 1 :].reshape(-1) + mtp_logits_proj = mtp_head(mtp_hidden) + mtp_logits = self.logit_softcap * torch.tanh(mtp_logits_proj / self.logit_softcap) + mtp_loss_sum = mtp_loss_sum + F.cross_entropy(mtp_logits.float(), mtp_targets, reduction="mean") + mtp_loss_count += 1 + if mtp_loss_count > 0: + main_loss = main_loss + self.mtp_loss_weight * (mtp_loss_sum / mtp_loss_count) + return main_loss + def forward_logits(self, input_ids: Tensor) -> Tensor: + """Return logits (bsz, seq_len, vocab) without computing loss.""" + x = self.tok_emb(input_ids) + if self.bigram is not None: + x = x + self.bigram(input_ids) + x = F.rms_norm(x, (x.size(-1),)) + x = self.smear(x) + x0 = x + skips: list[Tensor] = [] + ve_cache: dict = {} + for i in range(self.num_encoder_layers): + ve = self._get_ve(i, input_ids, ve_cache) + x = self.blocks[i](x, x0, v_embed=ve) + skips.append(x) + for i in range(self.num_decoder_layers): + bi = self.num_encoder_layers + i + if skips: + x = x + self.skip_weights[i].to(dtype=x.dtype)[None, None, :] * skips.pop() + ve = self._get_ve(bi, input_ids, ve_cache) + x = self.blocks[bi](x, x0, v_embed=ve) + x = self.final_norm(x) + if self.tie_embeddings: + logits_proj = F.linear(x, self.tok_emb.weight) + else: + logits_proj = self.lm_head(x) + if self.f1_corr_in is not None and self.f1_corr_out is not None and self.f1_corr_scale is not None: + corr_hidden = F.silu(self.f1_corr_in(x)) + corr_proj = self.f1_corr_out(corr_hidden) + logits_proj = logits_proj + self.f1_corr_scale.to(dtype=logits_proj.dtype) * corr_proj + return self.logit_softcap * torch.tanh(logits_proj / self.logit_softcap) +def eval_val_sliding( + args: Hyperparameters, + base_model: nn.Module, + rank: int, + world_size: int, + device: torch.device, + val_tokens: Tensor, + base_bytes_lut: Tensor, + has_leading_space_lut: Tensor, + is_boundary_token_lut: Tensor, + stride: int, + batch_seqs: int = 32, + eval_seq_len: int | None = None, +) -> tuple[float, float]: + """Sliding window evaluation: each token scored with maximum context.""" + seq_len = eval_seq_len or args.train_seq_len + total_tokens = val_tokens.numel() - 1 + window_starts = [ws for ws in range(0, total_tokens, stride) + if min(ws + seq_len, total_tokens) - ws >= 1] + total_windows = len(window_starts) + my_s = (total_windows * rank) // world_size + my_e = (total_windows * (rank + 1)) // world_size + my_windows = window_starts[my_s:my_e] + loss_sum = torch.zeros((), device=device, dtype=torch.float64) + token_count = torch.zeros((), device=device, dtype=torch.float64) + byte_count = torch.zeros((), device=device, dtype=torch.float64) + base_model.eval() + compiled_logits = maybe_torch_compile(base_model.forward_logits, args) + with torch.inference_mode(): + for bi in range(0, len(my_windows), batch_seqs): + batch_ws = my_windows[bi:bi + batch_seqs] + bsz = len(batch_ws) + x_batch = torch.zeros(bsz, seq_len, dtype=torch.int64, device=device) + y_batch = torch.zeros(bsz, seq_len, dtype=torch.int64, device=device) + wlens: list[int] = [] + for i, ws in enumerate(batch_ws): + end = min(ws + seq_len, total_tokens) + wlen = end - ws + wlens.append(wlen) + chunk = val_tokens[ws:end + 1].to(dtype=torch.int64, device=device) + x_batch[i, :wlen] = chunk[:-1] + y_batch[i, :wlen] = chunk[1:] + with torch.autocast(device_type="cuda", dtype=torch.bfloat16): + logits = compiled_logits(x_batch) + nll = F.cross_entropy( + logits.reshape(-1, logits.size(-1)).float(), + y_batch.reshape(-1), + reduction="none", + ).reshape(bsz, seq_len) + for i, ws in enumerate(batch_ws): + wlen = wlens[i] + s = 0 if ws == 0 else max(wlen - stride, 0) + scored_nll = nll[i, s:wlen].to(torch.float64) + loss_sum += scored_nll.sum() + token_count += float(wlen - s) + tgt = y_batch[i, s:wlen] + prev = x_batch[i, s:wlen] + tb = base_bytes_lut[tgt].to(torch.float64) + tb += (has_leading_space_lut[tgt] & ~is_boundary_token_lut[prev]).to(torch.float64) + byte_count += tb.sum() + if dist.is_available() and dist.is_initialized(): + dist.all_reduce(loss_sum, op=dist.ReduceOp.SUM) + dist.all_reduce(token_count, op=dist.ReduceOp.SUM) + dist.all_reduce(byte_count, op=dist.ReduceOp.SUM) + val_loss = (loss_sum / token_count).item() + bits_per_token = val_loss / math.log(2.0) + tokens_per_byte = token_count.item() / byte_count.item() + base_model.train() + return val_loss, bits_per_token * tokens_per_byte +def eval_val_sliding_hashed_ngram( + args: Hyperparameters, + base_model: nn.Module, + rank: int, + world_size: int, + device: torch.device, + val_tokens: Tensor, + base_bytes_lut: Tensor, + has_leading_space_lut: Tensor, + is_boundary_token_lut: Tensor, + stride: int, + order: int, + alpha: float, + min_count: int, + buckets: int, + max_seconds: float = 0.0, + batch_seqs: int = 32, + eval_seq_len: int | None = None, +) -> tuple[float, float, float]: + """Score-first sliding eval with fixed-weight hashed n-gram interpolation. + + Legal behavior: + - per-token score is computed before that token updates the cache + - no target-aware gating/min-NLL selection is used + """ + if order < 2: + raise ValueError(f"NGRAM_EVAL_ORDER must be >=2, got {order}") + if not (0.0 <= alpha <= 1.0): + raise ValueError(f"NGRAM_EVAL_ALPHA must be in [0, 1], got {alpha}") + if min_count < 1: + raise ValueError(f"NGRAM_EVAL_MIN_COUNT must be >=1, got {min_count}") + if buckets < 1024: + raise ValueError(f"NGRAM_EVAL_BUCKETS must be >=1024, got {buckets}") + if max_seconds < 0.0: + raise ValueError(f"NGRAM_EVAL_MAX_SECONDS must be >=0, got {max_seconds}") + + seq_len = eval_seq_len or args.train_seq_len + total_tokens = val_tokens.numel() - 1 + all_window_starts = [ws for ws in range(0, total_tokens, stride) if min(ws + seq_len, total_tokens) - ws >= 1] + total_scored_tokens = 0.0 + for ws in all_window_starts: + end = min(ws + seq_len, total_tokens) + wlen = end - ws + s = 0 if ws == 0 else max(wlen - stride, 0) + total_scored_tokens += float(max(wlen - s, 0)) + # Distribute windows across ranks + my_s = (len(all_window_starts) * rank) // world_size + my_e = (len(all_window_starts) * (rank + 1)) // world_size + window_starts = all_window_starts[my_s:my_e] + + val_np = val_tokens.numpy() + ctx_table = np.zeros((buckets,), dtype=np.uint32) + full_table = np.zeros((buckets,), dtype=np.uint32) + mask = np.uint64(buckets - 1) + primes = np.array( + [np.uint64(36313), np.uint64(27191), np.uint64(51647), np.uint64(81929), np.uint64(131071)], + dtype=np.uint64, + ) + + loss_sum = 0.0 + token_count = 0.0 + byte_count = 0.0 + + base_model.eval() + compiled_logits = maybe_torch_compile(base_model.forward_logits, args) + t0 = time.perf_counter() + deadline = (t0 + max_seconds) if max_seconds > 0.0 else None + cutoff_hit = False + with torch.inference_mode(): + for bi in range(0, len(window_starts), batch_seqs): + if deadline is not None and time.perf_counter() >= deadline: + cutoff_hit = True + break + batch_ws = window_starts[bi:bi + batch_seqs] + bsz = len(batch_ws) + x_batch = torch.zeros(bsz, seq_len, dtype=torch.int64, device=device) + y_batch = torch.zeros(bsz, seq_len, dtype=torch.int64, device=device) + wlens: list[int] = [] + for i, ws in enumerate(batch_ws): + end = min(ws + seq_len, total_tokens) + wlen = end - ws + wlens.append(wlen) + chunk = val_tokens[ws:end + 1].to(dtype=torch.int64, device=device) + x_batch[i, :wlen] = chunk[:-1] + y_batch[i, :wlen] = chunk[1:] + + with torch.autocast(device_type="cuda", dtype=torch.bfloat16): + logits = compiled_logits(x_batch) + nll = F.cross_entropy( + logits.reshape(-1, logits.size(-1)).float(), + y_batch.reshape(-1), + reduction="none", + ).reshape(bsz, seq_len) + + for i, ws in enumerate(batch_ws): + wlen = wlens[i] + s = 0 if ws == 0 else max(wlen - stride, 0) + seg_len = wlen - s + if seg_len <= 0: + continue + + seg_nll = nll[i, s:wlen].to(torch.float64).cpu().numpy() + seg_model_p = np.exp(-seg_nll) + + global_j = np.arange(ws + s + 1, ws + wlen + 1, dtype=np.int64) + valid = global_j >= (order - 1) + if valid.any(): + v_idx = np.nonzero(valid)[0] + jv = global_j[v_idx] + + ctx_hash = np.zeros((len(jv),), dtype=np.uint64) + ctx_width = order - 1 + for k in range(ctx_width): + tok = val_np[jv - (ctx_width - k)].astype(np.uint64) + ctx_hash ^= tok * primes[k % len(primes)] + ctx_key = (ctx_hash & mask).astype(np.int64) + + tgt_np = val_np[jv].astype(np.uint64) + full_key = ((ctx_hash ^ (tgt_np * primes[ctx_width % len(primes)])) & mask).astype(np.int64) + + ctx_counts = ctx_table[ctx_key].astype(np.float64) + full_counts = full_table[full_key].astype(np.float64) + can_mix = ctx_counts >= float(min_count) + if can_mix.any(): + # Collision-safe estimate: ensure n-gram probability stays in [0, 1]. + # With hashed sketches, full_counts can exceed ctx_counts due collisions. + p_ng = np.minimum(full_counts, ctx_counts) / np.maximum(ctx_counts, 1.0) + p_ng = np.clip(p_ng, 0.0, 1.0) + mixed = (1.0 - alpha) * seg_model_p[v_idx] + alpha * p_ng + seg_model_p[v_idx[can_mix]] = mixed[can_mix] + seg_nll = -np.log(np.clip(seg_model_p, 1e-12, 1.0)) + + # Score-first legality: update cache only after segment scoring. + np.add.at(ctx_table, ctx_key, 1) + np.add.at(full_table, full_key, 1) + + loss_sum += float(seg_nll.sum()) + token_count += float(seg_len) + tgt = y_batch[i, s:wlen] + prev = x_batch[i, s:wlen] + tb = base_bytes_lut[tgt].to(torch.float64) + tb += (has_leading_space_lut[tgt] & ~is_boundary_token_lut[prev]).to(torch.float64) + byte_count += float(tb.sum().item()) + + if (bi // batch_seqs) % 2000 == 0 and bi > 0: + elapsed = time.perf_counter() - t0 + prog = min((bi + bsz) / max(len(window_starts), 1), 1.0) + cur_bpb = (loss_sum / max(token_count, 1.0)) / math.log(2.0) * (token_count / max(byte_count, 1.0)) + print( + f"ngram_eval:progress windows={bi + bsz}/{len(window_starts)} " + f"({prog*100:.1f}%) bpb={cur_bpb:.6f} t={elapsed:.0f}s", + flush=True, + ) + # All-reduce across ranks + _loss = torch.tensor(loss_sum, device=device, dtype=torch.float64) + _toks = torch.tensor(token_count, device=device, dtype=torch.float64) + _bytes = torch.tensor(byte_count, device=device, dtype=torch.float64) + if dist.is_available() and dist.is_initialized(): + dist.all_reduce(_loss, op=dist.ReduceOp.SUM) + dist.all_reduce(_toks, op=dist.ReduceOp.SUM) + dist.all_reduce(_bytes, op=dist.ReduceOp.SUM) + loss_sum = _loss.item() + token_count = _toks.item() + byte_count = _bytes.item() + + coverage = token_count / max(total_scored_tokens, 1.0) + if cutoff_hit: + elapsed = time.perf_counter() - t0 + print( + f"ngram_eval:cutoff max_seconds={max_seconds:.1f} " + f"coverage={coverage*100:.2f}% elapsed={elapsed:.0f}s", + flush=True, + ) + + val_loss = loss_sum / max(token_count, 1.0) + val_bpb = val_loss / math.log(2.0) * (token_count / max(byte_count, 1.0)) + base_model.train() + return val_loss, val_bpb, coverage +def _classify_param(name: str) -> str: + if "tok_emb" in name or "lm_head" in name: + return "embed" + if "f1_corr_in" in name or "f1_corr_out" in name: + return "aux" + if ".mlp." in name: + return "mlp" + if ".attn." in name or (".proj." in name and ".mlp." not in name): + return "attn" + return "other" +def eval_val_sliding_ttt( + args: Hyperparameters, base_model: nn.Module, rank: int, world_size: int, + device: torch.device, val_tokens: Tensor, base_bytes_lut: Tensor, + has_leading_space_lut: Tensor, is_boundary_token_lut: Tensor, + stride: int, batch_seqs: int = 32, +) -> tuple[float, float]: + seq_len, total_tokens, ttt_chunk = args.train_seq_len, val_tokens.numel() - 1, args.ttt_chunk_tokens + master = (rank == 0) + log0 = (lambda msg: print(msg, flush=True)) if master else (lambda msg: None) + window_starts = [ws for ws in range(0, total_tokens, stride) if min(ws + seq_len, total_tokens) - ws >= stride or ws == 0] + num_chunks = (total_tokens + ttt_chunk - 1) // ttt_chunk + chunk_windows: list[list[int]] = [[] for _ in range(num_chunks)] + for ws in window_starts: + end, wlen = min(ws + seq_len, total_tokens), min(ws + seq_len, total_tokens) - ws + s = 0 if ws == 0 else max(wlen - stride, 0) + chunk_windows[min((ws + s) // ttt_chunk, num_chunks - 1)].append(ws) + log0(f"ttt_sliding:start chunks={num_chunks} windows={len(window_starts)} lr={args.ttt_lr} epochs={args.ttt_epochs} freeze={args.ttt_freeze_blocks}") + loss_sum = torch.zeros((), device=device, dtype=torch.float64) + token_count = torch.zeros((), device=device, dtype=torch.float64) + byte_count = torch.zeros((), device=device, dtype=torch.float64) + frozen_ids = set(range(min(args.ttt_freeze_blocks, len(base_model.blocks)))) + embed_names = {"tok_emb", "bigram", "ve_shared"} if args.ttt_freeze_embed else set() + ttt_params = [] + for name, p in base_model.named_parameters(): + if any(f"blocks.{bi}." in name for bi in frozen_ids): + p.requires_grad_(False) + elif any(en in name for en in embed_names): + p.requires_grad_(False) + else: + p.requires_grad_(True); ttt_params.append(p) + log0(f"ttt_sliding:unfrozen={sum(p.numel() for p in ttt_params)} freeze_embed={args.ttt_freeze_embed}") + optimizer = torch.optim.SGD(ttt_params, lr=args.ttt_lr, momentum=args.ttt_momentum) + # TTT-EMA: maintain smoothed weights for scoring + ema_decay = args.ttt_ema_decay + ema_state = None + raw_state = None + if ema_decay > 0: + ema_state = {n: p.data.clone() for n, p in base_model.named_parameters() if p.requires_grad} + raw_state = {n: torch.empty_like(p.data) for n, p in base_model.named_parameters() if n in ema_state} + log0(f"ttt_sliding:ema_decay={ema_decay} ema_params={len(ema_state)}") + t0 = time.perf_counter() + cur_lr = args.ttt_lr + for ci in range(num_chunks): + windows = chunk_windows[ci] + if not windows: + continue + chunk_start, chunk_end = ci * ttt_chunk, min((ci + 1) * ttt_chunk, total_tokens) + my_windows = windows[(len(windows) * rank) // world_size:(len(windows) * (rank + 1)) // world_size] + # Swap to EMA weights for scoring (if enabled and past first chunk) + if ema_state is not None and ci > 0: + for n, p in base_model.named_parameters(): + if n in ema_state: + raw_state[n].copy_(p.data) + p.data.copy_(ema_state[n]) + base_model.eval() + with torch.inference_mode(): + for bi in range(0, len(my_windows), batch_seqs): + batch_ws = my_windows[bi:bi + batch_seqs] + bsz = len(batch_ws) + x_batch = torch.zeros(bsz, seq_len, dtype=torch.int64, device=device) + y_batch = torch.zeros(bsz, seq_len, dtype=torch.int64, device=device) + wlens = [] + for i, ws in enumerate(batch_ws): + wlen = min(ws + seq_len, total_tokens) - ws; wlens.append(wlen) + ct = val_tokens[ws:ws + wlen + 1].to(dtype=torch.int64, device=device) + x_batch[i, :wlen] = ct[:-1]; y_batch[i, :wlen] = ct[1:] + with torch.autocast(device_type="cuda", dtype=torch.bfloat16): + logits = base_model.forward_logits(x_batch) + nll = F.cross_entropy( + logits.reshape(-1, logits.size(-1)).float(), + y_batch.reshape(-1), reduction="none", + ).reshape(bsz, seq_len) + for i, ws in enumerate(batch_ws): + wlen, s = wlens[i], 0 if ws == 0 else max(wlens[i] - stride, 0) + loss_sum += nll[i, s:wlen].to(torch.float64).sum(); token_count += float(wlen - s) + tgt, prev = y_batch[i, s:wlen], x_batch[i, s:wlen] + tb = base_bytes_lut[tgt].to(torch.float64) + (has_leading_space_lut[tgt] & ~is_boundary_token_lut[prev]).to(torch.float64) + byte_count += tb.sum() + # Restore raw weights after scoring (for training phase) + if ema_state is not None and ci > 0: + for n, p in base_model.named_parameters(): + if n in raw_state: + p.data.copy_(raw_state[n]) + # Phase 2: TRAIN on this chunk (already scored = legal) + if ci < num_chunks - 1 and ci < args.ttt_max_train_chunks and args.ttt_epochs > 0: + base_model.train() + chunk_seqs = (chunk_end - chunk_start) // seq_len + if chunk_seqs > 0: + cur_lr = args.ttt_lr * 0.5 * (1.0 + math.cos(math.pi * ci / max(args.ttt_max_train_chunks - 1, 1))) + for pg in optimizer.param_groups: + pg['lr'] = cur_lr + ms, me = (chunk_seqs * rank) // world_size, (chunk_seqs * (rank + 1)) // world_size + for _ep in range(args.ttt_epochs): + for bs in range(0, me - ms, args.ttt_batch_seqs): + be = min(bs + args.ttt_batch_seqs, me - ms) + start_tok = chunk_start + (ms + bs) * seq_len + end_tok = chunk_start + (ms + be) * seq_len + 1 + if end_tok > val_tokens.numel(): + continue + local = val_tokens[start_tok:end_tok].to(device=device, dtype=torch.int64) + x = local[:-1].reshape(-1, seq_len) + y = local[1:].reshape(-1, seq_len) + optimizer.zero_grad(set_to_none=True) + with torch.autocast(device_type="cuda", dtype=torch.bfloat16): + loss = base_model(x, y) + loss.backward() + if world_size > 1: + for p in ttt_params: + if p.grad is not None: + dist.all_reduce(p.grad, op=dist.ReduceOp.AVG) + torch.nn.utils.clip_grad_norm_(ttt_params, args.ttt_grad_clip) + optimizer.step() + # Update EMA after this chunk's training + if ema_state is not None: + with torch.no_grad(): + for n, p in base_model.named_parameters(): + if n in ema_state: + ema_state[n].mul_(ema_decay).add_(p.data, alpha=1.0 - ema_decay) + # Once training stops, load EMA weights permanently for remaining score-only chunks + if ema_state is not None and ci == args.ttt_max_train_chunks: + log0(f" ttt:loading EMA weights permanently at chunk {ci}") + for n, p in base_model.named_parameters(): + if n in ema_state: + p.data.copy_(ema_state[n]) + ema_state = None + raw_state = None + if master and (ci % 5 == 0 or ci == num_chunks - 1): + rl = loss_sum.item() / max(token_count.item(), 1) + cur_bpb = rl / math.log(2) * (token_count.item() / max(byte_count.item(), 1)) if token_count.item() > 0 else 0 + lr_str = f" lr={cur_lr:.6f}" if ci < args.ttt_max_train_chunks else " lr=done" + log0(f" ttt[{ci+1}/{num_chunks}] bpb={cur_bpb:.6f}{lr_str} t={time.perf_counter()-t0:.0f}s") + if dist.is_available() and dist.is_initialized(): + for t in [loss_sum, token_count, byte_count]: + dist.all_reduce(t, op=dist.ReduceOp.SUM) + val_loss = (loss_sum / token_count).item() + val_bpb = val_loss / math.log(2.0) * (token_count.item() / byte_count.item()) + for p in base_model.parameters(): + p.requires_grad_(True) + base_model.eval() + log0(f"ttt_sliding:done loss={val_loss:.6f} bpb={val_bpb:.6f} time={time.perf_counter()-t0:.0f}s") + return val_loss, val_bpb +# --------------------------------------------------------------------------- +# GPTQ: Hessian-aware quantization with column-wise error compensation +# --------------------------------------------------------------------------- +def _find_best_row_scales(W: Tensor, clip_range: int = 31) -> Tensor: + """Find optimal per-row scales by searching percentile clipping thresholds.""" + t32 = W.float() + best_s = t32.abs().amax(dim=1) / clip_range + best_s = best_s.clamp_min(1.0 / clip_range) + best_err = torch.full((t32.shape[0],), float('inf')) + for pct in [0.9990, 0.9995, 0.9999, 0.99999, 1.0]: + if pct < 1.0: + row_clip = torch.quantile(t32.abs(), pct, dim=1) + else: + row_clip = t32.abs().amax(dim=1) + s = (row_clip / clip_range).clamp_min(1.0 / clip_range) + q = torch.clamp(torch.round(t32 / s[:, None]), -clip_range, clip_range) + recon = q * s[:, None] + err = (t32 - recon).pow(2).mean(dim=1) + improved = err < best_err + best_s[improved] = s[improved] + best_err[improved] = err[improved] + return best_s +def gptq_quantize_weight(W: Tensor, H: Tensor, clip_range: int = 31, + block_size: int = 64, percdamp: float = 0.002) -> tuple[Tensor, Tensor]: + """GPTQ: quantize weight matrix W using Hessian H = X^T X for error compensation. + Uses pre-computed per-row scales and column reordering by Hessian diagonal. + Returns (quantized_int8, scale_fp16) in int6 range [-clip_range, clip_range].""" + W = W.float().clone() + rows, cols = W.shape + # Pre-compute optimal per-row scales from the original weight matrix + row_scale = _find_best_row_scales(W, clip_range) + H = H.float().clone() + damp = percdamp * H.diag().mean() + H.diagonal().add_(damp) + # Column reordering: process least-important columns first (ascending H_diag) + perm = torch.argsort(H.diag()) + invperm = torch.argsort(perm) + W = W[:, perm] + H = H[perm][:, perm] + try: + L = torch.linalg.cholesky(H) + Hinv = torch.cholesky_inverse(L) + except torch._C._LinAlgError: + Hinv = torch.diag(1.0 / H.diag().clamp_min(1e-6)) + Q = torch.zeros(rows, cols, dtype=torch.int8) + for i1 in range(0, cols, block_size): + i2 = min(i1 + block_size, cols) + W_block = W[:, i1:i2].clone() + Hinv_block = Hinv[i1:i2, i1:i2] + Err = torch.zeros_like(W_block) + for j in range(i2 - i1): + w_col = W_block[:, j] + h_inv_jj = Hinv_block[j, j].clamp_min(1e-8) + # Quantize using pre-computed per-row scales + q_col = torch.clamp(torch.round(w_col / row_scale), -clip_range, clip_range) + deq_col = q_col * row_scale + Q[:, i1 + j] = q_col.to(torch.int8) + err = (w_col - deq_col) / h_inv_jj + Err[:, j] = err + if j + 1 < i2 - i1: + W_block[:, j + 1:] -= err.unsqueeze(1) * Hinv_block[j, j + 1:].unsqueeze(0) + if i2 < cols: + W[:, i2:] -= Err @ Hinv[i1:i2, i2:] + # Undo column reordering + Q = Q[:, invperm] + return Q, row_scale.to(torch.float16) +def gptq_calibrate(model: nn.Module, train_pattern: str, device: torch.device, + n_samples: int = 256, seq_len: int = 2048) -> dict[str, Tensor]: + """Collect Hessian H = X^T X for each linear layer using training data.""" + hessians: dict[str, Tensor] = {} + n_seen: dict[str, int] = {} + hooks = [] + def make_hook(name: str): + def hook_fn(module, inp, out): + x = inp[0].detach().float() + if x.ndim == 3: + x = x.reshape(-1, x.shape[-1]) + if name not in hessians: + hessians[name] = torch.zeros(x.shape[1], x.shape[1], device=x.device, dtype=torch.float32) + n_seen[name] = 0 + hessians[name].addmm_(x.t(), x) + n_seen[name] += x.shape[0] + return hook_fn + for name, module in model.named_modules(): + if isinstance(module, (nn.Linear, CastedLinear)): + hooks.append(module.register_forward_hook(make_hook(name))) + stream = TokenStream(train_pattern) + model.eval() + with torch.no_grad(): + for _ in range(n_samples): + tokens = stream.take(seq_len + 1).to(device=device, dtype=torch.int64) + x = tokens[:-1].unsqueeze(0) + with torch.autocast(device_type="cuda", dtype=torch.bfloat16): + model.forward_logits(x) + for h in hooks: + h.remove() + for name in hessians: + hessians[name] /= max(n_seen[name], 1) + return hessians +def mixed_quantize_int6_gptq(state_dict: dict[str, Tensor], int6_cats: set[str], + hessians: dict[str, Tensor]) -> tuple[dict, dict]: + """Like mixed_quantize_int6 but uses GPTQ for int6 categories when Hessian available.""" + result: dict[str, Tensor] = {} + meta: dict[str, object] = {} + gptq_count, naive_count = 0, 0 + for name, tensor in state_dict.items(): + t = tensor.detach().cpu().contiguous() + cat = _classify_param(name) + if not t.is_floating_point() or t.numel() <= 65536: + result[name] = t.to(torch.float16) if t.is_floating_point() else t + meta[name] = "passthrough" + continue + if any(p in name for p in CONTROL_TENSOR_NAME_PATTERNS): + result[name] = t.float() + meta[name] = "passthrough_ctrl" + continue + if cat in int6_cats and t.ndim == 2: + module_name = name.rsplit(".weight", 1)[0] if name.endswith(".weight") else name + H = hessians.get(module_name) + if H is not None and H.shape[0] == t.shape[1]: + q, s = gptq_quantize_weight(t, H.cpu()) + gptq_count += 1 + else: + q, s = quantize_int6_per_row(t) + naive_count += 1 + result[name + ".q"] = q + result[name + ".scale"] = s + meta[name] = {"type": "int6"} + elif cat in int6_cats and t.ndim >= 1: + q, s = quantize_int6_per_row(t) + result[name + ".q"] = q + result[name + ".scale"] = s + meta[name] = {"type": "int6"} + naive_count += 1 + else: + q, s = quantize_float_tensor(t) + result[name + ".q"] = q + result[name + ".scale"] = s + meta[name] = {"type": "int8"} + print(f"gptq_quantize: {gptq_count} GPTQ layers, {naive_count} naive layers", flush=True) + return result, meta +def quantize_int6_per_row(t: Tensor, clip_range: int = 31) -> tuple[Tensor, Tensor]: + t32 = t.float() + if t32.ndim == 2: + best_q, best_s, best_err = None, None, float('inf') + for pct in [0.9990, 0.9995, 0.9999, 0.99999, 1.0]: + if pct < 1.0: + row_clip = torch.quantile(t32.abs(), pct, dim=1) + else: + row_clip = t32.abs().amax(dim=1) + s = (row_clip / clip_range).clamp_min(1.0 / clip_range).to(torch.float16) + q = torch.clamp(torch.round(t32 / s.float()[:, None]), -clip_range, clip_range).to(torch.int8) + recon = q.float() * s.float()[:, None] + err = (t32 - recon).pow(2).mean().item() + if err < best_err: + best_q, best_s, best_err = q, s, err + return best_q, best_s + amax = t32.abs().max().item() + scale = torch.tensor(amax / clip_range if amax > 0 else 1.0, dtype=torch.float16) + q = torch.clamp(torch.round(t32 / scale.float()), -clip_range, clip_range).to(torch.int8) + return q, scale +def mixed_quantize_int6(state_dict: dict[str, Tensor], int6_cats: set[str]): + num_layers_total = max( + (int(k.split(".")[1]) for k in state_dict if k.startswith("blocks.")), + default=0, + ) + 1 + late_k_layers = set(range(num_layers_total - 2, num_layers_total)) + result: dict[str, Tensor] = {} + meta: dict[str, object] = {} + for name, tensor in state_dict.items(): + t = tensor.detach().cpu().contiguous() + cat = _classify_param(name) + if not t.is_floating_point() or t.numel() <= 65536: + result[name] = t.to(torch.float16) if t.is_floating_point() else t + meta[name] = "passthrough" + continue + if any(p in name for p in CONTROL_TENSOR_NAME_PATTERNS): + result[name] = t.float() + meta[name] = "passthrough_ctrl" + continue + if cat in int6_cats and t.ndim >= 1: + q, s = quantize_int6_per_row(t) + result[name + ".q"] = q + result[name + ".scale"] = s + meta[name] = {"type": "int6"} + else: + q, s = quantize_float_tensor(t) + result[name + ".q"] = q + result[name + ".scale"] = s + meta[name] = {"type": "int8"} + return result, meta +def dequantize_mixed_int6(result: dict[str, Tensor], meta: dict[str, object], + template_sd: dict[str, Tensor]) -> dict[str, Tensor]: + out: dict[str, Tensor] = {} + for name, orig in template_sd.items(): + info = meta.get(name) + if info is None: + continue + orig_dtype = orig.dtype + if info in ("passthrough", "passthrough_ctrl", "passthrough_fp16"): + t = result[name] + if t.dtype == torch.float16 and orig_dtype in (torch.float32, torch.bfloat16): + t = t.to(orig_dtype) + out[name] = t + continue + q, s = result[name + ".q"], result[name + ".scale"] + if s.ndim > 0: + out[name] = (q.float() * s.float().view(q.shape[0], *([1] * (q.ndim - 1)))).to(orig_dtype) + else: + out[name] = (q.float() * float(s.item())).to(orig_dtype) + return out +def main() -> None: + global zeropower_via_newtonschulz5 + code = Path(__file__).read_text(encoding="utf-8") + args = Hyperparameters() + if args.compile_enabled: + zeropower_via_newtonschulz5 = torch.compile(zeropower_via_newtonschulz5) + distributed = "RANK" in os.environ and "WORLD_SIZE" in os.environ + rank = int(os.environ.get("RANK", "0")) + world_size = int(os.environ.get("WORLD_SIZE", "1")) + local_rank = int(os.environ.get("LOCAL_RANK", "0")) + if world_size <= 0: + raise ValueError(f"WORLD_SIZE must be positive, got {world_size}") + if 8 % world_size != 0: + raise ValueError(f"WORLD_SIZE={world_size} must divide 8 so grad_accum_steps stays integral") + grad_accum_steps = 8 // world_size + grad_scale = 1.0 / grad_accum_steps + if not torch.cuda.is_available(): + raise RuntimeError("CUDA is required") + device = torch.device("cuda", local_rank) + torch.cuda.set_device(device) + if distributed: + dist.init_process_group(backend="nccl", device_id=device) + dist.barrier() + master_process = rank == 0 + torch.backends.cuda.matmul.allow_tf32 = True + torch.backends.cudnn.allow_tf32 = True + from torch.backends.cuda import enable_cudnn_sdp, enable_flash_sdp, enable_math_sdp, enable_mem_efficient_sdp + enable_cudnn_sdp(False) + enable_flash_sdp(True) + enable_mem_efficient_sdp(False) + enable_math_sdp(False) + logfile = None + if master_process: + os.makedirs("logs", exist_ok=True) + logfile = f"logs/{args.run_id}.txt" + print(logfile) + def log0(msg: str, console: bool = True) -> None: + if not master_process: + return + if console: + print(msg) + if logfile is not None: + with open(logfile, "a", encoding="utf-8") as f: + print(msg, file=f) + log0(code, console=False) + log0("=" * 100, console=False) + log0(f"Running Python {sys.version}", console=False) + log0(f"Running PyTorch {torch.__version__}", console=False) + log0( + subprocess.run(["nvidia-smi"], stdout=subprocess.PIPE, stderr=subprocess.PIPE, text=True, check=False).stdout, + console=False, + ) + log0("=" * 100, console=False) + random.seed(args.seed) + np.random.seed(args.seed) + torch.manual_seed(args.seed) + torch.cuda.manual_seed_all(args.seed) + if not args.tokenizer_path.endswith(".model"): + raise ValueError(f"Script only setup for SentencePiece .model file: {args.tokenizer_path}") + sp = spm.SentencePieceProcessor(model_file=args.tokenizer_path) + if int(sp.vocab_size()) != args.vocab_size: + raise ValueError( + f"VOCAB_SIZE={args.vocab_size} does not match tokenizer vocab_size={int(sp.vocab_size())}" + ) + dataset_dir = Path(args.data_path).resolve() + actual_train_files = len(list(dataset_dir.glob("fineweb_train_*.bin"))) + effective_eval_seq_len = args.eval_seq_len if args.eval_seq_len > 0 else args.train_seq_len + val_seq_len = max(args.train_seq_len, effective_eval_seq_len) + val_tokens = load_validation_tokens(args.val_files, val_seq_len) + base_bytes_lut, has_leading_space_lut, is_boundary_token_lut = build_sentencepiece_luts( + sp, args.vocab_size, device + ) + log0(f"val_bpb:enabled tokenizer_kind=sentencepiece tokenizer_path={args.tokenizer_path}") + log0(f"train_loader:dataset:{dataset_dir.name} train_shards:{actual_train_files}") + log0(f"val_loader:shards pattern={args.val_files} tokens:{val_tokens.numel() - 1}") + CastedLinear._qat_enabled = args.qat_enabled + base_model = GPT( + vocab_size=args.vocab_size, + num_layers=args.num_layers, + model_dim=args.model_dim, + num_heads=args.num_heads, + num_kv_heads=args.num_kv_heads, + mlp_mult=args.mlp_mult, + tie_embeddings=args.tie_embeddings, + tied_embed_init_std=args.tied_embed_init_std, + logit_softcap=args.logit_softcap, + rope_base=args.rope_base, + qk_gain_init=args.qk_gain_init, + mtp_num_heads=args.mtp_num_heads, + mtp_loss_weight=args.mtp_loss_weight, + bigram_vocab_size=args.bigram_vocab_size, + bigram_dim=args.bigram_dim, + xsa_last_n=args.xsa_last_n, + rope_dims=args.rope_dims, + ln_scale=args.ln_scale, + dtg=args.dtg_enabled, + ve_enabled=args.ve_enabled, + ve_dim=args.ve_dim, + ve_layers=args.ve_layers, + mlp_act=args.mlp_act, + mlp_leaky_slope=args.mlp_leaky_slope, + f1_corr_rank=args.f1_corr_rank, + f1_corr_scale_init=args.f1_corr_scale_init, + ).to(device).bfloat16() + for module in base_model.modules(): + if isinstance(module, CastedLinear): + module.float() + restore_low_dim_params_to_fp32(base_model) + compiled_model = maybe_torch_compile(base_model, args) + model: nn.Module = DDP(compiled_model, device_ids=[local_rank], broadcast_buffers=False) if distributed else compiled_model + block_named_params = list(base_model.blocks.named_parameters()) + matrix_params = [ + p + for name, p in block_named_params + if p.ndim == 2 and not any(pattern in name for pattern in CONTROL_TENSOR_NAME_PATTERNS) + ] + if base_model.mtp_num_heads > 0: + matrix_params.extend([p for p in base_model.mtp_heads.parameters() if p.ndim == 2]) + if base_model.f1_corr_in is not None and base_model.f1_corr_out is not None: + matrix_params.append(base_model.f1_corr_in.weight) + matrix_params.append(base_model.f1_corr_out.weight) + scalar_params = [ + p + for name, p in block_named_params + if p.ndim < 2 or any(pattern in name for pattern in CONTROL_TENSOR_NAME_PATTERNS) + ] + if base_model.skip_weights.numel() > 0: + scalar_params.append(base_model.skip_weights) + scalar_params.append(base_model.smear.gate) + if base_model.bigram is not None: + scalar_params.append(base_model.bigram.scale) + if base_model.f1_corr_scale is not None: + scalar_params.append(base_model.f1_corr_scale) + token_lr = args.tied_embed_lr if args.tie_embeddings else args.embed_lr + tok_params = [{"params": [base_model.tok_emb.weight], "lr": token_lr, "base_lr": token_lr}] + if base_model.bigram is not None: + tok_params.append({"params": [base_model.bigram.embed.weight], "lr": token_lr, "base_lr": token_lr}) + if base_model.bigram.proj is not None: + matrix_params.append(base_model.bigram.proj.weight) + if base_model.ve_shared is not None: + tok_params.append({"params": [base_model.ve_shared.embed.weight], "lr": token_lr, "base_lr": token_lr}) + if base_model.ve_shared.proj is not None: + matrix_params.append(base_model.ve_shared.proj.weight) + scalar_params.append(base_model.ve_shared.scale) + for s in base_model.ve_layer_scales: + scalar_params.append(s) + optimizer_tok = torch.optim.AdamW( + tok_params, + betas=(args.beta1, args.beta2), + eps=args.adam_eps, + weight_decay=args.adam_wd, + fused=True, + ) + optimizer_muon = Muon( + matrix_params, + lr=args.matrix_lr, + momentum=args.muon_momentum, + backend_steps=args.muon_backend_steps, + weight_decay=args.muon_wd, + ) + for group in optimizer_muon.param_groups: + group["base_lr"] = args.matrix_lr + optimizer_scalar = torch.optim.AdamW( + [{"params": scalar_params, "lr": args.scalar_lr, "base_lr": args.scalar_lr}], + betas=(args.beta1, args.beta2), + eps=args.adam_eps, + weight_decay=args.adam_wd, + fused=True, + ) + optimizers: list[torch.optim.Optimizer] = [optimizer_tok, optimizer_muon, optimizer_scalar] + if base_model.lm_head is not None: + optimizer_head = torch.optim.Adam( + [{"params": [base_model.lm_head.weight], "lr": args.head_lr, "base_lr": args.head_lr}], + betas=(args.beta1, args.beta2), + eps=args.adam_eps, + fused=True, + ) + optimizers.insert(1, optimizer_head) + n_params = sum(p.numel() for p in base_model.parameters()) + f1_corr_params = 0 + if base_model.f1_corr_in is not None and base_model.f1_corr_out is not None: + f1_corr_params = int(base_model.f1_corr_in.weight.numel() + base_model.f1_corr_out.weight.numel()) + est_corr_int6_bytes = 0 + if args.f1_corr_rank > 0: + # int8 payload stores int6 values + per-row fp16 scales. + est_corr_int6_bytes = ( + args.f1_corr_rank * (args.model_dim + args.vocab_size) + + 2 * (args.f1_corr_rank + args.vocab_size) + ) + log0(f"model_params:{n_params}") + log0( + f"f1_corr:rank={args.f1_corr_rank} params={f1_corr_params} " + f"est_int6_bytes~{est_corr_int6_bytes}" + ) + log0(f"mlp_act:{args.mlp_act} mlp_leaky_slope:{args.mlp_leaky_slope}") + log0(f"XSA:last_{args.xsa_last_n} world_size:{world_size} grad_accum_steps:{grad_accum_steps}") + log0(f"num_heads:{args.num_heads} num_kv_heads:{args.num_kv_heads} embed_lr:{token_lr} matrix_lr:{args.matrix_lr}") + log0( + f"train_batch_tokens:{args.train_batch_tokens} train_seq_len:{args.train_seq_len} " + f"iterations:{args.iterations} warmup_steps:{args.warmup_steps} " + f"max_wallclock_seconds:{args.max_wallclock_seconds:.3f}" + ) + log0(f"compile:enabled={int(args.compile_enabled)} fullgraph={int(args.compile_fullgraph)}") + log0(f"seed:{args.seed}") + if args.ngram_eval_order >= 2: + log0( + f"ngram_eval:order={args.ngram_eval_order} alpha={args.ngram_eval_alpha} " + f"min_count={args.ngram_eval_min_count} buckets={args.ngram_eval_buckets}" + ) + train_loader = DistributedTokenLoader(args.train_files, rank, world_size, device) + def zero_grad_all() -> None: + for opt in optimizers: + opt.zero_grad(set_to_none=True) + max_wallclock_ms = 1000.0 * args.max_wallclock_seconds if args.max_wallclock_seconds > 0 else None + def lr_mul(step: int, elapsed_ms: float) -> float: + if args.warmdown_iters <= 0: + return 1.0 + if max_wallclock_ms is None: + warmdown_start = max(args.iterations - args.warmdown_iters, 0) + return max((args.iterations - step) / max(args.warmdown_iters, 1), 0.0) if warmdown_start <= step < args.iterations else 1.0 + step_ms = elapsed_ms / max(step, 1) + warmdown_ms = args.warmdown_iters * step_ms + remaining_ms = max(max_wallclock_ms - elapsed_ms, 0.0) + return remaining_ms / max(warmdown_ms, 1e-9) if remaining_ms <= warmdown_ms else 1.0 + if args.warmup_steps > 0: + initial_model_state = {name: tensor.detach().cpu().clone() for name, tensor in base_model.state_dict().items()} + initial_optimizer_states = [copy.deepcopy(opt.state_dict()) for opt in optimizers] + model.train() + for warmup_step in range(args.warmup_steps): + zero_grad_all() + for micro_step in range(grad_accum_steps): + if distributed: + model.require_backward_grad_sync = micro_step == grad_accum_steps - 1 + x, y = train_loader.next_batch(args.train_batch_tokens, args.train_seq_len, grad_accum_steps) + with torch.autocast(device_type="cuda", dtype=torch.bfloat16, enabled=True): + warmup_loss = model(x, y) + (warmup_loss * grad_scale).backward() + for opt in optimizers: + opt.step() + zero_grad_all() + if args.warmup_steps <= 20 or (warmup_step + 1) % 10 == 0 or warmup_step + 1 == args.warmup_steps: + log0(f"warmup_step:{warmup_step + 1}/{args.warmup_steps}") + base_model.load_state_dict(initial_model_state, strict=True) + for opt, state in zip(optimizers, initial_optimizer_states, strict=True): + opt.load_state_dict(state) + zero_grad_all() + if distributed: + model.require_backward_grad_sync = True + train_loader = DistributedTokenLoader(args.train_files, rank, world_size, device) + swa_state: dict[str, Tensor] | None = None + swa_count = 0 + ema_state = {name: t.detach().float().clone() for name, t in base_model.state_dict().items()} + ema_decay = 0.997 + training_time_ms = 0.0 + stop_after_step: int | None = None + torch.cuda.synchronize() + t0 = time.perf_counter() + step = 0 + while True: + last_step = step == args.iterations or (stop_after_step is not None and step >= stop_after_step) + should_validate = last_step or (args.val_loss_every > 0 and step % args.val_loss_every == 0) + if should_validate: + torch.cuda.synchronize() + training_time_ms += 1000.0 * (time.perf_counter() - t0) + val_loss, val_bpb = eval_val( + args, + model, + rank, + world_size, + device, + grad_accum_steps, + val_tokens, + base_bytes_lut, + has_leading_space_lut, + is_boundary_token_lut, + ) + log0( + f"step:{step}/{args.iterations} val_loss:{val_loss:.4f} val_bpb:{val_bpb:.4f} " + f"train_time:{training_time_ms:.0f}ms step_avg:{training_time_ms / max(step, 1):.2f}ms" + ) + torch.cuda.synchronize() + t0 = time.perf_counter() + if last_step: + if stop_after_step is not None and step < args.iterations: + log0( + f"stopping_early: wallclock_cap train_time:{training_time_ms:.0f}ms " + f"step:{step}/{args.iterations}" + ) + break + elapsed_ms = training_time_ms + 1000.0 * (time.perf_counter() - t0) + scale = lr_mul(step, elapsed_ms) + if args.late_qat_threshold > 0 and scale < args.late_qat_threshold and not CastedLinear._qat_enabled: + CastedLinear._qat_enabled = True + log0(f"late_qat:enabled step:{step} scale:{scale:.4f}") + zero_grad_all() + train_loss = torch.zeros((), device=device) + for micro_step in range(grad_accum_steps): + if distributed: + model.require_backward_grad_sync = micro_step == grad_accum_steps - 1 + x, y = train_loader.next_batch(args.train_batch_tokens, args.train_seq_len, grad_accum_steps) + with torch.autocast(device_type="cuda", dtype=torch.bfloat16, enabled=True): + loss = model(x, y) + train_loss += loss.detach() + loss.backward() + train_loss /= grad_accum_steps + frac = min(step / args.muon_momentum_warmup_steps, 1.0) if args.muon_momentum_warmup_steps > 0 else 1.0 + muon_momentum = (1 - frac) * args.muon_momentum_warmup_start + frac * args.muon_momentum + for group in optimizer_muon.param_groups: + group["momentum"] = muon_momentum + for opt in optimizers: + for group in opt.param_groups: + group["lr"] = group["base_lr"] * scale + if args.grad_clip_norm > 0: + torch.nn.utils.clip_grad_norm_(base_model.parameters(), args.grad_clip_norm) + for opt in optimizers: + opt.step() + zero_grad_all() + # EMA update + with torch.no_grad(): + for name, t in base_model.state_dict().items(): + ema_state[name].mul_(ema_decay).add_(t.detach().float(), alpha=1.0 - ema_decay) + step += 1 + approx_training_time_ms = training_time_ms + 1000.0 * (time.perf_counter() - t0) + if args.swa_enabled and scale < 0.2 and step % args.swa_every == 0: + if swa_state is None: + swa_state = {name: t.detach().cpu().clone() for name, t in base_model.state_dict().items()} + swa_count = 1 + log0(f"swa:start step:{step}") + else: + for name, t in base_model.state_dict().items(): + swa_state[name] += t.detach().cpu() + swa_count += 1 + should_log_train = ( + args.train_log_every > 0 + and (step <= 10 or step % args.train_log_every == 0 or stop_after_step is not None) + ) + if should_log_train: + log0( + f"step:{step}/{args.iterations} train_loss:{train_loss.item():.4f} " + f"train_time:{approx_training_time_ms:.0f}ms step_avg:{approx_training_time_ms / step:.2f}ms" + ) + reached_cap = max_wallclock_ms is not None and approx_training_time_ms >= max_wallclock_ms + if distributed and max_wallclock_ms is not None: + reached_cap_tensor = torch.tensor(int(reached_cap), device=device) + dist.all_reduce(reached_cap_tensor, op=dist.ReduceOp.MAX) + reached_cap = bool(reached_cap_tensor.item()) + if stop_after_step is None and reached_cap: + stop_after_step = step + log0( + f"peak memory allocated: {torch.cuda.max_memory_allocated() // 1024 // 1024} MiB " + f"reserved: {torch.cuda.max_memory_reserved() // 1024 // 1024} MiB" + ) + if args.distill_enabled and args.distill_steps > 0: + log0( + f"distill:start steps:{args.distill_steps} lr_factor:{args.distill_lr_factor} " + f"temp:{args.distill_temperature} alpha:{args.distill_alpha} kl_clip:{args.distill_kl_clip}" + ) + current_state = base_model.state_dict() + teacher_state = {name: t.to(dtype=current_state[name].dtype) for name, t in ema_state.items()} + teacher_model = GPT( + vocab_size=args.vocab_size, num_layers=args.num_layers, model_dim=args.model_dim, + num_heads=args.num_heads, num_kv_heads=args.num_kv_heads, mlp_mult=args.mlp_mult, + tie_embeddings=args.tie_embeddings, tied_embed_init_std=args.tied_embed_init_std, + logit_softcap=args.logit_softcap, rope_base=args.rope_base, qk_gain_init=args.qk_gain_init, + mtp_num_heads=args.mtp_num_heads, mtp_loss_weight=args.mtp_loss_weight, + bigram_vocab_size=args.bigram_vocab_size, bigram_dim=args.bigram_dim, + xsa_last_n=args.xsa_last_n, rope_dims=args.rope_dims, ln_scale=args.ln_scale, dtg=args.dtg_enabled, + ve_enabled=args.ve_enabled, ve_dim=args.ve_dim, ve_layers=args.ve_layers, + mlp_act=args.mlp_act, mlp_leaky_slope=args.mlp_leaky_slope, + f1_corr_rank=args.f1_corr_rank, f1_corr_scale_init=args.f1_corr_scale_init, + ).to(device).bfloat16() + for m in teacher_model.modules(): + if isinstance(m, CastedLinear): + m.float() + restore_low_dim_params_to_fp32(teacher_model) + teacher_model.load_state_dict(teacher_state, strict=True) + teacher_model.eval() + for p in teacher_model.parameters(): + p.requires_grad_(False) + compiled_teacher_logits = maybe_torch_compile(teacher_model.forward_logits, args) + model.train() + T = args.distill_temperature + alpha = args.distill_alpha + for d_step in range(args.distill_steps): + zero_grad_all() + for opt in optimizers: + for group in opt.param_groups: + group["lr"] = group["base_lr"] * args.distill_lr_factor + x, y = train_loader.next_batch(args.train_batch_tokens, args.train_seq_len, grad_accum_steps) + with torch.autocast(device_type="cuda", dtype=torch.bfloat16, enabled=True): + student_logits = base_model.forward_logits(x) + with torch.no_grad(): + teacher_logits = compiled_teacher_logits(x) + student_log_probs = F.log_softmax(student_logits.float() / T, dim=-1) + teacher_probs = F.softmax(teacher_logits.float() / T, dim=-1) + token_kl = F.kl_div(student_log_probs, teacher_probs, reduction="none").sum(dim=-1) + kl_loss = token_kl.mean() * (T * T) + if args.distill_kl_clip > 0: + kl_loss = torch.clamp(kl_loss, max=args.distill_kl_clip) + ce_loss = F.cross_entropy( + student_logits.reshape(-1, student_logits.size(-1)).float(), + y.reshape(-1), + reduction="mean", + ) + loss = alpha * kl_loss + (1.0 - alpha) * ce_loss + (loss * grad_scale).backward() + if world_size > 1: + for p in base_model.parameters(): + if p.grad is not None: + dist.all_reduce(p.grad, op=dist.ReduceOp.AVG) + if args.grad_clip_norm > 0: + torch.nn.utils.clip_grad_norm_(base_model.parameters(), args.grad_clip_norm) + for opt in optimizers: + opt.step() + zero_grad_all() + with torch.no_grad(): + for name, t in base_model.state_dict().items(): + ema_state[name].mul_(ema_decay).add_(t.detach().float(), alpha=1.0 - ema_decay) + if (d_step + 1) % 8 == 0 or d_step == 0: + log0( + f"distill:step:{d_step + 1}/{args.distill_steps} " + f"kl:{kl_loss.item():.4f} ce:{ce_loss.item():.4f} total:{loss.item():.4f}" + ) + del teacher_model, compiled_teacher_logits + torch.cuda.empty_cache() + log0("distill:done") + # Apply EMA weights (better than SWA alone per PR#401) + log0("ema:applying EMA weights") + current_state = base_model.state_dict() + avg_state = {name: t.to(dtype=current_state[name].dtype) for name, t in ema_state.items()} + base_model.load_state_dict(avg_state, strict=True) + torch.cuda.synchronize() + t_diag = time.perf_counter() + diag_val_loss, diag_val_bpb = eval_val( + args, compiled_model, rank, world_size, device, grad_accum_steps, + val_tokens, base_bytes_lut, has_leading_space_lut, is_boundary_token_lut, + ) + torch.cuda.synchronize() + log0( + f"DIAGNOSTIC post_ema val_loss:{diag_val_loss:.4f} val_bpb:{diag_val_bpb:.4f} " + f"eval_time:{1000.0 * (time.perf_counter() - t_diag):.0f}ms" + ) + full_state_dict = base_model.state_dict() + export_sd = {k: v for k, v in full_state_dict.items() if "mtp_heads" not in k} + excluded_mtp = sum(int(t.numel()) for k, t in full_state_dict.items() if "mtp_heads" in k) + if excluded_mtp > 0: + log0(f"export_excluding_mtp_params:{excluded_mtp}") + if master_process: + torch.save(export_sd, "final_model.pt") + model_bytes = os.path.getsize("final_model.pt") + code_bytes = len(code.encode("utf-8")) + log0(f"Serialized model: {model_bytes} bytes") + log0(f"Code size: {code_bytes} bytes") + sd_cpu = {k: v.detach().cpu() for k, v in export_sd.items()} + # GPTQ calibration: collect Hessians from training data + log0("gptq:calibrating with training data...") + t_gptq = time.perf_counter() + gptq_hessians = gptq_calibrate(base_model, args.train_files, device, n_samples=256, seq_len=args.train_seq_len) + log0(f"gptq:calibrated {len(gptq_hessians)} layers in {time.perf_counter()-t_gptq:.1f}s") + quant_result, quant_meta = mixed_quantize_int6_gptq(sd_cpu, {"mlp", "attn", "aux"}, gptq_hessians) + quant_buf = io.BytesIO() + torch.save({"w": quant_result, "m": quant_meta}, quant_buf) + quant_raw = quant_buf.getvalue() + quant_blob = zstandard.ZstdCompressor(level=22).compress(quant_raw) if _COMPRESSOR == "zstd" else zlib.compress(quant_raw, 9) + if master_process: + with open("final_model.int6.ptz", "wb") as f: + f.write(quant_blob) + quant_file_bytes = len(quant_blob) + code_bytes = len(code.encode("utf-8")) + log0(f"Serialized model int6+{_COMPRESSOR}: {quant_file_bytes} bytes") + log0(f"Total submission size int6+{_COMPRESSOR}: {quant_file_bytes + code_bytes} bytes") + log0(f"Total submission size int8+zlib: {quant_file_bytes + code_bytes} bytes") + if distributed: + dist.barrier() + with open("final_model.int6.ptz", "rb") as f: + quant_blob_disk = f.read() + quant_state = torch.load( + io.BytesIO(zstandard.ZstdDecompressor().decompress(quant_blob_disk) if _COMPRESSOR == "zstd" else zlib.decompress(quant_blob_disk)), + map_location="cpu", + ) + deq_state = dequantize_mixed_int6(quant_state["w"], quant_state["m"], sd_cpu) + eval_model = GPT( + vocab_size=args.vocab_size, num_layers=args.num_layers, model_dim=args.model_dim, + num_heads=args.num_heads, num_kv_heads=args.num_kv_heads, mlp_mult=args.mlp_mult, + tie_embeddings=args.tie_embeddings, tied_embed_init_std=args.tied_embed_init_std, + logit_softcap=args.logit_softcap, rope_base=args.rope_base, qk_gain_init=args.qk_gain_init, + mtp_num_heads=0, mtp_loss_weight=0.0, + bigram_vocab_size=args.bigram_vocab_size, bigram_dim=args.bigram_dim, + xsa_last_n=args.xsa_last_n, # must match training model + rope_dims=args.rope_dims, ln_scale=args.ln_scale, dtg=args.dtg_enabled, + ve_enabled=args.ve_enabled, ve_dim=args.ve_dim, ve_layers=args.ve_layers, + mlp_act=args.mlp_act, mlp_leaky_slope=args.mlp_leaky_slope, + f1_corr_rank=args.f1_corr_rank, f1_corr_scale_init=args.f1_corr_scale_init, + ).to(device).bfloat16() + for m in eval_model.modules(): + if isinstance(m, CastedLinear): + m.float() + restore_low_dim_params_to_fp32(eval_model) + eval_model.load_state_dict(deq_state, strict=True) + compiled_eval = maybe_torch_compile(eval_model, args) + torch.cuda.synchronize() + t_qeval = time.perf_counter() + q_val_loss, q_val_bpb = eval_val( + args, compiled_eval, rank, world_size, device, grad_accum_steps, + val_tokens, base_bytes_lut, has_leading_space_lut, is_boundary_token_lut, + eval_seq_len=effective_eval_seq_len, + ) + torch.cuda.synchronize() + log0( + f"final_int6_roundtrip val_loss:{q_val_loss:.4f} val_bpb:{q_val_bpb:.4f} " + f"eval_time:{1000.0 * (time.perf_counter() - t_qeval):.0f}ms" + ) + log0(f"final_int6_roundtrip_exact val_loss:{q_val_loss:.8f} val_bpb:{q_val_bpb:.8f}") + sw_seq_len = effective_eval_seq_len + if args.eval_stride > 0 and args.eval_stride < sw_seq_len: + torch.cuda.synchronize() + t_slide = time.perf_counter() + sw_val_loss, sw_val_bpb = eval_val_sliding( + args, eval_model, rank, world_size, device, + val_tokens, base_bytes_lut, has_leading_space_lut, is_boundary_token_lut, + stride=args.eval_stride, + eval_seq_len=sw_seq_len, + ) + torch.cuda.synchronize() + log0( + f"final_int6_sliding_window val_loss:{sw_val_loss:.4f} val_bpb:{sw_val_bpb:.4f} " + f"stride:{args.eval_stride} eval_time:{1000.0 * (time.perf_counter() - t_slide):.0f}ms" + ) + log0(f"final_int6_sliding_window_exact val_loss:{sw_val_loss:.8f} val_bpb:{sw_val_bpb:.8f}") + log0(f"final_int8_zlib_roundtrip_exact val_loss:{sw_val_loss:.8f} val_bpb:{sw_val_bpb:.8f}") + if args.ngram_eval_order >= 2: + if distributed: + dist.barrier() + torch.cuda.synchronize() + t_ng = time.perf_counter() + ng_loss, ng_bpb, ng_coverage = eval_val_sliding_hashed_ngram( + args, + eval_model, + rank, + world_size, + device, + val_tokens, + base_bytes_lut, + has_leading_space_lut, + is_boundary_token_lut, + stride=args.eval_stride, + order=args.ngram_eval_order, + alpha=args.ngram_eval_alpha, + min_count=args.ngram_eval_min_count, + buckets=args.ngram_eval_buckets, + max_seconds=args.ngram_eval_max_seconds, + eval_seq_len=sw_seq_len, + ) + if rank == 0: + torch.cuda.synchronize() + ng_eval_ms = 1000.0 * (time.perf_counter() - t_ng) + if ng_coverage >= 0.999999: + log0( + f"final_int6_sliding_window_ngram{args.ngram_eval_order} val_loss:{ng_loss:.4f} " + f"val_bpb:{ng_bpb:.4f} eval_time:{ng_eval_ms:.0f}ms" + ) + log0( + f"final_int6_sliding_window_ngram{args.ngram_eval_order}_exact " + f"val_loss:{ng_loss:.8f} val_bpb:{ng_bpb:.8f}" + ) + else: + log0( + f"final_int6_sliding_window_ngram{args.ngram_eval_order}_partial val_loss:{ng_loss:.4f} " + f"val_bpb:{ng_bpb:.4f} coverage:{ng_coverage:.4f} eval_time:{ng_eval_ms:.0f}ms" + ) + log0( + f"final_int6_sliding_window_ngram{args.ngram_eval_order}_partial_exact " + f"val_loss:{ng_loss:.8f} val_bpb:{ng_bpb:.8f} coverage:{ng_coverage:.8f}" + ) + if distributed: + dist.barrier() + # Legal score-first TTT eval + if args.ttt_eval_enabled: + torch.cuda.synchronize() + t_ttt = time.perf_counter() + ttt_loss, ttt_bpb = eval_val_sliding_ttt( + args, eval_model, rank, world_size, device, + val_tokens, base_bytes_lut, has_leading_space_lut, is_boundary_token_lut, + stride=args.eval_stride, + ) + torch.cuda.synchronize() + log0(f"legal_ttt val_loss:{ttt_loss:.4f} val_bpb:{ttt_bpb:.4f} " + f"eval_time:{1000.0 * (time.perf_counter() - t_ttt):.0f}ms") + log0(f"legal_ttt_exact val_loss:{ttt_loss:.8f} val_bpb:{ttt_bpb:.8f}") + if distributed: + dist.destroy_process_group() +if __name__ == "__main__": + main() diff --git a/records/track_10min_16mb/2026-03-25_Podracing_5gram_LeakyReLU2_8xH100/train_seed1337.log b/records/track_10min_16mb/2026-03-25_Podracing_5gram_LeakyReLU2_8xH100/train_seed1337.log new file mode 100644 index 000000000..64c138c94 --- /dev/null +++ b/records/track_10min_16mb/2026-03-25_Podracing_5gram_LeakyReLU2_8xH100/train_seed1337.log @@ -0,0 +1,103 @@ + _____ _ + | | + |_| +W0325 02:56:21.776000 303290 torch/distributed/run.py:803] +W0325 02:56:21.776000 303290 torch/distributed/run.py:803] ***************************************** +W0325 02:56:21.776000 303290 torch/distributed/run.py:803] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. +W0325 02:56:21.776000 303290 torch/distributed/run.py:803] ***************************************** +logs/f1_car02_iso_var_t2_rope24_ngram5_s1337_20260325_025620.txt +val_bpb:enabled tokenizer_kind=sentencepiece tokenizer_path=./data/tokenizers/fineweb_1024_bpe.model +train_loader:dataset:fineweb10B_sp1024 train_shards:80 +val_loader:shards pattern=./data/datasets/fineweb10B_sp1024/fineweb_val_*.bin tokens:62021632 +model_params:26928220 +f1_corr:rank=0 params=0 est_int6_bytes~0 +mlp_act:leaky_relu_sq mlp_leaky_slope:0.5 +XSA:last_4 world_size:8 grad_accum_steps:1 +num_heads:8 num_kv_heads:4 embed_lr:0.035 matrix_lr:0.025 +train_batch_tokens:786432 train_seq_len:2048 iterations:20000 warmup_steps:20 max_wallclock_seconds:600.000 +compile:enabled=1 fullgraph=0 +seed:1337 +ngram_eval:order=5 alpha=0.2 min_count=2 buckets=4194304 +warmup_step:1/20 +warmup_step:2/20 +warmup_step:3/20 +warmup_step:4/20 +warmup_step:5/20 +warmup_step:6/20 +warmup_step:7/20 +warmup_step:8/20 +warmup_step:9/20 +warmup_step:10/20 +warmup_step:11/20 +warmup_step:12/20 +warmup_step:13/20 +warmup_step:14/20 +warmup_step:15/20 +warmup_step:16/20 +warmup_step:17/20 +warmup_step:18/20 +warmup_step:19/20 +warmup_step:20/20 +step:0/20000 val_loss:6.9317 val_bpb:4.1054 train_time:0ms step_avg:0.02ms +step:1/20000 train_loss:6.9343 train_time:143ms step_avg:143.26ms +step:2/20000 train_loss:8.7867 train_time:225ms step_avg:112.32ms +step:3/20000 train_loss:7.9649 train_time:310ms step_avg:103.33ms +step:4/20000 train_loss:7.2290 train_time:396ms step_avg:98.99ms +step:5/20000 train_loss:6.9872 train_time:482ms step_avg:96.38ms +step:6/20000 train_loss:6.9480 train_time:568ms step_avg:94.60ms +step:7/20000 train_loss:6.8041 train_time:654ms step_avg:93.37ms +step:8/20000 train_loss:6.6817 train_time:739ms step_avg:92.43ms +step:9/20000 train_loss:6.3606 train_time:825ms step_avg:91.66ms +step:10/20000 train_loss:6.0478 train_time:910ms step_avg:91.00ms +step:500/20000 train_loss:2.3734 train_time:43806ms step_avg:87.61ms +step:1000/20000 train_loss:2.2560 train_time:87852ms step_avg:87.85ms +step:1500/20000 train_loss:2.2085 train_time:131828ms step_avg:87.89ms +step:2000/20000 train_loss:2.0473 train_time:175779ms step_avg:87.89ms +step:2500/20000 train_loss:2.1515 train_time:219751ms step_avg:87.90ms +step:3000/20000 train_loss:2.1464 train_time:263722ms step_avg:87.91ms +step:3500/20000 train_loss:2.1659 train_time:307675ms step_avg:87.91ms +step:4000/20000 train_loss:1.9575 train_time:351636ms step_avg:87.91ms +step:4000/20000 val_loss:2.0468 val_bpb:1.2122 train_time:351640ms step_avg:87.91ms +step:4500/20000 train_loss:2.1031 train_time:395591ms step_avg:87.91ms +step:5000/20000 train_loss:2.0869 train_time:439535ms step_avg:87.91ms +late_qat:enabled step:5076 scale:0.5000 +step:5500/20000 train_loss:1.9999 train_time:483471ms step_avg:87.90ms +step:6000/20000 train_loss:1.9245 train_time:527392ms step_avg:87.90ms +swa:start step:6150 +step:6500/20000 train_loss:2.0633 train_time:571582ms step_avg:87.94ms +step:6822/20000 val_loss:1.9217 val_bpb:1.1382 train_time:600026ms step_avg:87.95ms +stopping_early: wallclock_cap train_time:600026ms step:6822/20000 +peak memory allocated: 20672 MiB reserved: 20718 MiB +ema:applying EMA weights +DIAGNOSTIC post_ema val_loss:1.9201 val_bpb:1.1372 eval_time:2042ms +Serialized model: 106047497 bytes +Code size: 102835 bytes +gptq:calibrating with training data... +gptq:calibrated 68 layers in 3.5s +gptq_quantize: 66 GPTQ layers, 0 naive layers +gptq_quantize: 66 GPTQ layers, 0 naive layers +gptq_quantize: 66 GPTQ layers, 0 naive layers +gptq_quantize: 66 GPTQ layers, 0 naive layers +gptq_quantize: 66 GPTQ layers, 0 naive layers +gptq_quantize: 66 GPTQ layers, 0 naive layers +gptq_quantize: 66 GPTQ layers, 0 naive layers +gptq_quantize: 66 GPTQ layers, 0 naive layers +Serialized model int6+zstd: 15528630 bytes +Total submission size int6+zstd: 15631465 bytes +Total submission size int8+zlib: 15631465 bytes +final_int6_roundtrip val_loss:1.9293 val_bpb:1.1426 eval_time:36893ms +final_int6_roundtrip_exact val_loss:1.92926834 val_bpb:1.14262138 +final_int6_sliding_window val_loss:1.8894 val_bpb:1.1190 stride:64 eval_time:97693ms +final_int6_sliding_window_exact val_loss:1.88940527 val_bpb:1.11901519 +final_int8_zlib_roundtrip_exact val_loss:1.88940527 val_bpb:1.11901519 +ngram_eval:progress windows=64032/121136 (52.9%) bpb=1.109128 t=47s +ngram_eval:progress windows=64032/121136 (52.9%) bpb=1.091968 t=48s +ngram_eval:progress windows=64032/121136 (52.9%) bpb=1.089283 t=48s +ngram_eval:progress windows=64032/121136 (52.9%) bpb=1.096762 t=48s +ngram_eval:progress windows=64032/121136 (52.9%) bpb=1.085093 t=48s +ngram_eval:progress windows=64032/121136 (52.9%) bpb=1.084128 t=48s +ngram_eval:progress windows=64032/121136 (52.9%) bpb=1.105740 t=48s +ngram_eval:progress windows=64032/121136 (52.9%) bpb=1.071534 t=48s +final_int6_sliding_window_ngram5 val_loss:1.7646 val_bpb:1.0451 eval_time:91757ms +final_int6_sliding_window_ngram5_exact val_loss:1.76457796 val_bpb:1.04508523 +Connection to 100.65.33.119 closed. diff --git a/records/track_10min_16mb/2026-03-25_Podracing_5gram_LeakyReLU2_8xH100/train_seed2045.log b/records/track_10min_16mb/2026-03-25_Podracing_5gram_LeakyReLU2_8xH100/train_seed2045.log new file mode 100644 index 000000000..9519799d8 --- /dev/null +++ b/records/track_10min_16mb/2026-03-25_Podracing_5gram_LeakyReLU2_8xH100/train_seed2045.log @@ -0,0 +1,103 @@ + _____ _ + | | + |_| +W0325 03:31:35.065000 435589 torch/distributed/run.py:803] +W0325 03:31:35.065000 435589 torch/distributed/run.py:803] ***************************************** +W0325 03:31:35.065000 435589 torch/distributed/run.py:803] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. +W0325 03:31:35.065000 435589 torch/distributed/run.py:803] ***************************************** +logs/f1_car02_iso_var_t2_rope24_ngram5_s2045_20260325_033133.txt +val_bpb:enabled tokenizer_kind=sentencepiece tokenizer_path=./data/tokenizers/fineweb_1024_bpe.model +train_loader:dataset:fineweb10B_sp1024 train_shards:80 +val_loader:shards pattern=./data/datasets/fineweb10B_sp1024/fineweb_val_*.bin tokens:62021632 +model_params:26928220 +f1_corr:rank=0 params=0 est_int6_bytes~0 +mlp_act:leaky_relu_sq mlp_leaky_slope:0.5 +XSA:last_4 world_size:8 grad_accum_steps:1 +num_heads:8 num_kv_heads:4 embed_lr:0.035 matrix_lr:0.025 +train_batch_tokens:786432 train_seq_len:2048 iterations:20000 warmup_steps:20 max_wallclock_seconds:600.000 +compile:enabled=1 fullgraph=0 +seed:2045 +ngram_eval:order=5 alpha=0.2 min_count=2 buckets=4194304 +warmup_step:1/20 +warmup_step:2/20 +warmup_step:3/20 +warmup_step:4/20 +warmup_step:5/20 +warmup_step:6/20 +warmup_step:7/20 +warmup_step:8/20 +warmup_step:9/20 +warmup_step:10/20 +warmup_step:11/20 +warmup_step:12/20 +warmup_step:13/20 +warmup_step:14/20 +warmup_step:15/20 +warmup_step:16/20 +warmup_step:17/20 +warmup_step:18/20 +warmup_step:19/20 +warmup_step:20/20 +step:0/20000 val_loss:6.9303 val_bpb:4.1045 train_time:0ms step_avg:0.01ms +step:1/20000 train_loss:6.9322 train_time:144ms step_avg:144.47ms +step:2/20000 train_loss:8.7644 train_time:225ms step_avg:112.59ms +step:3/20000 train_loss:7.8957 train_time:311ms step_avg:103.69ms +step:4/20000 train_loss:7.1978 train_time:396ms step_avg:99.12ms +step:5/20000 train_loss:6.9515 train_time:481ms step_avg:96.30ms +step:6/20000 train_loss:6.9438 train_time:567ms step_avg:94.49ms +step:7/20000 train_loss:6.8028 train_time:653ms step_avg:93.25ms +step:8/20000 train_loss:6.6956 train_time:739ms step_avg:92.37ms +step:9/20000 train_loss:6.3689 train_time:825ms step_avg:91.63ms +step:10/20000 train_loss:6.0672 train_time:910ms step_avg:91.01ms +step:500/20000 train_loss:2.3814 train_time:43815ms step_avg:87.63ms +step:1000/20000 train_loss:2.2559 train_time:87771ms step_avg:87.77ms +step:1500/20000 train_loss:2.2061 train_time:131708ms step_avg:87.81ms +step:2000/20000 train_loss:2.0501 train_time:175692ms step_avg:87.85ms +step:2500/20000 train_loss:2.1548 train_time:219671ms step_avg:87.87ms +step:3000/20000 train_loss:2.1471 train_time:263638ms step_avg:87.88ms +step:3500/20000 train_loss:2.1656 train_time:307599ms step_avg:87.89ms +step:4000/20000 train_loss:1.9575 train_time:351557ms step_avg:87.89ms +step:4000/20000 val_loss:2.0474 val_bpb:1.2126 train_time:351562ms step_avg:87.89ms +step:4500/20000 train_loss:2.1057 train_time:395598ms step_avg:87.91ms +step:5000/20000 train_loss:2.0869 train_time:439521ms step_avg:87.90ms +late_qat:enabled step:5077 scale:0.4998 +step:5500/20000 train_loss:2.0005 train_time:483438ms step_avg:87.90ms +step:6000/20000 train_loss:1.9250 train_time:527343ms step_avg:87.89ms +swa:start step:6150 +step:6500/20000 train_loss:2.0644 train_time:571490ms step_avg:87.92ms +step:6823/20000 val_loss:1.9229 val_bpb:1.1389 train_time:600022ms step_avg:87.94ms +stopping_early: wallclock_cap train_time:600022ms step:6823/20000 +peak memory allocated: 20672 MiB reserved: 20718 MiB +ema:applying EMA weights +DIAGNOSTIC post_ema val_loss:1.9213 val_bpb:1.1379 eval_time:2043ms +Serialized model: 106047497 bytes +Code size: 102835 bytes +gptq:calibrating with training data... +gptq:calibrated 68 layers in 3.6s +gptq_quantize: 66 GPTQ layers, 0 naive layers +gptq_quantize: 66 GPTQ layers, 0 naive layers +gptq_quantize: 66 GPTQ layers, 0 naive layers +gptq_quantize: 66 GPTQ layers, 0 naive layers +gptq_quantize: 66 GPTQ layers, 0 naive layers +gptq_quantize: 66 GPTQ layers, 0 naive layers +gptq_quantize: 66 GPTQ layers, 0 naive layers +gptq_quantize: 66 GPTQ layers, 0 naive layers +Serialized model int6+zstd: 15537990 bytes +Total submission size int6+zstd: 15640825 bytes +Total submission size int8+zlib: 15640825 bytes +final_int6_roundtrip val_loss:1.9309 val_bpb:1.1436 eval_time:37187ms +final_int6_roundtrip_exact val_loss:1.93088788 val_bpb:1.14358056 +final_int6_sliding_window val_loss:1.8910 val_bpb:1.1200 stride:64 eval_time:98043ms +final_int6_sliding_window_exact val_loss:1.89104663 val_bpb:1.11998730 +final_int8_zlib_roundtrip_exact val_loss:1.89104663 val_bpb:1.11998730 +ngram_eval:progress windows=64032/121136 (52.9%) bpb=1.110070 t=47s +ngram_eval:progress windows=64032/121136 (52.9%) bpb=1.084993 t=48s +ngram_eval:progress windows=64032/121136 (52.9%) bpb=1.072233 t=48s +ngram_eval:progress windows=64032/121136 (52.9%) bpb=1.086047 t=48s +ngram_eval:progress windows=64032/121136 (52.9%) bpb=1.092938 t=48s +ngram_eval:progress windows=64032/121136 (52.9%) bpb=1.106821 t=48s +ngram_eval:progress windows=64032/121136 (52.9%) bpb=1.097940 t=48s +ngram_eval:progress windows=64032/121136 (52.9%) bpb=1.089815 t=49s +final_int6_sliding_window_ngram5 val_loss:1.7661 val_bpb:1.0460 eval_time:92842ms +final_int6_sliding_window_ngram5_exact val_loss:1.76610289 val_bpb:1.04598838 +Connection to 100.65.33.119 closed. diff --git a/records/track_10min_16mb/2026-03-25_Podracing_5gram_LeakyReLU2_8xH100/train_seed42.log b/records/track_10min_16mb/2026-03-25_Podracing_5gram_LeakyReLU2_8xH100/train_seed42.log new file mode 100644 index 000000000..f3db26dca --- /dev/null +++ b/records/track_10min_16mb/2026-03-25_Podracing_5gram_LeakyReLU2_8xH100/train_seed42.log @@ -0,0 +1,103 @@ + _____ _ + | | + |_| +W0325 03:13:59.013000 369417 torch/distributed/run.py:803] +W0325 03:13:59.013000 369417 torch/distributed/run.py:803] ***************************************** +W0325 03:13:59.013000 369417 torch/distributed/run.py:803] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. +W0325 03:13:59.013000 369417 torch/distributed/run.py:803] ***************************************** +logs/f1_car02_iso_var_t2_rope24_ngram5_s42_20260325_031357.txt +val_bpb:enabled tokenizer_kind=sentencepiece tokenizer_path=./data/tokenizers/fineweb_1024_bpe.model +train_loader:dataset:fineweb10B_sp1024 train_shards:80 +val_loader:shards pattern=./data/datasets/fineweb10B_sp1024/fineweb_val_*.bin tokens:62021632 +model_params:26928220 +f1_corr:rank=0 params=0 est_int6_bytes~0 +mlp_act:leaky_relu_sq mlp_leaky_slope:0.5 +XSA:last_4 world_size:8 grad_accum_steps:1 +num_heads:8 num_kv_heads:4 embed_lr:0.035 matrix_lr:0.025 +train_batch_tokens:786432 train_seq_len:2048 iterations:20000 warmup_steps:20 max_wallclock_seconds:600.000 +compile:enabled=1 fullgraph=0 +seed:42 +ngram_eval:order=5 alpha=0.2 min_count=2 buckets=4194304 +warmup_step:1/20 +warmup_step:2/20 +warmup_step:3/20 +warmup_step:4/20 +warmup_step:5/20 +warmup_step:6/20 +warmup_step:7/20 +warmup_step:8/20 +warmup_step:9/20 +warmup_step:10/20 +warmup_step:11/20 +warmup_step:12/20 +warmup_step:13/20 +warmup_step:14/20 +warmup_step:15/20 +warmup_step:16/20 +warmup_step:17/20 +warmup_step:18/20 +warmup_step:19/20 +warmup_step:20/20 +step:0/20000 val_loss:6.9293 val_bpb:4.1039 train_time:0ms step_avg:0.01ms +step:1/20000 train_loss:6.9307 train_time:145ms step_avg:144.74ms +step:2/20000 train_loss:8.6833 train_time:227ms step_avg:113.25ms +step:3/20000 train_loss:7.9060 train_time:312ms step_avg:104.16ms +step:4/20000 train_loss:7.2570 train_time:398ms step_avg:99.58ms +step:5/20000 train_loss:7.0185 train_time:484ms step_avg:96.80ms +step:6/20000 train_loss:6.8703 train_time:570ms step_avg:94.98ms +step:7/20000 train_loss:6.7342 train_time:655ms step_avg:93.62ms +step:8/20000 train_loss:6.6461 train_time:741ms step_avg:92.61ms +step:9/20000 train_loss:6.3717 train_time:826ms step_avg:91.82ms +step:10/20000 train_loss:6.0673 train_time:913ms step_avg:91.26ms +step:500/20000 train_loss:2.3771 train_time:43835ms step_avg:87.67ms +step:1000/20000 train_loss:2.2588 train_time:87833ms step_avg:87.83ms +step:1500/20000 train_loss:2.2073 train_time:131823ms step_avg:87.88ms +step:2000/20000 train_loss:2.0545 train_time:175840ms step_avg:87.92ms +step:2500/20000 train_loss:2.1549 train_time:219846ms step_avg:87.94ms +step:3000/20000 train_loss:2.1506 train_time:263922ms step_avg:87.97ms +step:3500/20000 train_loss:2.1715 train_time:307898ms step_avg:87.97ms +step:4000/20000 train_loss:1.9611 train_time:351877ms step_avg:87.97ms +step:4000/20000 val_loss:2.0500 val_bpb:1.2142 train_time:351882ms step_avg:87.97ms +step:4500/20000 train_loss:2.1072 train_time:395852ms step_avg:87.97ms +step:5000/20000 train_loss:2.0883 train_time:439818ms step_avg:87.96ms +late_qat:enabled step:5072 scale:0.4998 +step:5500/20000 train_loss:2.0034 train_time:483779ms step_avg:87.96ms +step:6000/20000 train_loss:1.9276 train_time:527740ms step_avg:87.96ms +swa:start step:6150 +step:6500/20000 train_loss:2.0657 train_time:571958ms step_avg:87.99ms +step:6817/20000 val_loss:1.9253 val_bpb:1.1403 train_time:600008ms step_avg:88.02ms +stopping_early: wallclock_cap train_time:600008ms step:6817/20000 +peak memory allocated: 20672 MiB reserved: 20718 MiB +ema:applying EMA weights +DIAGNOSTIC post_ema val_loss:1.9237 val_bpb:1.1393 eval_time:2041ms +Serialized model: 106047497 bytes +Code size: 102835 bytes +gptq:calibrating with training data... +gptq:calibrated 68 layers in 3.6s +gptq_quantize: 66 GPTQ layers, 0 naive layers +gptq_quantize: 66 GPTQ layers, 0 naive layers +gptq_quantize: 66 GPTQ layers, 0 naive layers +gptq_quantize: 66 GPTQ layers, 0 naive layers +gptq_quantize: 66 GPTQ layers, 0 naive layers +gptq_quantize: 66 GPTQ layers, 0 naive layers +gptq_quantize: 66 GPTQ layers, 0 naive layers +gptq_quantize: 66 GPTQ layers, 0 naive layers +Serialized model int6+zstd: 15487548 bytes +Total submission size int6+zstd: 15590383 bytes +Total submission size int8+zlib: 15590383 bytes +final_int6_roundtrip val_loss:1.9335 val_bpb:1.1451 eval_time:39681ms +final_int6_roundtrip_exact val_loss:1.93345159 val_bpb:1.14509894 +final_int6_sliding_window val_loss:1.8939 val_bpb:1.1217 stride:64 eval_time:100629ms +final_int6_sliding_window_exact val_loss:1.89385941 val_bpb:1.12165319 +final_int8_zlib_roundtrip_exact val_loss:1.89385941 val_bpb:1.12165319 +ngram_eval:progress windows=64032/121136 (52.9%) bpb=1.091392 t=48s +ngram_eval:progress windows=64032/121136 (52.9%) bpb=1.099295 t=48s +ngram_eval:progress windows=64032/121136 (52.9%) bpb=1.086410 t=48s +ngram_eval:progress windows=64032/121136 (52.9%) bpb=1.108322 t=48s +ngram_eval:progress windows=64032/121136 (52.9%) bpb=1.094407 t=48s +ngram_eval:progress windows=64032/121136 (52.9%) bpb=1.087306 t=48s +ngram_eval:progress windows=64032/121136 (52.9%) bpb=1.111408 t=48s +ngram_eval:progress windows=64032/121136 (52.9%) bpb=1.073701 t=49s +final_int6_sliding_window_ngram5 val_loss:1.7680 val_bpb:1.0471 eval_time:92720ms +final_int6_sliding_window_ngram5_exact val_loss:1.76796876 val_bpb:1.04709346 +Connection to 100.65.33.119 closed. diff --git a/train_gpt.py b/train_gpt.py index 651beb2b8..6566ad098 100644 --- a/train_gpt.py +++ b/train_gpt.py @@ -1,11 +1,4 @@ -""" -The `train_gpt.py` and `train_gpt_mlx.py` scripts are intended as good launching-off points for new participants, not SOTA configs. We'll accept PRs that tune, improve, or simplify these scripts without significantly increasing complexity, but competitive submissions should stay in the `/records` folder. - -Hard stop: To keep readable for newcomers, let's make sure `train_gpt.py` and `train_gpt_mlx.py` never are longer than 1500 lines. -""" - from __future__ import annotations - import copy import glob import io @@ -18,7 +11,11 @@ import uuid import zlib from pathlib import Path - +try: + import zstandard + _COMPRESSOR = "zstd" +except ImportError: + _COMPRESSOR = "zlib" import numpy as np import sentencepiece as spm import torch @@ -26,76 +23,106 @@ import torch.nn.functional as F from torch import Tensor, nn from torch.nn.parallel import DistributedDataParallel as DDP - -# ----------------------------- -# HYPERPARAMETERS -# ----------------------------- -# Default Simple Baseline run: -# - 9 transformer blocks at width 512 -# - 8 attention heads with 4 KV heads (GQA) and 2x MLP expansion -# - vocab size 1024, sequence length 1024, tied embeddings -# - 524,288 train tokens per step for 20,000 iterations with a ~10 minute cap - +from flash_attn_interface import flash_attn_func as flash_attn_3_func class Hyperparameters: - # Data paths are shard globs produced by the existing preprocessing pipeline. data_path = os.environ.get("DATA_PATH", "./data/datasets/fineweb10B_sp1024") train_files = os.path.join(data_path, "fineweb_train_*.bin") val_files = os.path.join(data_path, "fineweb_val_*.bin") tokenizer_path = os.environ.get("TOKENIZER_PATH", "./data/tokenizers/fineweb_1024_bpe.model") run_id = os.environ.get("RUN_ID", str(uuid.uuid4())) seed = int(os.environ.get("SEED", 1337)) - - # Validation cadence and batch size. Validation always uses the full fineweb_val split. val_batch_size = int(os.environ.get("VAL_BATCH_SIZE", 524_288)) - val_loss_every = int(os.environ.get("VAL_LOSS_EVERY", 1000)) - train_log_every = int(os.environ.get("TRAIN_LOG_EVERY", 200)) - - # Training length. + val_loss_every = int(os.environ.get("VAL_LOSS_EVERY", 4000)) + train_log_every = int(os.environ.get("TRAIN_LOG_EVERY", 500)) iterations = int(os.environ.get("ITERATIONS", 20000)) - warmdown_iters = int(os.environ.get("WARMDOWN_ITERS", 1200)) + warmdown_iters = int(os.environ.get("WARMDOWN_ITERS", 3500)) warmup_steps = int(os.environ.get("WARMUP_STEPS", 20)) - train_batch_tokens = int(os.environ.get("TRAIN_BATCH_TOKENS", 524_288)) - train_seq_len = int(os.environ.get("TRAIN_SEQ_LEN", 1024)) + train_batch_tokens = int(os.environ.get("TRAIN_BATCH_TOKENS", 786_432)) + train_seq_len = int(os.environ.get("TRAIN_SEQ_LEN", 2048)) + eval_seq_len = int(os.environ.get("EVAL_SEQ_LEN", 2048)) max_wallclock_seconds = float(os.environ.get("MAX_WALLCLOCK_SECONDS", 600.0)) qk_gain_init = float(os.environ.get("QK_GAIN_INIT", 1.5)) - - # Model shape. vocab_size = int(os.environ.get("VOCAB_SIZE", 1024)) - num_layers = int(os.environ.get("NUM_LAYERS", 9)) + num_layers = int(os.environ.get("NUM_LAYERS", 11)) num_kv_heads = int(os.environ.get("NUM_KV_HEADS", 4)) model_dim = int(os.environ.get("MODEL_DIM", 512)) num_heads = int(os.environ.get("NUM_HEADS", 8)) - mlp_mult = int(os.environ.get("MLP_MULT", 2)) + mlp_mult = float(os.environ.get("MLP_MULT", 3.0)) + mlp_act = os.environ.get("MLP_ACT", "relu_sq").lower() + mlp_leaky_slope = float(os.environ.get("MLP_LEAKY_SLOPE", 0.5)) tie_embeddings = bool(int(os.environ.get("TIE_EMBEDDINGS", "1"))) rope_base = float(os.environ.get("ROPE_BASE", 10000.0)) logit_softcap = float(os.environ.get("LOGIT_SOFTCAP", 30.0)) - - # Optimizer hyperparameters. embed_lr = float(os.environ.get("EMBED_LR", 0.6)) head_lr = float(os.environ.get("HEAD_LR", 0.008)) - tied_embed_lr = float(os.environ.get("TIED_EMBED_LR", 0.05)) + tied_embed_lr = float(os.environ.get("TIED_EMBED_LR", 0.035)) tied_embed_init_std = float(os.environ.get("TIED_EMBED_INIT_STD", 0.005)) - matrix_lr = float(os.environ.get("MATRIX_LR", 0.04)) - scalar_lr = float(os.environ.get("SCALAR_LR", 0.04)) - muon_momentum = float(os.environ.get("MUON_MOMENTUM", 0.95)) + matrix_lr = float(os.environ.get("MATRIX_LR", 0.025)) + scalar_lr = float(os.environ.get("SCALAR_LR", 0.025)) + muon_momentum = float(os.environ.get("MUON_MOMENTUM", 0.99)) muon_backend_steps = int(os.environ.get("MUON_BACKEND_STEPS", 5)) - muon_momentum_warmup_start = float(os.environ.get("MUON_MOMENTUM_WARMUP_START", 0.85)) - muon_momentum_warmup_steps = int(os.environ.get("MUON_MOMENTUM_WARMUP_STEPS", 500)) + muon_momentum_warmup_start = float(os.environ.get("MUON_MOMENTUM_WARMUP_START", 0.92)) + muon_momentum_warmup_steps = int(os.environ.get("MUON_MOMENTUM_WARMUP_STEPS", 1500)) beta1 = float(os.environ.get("BETA1", 0.9)) beta2 = float(os.environ.get("BETA2", 0.95)) adam_eps = float(os.environ.get("ADAM_EPS", 1e-8)) - grad_clip_norm = float(os.environ.get("GRAD_CLIP_NORM", 0.0)) - -# ----------------------------- -# MUON OPTIMIZER -# ----------------------------- -# -# As borrowed from modded-nanogpt -# Background on Muon: https://kellerjordan.github.io/posts/muon/ - + grad_clip_norm = float(os.environ.get("GRAD_CLIP_NORM", 0.3)) + eval_stride = int(os.environ.get("EVAL_STRIDE", 64)) + mtp_num_heads = int(os.environ.get("MTP_NUM_HEADS", 0)) + mtp_loss_weight = float(os.environ.get("MTP_LOSS_WEIGHT", 0.2)) + muon_beta2 = float(os.environ.get("MUON_BETA2", 0.95)) + swa_enabled = bool(int(os.environ.get("SWA_ENABLED", "1"))) + swa_every = int(os.environ.get("SWA_EVERY", 50)) # tighter: collect more recent checkpoints + muon_wd = float(os.environ.get("MUON_WD", 0.04)) + adam_wd = float(os.environ.get("ADAM_WD", 0.04)) + qat_enabled = bool(int(os.environ.get("QAT_ENABLED", "0"))) + bigram_vocab_size = int(os.environ.get("BIGRAM_VOCAB_SIZE", 2048)) + bigram_dim = int(os.environ.get("BIGRAM_DIM", 128)) + xsa_last_n = int(os.environ.get("XSA_LAST_N", 11)) # XSA on ALL 11 layers + rope_dims = int(os.environ.get("ROPE_DIMS", 16)) + ln_scale = bool(int(os.environ.get("LN_SCALE", "1"))) + dtg_enabled = bool(int(os.environ.get("DTG_ENABLED", "0"))) + late_qat_threshold = float(os.environ.get("LATE_QAT_THRESHOLD", 0.5)) + ve_enabled = bool(int(os.environ.get("VE_ENABLED", "1"))) + ve_dim = int(os.environ.get("VE_DIM", 128)) + ve_layers = os.environ.get("VE_LAYERS", "9,10") + # F1 capacity add-on: low-rank correction head (active at inference). + # Approx extra params ~= rank * (model_dim + vocab_size). + f1_corr_rank = int(os.environ.get("F1_CORR_RANK", 0)) + f1_corr_scale_init = float(os.environ.get("F1_CORR_SCALE_INIT", 0.10)) + # Post-train self-distillation: EMA teacher -> student. + distill_enabled = bool(int(os.environ.get("DISTILL_ENABLED", "0"))) + distill_steps = int(os.environ.get("DISTILL_STEPS", 24)) + distill_lr_factor = float(os.environ.get("DISTILL_LR_FACTOR", 0.02)) + distill_temperature = float(os.environ.get("DISTILL_TEMPERATURE", 1.5)) + distill_alpha = float(os.environ.get("DISTILL_ALPHA", 0.60)) + distill_kl_clip = float(os.environ.get("DISTILL_KL_CLIP", 10.0)) + # Legal score-first TTT eval (PR #461 recipe) + ttt_eval_enabled = bool(int(os.environ.get("TTT_EVAL_ENABLED", "1"))) + ttt_lr = float(os.environ.get("TTT_LR", 0.002)) + ttt_epochs = int(os.environ.get("TTT_EPOCHS", 3)) + ttt_chunk_tokens = int(os.environ.get("TTT_CHUNK_TOKENS", 32768)) + ttt_freeze_blocks = int(os.environ.get("TTT_FREEZE_BLOCKS", 2)) + ttt_momentum = float(os.environ.get("TTT_MOMENTUM", 0.9)) + ttt_batch_seqs = int(os.environ.get("TTT_BATCH_SEQS", 32)) + ttt_grad_clip = float(os.environ.get("TTT_GRAD_CLIP", 1.0)) + ttt_max_train_chunks = int(os.environ.get("TTT_MAX_TRAIN_CHUNKS", 200)) # stop training after N chunks, keep scoring + ttt_ema_decay = float(os.environ.get("TTT_EMA_DECAY", 0.995)) # EMA decay for TTT weight smoothing (0 = disabled) + ttt_freeze_embed = bool(int(os.environ.get("TTT_FREEZE_EMBED", "1"))) # freeze tok_emb/bigram/ve during TTT + # Optional legal score-first hashed n-gram interpolation at eval time. + # This path never uses label-aware gating: mix weight is fixed. + ngram_eval_order = int(os.environ.get("NGRAM_EVAL_ORDER", 0)) # 0=off, >=2 enables hashed n-gram + ngram_eval_alpha = float(os.environ.get("NGRAM_EVAL_ALPHA", 0.20)) + ngram_eval_min_count = int(os.environ.get("NGRAM_EVAL_MIN_COUNT", 2)) + ngram_eval_buckets = int(os.environ.get("NGRAM_EVAL_BUCKETS", 4_194_304)) + ngram_eval_max_seconds = float(os.environ.get("NGRAM_EVAL_MAX_SECONDS", 0.0)) + compile_enabled = bool(int(os.environ.get("COMPILE_ENABLED", "1"))) + compile_fullgraph = bool(int(os.environ.get("COMPILE_FULLGRAPH", "1"))) +def maybe_torch_compile(obj, args: Hyperparameters): + if not args.compile_enabled: + return obj + return torch.compile(obj, dynamic=False, fullgraph=args.compile_fullgraph) def zeropower_via_newtonschulz5(G: Tensor, steps: int = 10, eps: float = 1e-7) -> Tensor: - # Orthogonalize a 2D update matrix with a fast Newton-Schulz iteration. - # Muon uses this to normalize matrix-shaped gradients before applying them. a, b, c = (3.4445, -4.7750, 2.0315) X = G.bfloat16() X /= X.norm() + eps @@ -107,26 +134,23 @@ def zeropower_via_newtonschulz5(G: Tensor, steps: int = 10, eps: float = 1e-7) - B = b * A + c * A @ A X = a * X + B @ X return X.T if transposed else X - - class Muon(torch.optim.Optimizer): - def __init__(self, params, lr: float, momentum: float, backend_steps: int, nesterov: bool = True): + def __init__(self, params, lr: float, momentum: float, backend_steps: int, + nesterov: bool = True, weight_decay: float = 0.0): super().__init__( params, - dict(lr=lr, momentum=momentum, backend_steps=backend_steps, nesterov=nesterov), + dict(lr=lr, momentum=momentum, backend_steps=backend_steps, + nesterov=nesterov, weight_decay=weight_decay), ) - @torch.no_grad() def step(self, closure=None): loss = None if closure is not None: with torch.enable_grad(): loss = closure() - distributed = dist.is_available() and dist.is_initialized() world_size = dist.get_world_size() if distributed else 1 rank = dist.get_rank() if distributed else 0 - for group in self.param_groups: params = group["params"] if not params: @@ -135,10 +159,8 @@ def step(self, closure=None): momentum = group["momentum"] backend_steps = group["backend_steps"] nesterov = group["nesterov"] - total_params = sum(int(p.numel()) for p in params) updates_flat = torch.zeros(total_params, device=params[0].device, dtype=torch.bfloat16) - curr = 0 for i, p in enumerate(params): if i % world_size == rank and p.grad is not None: @@ -151,32 +173,20 @@ def step(self, closure=None): if nesterov: g = g.add(buf, alpha=momentum) g = zeropower_via_newtonschulz5(g, steps=backend_steps) - # Scale correction from Muon reference implementations. g *= max(1, g.size(0) / g.size(1)) ** 0.5 updates_flat[curr : curr + p.numel()] = g.reshape(-1) curr += p.numel() - if distributed: dist.all_reduce(updates_flat, op=dist.ReduceOp.SUM) - + wd = group.get("weight_decay", 0.0) curr = 0 for p in params: + if wd > 0.0: + p.data.mul_(1.0 - lr * wd) g = updates_flat[curr : curr + p.numel()].view_as(p).to(dtype=p.dtype) p.add_(g, alpha=-lr) curr += p.numel() - return loss - - -# ----------------------------- -# TOKENIZER-AGNOSTIC EVALUATION SETUP -# ----------------------------- -# -# It's common for small models have a large fraction of their parameters be embeddings, since the 2 * d_model * d_vocab vectors can be gigantic. -# Instead of locking the tokenizer, we let you bring your own and calculate our validation metrics on the average compression of the validation set. -# We calculate BPB (bits-per-byte) instead of validation loss, so we need methods to count the number of bits per token in the tokenizer. -# Note: Submissions that edit the tokenizer will be examined more carefully, since screwing this up might unjustly improve your score. - def build_sentencepiece_luts( sp: spm.SentencePieceProcessor, vocab_size: int, device: torch.device ) -> tuple[Tensor, Tensor, Tensor]: @@ -202,20 +212,15 @@ def build_sentencepiece_luts( torch.tensor(has_leading_space_np, dtype=torch.bool, device=device), torch.tensor(is_boundary_token_np, dtype=torch.bool, device=device), ) - - def load_validation_tokens(pattern: str, seq_len: int) -> Tensor: files = [Path(p) for p in sorted(glob.glob(pattern))] if not files: raise FileNotFoundError(f"No files found for pattern: {pattern}") - # The export pipeline writes the fixed first-50k-doc validation set to fineweb_val_*. tokens = torch.cat([load_data_shard(file) for file in files]).contiguous() usable = ((tokens.numel() - 1) // seq_len) * seq_len if usable <= 0: raise ValueError(f"Validation split is too short for TRAIN_SEQ_LEN={seq_len}") return tokens[: usable + 1] - - def eval_val( args: Hyperparameters, model: nn.Module, @@ -227,34 +232,32 @@ def eval_val( base_bytes_lut: Tensor, has_leading_space_lut: Tensor, is_boundary_token_lut: Tensor, + eval_seq_len: int | None = None, ) -> tuple[float, float]: - # Validation computes two metrics: - # - val_loss: token cross-entropy (natural log) - # - val_bpb: tokenizer-agnostic compression metric used by the challenge + seq_len = eval_seq_len or args.train_seq_len local_batch_tokens = args.val_batch_size // (world_size * grad_accum_steps) - if local_batch_tokens < args.train_seq_len: + if local_batch_tokens < seq_len: raise ValueError( "VAL_BATCH_SIZE must provide at least one sequence per rank; " f"got VAL_BATCH_SIZE={args.val_batch_size}, WORLD_SIZE={world_size}, " - f"GRAD_ACCUM_STEPS={grad_accum_steps}, TRAIN_SEQ_LEN={args.train_seq_len}" + f"GRAD_ACCUM_STEPS={grad_accum_steps}, seq_len={seq_len}" ) - local_batch_seqs = local_batch_tokens // args.train_seq_len - total_seqs = (val_tokens.numel() - 1) // args.train_seq_len + local_batch_seqs = local_batch_tokens // seq_len + total_seqs = (val_tokens.numel() - 1) // seq_len seq_start = (total_seqs * rank) // world_size seq_end = (total_seqs * (rank + 1)) // world_size val_loss_sum = torch.zeros((), device=device, dtype=torch.float64) val_token_count = torch.zeros((), device=device, dtype=torch.float64) val_byte_count = torch.zeros((), device=device, dtype=torch.float64) - model.eval() with torch.inference_mode(): for batch_seq_start in range(seq_start, seq_end, local_batch_seqs): batch_seq_end = min(batch_seq_start + local_batch_seqs, seq_end) - raw_start = batch_seq_start * args.train_seq_len - raw_end = batch_seq_end * args.train_seq_len + 1 + raw_start = batch_seq_start * seq_len + raw_end = batch_seq_end * seq_len + 1 local = val_tokens[raw_start:raw_end].to(device=device, dtype=torch.int64, non_blocking=True) - x = local[:-1].reshape(-1, args.train_seq_len) - y = local[1:].reshape(-1, args.train_seq_len) + x = local[:-1].reshape(-1, seq_len) + y = local[1:].reshape(-1, seq_len) with torch.autocast(device_type="cuda", dtype=torch.bfloat16, enabled=True): batch_loss = model(x, y).detach() batch_token_count = float(y.numel()) @@ -265,31 +268,20 @@ def eval_val( token_bytes = base_bytes_lut[tgt_ids].to(dtype=torch.int16) token_bytes += (has_leading_space_lut[tgt_ids] & ~is_boundary_token_lut[prev_ids]).to(dtype=torch.int16) val_byte_count += token_bytes.to(torch.float64).sum() - if dist.is_available() and dist.is_initialized(): dist.all_reduce(val_loss_sum, op=dist.ReduceOp.SUM) dist.all_reduce(val_token_count, op=dist.ReduceOp.SUM) dist.all_reduce(val_byte_count, op=dist.ReduceOp.SUM) - val_loss = val_loss_sum / val_token_count bits_per_token = val_loss.item() / math.log(2.0) tokens_per_byte = val_token_count.item() / val_byte_count.item() model.train() return float(val_loss.item()), float(bits_per_token * tokens_per_byte) - -# ----------------------------- -# POST-TRAINING QUANTIZATION -# ----------------------------- -# -# It's silly to export our model, which is trained in bf16 and fp32, at that same precision. -# Instead, we get approximately the same model (with a small hit) by quantizing the model to int8 & zlib compressing. -# We can then decompress the model and run in higher precision for evaluation, after closing in under the size limit. - CONTROL_TENSOR_NAME_PATTERNS = tuple( pattern for pattern in os.environ.get( "CONTROL_TENSOR_NAME_PATTERNS", - "attn_scale,attn_scales,mlp_scale,mlp_scales,resid_mix,resid_mixes,q_gain,skip_weight,skip_weights", + "attn_scale,attn_scales,mlp_scale,mlp_scales,resid_mix,resid_mixes,q_gain,skip_weight,skip_weights,smear,dtg_gate,ve_layer_scales,ve_shared.scale", ).split(",") if pattern ) @@ -306,10 +298,8 @@ def eval_val( INT8_PER_ROW_SCALE_DTYPE = torch.float16 INT8_CLIP_PERCENTILE = 99.99984 INT8_CLIP_Q = INT8_CLIP_PERCENTILE / 100.0 - def tensor_nbytes(t: Tensor) -> int: return int(t.numel()) * int(t.element_size()) - def keep_float_tensor(name: str, t: Tensor, passthrough_orig_dtypes: dict[str, str]) -> Tensor: if any(pattern in name for pattern in INT8_KEEP_FLOAT_FP32_NAME_PATTERNS): return t.float().contiguous() @@ -317,12 +307,9 @@ def keep_float_tensor(name: str, t: Tensor, passthrough_orig_dtypes: dict[str, s passthrough_orig_dtypes[name] = str(t.dtype).removeprefix("torch.") return t.to(dtype=INT8_KEEP_FLOAT_STORE_DTYPE).contiguous() return t - def quantize_float_tensor(t: Tensor) -> tuple[Tensor, Tensor]: t32 = t.float() if t32.ndim == 2: - # Matrices get one scale per row, which usually tracks output-channel - # ranges much better than a single tensor-wide scale. clip_abs = ( torch.quantile(t32.abs(), INT8_CLIP_Q, dim=1) if t32.numel() @@ -332,19 +319,11 @@ def quantize_float_tensor(t: Tensor) -> tuple[Tensor, Tensor]: scale = (clip_abs / 127.0).clamp_min(1.0 / 127.0) q = torch.clamp(torch.round(clipped / scale[:, None]), -127, 127).to(torch.int8).contiguous() return q, scale.to(dtype=INT8_PER_ROW_SCALE_DTYPE).contiguous() - - # Vectors / scalars use a simpler per-tensor scale. clip_abs = float(torch.quantile(t32.abs().flatten(), INT8_CLIP_Q).item()) if t32.numel() else 0.0 scale = torch.tensor(clip_abs / 127.0 if clip_abs > 0 else 1.0, dtype=torch.float32) q = torch.clamp(torch.round(torch.clamp(t32, -clip_abs, clip_abs) / scale), -127, 127).to(torch.int8).contiguous() return q, scale - def quantize_state_dict_int8(state_dict: dict[str, Tensor]): - # Single supported clean-script export format: - # - per-row int8 for 2D float tensors - # - per-tensor int8 for other float tensors - # - exact passthrough for non-floats - # - passthrough for small float tensors, stored as fp16 to save bytes quantized: dict[str, Tensor] = {} scales: dict[str, Tensor] = {} dtypes: dict[str, str] = {} @@ -355,27 +334,21 @@ def quantize_state_dict_int8(state_dict: dict[str, Tensor]): ("param_count", "num_tensors", "num_float_tensors", "num_nonfloat_tensors", "baseline_tensor_bytes", "int8_payload_bytes"), 0, ) - for name, tensor in state_dict.items(): t = tensor.detach().to("cpu").contiguous() stats["param_count"] += int(t.numel()) stats["num_tensors"] += 1 stats["baseline_tensor_bytes"] += tensor_nbytes(t) - if not t.is_floating_point(): stats["num_nonfloat_tensors"] += 1 passthrough[name] = t stats["int8_payload_bytes"] += tensor_nbytes(t) continue - - # Small float tensors are cheap enough to keep directly. We still downcast - # fp32/bf16 passthrough tensors to fp16 so metadata does not dominate size. if t.numel() <= INT8_KEEP_FLOAT_MAX_NUMEL: kept = keep_float_tensor(name, t, passthrough_orig_dtypes) passthrough[name] = kept stats["int8_payload_bytes"] += tensor_nbytes(kept) continue - stats["num_float_tensors"] += 1 q, s = quantize_float_tensor(t) if s.ndim > 0: @@ -384,7 +357,6 @@ def quantize_state_dict_int8(state_dict: dict[str, Tensor]): scales[name] = s dtypes[name] = str(t.dtype).removeprefix("torch.") stats["int8_payload_bytes"] += tensor_nbytes(q) + tensor_nbytes(s) - obj: dict[str, object] = { "__quant_format__": "int8_clean_per_row_v1", "quantized": quantized, @@ -397,7 +369,6 @@ def quantize_state_dict_int8(state_dict: dict[str, Tensor]): if passthrough_orig_dtypes: obj["passthrough_orig_dtypes"] = passthrough_orig_dtypes return obj, stats - def dequantize_state_dict_int8(obj: dict[str, object]) -> dict[str, Tensor]: out: dict[str, Tensor] = {} qmeta = obj.get("qmeta", {}) @@ -407,30 +378,21 @@ def dequantize_state_dict_int8(obj: dict[str, object]) -> dict[str, Tensor]: s = obj["scales"][name] if qmeta.get(name, {}).get("scheme") == "per_row" or s.ndim > 0: s = s.to(dtype=torch.float32) - # Broadcast the saved row scale back across trailing dimensions. out[name] = (q.float() * s.view(q.shape[0], *([1] * (q.ndim - 1)))).to(dtype=dtype).contiguous() else: scale = float(s.item()) out[name] = (q.float() * scale).to(dtype=dtype).contiguous() for name, t in obj["passthrough"].items(): - # Restore small tensors, undoing the temporary fp16 storage cast if needed. out_t = t.detach().to("cpu").contiguous() orig_dtype = passthrough_orig_dtypes.get(name) if isinstance(orig_dtype, str): out_t = out_t.to(dtype=getattr(torch, orig_dtype)).contiguous() out[name] = out_t return out - - -# ----------------------------- -# DATA LOADING -# ----------------------------- - def load_data_shard(file: Path) -> Tensor: header_bytes = 256 * np.dtype(" Tensor: if tokens_np.size != num_tokens: raise ValueError(f"Short read for {file}") return torch.from_numpy(tokens_np.astype(np.uint16, copy=False)) - - class TokenStream: - # Reads shards sequentially and wraps around forever. The training loop therefore - # has deterministic, simple streaming behavior with no sampling or workers. def __init__(self, pattern: str): self.files = [Path(p) for p in sorted(glob.glob(pattern))] if not self.files: @@ -453,12 +411,10 @@ def __init__(self, pattern: str): self.file_idx = 0 self.tokens = load_data_shard(self.files[0]) self.pos = 0 - def _advance_file(self) -> None: self.file_idx = (self.file_idx + 1) % len(self.files) self.tokens = load_data_shard(self.files[self.file_idx]) self.pos = 0 - def take(self, n: int) -> Tensor: chunks: list[Tensor] = [] remaining = n @@ -472,17 +428,12 @@ def take(self, n: int) -> Tensor: self.pos += k remaining -= k return chunks[0] if len(chunks) == 1 else torch.cat(chunks) - - class DistributedTokenLoader: - # Each call consumes a contiguous chunk from the shared token stream, then slices out - # one disjoint span per rank. The extra "+1" token lets us build (x, y) by shifting. def __init__(self, pattern: str, rank: int, world_size: int, device: torch.device): self.rank = rank self.world_size = world_size self.device = device self.stream = TokenStream(pattern) - def next_batch(self, global_tokens: int, seq_len: int, grad_accum_steps: int) -> tuple[Tensor, Tensor]: local_tokens = global_tokens // (self.world_size * grad_accum_steps) per_rank_span = local_tokens + 1 @@ -492,45 +443,43 @@ def next_batch(self, global_tokens: int, seq_len: int, grad_accum_steps: int) -> x = local[:-1].reshape(-1, seq_len) y = local[1:].reshape(-1, seq_len) return x.to(self.device, non_blocking=True), y.to(self.device, non_blocking=True) - -# ----------------------------- -# TRANSFORMER MODULES -# ----------------------------- - class RMSNorm(nn.Module): def __init__(self, eps: float | None = None): super().__init__() self.eps = eps - def forward(self, x: Tensor) -> Tensor: return F.rms_norm(x, (x.size(-1),), eps=self.eps) - - class CastedLinear(nn.Linear): - # Keep weights in fp32 for optimizer/state quality, cast at matmul time for bf16 compute. + _qat_enabled: bool = False def forward(self, x: Tensor) -> Tensor: + w = self.weight.to(x.dtype) + if CastedLinear._qat_enabled and self.training and w.ndim == 2: + with torch.no_grad(): + w32 = self.weight.float() + # Use 99.95th percentile clipping to match GPTQ export quantizer + row_clip = torch.quantile(w32.abs(), 0.9995, dim=1) + scale = (row_clip / 31.0).clamp_min(1.0 / 31.0) + w_q = (torch.clamp(torch.round(w32 / scale[:, None]), -32, 31) * scale[:, None]).to(x.dtype) + w = w + (w_q - w).detach() bias = self.bias.to(x.dtype) if self.bias is not None else None - return F.linear(x, self.weight.to(x.dtype), bias) - - + return F.linear(x, w, bias) def restore_low_dim_params_to_fp32(module: nn.Module) -> None: - # Keep small/control parameters in fp32 even when the model body runs in bf16. with torch.no_grad(): for name, param in module.named_parameters(): if (param.ndim < 2 or any(pattern in name for pattern in CONTROL_TENSOR_NAME_PATTERNS)) and param.dtype != torch.float32: param.data = param.data.float() - - class Rotary(nn.Module): - # Caches cos/sin tables per sequence length on the current device. - def __init__(self, dim: int, base: float = 10000.0): + def __init__(self, dim: int, base: float = 10000.0, train_seq_len: int = 1024, rope_dims: int = 0): super().__init__() - inv_freq = 1.0 / (base ** (torch.arange(0, dim, 2, dtype=torch.float32) / dim)) + self.dim = dim + self.base = base + self.train_seq_len = train_seq_len + self.rope_dims = rope_dims if rope_dims > 0 else dim + inv_freq = 1.0 / (base ** (torch.arange(0, self.rope_dims, 2, dtype=torch.float32) / self.rope_dims)) self.register_buffer("inv_freq", inv_freq, persistent=False) self._seq_len_cached = 0 self._cos_cached: Tensor | None = None self._sin_cached: Tensor | None = None - def forward(self, seq_len: int, device: torch.device, dtype: torch.dtype) -> tuple[Tensor, Tensor]: if ( self._cos_cached is None @@ -538,20 +487,29 @@ def forward(self, seq_len: int, device: torch.device, dtype: torch.dtype) -> tup or self._seq_len_cached != seq_len or self._cos_cached.device != device ): - t = torch.arange(seq_len, device=device, dtype=self.inv_freq.dtype) - freqs = torch.outer(t, self.inv_freq.to(device)) - self._cos_cached = freqs.cos()[None, None, :, :] - self._sin_cached = freqs.sin()[None, None, :, :] + rd = self.rope_dims + if seq_len > self.train_seq_len: + scale = seq_len / self.train_seq_len + new_base = self.base * (scale ** (rd / (rd - 2))) + inv_freq = 1.0 / (new_base ** (torch.arange(0, rd, 2, dtype=torch.float32, device=device) / rd)) + else: + inv_freq = self.inv_freq.to(device) + t = torch.arange(seq_len, device=device, dtype=inv_freq.dtype) + freqs = torch.outer(t, inv_freq) + self._cos_cached = freqs.cos()[None, :, None, :] + self._sin_cached = freqs.sin()[None, :, None, :] self._seq_len_cached = seq_len return self._cos_cached.to(dtype=dtype), self._sin_cached.to(dtype=dtype) - - -def apply_rotary_emb(x: Tensor, cos: Tensor, sin: Tensor) -> Tensor: +def apply_rotary_emb(x: Tensor, cos: Tensor, sin: Tensor, rope_dims: int = 0) -> Tensor: + if rope_dims > 0 and rope_dims < x.size(-1): + x_rope, x_pass = x[..., :rope_dims], x[..., rope_dims:] + half = rope_dims // 2 + x1, x2 = x_rope[..., :half], x_rope[..., half:] + x_rope = torch.cat((x1 * cos + x2 * sin, x1 * (-sin) + x2 * cos), dim=-1) + return torch.cat((x_rope, x_pass), dim=-1) half = x.size(-1) // 2 x1, x2 = x[..., :half], x[..., half:] return torch.cat((x1 * cos + x2 * sin, x1 * (-sin) + x2 * cos), dim=-1) - - class CausalSelfAttention(nn.Module): def __init__( self, @@ -578,45 +536,102 @@ def __init__( self.proj = CastedLinear(dim, dim, bias=False) self.proj._zero_init = True self.q_gain = nn.Parameter(torch.full((num_heads,), qk_gain_init, dtype=torch.float32)) - self.rotary = Rotary(self.head_dim, base=rope_base) - - def forward(self, x: Tensor) -> Tensor: + self.rope_dims = 0 # set by GPT.__init__ for partial RoPE + self.rotary = Rotary(self.head_dim, base=rope_base, train_seq_len=1024) + self.use_xsa = False # set by GPT.__init__ for deep layers only + def _xsa_efficient(self, y: Tensor, v: Tensor) -> Tensor: + """Efficient XSA: subtract self-value projection via GQA-aware reshape (no repeat_interleave). + y: [B, T, H, D], v: [B, T, Hkv, D]. H must be divisible by Hkv.""" + B, T, H, D = y.shape + Hkv = v.size(-2) + group = H // Hkv + y_g = y.reshape(B, T, Hkv, group, D) # [B, T, Hkv, group, D] + vn = F.normalize(v, dim=-1).unsqueeze(-2) # [B, T, Hkv, 1, D] — broadcast ready + proj = (y_g * vn).sum(dim=-1, keepdim=True) * vn + return (y_g - proj).reshape(B, T, H, D) + def forward(self, x: Tensor, v_embed: Tensor | None = None) -> Tensor: bsz, seqlen, dim = x.shape - q = self.c_q(x).reshape(bsz, seqlen, self.num_heads, self.head_dim).transpose(1, 2) - k = self.c_k(x).reshape(bsz, seqlen, self.num_kv_heads, self.head_dim).transpose(1, 2) - v = self.c_v(x).reshape(bsz, seqlen, self.num_kv_heads, self.head_dim).transpose(1, 2) + q = self.c_q(x).reshape(bsz, seqlen, self.num_heads, self.head_dim) + k = self.c_k(x).reshape(bsz, seqlen, self.num_kv_heads, self.head_dim) + v = self.c_v(x) + if v_embed is not None: + v = v + v_embed + v = v.reshape(bsz, seqlen, self.num_kv_heads, self.head_dim) q = F.rms_norm(q, (q.size(-1),)) k = F.rms_norm(k, (k.size(-1),)) cos, sin = self.rotary(seqlen, x.device, q.dtype) - q = apply_rotary_emb(q, cos, sin) - k = apply_rotary_emb(k, cos, sin) - q = q * self.q_gain.to(dtype=q.dtype)[None, :, None, None] - y = F.scaled_dot_product_attention( - q, - k, - v, - attn_mask=None, - is_causal=True, - enable_gqa=(self.num_kv_heads != self.num_heads), - ) - y = y.transpose(1, 2).contiguous().reshape(bsz, seqlen, dim) + q = apply_rotary_emb(q, cos, sin, self.rope_dims) + k = apply_rotary_emb(k, cos, sin, self.rope_dims) + q = q * self.q_gain.to(dtype=q.dtype)[None, None, :, None] + y = flash_attn_3_func(q, k, v, causal=True) + if self.use_xsa: + y = self._xsa_efficient(y, v) + y = y.reshape(bsz, seqlen, dim) return self.proj(y) - - +class SmearGate(nn.Module): + def __init__(self, dim: int): + super().__init__() + self.gate = nn.Parameter(torch.zeros(dim, dtype=torch.float32)) + def forward(self, x: Tensor) -> Tensor: + g = torch.sigmoid(self.gate.to(dtype=x.dtype))[None, None, :] + x_prev = torch.cat([torch.zeros_like(x[:, :1]), x[:, :-1]], dim=1) + return (1 - g) * x + g * x_prev +class BigramHashEmbedding(nn.Module): + def __init__(self, bigram_vocab_size: int, bigram_dim: int, model_dim: int): + super().__init__() + self.bigram_vocab_size = bigram_vocab_size + self.embed = nn.Embedding(bigram_vocab_size, bigram_dim) + nn.init.zeros_(self.embed.weight) + self.proj = CastedLinear(bigram_dim, model_dim, bias=False) if bigram_dim != model_dim else None + if self.proj is not None: + nn.init.zeros_(self.proj.weight) + self.scale = nn.Parameter(torch.tensor(0.05, dtype=torch.float32)) + def bigram_hash(self, tokens: Tensor) -> Tensor: + t = tokens.to(torch.int32) + mod = self.bigram_vocab_size - 1 + out = torch.empty_like(t) + out[..., 0] = mod + out[..., 1:] = torch.bitwise_xor(36313 * t[..., 1:], 27191 * t[..., :-1]) % mod + return out.long() + def forward(self, token_ids: Tensor) -> Tensor: + h = self.embed(self.bigram_hash(token_ids)) + if self.proj is not None: + h = self.proj(h) + return h * self.scale.to(dtype=h.dtype) +class ValueEmbedding(nn.Module): + """Reinject token identity into attention values at specific layers. + Each table maps vocab tokens to a low-dim embedding, projected to model_dim.""" + def __init__(self, vocab_size: int, ve_dim: int, model_dim: int): + super().__init__() + self.embed = nn.Embedding(vocab_size, ve_dim) + nn.init.normal_(self.embed.weight, std=0.01) + self.proj = CastedLinear(ve_dim, model_dim, bias=False) if ve_dim != model_dim else None + if self.proj is not None: + nn.init.zeros_(self.proj.weight) + self.scale = nn.Parameter(torch.tensor(0.1, dtype=torch.float32)) + def forward(self, token_ids: Tensor) -> Tensor: + h = self.embed(token_ids) + if self.proj is not None: + h = self.proj(h) + return h * self.scale.to(dtype=h.dtype) class MLP(nn.Module): - # relu^2 MLP from the original modded-nanogpt setup - def __init__(self, dim: int, mlp_mult: int): + def __init__(self, dim: int, mlp_mult: int, mlp_act: str = "relu_sq", mlp_leaky_slope: float = 0.5): super().__init__() - hidden = mlp_mult * dim + hidden = int(mlp_mult * dim) self.fc = CastedLinear(dim, hidden, bias=False) self.proj = CastedLinear(hidden, dim, bias=False) self.proj._zero_init = True - + self.mlp_act = mlp_act + self.mlp_leaky_slope = mlp_leaky_slope + if self.mlp_act not in {"relu_sq", "leaky_relu_sq"}: + raise ValueError(f"Unsupported MLP_ACT '{self.mlp_act}'. Use 'relu_sq' or 'leaky_relu_sq'.") def forward(self, x: Tensor) -> Tensor: - x = torch.relu(self.fc(x)) + x = self.fc(x) + if self.mlp_act == "leaky_relu_sq": + x = F.leaky_relu(x, negative_slope=self.mlp_leaky_slope) + else: + x = F.relu(x) return self.proj(x.square()) - - class Block(nn.Module): def __init__( self, @@ -626,25 +641,37 @@ def __init__( mlp_mult: int, rope_base: float, qk_gain_init: float, + layer_idx: int = 0, + ln_scale: bool = False, + dtg: bool = False, + mlp_act: str = "relu_sq", + mlp_leaky_slope: float = 0.5, ): super().__init__() self.attn_norm = RMSNorm() self.mlp_norm = RMSNorm() self.attn = CausalSelfAttention(dim, num_heads, num_kv_heads, rope_base, qk_gain_init) - self.mlp = MLP(dim, mlp_mult) + self.mlp = MLP(dim, mlp_mult, mlp_act=mlp_act, mlp_leaky_slope=mlp_leaky_slope) self.attn_scale = nn.Parameter(torch.ones(dim, dtype=torch.float32)) self.mlp_scale = nn.Parameter(torch.ones(dim, dtype=torch.float32)) self.resid_mix = nn.Parameter(torch.stack((torch.ones(dim), torch.zeros(dim))).float()) - - def forward(self, x: Tensor, x0: Tensor) -> Tensor: + self.ln_scale_factor = 1.0 / math.sqrt(layer_idx + 1) if ln_scale else 1.0 + if dtg: + self.dtg_gate = nn.Linear(dim, 1, bias=True) + nn.init.zeros_(self.dtg_gate.weight) + nn.init.constant_(self.dtg_gate.bias, 2.0) + else: + self.dtg_gate = None + def forward(self, x: Tensor, x0: Tensor, v_embed: Tensor | None = None) -> Tensor: mix = self.resid_mix.to(dtype=x.dtype) - x = mix[0][None, None, :] * x + mix[1][None, None, :] * x0 - attn_out = self.attn(self.attn_norm(x)) - x = x + self.attn_scale.to(dtype=x.dtype)[None, None, :] * attn_out - x = x + self.mlp_scale.to(dtype=x.dtype)[None, None, :] * self.mlp(self.mlp_norm(x)) - return x - - + x_in = mix[0][None, None, :] * x + mix[1][None, None, :] * x0 + attn_out = self.attn(self.attn_norm(x_in) * self.ln_scale_factor, v_embed=v_embed) + x_out = x_in + self.attn_scale.to(dtype=x_in.dtype)[None, None, :] * attn_out + x_out = x_out + self.mlp_scale.to(dtype=x_out.dtype)[None, None, :] * self.mlp(self.mlp_norm(x_out) * self.ln_scale_factor) + if self.dtg_gate is not None: + gate = torch.sigmoid(self.dtg_gate(x_in.detach())) + x_out = x_in + gate * (x_out - x_in) + return x_out class GPT(nn.Module): def __init__( self, @@ -659,14 +686,34 @@ def __init__( logit_softcap: float, rope_base: float, qk_gain_init: float, + mtp_num_heads: int = 0, + mtp_loss_weight: float = 0.1, + bigram_vocab_size: int = 0, + bigram_dim: int = 128, + xsa_last_n: int = 0, + rope_dims: int = 0, + ln_scale: bool = False, + dtg: bool = False, + ve_enabled: bool = False, + ve_dim: int = 128, + ve_layers: str = "9,10", + mlp_act: str = "relu_sq", + mlp_leaky_slope: float = 0.5, + f1_corr_rank: int = 0, + f1_corr_scale_init: float = 0.10, ): super().__init__() + self._ve_target_dim = num_kv_heads * (model_dim // num_heads) # kv_dim for value projection if logit_softcap <= 0.0: raise ValueError(f"logit_softcap must be positive, got {logit_softcap}") self.tie_embeddings = tie_embeddings self.tied_embed_init_std = tied_embed_init_std self.logit_softcap = logit_softcap + self.mtp_num_heads = mtp_num_heads + self.mtp_loss_weight = mtp_loss_weight self.tok_emb = nn.Embedding(vocab_size, model_dim) + self.bigram = BigramHashEmbedding(bigram_vocab_size, bigram_dim, model_dim) if bigram_vocab_size > 0 else None + self.smear = SmearGate(model_dim) self.num_encoder_layers = num_layers // 2 self.num_decoder_layers = num_layers - self.num_encoder_layers self.num_skip_weights = min(self.num_encoder_layers, self.num_decoder_layers) @@ -680,65 +727,773 @@ def __init__( mlp_mult, rope_base, qk_gain_init, + layer_idx=i, + ln_scale=ln_scale, + dtg=dtg, + mlp_act=mlp_act, + mlp_leaky_slope=mlp_leaky_slope, ) for i in range(num_layers) ] ) + if rope_dims > 0: + head_dim = model_dim // num_heads + for block in self.blocks: + block.attn.rope_dims = rope_dims + block.attn.rotary = Rotary(head_dim, base=rope_base, train_seq_len=1024, rope_dims=rope_dims) + self.ve_layer_indices = [int(x) for x in ve_layers.split(",") if x.strip()] if ve_enabled else [] + kv_dim = self._ve_target_dim + if self.ve_layer_indices: + self.ve_shared = ValueEmbedding(vocab_size, ve_dim, kv_dim) + self.ve_layer_scales = nn.ParameterList( + [nn.Parameter(torch.ones(1, dtype=torch.float32)) for _ in self.ve_layer_indices] + ) + else: + self.ve_shared = None + self.ve_layer_scales = nn.ParameterList() + self.value_embeds = nn.ModuleList() # keep empty for compat self.final_norm = RMSNorm() self.lm_head = None if tie_embeddings else CastedLinear(model_dim, vocab_size, bias=False) if self.lm_head is not None: self.lm_head._zero_init = True + self.mtp_heads = nn.ModuleList( + [CastedLinear(model_dim, vocab_size, bias=False) for _ in range(mtp_num_heads)] + ) + for head in self.mtp_heads: + head._zero_init = True + # Low-rank correction path for extra capacity under size budget. + self.f1_corr_rank = f1_corr_rank + if f1_corr_rank > 0: + self.f1_corr_in = CastedLinear(model_dim, f1_corr_rank, bias=False) + self.f1_corr_out = CastedLinear(f1_corr_rank, vocab_size, bias=False) + self.f1_corr_out._zero_init = True + self.f1_corr_scale = nn.Parameter(torch.tensor(f1_corr_scale_init, dtype=torch.float32)) + else: + self.f1_corr_in = None + self.f1_corr_out = None + self.f1_corr_scale = None + if xsa_last_n > 0: + for i in range(max(0, num_layers - xsa_last_n), num_layers): + self.blocks[i].attn.use_xsa = True self._init_weights() - def _init_weights(self) -> None: if self.tie_embeddings: nn.init.normal_(self.tok_emb.weight, mean=0.0, std=self.tied_embed_init_std) - for module in self.modules(): - if isinstance(module, nn.Linear) and getattr(module, "_zero_init", False): - nn.init.zeros_(module.weight) - + num_layers = len(self.blocks) + for name, module in self.named_modules(): + if isinstance(module, nn.Linear): + if getattr(module, "_zero_init", False): + nn.init.zeros_(module.weight) + elif module.weight.ndim == 2 and module.weight.shape[0] >= 64 and module.weight.shape[1] >= 64: + nn.init.orthogonal_(module.weight, gain=1.0) + if ".proj." in name or name.endswith(".proj"): + with torch.no_grad(): + module.weight.mul_(1.0 / math.sqrt(2 * num_layers)) + def _get_ve(self, layer_idx: int, input_ids: Tensor, ve_cache: dict | None = None) -> Tensor | None: + """Get value embedding for a specific layer using shared table + per-layer scale.""" + if self.ve_shared is None or layer_idx not in self.ve_layer_indices: + return None + if ve_cache is not None and 've' not in ve_cache: + ve_cache['ve'] = self.ve_shared(input_ids) + ve_base = ve_cache['ve'] if ve_cache is not None else self.ve_shared(input_ids) + ve_idx = self.ve_layer_indices.index(layer_idx) + return ve_base * self.ve_layer_scales[ve_idx].to(dtype=ve_base.dtype) def forward(self, input_ids: Tensor, target_ids: Tensor) -> Tensor: x = self.tok_emb(input_ids) + if self.bigram is not None: + x = x + self.bigram(input_ids) x = F.rms_norm(x, (x.size(-1),)) + x = self.smear(x) x0 = x skips: list[Tensor] = [] - - # First half stores skips; second half reuses them in reverse order. + ve_cache: dict = {} for i in range(self.num_encoder_layers): - x = self.blocks[i](x, x0) + ve = self._get_ve(i, input_ids, ve_cache) + x = self.blocks[i](x, x0, v_embed=ve) skips.append(x) for i in range(self.num_decoder_layers): + bi = self.num_encoder_layers + i if skips: x = x + self.skip_weights[i].to(dtype=x.dtype)[None, None, :] * skips.pop() - x = self.blocks[self.num_encoder_layers + i](x, x0) - - x = self.final_norm(x).reshape(-1, x.size(-1)) + ve = self._get_ve(bi, input_ids, ve_cache) + x = self.blocks[bi](x, x0, v_embed=ve) + x = self.final_norm(x) + x_flat = x.reshape(-1, x.size(-1)) targets = target_ids.reshape(-1) if self.tie_embeddings: - logits_proj = F.linear(x, self.tok_emb.weight) + logits_proj = F.linear(x_flat, self.tok_emb.weight) else: if self.lm_head is None: raise RuntimeError("lm_head is required when tie_embeddings=False") - logits_proj = self.lm_head(x) + logits_proj = self.lm_head(x_flat) + if self.f1_corr_in is not None and self.f1_corr_out is not None and self.f1_corr_scale is not None: + corr_hidden = F.silu(self.f1_corr_in(x_flat)) + corr_proj = self.f1_corr_out(corr_hidden) + logits_proj = logits_proj + self.f1_corr_scale.to(dtype=logits_proj.dtype) * corr_proj logits = self.logit_softcap * torch.tanh(logits_proj / self.logit_softcap) - return F.cross_entropy(logits.float(), targets, reduction="mean") + main_loss = F.cross_entropy(logits.float(), targets, reduction="mean") + if self.training and self.mtp_num_heads > 0 and self.mtp_loss_weight > 0.0: + _, seqlen, dim = x.shape + mtp_loss_sum = x.new_zeros(()) + mtp_loss_count = 0 + for k, mtp_head in enumerate(self.mtp_heads): + valid_t = seqlen - (k + 1) + if valid_t <= 0: + continue + mtp_hidden = x[:, :valid_t, :].reshape(-1, dim) + mtp_targets = target_ids[:, k + 1 :].reshape(-1) + mtp_logits_proj = mtp_head(mtp_hidden) + mtp_logits = self.logit_softcap * torch.tanh(mtp_logits_proj / self.logit_softcap) + mtp_loss_sum = mtp_loss_sum + F.cross_entropy(mtp_logits.float(), mtp_targets, reduction="mean") + mtp_loss_count += 1 + if mtp_loss_count > 0: + main_loss = main_loss + self.mtp_loss_weight * (mtp_loss_sum / mtp_loss_count) + return main_loss + def forward_logits(self, input_ids: Tensor) -> Tensor: + """Return logits (bsz, seq_len, vocab) without computing loss.""" + x = self.tok_emb(input_ids) + if self.bigram is not None: + x = x + self.bigram(input_ids) + x = F.rms_norm(x, (x.size(-1),)) + x = self.smear(x) + x0 = x + skips: list[Tensor] = [] + ve_cache: dict = {} + for i in range(self.num_encoder_layers): + ve = self._get_ve(i, input_ids, ve_cache) + x = self.blocks[i](x, x0, v_embed=ve) + skips.append(x) + for i in range(self.num_decoder_layers): + bi = self.num_encoder_layers + i + if skips: + x = x + self.skip_weights[i].to(dtype=x.dtype)[None, None, :] * skips.pop() + ve = self._get_ve(bi, input_ids, ve_cache) + x = self.blocks[bi](x, x0, v_embed=ve) + x = self.final_norm(x) + if self.tie_embeddings: + logits_proj = F.linear(x, self.tok_emb.weight) + else: + logits_proj = self.lm_head(x) + if self.f1_corr_in is not None and self.f1_corr_out is not None and self.f1_corr_scale is not None: + corr_hidden = F.silu(self.f1_corr_in(x)) + corr_proj = self.f1_corr_out(corr_hidden) + logits_proj = logits_proj + self.f1_corr_scale.to(dtype=logits_proj.dtype) * corr_proj + return self.logit_softcap * torch.tanh(logits_proj / self.logit_softcap) +def eval_val_sliding( + args: Hyperparameters, + base_model: nn.Module, + rank: int, + world_size: int, + device: torch.device, + val_tokens: Tensor, + base_bytes_lut: Tensor, + has_leading_space_lut: Tensor, + is_boundary_token_lut: Tensor, + stride: int, + batch_seqs: int = 32, + eval_seq_len: int | None = None, +) -> tuple[float, float]: + """Sliding window evaluation: each token scored with maximum context.""" + seq_len = eval_seq_len or args.train_seq_len + total_tokens = val_tokens.numel() - 1 + window_starts = [ws for ws in range(0, total_tokens, stride) + if min(ws + seq_len, total_tokens) - ws >= 1] + total_windows = len(window_starts) + my_s = (total_windows * rank) // world_size + my_e = (total_windows * (rank + 1)) // world_size + my_windows = window_starts[my_s:my_e] + loss_sum = torch.zeros((), device=device, dtype=torch.float64) + token_count = torch.zeros((), device=device, dtype=torch.float64) + byte_count = torch.zeros((), device=device, dtype=torch.float64) + base_model.eval() + compiled_logits = maybe_torch_compile(base_model.forward_logits, args) + with torch.inference_mode(): + for bi in range(0, len(my_windows), batch_seqs): + batch_ws = my_windows[bi:bi + batch_seqs] + bsz = len(batch_ws) + x_batch = torch.zeros(bsz, seq_len, dtype=torch.int64, device=device) + y_batch = torch.zeros(bsz, seq_len, dtype=torch.int64, device=device) + wlens: list[int] = [] + for i, ws in enumerate(batch_ws): + end = min(ws + seq_len, total_tokens) + wlen = end - ws + wlens.append(wlen) + chunk = val_tokens[ws:end + 1].to(dtype=torch.int64, device=device) + x_batch[i, :wlen] = chunk[:-1] + y_batch[i, :wlen] = chunk[1:] + with torch.autocast(device_type="cuda", dtype=torch.bfloat16): + logits = compiled_logits(x_batch) + nll = F.cross_entropy( + logits.reshape(-1, logits.size(-1)).float(), + y_batch.reshape(-1), + reduction="none", + ).reshape(bsz, seq_len) + for i, ws in enumerate(batch_ws): + wlen = wlens[i] + s = 0 if ws == 0 else max(wlen - stride, 0) + scored_nll = nll[i, s:wlen].to(torch.float64) + loss_sum += scored_nll.sum() + token_count += float(wlen - s) + tgt = y_batch[i, s:wlen] + prev = x_batch[i, s:wlen] + tb = base_bytes_lut[tgt].to(torch.float64) + tb += (has_leading_space_lut[tgt] & ~is_boundary_token_lut[prev]).to(torch.float64) + byte_count += tb.sum() + if dist.is_available() and dist.is_initialized(): + dist.all_reduce(loss_sum, op=dist.ReduceOp.SUM) + dist.all_reduce(token_count, op=dist.ReduceOp.SUM) + dist.all_reduce(byte_count, op=dist.ReduceOp.SUM) + val_loss = (loss_sum / token_count).item() + bits_per_token = val_loss / math.log(2.0) + tokens_per_byte = token_count.item() / byte_count.item() + base_model.train() + return val_loss, bits_per_token * tokens_per_byte +def eval_val_sliding_hashed_ngram( + args: Hyperparameters, + base_model: nn.Module, + rank: int, + world_size: int, + device: torch.device, + val_tokens: Tensor, + base_bytes_lut: Tensor, + has_leading_space_lut: Tensor, + is_boundary_token_lut: Tensor, + stride: int, + order: int, + alpha: float, + min_count: int, + buckets: int, + max_seconds: float = 0.0, + batch_seqs: int = 32, + eval_seq_len: int | None = None, +) -> tuple[float, float, float]: + """Score-first sliding eval with fixed-weight hashed n-gram interpolation. + + Legal behavior: + - per-token score is computed before that token updates the cache + - no target-aware gating/min-NLL selection is used + """ + if order < 2: + raise ValueError(f"NGRAM_EVAL_ORDER must be >=2, got {order}") + if not (0.0 <= alpha <= 1.0): + raise ValueError(f"NGRAM_EVAL_ALPHA must be in [0, 1], got {alpha}") + if min_count < 1: + raise ValueError(f"NGRAM_EVAL_MIN_COUNT must be >=1, got {min_count}") + if buckets < 1024: + raise ValueError(f"NGRAM_EVAL_BUCKETS must be >=1024, got {buckets}") + if max_seconds < 0.0: + raise ValueError(f"NGRAM_EVAL_MAX_SECONDS must be >=0, got {max_seconds}") + + seq_len = eval_seq_len or args.train_seq_len + total_tokens = val_tokens.numel() - 1 + all_window_starts = [ws for ws in range(0, total_tokens, stride) if min(ws + seq_len, total_tokens) - ws >= 1] + total_scored_tokens = 0.0 + for ws in all_window_starts: + end = min(ws + seq_len, total_tokens) + wlen = end - ws + s = 0 if ws == 0 else max(wlen - stride, 0) + total_scored_tokens += float(max(wlen - s, 0)) + # Distribute windows across ranks + my_s = (len(all_window_starts) * rank) // world_size + my_e = (len(all_window_starts) * (rank + 1)) // world_size + window_starts = all_window_starts[my_s:my_e] + + val_np = val_tokens.numpy() + ctx_table = np.zeros((buckets,), dtype=np.uint32) + full_table = np.zeros((buckets,), dtype=np.uint32) + mask = np.uint64(buckets - 1) + primes = np.array( + [np.uint64(36313), np.uint64(27191), np.uint64(51647), np.uint64(81929), np.uint64(131071)], + dtype=np.uint64, + ) + loss_sum = 0.0 + token_count = 0.0 + byte_count = 0.0 -# ----------------------------- -# TRAINING -# ----------------------------- + base_model.eval() + compiled_logits = maybe_torch_compile(base_model.forward_logits, args) + t0 = time.perf_counter() + deadline = (t0 + max_seconds) if max_seconds > 0.0 else None + cutoff_hit = False + with torch.inference_mode(): + for bi in range(0, len(window_starts), batch_seqs): + if deadline is not None and time.perf_counter() >= deadline: + cutoff_hit = True + break + batch_ws = window_starts[bi:bi + batch_seqs] + bsz = len(batch_ws) + x_batch = torch.zeros(bsz, seq_len, dtype=torch.int64, device=device) + y_batch = torch.zeros(bsz, seq_len, dtype=torch.int64, device=device) + wlens: list[int] = [] + for i, ws in enumerate(batch_ws): + end = min(ws + seq_len, total_tokens) + wlen = end - ws + wlens.append(wlen) + chunk = val_tokens[ws:end + 1].to(dtype=torch.int64, device=device) + x_batch[i, :wlen] = chunk[:-1] + y_batch[i, :wlen] = chunk[1:] + + with torch.autocast(device_type="cuda", dtype=torch.bfloat16): + logits = compiled_logits(x_batch) + nll = F.cross_entropy( + logits.reshape(-1, logits.size(-1)).float(), + y_batch.reshape(-1), + reduction="none", + ).reshape(bsz, seq_len) + + for i, ws in enumerate(batch_ws): + wlen = wlens[i] + s = 0 if ws == 0 else max(wlen - stride, 0) + seg_len = wlen - s + if seg_len <= 0: + continue + + seg_nll = nll[i, s:wlen].to(torch.float64).cpu().numpy() + seg_model_p = np.exp(-seg_nll) + + global_j = np.arange(ws + s + 1, ws + wlen + 1, dtype=np.int64) + valid = global_j >= (order - 1) + if valid.any(): + v_idx = np.nonzero(valid)[0] + jv = global_j[v_idx] + + ctx_hash = np.zeros((len(jv),), dtype=np.uint64) + ctx_width = order - 1 + for k in range(ctx_width): + tok = val_np[jv - (ctx_width - k)].astype(np.uint64) + ctx_hash ^= tok * primes[k % len(primes)] + ctx_key = (ctx_hash & mask).astype(np.int64) + + tgt_np = val_np[jv].astype(np.uint64) + full_key = ((ctx_hash ^ (tgt_np * primes[ctx_width % len(primes)])) & mask).astype(np.int64) + + ctx_counts = ctx_table[ctx_key].astype(np.float64) + full_counts = full_table[full_key].astype(np.float64) + can_mix = ctx_counts >= float(min_count) + if can_mix.any(): + # Collision-safe estimate: ensure n-gram probability stays in [0, 1]. + # With hashed sketches, full_counts can exceed ctx_counts due collisions. + p_ng = np.minimum(full_counts, ctx_counts) / np.maximum(ctx_counts, 1.0) + p_ng = np.clip(p_ng, 0.0, 1.0) + mixed = (1.0 - alpha) * seg_model_p[v_idx] + alpha * p_ng + seg_model_p[v_idx[can_mix]] = mixed[can_mix] + seg_nll = -np.log(np.clip(seg_model_p, 1e-12, 1.0)) + + # Score-first legality: update cache only after segment scoring. + np.add.at(ctx_table, ctx_key, 1) + np.add.at(full_table, full_key, 1) + + loss_sum += float(seg_nll.sum()) + token_count += float(seg_len) + tgt = y_batch[i, s:wlen] + prev = x_batch[i, s:wlen] + tb = base_bytes_lut[tgt].to(torch.float64) + tb += (has_leading_space_lut[tgt] & ~is_boundary_token_lut[prev]).to(torch.float64) + byte_count += float(tb.sum().item()) + + if (bi // batch_seqs) % 2000 == 0 and bi > 0: + elapsed = time.perf_counter() - t0 + prog = min((bi + bsz) / max(len(window_starts), 1), 1.0) + cur_bpb = (loss_sum / max(token_count, 1.0)) / math.log(2.0) * (token_count / max(byte_count, 1.0)) + print( + f"ngram_eval:progress windows={bi + bsz}/{len(window_starts)} " + f"({prog*100:.1f}%) bpb={cur_bpb:.6f} t={elapsed:.0f}s", + flush=True, + ) + # All-reduce across ranks + _loss = torch.tensor(loss_sum, device=device, dtype=torch.float64) + _toks = torch.tensor(token_count, device=device, dtype=torch.float64) + _bytes = torch.tensor(byte_count, device=device, dtype=torch.float64) + if dist.is_available() and dist.is_initialized(): + dist.all_reduce(_loss, op=dist.ReduceOp.SUM) + dist.all_reduce(_toks, op=dist.ReduceOp.SUM) + dist.all_reduce(_bytes, op=dist.ReduceOp.SUM) + loss_sum = _loss.item() + token_count = _toks.item() + byte_count = _bytes.item() + + coverage = token_count / max(total_scored_tokens, 1.0) + if cutoff_hit: + elapsed = time.perf_counter() - t0 + print( + f"ngram_eval:cutoff max_seconds={max_seconds:.1f} " + f"coverage={coverage*100:.2f}% elapsed={elapsed:.0f}s", + flush=True, + ) + val_loss = loss_sum / max(token_count, 1.0) + val_bpb = val_loss / math.log(2.0) * (token_count / max(byte_count, 1.0)) + base_model.train() + return val_loss, val_bpb, coverage +def _classify_param(name: str) -> str: + if "tok_emb" in name or "lm_head" in name: + return "embed" + if "f1_corr_in" in name or "f1_corr_out" in name: + return "aux" + if ".mlp." in name: + return "mlp" + if ".attn." in name or (".proj." in name and ".mlp." not in name): + return "attn" + return "other" +def eval_val_sliding_ttt( + args: Hyperparameters, base_model: nn.Module, rank: int, world_size: int, + device: torch.device, val_tokens: Tensor, base_bytes_lut: Tensor, + has_leading_space_lut: Tensor, is_boundary_token_lut: Tensor, + stride: int, batch_seqs: int = 32, +) -> tuple[float, float]: + seq_len, total_tokens, ttt_chunk = args.train_seq_len, val_tokens.numel() - 1, args.ttt_chunk_tokens + master = (rank == 0) + log0 = (lambda msg: print(msg, flush=True)) if master else (lambda msg: None) + window_starts = [ws for ws in range(0, total_tokens, stride) if min(ws + seq_len, total_tokens) - ws >= stride or ws == 0] + num_chunks = (total_tokens + ttt_chunk - 1) // ttt_chunk + chunk_windows: list[list[int]] = [[] for _ in range(num_chunks)] + for ws in window_starts: + end, wlen = min(ws + seq_len, total_tokens), min(ws + seq_len, total_tokens) - ws + s = 0 if ws == 0 else max(wlen - stride, 0) + chunk_windows[min((ws + s) // ttt_chunk, num_chunks - 1)].append(ws) + log0(f"ttt_sliding:start chunks={num_chunks} windows={len(window_starts)} lr={args.ttt_lr} epochs={args.ttt_epochs} freeze={args.ttt_freeze_blocks}") + loss_sum = torch.zeros((), device=device, dtype=torch.float64) + token_count = torch.zeros((), device=device, dtype=torch.float64) + byte_count = torch.zeros((), device=device, dtype=torch.float64) + frozen_ids = set(range(min(args.ttt_freeze_blocks, len(base_model.blocks)))) + embed_names = {"tok_emb", "bigram", "ve_shared"} if args.ttt_freeze_embed else set() + ttt_params = [] + for name, p in base_model.named_parameters(): + if any(f"blocks.{bi}." in name for bi in frozen_ids): + p.requires_grad_(False) + elif any(en in name for en in embed_names): + p.requires_grad_(False) + else: + p.requires_grad_(True); ttt_params.append(p) + log0(f"ttt_sliding:unfrozen={sum(p.numel() for p in ttt_params)} freeze_embed={args.ttt_freeze_embed}") + optimizer = torch.optim.SGD(ttt_params, lr=args.ttt_lr, momentum=args.ttt_momentum) + # TTT-EMA: maintain smoothed weights for scoring + ema_decay = args.ttt_ema_decay + ema_state = None + raw_state = None + if ema_decay > 0: + ema_state = {n: p.data.clone() for n, p in base_model.named_parameters() if p.requires_grad} + raw_state = {n: torch.empty_like(p.data) for n, p in base_model.named_parameters() if n in ema_state} + log0(f"ttt_sliding:ema_decay={ema_decay} ema_params={len(ema_state)}") + t0 = time.perf_counter() + cur_lr = args.ttt_lr + for ci in range(num_chunks): + windows = chunk_windows[ci] + if not windows: + continue + chunk_start, chunk_end = ci * ttt_chunk, min((ci + 1) * ttt_chunk, total_tokens) + my_windows = windows[(len(windows) * rank) // world_size:(len(windows) * (rank + 1)) // world_size] + # Swap to EMA weights for scoring (if enabled and past first chunk) + if ema_state is not None and ci > 0: + for n, p in base_model.named_parameters(): + if n in ema_state: + raw_state[n].copy_(p.data) + p.data.copy_(ema_state[n]) + base_model.eval() + with torch.inference_mode(): + for bi in range(0, len(my_windows), batch_seqs): + batch_ws = my_windows[bi:bi + batch_seqs] + bsz = len(batch_ws) + x_batch = torch.zeros(bsz, seq_len, dtype=torch.int64, device=device) + y_batch = torch.zeros(bsz, seq_len, dtype=torch.int64, device=device) + wlens = [] + for i, ws in enumerate(batch_ws): + wlen = min(ws + seq_len, total_tokens) - ws; wlens.append(wlen) + ct = val_tokens[ws:ws + wlen + 1].to(dtype=torch.int64, device=device) + x_batch[i, :wlen] = ct[:-1]; y_batch[i, :wlen] = ct[1:] + with torch.autocast(device_type="cuda", dtype=torch.bfloat16): + logits = base_model.forward_logits(x_batch) + nll = F.cross_entropy( + logits.reshape(-1, logits.size(-1)).float(), + y_batch.reshape(-1), reduction="none", + ).reshape(bsz, seq_len) + for i, ws in enumerate(batch_ws): + wlen, s = wlens[i], 0 if ws == 0 else max(wlens[i] - stride, 0) + loss_sum += nll[i, s:wlen].to(torch.float64).sum(); token_count += float(wlen - s) + tgt, prev = y_batch[i, s:wlen], x_batch[i, s:wlen] + tb = base_bytes_lut[tgt].to(torch.float64) + (has_leading_space_lut[tgt] & ~is_boundary_token_lut[prev]).to(torch.float64) + byte_count += tb.sum() + # Restore raw weights after scoring (for training phase) + if ema_state is not None and ci > 0: + for n, p in base_model.named_parameters(): + if n in raw_state: + p.data.copy_(raw_state[n]) + # Phase 2: TRAIN on this chunk (already scored = legal) + if ci < num_chunks - 1 and ci < args.ttt_max_train_chunks and args.ttt_epochs > 0: + base_model.train() + chunk_seqs = (chunk_end - chunk_start) // seq_len + if chunk_seqs > 0: + cur_lr = args.ttt_lr * 0.5 * (1.0 + math.cos(math.pi * ci / max(args.ttt_max_train_chunks - 1, 1))) + for pg in optimizer.param_groups: + pg['lr'] = cur_lr + ms, me = (chunk_seqs * rank) // world_size, (chunk_seqs * (rank + 1)) // world_size + for _ep in range(args.ttt_epochs): + for bs in range(0, me - ms, args.ttt_batch_seqs): + be = min(bs + args.ttt_batch_seqs, me - ms) + start_tok = chunk_start + (ms + bs) * seq_len + end_tok = chunk_start + (ms + be) * seq_len + 1 + if end_tok > val_tokens.numel(): + continue + local = val_tokens[start_tok:end_tok].to(device=device, dtype=torch.int64) + x = local[:-1].reshape(-1, seq_len) + y = local[1:].reshape(-1, seq_len) + optimizer.zero_grad(set_to_none=True) + with torch.autocast(device_type="cuda", dtype=torch.bfloat16): + loss = base_model(x, y) + loss.backward() + if world_size > 1: + for p in ttt_params: + if p.grad is not None: + dist.all_reduce(p.grad, op=dist.ReduceOp.AVG) + torch.nn.utils.clip_grad_norm_(ttt_params, args.ttt_grad_clip) + optimizer.step() + # Update EMA after this chunk's training + if ema_state is not None: + with torch.no_grad(): + for n, p in base_model.named_parameters(): + if n in ema_state: + ema_state[n].mul_(ema_decay).add_(p.data, alpha=1.0 - ema_decay) + # Once training stops, load EMA weights permanently for remaining score-only chunks + if ema_state is not None and ci == args.ttt_max_train_chunks: + log0(f" ttt:loading EMA weights permanently at chunk {ci}") + for n, p in base_model.named_parameters(): + if n in ema_state: + p.data.copy_(ema_state[n]) + ema_state = None + raw_state = None + if master and (ci % 5 == 0 or ci == num_chunks - 1): + rl = loss_sum.item() / max(token_count.item(), 1) + cur_bpb = rl / math.log(2) * (token_count.item() / max(byte_count.item(), 1)) if token_count.item() > 0 else 0 + lr_str = f" lr={cur_lr:.6f}" if ci < args.ttt_max_train_chunks else " lr=done" + log0(f" ttt[{ci+1}/{num_chunks}] bpb={cur_bpb:.6f}{lr_str} t={time.perf_counter()-t0:.0f}s") + if dist.is_available() and dist.is_initialized(): + for t in [loss_sum, token_count, byte_count]: + dist.all_reduce(t, op=dist.ReduceOp.SUM) + val_loss = (loss_sum / token_count).item() + val_bpb = val_loss / math.log(2.0) * (token_count.item() / byte_count.item()) + for p in base_model.parameters(): + p.requires_grad_(True) + base_model.eval() + log0(f"ttt_sliding:done loss={val_loss:.6f} bpb={val_bpb:.6f} time={time.perf_counter()-t0:.0f}s") + return val_loss, val_bpb +# --------------------------------------------------------------------------- +# GPTQ: Hessian-aware quantization with column-wise error compensation +# --------------------------------------------------------------------------- +def _find_best_row_scales(W: Tensor, clip_range: int = 31) -> Tensor: + """Find optimal per-row scales by searching percentile clipping thresholds.""" + t32 = W.float() + best_s = t32.abs().amax(dim=1) / clip_range + best_s = best_s.clamp_min(1.0 / clip_range) + best_err = torch.full((t32.shape[0],), float('inf')) + for pct in [0.9990, 0.9995, 0.9999, 0.99999, 1.0]: + if pct < 1.0: + row_clip = torch.quantile(t32.abs(), pct, dim=1) + else: + row_clip = t32.abs().amax(dim=1) + s = (row_clip / clip_range).clamp_min(1.0 / clip_range) + q = torch.clamp(torch.round(t32 / s[:, None]), -clip_range, clip_range) + recon = q * s[:, None] + err = (t32 - recon).pow(2).mean(dim=1) + improved = err < best_err + best_s[improved] = s[improved] + best_err[improved] = err[improved] + return best_s +def gptq_quantize_weight(W: Tensor, H: Tensor, clip_range: int = 31, + block_size: int = 64, percdamp: float = 0.002) -> tuple[Tensor, Tensor]: + """GPTQ: quantize weight matrix W using Hessian H = X^T X for error compensation. + Uses pre-computed per-row scales and column reordering by Hessian diagonal. + Returns (quantized_int8, scale_fp16) in int6 range [-clip_range, clip_range].""" + W = W.float().clone() + rows, cols = W.shape + # Pre-compute optimal per-row scales from the original weight matrix + row_scale = _find_best_row_scales(W, clip_range) + H = H.float().clone() + damp = percdamp * H.diag().mean() + H.diagonal().add_(damp) + # Column reordering: process least-important columns first (ascending H_diag) + perm = torch.argsort(H.diag()) + invperm = torch.argsort(perm) + W = W[:, perm] + H = H[perm][:, perm] + try: + L = torch.linalg.cholesky(H) + Hinv = torch.cholesky_inverse(L) + except torch._C._LinAlgError: + Hinv = torch.diag(1.0 / H.diag().clamp_min(1e-6)) + Q = torch.zeros(rows, cols, dtype=torch.int8) + for i1 in range(0, cols, block_size): + i2 = min(i1 + block_size, cols) + W_block = W[:, i1:i2].clone() + Hinv_block = Hinv[i1:i2, i1:i2] + Err = torch.zeros_like(W_block) + for j in range(i2 - i1): + w_col = W_block[:, j] + h_inv_jj = Hinv_block[j, j].clamp_min(1e-8) + # Quantize using pre-computed per-row scales + q_col = torch.clamp(torch.round(w_col / row_scale), -clip_range, clip_range) + deq_col = q_col * row_scale + Q[:, i1 + j] = q_col.to(torch.int8) + err = (w_col - deq_col) / h_inv_jj + Err[:, j] = err + if j + 1 < i2 - i1: + W_block[:, j + 1:] -= err.unsqueeze(1) * Hinv_block[j, j + 1:].unsqueeze(0) + if i2 < cols: + W[:, i2:] -= Err @ Hinv[i1:i2, i2:] + # Undo column reordering + Q = Q[:, invperm] + return Q, row_scale.to(torch.float16) +def gptq_calibrate(model: nn.Module, train_pattern: str, device: torch.device, + n_samples: int = 256, seq_len: int = 2048) -> dict[str, Tensor]: + """Collect Hessian H = X^T X for each linear layer using training data.""" + hessians: dict[str, Tensor] = {} + n_seen: dict[str, int] = {} + hooks = [] + def make_hook(name: str): + def hook_fn(module, inp, out): + x = inp[0].detach().float() + if x.ndim == 3: + x = x.reshape(-1, x.shape[-1]) + if name not in hessians: + hessians[name] = torch.zeros(x.shape[1], x.shape[1], device=x.device, dtype=torch.float32) + n_seen[name] = 0 + hessians[name].addmm_(x.t(), x) + n_seen[name] += x.shape[0] + return hook_fn + for name, module in model.named_modules(): + if isinstance(module, (nn.Linear, CastedLinear)): + hooks.append(module.register_forward_hook(make_hook(name))) + stream = TokenStream(train_pattern) + model.eval() + with torch.no_grad(): + for _ in range(n_samples): + tokens = stream.take(seq_len + 1).to(device=device, dtype=torch.int64) + x = tokens[:-1].unsqueeze(0) + with torch.autocast(device_type="cuda", dtype=torch.bfloat16): + model.forward_logits(x) + for h in hooks: + h.remove() + for name in hessians: + hessians[name] /= max(n_seen[name], 1) + return hessians +def mixed_quantize_int6_gptq(state_dict: dict[str, Tensor], int6_cats: set[str], + hessians: dict[str, Tensor]) -> tuple[dict, dict]: + """Like mixed_quantize_int6 but uses GPTQ for int6 categories when Hessian available.""" + result: dict[str, Tensor] = {} + meta: dict[str, object] = {} + gptq_count, naive_count = 0, 0 + for name, tensor in state_dict.items(): + t = tensor.detach().cpu().contiguous() + cat = _classify_param(name) + if not t.is_floating_point() or t.numel() <= 65536: + result[name] = t.to(torch.float16) if t.is_floating_point() else t + meta[name] = "passthrough" + continue + if any(p in name for p in CONTROL_TENSOR_NAME_PATTERNS): + result[name] = t.float() + meta[name] = "passthrough_ctrl" + continue + if cat in int6_cats and t.ndim == 2: + module_name = name.rsplit(".weight", 1)[0] if name.endswith(".weight") else name + H = hessians.get(module_name) + if H is not None and H.shape[0] == t.shape[1]: + q, s = gptq_quantize_weight(t, H.cpu()) + gptq_count += 1 + else: + q, s = quantize_int6_per_row(t) + naive_count += 1 + result[name + ".q"] = q + result[name + ".scale"] = s + meta[name] = {"type": "int6"} + elif cat in int6_cats and t.ndim >= 1: + q, s = quantize_int6_per_row(t) + result[name + ".q"] = q + result[name + ".scale"] = s + meta[name] = {"type": "int6"} + naive_count += 1 + else: + q, s = quantize_float_tensor(t) + result[name + ".q"] = q + result[name + ".scale"] = s + meta[name] = {"type": "int8"} + print(f"gptq_quantize: {gptq_count} GPTQ layers, {naive_count} naive layers", flush=True) + return result, meta +def quantize_int6_per_row(t: Tensor, clip_range: int = 31) -> tuple[Tensor, Tensor]: + t32 = t.float() + if t32.ndim == 2: + best_q, best_s, best_err = None, None, float('inf') + for pct in [0.9990, 0.9995, 0.9999, 0.99999, 1.0]: + if pct < 1.0: + row_clip = torch.quantile(t32.abs(), pct, dim=1) + else: + row_clip = t32.abs().amax(dim=1) + s = (row_clip / clip_range).clamp_min(1.0 / clip_range).to(torch.float16) + q = torch.clamp(torch.round(t32 / s.float()[:, None]), -clip_range, clip_range).to(torch.int8) + recon = q.float() * s.float()[:, None] + err = (t32 - recon).pow(2).mean().item() + if err < best_err: + best_q, best_s, best_err = q, s, err + return best_q, best_s + amax = t32.abs().max().item() + scale = torch.tensor(amax / clip_range if amax > 0 else 1.0, dtype=torch.float16) + q = torch.clamp(torch.round(t32 / scale.float()), -clip_range, clip_range).to(torch.int8) + return q, scale +def mixed_quantize_int6(state_dict: dict[str, Tensor], int6_cats: set[str]): + num_layers_total = max( + (int(k.split(".")[1]) for k in state_dict if k.startswith("blocks.")), + default=0, + ) + 1 + late_k_layers = set(range(num_layers_total - 2, num_layers_total)) + result: dict[str, Tensor] = {} + meta: dict[str, object] = {} + for name, tensor in state_dict.items(): + t = tensor.detach().cpu().contiguous() + cat = _classify_param(name) + if not t.is_floating_point() or t.numel() <= 65536: + result[name] = t.to(torch.float16) if t.is_floating_point() else t + meta[name] = "passthrough" + continue + if any(p in name for p in CONTROL_TENSOR_NAME_PATTERNS): + result[name] = t.float() + meta[name] = "passthrough_ctrl" + continue + if cat in int6_cats and t.ndim >= 1: + q, s = quantize_int6_per_row(t) + result[name + ".q"] = q + result[name + ".scale"] = s + meta[name] = {"type": "int6"} + else: + q, s = quantize_float_tensor(t) + result[name + ".q"] = q + result[name + ".scale"] = s + meta[name] = {"type": "int8"} + return result, meta +def dequantize_mixed_int6(result: dict[str, Tensor], meta: dict[str, object], + template_sd: dict[str, Tensor]) -> dict[str, Tensor]: + out: dict[str, Tensor] = {} + for name, orig in template_sd.items(): + info = meta.get(name) + if info is None: + continue + orig_dtype = orig.dtype + if info in ("passthrough", "passthrough_ctrl", "passthrough_fp16"): + t = result[name] + if t.dtype == torch.float16 and orig_dtype in (torch.float32, torch.bfloat16): + t = t.to(orig_dtype) + out[name] = t + continue + q, s = result[name + ".q"], result[name + ".scale"] + if s.ndim > 0: + out[name] = (q.float() * s.float().view(q.shape[0], *([1] * (q.ndim - 1)))).to(orig_dtype) + else: + out[name] = (q.float() * float(s.item())).to(orig_dtype) + return out def main() -> None: global zeropower_via_newtonschulz5 - code = Path(__file__).read_text(encoding="utf-8") args = Hyperparameters() - zeropower_via_newtonschulz5 = torch.compile(zeropower_via_newtonschulz5) - - # ----------------------------- - # DISTRIBUTED + CUDA SETUP - # ----------------------------- - + if args.compile_enabled: + zeropower_via_newtonschulz5 = torch.compile(zeropower_via_newtonschulz5) distributed = "RANK" in os.environ and "WORLD_SIZE" in os.environ rank = int(os.environ.get("RANK", "0")) world_size = int(os.environ.get("WORLD_SIZE", "1")) @@ -757,23 +1512,18 @@ def main() -> None: dist.init_process_group(backend="nccl", device_id=device) dist.barrier() master_process = rank == 0 - - # Fast math knobs torch.backends.cuda.matmul.allow_tf32 = True torch.backends.cudnn.allow_tf32 = True from torch.backends.cuda import enable_cudnn_sdp, enable_flash_sdp, enable_math_sdp, enable_mem_efficient_sdp - enable_cudnn_sdp(False) enable_flash_sdp(True) enable_mem_efficient_sdp(False) enable_math_sdp(False) - logfile = None if master_process: os.makedirs("logs", exist_ok=True) logfile = f"logs/{args.run_id}.txt" print(logfile) - def log0(msg: str, console: bool = True) -> None: if not master_process: return @@ -782,7 +1532,6 @@ def log0(msg: str, console: bool = True) -> None: if logfile is not None: with open(logfile, "a", encoding="utf-8") as f: print(msg, file=f) - log0(code, console=False) log0("=" * 100, console=False) log0(f"Running Python {sys.version}", console=False) @@ -792,16 +1541,10 @@ def log0(msg: str, console: bool = True) -> None: console=False, ) log0("=" * 100, console=False) - - # ----------------------------- - # TOKENIZER + VALIDATION METRIC SETUP - # ----------------------------- - random.seed(args.seed) np.random.seed(args.seed) torch.manual_seed(args.seed) torch.cuda.manual_seed_all(args.seed) - if not args.tokenizer_path.endswith(".model"): raise ValueError(f"Script only setup for SentencePiece .model file: {args.tokenizer_path}") sp = spm.SentencePieceProcessor(model_file=args.tokenizer_path) @@ -811,18 +1554,16 @@ def log0(msg: str, console: bool = True) -> None: ) dataset_dir = Path(args.data_path).resolve() actual_train_files = len(list(dataset_dir.glob("fineweb_train_*.bin"))) - val_tokens = load_validation_tokens(args.val_files, args.train_seq_len) + effective_eval_seq_len = args.eval_seq_len if args.eval_seq_len > 0 else args.train_seq_len + val_seq_len = max(args.train_seq_len, effective_eval_seq_len) + val_tokens = load_validation_tokens(args.val_files, val_seq_len) base_bytes_lut, has_leading_space_lut, is_boundary_token_lut = build_sentencepiece_luts( sp, args.vocab_size, device ) log0(f"val_bpb:enabled tokenizer_kind=sentencepiece tokenizer_path={args.tokenizer_path}") log0(f"train_loader:dataset:{dataset_dir.name} train_shards:{actual_train_files}") log0(f"val_loader:shards pattern={args.val_files} tokens:{val_tokens.numel() - 1}") - - # ----------------------------- - # MODEL + OPTIMIZER SETUP - # ----------------------------- - + CastedLinear._qat_enabled = args.qat_enabled base_model = GPT( vocab_size=args.vocab_size, num_layers=args.num_layers, @@ -835,25 +1576,39 @@ def log0(msg: str, console: bool = True) -> None: logit_softcap=args.logit_softcap, rope_base=args.rope_base, qk_gain_init=args.qk_gain_init, + mtp_num_heads=args.mtp_num_heads, + mtp_loss_weight=args.mtp_loss_weight, + bigram_vocab_size=args.bigram_vocab_size, + bigram_dim=args.bigram_dim, + xsa_last_n=args.xsa_last_n, + rope_dims=args.rope_dims, + ln_scale=args.ln_scale, + dtg=args.dtg_enabled, + ve_enabled=args.ve_enabled, + ve_dim=args.ve_dim, + ve_layers=args.ve_layers, + mlp_act=args.mlp_act, + mlp_leaky_slope=args.mlp_leaky_slope, + f1_corr_rank=args.f1_corr_rank, + f1_corr_scale_init=args.f1_corr_scale_init, ).to(device).bfloat16() for module in base_model.modules(): if isinstance(module, CastedLinear): module.float() restore_low_dim_params_to_fp32(base_model) - compiled_model = torch.compile(base_model, dynamic=False, fullgraph=True) + compiled_model = maybe_torch_compile(base_model, args) model: nn.Module = DDP(compiled_model, device_ids=[local_rank], broadcast_buffers=False) if distributed else compiled_model - - # Optimizer split: - # - token embedding (Adam) uses EMBED_LR - # - untied lm_head (Adam) uses HEAD_LR - # - matrix params in transformer blocks use MATRIX_LR via Muon - # - vectors/scalars use SCALAR_LR via Adam block_named_params = list(base_model.blocks.named_parameters()) matrix_params = [ p for name, p in block_named_params if p.ndim == 2 and not any(pattern in name for pattern in CONTROL_TENSOR_NAME_PATTERNS) ] + if base_model.mtp_num_heads > 0: + matrix_params.extend([p for p in base_model.mtp_heads.parameters() if p.ndim == 2]) + if base_model.f1_corr_in is not None and base_model.f1_corr_out is not None: + matrix_params.append(base_model.f1_corr_in.weight) + matrix_params.append(base_model.f1_corr_out.weight) scalar_params = [ p for name, p in block_named_params @@ -861,11 +1616,29 @@ def log0(msg: str, console: bool = True) -> None: ] if base_model.skip_weights.numel() > 0: scalar_params.append(base_model.skip_weights) + scalar_params.append(base_model.smear.gate) + if base_model.bigram is not None: + scalar_params.append(base_model.bigram.scale) + if base_model.f1_corr_scale is not None: + scalar_params.append(base_model.f1_corr_scale) token_lr = args.tied_embed_lr if args.tie_embeddings else args.embed_lr - optimizer_tok = torch.optim.Adam( - [{"params": [base_model.tok_emb.weight], "lr": token_lr, "base_lr": token_lr}], + tok_params = [{"params": [base_model.tok_emb.weight], "lr": token_lr, "base_lr": token_lr}] + if base_model.bigram is not None: + tok_params.append({"params": [base_model.bigram.embed.weight], "lr": token_lr, "base_lr": token_lr}) + if base_model.bigram.proj is not None: + matrix_params.append(base_model.bigram.proj.weight) + if base_model.ve_shared is not None: + tok_params.append({"params": [base_model.ve_shared.embed.weight], "lr": token_lr, "base_lr": token_lr}) + if base_model.ve_shared.proj is not None: + matrix_params.append(base_model.ve_shared.proj.weight) + scalar_params.append(base_model.ve_shared.scale) + for s in base_model.ve_layer_scales: + scalar_params.append(s) + optimizer_tok = torch.optim.AdamW( + tok_params, betas=(args.beta1, args.beta2), eps=args.adam_eps, + weight_decay=args.adam_wd, fused=True, ) optimizer_muon = Muon( @@ -873,13 +1646,15 @@ def log0(msg: str, console: bool = True) -> None: lr=args.matrix_lr, momentum=args.muon_momentum, backend_steps=args.muon_backend_steps, + weight_decay=args.muon_wd, ) for group in optimizer_muon.param_groups: group["base_lr"] = args.matrix_lr - optimizer_scalar = torch.optim.Adam( + optimizer_scalar = torch.optim.AdamW( [{"params": scalar_params, "lr": args.scalar_lr, "base_lr": args.scalar_lr}], betas=(args.beta1, args.beta2), eps=args.adam_eps, + weight_decay=args.adam_wd, fused=True, ) optimizers: list[torch.optim.Optimizer] = [optimizer_tok, optimizer_muon, optimizer_scalar] @@ -891,36 +1666,42 @@ def log0(msg: str, console: bool = True) -> None: fused=True, ) optimizers.insert(1, optimizer_head) - n_params = sum(p.numel() for p in base_model.parameters()) + f1_corr_params = 0 + if base_model.f1_corr_in is not None and base_model.f1_corr_out is not None: + f1_corr_params = int(base_model.f1_corr_in.weight.numel() + base_model.f1_corr_out.weight.numel()) + est_corr_int6_bytes = 0 + if args.f1_corr_rank > 0: + # int8 payload stores int6 values + per-row fp16 scales. + est_corr_int6_bytes = ( + args.f1_corr_rank * (args.model_dim + args.vocab_size) + + 2 * (args.f1_corr_rank + args.vocab_size) + ) log0(f"model_params:{n_params}") - log0(f"world_size:{world_size} grad_accum_steps:{grad_accum_steps}") - log0("sdp_backends:cudnn=False flash=True mem_efficient=False math=False") - log0(f"attention_mode:gqa num_heads:{args.num_heads} num_kv_heads:{args.num_kv_heads}") log0( - f"tie_embeddings:{args.tie_embeddings} embed_lr:{token_lr} " - f"head_lr:{args.head_lr if base_model.lm_head is not None else 0.0} " - f"matrix_lr:{args.matrix_lr} scalar_lr:{args.scalar_lr}" + f"f1_corr:rank={args.f1_corr_rank} params={f1_corr_params} " + f"est_int6_bytes~{est_corr_int6_bytes}" ) + log0(f"mlp_act:{args.mlp_act} mlp_leaky_slope:{args.mlp_leaky_slope}") + log0(f"XSA:last_{args.xsa_last_n} world_size:{world_size} grad_accum_steps:{grad_accum_steps}") + log0(f"num_heads:{args.num_heads} num_kv_heads:{args.num_kv_heads} embed_lr:{token_lr} matrix_lr:{args.matrix_lr}") log0( f"train_batch_tokens:{args.train_batch_tokens} train_seq_len:{args.train_seq_len} " f"iterations:{args.iterations} warmup_steps:{args.warmup_steps} " f"max_wallclock_seconds:{args.max_wallclock_seconds:.3f}" ) + log0(f"compile:enabled={int(args.compile_enabled)} fullgraph={int(args.compile_fullgraph)}") log0(f"seed:{args.seed}") - - # ----------------------------- - # DATA LOADER & MODEL WARMUP - # ----------------------------- - + if args.ngram_eval_order >= 2: + log0( + f"ngram_eval:order={args.ngram_eval_order} alpha={args.ngram_eval_alpha} " + f"min_count={args.ngram_eval_min_count} buckets={args.ngram_eval_buckets}" + ) train_loader = DistributedTokenLoader(args.train_files, rank, world_size, device) - def zero_grad_all() -> None: for opt in optimizers: opt.zero_grad(set_to_none=True) - max_wallclock_ms = 1000.0 * args.max_wallclock_seconds if args.max_wallclock_seconds > 0 else None - def lr_mul(step: int, elapsed_ms: float) -> float: if args.warmdown_iters <= 0: return 1.0 @@ -931,9 +1712,6 @@ def lr_mul(step: int, elapsed_ms: float) -> float: warmdown_ms = args.warmdown_iters * step_ms remaining_ms = max(max_wallclock_ms - elapsed_ms, 0.0) return remaining_ms / max(warmdown_ms, 1e-9) if remaining_ms <= warmdown_ms else 1.0 - - # Warmup primes the compiled forward/backward/optimizer paths, then we restore the - # initial weights/optimizer state so measured training starts from the true init. if args.warmup_steps > 0: initial_model_state = {name: tensor.detach().cpu().clone() for name, tensor in base_model.state_dict().items()} initial_optimizer_states = [copy.deepcopy(opt.state_dict()) for opt in optimizers] @@ -959,20 +1737,17 @@ def lr_mul(step: int, elapsed_ms: float) -> float: if distributed: model.require_backward_grad_sync = True train_loader = DistributedTokenLoader(args.train_files, rank, world_size, device) - - # ----------------------------- - # MAIN TRAINING LOOP - # ----------------------------- - + swa_state: dict[str, Tensor] | None = None + swa_count = 0 + ema_state = {name: t.detach().float().clone() for name, t in base_model.state_dict().items()} + ema_decay = 0.997 training_time_ms = 0.0 stop_after_step: int | None = None torch.cuda.synchronize() t0 = time.perf_counter() - step = 0 while True: last_step = step == args.iterations or (stop_after_step is not None and step >= stop_after_step) - should_validate = last_step or (args.val_loss_every > 0 and step % args.val_loss_every == 0) if should_validate: torch.cuda.synchronize() @@ -995,7 +1770,6 @@ def lr_mul(step: int, elapsed_ms: float) -> float: ) torch.cuda.synchronize() t0 = time.perf_counter() - if last_step: if stop_after_step is not None and step < args.iterations: log0( @@ -1003,9 +1777,11 @@ def lr_mul(step: int, elapsed_ms: float) -> float: f"step:{step}/{args.iterations}" ) break - elapsed_ms = training_time_ms + 1000.0 * (time.perf_counter() - t0) scale = lr_mul(step, elapsed_ms) + if args.late_qat_threshold > 0 and scale < args.late_qat_threshold and not CastedLinear._qat_enabled: + CastedLinear._qat_enabled = True + log0(f"late_qat:enabled step:{step} scale:{scale:.4f}") zero_grad_all() train_loss = torch.zeros((), device=device) for micro_step in range(grad_accum_steps): @@ -1015,26 +1791,35 @@ def lr_mul(step: int, elapsed_ms: float) -> float: with torch.autocast(device_type="cuda", dtype=torch.bfloat16, enabled=True): loss = model(x, y) train_loss += loss.detach() - (loss * grad_scale).backward() + loss.backward() train_loss /= grad_accum_steps - frac = min(step / args.muon_momentum_warmup_steps, 1.0) if args.muon_momentum_warmup_steps > 0 else 1.0 muon_momentum = (1 - frac) * args.muon_momentum_warmup_start + frac * args.muon_momentum for group in optimizer_muon.param_groups: group["momentum"] = muon_momentum - for opt in optimizers: for group in opt.param_groups: group["lr"] = group["base_lr"] * scale - if args.grad_clip_norm > 0: torch.nn.utils.clip_grad_norm_(base_model.parameters(), args.grad_clip_norm) for opt in optimizers: opt.step() zero_grad_all() - + # EMA update + with torch.no_grad(): + for name, t in base_model.state_dict().items(): + ema_state[name].mul_(ema_decay).add_(t.detach().float(), alpha=1.0 - ema_decay) step += 1 approx_training_time_ms = training_time_ms + 1000.0 * (time.perf_counter() - t0) + if args.swa_enabled and scale < 0.2 and step % args.swa_every == 0: + if swa_state is None: + swa_state = {name: t.detach().cpu().clone() for name, t in base_model.state_dict().items()} + swa_count = 1 + log0(f"swa:start step:{step}") + else: + for name, t in base_model.state_dict().items(): + swa_state[name] += t.detach().cpu() + swa_count += 1 should_log_train = ( args.train_log_every > 0 and (step <= 10 or step % args.train_log_every == 0 or stop_after_step is not None) @@ -1044,8 +1829,6 @@ def lr_mul(step: int, elapsed_ms: float) -> float: f"step:{step}/{args.iterations} train_loss:{train_loss.item():.4f} " f"train_time:{approx_training_time_ms:.0f}ms step_avg:{approx_training_time_ms / step:.2f}ms" ) - - # Needed to sync whether we've reached the wallclock cap. reached_cap = max_wallclock_ms is not None and approx_training_time_ms >= max_wallclock_ms if distributed and max_wallclock_ms is not None: reached_cap_tensor = torch.tensor(int(reached_cap), device=device) @@ -1053,74 +1836,248 @@ def lr_mul(step: int, elapsed_ms: float) -> float: reached_cap = bool(reached_cap_tensor.item()) if stop_after_step is None and reached_cap: stop_after_step = step - log0( f"peak memory allocated: {torch.cuda.max_memory_allocated() // 1024 // 1024} MiB " f"reserved: {torch.cuda.max_memory_reserved() // 1024 // 1024} MiB" ) - - # ----------------------------- - # SERIALIZATION + ROUNDTRIP VALIDATION - # ----------------------------- - # Save the raw state (useful for debugging/loading in PyTorch directly), then always produce - # the compressed int8+zlib artifact and validate the round-tripped weights. - + if args.distill_enabled and args.distill_steps > 0: + log0( + f"distill:start steps:{args.distill_steps} lr_factor:{args.distill_lr_factor} " + f"temp:{args.distill_temperature} alpha:{args.distill_alpha} kl_clip:{args.distill_kl_clip}" + ) + current_state = base_model.state_dict() + teacher_state = {name: t.to(dtype=current_state[name].dtype) for name, t in ema_state.items()} + teacher_model = GPT( + vocab_size=args.vocab_size, num_layers=args.num_layers, model_dim=args.model_dim, + num_heads=args.num_heads, num_kv_heads=args.num_kv_heads, mlp_mult=args.mlp_mult, + tie_embeddings=args.tie_embeddings, tied_embed_init_std=args.tied_embed_init_std, + logit_softcap=args.logit_softcap, rope_base=args.rope_base, qk_gain_init=args.qk_gain_init, + mtp_num_heads=args.mtp_num_heads, mtp_loss_weight=args.mtp_loss_weight, + bigram_vocab_size=args.bigram_vocab_size, bigram_dim=args.bigram_dim, + xsa_last_n=args.xsa_last_n, rope_dims=args.rope_dims, ln_scale=args.ln_scale, dtg=args.dtg_enabled, + ve_enabled=args.ve_enabled, ve_dim=args.ve_dim, ve_layers=args.ve_layers, + mlp_act=args.mlp_act, mlp_leaky_slope=args.mlp_leaky_slope, + f1_corr_rank=args.f1_corr_rank, f1_corr_scale_init=args.f1_corr_scale_init, + ).to(device).bfloat16() + for m in teacher_model.modules(): + if isinstance(m, CastedLinear): + m.float() + restore_low_dim_params_to_fp32(teacher_model) + teacher_model.load_state_dict(teacher_state, strict=True) + teacher_model.eval() + for p in teacher_model.parameters(): + p.requires_grad_(False) + compiled_teacher_logits = maybe_torch_compile(teacher_model.forward_logits, args) + model.train() + T = args.distill_temperature + alpha = args.distill_alpha + for d_step in range(args.distill_steps): + zero_grad_all() + for opt in optimizers: + for group in opt.param_groups: + group["lr"] = group["base_lr"] * args.distill_lr_factor + x, y = train_loader.next_batch(args.train_batch_tokens, args.train_seq_len, grad_accum_steps) + with torch.autocast(device_type="cuda", dtype=torch.bfloat16, enabled=True): + student_logits = base_model.forward_logits(x) + with torch.no_grad(): + teacher_logits = compiled_teacher_logits(x) + student_log_probs = F.log_softmax(student_logits.float() / T, dim=-1) + teacher_probs = F.softmax(teacher_logits.float() / T, dim=-1) + token_kl = F.kl_div(student_log_probs, teacher_probs, reduction="none").sum(dim=-1) + kl_loss = token_kl.mean() * (T * T) + if args.distill_kl_clip > 0: + kl_loss = torch.clamp(kl_loss, max=args.distill_kl_clip) + ce_loss = F.cross_entropy( + student_logits.reshape(-1, student_logits.size(-1)).float(), + y.reshape(-1), + reduction="mean", + ) + loss = alpha * kl_loss + (1.0 - alpha) * ce_loss + (loss * grad_scale).backward() + if world_size > 1: + for p in base_model.parameters(): + if p.grad is not None: + dist.all_reduce(p.grad, op=dist.ReduceOp.AVG) + if args.grad_clip_norm > 0: + torch.nn.utils.clip_grad_norm_(base_model.parameters(), args.grad_clip_norm) + for opt in optimizers: + opt.step() + zero_grad_all() + with torch.no_grad(): + for name, t in base_model.state_dict().items(): + ema_state[name].mul_(ema_decay).add_(t.detach().float(), alpha=1.0 - ema_decay) + if (d_step + 1) % 8 == 0 or d_step == 0: + log0( + f"distill:step:{d_step + 1}/{args.distill_steps} " + f"kl:{kl_loss.item():.4f} ce:{ce_loss.item():.4f} total:{loss.item():.4f}" + ) + del teacher_model, compiled_teacher_logits + torch.cuda.empty_cache() + log0("distill:done") + # Apply EMA weights (better than SWA alone per PR#401) + log0("ema:applying EMA weights") + current_state = base_model.state_dict() + avg_state = {name: t.to(dtype=current_state[name].dtype) for name, t in ema_state.items()} + base_model.load_state_dict(avg_state, strict=True) + torch.cuda.synchronize() + t_diag = time.perf_counter() + diag_val_loss, diag_val_bpb = eval_val( + args, compiled_model, rank, world_size, device, grad_accum_steps, + val_tokens, base_bytes_lut, has_leading_space_lut, is_boundary_token_lut, + ) + torch.cuda.synchronize() + log0( + f"DIAGNOSTIC post_ema val_loss:{diag_val_loss:.4f} val_bpb:{diag_val_bpb:.4f} " + f"eval_time:{1000.0 * (time.perf_counter() - t_diag):.0f}ms" + ) + full_state_dict = base_model.state_dict() + export_sd = {k: v for k, v in full_state_dict.items() if "mtp_heads" not in k} + excluded_mtp = sum(int(t.numel()) for k, t in full_state_dict.items() if "mtp_heads" in k) + if excluded_mtp > 0: + log0(f"export_excluding_mtp_params:{excluded_mtp}") if master_process: - torch.save(base_model.state_dict(), "final_model.pt") + torch.save(export_sd, "final_model.pt") model_bytes = os.path.getsize("final_model.pt") code_bytes = len(code.encode("utf-8")) log0(f"Serialized model: {model_bytes} bytes") log0(f"Code size: {code_bytes} bytes") - log0(f"Total submission size: {model_bytes + code_bytes} bytes") - - quant_obj, quant_stats = quantize_state_dict_int8(base_model.state_dict()) + sd_cpu = {k: v.detach().cpu() for k, v in export_sd.items()} + # GPTQ calibration: collect Hessians from training data + log0("gptq:calibrating with training data...") + t_gptq = time.perf_counter() + gptq_hessians = gptq_calibrate(base_model, args.train_files, device, n_samples=256, seq_len=args.train_seq_len) + log0(f"gptq:calibrated {len(gptq_hessians)} layers in {time.perf_counter()-t_gptq:.1f}s") + quant_result, quant_meta = mixed_quantize_int6_gptq(sd_cpu, {"mlp", "attn", "aux"}, gptq_hessians) quant_buf = io.BytesIO() - torch.save(quant_obj, quant_buf) + torch.save({"w": quant_result, "m": quant_meta}, quant_buf) quant_raw = quant_buf.getvalue() - quant_blob = zlib.compress(quant_raw, level=9) - quant_raw_bytes = len(quant_raw) + quant_blob = zstandard.ZstdCompressor(level=22).compress(quant_raw) if _COMPRESSOR == "zstd" else zlib.compress(quant_raw, 9) if master_process: - with open("final_model.int8.ptz", "wb") as f: + with open("final_model.int6.ptz", "wb") as f: f.write(quant_blob) - quant_file_bytes = os.path.getsize("final_model.int8.ptz") + quant_file_bytes = len(quant_blob) code_bytes = len(code.encode("utf-8")) - ratio = quant_stats["baseline_tensor_bytes"] / max(quant_stats["int8_payload_bytes"], 1) - log0( - f"Serialized model int8+zlib: {quant_file_bytes} bytes " - f"(payload:{quant_stats['int8_payload_bytes']} raw_torch:{quant_raw_bytes} payload_ratio:{ratio:.2f}x)" - ) + log0(f"Serialized model int6+{_COMPRESSOR}: {quant_file_bytes} bytes") + log0(f"Total submission size int6+{_COMPRESSOR}: {quant_file_bytes + code_bytes} bytes") log0(f"Total submission size int8+zlib: {quant_file_bytes + code_bytes} bytes") - if distributed: dist.barrier() - with open("final_model.int8.ptz", "rb") as f: + with open("final_model.int6.ptz", "rb") as f: quant_blob_disk = f.read() - quant_state = torch.load(io.BytesIO(zlib.decompress(quant_blob_disk)), map_location="cpu") - base_model.load_state_dict(dequantize_state_dict_int8(quant_state), strict=True) + quant_state = torch.load( + io.BytesIO(zstandard.ZstdDecompressor().decompress(quant_blob_disk) if _COMPRESSOR == "zstd" else zlib.decompress(quant_blob_disk)), + map_location="cpu", + ) + deq_state = dequantize_mixed_int6(quant_state["w"], quant_state["m"], sd_cpu) + eval_model = GPT( + vocab_size=args.vocab_size, num_layers=args.num_layers, model_dim=args.model_dim, + num_heads=args.num_heads, num_kv_heads=args.num_kv_heads, mlp_mult=args.mlp_mult, + tie_embeddings=args.tie_embeddings, tied_embed_init_std=args.tied_embed_init_std, + logit_softcap=args.logit_softcap, rope_base=args.rope_base, qk_gain_init=args.qk_gain_init, + mtp_num_heads=0, mtp_loss_weight=0.0, + bigram_vocab_size=args.bigram_vocab_size, bigram_dim=args.bigram_dim, + xsa_last_n=args.xsa_last_n, # must match training model + rope_dims=args.rope_dims, ln_scale=args.ln_scale, dtg=args.dtg_enabled, + ve_enabled=args.ve_enabled, ve_dim=args.ve_dim, ve_layers=args.ve_layers, + mlp_act=args.mlp_act, mlp_leaky_slope=args.mlp_leaky_slope, + f1_corr_rank=args.f1_corr_rank, f1_corr_scale_init=args.f1_corr_scale_init, + ).to(device).bfloat16() + for m in eval_model.modules(): + if isinstance(m, CastedLinear): + m.float() + restore_low_dim_params_to_fp32(eval_model) + eval_model.load_state_dict(deq_state, strict=True) + compiled_eval = maybe_torch_compile(eval_model, args) torch.cuda.synchronize() t_qeval = time.perf_counter() q_val_loss, q_val_bpb = eval_val( - args, - model, - rank, - world_size, - device, - grad_accum_steps, - val_tokens, - base_bytes_lut, - has_leading_space_lut, - is_boundary_token_lut, + args, compiled_eval, rank, world_size, device, grad_accum_steps, + val_tokens, base_bytes_lut, has_leading_space_lut, is_boundary_token_lut, + eval_seq_len=effective_eval_seq_len, ) torch.cuda.synchronize() log0( - f"final_int8_zlib_roundtrip val_loss:{q_val_loss:.4f} val_bpb:{q_val_bpb:.4f} " + f"final_int6_roundtrip val_loss:{q_val_loss:.4f} val_bpb:{q_val_bpb:.4f} " f"eval_time:{1000.0 * (time.perf_counter() - t_qeval):.0f}ms" ) - log0(f"final_int8_zlib_roundtrip_exact val_loss:{q_val_loss:.8f} val_bpb:{q_val_bpb:.8f}") - + log0(f"final_int6_roundtrip_exact val_loss:{q_val_loss:.8f} val_bpb:{q_val_bpb:.8f}") + sw_seq_len = effective_eval_seq_len + if args.eval_stride > 0 and args.eval_stride < sw_seq_len: + torch.cuda.synchronize() + t_slide = time.perf_counter() + sw_val_loss, sw_val_bpb = eval_val_sliding( + args, eval_model, rank, world_size, device, + val_tokens, base_bytes_lut, has_leading_space_lut, is_boundary_token_lut, + stride=args.eval_stride, + eval_seq_len=sw_seq_len, + ) + torch.cuda.synchronize() + log0( + f"final_int6_sliding_window val_loss:{sw_val_loss:.4f} val_bpb:{sw_val_bpb:.4f} " + f"stride:{args.eval_stride} eval_time:{1000.0 * (time.perf_counter() - t_slide):.0f}ms" + ) + log0(f"final_int6_sliding_window_exact val_loss:{sw_val_loss:.8f} val_bpb:{sw_val_bpb:.8f}") + log0(f"final_int8_zlib_roundtrip_exact val_loss:{sw_val_loss:.8f} val_bpb:{sw_val_bpb:.8f}") + if args.ngram_eval_order >= 2: + if distributed: + dist.barrier() + torch.cuda.synchronize() + t_ng = time.perf_counter() + ng_loss, ng_bpb, ng_coverage = eval_val_sliding_hashed_ngram( + args, + eval_model, + rank, + world_size, + device, + val_tokens, + base_bytes_lut, + has_leading_space_lut, + is_boundary_token_lut, + stride=args.eval_stride, + order=args.ngram_eval_order, + alpha=args.ngram_eval_alpha, + min_count=args.ngram_eval_min_count, + buckets=args.ngram_eval_buckets, + max_seconds=args.ngram_eval_max_seconds, + eval_seq_len=sw_seq_len, + ) + if rank == 0: + torch.cuda.synchronize() + ng_eval_ms = 1000.0 * (time.perf_counter() - t_ng) + if ng_coverage >= 0.999999: + log0( + f"final_int6_sliding_window_ngram{args.ngram_eval_order} val_loss:{ng_loss:.4f} " + f"val_bpb:{ng_bpb:.4f} eval_time:{ng_eval_ms:.0f}ms" + ) + log0( + f"final_int6_sliding_window_ngram{args.ngram_eval_order}_exact " + f"val_loss:{ng_loss:.8f} val_bpb:{ng_bpb:.8f}" + ) + else: + log0( + f"final_int6_sliding_window_ngram{args.ngram_eval_order}_partial val_loss:{ng_loss:.4f} " + f"val_bpb:{ng_bpb:.4f} coverage:{ng_coverage:.4f} eval_time:{ng_eval_ms:.0f}ms" + ) + log0( + f"final_int6_sliding_window_ngram{args.ngram_eval_order}_partial_exact " + f"val_loss:{ng_loss:.8f} val_bpb:{ng_bpb:.8f} coverage:{ng_coverage:.8f}" + ) + if distributed: + dist.barrier() + # Legal score-first TTT eval + if args.ttt_eval_enabled: + torch.cuda.synchronize() + t_ttt = time.perf_counter() + ttt_loss, ttt_bpb = eval_val_sliding_ttt( + args, eval_model, rank, world_size, device, + val_tokens, base_bytes_lut, has_leading_space_lut, is_boundary_token_lut, + stride=args.eval_stride, + ) + torch.cuda.synchronize() + log0(f"legal_ttt val_loss:{ttt_loss:.4f} val_bpb:{ttt_bpb:.4f} " + f"eval_time:{1000.0 * (time.perf_counter() - t_ttt):.0f}ms") + log0(f"legal_ttt_exact val_loss:{ttt_loss:.8f} val_bpb:{ttt_bpb:.8f}") if distributed: dist.destroy_process_group() - - if __name__ == "__main__": main()