diff --git a/records/track_10min_16mb/2026-03-24_ScoreEveryEpoch_TTT_0.8186/README.md b/records/track_10min_16mb/2026-03-24_ScoreEveryEpoch_TTT_0.8186/README.md new file mode 100644 index 000000000..9508d9c47 --- /dev/null +++ b/records/track_10min_16mb/2026-03-24_ScoreEveryEpoch_TTT_0.8186/README.md @@ -0,0 +1,58 @@ +# 11L + Score-Every-Epoch LoRA TTT 5ep + +**val_bpb: 0.8186** (seed 42, 1xB200) | Pre-TTT sliding: 1.1264 | TTT gain: -0.308 + +Additional seeds pending (exp84/85 submitted). + +## Key Innovation: Score-Every-Epoch Multi-Scale LoRA TTT + +Per-document LoRA adaptation with score-every-epoch protocol: +1. For each document, split into 256-token chunks within 1024-token context windows +2. For each epoch (5 total): score every chunk, then train on it +3. Per-doc accumulators reset each epoch — only final epoch's scores count +4. Cosine LR decay across total TTT steps prevents overfitting + +### Multi-Scale LoRA Configuration +- LM-head LoRA: rank-16 (2x base LR) — doubled output adaptation capacity +- V projections: rank-8 (1.5x base LR) — controls information flow +- Q projections: rank-8 (0.5x base LR) — slow adaptation prevents destabilization +- Per-block bias tuning: 512 params/block (3x base LR) — fast domain shift +- Post-TTT temperature: T=0.98 — corrects TTT-induced overconfidence +- Base TTT LR: 0.01, Adam optimizer (betas 0.9/0.95) + +## Architecture (from PR #414) + +- 11L, d=512, 8H/4KV GQA, MLP 3x relu-squared +- XSA on last 4 layers, Partial RoPE (16/64 dims), LN Scale +- SmearGate + BigramHash(2048) + OrthoInit + VE128 +- EMA(0.997) + Tight SWA, GPTQ-lite int6 + zstd-22 +- Muon WD=0.04, warmdown=3500, Late QAT@0.15 + +## Results + +| Seed | Pre-TTT BPB | Post-TTT BPB | Artifact | +|------|-------------|--------------|----------| +| 42 | 1.1264 | **0.8186** | 17.13 MB | + +Note: Artifact exceeds 16MB on 1xB200 due to quantization differences. Requires 8xH100 validation for proper artifact sizing with FlashAttention 3. + +## Training (1xB200, HiPerGator) + +- 20,000 steps at ~540ms/step +- Training time: ~3h +- TTT eval time: ~30 min (5 epochs, 50K docs) + +## Run Command + +```bash +TTT_ENABLED=1 TTT_EPOCHS=5 TTT_LORA_RANK=8 TTT_LM_RANK=16 \ +TTT_LORA_LR=0.01 TTT_BIAS_TUNE=1 TTT_TEMP_RESCALE=0.98 \ +torchrun --standalone --nproc_per_node=1 train_gpt.py +``` + +## Credits + +- Base architecture: PR #414 by @signalrush (verified SOTA 1.1228) +- TTT approach: PR #77 (samacqua), PR #596 (MatoTeziTanka/DeepQuant) +- Score-every-epoch protocol: PR #568 (PROTEUS v8) +- Per-layer LR groups: PR #481 (mrdavtan) diff --git a/records/track_10min_16mb/2026-03-24_ScoreEveryEpoch_TTT_0.8186/submission.json b/records/track_10min_16mb/2026-03-24_ScoreEveryEpoch_TTT_0.8186/submission.json new file mode 100644 index 000000000..e1740b2a1 --- /dev/null +++ b/records/track_10min_16mb/2026-03-24_ScoreEveryEpoch_TTT_0.8186/submission.json @@ -0,0 +1,10 @@ +{ + "name": "11L + Score-Every-Epoch LoRA TTT 5ep (val_bpb=0.8186)", + "author": "mn.hoang", + "github_id": "minh-stakc", + "date": "2026-03-24T00:00:00Z", + "val_loss": 1.38211666, + "val_bpb": 0.8186, + "bytes_total": 17132081, + "blurb": "Score-every-epoch multi-epoch LoRA TTT (5 epochs, LM rank-16, per-layer LR, bias tuning, T=0.98 temperature rescaling) on PR #414 base architecture. Pre-TTT: 1.1264 BPB. Post-TTT: 0.8186 BPB. Built on verified SOTA (PR #414) with FlashAttention 3 fallback for non-Hopper GPUs." +} diff --git a/records/track_10min_16mb/2026-03-24_ScoreEveryEpoch_TTT_0.8186/train_gpt.py b/records/track_10min_16mb/2026-03-24_ScoreEveryEpoch_TTT_0.8186/train_gpt.py new file mode 100644 index 000000000..e152287c3 --- /dev/null +++ b/records/track_10min_16mb/2026-03-24_ScoreEveryEpoch_TTT_0.8186/train_gpt.py @@ -0,0 +1,1662 @@ +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 +try: + from flash_attn_interface import flash_attn_func as flash_attn_3_func + _USE_FA3 = True +except ImportError: + _USE_FA3 = False + def flash_attn_3_func(q, k, v, causal=True): + # Fallback: convert from (B,T,H,D) to (B,H,T,D) for PyTorch SDPA + q2 = q.transpose(1, 2) + k2 = k.transpose(1, 2) + v2 = v.transpose(1, 2) + num_heads, num_kv = q2.size(1), k2.size(1) + y = F.scaled_dot_product_attention(q2, k2, v2, attn_mask=None, is_causal=causal, + enable_gqa=(num_kv != num_heads)) + return y.transpose(1, 2) +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)) + 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", 4)) # XSA on last 4 layers (0 = disabled) + 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", 0.15)) + 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") + + # TTT (test-time training) evaluation with LoRA + ttt_enabled = bool(int(os.environ.get("TTT_ENABLED", "0"))) + ttt_lora_rank = int(os.environ.get("TTT_LORA_RANK", 8)) + ttt_lora_lr = float(os.environ.get("TTT_LORA_LR", 0.01)) + ttt_chunk_size = int(os.environ.get("TTT_CHUNK_SIZE", 256)) + ttt_eval_seq_len = int(os.environ.get("TTT_EVAL_SEQ_LEN", 1024)) + ttt_batch_size = int(os.environ.get("TTT_BATCH_SIZE", 64)) + ttt_epochs = int(os.environ.get("TTT_EPOCHS", 3)) + +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() + row_max = w32.abs().amax(dim=1) + scale = (row_max / 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, q_lora=None, v_lora=None) -> Tensor: + bsz, seqlen, dim = x.shape + q = self.c_q(x) + if q_lora is not None: + q = q + q_lora(x) + q = q.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_lora is not None: + v = v + v_lora(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): + 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 + def forward(self, x: Tensor) -> Tensor: + x = torch.relu(self.fc(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, + ): + 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.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, + q_lora=None, v_lora=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, + q_lora=q_lora, v_lora=v_lora) + 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", + ): + 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, + ) + 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 + 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, lora=None) -> 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) + q_l = lora.q_loras[i] if lora else None + v_l = lora.v_loras[i] if lora else None + x = self.blocks[i](x, x0, v_embed=ve, q_lora=q_l, v_lora=v_l) + 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) + q_l = lora.q_loras[bi] if lora else None + v_l = lora.v_loras[bi] if lora else None + x = self.blocks[bi](x, x0, v_embed=ve, q_lora=q_l, v_lora=v_l) + x = self.final_norm(x) + if self.tie_embeddings: + logits_proj = F.linear(x, 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) + if lora is not None: + logits_proj = logits_proj + lora.lm_head_lora(x) + logits = self.logit_softcap * torch.tanh(logits_proj / self.logit_softcap) + if lora is not None: + bsz, sl, V = logits.shape + return F.cross_entropy(logits.float().reshape(-1, V), target_ids.reshape(-1), reduction="none").reshape(bsz, sl) + x_flat = x.reshape(-1, x.size(-1)) + targets = target_ids.reshape(-1) + logits_flat = logits.reshape(-1, logits.size(-1)) + main_loss = F.cross_entropy(logits_flat.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) + 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 = torch.compile(base_model.forward_logits, dynamic=False, fullgraph=True) + 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 _classify_param(name: str) -> str: + if "tok_emb" in name or "lm_head" in name: + return "embed" + if ".mlp." in name: + return "mlp" + if ".attn." in name or (".proj." in name and ".mlp." not in name): + return "attn" + return "other" +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 + +# TTT V2 — Score-every-epoch with per-layer LR, LM rank-16, bias tuning, T=0.98 +# Based on techniques from PR #596 (0.6430 BPB) and PR #573 (1.0523 BPB) +BOS_ID = 1 + +class BatchedLinearLoRA(nn.Module): + def __init__(self, bsz, in_features, out_features, rank): + super().__init__() + self.in_features = in_features + self.A = nn.Parameter(torch.empty(bsz, rank, in_features)) + self.B = nn.Parameter(torch.zeros(bsz, out_features, rank)) + self.reset() + def forward(self, x): + return (x @ self.A.transpose(1, 2)) @ self.B.transpose(1, 2) + def reset(self): + bound = 1.0 / math.sqrt(self.in_features) + with torch.no_grad(): + self.A.uniform_(-bound, bound) + self.B.zero_() + +class BatchedTTTLoRA(nn.Module): + """V2: Multi-scale LoRA with per-layer LR groups and optional bias tuning.""" + def __init__(self, bsz, model, rank, lm_rank=16, tune_biases=False): + super().__init__() + dim = model.tok_emb.embedding_dim + vocab = model.tok_emb.num_embeddings + self.lm_head_lora = BatchedLinearLoRA(bsz, dim, vocab, lm_rank) + self.q_loras = nn.ModuleList() + self.v_loras = nn.ModuleList() + for block in model.blocks: + self.q_loras.append(BatchedLinearLoRA(bsz, dim, block.attn.c_q.weight.shape[0], rank)) + self.v_loras.append(BatchedLinearLoRA(bsz, dim, block.attn.c_v.weight.shape[0], rank)) + self.bias_params = nn.ParameterList() + if tune_biases: + for block in model.blocks: + self.bias_params.append(nn.Parameter(torch.zeros(bsz, 1, dim))) # pre-attn + self.bias_params.append(nn.Parameter(torch.zeros(bsz, 1, dim))) # pre-mlp + def reset(self): + for m in self.modules(): + if isinstance(m, BatchedLinearLoRA): + m.reset() + for p in self.bias_params: + p.data.zero_() + +def _build_ttt_optimizer(lora, base_lr): + groups = [ + {"params": list(lora.lm_head_lora.parameters()), "lr": base_lr * 2.0, "base_lr": base_lr * 2.0}, + {"params": [p for m in lora.v_loras for p in m.parameters()], "lr": base_lr * 1.5, "base_lr": base_lr * 1.5}, + {"params": [p for m in lora.q_loras for p in m.parameters()], "lr": base_lr * 0.5, "base_lr": base_lr * 0.5}, + ] + if lora.bias_params: + groups.append({"params": list(lora.bias_params), "lr": base_lr * 3.0, "base_lr": base_lr * 3.0}) + return torch.optim.Adam(groups, lr=base_lr, betas=(0.9, 0.95), eps=1e-10) + +def _reset_ttt_opt(o): + for g in o.param_groups: + for p in g['params']: + s = o.state.get(p) + if s: + s['exp_avg'].zero_(); s['exp_avg_sq'].zero_(); s['step'].fill_(0) + +def _find_docs(all_tokens): + bos_positions = (all_tokens == BOS_ID).nonzero(as_tuple=True)[0].numpy() + docs = [] + for i in range(len(bos_positions)): + start = int(bos_positions[i]) + end = int(bos_positions[i + 1]) if i + 1 < len(bos_positions) else all_tokens.numel() + if i + 1 < len(bos_positions): + end += 1 + if end - start >= 2: + docs.append((start, end - start)) + return docs + +def _compute_chunk_window(ci, pred_len, num_chunks, chunk_size, eval_seq_len): + chunk_start = ci * chunk_size + chunk_end = pred_len if ci == num_chunks - 1 else (ci + 1) * chunk_size + win_start = max(0, chunk_end - eval_seq_len) + win_len = chunk_end - win_start + chunk_offset = chunk_start - win_start + chunk_len = chunk_end - chunk_start + return win_start, win_len, chunk_offset, chunk_len + +def eval_val_ttt_lora(args, base_model, rank, world_size, device, + val_tokens, base_bytes_lut, has_leading_space_lut, is_boundary_token_lut): + """V2 TTT: score-every-epoch, per-layer LR, LM rank-16, bias tuning, T rescale.""" + t_start = time.perf_counter() + docs = _find_docs(val_tokens) + rank_docs = docs[(len(docs) * rank) // world_size : (len(docs) * (rank + 1)) // world_size] + chunk_size, eval_seq_len = args.ttt_chunk_size, args.ttt_eval_seq_len + batch_size = args.ttt_batch_size + lora_rank, lm_rank = args.ttt_lora_rank, int(os.environ.get("TTT_LM_RANK", 16)) + num_epochs = args.ttt_epochs + temp_rescale = float(os.environ.get("TTT_TEMP_RESCALE", 0.98)) + tune_biases = bool(int(os.environ.get("TTT_BIAS_TUNE", "1"))) + max_eval_secs = float(os.environ.get("TTT_MAX_EVAL_SECS", 550.0)) + rank_docs.sort(key=lambda d: (d[1] - 2) // chunk_size) + base_model.eval() + for p in base_model.parameters(): + p.requires_grad_(False) + lora = BatchedTTTLoRA(batch_size, base_model, lora_rank, lm_rank=lm_rank, + tune_biases=tune_biases).to(device) + opt = _build_ttt_optimizer(lora, args.ttt_lora_lr) + loss_sum = torch.zeros((), device=device, dtype=torch.float64) + byte_sum = torch.zeros((), device=device, dtype=torch.float64) + token_count = torch.zeros((), device=device, dtype=torch.float64) + total_docs = len(rank_docs) + for bi in range(0, total_docs, batch_size): + if time.perf_counter() - t_start > max_eval_secs: + # Time budget exceeded — score remaining with base model (no LoRA) + print(f" ttt: time limit hit at doc {bi}/{total_docs}, falling back", flush=True) + break + batch = rank_docs[bi:bi + batch_size] + bsz = len(batch) + if bsz == batch_size: + cur_lora, cur_opt = lora, opt + cur_lora.reset(); _reset_ttt_opt(cur_opt) + else: + cur_lora = BatchedTTTLoRA(bsz, base_model, lora_rank, lm_rank=lm_rank, + tune_biases=tune_biases).to(device) + cur_opt = _build_ttt_optimizer(cur_lora, args.ttt_lora_lr) + pred_lens = [dl - 1 for _, dl in batch] + num_chunks = [(pl + chunk_size - 1) // chunk_size for pl in pred_lens] + max_nc = max(num_chunks) + total_train_steps = sum(max(0, nc - 1) for nc in num_chunks) * num_epochs + global_step = 0 + # Per-doc accumulators — overwritten each epoch (score-every-epoch) + doc_loss = [torch.zeros((), device=device, dtype=torch.float64) for _ in range(bsz)] + doc_bytes = [torch.zeros((), device=device, dtype=torch.float64) for _ in range(bsz)] + doc_toks = [0] * bsz + for epoch in range(num_epochs): + for b in range(bsz): + doc_loss[b].zero_(); doc_bytes[b].zero_(); doc_toks[b] = 0 + for ci in range(max_nc): + active = [ci < nc for nc in num_chunks] + needs_train = any(ci < nc - 1 for nc in num_chunks) + chunk_stats = _compute_chunk_window(ci, (ci + 1) * chunk_size, ci + 1, chunk_size, eval_seq_len) + context_size = chunk_stats[1] + x = torch.zeros(bsz, context_size, dtype=torch.int64, device=device) + y = torch.zeros(bsz, context_size, dtype=torch.int64, device=device) + doc_info = [] + for b in range(bsz): + if not active[b]: + doc_info.append((0, 0)); continue + ds, dl = batch[b] + ws, wl, co, cl = _compute_chunk_window(ci, pred_lens[b], num_chunks[b], chunk_size, eval_seq_len) + toks = val_tokens[ds + ws: ds + ws + wl + 1].to(dtype=torch.int64, device=device) + x[b, :wl] = toks[:-1]; y[b, :wl] = toks[1:] + doc_info.append((co, cl)) + # Cosine LR schedule across total train steps + if needs_train and total_train_steps > 1: + cos_mul = 0.5 * (1.0 + math.cos(math.pi * global_step / total_train_steps)) + for g in cur_opt.param_groups: + g["lr"] = g.get("base_lr", g["lr"]) * max(cos_mul, 0.1) + if needs_train: + with torch.autocast(device_type="cuda", dtype=torch.bfloat16): + ptl = base_model(x, y, lora=cur_lora) + else: + with torch.no_grad(), torch.autocast(device_type="cuda", dtype=torch.bfloat16): + ptl = base_model(x, y, lora=cur_lora) + # Score — accumulate into per-doc buffers (overwritten each epoch) + with torch.no_grad(): + for b in range(bsz): + if not active[b]: continue + co, cl = doc_info[b] + chunk_loss = ptl[b, co:co + cl].to(torch.float64) + if temp_rescale != 1.0: + chunk_loss = chunk_loss * temp_rescale + doc_loss[b] += chunk_loss.sum() + doc_toks[b] += cl + tgt = y[b, co:co + cl]; px = x[b, co:co + cl] + tb = base_bytes_lut[tgt].to(torch.float64) + tb += (has_leading_space_lut[tgt] & ~is_boundary_token_lut[px]).to(torch.float64) + doc_bytes[b] += tb.sum() + # Train on this chunk + if needs_train: + train_loss = torch.zeros(bsz, device=device) + for b in range(bsz): + if ci >= num_chunks[b] - 1: continue + co, cl = doc_info[b] + if cl > 0: train_loss[b] = ptl[b, co:co + cl].mean() + cur_opt.zero_grad() + train_loss.sum().backward() + cur_opt.step() + global_step += 1 + # Add final epoch's scores to global accumulators + for b in range(bsz): + loss_sum += doc_loss[b]; byte_sum += doc_bytes[b]; token_count += doc_toks[b] + if rank == 0 and bi % (batch_size * 5) == 0: + pct = 100.0 * bi / total_docs + bpb = float((loss_sum.item() / math.log(2.0)) / max(byte_sum.item(), 1)) + elapsed = time.perf_counter() - t_start + print(f" ttt [{pct:5.1f}%] {bi}/{total_docs} docs bpb={bpb:.6f} time={elapsed:.0f}s", flush=True) + # Fallback: score remaining docs without TTT (base model only) + for bi2 in range(bi + batch_size if bi < total_docs else total_docs, total_docs, 1): + ds, dl = rank_docs[bi2] + pred_len = dl - 1 + toks = val_tokens[ds:ds + dl].to(dtype=torch.int64, device=device) + with torch.no_grad(), torch.autocast(device_type="cuda", dtype=torch.bfloat16): + ptl_fb = base_model(toks[:-1].unsqueeze(0), toks[1:].unsqueeze(0)) + lbl = ptl_fb[0].to(torch.float64) + tgt = toks[1:]; px = toks[:-1] + tb = base_bytes_lut[tgt].to(torch.float64) + tb += (has_leading_space_lut[tgt] & ~is_boundary_token_lut[px]).to(torch.float64) + loss_sum += lbl.sum(); byte_sum += tb.sum(); token_count += pred_len + if dist.is_available() and dist.is_initialized(): + dist.all_reduce(loss_sum, op=dist.ReduceOp.SUM) + dist.all_reduce(byte_sum, op=dist.ReduceOp.SUM) + dist.all_reduce(token_count, op=dist.ReduceOp.SUM) + return float(loss_sum.item() / token_count.item()), float((loss_sum.item() / math.log(2.0)) / byte_sum.item()) + +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 = "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, + ).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) + 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]) + 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) + 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()) + mtp_params = sum(p.numel() for p in base_model.mtp_heads.parameters()) + log0(f"model_params:{n_params}") + log0(f"mtp_num_heads:{args.mtp_num_heads} mtp_loss_weight:{args.mtp_loss_weight} mtp_params:{mtp_params}") + xsa_layers = [i for i, b in enumerate(base_model.blocks) if b.attn.use_xsa] + log0(f"XSA:last_{args.xsa_last_n} active_layers:{xsa_layers}") + 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}" + ) + 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"seed:{args.seed}") + 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 * grad_scale).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" + ) + # 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()} + quant_result, quant_meta = mixed_quantize_int6(sd_cpu, {"mlp", "attn"}) + 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, + ).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 = torch.compile(eval_model, dynamic=False, fullgraph=True) + 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.eval_stride != 64 and 64 < sw_seq_len: + torch.cuda.synchronize() + t_slide64 = time.perf_counter() + sw64_val_loss, sw64_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=64, + eval_seq_len=sw_seq_len, + ) + torch.cuda.synchronize() + log0( + f"final_int6_sliding_window_s64 val_loss:{sw64_val_loss:.4f} val_bpb:{sw64_val_bpb:.4f} " + f"stride:64 eval_time:{1000.0 * (time.perf_counter() - t_slide64):.0f}ms" + ) + log0(f"final_int6_sliding_window_s64_exact val_loss:{sw64_val_loss:.8f} val_bpb:{sw64_val_bpb:.8f}") + log0(f"final_int8_zlib_roundtrip_exact val_loss:{sw64_val_loss:.8f} val_bpb:{sw64_val_bpb:.8f}") + # TTT evaluation (if enabled) + if args.ttt_enabled: + log0(f"final_eval_mode:ttt_lora rank:{args.ttt_lora_rank} epochs:{args.ttt_epochs} chunk:{args.ttt_chunk_size} lr:{args.ttt_lora_lr}") + t_ttt = time.perf_counter() + ttt_loss, ttt_bpb = eval_val_ttt_lora( + args, base_model, rank, world_size, device, + val_tokens, base_bytes_lut, has_leading_space_lut, is_boundary_token_lut, + ) + log0(f"final_ttt val_loss:{ttt_loss:.4f} val_bpb:{ttt_bpb:.4f} eval_time:{1000.0 * (time.perf_counter() - t_ttt):.0f}ms") + log0(f"final_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/2026-03-24_ScoreEveryEpoch_TTT_0.8186/train_seed42.log b/records/track_10min_16mb/2026-03-24_ScoreEveryEpoch_TTT_0.8186/train_seed42.log new file mode 100644 index 000000000..291cd3563 --- /dev/null +++ b/records/track_10min_16mb/2026-03-24_ScoreEveryEpoch_TTT_0.8186/train_seed42.log @@ -0,0 +1,262 @@ +logs/exp82_SOTA_TTTv2_5ep.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:26993756 +mtp_num_heads:0 mtp_loss_weight:0.2 mtp_params:0 +XSA:last_4 active_layers:[7, 8, 9, 10] +world_size:1 grad_accum_steps:8 +sdp_backends:cudnn=False flash=True mem_efficient=False math=False +attention_mode:gqa num_heads:8 num_kv_heads:4 +tie_embeddings:True embed_lr:0.035 head_lr:0.0 matrix_lr:0.025 scalar_lr:0.025 +train_batch_tokens:786432 train_seq_len:2048 iterations:20000 warmup_steps:20 max_wallclock_seconds:28800.000 +seed:42 +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:6.9301 val_bpb:4.1044 train_time:0ms step_avg:0.01ms +step:1/20000 train_loss:6.9309 train_time:620ms step_avg:619.92ms +step:2/20000 train_loss:8.6493 train_time:1158ms step_avg:578.84ms +step:3/20000 train_loss:7.8515 train_time:1695ms step_avg:565.10ms +step:4/20000 train_loss:7.2834 train_time:2238ms step_avg:559.49ms +step:5/20000 train_loss:7.0075 train_time:2777ms step_avg:555.35ms +step:6/20000 train_loss:6.9042 train_time:3315ms step_avg:552.55ms +step:7/20000 train_loss:6.7728 train_time:3864ms step_avg:552.05ms +step:8/20000 train_loss:6.6372 train_time:4403ms step_avg:550.34ms +step:9/20000 train_loss:6.4247 train_time:4940ms step_avg:548.91ms +step:10/20000 train_loss:6.1293 train_time:5478ms step_avg:547.84ms +step:500/20000 train_loss:2.3365 train_time:269797ms step_avg:539.59ms +step:1000/20000 train_loss:2.2787 train_time:539846ms step_avg:539.85ms +step:1500/20000 train_loss:2.1462 train_time:809275ms step_avg:539.52ms +step:2000/20000 train_loss:2.0608 train_time:1078630ms step_avg:539.31ms +step:2500/20000 train_loss:2.1064 train_time:1348124ms step_avg:539.25ms +step:3000/20000 train_loss:2.0881 train_time:1618678ms step_avg:539.56ms +step:3500/20000 train_loss:2.0826 train_time:1890189ms step_avg:540.05ms +step:4000/20000 train_loss:2.1590 train_time:2161106ms step_avg:540.28ms +step:4000/20000 val_loss:2.0695 val_bpb:1.2257 train_time:2161107ms step_avg:540.28ms +step:4500/20000 train_loss:2.1635 train_time:2430332ms step_avg:540.07ms +step:5000/20000 train_loss:2.0850 train_time:2699671ms step_avg:539.93ms +step:5500/20000 train_loss:2.1032 train_time:2969143ms step_avg:539.84ms +step:6000/20000 train_loss:2.0135 train_time:3238597ms step_avg:539.77ms +step:6500/20000 train_loss:2.1377 train_time:3508121ms step_avg:539.71ms +step:7000/20000 train_loss:1.9553 train_time:3777673ms step_avg:539.67ms +step:7500/20000 train_loss:2.0262 train_time:4047143ms step_avg:539.62ms +step:8000/20000 train_loss:1.9944 train_time:4316325ms step_avg:539.54ms +step:8000/20000 val_loss:2.0372 val_bpb:1.2065 train_time:4316327ms step_avg:539.54ms +step:8500/20000 train_loss:1.9771 train_time:4585755ms step_avg:539.50ms +step:9000/20000 train_loss:2.0287 train_time:4854780ms step_avg:539.42ms +step:9500/20000 train_loss:2.1483 train_time:5123901ms step_avg:539.36ms +step:10000/20000 train_loss:2.0645 train_time:5393087ms step_avg:539.31ms +step:10500/20000 train_loss:2.0524 train_time:5662241ms step_avg:539.26ms +step:11000/20000 train_loss:2.0164 train_time:5932487ms step_avg:539.32ms +step:11500/20000 train_loss:2.0095 train_time:6202903ms step_avg:539.38ms +step:12000/20000 train_loss:2.0306 train_time:6473571ms step_avg:539.46ms +step:12000/20000 val_loss:2.0274 val_bpb:1.2008 train_time:6473572ms step_avg:539.46ms +step:12500/20000 train_loss:2.0244 train_time:6742627ms step_avg:539.41ms +step:13000/20000 train_loss:1.9829 train_time:7011689ms step_avg:539.36ms +step:13500/20000 train_loss:2.0434 train_time:7280671ms step_avg:539.31ms +step:14000/20000 train_loss:1.9777 train_time:7549684ms step_avg:539.26ms +step:14500/20000 train_loss:2.0642 train_time:7818540ms step_avg:539.21ms +step:15000/20000 train_loss:2.0342 train_time:8087482ms step_avg:539.17ms +step:15500/20000 train_loss:2.0133 train_time:8356529ms step_avg:539.13ms +step:16000/20000 train_loss:2.0044 train_time:8625790ms step_avg:539.11ms +step:16000/20000 val_loss:2.0211 val_bpb:1.1970 train_time:8625791ms step_avg:539.11ms +step:16500/20000 train_loss:2.1391 train_time:8895088ms step_avg:539.10ms +step:17000/20000 train_loss:2.0137 train_time:9164358ms step_avg:539.08ms +step:17500/20000 train_loss:2.0121 train_time:9433593ms step_avg:539.06ms +step:18000/20000 train_loss:2.0572 train_time:9702830ms step_avg:539.05ms +step:18500/20000 train_loss:1.9100 train_time:9972129ms step_avg:539.03ms +step:19000/20000 train_loss:1.9059 train_time:10246622ms step_avg:539.30ms +step:19500/20000 train_loss:2.1205 train_time:10518165ms step_avg:539.39ms +step:20000/20000 train_loss:2.0205 train_time:10790324ms step_avg:539.52ms +step:20000/20000 val_loss:2.0138 val_bpb:1.1927 train_time:10790325ms step_avg:539.52ms +peak memory allocated: 20915 MiB reserved: 22058 MiB +ema:applying EMA weights +DIAGNOSTIC post_ema val_loss:1.9375 val_bpb:1.1475 eval_time:12018ms +Serialized model: 106178569 bytes +Code size: 81323 bytes +Serialized model int6+zstd: 17050758 bytes +Total submission size int6+zstd: 17132081 bytes +Total submission size int8+zlib: 17132081 bytes +final_int6_roundtrip val_loss:1.9421 val_bpb:1.1502 eval_time:44884ms +final_int6_roundtrip_exact val_loss:1.94206549 val_bpb:1.15020057 +final_int6_sliding_window val_loss:1.9019 val_bpb:1.1264 stride:64 eval_time:485779ms +final_int6_sliding_window_exact val_loss:1.90191176 val_bpb:1.12642226 +final_int8_zlib_roundtrip_exact val_loss:1.90191176 val_bpb:1.12642226 +final_eval_mode:ttt_lora rank:8 epochs:5 chunk:256 lr:0.01 + ttt [ 0.0%] 0/50000 docs bpb=1.275697 time=0s + ttt [ 0.6%] 320/50000 docs bpb=1.274202 time=1s + ttt [ 1.3%] 640/50000 docs bpb=1.283712 time=1s + ttt [ 1.9%] 960/50000 docs bpb=1.290907 time=2s + ttt [ 2.6%] 1280/50000 docs bpb=1.285963 time=2s + ttt [ 3.2%] 1600/50000 docs bpb=1.281384 time=3s + ttt [ 3.8%] 1920/50000 docs bpb=1.273987 time=3s + ttt [ 4.5%] 2240/50000 docs bpb=1.269887 time=4s + ttt [ 5.1%] 2560/50000 docs bpb=1.269091 time=4s + ttt [ 5.8%] 2880/50000 docs bpb=1.271943 time=5s + ttt [ 6.4%] 3200/50000 docs bpb=1.275074 time=5s + ttt [ 7.0%] 3520/50000 docs bpb=1.274511 time=6s + ttt [ 7.7%] 3840/50000 docs bpb=1.272647 time=6s + ttt [ 8.3%] 4160/50000 docs bpb=1.269977 time=7s + ttt [ 9.0%] 4480/50000 docs bpb=1.268232 time=7s + ttt [ 9.6%] 4800/50000 docs bpb=1.268124 time=8s + ttt [ 10.2%] 5120/50000 docs bpb=1.269726 time=8s + ttt [ 10.9%] 5440/50000 docs bpb=1.271132 time=8s + ttt [ 11.5%] 5760/50000 docs bpb=1.271848 time=9s + ttt [ 12.2%] 6080/50000 docs bpb=1.271271 time=9s + ttt [ 12.8%] 6400/50000 docs bpb=1.270564 time=10s + ttt [ 13.4%] 6720/50000 docs bpb=1.266495 time=11s + ttt [ 14.1%] 7040/50000 docs bpb=1.258133 time=13s + ttt [ 14.7%] 7360/50000 docs bpb=1.250479 time=15s + ttt [ 15.4%] 7680/50000 docs bpb=1.241813 time=17s + ttt [ 16.0%] 8000/50000 docs bpb=1.234906 time=18s + ttt [ 16.6%] 8320/50000 docs bpb=1.227923 time=20s + ttt [ 17.3%] 8640/50000 docs bpb=1.226089 time=22s + ttt [ 17.9%] 8960/50000 docs bpb=1.222978 time=23s + ttt [ 18.6%] 9280/50000 docs bpb=1.217660 time=25s + ttt [ 19.2%] 9600/50000 docs bpb=1.211861 time=27s + ttt [ 19.8%] 9920/50000 docs bpb=1.209144 time=29s + ttt [ 20.5%] 10240/50000 docs bpb=1.206229 time=30s + ttt [ 21.1%] 10560/50000 docs bpb=1.201988 time=32s + ttt [ 21.8%] 10880/50000 docs bpb=1.198494 time=34s + ttt [ 22.4%] 11200/50000 docs bpb=1.195895 time=36s + ttt [ 23.0%] 11520/50000 docs bpb=1.194860 time=37s + ttt [ 23.7%] 11840/50000 docs bpb=1.193412 time=39s + ttt [ 24.3%] 12160/50000 docs bpb=1.191963 time=41s + ttt [ 25.0%] 12480/50000 docs bpb=1.189929 time=42s + ttt [ 25.6%] 12800/50000 docs bpb=1.189758 time=44s + ttt [ 26.2%] 13120/50000 docs bpb=1.188505 time=46s + ttt [ 26.9%] 13440/50000 docs bpb=1.187348 time=48s + ttt [ 27.5%] 13760/50000 docs bpb=1.185899 time=49s + ttt [ 28.2%] 14080/50000 docs bpb=1.184577 time=51s + ttt [ 28.8%] 14400/50000 docs bpb=1.182756 time=53s + ttt [ 29.4%] 14720/50000 docs bpb=1.180416 time=55s + ttt [ 30.1%] 15040/50000 docs bpb=1.179967 time=56s + ttt [ 30.7%] 15360/50000 docs bpb=1.178523 time=58s + ttt [ 31.4%] 15680/50000 docs bpb=1.178339 time=60s + ttt [ 32.0%] 16000/50000 docs bpb=1.178676 time=61s + ttt [ 32.6%] 16320/50000 docs bpb=1.177922 time=63s + ttt [ 33.3%] 16640/50000 docs bpb=1.177328 time=65s + ttt [ 33.9%] 16960/50000 docs bpb=1.176235 time=67s + ttt [ 34.6%] 17280/50000 docs bpb=1.175264 time=68s + ttt [ 35.2%] 17600/50000 docs bpb=1.173989 time=70s + ttt [ 35.8%] 17920/50000 docs bpb=1.172907 time=72s + ttt [ 36.5%] 18240/50000 docs bpb=1.166850 time=75s + ttt [ 37.1%] 18560/50000 docs bpb=1.159179 time=78s + ttt [ 37.8%] 18880/50000 docs bpb=1.150826 time=82s + ttt [ 38.4%] 19200/50000 docs bpb=1.143902 time=86s + ttt [ 39.0%] 19520/50000 docs bpb=1.137498 time=89s + ttt [ 39.7%] 19840/50000 docs bpb=1.130720 time=93s + ttt [ 40.3%] 20160/50000 docs bpb=1.123809 time=96s + ttt [ 41.0%] 20480/50000 docs bpb=1.117963 time=100s + ttt [ 41.6%] 20800/50000 docs bpb=1.111941 time=104s + ttt [ 42.2%] 21120/50000 docs bpb=1.106958 time=107s + ttt [ 42.9%] 21440/50000 docs bpb=1.102852 time=111s + ttt [ 43.5%] 21760/50000 docs bpb=1.098628 time=114s + ttt [ 44.2%] 22080/50000 docs bpb=1.095044 time=118s + ttt [ 44.8%] 22400/50000 docs bpb=1.090863 time=122s + ttt [ 45.4%] 22720/50000 docs bpb=1.086984 time=125s + ttt [ 46.1%] 23040/50000 docs bpb=1.082653 time=129s + ttt [ 46.7%] 23360/50000 docs bpb=1.078375 time=133s + ttt [ 47.4%] 23680/50000 docs bpb=1.075058 time=136s + ttt [ 48.0%] 24000/50000 docs bpb=1.072147 time=140s + ttt [ 48.6%] 24320/50000 docs bpb=1.069409 time=144s + ttt [ 49.3%] 24640/50000 docs bpb=1.066548 time=147s + ttt [ 49.9%] 24960/50000 docs bpb=1.063903 time=151s + ttt [ 50.6%] 25280/50000 docs bpb=1.060120 time=154s + ttt [ 51.2%] 25600/50000 docs bpb=1.057388 time=158s + ttt [ 51.8%] 25920/50000 docs bpb=1.053572 time=162s + ttt [ 52.5%] 26240/50000 docs bpb=1.046358 time=168s + ttt [ 53.1%] 26560/50000 docs bpb=1.040617 time=175s + ttt [ 53.8%] 26880/50000 docs bpb=1.034678 time=181s + ttt [ 54.4%] 27200/50000 docs bpb=1.028635 time=187s + ttt [ 55.0%] 27520/50000 docs bpb=1.022887 time=193s + ttt [ 55.7%] 27840/50000 docs bpb=1.016878 time=199s + ttt [ 56.3%] 28160/50000 docs bpb=1.012101 time=206s + ttt [ 57.0%] 28480/50000 docs bpb=1.007559 time=212s + ttt [ 57.6%] 28800/50000 docs bpb=1.003629 time=218s + ttt [ 58.2%] 29120/50000 docs bpb=0.998954 time=224s + ttt [ 58.9%] 29440/50000 docs bpb=0.994546 time=230s + ttt [ 59.5%] 29760/50000 docs bpb=0.990149 time=237s + ttt [ 60.2%] 30080/50000 docs bpb=0.986628 time=243s + ttt [ 60.8%] 30400/50000 docs bpb=0.982818 time=249s + ttt [ 61.4%] 30720/50000 docs bpb=0.979892 time=255s + ttt [ 62.1%] 31040/50000 docs bpb=0.976441 time=261s + ttt [ 62.7%] 31360/50000 docs bpb=0.972690 time=268s + ttt [ 63.4%] 31680/50000 docs bpb=0.968719 time=275s + ttt [ 64.0%] 32000/50000 docs bpb=0.964100 time=284s + ttt [ 64.6%] 32320/50000 docs bpb=0.959533 time=293s + ttt [ 65.3%] 32640/50000 docs bpb=0.954839 time=303s + ttt [ 65.9%] 32960/50000 docs bpb=0.950151 time=312s + ttt [ 66.6%] 33280/50000 docs bpb=0.945582 time=321s + ttt [ 67.2%] 33600/50000 docs bpb=0.941444 time=330s + ttt [ 67.8%] 33920/50000 docs bpb=0.937960 time=339s + ttt [ 68.5%] 34240/50000 docs bpb=0.934232 time=349s + ttt [ 69.1%] 34560/50000 docs bpb=0.930180 time=358s + ttt [ 69.8%] 34880/50000 docs bpb=0.926866 time=367s + ttt [ 70.4%] 35200/50000 docs bpb=0.923682 time=376s + ttt [ 71.0%] 35520/50000 docs bpb=0.920468 time=385s + ttt [ 71.7%] 35840/50000 docs bpb=0.916893 time=395s + ttt [ 72.3%] 36160/50000 docs bpb=0.912889 time=407s + ttt [ 73.0%] 36480/50000 docs bpb=0.909177 time=420s + ttt [ 73.6%] 36800/50000 docs bpb=0.904685 time=432s + ttt [ 74.2%] 37120/50000 docs bpb=0.900854 time=444s + ttt [ 74.9%] 37440/50000 docs bpb=0.897448 time=456s + ttt [ 75.5%] 37760/50000 docs bpb=0.894047 time=468s + ttt [ 76.2%] 38080/50000 docs bpb=0.890374 time=481s + ttt [ 76.8%] 38400/50000 docs bpb=0.887212 time=493s + ttt [ 77.4%] 38720/50000 docs bpb=0.884451 time=505s + ttt [ 78.1%] 39040/50000 docs bpb=0.881149 time=519s + ttt [ 78.7%] 39360/50000 docs bpb=0.877785 time=534s + ttt [ 79.4%] 39680/50000 docs bpb=0.874174 time=549s + ttt [ 80.0%] 40000/50000 docs bpb=0.870655 time=564s + ttt [ 80.6%] 40320/50000 docs bpb=0.867465 time=580s + ttt [ 81.3%] 40640/50000 docs bpb=0.864086 time=595s + ttt [ 81.9%] 40960/50000 docs bpb=0.861296 time=610s + ttt [ 82.6%] 41280/50000 docs bpb=0.858177 time=626s + ttt [ 83.2%] 41600/50000 docs bpb=0.855110 time=645s + ttt [ 83.8%] 41920/50000 docs bpb=0.851826 time=663s + ttt [ 84.5%] 42240/50000 docs bpb=0.848757 time=681s + ttt [ 85.1%] 42560/50000 docs bpb=0.845685 time=699s + ttt [ 85.8%] 42880/50000 docs bpb=0.843201 time=717s + ttt [ 86.4%] 43200/50000 docs bpb=0.840483 time=738s + ttt [ 87.0%] 43520/50000 docs bpb=0.837364 time=759s + ttt [ 87.7%] 43840/50000 docs bpb=0.834642 time=780s + ttt [ 88.3%] 44160/50000 docs bpb=0.831938 time=801s + ttt [ 89.0%] 44480/50000 docs bpb=0.829165 time=824s + ttt [ 89.6%] 44800/50000 docs bpb=0.826148 time=849s + ttt [ 90.2%] 45120/50000 docs bpb=0.823377 time=873s + ttt [ 90.9%] 45440/50000 docs bpb=0.820912 time=898s + ttt [ 91.5%] 45760/50000 docs bpb=0.818454 time=925s + ttt [ 92.2%] 46080/50000 docs bpb=0.816154 time=953s + ttt [ 92.8%] 46400/50000 docs bpb=0.814358 time=983s + ttt [ 93.4%] 46720/50000 docs bpb=0.812441 time=1015s + ttt [ 94.1%] 47040/50000 docs bpb=0.810758 time=1048s + ttt [ 94.7%] 47360/50000 docs bpb=0.809467 time=1084s + ttt [ 95.4%] 47680/50000 docs bpb=0.808880 time=1123s + ttt [ 96.0%] 48000/50000 docs bpb=0.807946 time=1167s + ttt [ 96.6%] 48320/50000 docs bpb=0.807166 time=1213s + ttt [ 97.3%] 48640/50000 docs bpb=0.806905 time=1267s + ttt [ 97.9%] 48960/50000 docs bpb=0.807058 time=1329s + ttt [ 98.6%] 49280/50000 docs bpb=0.807168 time=1404s + ttt [ 99.2%] 49600/50000 docs bpb=0.807405 time=1506s + ttt [ 99.8%] 49920/50000 docs bpb=0.815240 time=1735s +final_ttt val_loss:1.3821 val_bpb:0.8186 eval_time:1819176ms +final_ttt_exact val_loss:1.38211666 val_bpb:0.81856733