train_gpt.py

import os
import sys
# ----------------- CLI arguments -----------------
import argparse

parser = argparse.ArgumentParser(description="modded-nanogpt trainer")
parser.add_argument("--lr", type=float, default=1, help="Learning rate")
parser.add_argument("--p", type=int, default=1, help="Schatten constraint")
# parse_known_args lets torchrun pass its own flags (like --local_rank)
cli_args, _ = parser.parse_known_args()

with open(sys.argv[0]) as f:
    code = f.read() # read the code of this file ASAP, for logging
import uuid
import time
import copy
import glob
from dataclasses import dataclass
from functools import lru_cache, partial # Added partial for hook registration
from pathlib import Path

os.environ["PYTORCH_CUDA_ALLOC_CONF"] = "expandable_segments:True"
import torch
torch.empty(1, device="cuda", requires_grad=True).backward() # prevents a bug on some systems
from torch import Tensor, nn
import torch.nn.functional as F
import torch.distributed as dist
# use of FlexAttention contributed by @KoszarskyB
from torch.nn.attention.flex_attention import BlockMask, flex_attention
#torch._inductor.config.coordinate_descent_tuning = True # we have banned this flag for new records because it causes compilation to take 30min

# -----------------------------------------------------------------------------
# Custom operators: FP8 matmul by @YouJiacheng

@torch.library.custom_op("nanogpt::mm", mutates_args=())
def mm_op(x: Tensor, w: Tensor, x_s: float, w_s: float, grad_s: float) -> tuple[Tensor, Tensor, Tensor]:
    @torch.compile
    def impl(x: Tensor, w: Tensor):
        assert x.is_contiguous() and w.is_contiguous()
        x_f8 = x.div(x_s).to(torch.float8_e4m3fn)
        w_f8 = w.div(w_s).to(torch.float8_e4m3fn)
        out = torch._scaled_mm(
            x_f8,
            w_f8.T,
            out_dtype=torch.bfloat16,
            scale_a=x.new_tensor(x_s, dtype=torch.float32),
            scale_b=x.new_tensor(w_s, dtype=torch.float32),
            use_fast_accum=True,
        )
        return out, x_f8, w_f8

    return impl(x, w)

@mm_op.register_fake
def _(x: Tensor, w: Tensor, *_):
    assert x.ndim == w.ndim == 2
    assert x.shape[1] == w.shape[1]
    assert x.device == w.device
    assert x.is_contiguous() and w.is_contiguous()
    return x @ w.T, x.to(torch.float8_e4m3fn), w.to(torch.float8_e4m3fn)

@torch.library.custom_op("nanogpt::mm_backward", mutates_args=())
def mm_backward_op(g: Tensor, x_f8: Tensor, w_f8: Tensor, x_s: float, w_s: float, grad_s: float) -> tuple[Tensor, Tensor]:
    @torch.compile
    def impl(grad: Tensor, x_f8: Tensor, w_f8: Tensor):
        assert grad.is_contiguous()
        x_inv_s = grad.new_tensor(x_s, dtype=torch.float32)
        w_inv_s = grad.new_tensor(w_s, dtype=torch.float32)
        grad_inv_s = grad.new_tensor(grad_s, dtype=torch.float32)
        grad_f8 = grad.div(grad_s).to(torch.float8_e5m2)
        grad_x = torch._scaled_mm(
            grad_f8,
            w_f8.T.contiguous().T,
            out_dtype=torch.bfloat16,
            scale_a=grad_inv_s,
            scale_b=w_inv_s,
            use_fast_accum=False,
        )
        # faster than grad_f8_t @ x_f8, for (d_out, d_in) == (50304, 768)
        grad_w = torch._scaled_mm(
            x_f8.T.contiguous(),
            grad_f8.T.contiguous().T,
            out_dtype=torch.float32,
            scale_a=x_inv_s,
            scale_b=grad_inv_s,
            use_fast_accum=False,
        ).T
        return grad_x, grad_w

    return impl(g, x_f8, w_f8)

@mm_backward_op.register_fake
def _(g: Tensor, x_f8: Tensor, w_f8: Tensor, *_):
    return x_f8.to(torch.bfloat16), w_f8.T.contiguous().T.to(torch.float32)

def backward(ctx, grad_out: Tensor, *_):
    x_f8, w_f8 = ctx.saved_tensors
    x_s, w_s, grad_s = ctx.scales
    grad_x, grad_w = torch.ops.nanogpt.mm_backward(
        grad_out, x_f8, w_f8, x_s, w_s, grad_s
    )
    return grad_x, grad_w, None, None, None

def setup_context(ctx: torch.autograd.function.FunctionCtx, inputs, output):
    *_, x_s, w_s, grad_s = inputs
    _, x_f8, w_f8 = output
    ctx.save_for_backward(x_f8, w_f8)
    ctx.scales = x_s, w_s, grad_s
    ctx.set_materialize_grads(False)

mm_op.register_autograd(backward, setup_context=setup_context)

# -----------------------------------------------------------------------------
# Muon optimizer

@torch.compile
def zeropower_via_newtonschulz5(G: Tensor, steps: int) -> Tensor:
    """
    Newton-Schulz iteration to compute the zeroth power / orthogonalization of G. We opt to use a
    quintic iteration whose coefficients are selected to maximize the slope at zero. For the purpose
    of minimizing steps, it turns out to be empirically effective to keep increasing the slope at
    zero even beyond the point where the iteration no longer converges all the way to one everywhere
    on the interval. This iteration therefore does not produce UV^T but rather something like US'V^T
    where S' is diagonal with S_{ii}' ~ Uniform(0.5, 1.5), which turns out not to hurt model
    performance at all relative to UV^T, where USV^T = G is the SVD.
    """

    assert G.ndim >= 2 # batched Muon implementation by @scottjmaddox, and put into practice in the record by @YouJiacheng
    a, b, c = (3.4445, -4.7750,  2.0315)
    X = G
    if G.size(-2) > G.size(-1):
        X = X.mT

    # Ensure spectral norm is at most 1
    X = X / (X.norm(dim=(-2, -1), keepdim=True) + 1e-7)
    # Perform the NS iterations
    for _ in range(steps):
        A = X @ X.mT
        B = b * A + c * A @ A # quintic computation strategy adapted from suggestion by @jxbz, @leloykun, and @YouJiacheng
        X = a * X + B @ X

    if G.size(-2) > G.size(-1):
        X = X.mT
    return X
    # Compute the SVD of G and construct the nearest orthogonal matrix.
    # For G = U Σ Vᵀ, the zeroth-power (orthogonal) component is U Vᵀ.
    # torch.linalg.svd works with batched inputs, so this is fully vectorised.
    if(G.allclose(torch.zeros_like(G))):
        return G
    U, S, Vh = torch.linalg.svd(G.float(), full_matrices=False)
    # √σ
    S_half = S.to(G.dtype) ** (1.0/(cli_args.p - 1))
     # --- Energy‑matching scale γ  -----------------------------------
    r = min(G.shape[-2], G.shape[-1])                     # rank
    energy = (S_half ** cli_args.p).sum()
    gamma = (r / energy) ** (1.0/cli_args.p)                 # scalar
    # Reconstruct Z = U Σ½ Vᵀ
    X = (U.to(G.dtype) * S_half.unsqueeze(-2)) @ Vh.to(G.dtype) * gamma
    #X = U.to(G.dtype) @ Vh.to(G.dtype)
    return X * cli_args.lr

class Muon(torch.optim.Optimizer):
    """
    Muon - MomentUm Orthogonalized by Newton-schulz

    https://kellerjordan.github.io/posts/muon/

    Muon internally runs standard SGD-momentum, and then performs an orthogonalization post-
    processing step, in which each 2D parameter's update is replaced with the nearest orthogonal
    matrix. To efficiently orthogonalize each update, we use a Newton-Schulz iteration, which has
    the advantage that it can be stably run in bfloat16 on the GPU.

    Warning: This optimizer should not be used for the embedding layer, the final fully connected layer,
    or any {0,1}-D parameters; those should all be optimized by a standard method (e.g., AdamW).
    """
    def __init__(self, params, lr=0.02, weight_decay=0.01, momentum=0.95):
        defaults = dict(lr=lr, weight_decay=weight_decay, momentum=momentum)
        params = list(params)
        sizes = {p.shape for p in params}
        # create one buffer per unique parameter-size
        param_groups = []
        for size in sizes:
            group_params = [p for p in params if p.shape == size]
            param_groups.append(dict(params=group_params))
        super().__init__(param_groups, defaults)

    @torch.no_grad()
    def step(self):
        # Efficient systems-wise implementation of step developed by @YouJiacheng,
        # @KonstantinWilleke, @alexrgilbert, @adricarda, @tuttyfrutyee, @vdlad,
        # @ryanyang0, and @vagrawal.
        rank = dist.get_rank()
        world_size = dist.get_world_size()
        reduce_scatter_futures: list[torch.Future] = []
        all_reduce_futures: list[torch.Future] = []
        for group in self.param_groups:
            params: list[Tensor] = group["params"]
            grad = torch.empty_like(params[-1])
            grad_pad = [param.grad for param in params] + [torch.zeros_like(params[-1])] * world_size
            for base_i in range(0, len(params), world_size):
                if base_i + rank < len(params):
                    grad = params[base_i + rank].grad
                # This gives strange dynamo warnings
                reduce_scatter_futures.append(dist.reduce_scatter(grad, grad_pad[base_i:base_i + world_size], op=dist.ReduceOp.AVG, async_op=True).get_future())

        idx = 0
        for group in self.param_groups:
            params: list[Tensor] = group["params"]
            params_pad = params + [torch.empty_like(params[-1])] * world_size
            momentum = group["momentum"]
            for base_i in range(0, len(params), world_size):
                reduce_scatter_futures[idx].wait()
                if base_i + rank < len(params):
                    p = params[base_i + rank]
                    grad = p.grad
                    eff_lr = group["lr"] * max(1, p.size(-2) / p.size(-1)) ** 0.5 * getattr(p, "lr_mul", 1.0)
                    eff_weight_decay = group["lr"] * group["weight_decay"] * getattr(p, "wd_mul", 1.0)
                    state = self.state[p]
                    if len(state) == 0:
                        state["momentum_buffer"] = torch.zeros_like(grad)
                    momentum_buffer = state["momentum_buffer"]
                    p.mul_(1 - eff_weight_decay)
                    momentum_buffer.lerp_(grad, 1 - momentum)
                    grad = grad.lerp_(momentum_buffer, momentum)
                    if p.grad.max() < 1e-10:
                        v = grad
                    else:
                        v = zeropower_via_newtonschulz5(grad.bfloat16(), 5)
                    p.add_(other=v, alpha=-eff_lr)
                idx += 1
                all_reduce_futures.append(dist.all_gather(params_pad[base_i:base_i + world_size], params_pad[base_i + rank], async_op=True).get_future())
        torch.futures.collect_all(all_reduce_futures).wait()

class DistAdam(torch.optim.Optimizer):
    def __init__(self, params, lr: float = 1e-3, betas: tuple[float, float] = (0.9, 0.999), eps: float = 1e-8, weight_decay: float = 0.01):
        defaults = dict(lr=lr, betas=betas, eps=eps, weight_decay=weight_decay)
        params = list(params)
        sizes = {p.shape for p in params}
        # create one buffer per unique parameter-size
        param_groups = []
        for size in sizes:
            group_params = [p for p in params if p.shape == size]
            param_groups.append(dict(params=group_params))
        super().__init__(param_groups, defaults)
        # DistributedAdam implementation by @vagrawal

    @torch.compile
    @torch.no_grad()
    def step(self):
        rank = dist.get_rank()
        world_size = dist.get_world_size()
        reduce_scatter_futures: list[torch.Future] = []
        all_reduce_futures: list[torch.Future] = []
        grad_slices = []
        for group in self.param_groups:
            params: list[Tensor] = group["params"]
            grad = torch.empty_like(params[-1])
            for base_i in range(len(params)):
                grad = params[base_i].grad
                rank_size = grad.shape[0] // world_size
                grad_slice = torch.empty_like(grad[:rank_size])
                reduce_scatter_futures.append(dist.reduce_scatter_tensor(grad_slice, grad, op=dist.ReduceOp.AVG, async_op=True).get_future())
                grad_slices.append(grad_slice)

        idx = 0
        for group in self.param_groups:
            beta1, beta2 = group['betas']
            eps = group['eps']
            wd = group['weight_decay']
            params = group['params']
            for base in range(len(params)):
                reduce_scatter_futures[idx].wait()
                p = params[base]
                rank_size = p.shape[0] // world_size
                p_slice = p[rank * rank_size:(rank + 1) * rank_size]
                lr = group['lr'] * getattr(p, "lr_mul", 1.0)
                state = self.state[p]
                g_slice = grad_slices[idx]
                # State init
                if not state:
                    state['step'] = torch.tensor(0, dtype=torch.int64, device=p.device)
                    state['exp_avg'] = torch.zeros_like(p_slice)
                    state['exp_avg_sq'] = torch.zeros_like(p_slice)
                exp_avg = state['exp_avg']
                exp_avg_sq = state['exp_avg_sq']
                state['step'] += 1
                t = state['step']
                # weight decay
                if wd != 0:
                    eff_weight_decay = lr * wd * getattr(p, "wd_mul", 1.0)
                    p_slice.mul_(1 - eff_weight_decay)
                # update running averages
                exp_avg.mul_(beta1).add_(g_slice, alpha=1 - beta1)
                exp_avg_sq.mul_(beta2).addcmul_(g_slice, g_slice, value=1 - beta2)
                # bias corrections
                bias1 = 1 - beta1 ** t
                bias2 = 1 - beta2 ** t
                # compute step
                denom = exp_avg_sq.sqrt().add_(eps)
                step_size = lr * (torch.sqrt(bias2) / bias1)
                update = exp_avg.div(denom).mul_(step_size)
                p_slice.add_(other=update, alpha=-1.0)
                idx += 1
                all_reduce_futures.append(dist.all_gather_into_tensor(p, p_slice, async_op=True).get_future())
        torch.futures.collect_all(all_reduce_futures).wait()

# -----------------------------------------------------------------------------
# PyTorch nn.Module definitions for the model

def norm(x: Tensor):
    return F.rms_norm(x, (x.size(-1),))

class CastedLinear(nn.Linear):
    def __init__(self, in_features: int, out_features: int, use_fp8=False, x_s=1.0, w_s=1.0, grad_s=1.0):
        super().__init__(in_features, out_features, bias=False)
        self.use_fp8 = use_fp8
        self.x_s = x_s
        self.w_s = w_s
        self.grad_s = grad_s

    def reset_parameters(self) -> None:
        std = 0.5 * (self.in_features ** -0.5) # 0.5 is a bit better than the default 1/sqrt(3)
        bound = (3 ** 0.5) * std
        with torch.no_grad():
            self.weight.uniform_(-bound, bound)

    def forward(self, x: Tensor):
        if self.use_fp8 and self.training:
            _x = x.flatten(0, -2)
            out: Tensor = torch.ops.nanogpt.mm(_x, self.weight, x_s=self.x_s, w_s=self.w_s, grad_s=self.grad_s)[0]
            return out.reshape(*x.shape[:-1], -1)
        else:
            return F.linear(x, self.weight.type_as(x))

class Rotary(nn.Module):
    def __init__(self, dim: int, max_seq_len: int):
        super().__init__()
        # half-truncate RoPE by @YouJiacheng (w/ base freq tuning)
        angular_freq = (1 / 1024) ** torch.linspace(0, 1, steps=dim//4, dtype=torch.float32)
        angular_freq = torch.cat([angular_freq, angular_freq.new_zeros(dim//4)])
        t = torch.arange(max_seq_len, dtype=torch.float32)
        theta = torch.einsum("i,j -> ij", t, angular_freq)
        self.cos = nn.Buffer(theta.cos(), persistent=False)
        self.sin = nn.Buffer(theta.sin(), persistent=False)

    def forward(self, x_BTHD: Tensor):
        assert self.cos.size(0) >= x_BTHD.size(-3)
        cos, sin = self.cos[None, :x_BTHD.size(-3), None, :], self.sin[None, :x_BTHD.size(-3), None, :]
        x1, x2 = x_BTHD.to(dtype=torch.float32).chunk(2, dim=-1)
        y1 = x1 * cos + x2 * sin
        y2 = x1 * (-sin) + x2 * cos
        return torch.cat((y1, y2), 3).type_as(x_BTHD)

class CausalSelfAttention(nn.Module):
    def __init__(self, dim: int, num_heads: int, max_seq_len: int, head_dim=128):
        super().__init__()
        self.num_heads = num_heads
        self.head_dim = head_dim
        hdim = num_heads * head_dim
        std = 0.5 * (dim ** -0.5)
        bound = (3 ** 0.5) * std # improved init scale by @YouJiacheng
        # merged QKV weights: suggested by many, implemented by @fernbear.bsky.social, and further improved by @YouJiacheng
        # https://x.com/hi_tysam/status/1879699187107033311
        self.qkv_w = nn.Parameter(torch.empty(3, hdim, dim).uniform_(-bound, bound))
        self.rotary = Rotary(head_dim, max_seq_len)
        self.c_proj = CastedLinear(hdim, dim)
        self.c_proj.weight.detach().zero_() # zero init suggested by @Grad62304977
        # scale the attention logits by given constant, instead of the default head_dim**-0.5, by @leloykun
        # inspired by learnable scalars used by @brendanh0gan https://x.com/hi_tysam/status/1879693583898591283
        self.attn_scale = 0.12

    def forward(self, x: Tensor, ve: Tensor | None, lambdas: Tensor, block_mask: BlockMask):
        B, T = x.size(0), x.size(1) # batch size, sequence length
        assert B == 1, "Must use batch size = 1 for FlexAttention"
        q, k, v = F.linear(x, self.qkv_w.flatten(end_dim=1).type_as(x)).view(B, T, 3 * self.num_heads, self.head_dim).chunk(3, dim=-2)
        q, k = norm(q), norm(k) # QK norm @Grad62304977
        q, k = self.rotary(q), self.rotary(k)
        if ve is not None:
            v = lambdas[0] * v + lambdas[1] * ve.view_as(v) # @KoszarskyB & @Grad62304977
        else: # skip mid-layers token value embeddings by @YouJiacheng
            v = lambdas[0] * v
        y = flex_attention(q.transpose(1, 2), k.transpose(1, 2), v.transpose(1, 2), block_mask=block_mask, scale=self.attn_scale).transpose(1, 2)
        y = y.contiguous().view(B, T, self.num_heads * self.head_dim) # re-assemble all head outputs side by side
        y = self.c_proj(y)
        return y

class MLP(nn.Module):
    def __init__(self, dim: int):
        super().__init__()
        hdim = 4 * dim
        self.c_fc = CastedLinear(dim, hdim)
        self.c_proj = CastedLinear(hdim, dim)
        self.c_proj.weight.detach().zero_() # zero init suggested by @Grad62304977

    def forward(self, x: Tensor):
        x = self.c_fc(x)
        x = F.relu(x).square() # https://arxiv.org/abs/2109.08668v2; ~1-2% better than GELU; suggested by @SKYLINEZ007 and @Grad62304977
        x = self.c_proj(x)
        return x

class Block(nn.Module):
    def __init__(self, dim: int, num_heads: int, max_seq_len: int, layer_idx: int):
        super().__init__()
        # skip attention of blocks.7 (the 8th layer) by @YouJiacheng
        self.attn = CausalSelfAttention(dim, num_heads, max_seq_len) if layer_idx != 7 else None
        self.mlp = MLP(dim)

    def forward(self, x: Tensor, ve: Tensor | None, x0: Tensor, lambdas: Tensor, sa_lambdas: Tensor, block_mask: BlockMask):
        x = lambdas[0] * x + lambdas[1] * x0
        if self.attn is not None:
            x = x + self.attn(norm(x), ve, sa_lambdas, block_mask)
        x = x + self.mlp(norm(x))
        return x

# -----------------------------------------------------------------------------
# The main model

def next_multiple_of_n(v: float | int, *, n: int):
    return next(x for x in range(n, int(v) + 1 + n, n) if x >= v)

class GPT(nn.Module):
    def __init__(self, vocab_size: int, num_layers: int, num_heads: int, model_dim: int, max_seq_len: int):
        super().__init__()
        vocab_size = next_multiple_of_n(vocab_size, n=128)
        self.embed = nn.Embedding(vocab_size, model_dim)
        # token value embeddings by @KoszarskyB - inspired by @Grad62304977's value residual implementation following https://arxiv.org/abs/2410.17897
        # value embedding code simplification inspired by @ragulpr https://github.com/KellerJordan/modded-nanogpt/pull/78
        self.value_embeds = nn.ModuleList([nn.Embedding(vocab_size, model_dim) for _ in range(3)])
        self.blocks = nn.ModuleList([Block(model_dim, num_heads, max_seq_len, i) for i in range(num_layers)])
        # there are only 50257 unique GPT-2 tokens; we extend to nearest multiple of 128 for efficiency.
        # suggested to me by @Grad62304977. this originates from Karpathy's experiments.
        self.lm_head = CastedLinear(model_dim, vocab_size, use_fp8=True, x_s=(model_dim**0.5)/448, w_s=24/448, grad_s=1/448)
        self.lm_head.weight.detach().zero_() # @Grad62304977
        # Add learnable skip connection weights for decoder layers
        assert num_layers % 2 == 0
        pad = (-num_layers * 5) % dist.get_world_size()
        self.scalars = nn.Parameter(torch.cat([
            torch.ones(num_layers), # skip_weights
            *[torch.tensor([1.0, 0.0]) for _ in range(num_layers)], # block lambdas
            *[torch.tensor([0.5, 0.5]) for _ in range(num_layers)], # SA lambdas
            torch.ones(pad),
        ]))
        # set learning rates
        for param in self.embed.parameters():
            param.lr_mul = 75.
        for param in self.value_embeds.parameters():
            param.lr_mul = 75.
        self.lm_head.weight.lr_mul = 27.5
        self.scalars.lr_mul = 5.0

    def create_blockmasks(self, input_seq: Tensor, sliding_window_num_blocks: Tensor):
        BLOCK_SIZE = 128
        docs = (input_seq == 50256).cumsum(0)

        def document_causal(b, h, q_idx, kv_idx):
            causal_mask = q_idx >= kv_idx
            document_mask = docs[q_idx] == docs[kv_idx]
            return causal_mask & document_mask

        def dense_to_ordered(dense_blockmask: Tensor):
            num_blocks = dense_blockmask.sum(dim=-1, dtype=torch.int32)
            indices = dense_blockmask.argsort(dim=-1, descending=False, stable=True).flip(-1).to(torch.int32)
            return num_blocks[None, None].contiguous(), indices[None, None].contiguous()

        # manual block mask creation by @YouJiacheng
        assert len(input_seq) % BLOCK_SIZE == 0
        NUM_BLOCKS = len(input_seq) // BLOCK_SIZE
        block_idx = torch.arange(NUM_BLOCKS, dtype=torch.int32, device="cuda")
        causal_blockmask_any = block_idx[:, None] >= block_idx
        causal_blockmask_all = block_idx[:, None] > block_idx
        docs_low = docs.view(-1, BLOCK_SIZE)[:, 0].contiguous()
        docs_high = docs.view(-1, BLOCK_SIZE)[:, -1].contiguous()
        document_blockmask_any = (docs_low[:, None] <= docs_high) & (docs_high[:, None] >= docs_low)
        document_blockmask_all = (docs_low[:, None] == docs_high) & (docs_high[:, None] == docs_low)
        blockmask_any = causal_blockmask_any & document_blockmask_any
        blockmask_all = causal_blockmask_all & document_blockmask_all
        partial_kv_num_blocks, partial_kv_indices = dense_to_ordered(blockmask_any & ~blockmask_all)
        full_kv_num_blocks, full_kv_indices = dense_to_ordered(blockmask_all)
        def build_bm(window_size_blocks: Tensor) -> BlockMask:
            return BlockMask.from_kv_blocks(
                torch.clamp_max(partial_kv_num_blocks, torch.clamp_min(window_size_blocks - full_kv_num_blocks, 1)),
                partial_kv_indices,
                torch.clamp_max(full_kv_num_blocks, window_size_blocks - 1),
                full_kv_indices,
                BLOCK_SIZE=BLOCK_SIZE,
                mask_mod=document_causal,
            )
        # Long-short SWA block masks by @leloykun & @YouJiacheng, adapated from suggestion by @Grad62304977, following Gemma 2 paper
        return build_bm(sliding_window_num_blocks), build_bm(sliding_window_num_blocks // 2)

    def forward(self, input_seq: Tensor, target_seq: Tensor, sliding_window_num_blocks: Tensor):
        assert input_seq.ndim == 1

        ve = [value_embed(input_seq) for value_embed in self.value_embeds]
        # 012 ... 012 structure on token value embeddings by @YouJiacheng, improved on @leloykun's U-net structure
        ve = [ve[0], ve[1], ve[2]] + [None] * (len(self.blocks) - 6) + [ve[0], ve[1], ve[2]]
        assert len(ve) == len(self.blocks)

        long_bm, short_bm = self.create_blockmasks(input_seq, sliding_window_num_blocks)
        block_masks = [long_bm, short_bm, short_bm, short_bm, long_bm, short_bm, short_bm, long_bm, short_bm, short_bm, short_bm, long_bm]
        assert len(block_masks) == len(self.blocks)

        x = x0 = norm(self.embed(input_seq)[None]) # use of norm here by @Grad62304977

        # U-net design by @brendanh0gan
        skip_connections = []
        skip_weights = self.scalars[:(len(self.blocks) // 2)]
        lambdas = self.scalars[1 * len(self.blocks): 3 * len(self.blocks)].view(-1, 2)
        sa_lambdas = self.scalars[3 * len(self.blocks): 5 * len(self.blocks)].view(-1, 2)

        n = len(self.blocks) // 2

        for i in range(len(self.blocks)):
            if i >= n:
                x = x + skip_weights[i - n] * skip_connections.pop()
            x = self.blocks[i](x, ve[i], x0, lambdas[i], sa_lambdas[i], block_masks[i])
            if i < n:
                skip_connections.append(x)

        x = norm(x)
        logits = self.lm_head(x).float()
        # @Grad62304977 added tanh softcapping following Gemma 2 paper, @KoszarskyB reduced it from 30 to 15, @YouJiacheng shifted it by +15 (2*sigmoid(2*x)=tanh(x)+1)
        logits = 30 * torch.sigmoid(logits / (7.5 * x.size(-1)**0.5))
        loss = F.cross_entropy(logits.view(-1, logits.size(-1)), target_seq, reduction="sum" if self.training else "mean")
        return loss

# -----------------------------------------------------------------------------
# Distributed data loader

def _load_data_shard(file: Path):
    header = torch.from_file(str(file), False, 256, dtype=torch.int32) # header is 256 int32
    assert header[0] == 20240520, "magic number mismatch in the data .bin file"
    assert header[1] == 1, "unsupported version"
    num_tokens = int(header[2]) # number of tokens (claimed)
    with file.open("rb", buffering=0) as f:
        tokens = torch.empty(num_tokens, dtype=torch.uint16, pin_memory=True) # avoid pin_memory copy by @YouJiacheng
        f.seek(256 * 4)
        nbytes = f.readinto(tokens.numpy()) # avoid bytes->array copy by @YouJiacheng
        assert nbytes == 2 * num_tokens, "number of tokens read does not match header"
    return tokens

# find world_size starting indicies, such that each begins with token 50256 and local_batches don't overlap
def find_batch_starts(tokens: Tensor, pos: int, local_batch_size: int, max_batch_span: int):
    boundary_mask = tokens[pos : pos + max_batch_span] == 50256
    boundary_positions = torch.nonzero(boundary_mask, as_tuple=False).squeeze(-1) + pos
    start = boundary_positions[0].item()
    starts = []
    for i in range(1, len(boundary_positions)):
        end = boundary_positions[i].item() 
        if end - start >= local_batch_size:
            starts.append(start) # append start once end pos is confirmed
            if len(starts) == dist.get_world_size():
                return starts, end - pos
            start = end
    assert False # increase max_batch_span if necessary

def distributed_data_generator(filename_pattern: str, batch_size: int, align_to_bos: bool):
    rank = dist.get_rank()
    world_size = dist.get_world_size()
    files = [Path(file) for file in sorted(glob.glob(filename_pattern))]
    assert batch_size % world_size == 0
    local_batch_size = batch_size // world_size
    file_iter = iter(files) # use itertools.cycle(files) instead if you want to do multi-epoch training
    tokens, pos = _load_data_shard(next(file_iter)), 0
    max_batch_span = 2 * batch_size if align_to_bos else batch_size # provide buffer to handle samples up to length local_batch_size
    while True:
        if pos + max_batch_span + 1 >= len(tokens):
            tokens, pos = _load_data_shard(next(file_iter)), 0
        if align_to_bos:
            batch_starts, batch_span = find_batch_starts(tokens, pos, local_batch_size, max_batch_span)
            start_idx = batch_starts[rank]
        else:
            batch_span = batch_size
            start_idx = pos + rank * local_batch_size
        buf = tokens[start_idx:][:local_batch_size + 1]
        inputs = buf[:-1].to(device="cuda", dtype=torch.int32, non_blocking=True) # no sync on host side;
        targets = buf[1:].to(device="cuda", dtype=torch.int64, non_blocking=True) # H2D in another stream isn't helpful.
        pos += batch_span
        yield inputs, targets

# -----------------------------------------------------------------------------
# int main

@dataclass
class Hyperparameters:
    # data
    train_files = "data/fineweb10B/fineweb_train_*.bin" # input .bin to train on
    val_files = "data/fineweb10B/fineweb_val_*.bin" # input .bin to eval validation loss on
    val_tokens = 10485760 # how many tokens of validation data? it's important to keep this fixed for consistent comparisons
    train_seq_len = 48*1024 # FlexAttention sequence length
    val_seq_len = 4*64*1024 # FlexAttention sequence length for validation
    # optimization
    num_iterations = 1750 # number of iterations to run
    cooldown_frac = 0.45 # fraction of training spent cooling down the learning rate
    # evaluation and logging
    val_loss_every = 125 # every how many steps to evaluate val loss? 0 for only at the end
    save_checkpoint = False
args = Hyperparameters()

# torchrun sets these env variables
rank = int(os.environ["RANK"])
world_size = int(os.environ["WORLD_SIZE"])
assert world_size == 8 # this code is designed for 8xH100
assert torch.cuda.is_available()
device = torch.device("cuda", int(os.environ["LOCAL_RANK"]))
torch.cuda.set_device(device)
dist.init_process_group(backend="nccl", device_id=device)
dist.barrier()
master_process = (rank == 0) # this process will do logging, checkpointing etc.

# begin logging
logfile = None
if master_process:
    run_id = uuid.uuid4()
    os.makedirs("logs", exist_ok=True)
    logfile = f"logs/p_{cli_args.p}_{cli_args.lr}.txt"
    print(logfile)
def print0(s, console=False):
    if master_process:
        with open(logfile, "a") as f:
            if console:
                print(s)
            print(s, file=f)

# begin by printing this file (the Python code)
print0(code)
print0("="*100)
# log information about the hardware/software environment this is running on
print0(f"Running Python {sys.version}")
print0(f"Running PyTorch {torch.version.__version__} compiled for CUDA {torch.version.cuda}")
def nvidia_smi():
    import subprocess  # avoid top level import
    return subprocess.run(["nvidia-smi"], stdout=subprocess.PIPE, stderr=subprocess.PIPE, text=True).stdout
print0(nvidia_smi())
print0("="*100)

model: nn.Module = GPT(vocab_size=50257, num_layers=12, num_heads=6, model_dim=768, max_seq_len=max(args.train_seq_len, args.val_seq_len)).cuda()
for m in model.modules():
    if isinstance(m, nn.Embedding):
        m.bfloat16()
for param in model.parameters():
    dist.broadcast(param.detach(), 0)

# collect the parameters to optimize
hidden_matrix_params = [p for n, p in model.blocks.named_parameters() if p.ndim >= 2 and "embed" not in n]
embed_params = [p for n, p in model.named_parameters() if "embed" in n]
scalar_params = [p for p in model.parameters() if p.ndim < 2]
head_params = [model.lm_head.weight]

# init the optimizer(s)
# small adam epsilon by @YouJiacheng. this is an alternate method of fixing the world_size dependence
# discovered by @fernbear.bsky.social https://x.com/hi_tysam/status/1879692937589875094
optimizer1 = DistAdam(scalar_params + head_params + embed_params, lr=0.008, betas=(0.8, 0.95), eps=1e-10, weight_decay=0.0)
optimizer2 = Muon(hidden_matrix_params, lr=0.05, momentum=0.95, weight_decay=0.0)
optimizers = [optimizer1, optimizer2]
for opt in optimizers:
    for group in opt.param_groups:
        group["initial_lr"] = group["lr"]

# learning rate schedule: stable then decay
def get_lr(step: int):
    x = step / args.num_iterations # progress in training
    assert 0 <= x < 1
    if x < 1 - args.cooldown_frac:
        return 1.0
    else:
        w = (1 - x) / args.cooldown_frac
        return w * 1.0 + (1 - w) * 0.1

# attention window size schedule: linearly increase
@lru_cache(1)
def get_window_size_blocks_helper(window_size: int):
    return torch.tensor(window_size // 128, dtype=torch.int32, pin_memory=True).cuda(non_blocking=True)
def get_window_size_blocks(step: int):
    x = step / args.num_iterations # progress in training
    assert 0 <= x <= 1
    # Linearly increase the block-wise sliding window size over training 128 -> 1792
    # increase by @fernbear.bsky.social; block-wise by @YouJiacheng
    window_size = next_multiple_of_n(1728 * x, n=128)
    return get_window_size_blocks_helper(window_size)

model: nn.Module = torch.compile(model, dynamic=False)

########################################
#            Warmup kernels            #
########################################

# Warmup the training kernels, then re-initialize the state so we aren't cheating
warmup_steps = 10
initial_state = dict(model=copy.deepcopy(model.state_dict()),
                     optimizers=[copy.deepcopy(opt.state_dict()) for opt in optimizers]) # save the initial state
train_loader = distributed_data_generator(args.train_files, world_size * args.train_seq_len, align_to_bos=True)
for _ in range(warmup_steps):
    inputs, targets = next(train_loader)
    model(inputs, targets, get_window_size_blocks(1)).backward()
    for opt in optimizers:
        opt.step()
    model.zero_grad(set_to_none=True)
model.load_state_dict(initial_state["model"])
for opt, opt_state in zip(optimizers, initial_state["optimizers"]):
    opt.load_state_dict(opt_state)
del train_loader, initial_state

########################################
#        Training and validation       #
########################################

train_loader = distributed_data_generator(args.train_files, world_size * args.train_seq_len, align_to_bos=True)
training_time_ms = 0
# start the clock
torch.cuda.synchronize()
t0 = time.perf_counter()
# begin training
train_steps = args.num_iterations
for step in range(train_steps + 1):
    last_step = (step == train_steps)

    # --------------- VALIDATION SECTION -----------------
    if last_step or (args.val_loss_every > 0 and step % args.val_loss_every == 0):
        # stop the clock
        torch.cuda.synchronize()
        training_time_ms += 1000 * (time.perf_counter() - t0)
        model.eval()
        val_batch_size = world_size * args.val_seq_len
        assert args.val_tokens % val_batch_size == 0
        val_steps = args.val_tokens // val_batch_size
        val_loader = distributed_data_generator(args.val_files, val_batch_size, align_to_bos=False)
        val_loss = 0
        with torch.no_grad():
            for _ in range(val_steps):
                inputs, targets = next(val_loader)
                val_loss += model(inputs, targets, get_window_size_blocks(step))
        val_loss /= val_steps
        del val_loader
        dist.all_reduce(val_loss, op=dist.ReduceOp.AVG)
        print0(f"step:{step}/{train_steps} val_loss:{val_loss:.4f} train_time:{training_time_ms:.0f}ms step_avg:{training_time_ms/max(step, 1):.2f}ms", console=True)
        model.train()
        # start the clock again
        torch.cuda.synchronize()
        t0 = time.perf_counter()

    if last_step:
        if master_process and args.save_checkpoint:
            log = dict(step=step, code=code, model=model.state_dict(), optimizers=[opt.state_dict() for opt in optimizers])
            os.makedirs(f"logs/{run_id}", exist_ok=True)
            torch.save(log, f"logs/{run_id}/state_step{step:06d}.pt")
        # the last step only has the validation loop, so break to avoid training
        break

    # --------------- TRAINING SECTION -----------------
    inputs, targets = next(train_loader)
    model(inputs, targets, get_window_size_blocks(step)).backward()
    # set optimization hyperparameters
    for opt in optimizers:
        for group in opt.param_groups:
            group["lr"] = group["initial_lr"] * get_lr(step)
    for group in optimizer2.param_groups:
        frac = min(step / 300, 1) # momentum warmup for muon
        group["momentum"] = (1 - frac) * 0.85 + frac * 0.95
    # step the optimizers
    for opt in optimizers:
        opt.step()
    # null the gradients
    model.zero_grad(set_to_none=True)
    # logging
    approx_training_time_ms = training_time_ms + 1000 * (time.perf_counter() - t0)
    print0(f"step:{step+1}/{train_steps} train_time:{approx_training_time_ms:.0f}ms step_avg:{approx_training_time_ms/(step + 1):.2f}ms", console=True)

print0(f"peak memory allocated: {torch.cuda.max_memory_allocated() // 1024 // 1024} MiB "
       f"reserved: {torch.cuda.max_memory_reserved() // 1024 // 1024} MiB", console=True)
dist.destroy_process_group()