import math
from dataclasses import dataclass
from typing import Tuple, Optional, Literal
from functools import lru_cache
from contextlib import contextmanager

import torch
from torch import nn
import torch.nn.functional as F
import torch.distributed as dist

from kernel import act_quant, fp4_act_quant, fp8_gemm, fp4_gemm, sparse_attn, hc_split_sinkhorn


world_size = 1
rank = 0
block_size = 128
fp4_block_size = 32
w4a16_group_size = 128
default_dtype = torch.bfloat16
scale_fmt = None
scale_dtype = torch.float32
w4a16_mode = False  # set in Transformer.__init__ when args.dtype == "w4a16"


def dequantize_w4a16(qweight: torch.Tensor, qzeros: torch.Tensor, scales: torch.Tensor,
                     group_size: int = 128) -> torch.Tensor:
    """Auto-round / auto_gptq W4A16 packing -> BF16 weight [out, in].

    qweight: int32 [in/8, out], LSB-first 4-bit packed along input dim
    qzeros : int32 [in/g, out/8], LSB-first 4-bit packed along output dim
    scales : bf16 [in/g, out]
    """
    in_packed, out_features = qweight.shape
    in_features = in_packed * 8
    n_groups = scales.shape[0]
    device = qweight.device
    shifts = torch.arange(0, 32, 4, device=device, dtype=torch.int32)
    w = (qweight.unsqueeze(1) >> shifts.view(1, 8, 1)) & 0xF       # [in/8, 8, out]
    w = w.reshape(in_features, out_features).to(torch.float32)
    z = (qzeros.unsqueeze(2) >> shifts.view(1, 1, 8)) & 0xF        # [in/g, out/8, 8]
    z = z.reshape(n_groups, out_features).to(torch.float32) + 1.0  # GPTQ stores zero - 1
    s = scales.to(torch.float32)
    w = w.view(n_groups, group_size, out_features)
    deq = (w - z.unsqueeze(1)) * s.unsqueeze(1)
    deq = deq.view(in_features, out_features)
    return deq.t().contiguous().to(torch.bfloat16)


@contextmanager
def set_dtype(dtype):
    """Temporarily override torch default dtype, restoring it on exit (even if an exception occurs)."""
    prev = torch.get_default_dtype()
    torch.set_default_dtype(dtype)
    try:
        yield
    finally:
        torch.set_default_dtype(prev)

@dataclass
class ModelArgs:
    """Model hyperparameters. Field names match the config JSON keys."""
    max_batch_size: int = 4
    max_seq_len: int = 4096
    dtype: Literal["bf16", "fp8", "w4a16"] = "fp8"
    scale_fmt: Literal[None, "ue8m0"] = "ue8m0"
    expert_dtype: Literal[None, "fp4"] = None
    scale_dtype: Literal["fp32", "fp8"] = "fp8"
    vocab_size: int = 129280
    dim: int = 4096
    moe_inter_dim: int = 4096
    n_layers: int = 7
    n_hash_layers: int = 0
    n_mtp_layers: int = 1
    n_heads: int = 64
    # moe
    n_routed_experts: int = 8
    n_shared_experts: int = 1
    n_activated_experts: int = 2
    score_func: Literal["softmax", "sigmoid", "sqrtsoftplus"] = "sqrtsoftplus"
    route_scale: float = 1.
    swiglu_limit: float = 0.
    # mqa
    q_lora_rank: int = 1024
    head_dim: int = 512
    rope_head_dim: int = 64
    norm_eps: float = 1e-6
    o_groups: int = 8
    o_lora_rank: int = 1024
    window_size: int = 128
    compress_ratios: Tuple[int] = (0, 0, 4, 128, 4, 128, 4, 0)
    # yarn
    compress_rope_theta: float = 40000.0
    original_seq_len: int = 0
    rope_theta: float = 10000.0
    rope_factor: float = 40
    beta_fast: int = 32
    beta_slow: int = 1
    # index
    index_n_heads: int = 64
    index_head_dim: int = 128
    index_topk: int = 512
    # hc
    hc_mult: int = 4
    hc_sinkhorn_iters: int = 20
    hc_eps: float = 1e-6


class ParallelEmbedding(nn.Module):
    """Embedding sharded along the vocab dimension. Each rank holds vocab_size // world_size rows.
    Out-of-range indices are zero-masked before all_reduce to combine partial embeddings."""
    def __init__(self, vocab_size: int, dim: int):
        super().__init__()
        self.vocab_size = vocab_size
        self.dim = dim
        assert vocab_size % world_size == 0, f"Vocabulary size must be divisible by world size (world_size={world_size})"
        self.part_vocab_size = (vocab_size // world_size)
        self.vocab_start_idx = rank * self.part_vocab_size
        self.vocab_end_idx = self.vocab_start_idx + self.part_vocab_size
        self.weight = nn.Parameter(torch.empty(self.part_vocab_size, self.dim))

    def forward(self, x: torch.Tensor) -> torch.Tensor:
        if world_size > 1:
            mask = (x < self.vocab_start_idx) | (x >= self.vocab_end_idx)
            x = x - self.vocab_start_idx
            x[mask] = 0
        y = F.embedding(x, self.weight)
        if world_size > 1:
            y[mask] = 0
            dist.all_reduce(y)
        return y


def linear(x: torch.Tensor, weight: torch.Tensor, bias: Optional[torch.Tensor] = None) -> torch.Tensor:
    """Dispatches to fp4_gemm / fp8_gemm / F.linear based on weight dtype.
    For quantized weights, x is first quantized to FP8 via act_quant."""
    assert bias is None

    if weight.dtype == torch.float4_e2m1fn_x2:
        x, s = act_quant(x, block_size, scale_fmt, scale_dtype)
        return fp4_gemm(x, s, weight, weight.scale, scale_dtype)
    elif weight.dtype == torch.float8_e4m3fn:
        x, s = act_quant(x, block_size, scale_fmt, scale_dtype)
        return fp8_gemm(x, s, weight, weight.scale, scale_dtype)
    else:
        return F.linear(x, weight)


class Linear(nn.Module):
    """Linear layer supporting BF16, FP8, and FP4 weight formats with per-block scaling."""

    def __init__(self, in_features: int, out_features: int, bias: bool = False, dtype = None):
        super().__init__()
        self.in_features = in_features
        self.out_features = out_features
        # In a W4A16 build every Linear becomes W4A16 regardless of the dtype the
        # original FP8/FP4 model wanted. The non-quant special cases (RMSNorm,
        # embed, attn_sink, etc.) are NOT instances of `Linear`, so they are
        # untouched.
        if w4a16_mode:
            dtype = "w4a16"
        else:
            dtype = dtype or default_dtype
        self.is_w4a16 = (dtype == "w4a16")
        if self.is_w4a16:
            assert in_features % 8 == 0 and in_features % w4a16_group_size == 0
            assert out_features % 8 == 0
            self.group_size = w4a16_group_size
            self.qweight = nn.Parameter(
                torch.empty(in_features // 8, out_features, dtype=torch.int32),
                requires_grad=False,
            )
            self.qzeros = nn.Parameter(
                torch.empty(in_features // self.group_size, out_features // 8, dtype=torch.int32),
                requires_grad=False,
            )
            self.scales = nn.Parameter(
                torch.empty(in_features // self.group_size, out_features, dtype=torch.bfloat16),
                requires_grad=False,
            )
            self.register_parameter("weight", None)
            self.register_parameter("scale", None)
        elif dtype == torch.float4_e2m1fn_x2:
            # FP4: weight is [out, in//2] in float4_e2m1fn_x2, logically [out, in] in fp4
            # Scale is [out, in//32] in float8_e8m0fnu (1 scale per 32 fp4 elements along K)
            self.weight = nn.Parameter(torch.empty(out_features, in_features // 2, dtype=torch.float4_e2m1fn_x2))
            scale_out_features = out_features
            scale_in_features = in_features // fp4_block_size
            self.weight.scale = self.scale = nn.Parameter(torch.empty(scale_out_features, scale_in_features, dtype=torch.float8_e8m0fnu))
        elif dtype == torch.float8_e4m3fn:
            self.weight = nn.Parameter(torch.empty(out_features, in_features, dtype=dtype))
            scale_out_features = (out_features + block_size - 1) // block_size
            scale_in_features = (in_features + block_size - 1) // block_size
            self.weight.scale = self.scale = nn.Parameter(torch.empty(scale_out_features, scale_in_features, dtype=torch.float8_e8m0fnu))
        else:
            self.weight = nn.Parameter(torch.empty(out_features, in_features, dtype=dtype))
            self.register_parameter("scale", None)
        if bias:
            self.bias = nn.Parameter(torch.empty(out_features))
        else:
            self.register_parameter("bias", None)

    def init_woq(self, QuantLinear):
        """Create a QuantLinear from loaded GPTQ parameters."""
        if not self.is_w4a16:
            return
        # Marlin requires out_features % 64 == 0; fall back to manual dequant
        if self.out_features % 64 != 0:
            self._woq = None
            return
        dev = self.qweight.device
        layer = QuantLinear(
            bits=4, group_size=self.group_size,
            in_features=self.in_features, out_features=self.out_features,
            bias=False, desc_act=False, sym=True, register_buffers=True,
        )
        layer = layer.to(dev)
        layer.qweight.copy_(self.qweight.data)
        layer.qzeros.copy_(self.qzeros.data)
        layer.scales.copy_(self.scales.to(layer.scales.dtype).data)
        layer.g_idx.copy_(torch.arange(self.in_features, dtype=torch.int32, device=dev) // self.group_size)
        layer.post_init()
        self._woq = layer
        # Free original parameters to save memory
        self.qweight = None
        self.qzeros = None
        self.scales = None

    def get_weight(self) -> torch.Tensor:
        """Return the dequantised BF16 weight [out, in]. For non-W4A16 modes
        returns ``self.weight`` unchanged. Used only for wo_a einsum path."""
        if self.is_w4a16:
            if self._woq is not None:
                return dequantize_w4a16(self._woq.qweight, self._woq.qzeros, self._woq.scales, self.group_size)
            return dequantize_w4a16(self.qweight, self.qzeros, self.scales, self.group_size)
        return self.weight

    def forward(self, x: torch.Tensor) -> torch.Tensor:
        if self.is_w4a16:
            if hasattr(self, '_woq') and self._woq is not None:
                y = self._woq(x.to(torch.bfloat16))
            else:
                w = dequantize_w4a16(self.qweight, self.qzeros, self.scales, self.group_size)
                y = F.linear(x.to(w.dtype), w)
            if self.bias is not None:
                y = y + self.bias
            return y.type_as(x)
        return linear(x, self.weight, self.bias)


class ColumnParallelLinear(Linear):
    """Shards output dim across TP ranks. No all-reduce needed on output."""
    def __init__(self, in_features: int, out_features: int, bias: bool = False, dtype = None):
        assert out_features % world_size == 0, f"Output features must be divisible by world size (world_size={world_size})"
        self.part_out_features = out_features // world_size
        super().__init__(in_features, self.part_out_features, bias, dtype)

    def forward(self, x: torch.Tensor) -> torch.Tensor:
        if self.is_w4a16:
            return Linear.forward(self, x)
        return linear(x, self.weight, self.bias)


class RowParallelLinear(Linear):
    """Shards input dim across TP ranks. All-reduce on output to sum partial results."""
    def __init__(self, in_features: int, out_features: int, bias: bool = False, dtype = None):
        assert in_features % world_size == 0, f"Input features must be divisible by world size (world_size={world_size})"
        self.part_in_features = in_features // world_size
        super().__init__(self.part_in_features, out_features, bias, dtype)

    def forward(self, x: torch.Tensor) -> torch.Tensor:
        if self.is_w4a16:
            if hasattr(self, '_woq') and self._woq is not None:
                y = self._woq(x.to(torch.bfloat16))
            else:
                w = dequantize_w4a16(self.qweight, self.qzeros, self.scales, self.group_size)
                y = F.linear(x.to(w.dtype), w)
        else:
            y = linear(x, self.weight, None)
        if world_size > 1:
            y = y.float()
            dist.all_reduce(y)
        if self.bias is not None:
            y += self.bias
        return y.type_as(x)


class RMSNorm(nn.Module):
    def __init__(self, dim: int, eps: float = 1e-6):
        super().__init__()
        self.dim = dim
        self.eps = eps
        # rmsnorm in the checkpoint is stored in bf16, while the parameter here is stored in fp32 for convenient.
        self.weight = nn.Parameter(torch.ones(dim, dtype=torch.float32))

    def forward(self, x: torch.Tensor):
        dtype = x.dtype
        x = x.float()
        var = x.square().mean(-1, keepdim=True)
        x = x * torch.rsqrt(var + self.eps)
        return (self.weight * x).to(dtype)


@lru_cache(2)
def precompute_freqs_cis(dim, seqlen, original_seq_len, base, factor, beta_fast, beta_slow) -> torch.Tensor:
    """Precomputes complex exponentials for rotary embeddings with YaRN scaling.
    When original_seq_len > 0, applies frequency interpolation with a smooth
    linear ramp between beta_fast and beta_slow correction ranges."""

    def find_correction_dim(num_rotations, dim, base, max_seq_len):
        return dim * math.log(max_seq_len / (num_rotations * 2 * math.pi)) / (2 * math.log(base))

    def find_correction_range(low_rot, high_rot, dim, base, max_seq_len):
        low = math.floor(find_correction_dim(low_rot, dim, base, max_seq_len))
        high = math.ceil(find_correction_dim(high_rot, dim, base, max_seq_len))
        return max(low, 0), min(high, dim-1)

    def linear_ramp_factor(min, max, dim):
        if min == max:
            max += 0.001
        linear_func = (torch.arange(dim, dtype=torch.float32) - min) / (max - min)
        ramp_func = torch.clamp(linear_func, 0, 1)
        return ramp_func

    freqs = 1.0 / (base ** (torch.arange(0, dim, 2, dtype=torch.float32) / dim))
    if original_seq_len > 0:
        low, high = find_correction_range(beta_fast, beta_slow, dim, base, original_seq_len)
        smooth = 1 - linear_ramp_factor(low, high, dim // 2)
        freqs = freqs / factor * (1 - smooth) + freqs * smooth

    t = torch.arange(seqlen)
    freqs = torch.outer(t, freqs)
    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)
    return freqs_cis


def apply_rotary_emb(x: torch.Tensor, freqs_cis: torch.Tensor, inverse: bool = False) -> torch.Tensor:
    """Applies rotary positional embeddings in-place. Uses conjugate for inverse (de-rotation)."""
    y = x
    x = torch.view_as_complex(x.float().unflatten(-1, (-1, 2)))
    if inverse:
        freqs_cis = freqs_cis.conj()
    if x.ndim == 3:
        freqs_cis = freqs_cis.view(1, x.size(1), x.size(-1))
    else:
        freqs_cis = freqs_cis.view(1, x.size(1), 1, x.size(-1))
    x = torch.view_as_real(x * freqs_cis).flatten(-2)
    y.copy_(x)
    return y


def rotate_activation(x: torch.Tensor) -> torch.Tensor:
    """Applies randomized Hadamard rotation to spread information across dims before FP8 quant."""
    assert x.dtype == torch.bfloat16
    from fast_hadamard_transform import hadamard_transform
    return hadamard_transform(x, scale=x.size(-1) ** -0.5)


@lru_cache(1)
def get_window_topk_idxs(window_size: int, bsz: int, seqlen: int, start_pos: int):
    if start_pos >= window_size - 1:
        start_pos %= window_size
        matrix = torch.cat([torch.arange(start_pos + 1, window_size),  torch.arange(0, start_pos + 1)], dim=0)
    elif start_pos > 0:
        matrix = F.pad(torch.arange(start_pos + 1), (0, window_size - start_pos - 1), value=-1)
    else:
        base = torch.arange(seqlen).unsqueeze(1)
        matrix = (base - window_size + 1).clamp(0) + torch.arange(min(seqlen, window_size))
        matrix = torch.where(matrix > base, -1, matrix)
    return matrix.unsqueeze(0).expand(bsz, -1, -1)


@lru_cache(2)
def get_compress_topk_idxs(ratio: int, bsz: int, seqlen: int, start_pos: int, offset: int):
    if start_pos > 0:
        matrix = torch.arange(0, (start_pos + 1) // ratio) + offset
    else:
        matrix = torch.arange(seqlen // ratio).repeat(seqlen, 1)
        mask = matrix >= torch.arange(1, seqlen + 1).unsqueeze(1) // ratio
        matrix = torch.where(mask, -1, matrix + offset)
    return matrix.unsqueeze(0).expand(bsz, -1, -1)


class Compressor(nn.Module):
    """Compresses KV cache via learned gated pooling over `compress_ratio` consecutive tokens.
    When overlap=True (ratio==4), uses overlapping windows for smoother compression boundaries."""

    def __init__(self, args: ModelArgs, compress_ratio: int = 4, head_dim: int = 512, rotate: bool = False):
        super().__init__()
        self.dim = args.dim
        self.head_dim = head_dim
        self.rope_head_dim = args.rope_head_dim
        self.nope_head_dim = head_dim - args.rope_head_dim
        self.compress_ratio = compress_ratio
        self.overlap = compress_ratio == 4
        self.rotate = rotate
        coff = 1 + self.overlap

        self.ape = nn.Parameter(torch.empty(compress_ratio, coff * self.head_dim, dtype=torch.float32))
        # wkv and wgate in the checkpoint is stored in bf16, while the parameter here is stored in fp32 for convenient.
        # When overlap, the first half of dims is for overlapping compression, second half for normal.
        self.wkv = Linear(self.dim, coff * self.head_dim, dtype=torch.float32)
        self.wgate = Linear(self.dim, coff * self.head_dim, dtype=torch.float32)
        self.norm = RMSNorm(self.head_dim, args.norm_eps)
        self.kv_cache: torch.Tensor = None  # assigned lazily from Attention.kv_cache
        # State buffers for decode-phase incremental compression.
        # With overlap: state[:, :ratio] = overlapping window, state[:, ratio:] = current window.
        self.register_buffer("kv_state", torch.zeros(args.max_batch_size, coff * compress_ratio, coff * self.head_dim, dtype=torch.float32), persistent=False)
        self.register_buffer("score_state", torch.full((args.max_batch_size, coff * compress_ratio, coff * self.head_dim), float("-inf"), dtype=torch.float32), persistent=False)
        self.freqs_cis: torch.Tensor = None

    def overlap_transform(self, tensor: torch.Tensor, value=0):
        # tensor: [b,s,r,2d]
        b, s, _, _ = tensor.size()
        ratio, d = self.compress_ratio, self.head_dim
        new_tensor = tensor.new_full((b, s, 2 * ratio, d), value)
        new_tensor[:, :, ratio:] = tensor[:, :, :, d:]
        new_tensor[:, 1:, :ratio] = tensor[:, :-1, :, :d]
        return new_tensor

    def forward(self, x: torch.Tensor, start_pos: int):
        assert self.kv_cache is not None
        bsz, seqlen, _ = x.size()
        ratio, overlap, d, rd = self.compress_ratio, self.overlap, self.head_dim, self.rope_head_dim
        dtype = x.dtype
        # compression need fp32
        x = x.float()
        kv = self.wkv(x)
        score = self.wgate(x)
        if start_pos == 0:
            should_compress = seqlen >= ratio
            remainder = seqlen % ratio
            cutoff = seqlen - remainder
            offset = ratio if overlap else 0
            if overlap and cutoff >= ratio:
                self.kv_state[:bsz, :ratio] = kv[:, cutoff-ratio : cutoff]
                self.score_state[:bsz, :ratio] = score[:, cutoff-ratio : cutoff] + self.ape
            if remainder > 0:
                kv, self.kv_state[:bsz, offset : offset+remainder] = kv.split([cutoff, remainder], dim=1)
                self.score_state[:bsz, offset : offset+remainder] = score[:, cutoff:] + self.ape[:remainder]
                score = score[:, :cutoff]
            kv = kv.unflatten(1, (-1, ratio))
            score = score.unflatten(1, (-1, ratio)) + self.ape
            if overlap:
                kv = self.overlap_transform(kv, 0)
                score = self.overlap_transform(score, float("-inf"))
            kv = (kv * score.softmax(dim=2)).sum(dim=2)
        else:
            should_compress = (start_pos + 1) % self.compress_ratio == 0
            score += self.ape[start_pos % ratio]
            if overlap:
                self.kv_state[:bsz, ratio + start_pos % ratio] = kv.squeeze(1)
                self.score_state[:bsz, ratio + start_pos % ratio] = score.squeeze(1)
                if should_compress:
                    kv_state = torch.cat([self.kv_state[:bsz, :ratio, :d], self.kv_state[:bsz, ratio:, d:]], dim=1)
                    score_state = torch.cat([self.score_state[:bsz, :ratio, :d], self.score_state[:bsz, ratio:, d:]], dim=1)
                    kv = (kv_state * score_state.softmax(dim=1)).sum(dim=1, keepdim=True)
                    self.kv_state[:bsz, :ratio] = self.kv_state[:bsz, ratio:]
                    self.score_state[:bsz, :ratio] = self.score_state[:bsz, ratio:]
            else:
                self.kv_state[:bsz, start_pos % ratio] = kv.squeeze(1)
                self.score_state[:bsz, start_pos % ratio] = score.squeeze(1)
                if should_compress:
                    kv = (self.kv_state[:bsz] * self.score_state[:bsz].softmax(dim=1)).sum(dim=1, keepdim=True)
        if not should_compress:
            return
        kv = self.norm(kv.to(dtype))
        if start_pos == 0:
            freqs_cis = self.freqs_cis[:cutoff:ratio]
        else:
            freqs_cis = self.freqs_cis[start_pos + 1 - self.compress_ratio].unsqueeze(0)
        apply_rotary_emb(kv[..., -rd:], freqs_cis)
        if self.rotate:
            kv = rotate_activation(kv)
            fp4_act_quant(kv, fp4_block_size, True)
        else:
            act_quant(kv[..., :-rd], 64, scale_fmt, scale_dtype, True)
        if start_pos == 0:
            self.kv_cache[:bsz, :seqlen // ratio] = kv
        else:
            self.kv_cache[:bsz, start_pos // ratio] = kv.squeeze(1)
        return kv


class Indexer(torch.nn.Module):
    """Selects top-k compressed KV positions for sparse attention via learned scoring.
    Has its own Compressor (with Hadamard rotation) to build compressed KV for scoring."""

    def __init__(self, args: ModelArgs, compress_ratio: int = 4):
        super().__init__()
        self.dim = args.dim
        self.n_heads = args.index_n_heads
        self.n_local_heads = args.index_n_heads // world_size
        self.head_dim = args.index_head_dim
        self.rope_head_dim = args.rope_head_dim
        self.index_topk = args.index_topk
        self.q_lora_rank = args.q_lora_rank
        self.wq_b = ColumnParallelLinear(self.q_lora_rank, self.n_heads * self.head_dim)
        self.weights_proj = ColumnParallelLinear(self.dim, self.n_heads, dtype=torch.bfloat16)
        self.softmax_scale = self.head_dim ** -0.5
        self.compress_ratio = compress_ratio

        self.compressor = Compressor(args, compress_ratio, self.head_dim, True)
        self.register_buffer("kv_cache", torch.zeros(args.max_batch_size, args.max_seq_len // compress_ratio, self.head_dim), persistent=False)
        self.freqs_cis = None

    def forward(self, x: torch.Tensor, qr: torch.Tensor, start_pos: int, offset: int):
        bsz, seqlen, _ = x.size()
        freqs_cis = self.freqs_cis[start_pos:start_pos+seqlen]
        ratio = self.compress_ratio
        rd = self.rope_head_dim
        end_pos = start_pos + seqlen
        if self.compressor.kv_cache is None:
            self.compressor.kv_cache = self.kv_cache
            self.compressor.freqs_cis = self.freqs_cis
        q = self.wq_b(qr)
        q = q.unflatten(-1, (self.n_local_heads, self.head_dim))
        apply_rotary_emb(q[..., -rd:], freqs_cis)
        q = rotate_activation(q)
        # use fp4 simulation for q and kv in indexer
        fp4_act_quant(q, fp4_block_size, True)
        self.compressor(x, start_pos)
        weights = self.weights_proj(x) * (self.softmax_scale * self.n_heads ** -0.5)
        # We performed QAT here, kv could also use fp8 format, though current implementation uses bf16
        index_score = torch.einsum("bshd,btd->bsht", q, self.kv_cache[:bsz, :end_pos // ratio])
        index_score = (index_score.relu_() * weights.unsqueeze(-1)).sum(dim=2)
        if world_size > 1:
            dist.all_reduce(index_score)
        if start_pos == 0:
            mask = torch.arange(seqlen // ratio).repeat(seqlen, 1) >= torch.arange(1, seqlen + 1).unsqueeze(1) // ratio
            index_score += torch.where(mask, float("-inf"), 0)
        topk_idxs = index_score.topk(min(self.index_topk, end_pos // ratio), dim=-1)[1]
        if start_pos == 0:
            mask = topk_idxs >= torch.arange(1, seqlen + 1).unsqueeze(1) // ratio
            topk_idxs = torch.where(mask, -1, topk_idxs + offset)
        else:
            topk_idxs += offset
        return topk_idxs


class Attention(nn.Module):
    """Multi-head Latent Attention (MLA) with sliding window + optional KV compression.
    Uses low-rank Q projection (wq_a -> q_norm -> wq_b) and grouped low-rank O projection."""
    def __init__(self, layer_id: int, args: ModelArgs):
        super().__init__()
        self.layer_id = layer_id
        self.dim = args.dim
        self.n_heads = args.n_heads
        self.n_local_heads = args.n_heads // world_size
        self.q_lora_rank = args.q_lora_rank
        self.o_lora_rank = args.o_lora_rank
        self.head_dim = args.head_dim
        self.rope_head_dim = args.rope_head_dim
        self.nope_head_dim = args.head_dim - args.rope_head_dim
        self.n_groups = args.o_groups
        self.n_local_groups = self.n_groups // world_size
        self.window_size = args.window_size
        self.compress_ratio = args.compress_ratios[layer_id]
        self.eps = args.norm_eps

        self.attn_sink = nn.Parameter(torch.empty(self.n_local_heads, dtype=torch.float32))
        self.wq_a = Linear(self.dim, self.q_lora_rank)
        self.q_norm = RMSNorm(self.q_lora_rank, self.eps)
        self.wq_b = ColumnParallelLinear(self.q_lora_rank, self.n_heads * self.head_dim)
        self.wkv = Linear(self.dim, self.head_dim)
        self.kv_norm = RMSNorm(self.head_dim, self.eps)
        self.wo_a = ColumnParallelLinear(self.n_heads * self.head_dim // self.n_groups, self.n_groups * args.o_lora_rank, dtype=torch.bfloat16)
        self.wo_b = RowParallelLinear(self.n_groups * args.o_lora_rank, self.dim)
        self.softmax_scale = self.head_dim ** -0.5

        if self.compress_ratio:
            self.compressor = Compressor(args, self.compress_ratio, self.head_dim)
            if self.compress_ratio == 4:
                self.indexer = Indexer(args, self.compress_ratio)
            else:
                self.indexer = None

        kv_cache_size = args.window_size + (args.max_seq_len // self.compress_ratio if self.compress_ratio else 0)
        self.register_buffer("kv_cache", torch.zeros(args.max_batch_size, kv_cache_size, self.head_dim), persistent=False)
        if self.compress_ratio:
            original_seq_len, rope_theta = args.original_seq_len, args.compress_rope_theta
        else:
            # disable YaRN and use base rope_theta in pure sliding-window attention
            original_seq_len, rope_theta = 0, args.rope_theta
        freqs_cis = precompute_freqs_cis(self.rope_head_dim, args.max_seq_len, original_seq_len,
                                         rope_theta, args.rope_factor, args.beta_fast, args.beta_slow)
        self.register_buffer("freqs_cis", freqs_cis, persistent=False)

    def forward(self, x: torch.Tensor, start_pos: int):
        bsz, seqlen, _ = x.size()
        freqs_cis = self.freqs_cis[start_pos:start_pos+seqlen]
        win = self.window_size
        ratio = self.compress_ratio
        rd = self.rope_head_dim
        if self.compress_ratio and self.compressor.kv_cache is None:
            self.compressor.kv_cache = self.kv_cache[:, win:]
            self.compressor.freqs_cis = self.freqs_cis
            if self.indexer is not None:
                self.indexer.freqs_cis = self.freqs_cis
        # q
        qr = q = self.q_norm(self.wq_a(x))
        q = self.wq_b(q).unflatten(-1, (self.n_local_heads, self.head_dim))
        q *= torch.rsqrt(q.square().mean(-1, keepdim=True) + self.eps)
        apply_rotary_emb(q[..., -rd:], freqs_cis)

        # win kv & topk_idxs
        kv = self.wkv(x)
        kv = self.kv_norm(kv)
        apply_rotary_emb(kv[..., -rd:], freqs_cis)
        # FP8-simulate non-rope dims to match QAT; rope dims stay bf16 for positional precision
        act_quant(kv[..., :-rd], 64, scale_fmt, scale_dtype, True)
        topk_idxs = get_window_topk_idxs(win, bsz, seqlen, start_pos)
        if self.compress_ratio:
            offset = kv.size(1) if start_pos == 0 else win
            if self.indexer is not None:
                compress_topk_idxs = self.indexer(x, qr, start_pos, offset)
            else:
                compress_topk_idxs = get_compress_topk_idxs(ratio, bsz, seqlen, start_pos, offset)
            topk_idxs = torch.cat([topk_idxs, compress_topk_idxs], dim=-1)
        topk_idxs = topk_idxs.int()

        # compress kv & attn
        if start_pos == 0:
            if seqlen <= win:
                self.kv_cache[:bsz, :seqlen] = kv
            else:
                cutoff = seqlen % win
                self.kv_cache[:bsz, cutoff: win], self.kv_cache[:bsz, :cutoff] = kv[:, -win:].split([win - cutoff, cutoff], dim=1)
            if self.compress_ratio:
                if (kv_compress := self.compressor(x, start_pos)) is not None:
                    kv = torch.cat([kv, kv_compress], dim=1)
            # We performed QAT here, kv could also use fp8 format, though current implementation uses bf16
            o = sparse_attn(q, kv, self.attn_sink, topk_idxs, self.softmax_scale)
        else:
            self.kv_cache[:bsz, start_pos % win] = kv.squeeze(1)
            if self.compress_ratio:
                self.compressor(x, start_pos)
            o = sparse_attn(q, self.kv_cache[:bsz], self.attn_sink, topk_idxs, self.softmax_scale)
        apply_rotary_emb(o[..., -rd:], freqs_cis, True)

        # o: apply wo_a per-group projection then wo_b
        # Flatten groups into the feature dim, call wo_a as a normal linear, then reshape back.
        # Equivalent to the per-group einsum when wo_a weight is block-diagonal across groups
        # (always true here since n_local_groups = n_groups/world_size = 1 for 8-GPU deploy).
        o = o.view(bsz, seqlen, self.n_local_groups, -1)
        o = self.wo_a(o.flatten(2)).view(bsz, seqlen, self.n_local_groups, self.o_lora_rank)
        x = self.wo_b(o.flatten(2))
        return x


class Gate(nn.Module):
    """MoE gating: computes expert routing scores and selects top-k experts.
    Supports hash-based routing (first n_hash_layers) where expert indices are
    predetermined per token ID, and score-based routing (remaining layers)."""
    def __init__(self, layer_id: int, args: ModelArgs):
        super().__init__()
        self.dim = args.dim
        self.topk = args.n_activated_experts
        self.score_func = args.score_func
        self.route_scale = args.route_scale
        self.hash = layer_id < args.n_hash_layers
        self.is_w4a16 = w4a16_mode
        if self.is_w4a16:
            in_f, out_f = args.dim, args.n_routed_experts
            assert in_f % w4a16_group_size == 0 and out_f % 8 == 0
            self.group_size = w4a16_group_size
            self.qweight = nn.Parameter(
                torch.empty(in_f // 8, out_f, dtype=torch.int32), requires_grad=False)
            self.qzeros = nn.Parameter(
                torch.empty(in_f // self.group_size, out_f // 8, dtype=torch.int32), requires_grad=False)
            self.scales = nn.Parameter(
                torch.empty(in_f // self.group_size, out_f, dtype=torch.bfloat16), requires_grad=False)
            self.register_parameter("weight", None)
        else:
            self.weight = nn.Parameter(torch.empty(args.n_routed_experts, args.dim))
        if self.hash:
            self.tid2eid = nn.Parameter(torch.empty(args.vocab_size, args.n_activated_experts, dtype=torch.int32), requires_grad=False)
            self.bias = None
        else:
            self.bias = nn.Parameter(torch.empty(args.n_routed_experts, dtype=torch.float32))

    def init_woq(self, QuantLinear):
        if not self.is_w4a16:
            return
        dev = self.qweight.device
        in_f, out_f = self.dim, self.qweight.shape[1]
        if out_f % 64 != 0:
            self._woq = None
            return
        layer = QuantLinear(
            bits=4, group_size=self.group_size,
            in_features=in_f, out_features=out_f,
            bias=False, desc_act=False, sym=True, register_buffers=True,
        )
        layer = layer.to(dev)
        layer.qweight.copy_(self.qweight.data)
        layer.qzeros.copy_(self.qzeros.data)
        layer.scales.copy_(self.scales.to(layer.scales.dtype).data)
        layer.g_idx.copy_(torch.arange(in_f, dtype=torch.int32, device=dev) // self.group_size)
        layer.post_init()
        self._woq = layer
        self.qweight = None
        self.qzeros = None
        self.scales = None

    def forward(self, x: torch.Tensor, input_ids: Optional[torch.Tensor] = None) -> Tuple[torch.Tensor, torch.Tensor]:
        if self.is_w4a16:
            if hasattr(self, '_woq') and self._woq is not None:
                scores = self._woq(x.to(torch.bfloat16)).float()
            else:
                w = dequantize_w4a16(self.qweight, self.qzeros, self.scales, self.group_size)
                scores = F.linear(x.to(w.dtype), w).float()
        else:
            scores = linear(x.float(), self.weight.float())
        if self.score_func == "softmax":
            scores = scores.softmax(dim=-1)
        elif self.score_func == "sigmoid":
            scores = scores.sigmoid()
        else:
            scores = F.softplus(scores).sqrt()
        original_scores = scores
        # Bias shifts scores for expert selection (topk) but does not affect routing weights.
        if self.bias is not None:
            scores = scores + self.bias
        if self.hash:
            indices = self.tid2eid[input_ids]
        else:
            indices = scores.topk(self.topk, dim=-1)[1]
        weights = original_scores.gather(1, indices)
        if self.score_func != "softmax":
            weights /= weights.sum(dim=-1, keepdim=True)
        weights *= self.route_scale
        return weights, indices


class Expert(nn.Module):
    """Single MoE expert: SwiGLU FFN (w1, w2, w3). Computation in float32 for stability."""
    def __init__(self, dim: int, inter_dim: int, dtype=None, swiglu_limit=0):
        super().__init__()
        self.w1 = Linear(dim, inter_dim, dtype=dtype)
        self.w2 = Linear(inter_dim, dim, dtype=dtype)
        self.w3 = Linear(dim, inter_dim, dtype=dtype)
        self.swiglu_limit = swiglu_limit

    def forward(self, x: torch.Tensor, weights: Optional[torch.Tensor] = None) -> torch.Tensor:
        dtype = x.dtype
        gate = self.w1(x).float()
        up = self.w3(x).float()
        if self.swiglu_limit > 0:
            up = torch.clamp(up, min=-self.swiglu_limit, max=self.swiglu_limit)
            gate = torch.clamp(gate, max=self.swiglu_limit)
        x = F.silu(gate) * up
        if weights is not None:
            x = weights * x
        return self.w2(x.to(dtype))


class MoE(nn.Module):
    """Mixture-of-Experts: gate routes each token to top-k routed experts + 1 shared expert.
    Experts are sharded across TP ranks; each rank handles n_routed_experts // world_size experts."""
    def __init__(self, layer_id: int, args: ModelArgs):
        super().__init__()
        self.layer_id = layer_id
        self.dim = args.dim
        assert args.n_routed_experts % world_size == 0, f"Number of experts must be divisible by world size (world_size={world_size})"
        self.n_routed_experts = args.n_routed_experts
        self.n_local_experts = args.n_routed_experts // world_size
        self.n_activated_experts = args.n_activated_experts
        self.experts_start_idx = rank * self.n_local_experts
        self.experts_end_idx = self.experts_start_idx + self.n_local_experts
        self.gate = Gate(layer_id, args)
        expert_dtype = torch.float4_e2m1fn_x2 if args.expert_dtype == "fp4" else None
        self.experts = nn.ModuleList([Expert(args.dim, args.moe_inter_dim, dtype=expert_dtype, swiglu_limit=args.swiglu_limit) if self.experts_start_idx <= i < self.experts_end_idx else None
                                       for i in range(self.n_routed_experts)])
        assert args.n_shared_experts == 1
        # no swiglu_limit
        self.shared_experts = Expert(args.dim, args.moe_inter_dim, swiglu_limit=args.swiglu_limit)

    def forward(self, x: torch.Tensor, input_ids: torch.Tensor) -> torch.Tensor:
        shape = x.size()
        x = x.view(-1, self.dim)
        weights, indices = self.gate(x, input_ids.flatten())
        y = torch.zeros_like(x, dtype=torch.float32)
        counts = torch.bincount(indices.flatten(), minlength=self.n_routed_experts).tolist()
        for i in range(self.experts_start_idx, self.experts_end_idx):
            if counts[i] == 0:
                continue
            expert = self.experts[i]
            idx, top = torch.where(indices == i)
            y[idx] += expert(x[idx], weights[idx, top, None])
        if world_size > 1:
            dist.all_reduce(y)
        y += self.shared_experts(x)
        return y.type_as(x).view(shape)


class Block(nn.Module):
    """Transformer block with Hyper-Connections (HC) mixing.
    Instead of a simple residual, HC maintains `hc_mult` copies of the hidden state.
    hc_pre: reduces hc copies -> 1 via learned weighted sum (pre-weights from Sinkhorn).
    hc_post: expands 1 -> hc copies via learned post-weights + combination matrix."""
    def __init__(self, layer_id: int, args: ModelArgs):
        super().__init__()
        self.layer_id = layer_id
        self.norm_eps = args.norm_eps
        self.attn = Attention(layer_id, args)
        self.ffn = MoE(layer_id, args)
        self.attn_norm = RMSNorm(args.dim, self.norm_eps)
        self.ffn_norm = RMSNorm(args.dim, self.norm_eps)
        self.hc_mult = hc_mult = args.hc_mult
        self.hc_sinkhorn_iters = args.hc_sinkhorn_iters
        self.hc_eps = args.hc_eps
        mix_hc = (2 + hc_mult) * hc_mult
        hc_dim = hc_mult * args.dim
        with set_dtype(torch.float32):
            self.hc_attn_fn = nn.Parameter(torch.empty(mix_hc, hc_dim))
            self.hc_ffn_fn = nn.Parameter(torch.empty(mix_hc, hc_dim))
            self.hc_attn_base = nn.Parameter(torch.empty(mix_hc))
            self.hc_ffn_base = nn.Parameter(torch.empty(mix_hc))
            self.hc_attn_scale = nn.Parameter(torch.empty(3))
            self.hc_ffn_scale = nn.Parameter(torch.empty(3))

    def hc_pre(self, x: torch.Tensor, hc_fn: torch.Tensor, hc_scale: torch.Tensor, hc_base: torch.Tensor):
        # x: [b,s,hc,d], hc_fn: [mix_hc,hc*d], hc_scale: [3], hc_base: [mix_hc], y: [b,s,hc,d]
        shape, dtype = x.size(), x.dtype
        x = x.flatten(2).float()
        rsqrt = torch.rsqrt(x.square().mean(-1, keepdim=True) + self.norm_eps)
        mixes = F.linear(x, hc_fn) * rsqrt
        pre, post, comb = hc_split_sinkhorn(mixes, hc_scale, hc_base, self.hc_mult, self.hc_sinkhorn_iters, self.hc_eps)
        y = torch.sum(pre.unsqueeze(-1) * x.view(shape), dim=2)
        return y.to(dtype), post, comb

    def hc_post(self, x: torch.Tensor, residual: torch.Tensor, post: torch.Tensor, comb: torch.Tensor):
        # x: [b,s,d], residual: [b,s,hc,d], post: [b,s,hc], comb: [b,s,hc,hc], y: [b,s,hc,d]
        y = post.unsqueeze(-1) * x.unsqueeze(-2) + torch.sum(comb.unsqueeze(-1) * residual.unsqueeze(-2), dim=2)
        return y.type_as(x)

    def forward(self, x: torch.Tensor, start_pos: int, input_ids: Optional[torch.Tensor]) -> torch.Tensor:
        residual = x
        x, post, comb = self.hc_pre(x, self.hc_attn_fn, self.hc_attn_scale, self.hc_attn_base)
        x = self.attn_norm(x)
        x = self.attn(x, start_pos)
        x = self.hc_post(x, residual, post, comb)

        residual = x
        x, post, comb = self.hc_pre(x, self.hc_ffn_fn, self.hc_ffn_scale, self.hc_ffn_base)
        x = self.ffn_norm(x)
        x = self.ffn(x, input_ids)
        x = self.hc_post(x, residual, post, comb)
        return x


class ParallelHead(nn.Module):

    def __init__(self, vocab_size: int, dim: int, norm_eps: float = 1e-6, hc_eps: float = 1e-6):
        super().__init__()
        self.vocab_size = vocab_size
        self.dim = dim
        self.norm_eps = norm_eps
        self.hc_eps = hc_eps
        self.part_vocab_size = (vocab_size // world_size)
        # lm_head is always stored as bf16 (even in W4A16 checkpoints); use fp32 for logit precision
        self.weight = nn.Parameter(torch.empty(self.part_vocab_size, self.dim, dtype=torch.float32))

    def get_logits(self, x):
        return F.linear(x[:, -1].float(), self.weight)

    def forward(self, x: torch.Tensor, hc_fn: torch.Tensor, hc_scale: torch.Tensor, hc_base: torch.Tensor, norm: RMSNorm):
        # x: [b,s,hc,d]
        x = self.hc_head(x, hc_fn, hc_scale, hc_base)
        logits = self.get_logits(norm(x))
        if world_size > 1:
            all_logits = [torch.empty_like(logits) for _ in range(world_size)]
            dist.all_gather(all_logits, logits)
            logits = torch.cat(all_logits, dim=-1)
        return logits

    def hc_head(self, x: torch.Tensor, hc_fn: torch.Tensor, hc_scale: torch.Tensor, hc_base: torch.Tensor):
        shape, dtype = x.size(), x.dtype
        x = x.flatten(2).float()
        rsqrt = torch.rsqrt(x.square().mean(-1, keepdim=True) + self.norm_eps)
        mixes = F.linear(x, hc_fn) * rsqrt
        pre = torch.sigmoid(mixes * hc_scale + hc_base) + self.hc_eps
        y = torch.sum(pre.unsqueeze(-1) * x.view(shape), dim=2)
        return y.to(dtype)


class MTPBlock(Block):

    def __init__(self, layer_id: int, args: ModelArgs):
        super().__init__(layer_id, args)
        self.e_proj = Linear(args.dim, args.dim)
        self.h_proj = Linear(args.dim, args.dim)
        self.enorm = RMSNorm(args.dim, args.norm_eps)
        self.hnorm = RMSNorm(args.dim, args.norm_eps)
        self.norm = RMSNorm(args.dim, args.norm_eps)
        self.hc_mult = hc_mult = args.hc_mult
        hc_dim = hc_mult * args.dim
        with set_dtype(torch.float32):
            self.hc_head_fn = nn.Parameter(torch.empty(hc_mult, hc_dim))
            self.hc_head_base = nn.Parameter(torch.empty(hc_mult))
            self.hc_head_scale = nn.Parameter(torch.empty(1))
        self.embed: ParallelEmbedding = None
        self.head: ParallelHead = None

    @torch.inference_mode()
    def forward(self, x: torch.Tensor, start_pos: int, input_ids: torch.Tensor) -> torch.Tensor:
        # x: [b,s,hc,d]
        assert self.embed is not None and self.head is not None
        e = self.embed(input_ids)
        e = self.enorm(e)
        x = self.hnorm(x)
        x = self.e_proj(e).unsqueeze(2) + self.h_proj(x)
        x = super().forward(x, start_pos, input_ids)
        logits = self.head(x, self.hc_head_fn, self.hc_head_scale, self.hc_head_base, self.norm)
        return logits


class Transformer(nn.Module):
    """Full DeepSeek-V4 model: embed -> HC-expand -> N blocks -> HC-head -> logits.
    Sets global state (world_size, rank, default_dtype, scale_fmt, scale_dtype) in __init__."""
    def __init__(self, args: ModelArgs):
        global world_size, rank, default_dtype, scale_fmt, scale_dtype, w4a16_mode
        world_size = dist.get_world_size() if dist.is_initialized() else 1
        rank = dist.get_rank() if dist.is_initialized() else 0
        w4a16_mode = (args.dtype == "w4a16")
        if w4a16_mode:
            default_dtype = torch.bfloat16
            scale_fmt = None
            scale_dtype = torch.float32
        else:
            default_dtype = torch.float8_e4m3fn if args.dtype == "fp8" else torch.bfloat16
            scale_fmt = "ue8m0" if args.scale_dtype == "fp8" else args.scale_fmt
            scale_dtype = torch.float8_e8m0fnu if args.scale_dtype == "fp8" else torch.float32
        super().__init__()
        self.max_seq_len = args.max_seq_len
        self.norm_eps = args.norm_eps
        self.hc_eps = args.hc_eps
        self.embed = ParallelEmbedding(args.vocab_size, args.dim)
        self.layers = torch.nn.ModuleList()
        for layer_id in range(args.n_layers):
            self.layers.append(Block(layer_id, args))
        self.norm = RMSNorm(args.dim, self.norm_eps)
        self.head = ParallelHead(args.vocab_size, args.dim, self.norm_eps, self.hc_eps)
        self.mtp = torch.nn.ModuleList()
        for layer_id in range(args.n_mtp_layers):
            self.mtp.append(MTPBlock(args.n_layers + layer_id, args))
            self.mtp[-1].embed = self.embed
            self.mtp[-1].head = self.head
        self.hc_mult = hc_mult = args.hc_mult
        hc_dim = hc_mult * args.dim
        with set_dtype(torch.float32):
            self.hc_head_fn = nn.Parameter(torch.empty(hc_mult, hc_dim))
            self.hc_head_base = nn.Parameter(torch.empty(hc_mult))
            self.hc_head_scale = nn.Parameter(torch.empty(1))

    def init_woq_layers(self):
        """After load_model(), convert all W4A16 parameters into QuantLinear layers."""
        # from gptqmodel.nn_modules.qlinear.tritonv2 import TritonV2QuantLinear as QuantLinear
        from gptqmodel.nn_modules.qlinear.marlin import MarlinQuantLinear as QuantLinear
        for module in self.modules():
            if hasattr(module, 'init_woq') and module is not self:
                module.init_woq(QuantLinear)
        torch.cuda.empty_cache()

    @torch.inference_mode()
    def forward(self, input_ids: torch.Tensor, start_pos: int = 0):
        h = self.embed(input_ids)
        # Expand to hc_mult copies for Hyper-Connections
        h = h.unsqueeze(2).repeat(1, 1, self.hc_mult, 1)
        for layer in self.layers:
            h = layer(h, start_pos, input_ids)
        logits = self.head(h, self.hc_head_fn, self.hc_head_scale, self.hc_head_base, self.norm)
        return logits


if __name__ == "__main__":
    torch.set_default_dtype(torch.bfloat16)
    torch.set_default_device("cuda")
    torch.manual_seed(0)
    args = ModelArgs(n_hash_layers=0)
    x = torch.randint(0, args.vocab_size, (2, 128))
    model = Transformer(args)

    print(model(x).size())
    for i in range(128, 150):
        print(i, model(x[:, 0:1], i).size())

    h = torch.randn(2, 128, args.hc_mult, args.dim)
    mtp = model.mtp[0]
    print(mtp(h, 0, x).size())
    print(mtp(h[:, 0:1], 1, x[:, 0:1]).size())