fla/ops/rwkv7/fused_recurrent.py

# Copyright (c) 2023-2025, Songlin Yang, Yu Zhang

import warnings

import torch
import triton
import triton.language as tl

from fla.ops.generalized_delta_rule import fused_recurrent_dplr_delta_rule
from fla.ops.utils.op import exp
from fla.utils import USE_CUDA_GRAPH, autotune_cache_kwargs, input_guard


@triton.heuristics({
    'USE_INITIAL_STATE': lambda args: args['h0'] is not None,
    'STORE_FINAL_STATE': lambda args: args['ht'] is not None,
    'IS_VARLEN': lambda args: args['cu_seqlens'] is not None,
})
@triton.autotune(
    configs=[
        triton.Config({'BV': BV}, num_warps=num_warps, num_stages=num_stages)
        for BV in [16, 32, 64]
        for num_warps in [2, 4, 8, 16]
        for num_stages in [2, 3, 4]
    ],
    key=['BK'],
    use_cuda_graph=USE_CUDA_GRAPH,
    **autotune_cache_kwargs,
)
@triton.jit(do_not_specialize=['T'])
def fused_recurrent_rwkv7_fwd_kernel(
    r,
    w,
    k,
    v,
    kk,
    a,
    o,
    h0,
    ht,
    cu_seqlens,
    scale,
    T,
    B: tl.constexpr,
    H: tl.constexpr,
    K: tl.constexpr,
    V: tl.constexpr,
    BK: tl.constexpr,
    BV: tl.constexpr,
    REVERSE: tl.constexpr,
    USE_INITIAL_STATE: tl.constexpr,
    STORE_FINAL_STATE: tl.constexpr,
    IS_VARLEN: tl.constexpr,
    IS_DECODE: tl.constexpr,
):
    i_v, i_nh = tl.program_id(0).to(tl.int64), tl.program_id(1).to(tl.int64)
    i_n, i_h = i_nh // H, i_nh % H

    if IS_VARLEN:
        bos, eos = tl.load(cu_seqlens + i_n).to(tl.int64), tl.load(cu_seqlens + i_n + 1).to(tl.int64)
        T = eos - bos
    else:
        bos, eos = i_n * T, i_n * T + T

    o_k = tl.arange(0, BK)
    o_v = i_v * BV + tl.arange(0, BV)
    p_r = r + (bos + ((T - 1) if REVERSE else 0)) * H*K + i_h * K + o_k
    p_w = w + (bos + ((T - 1) if REVERSE else 0)) * H*K + i_h * K + o_k
    p_k = k + (bos + ((T - 1) if REVERSE else 0)) * H*K + i_h * K + o_k
    p_v = v + (bos + ((T - 1) if REVERSE else 0)) * H*V + i_h * V + o_v
    p_a = a + (bos + ((T - 1) if REVERSE else 0)) * H*K + i_h * K + o_k
    p_kk = kk + (bos + ((T - 1) if REVERSE else 0)) * H*K + i_h * K + o_k

    p_o = o + (bos + ((T - 1) if REVERSE else 0)) * H*V + i_h * V + o_v

    mask_k = o_k < K
    mask_v = o_v < V
    mask_h = mask_k[:, None] & mask_v[None, :]
    b_h = tl.zeros([BK, BV], dtype=tl.float32)

    if USE_INITIAL_STATE:
        p_h0 = h0 + i_nh * K*V + o_k[:, None] * V + o_v
        b_h += tl.load(p_h0, mask=mask_h, other=0).to(tl.float32)

    if IS_DECODE:
        b_r = tl.load(p_r, mask=mask_k, other=0).to(tl.float32) * scale
        b_w = tl.load(p_w, mask=mask_k, other=0).to(tl.float32)
        b_k = tl.load(p_k, mask=mask_k, other=0).to(tl.float32)
        b_v = tl.load(p_v, mask=mask_v, other=0).to(tl.float32)
        b_a = tl.load(p_a, mask=mask_k, other=0).to(tl.float32)
        b_kk = tl.load(p_kk, mask=mask_k, other=0).to(tl.float32)
        b_act_a = -b_kk
        b_b = b_kk * b_a

        b_h = exp(b_w)[:, None] * b_h + b_b[:, None] * tl.sum(b_act_a[:, None] * b_h, 0)[None, :]
        b_h += b_k[:, None] * b_v[None, :]
        b_o = tl.sum(b_h * b_r[:, None], 0)

        tl.store(p_o, b_o.to(p_o.dtype.element_ty), mask=mask_v)
    else:
        for _ in range(0, T):
            b_r = tl.load(p_r, mask=mask_k, other=0).to(tl.float32) * scale
            b_w = tl.load(p_w, mask=mask_k, other=0).to(tl.float32)
            b_k = tl.load(p_k, mask=mask_k, other=0).to(tl.float32)
            b_v = tl.load(p_v, mask=mask_v, other=0).to(tl.float32)
            b_a = tl.load(p_a, mask=mask_k, other=0).to(tl.float32)
            b_kk = tl.load(p_kk, mask=mask_k, other=0).to(tl.float32)
            b_act_a = -b_kk
            b_b = b_kk * b_a

            b_h = exp(b_w)[:, None] * b_h + b_b[:, None] * tl.sum(b_act_a[:, None] * b_h, 0)[None, :]
            b_h += b_k[:, None] * b_v[None, :]
            b_o = tl.sum(b_h * b_r[:, None], 0)

            tl.store(p_o, b_o.to(p_o.dtype.element_ty), mask=mask_v)
            p_r += (-1 if REVERSE else 1) * H*K
            p_w += (-1 if REVERSE else 1) * H*K
            p_k += (-1 if REVERSE else 1) * H*K
            p_v += (-1 if REVERSE else 1) * H*V
            p_a += (-1 if REVERSE else 1) * H*K
            p_kk += (-1 if REVERSE else 1) * H*K
            p_o += (-1 if REVERSE else 1) * H*V

    if STORE_FINAL_STATE:
        p_ht = ht + i_nh * K*V + o_k[:, None] * V + o_v
        tl.store(p_ht, b_h.to(p_ht.dtype.element_ty), mask=mask_h)


@input_guard
def fused_recurrent_rwkv7_fwd(
    r: torch.Tensor,
    w: torch.Tensor,
    k: torch.Tensor,
    v: torch.Tensor,
    kk: torch.Tensor,
    a: torch.Tensor,
    scale: float | None = 1.0,
    initial_state: torch.Tensor | None = None,
    output_final_state: bool = False,
    reverse: bool = False,
    cu_seqlens: torch.LongTensor | None = None,
):
    B, T, H, K, V = *k.shape, v.shape[-1]
    N = B if cu_seqlens is None else len(cu_seqlens) - 1
    BK = triton.next_power_of_2(K)
    IS_DECODE = (T == 1)

    h0 = initial_state
    if not output_final_state:
        ht = None
    else:
        ht = r.new_empty(N, H, K, V, dtype=torch.float32)
    o = torch.empty_like(v)

    def grid(meta): return (triton.cdiv(V, meta['BV']), N * H)
    fused_recurrent_rwkv7_fwd_kernel[grid](
        r,
        w,
        k,
        v,
        kk,
        a,
        o,
        h0,
        ht,
        cu_seqlens,
        scale,
        T=T,
        B=B,
        H=H,
        K=K,
        V=V,
        BK=BK,
        REVERSE=reverse,
        IS_DECODE=IS_DECODE,
    )
    return o, ht


def fused_recurrent_rwkv7(
    r: torch.Tensor,
    w: torch.Tensor,
    k: torch.Tensor,
    v: torch.Tensor,
    a: torch.Tensor,
    b: torch.Tensor,
    scale: float | None = None,
    initial_state: torch.Tensor = None,
    output_final_state: bool = True,
    cu_seqlens: torch.LongTensor | None = None,
    head_first: bool = False,
):
    """
    Args:
        r (torch.Tensor):
            r of shape `[B, T, H, K]`.
        w (torch.Tensor):
            log decay of shape `[B, T, H, K]`.
        k (torch.Tensor):
            k of shape `[B, T, H, K]`.
        v (torch.Tensor):
            v of shape `[B, T, H, V]`.
        a (torch.Tensor):
            a of shape `[B, T, H, K]`.
        b (torch.Tensor):
            b of shape `[B, T, H, K]`.
        scale (float):
            scale of the attention.
            If not provided, it will default to `1 / sqrt(K)`. Default: `None`.
        initial_state (torch.Tensor):
            initial state of shape `[B, H, K, V]` if cu_seqlens is None else `[N, H, K, V]` where N = len(cu_seqlens) - 1.
        output_final_state (bool):
            whether to output the final state.
        cu_seqlens (torch.LongTensor):
            Cumulative sequence lengths of shape `[N+1]` used for variable-length training,
            consistent with the FlashAttention API.
        head_first (Optional[bool]):
            Whether the inputs are in the head-first format. Default: `False`.
            This argument has been deprecated.
    """
    if head_first:
        raise DeprecationWarning(
            "head_first is deprecated and will be removed in a future version. "
            "Please use head_first=False for now instead.",
        )
    elif r.shape[1] < r.shape[2]:
        warnings.warn(
            f"Input tensor shape suggests potential format mismatch: seq_len ({r.shape[1]}) < num_heads ({r.shape[2]}). "
            "This may indicate the inputs were passed in head-first format [B, H, T, ...] "
            "when head_first=False was specified. "
            "Please verify your input tensor format matches the expected shape [B, T, H, ...].",
        )
    return fused_recurrent_dplr_delta_rule(
        q=r,
        k=k,
        v=v,
        a=a,
        b=b,
        gk=w,
        scale=scale,
        initial_state=initial_state,
        output_final_state=output_final_state,
        cu_seqlens=cu_seqlens,
    )


def fused_mul_recurrent_rwkv7(
    r: torch.Tensor,
    w: torch.Tensor,
    k: torch.Tensor,
    v: torch.Tensor,
    kk: torch.Tensor,
    a: torch.Tensor,
    scale: float | None = 1.0,
    initial_state: torch.Tensor | None = None,
    output_final_state: bool = False,
    reverse: bool = False,
    cu_seqlens: torch.Tensor | None = None,
    head_first: bool = False,
) -> tuple[torch.Tensor, torch.Tensor]:
    r"""
    This function computes the recurrence S_t = S_t @ (I + a_t b_t^T) + v_t k_t^T in a recurrent manner.

    Args:
        r (torch.Tensor):
            queries of shape `[B, T, H, K]`.
        w (torch.Tensor):
            keys of shape `[B, T, H, K]`.
        k (torch.Tensor):
            values of shape `[B, T, H, V]`.
        v (torch.Tensor):
            a of shape `[B, T, H, K]`.
        kk (torch.Tensor):
            b of shape `[B, T, H, K]`.
        a (torch.Tensor):
            gk of shape `[B, T, H, K]`. decay term in log space!
        scale (Optional[float]):
            Scale factor for the RetNet attention scores.
            If not provided, it will default to `1 / sqrt(K)`. Default: 1.
        initial_state (Optional[torch.Tensor]):
            Initial state of shape `[N, H, K, V]` for `N` input sequences.
            For equal-length input sequences, `N` equals the batch size `B`.
            Default: `None`.
        output_final_state (Optional[bool]):
            Whether to output the final state of shape `[N, H, K, V]`. Default: `False`.
        reverse (Optional[bool]):
            If `True`, process the state passing in reverse order. Default: `False`.
        cu_seqlens (Optional[torch.Tensor]):
            Cumulative sequence lengths of shape `[N + 1]` used for variable-length training,
            consistent with the FlashAttention API.
        head_first (Optional[bool]):
            Whether the inputs are in the head-first format. Default: `False`.
            This argument has been deprecated.
    """
    if head_first:
        raise DeprecationWarning(
            "head_first is deprecated and will be removed in a future version. "
            "Please use head_first=False for now instead.",
        )
    elif r.shape[1] < r.shape[2]:
        warnings.warn(
            f"Input tensor shape suggests potential format mismatch: seq_len ({r.shape[1]}) < num_heads ({r.shape[2]}). "
            "This may indicate the inputs were passed in head-first format [B, H, T, ...] "
            "when head_first=False was specified. "
            "Please verify your input tensor format matches the expected shape [B, T, H, ...].",
        )
    if cu_seqlens is not None:
        if r.shape[0] != 1:
            raise ValueError(
                f"The batch size is expected to be 1 rather than {r.shape[0]} when using `cu_seqlens`."
                f"Please flatten variable-length inputs before processing.",
            )
        if initial_state is not None and initial_state.shape[0] != len(cu_seqlens) - 1:
            raise ValueError(
                f"The number of initial states is expected to be equal to the number of input sequences, "
                f"i.e., {len(cu_seqlens) - 1} rather than {initial_state.shape[0]}.",
            )
    if scale is None:
        scale = r.shape[-1] ** -0.5
    o, final_state = fused_recurrent_rwkv7_fwd(
        r,
        w,
        k,
        v,
        kk,
        a,
        scale,
        initial_state,
        output_final_state,
        reverse,
        cu_seqlens,
    )
    return o, final_state