modules/SwissArmyTransformer/sat/model/position_embedding/triton_rotary.py

from typing import Optional, Union

import torch

import triton
import triton.language as tl


# @triton.autotune(
#     configs=[
#         triton.Config({"BLOCK_M": 2}),
#         triton.Config({"BLOCK_M": 4}),
#         triton.Config({"BLOCK_M": 8}),
#         triton.Config({"BLOCK_M": 16}),
#     ],
#     key=["CACHE_KEY_SEQLEN", "BLOCK_K", "INTERLEAVED"],
# )
@triton.jit
def rotary_kernel(
    OUT,  # Pointers to matrices
    X,
    COS,
    SIN,
    POSITIONS,
    CU_SEQLENS,
    SEQLEN_OFFSETS,  # this could be int or a pointer
    # Matrix dimensions
    seqlen,
    nheads,
    rotary_dim,
    seqlen_ro,
    CACHE_KEY_SEQLEN,
    # strides
    stride_out_batch,
    stride_out_nheads,
    stride_out_seqlen,
    stride_out_headdim,
    stride_x_batch,
    stride_x_nheads,
    stride_x_seqlen,
    stride_x_headdim,
    stride_p_batch,
    # Meta-parameters
    BLOCK_K: tl.constexpr,
    IS_SEQLEN_OFFSETS_TENSOR: tl.constexpr,
    IS_VARLEN: tl.constexpr,
    INTERLEAVED: tl.constexpr,
    CONJUGATE: tl.constexpr,
    BLOCK_M: tl.constexpr,
):
    pid_m = tl.program_id(axis=0)
    pid_batch = tl.program_id(axis=1)
    pid_head = tl.program_id(axis=2)
    rotary_dim_half = rotary_dim // 2

    
    X = X + pid_batch * stride_x_batch + pid_head * stride_x_nheads
    OUT = OUT + pid_batch * stride_out_batch + pid_head * stride_out_nheads
    POS = POSITIONS + pid_batch * stride_p_batch
    
    if pid_m * BLOCK_M >= seqlen:
        return
    rm = pid_m * BLOCK_M + tl.arange(0, BLOCK_M)
    rm_cs = tl.load(POS + rm, mask=rm < seqlen, other=0)
    
    #tl.device_print("rm rm_cs, ", rm, rm_cs)
    
    if not IS_SEQLEN_OFFSETS_TENSOR:
        rm_cs = rm_cs + SEQLEN_OFFSETS
    else:
        rm_cs = rm_cs + tl.load(SEQLEN_OFFSETS + pid_batch)
    rk = tl.arange(0, BLOCK_K)
    rk_half = tl.arange(0, BLOCK_K // 2)

    if not INTERLEAVED:
        # Load the 1st and 2nd halves of X, do calculation, then store to 1st and 2nd halves of OUT
        X = X + (rm[:, None] * stride_x_seqlen + rk_half[None, :] * stride_x_headdim)
        COS = COS + (rm_cs[:, None] * rotary_dim_half + rk_half[None, :])
        SIN = SIN + (rm_cs[:, None] * rotary_dim_half + rk_half[None, :])
        cos = tl.load(
            COS, mask=(rm_cs[:, None] < seqlen_ro) & (rk_half[None, :] < rotary_dim_half), other=1.0
        )
        sin = tl.load(
            SIN, mask=(rm_cs[:, None] < seqlen_ro) & (rk_half[None, :] < rotary_dim_half), other=0.0
        )
        x0 = tl.load(
            X, mask=(rm[:, None] < seqlen) & (rk_half[None, :] < rotary_dim_half), other=0.0
        )
        x1 = tl.load(
            X + rotary_dim_half * stride_x_headdim,
            mask=(rm[:, None] < seqlen) & (rk_half[None, :] < rotary_dim_half),
            other=0.0,
        )
        if CONJUGATE:
            sin = -sin
        o0 = x0 * cos - x1 * sin
        o1 = x0 * sin + x1 * cos
        # write back result
        OUT = OUT + (rm[:, None] * stride_out_seqlen + rk_half[None, :] * stride_out_headdim)
        tl.store(OUT, o0, mask=(rm[:, None] < seqlen) & (rk_half[None, :] < rotary_dim_half))
        tl.store(
            OUT + rotary_dim_half * stride_out_headdim,
            o1,
            mask=(rm[:, None] < seqlen) & (rk_half[None, :] < rotary_dim_half),
        )
    else:
        # We don't want to load X[0, 2, 4, ...] and X[1, 3, 5, ...] separately since both are slow.
        # Instead, we load x0 = X[0, 1, 2, 3, ...] and x1 = X[1, 0, 3, 2, ...].
        # Loading x0 will be fast but x1 will be slow.
        # Then we load cos = COS[0, 0, 1, 1, ...] and sin = SIN[0, 0, 1, 1, ...].
        # Then we do the calculation and use tl.where to pick put the right outputs for the even
        # and for the odd indices.
        rk_swap = rk + ((rk + 1) % 2) * 2 - 1  # 1, 0, 3, 2, 5, 4, ...
        rk_repeat = tl.arange(0, BLOCK_K) // 2
        X0 = X + (rm[:, None] * stride_x_seqlen + rk[None, :] * stride_x_headdim)
        X1 = X + (rm[:, None] * stride_x_seqlen + rk_swap[None, :] * stride_x_headdim)
        COS = COS + (rm_cs[:, None] * rotary_dim_half + rk_repeat[None, :])
        SIN = SIN + (rm_cs[:, None] * rotary_dim_half + rk_repeat[None, :])
        cos = tl.load(
            COS,
            mask=(rm_cs[:, None] < seqlen_ro) & (rk_repeat[None, :] < rotary_dim_half),
            other=1.0,
        )
        sin = tl.load(
            SIN,
            mask=(rm_cs[:, None] < seqlen_ro) & (rk_repeat[None, :] < rotary_dim_half),
            other=0.0,
        )
        x0 = tl.load(X0, mask=(rm[:, None] < seqlen) & (rk[None, :] < rotary_dim), other=0.0)
        x1 = tl.load(X1, mask=(rm[:, None] < seqlen) & (rk_swap[None, :] < rotary_dim), other=0.0)
        if CONJUGATE:
            sin = -sin
        x0_cos = x0 * cos
        x1_sin = x1 * sin
        out = tl.where(rk[None, :] % 2 == 0, x0_cos - x1_sin, x0_cos + x1_sin)
        OUT = OUT + (rm[:, None] * stride_out_seqlen + rk[None, :] * stride_out_headdim)
        tl.store(OUT, out, mask=(rm[:, None] < seqlen) & (rk[None, :] < rotary_dim))


def apply_rotary(
    x: torch.Tensor,
    cos: torch.Tensor,
    sin: torch.Tensor,
    position_ids: torch.Tensor,
    seqlen_offsets: Union[int, torch.Tensor] = 0,
    cu_seqlens: Optional[torch.Tensor] = None,
    max_seqlen: Optional[int] = None,
    interleaved=False,
    inplace=False,
    conjugate=False,
) -> torch.Tensor:
    """
    Arguments:
        x: (batch, nheads, seqlen, headdim) if cu_seqlens is None
            else (total_seqlen, nheads, headdim).
        cos: (seqlen_ro, rotary_dim / 2)
        sin: (seqlen_ro, rotary_dim / 2)
        seqlen_offsets: integer or integer tensor of size (batch,)
        cu_seqlens: (batch + 1,) or None
        max_seqlen: int
    Returns:
        y: (batch, nheads, seqlen, headdim)
    """
    
    
    batch, nheads, seqlen, headdim = x.shape
    
    seqlen_ro, rotary_dim = cos.shape

    batch_p, seqlen_p = position_ids.shape
    
    assert batch_p == batch and seqlen_p == seqlen
    assert sin.shape == cos.shape
    rotary_dim *= 2
    assert rotary_dim <= headdim, "rotary_dim must be <= headdim"
    assert headdim <= 256, "Only support headdim <= 256"

    assert seqlen_ro >= max_seqlen, "seqlen_ro must be >= max_seqlen"

    assert (
        cos.dtype == sin.dtype
    ), f"cos and sin must have the same dtype, got {cos.dtype} and {sin.dtype}"
    assert (
        x.dtype == cos.dtype
    ), f"Input and cos/sin must have the same dtype, got {x.dtype} and {cos.dtype}"

    cos, sin = cos.contiguous(), sin.contiguous()
    
    if isinstance(seqlen_offsets, torch.Tensor):
        assert seqlen_offsets.shape == (batch,)
        assert seqlen_offsets.dtype in [torch.int32, torch.int64]
        seqlen_offsets = seqlen_offsets.contiguous()
    else:
        assert seqlen_offsets + max_seqlen <= seqlen_ro

    output = torch.empty_like(x) if not inplace else x
    if rotary_dim < headdim and not inplace:
        output[..., rotary_dim:].copy_(x[..., rotary_dim:])

    BLOCK_K = (
        32
        if rotary_dim <= 32
        else (64 if rotary_dim <= 64 else (128 if rotary_dim <= 128 else 256))
    )
    grid = lambda META: (triton.cdiv(seqlen, META["BLOCK_M"]), batch, nheads)  # noqa
    BLOCK_M = 4 if interleaved else (8 if rotary_dim <= 64 else 4)

    # Need this, otherwise Triton tries to launch from cuda:0 and we get
    # ValueError: Pointer argument (at 0) cannot be accessed from Triton (cpu tensor?)
    with torch.cuda.device(x.device.index):
        rotary_kernel[grid](
            output,  # data ptrs
            x,
            cos,
            sin,
            position_ids,
            cu_seqlens,
            seqlen_offsets,
            seqlen,  # shapes
            nheads,
            rotary_dim,
            seqlen_ro,
            seqlen // 128,  # key for triton cache (limit number of compilations)
            output.stride(0),  # batch_strides
            output.stride(-3),  # nheads_stride
            output.stride(-2),  # seqlen_stride
            output.stride(-1),  # headdim_stride
            x.stride(0),  # batch_strides
            x.stride(-3),  # nheads stride
            x.stride(-2),  # seqlen stride
            x.stride(-1),  # headdim stride
            position_ids.stride(0), # batch_strides
            BLOCK_K,
            isinstance(seqlen_offsets, torch.Tensor),
            False,
            interleaved,
            conjugate,
            BLOCK_M,
        )
    return output