vllm.lora.ops.triton_ops.fused_moe_lora_op ¶

@torch.inference_mode()
def _fused_moe_lora(
    output: torch.Tensor,  # (num_tokens, top_k_num, N*len(lora_a_stacked),)
    qcurr_hidden_states: torch.Tensor,  # (num_tokens, K,)
    lora_a_stacked: list[
        torch.Tensor
    ],  # [(max_loras, num_experts, max_lora_rank, K,),...]
    lora_b_stacked: list[
        torch.Tensor
    ],  # [(max_loras, num_experts, N, max_lora_rank,),...]
    topk_weights: torch.Tensor,  # (num_tokens, top_k_num)
    sorted_token_ids: torch.Tensor,  # (max_loras, _)
    expert_ids: torch.Tensor,  # (max_loras, _ ,)
    num_tokens_post_padded: torch.Tensor,  # (max_loras, )
    max_lora_rank: int,
    top_k_num: int,
    block_size_m: int,
    block_size_n: int,
    block_size_k: int,
    group_size_m: int,
    split_k: int,
    mul_routed_weight: bool = False,
) -> None:
    assert len(lora_a_stacked) == len(lora_b_stacked) > 0
    assert (
        sorted_token_ids.dim()
        == expert_ids.dim()
        == topk_weights.dim()
        == qcurr_hidden_states.dim()
        == 2
    )
    assert (
        sorted_token_ids.shape[0]
        == expert_ids.shape[0]
        == num_tokens_post_padded.shape[0]
    )
    assert len(lora_b_stacked) * lora_b_stacked[0].shape[-2] == output.shape[-1]
    assert output.shape[0] == topk_weights.shape[0]
    assert top_k_num == topk_weights.shape[1]

    for lora_a, lora_b in zip(lora_a_stacked, lora_b_stacked):
        assert lora_a.dtype == lora_b.dtype == output.dtype == qcurr_hidden_states.dtype
        assert lora_a.dtype in [torch.float16, torch.bfloat16]

    device = qcurr_hidden_states.device
    num_slices = len(lora_a_stacked)

    config = {
        "BLOCK_SIZE_M": block_size_m,
        "BLOCK_SIZE_N": block_size_n,
        "BLOCK_SIZE_K": block_size_k,
        "GROUP_SIZE_M": group_size_m,
        "SPLIT_K": split_k,
    }

    w1_lora_a_stacked = lora_a_stacked[0]
    w1_lora_b_stacked = lora_b_stacked[0]
    num_experts = lora_a_stacked[0].shape[1]

    N = max_lora_rank
    M = topk_weights.shape[0]
    EM = sorted_token_ids.shape[1]
    K = qcurr_hidden_states.shape[1]
    num_tokens = M * top_k_num
    w1_output_dim_size = w1_lora_b_stacked.shape[2]

    lora_intermediate_cache1 = torch.empty(
        (num_slices * M * top_k_num * (max_lora_rank + w1_output_dim_size)),
        dtype=output.dtype,
        device=device,
    )

    # slices
    a_intermediate_size = num_slices * M * top_k_num * max_lora_rank
    a_intermediate_cache1 = lora_intermediate_cache1[:a_intermediate_size].view(
        num_slices, M, top_k_num, max_lora_rank
    )
    b_intermediate_cache1 = lora_intermediate_cache1[a_intermediate_size:].view(
        num_slices, M, top_k_num, w1_output_dim_size
    )

    b_ptr = _get_ptr(lora_a_stacked, device)

    grid = lambda META: (
        split_k
        * triton.cdiv(EM, META["BLOCK_SIZE_M"])
        * triton.cdiv(N, META["BLOCK_SIZE_N"]),
        len(lora_a_stacked),
        lora_a_stacked[0].shape[0],
    )

    _fused_moe_lora_kernel[grid](
        qcurr_hidden_states,
        b_ptr,
        a_intermediate_cache1,
        topk_weights,
        sorted_token_ids,
        expert_ids,
        num_tokens_post_padded,
        N,
        K,
        EM,
        num_tokens,
        num_experts,
        qcurr_hidden_states.stride(0),
        qcurr_hidden_states.stride(1),
        w1_lora_a_stacked.stride(0),
        w1_lora_a_stacked.stride(1),
        w1_lora_a_stacked.stride(3),
        w1_lora_a_stacked.stride(2),
        a_intermediate_cache1.stride(2),
        a_intermediate_cache1.stride(3),
        sorted_token_ids.stride(0),
        expert_ids.stride(0),
        slice_a_size=qcurr_hidden_states.numel(),
        slice_c_size=a_intermediate_cache1.numel() // num_slices,
        num_slice_a=1,
        num_slice_c=num_slices,
        top_k=1 if mul_routed_weight else top_k_num,
        MUL_ROUTED_WEIGHT=False,
        **config,
    )

    b_ptr = _get_ptr(lora_b_stacked, device)
    K = max_lora_rank
    N = w1_output_dim_size

    a_intermediate_cache1 = a_intermediate_cache1.view(
        -1, a_intermediate_cache1.shape[3]
    )

    # Set split_k = 1 for expand calls
    config["SPLIT_K"] = 1
    grid = lambda META: (
        triton.cdiv(EM, META["BLOCK_SIZE_M"]) * triton.cdiv(N, META["BLOCK_SIZE_N"]),
        len(lora_b_stacked),
        lora_b_stacked[0].shape[0],
    )
    _fused_moe_lora_kernel[grid](
        a_intermediate_cache1,
        b_ptr,
        b_intermediate_cache1,
        topk_weights,
        sorted_token_ids,
        expert_ids,
        num_tokens_post_padded,
        N,
        K,
        EM,
        num_tokens,
        num_experts,
        a_intermediate_cache1.stride(0),
        a_intermediate_cache1.stride(1),
        w1_lora_b_stacked.stride(0),
        w1_lora_b_stacked.stride(1),
        w1_lora_b_stacked.stride(3),
        w1_lora_b_stacked.stride(2),
        b_intermediate_cache1.stride(2),
        b_intermediate_cache1.stride(3),
        sorted_token_ids.stride(0),
        expert_ids.stride(0),
        slice_a_size=a_intermediate_cache1.numel() // num_slices,
        slice_c_size=b_intermediate_cache1.numel() // num_slices,
        num_slice_a=num_slices,
        num_slice_c=num_slices,
        top_k=1,
        MUL_ROUTED_WEIGHT=mul_routed_weight,
        **config,
    )
    for i in range(num_slices):
        output[:, :, i * N : (i + 1) * N] += b_intermediate_cache1[i]

def _fused_moe_lora_fake(
    output: torch.Tensor,
    qcurr_hidden_states: torch.Tensor,
    lora_a_stacked: list[torch.Tensor],
    lora_b_stacked: list[torch.Tensor],
    topk_weights: torch.Tensor,
    sorted_token_ids: torch.Tensor,
    expert_ids: torch.Tensor,
    num_tokens_post_padded: torch.Tensor,
    max_lora_rank: int,
    top_k_num: int,
    block_size_m: int,
    block_size_n: int,
    block_size_k: int,
    group_size_m: int,
    mul_routed_weight: bool = False,
) -> None:
    return

@triton.jit(
    do_not_specialize=[
        "num_valid_tokens",
        "EM",
        "stride_tl",
        "stride_el",
        "slice_a_size",
        "slice_c_size",
    ]
)
def _fused_moe_lora_kernel(
    a_ptr,
    b_ptr,
    c_ptr,
    topk_weights_ptr,
    sorted_token_ids_ptr,
    expert_ids_ptr,
    num_tokens_post_padded_ptr,
    # Matrix dimensions
    N,
    K,
    EM,
    num_valid_tokens,
    num_experts,
    # The stride variables represent how much to increase the ptr by when
    # moving by 1 element in a particular dimension. E.g. `stride_am` is
    # how much to increase `a_ptr` by to get the element one row down
    # (A has M rows).
    stride_am,
    stride_ak,
    stride_bl,
    stride_be,
    stride_bk,
    stride_bn,
    stride_cm,
    stride_cn,
    stride_tl,
    stride_el,
    slice_a_size,
    slice_c_size,
    # Meta-parameters
    num_slice_a: tl.constexpr,
    num_slice_c: tl.constexpr,
    top_k: tl.constexpr,
    MUL_ROUTED_WEIGHT: tl.constexpr,
    BLOCK_SIZE_M: tl.constexpr,
    BLOCK_SIZE_N: tl.constexpr,
    BLOCK_SIZE_K: tl.constexpr,
    GROUP_SIZE_M: tl.constexpr,
    SPLIT_K: tl.constexpr,
):
    pid = tl.program_id(axis=0)
    slice_id = tl.program_id(axis=1)
    lora_idx = tl.program_id(axis=2)
    max_loras = tl.num_programs(axis=2)
    grid_k = tl.cdiv(K, BLOCK_SIZE_K * SPLIT_K)

    # calculate pid_m,pid_n
    num_pid_m = tl.cdiv(EM, BLOCK_SIZE_M)
    num_pid_n = tl.cdiv(N, BLOCK_SIZE_N)
    num_pid_in_group = GROUP_SIZE_M * num_pid_n
    group_id = pid // num_pid_in_group
    first_pid_m = group_id * GROUP_SIZE_M
    group_size_m = min(num_pid_m - first_pid_m, GROUP_SIZE_M)
    pid_m = first_pid_m + ((pid % num_pid_in_group) % group_size_m)
    pid_n = (pid % num_pid_in_group) // group_size_m

    num_tokens_post_padded = tl.load(num_tokens_post_padded_ptr + lora_idx)
    if pid_m * BLOCK_SIZE_M >= num_tokens_post_padded:
        return

    # get the expert_id to process curr shard
    ind = lora_idx * stride_el + pid_m
    expert_id = tl.load(expert_ids_ptr + ind, ind < max_loras * stride_el, -1)
    if expert_id == -1:
        return

    # get a_ptr,b_ptr,c_ptr
    cur_a_ptr = a_ptr + (slice_id % num_slice_a) * slice_a_size
    cur_b_ptr = tl.load(b_ptr + slice_id).to(tl.pointer_type(c_ptr.dtype.element_ty))
    cur_c_ptr = c_ptr + (slice_id % num_slice_c) * slice_c_size

    offs_bn = (pid_n * BLOCK_SIZE_N + tl.arange(0, BLOCK_SIZE_N).to(tl.int64)) % N
    offs_k = tl.arange(0, BLOCK_SIZE_K)

    offs_token_id = pid_m * BLOCK_SIZE_M + tl.arange(0, BLOCK_SIZE_M).to(tl.int64)
    token_ind = stride_tl * lora_idx + offs_token_id
    offs_token = tl.load(
        sorted_token_ids_ptr + token_ind, token_ind < max_loras * stride_tl, 0
    )
    token_mask = offs_token < num_valid_tokens

    # get a_ptrs,b_ptrs
    a_ptrs = cur_a_ptr + (
        offs_token[:, None] // top_k * stride_am + offs_k[None, :] * stride_ak
    )

    b_ptrs = (
        cur_b_ptr
        + lora_idx * stride_bl
        + expert_id * stride_be
        + (offs_k[:, None] * stride_bk + offs_bn[None, :] * stride_bn)
    )

    # accumulator
    accumulator = tl.zeros((BLOCK_SIZE_M, BLOCK_SIZE_N), dtype=tl.float32)
    for k in range(0, grid_k):
        k_remaining = K - k * (BLOCK_SIZE_K * SPLIT_K)
        a = tl.load(
            a_ptrs,
            mask=token_mask[:, None] & (offs_k[None, :] < k_remaining),
            other=0.0,
        )
        b = tl.load(b_ptrs, mask=offs_k[:, None] < k_remaining, other=0.0)
        accumulator += tl.dot(a, b)
        # Advance the ptrs to the next K block.
        a_ptrs += BLOCK_SIZE_K * SPLIT_K * stride_ak
        b_ptrs += BLOCK_SIZE_K * SPLIT_K * stride_bk

    if MUL_ROUTED_WEIGHT:
        moe_weight = tl.load(topk_weights_ptr + offs_token, mask=token_mask, other=0)
        accumulator = accumulator * moe_weight[:, None]

    accumulator = accumulator.to(c_ptr.dtype.element_ty)
    # Write back the block of the output
    offs_cn = pid_n * BLOCK_SIZE_N + tl.arange(0, BLOCK_SIZE_N)
    c_ptrs = cur_c_ptr + stride_cm * offs_token[:, None] + stride_cn * offs_cn[None, :]
    c_mask = token_mask[:, None] & (offs_cn[None, :] < N)
    if SPLIT_K == 1:
        tl.store(c_ptrs, accumulator, mask=c_mask)
    else:
        tl.atomic_add(c_ptrs, accumulator, mask=c_mask, sem="relaxed")

def _get_ptr(lora_weights: list[torch.Tensor], device: torch.device):
    """
    `_LORA_PTR_DICT` collects the required information during `profile_run`,
    After this, it remains constant and subsequent usage is through LUT.
    Refer to:
    https://github.com/triton-lang/triton/blob/release/3.1.x/python/tutorials/08-grouped-gemm.py
    """
    key = tuple(lora_weight.data_ptr() for lora_weight in lora_weights)

    if (ptr_tensor := _LORA_PTR_DICT.get(key)) is not None:
        return ptr_tensor

    tensor_ptrs = []
    for lora_weight in lora_weights:
        tensor_ptrs.append(lora_weight.data_ptr())
    ptr_tensor = torch.tensor(tensor_ptrs, device=device)

    _LORA_PTR_DICT[key] = ptr_tensor
    return _LORA_PTR_DICT.get(key)