vllm/tests/compile/test_fusion.py

# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project


import pytest
import torch

import vllm.config
import vllm.plugins
from vllm._aiter_ops import IS_AITER_FOUND, rocm_aiter_ops
from vllm.compilation.fusion import FUSED_OPS, FusedRMSQuantKey, RMSNormQuantFusionPass
from vllm.compilation.fx_utils import find_op_nodes
from vllm.compilation.matcher_utils import QUANT_OPS
from vllm.compilation.noop_elimination import NoOpEliminationPass
from vllm.compilation.post_cleanup import PostCleanupPass
from vllm.config import (
    CompilationConfig,
    CompilationMode,
    ModelConfig,
    PassConfig,
    VllmConfig,
)
from vllm.model_executor.layers.layernorm import RMSNorm
from vllm.model_executor.layers.quantization.kernels.scaled_mm.cutlass import (
    CutlassFP8ScaledMMLinearKernel,
)
from vllm.model_executor.layers.quantization.kernels.scaled_mm.flashinfer import (
    FlashInferScaledMMLinearKernel,
)
from vllm.model_executor.layers.quantization.kernels.scaled_mm.pytorch import (
    ChannelWiseTorchScaledMMLinearKernel,
    PerTensorTorchScaledMMLinearKernel,
    RowWiseTorchScaledMMLinearKernel,
)
from vllm.model_executor.layers.quantization.kernels.scaled_mm.rocm import (
    ROCmScaledMMLinearKernel,
)
from vllm.model_executor.layers.quantization.kernels.scaled_mm.ScaledMMLinearKernel import (  # noqa: E501
    FP8ScaledMMLinearKernel,
)
from vllm.model_executor.layers.quantization.utils.quant_utils import (
    GroupShape,
    QuantKey,
    ScaleDesc,
)
from vllm.model_executor.layers.quantization.utils.w8a8_utils import (
    cutlass_block_fp8_supported,
)
from vllm.platforms import current_platform
from vllm.utils.deep_gemm import (
    is_deep_gemm_supported,
)

from ..utils import TestBlockFP8Layer, TestFP8Layer
from .backend import TestBackend

FP8_DTYPE = current_platform.fp8_dtype()

RMS_OP = torch.ops._C.rms_norm.default
RMS_ADD_OP = torch.ops._C.fused_add_rms_norm.default

# Kernel and group_shape combinations: (kernel, group_shape)
# CUDA kernels
CUDA_KERNEL_GROUPSHAPE_COMBINATIONS = [
    # FlashInferScaledMMLinearKernel supports both per-tensor and per-token
    (FlashInferScaledMMLinearKernel, GroupShape.PER_TOKEN),
    (FlashInferScaledMMLinearKernel, GroupShape.PER_TENSOR),
    # CutlassFP8ScaledMMLinearKernel supports both per-tensor and per-token
    (CutlassFP8ScaledMMLinearKernel, GroupShape.PER_TOKEN),
    (CutlassFP8ScaledMMLinearKernel, GroupShape.PER_TENSOR),
    # PerTensorTorchScaledMMLinearKernel only supports per-tensor
    (PerTensorTorchScaledMMLinearKernel, GroupShape.PER_TENSOR),
    # ChannelWiseTorchScaledMMLinearKernel only supports per-token
    (ChannelWiseTorchScaledMMLinearKernel, GroupShape.PER_TOKEN),
    # Blockwise group shapes (no kernel abstraction)
    (None, GroupShape(1, 128)),
    (None, GroupShape(1, 64)),
]

# ROCm kernels
ROCM_KERNEL_GROUPSHAPE_COMBINATIONS = [
    # ROCmScaledMMLinearKernel supports both per-tensor and per-token
    (ROCmScaledMMLinearKernel, GroupShape.PER_TOKEN),
    (ROCmScaledMMLinearKernel, GroupShape.PER_TENSOR),
    # RowWiseTorchScaledMMLinearKernel only supports per-token
    (RowWiseTorchScaledMMLinearKernel, GroupShape.PER_TOKEN),
    # ChannelWiseTorchScaledMMLinearKernel only supports per-token
    (ChannelWiseTorchScaledMMLinearKernel, GroupShape.PER_TOKEN),
    # Blockwise group shapes (no kernel abstraction)
    (None, GroupShape(1, 128)),
    (None, GroupShape(1, 64)),
]

KERNEL_GROUPSHAPE_COMBINATIONS = (
    CUDA_KERNEL_GROUPSHAPE_COMBINATIONS
    if current_platform.is_cuda()
    else ROCM_KERNEL_GROUPSHAPE_COMBINATIONS
)

# For Aiter tests we toggle use_aiter_quant_op
AITER_KERNEL_GROUPSHAPE_COMBINATIONS = [
    # Per-token with ROCmScaledMMLinearKernel
    (ROCmScaledMMLinearKernel, GroupShape.PER_TOKEN, True),
    (ROCmScaledMMLinearKernel, GroupShape.PER_TOKEN, False),
    # Per-token with RowWiseTorchScaledMMLinearKernel
    (RowWiseTorchScaledMMLinearKernel, GroupShape.PER_TOKEN, True),
    (RowWiseTorchScaledMMLinearKernel, GroupShape.PER_TOKEN, False),
    # Per-token with ChannelWiseTorchScaledMMLinearKernel
    (ChannelWiseTorchScaledMMLinearKernel, GroupShape.PER_TOKEN, True),
    (ChannelWiseTorchScaledMMLinearKernel, GroupShape.PER_TOKEN, False),
    # Blockwise (no kernel abstraction)
    (None, GroupShape(1, 128), True),
]


class TestModel(torch.nn.Module):
    def __init__(
        self,
        hidden_size: int,
        eps: float,
        force_kernel: FP8ScaledMMLinearKernel | None,
        group_shape: GroupShape,
        use_aiter_fusion: bool = False,
        use_aiter_quant: bool = False,
        *args,
        **kwargs,
    ):
        super().__init__(*args, **kwargs)
        self.fp8_linear_layers: list[torch.nn.Module]
        self.group_shape = group_shape
        self.use_aiter_quant_op = use_aiter_quant
        self.use_aiter_fusion = use_aiter_fusion
        self.norm = [RMSNorm(hidden_size, eps) for _ in range(4)]
        self.enable_rms_norm_custom_op = self.norm[0].enabled()

        # Determine if blockwise based on group_shape
        is_blockwise = group_shape.is_per_group()

        if is_blockwise:
            self._init_blockwise(
                hidden_size, group_shape, use_aiter_fusion, use_aiter_quant
            )
        else:
            self._init_nonblockwise(
                hidden_size, group_shape, force_kernel, use_aiter_quant
            )

    def _init_nonblockwise(
        self,
        hidden_size: int,
        group_shape: GroupShape,
        force_kernel: FP8ScaledMMLinearKernel | None,
        use_aiter_quant: bool,
    ):
        """Initialize non-blockwise (per-tensor/per-token) FP8 layers."""
        is_static = group_shape == GroupShape.PER_TENSOR
        act_quant_scale_desc = ScaleDesc(torch.float32, is_static, group_shape)
        w_quant_scale_desc = ScaleDesc(torch.float32, True, group_shape)
        self.activation_quant_key = QuantKey(
            dtype=FP8_DTYPE, scale=act_quant_scale_desc, symmetric=True
        )
        self.weight_quant_key = QuantKey(
            dtype=FP8_DTYPE, scale=w_quant_scale_desc, symmetric=True
        )

        # Setup weight scales
        wscale_shape = (1,) if group_shape.is_per_tensor() else (hidden_size, 1)
        self.wscale = [torch.rand(wscale_shape, dtype=torch.float32) for _ in range(3)]

        self.act_scale = (
            [torch.rand(1, dtype=torch.float32) for _ in range(3)]
            if is_static
            else [None for _ in range(3)]
        )

        # Initialize weights (transposed for non-blockwise)
        self.w = [
            torch.rand(hidden_size, hidden_size).to(dtype=FP8_DTYPE).t()
            for _ in range(3)
        ]

        # Setup FP8 linear layers with kernel abstraction
        self.fp8_linear_layers = [
            TestFP8Layer(
                self.activation_quant_key,
                self.weight_quant_key,
                self.w[i],
                self.wscale[i],
                input_scale=self.act_scale[i],
                force_kernel=force_kernel,
            )
            for i in range(3)
        ]

        # Enable aiter quantization if requested
        for layer in self.fp8_linear_layers:
            layer.kernel.quant_fp8.use_aiter = use_aiter_quant

        self.enable_quant_fp8_custom_op = self.fp8_linear_layers[
            0
        ].is_quant_fp8_enabled()

    def _init_blockwise(
        self,
        hidden_size: int,
        group_shape: GroupShape,
        use_aiter_fusion: bool,
        use_aiter_quant: bool,
    ):
        """Initialize blockwise FP8 layers."""
        act_quant_scale_desc = ScaleDesc(torch.float32, False, group_shape)
        self.activation_quant_key = QuantKey(
            dtype=FP8_DTYPE, scale=act_quant_scale_desc, symmetric=True
        )

        # Setup weight scales (for blockwise quantization)
        # Use aiter block size if aiter fusion is enabled
        scale_size = (
            (hidden_size + 128 - 1) // 128
            if use_aiter_fusion
            else hidden_size // group_shape[1]
        )
        wscale_shape = (scale_size, scale_size)
        self.wscale = [torch.rand(wscale_shape, dtype=torch.float32) for _ in range(3)]

        # Initialize weights (transposed if using aiter, otherwise not)
        self.w = [
            torch.rand(hidden_size, hidden_size).to(dtype=FP8_DTYPE) for _ in range(3)
        ]
        if use_aiter_fusion:
            self.w = [w.t() for w in self.w]

        self.fp8_linear_layers = [
            TestBlockFP8Layer(
                group_shape=group_shape,
                weight=self.w[i],
                weight_scale=self.wscale[i],
                input_scale=None,  # Dynamic quantization for blockwise
                cutlass_block_fp8_supported=cutlass_block_fp8_supported(),
                use_aiter_and_is_supported=use_aiter_quant,
            )
            for i in range(3)
        ]

        self.enable_quant_fp8_custom_op = (
            False
            if use_aiter_quant
            else self.fp8_linear_layers[0].linear_op.input_quant_op.enabled()
        )

    def forward(self, x):
        # avoid having graph input be an arg to a pattern directly
        x = resid = torch.relu(x)
        y = self.norm[0](x)

        x2 = self.fp8_linear_layers[0](y)
        # make sure resid is used for replacement to work
        y2, resid = self.norm[1](x2, resid)

        x3 = self.fp8_linear_layers[1](y2)

        y3, resid = self.norm[2](x3, resid)  # use resid here

        x4 = self.fp8_linear_layers[2](y3)

        y4, resid = self.norm[3](x4, resid)  # use resid here
        return y4

    def ops_in_model_before(self):
        if self.group_shape.is_per_group():
            # Blockwise path
            if self.use_aiter_fusion and self.use_aiter_quant_op:
                return [rocm_aiter_ops.get_group_quant_op()]
            if self.use_aiter_fusion:
                return [torch.ops.vllm.triton_per_token_group_quant_fp8.default]
        else:
            if self.use_aiter_quant_op:
                return [rocm_aiter_ops.get_per_token_quant_op()]

        # Common path
        return (
            [QUANT_OPS[self.activation_quant_key]]
            if self.enable_quant_fp8_custom_op
            else [torch.ops.aten.reciprocal]
        )

    def ops_in_model_after(self):
        if self.use_aiter_fusion:
            if self.group_shape.is_per_group():
                # Blockwise aiter fusion
                from vllm.compilation.rocm_aiter_fusion import (
                    AiterFusedAddRMSFp8GroupQuantPattern,
                    AiterRMSFp8GroupQuantPattern,
                )

                return [
                    AiterFusedAddRMSFp8GroupQuantPattern.FUSED_OP,
                    AiterRMSFp8GroupQuantPattern.FUSED_OP,
                ]
            else:
                # Per-token aiter fusion
                from vllm.compilation.rocm_aiter_fusion import (
                    AiterFusedAddRMSNormDynamicQuantPattern,
                    AiterRMSNormDynamicQuantPattern,
                )

                return [
                    AiterFusedAddRMSNormDynamicQuantPattern.FUSED_OP,
                    AiterRMSNormDynamicQuantPattern.FUSED_OP,
                ]

        # Regular fusion
        return [
            FUSED_OPS[FusedRMSQuantKey(self.activation_quant_key, True)],
            FUSED_OPS[FusedRMSQuantKey(self.activation_quant_key, False)],
        ]

    def ops_in_model_before_partial(self):
        return (
            [RMS_OP, RMS_ADD_OP]
            if self.enable_rms_norm_custom_op
            else [torch.ops.aten.rsqrt]
        )


def _run_fusion_test(
    model,
    fusion_pass,
    vllm_config,
    dtype,
    hidden_size,
    num_tokens,
):
    """Helper function for common fusion test logic.

    Must be called within vllm_config context.
    """
    noop_pass = NoOpEliminationPass(vllm_config)
    cleanup_pass = PostCleanupPass(vllm_config)

    backend = TestBackend(noop_pass, fusion_pass, cleanup_pass)
    backend2 = TestBackend(noop_pass, cleanup_pass)

    x = torch.rand(num_tokens, hidden_size)
    torch._dynamo.mark_dynamic(x, 0)

    model_fused = torch.compile(model, backend=backend)
    result_fused = model_fused(x)

    model_unfused = torch.compile(model, backend=backend2)
    result_unfused = model_unfused(x)

    if dtype == torch.float16:
        ATOL, RTOL = (2e-3, 2e-3)
    else:
        ATOL, RTOL = (1e-2, 1e-2)

    torch.testing.assert_close(result_fused, result_unfused, atol=ATOL, rtol=RTOL)

    assert fusion_pass.matched_count == 3
    backend.check_before_ops(model.ops_in_model_before())
    backend.check_after_ops(model.ops_in_model_after())

    return backend, backend2


@pytest.mark.parametrize("dtype", [torch.float16, torch.bfloat16])
@pytest.mark.parametrize("hidden_size", [256])
@pytest.mark.parametrize("num_tokens", [257])
@pytest.mark.parametrize("eps", [1e-5, 1e-6])
@pytest.mark.parametrize("kernel_groupshape", KERNEL_GROUPSHAPE_COMBINATIONS)
@pytest.mark.parametrize("enable_rms_norm_custom_op", [True, False])
@pytest.mark.parametrize("enable_quant_fp8_custom_op", [True, False])
@pytest.mark.skipif(
    not current_platform.is_cuda_alike(), reason="Only test on CUDA and ROCm"
)
def test_fusion_rmsnorm_quant(
    dtype,
    hidden_size,
    num_tokens,
    eps,
    kernel_groupshape,
    enable_rms_norm_custom_op,
    enable_quant_fp8_custom_op,
):
    force_kernel, group_shape = kernel_groupshape

    if not enable_quant_fp8_custom_op and group_shape.is_per_group():
        pytest.skip("Unsupported unwrapped quant fp8 op for blockwise quantization")

    if group_shape == GroupShape(1, 64) and (
        cutlass_block_fp8_supported() or is_deep_gemm_supported()
    ):
        pytest.skip("Unsupported group shape 64 for CUTLASS/DeepGemm")

    custom_ops = []
    if enable_rms_norm_custom_op:
        custom_ops.append("+rms_norm")
    if enable_quant_fp8_custom_op:
        custom_ops.append("+quant_fp8")

    vllm_config = VllmConfig(
        model_config=ModelConfig(dtype=dtype),
        compilation_config=CompilationConfig(
            mode=CompilationMode.VLLM_COMPILE,
            custom_ops=custom_ops,
            pass_config=PassConfig(
                fuse_norm_quant=True, fuse_act_quant=True, eliminate_noops=True
            ),
        ),
    )

    with vllm.config.set_current_vllm_config(vllm_config):
        # Setup device before model creation
        torch.set_default_device("cuda")
        torch.set_default_dtype(dtype)
        torch.manual_seed(1)

        fusion_pass = RMSNormQuantFusionPass(vllm_config)

        model = TestModel(
            hidden_size=hidden_size,
            eps=eps,
            force_kernel=force_kernel,
            group_shape=group_shape,
            use_aiter_fusion=False,
            use_aiter_quant=False,
        )

        backend, _ = _run_fusion_test(
            model, fusion_pass, vllm_config, dtype, hidden_size, num_tokens
        )
        backend.check_before_ops(
            model.ops_in_model_before_partial(), fully_replaced=False
        )

        # If RMSNorm custom op is disabled (native/torch impl used),
        # there's a risk that the fused add doesn't get included in the
        # replacement and only the rms part gets fused with quant.
        # Hence, we check only 2 add nodes are left (final fused rmsnorm add).
        if not enable_rms_norm_custom_op:
            n_add_nodes = lambda g: sum(1 for _ in find_op_nodes(torch.ops.aten.add, g))
            # 7 = 1 (RMS) + 3x2 (3xRMS_ADD, 2 each)
            assert n_add_nodes(backend.graph_pre_pass) == 7
            assert n_add_nodes(backend.graph_post_pass) == 2


@pytest.mark.parametrize("dtype", [torch.bfloat16])
@pytest.mark.parametrize("hidden_size", [256])
@pytest.mark.parametrize("num_tokens", [257])
@pytest.mark.parametrize("eps", [1e-5, 1e-6])
@pytest.mark.parametrize(
    "kernel_groupshape_quant", AITER_KERNEL_GROUPSHAPE_COMBINATIONS
)
@pytest.mark.skipif(
    (not current_platform.is_rocm() or not IS_AITER_FOUND),
    reason="Only test on ROCm with aiter package installed",
)
def test_aiter_fusion_rmsnorm_quant(
    dtype: torch.dtype,
    hidden_size: int,
    num_tokens: int,
    eps: float,
    kernel_groupshape_quant: tuple,
    monkeypatch: pytest.MonkeyPatch,
):
    force_kernel, group_shape, use_aiter_quant_op = kernel_groupshape_quant
    vllm_config = VllmConfig(
        model_config=ModelConfig(dtype=dtype),
        compilation_config=CompilationConfig(
            mode=CompilationMode.VLLM_COMPILE,
            custom_ops=["+rms_norm", "+quant_fp8"],
            pass_config=PassConfig(fuse_norm_quant=True, eliminate_noops=True),
        ),
    )

    with vllm.config.set_current_vllm_config(vllm_config), monkeypatch.context() as m:
        from vllm.compilation.rocm_aiter_fusion import RocmAiterRMSNormFusionPass

        m.setenv("VLLM_ROCM_USE_AITER", "1")

        rocm_aiter_ops.refresh_env_variables()

        torch.set_default_device("cuda")
        torch.set_default_dtype(dtype)
        torch.manual_seed(1)

        fusion_pass = RocmAiterRMSNormFusionPass(vllm_config)

        model = TestModel(
            hidden_size=hidden_size,
            eps=eps,
            force_kernel=force_kernel,
            group_shape=group_shape,
            use_aiter_fusion=True,  # Always use aiter fusion ops in aiter test
            use_aiter_quant=use_aiter_quant_op,  # Toggle aiter quantization
        )

        _run_fusion_test(
            model, fusion_pass, vllm_config, dtype, hidden_size, num_tokens
        )