linear.py

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

import itertools
from abc import abstractmethod
from typing import Any, Literal, Optional, Union

import torch
import torch.nn as nn
from torch.nn.parameter import Parameter, UninitializedParameter

from vllm.distributed import (
    divide,
    get_tensor_model_parallel_rank,
    get_tensor_model_parallel_world_size,
    split_tensor_along_last_dim,
    tensor_model_parallel_all_gather,
    tensor_model_parallel_all_reduce,
)
from vllm.logger import init_logger
from vllm.model_executor.custom_op import CustomOp
from vllm.model_executor.layers.quantization.base_config import (
    QuantizationConfig,
    QuantizeMethodBase,
)
from vllm.model_executor.layers.utils import dispatch_unquantized_gemm

# yapf: disable
from vllm.model_executor.parameter import (
    BasevLLMParameter,
    BlockQuantScaleParameter,
    ModelWeightParameter,
    PackedColumnParameter,
    PackedvLLMParameter,
    PerTensorScaleParameter,
    RowvLLMParameter,
)

# yapf: enable
from vllm.model_executor.utils import set_weight_attrs
from vllm.platforms import current_platform
from vllm.utils import GiB_bytes

logger = init_logger(__name__)

WEIGHT_LOADER_V2_SUPPORTED = [
    "UnquantizedLinearMethod",
    "CompressedTensorsLinearMethod",
    "CompressedTensorsLinearTransformMethod",
    "BitBLASLinearMethod",
    "GPTQBitBLASLinearMethod",
    "AWQMarlinLinearMethod",
    "AWQLinearMethod",
    "GPTQMarlinLinearMethod",
    "Fp8LinearMethod",
    "MarlinLinearMethod",
    "GPTQMarlin24LinearMethod",
    "TPUInt8LinearMethod",
    "GPTQLinearMethod",
    "FBGEMMFp8LinearMethod",
    "ModelOptFp8LinearMethod",
    "IPEXAWQLinearMethod",
    "IPEXGPTQLinearMethod",
    "HQQMarlinMethod",
    "QuarkLinearMethod",
    "ModelOptNvFp4LinearMethod",
    "PetitNvFp4LinearMethod",
]


def adjust_bitblas_shard(param, shard_size, shard_offset):
    bitblas_tile_size = getattr(param, "bitblas_tile_size", None)
    if bitblas_tile_size is not None:
        return (shard_size // bitblas_tile_size, shard_offset // bitblas_tile_size)

    return shard_size, shard_offset


def adjust_marlin_shard(param, shard_size, shard_offset):
    marlin_tile_size = getattr(param, "marlin_tile_size", None)
    if marlin_tile_size is None:
        return shard_size, shard_offset

    return shard_size * marlin_tile_size, shard_offset * marlin_tile_size


def adjust_bitsandbytes_4bit_shard(
    param: Parameter, shard_offsets: dict[str, tuple[int, int]], loaded_shard_id: str
) -> tuple[int, int]:
    """Adjust the quantization offsets and sizes for BitsAndBytes sharding."""

    total, _ = shard_offsets["total"]
    orig_offset, orig_size = shard_offsets[loaded_shard_id]

    quantized_total = param.data.shape[0]
    quantized_offset = orig_offset * quantized_total // total
    quantized_size = orig_size * quantized_total // total

    return quantized_size, quantized_offset


def adjust_scalar_to_fused_array(param, loaded_weight, shard_id):
    """For fused modules (QKV and MLP) we have an array of length
    N that holds 1 scale for each "logical" matrix. So the param
    is an array of length N. The loaded_weight corresponds to
    one of the shards on disk. Here, we slice the param based on
    the shard_id for loading.
    """
    qkv_idxs = {"q": 0, "k": 1, "v": 2}

    if isinstance(shard_id, str):
        shard_id = qkv_idxs[shard_id]
    elif not isinstance(shard_id, int):
        raise ValueError(f"Unknown Shard Id {shard_id}")

    # AutoFP8 scales do not have a shape
    # compressed-tensors scales do have a shape
    if len(loaded_weight.shape) != 0:
        assert loaded_weight.shape[0] == 1
        loaded_weight = loaded_weight[0]

    return param[shard_id], loaded_weight


# TODO(Isotr0py): We might need a more flexible structure to handle
# bitsandbytes shard offsets.
def left_shift_bitsandbytes_4bit_shard(bnb_weight_attrs: dict[str, Any]):
    """
    Separate the BitsAndBytes 4-bit shard.

    For example, given bnb weight attributes as below:
    {
        'bnb_shard_offsets': array([0, 4, 8, 16]),
        'bnb_quant_state': {0: ..., 1: ..., 2: ...},
    }

    The function will return:
    {
        'bnb_shard_offsets': array([0, 4]),
        'bnb_quant_state': {0: ...},
    }
    and
    {
        'bnb_shard_offsets': array([0, 4, 12]),
        'bnb_quant_state': {0: ..., 1: ...},
    }
    """
    shard_offsets = bnb_weight_attrs["bnb_shard_offsets"]
    offset_l = shard_offsets[:2]
    offset_r = shard_offsets[1:] - shard_offsets[1]
    quant_state_l = {0: bnb_weight_attrs["bnb_quant_state"][0]}
    quant_state_r = {
        i - 1: bnb_weight_attrs["bnb_quant_state"][i]
        for i in range(1, len(shard_offsets) - 1)
    }
    left = dict(bnb_shard_offsets=offset_l, bnb_quant_state=quant_state_l)
    right = dict(bnb_shard_offsets=offset_r, bnb_quant_state=quant_state_r)
    return left, right


class LinearMethodBase(QuantizeMethodBase):
    """Base class for different (maybe quantized) linear methods."""

    @abstractmethod
    def create_weights(
        self,
        layer: torch.nn.Module,
        input_size_per_partition: int,
        output_partition_sizes: list[int],
        input_size: int,
        output_size: int,
        params_dtype: torch.dtype,
        **extra_weight_attrs,
    ):
        """Create weights for a linear layer.
           The weights will be set as attributes of the layer.

        Args:
            layer: The layer that is using the LinearMethodBase factory.
            input_size_per_partition: Size of the weight input dim on rank X.
            output_partition_sizes: Sizes of the output dim of each logical
                weight on rank X. E.g., output_partition_sizes for QKVLinear
                is a list contains the width of Wq, Wk, Wv on rank X.
            input_size: Size of the input dim of the weight across all ranks.
            output_size: Size of the output dim of the weight across all ranks.
            params_dtype: Datatype of the parameters.
        """
        raise NotImplementedError

    @abstractmethod
    def apply(
        self,
        layer: torch.nn.Module,
        x: torch.Tensor,
        bias: Optional[torch.Tensor] = None,
    ) -> torch.Tensor:
        """Apply the weights in layer to the input tensor.
        Expects create_weights to have been called before on the layer."""
        raise NotImplementedError


class UnquantizedLinearMethod(LinearMethodBase):
    """Linear method without quantization."""

    def create_weights(
        self,
        layer: torch.nn.Module,
        input_size_per_partition: int,
        output_partition_sizes: list[int],
        input_size: int,
        output_size: int,
        params_dtype: torch.dtype,
        **extra_weight_attrs,
    ):
        # This method creates unquantized linear weights.
        # The weights are not quantized, and they are not sharded.
        # The amount of memory allocated for the weights is
        # sum(output_partition_sizes) * input_size_per_partition.
        try:
            weight_loader = extra_weight_attrs.pop("weight_loader")
            weight = ModelWeightParameter(
                data=torch.empty(
                    sum(output_partition_sizes),
                    input_size_per_partition,
                    dtype=params_dtype,
                ),
                input_dim=1,
                output_dim=0,
                weight_loader=weight_loader,
            )
        except torch.cuda.OutOfMemoryError as e:
            logger.error("Failed to create unquantized linear weights: %s", e)
            if torch.cuda.is_available():
                logger.debug("CUDA device: %s", torch.cuda.current_device())
                logger.debug(
                    "Allocated: %.2f GiB", torch.cuda.memory_allocated() / GiB_bytes
                )
                logger.debug(
                    "Reserved: %.2f GiB", torch.cuda.memory_reserved() / GiB_bytes
                )
            raise RuntimeError(
                "Failed to create unquantized linear weights. "
                "This may be caused by insufficient memory to allocate "
                "the weight."
            ) from e

        layer.register_parameter("weight", weight)
        set_weight_attrs(weight, extra_weight_attrs)

    def process_weights_after_loading(self, layer: torch.nn.Module) -> None:
        if current_platform.is_cpu():
            from vllm.model_executor.layers.utils import dispatch_cpu_unquantized_gemm

            dispatch_cpu_unquantized_gemm(layer, remove_weight=True)

    def apply(
        self,
        layer: torch.nn.Module,
        x: torch.Tensor,
        bias: Optional[torch.Tensor] = None,
    ) -> torch.Tensor:
        return dispatch_unquantized_gemm()(layer, x, layer.weight, bias)


class LinearBase(CustomOp):
    """Base linear layer.

    Args:
        input_size: input dimension of the linear layer.
        output_size: output dimension of the linear layer.
        skip_bias_add: If true, skip adding bias but instead return it.
        params_dtype: Data type for the parameters.
        quant_config: Quantization configure.
        prefix: Prefix for parameter names.
        return_bias: If true, return bias together with outputs in forward pass.
        disable_tp: If true, tensor parallelism will be disabled for this layer.
    """

    def __init__(
        self,
        input_size: int,
        output_size: int,
        skip_bias_add: bool = False,
        params_dtype: Optional[torch.dtype] = None,
        quant_config: Optional[QuantizationConfig] = None,
        prefix: str = "",
        *,
        return_bias: bool = True,
        disable_tp: bool = False,
    ):
        super().__init__()

        # Keep input parameters
        self.input_size = input_size
        self.output_size = output_size
        self.skip_bias_add = skip_bias_add
        if params_dtype is None:
            params_dtype = torch.get_default_dtype()
        self.params_dtype = params_dtype
        self.quant_config = quant_config
        self.prefix = prefix
        if quant_config is None:
            self.quant_method: Optional[QuantizeMethodBase] = UnquantizedLinearMethod()
        else:
            self.quant_method = quant_config.get_quant_method(self, prefix=prefix)
        self.return_bias = return_bias
        self.disable_tp = disable_tp
        self.tp_rank = get_tensor_model_parallel_rank() if not disable_tp else 0
        self.tp_size = get_tensor_model_parallel_world_size() if not disable_tp else 1

    def update_param_tp_status(self):
        for param in self.parameters():
            if isinstance(param, BasevLLMParameter):
                param.tp_rank = self.tp_rank
                param.tp_size = self.tp_size


@CustomOp.register("replicated_linear")
class ReplicatedLinear(LinearBase):
    """Replicated linear layer.

    Args:
        input_size: input dimension of the linear layer.
        output_size: output dimension of the linear layer.
        bias: If true, add bias.
        skip_bias_add: If true, skip adding bias but instead return it.
        params_dtype: Data type for the parameters.
        quant_config: Quantization configure.
        prefix: The name of the layer in the state dict, including all parents
                        (e.g. model.layers.0.qkv_proj)
        return_bias: If true, return bias together with outputs in forward pass.
        disable_tp: Take no effect for replicated linear layers.
    """

    def __init__(
        self,
        input_size: int,
        output_size: int,
        bias: bool = True,
        skip_bias_add: bool = False,
        params_dtype: Optional[torch.dtype] = None,
        quant_config: Optional[QuantizationConfig] = None,
        prefix: str = "",
        *,
        return_bias: bool = True,
        disable_tp: bool = False,
    ):
        # If MergedReplicatedLinear, use output size of each partition.
        if hasattr(self, "output_sizes"):
            self.output_partition_sizes = self.output_sizes
        else:
            self.output_partition_sizes = [output_size]

        super().__init__(
            input_size,
            output_size,
            skip_bias_add,
            params_dtype,
            quant_config,
            prefix=prefix,
            return_bias=return_bias,
            disable_tp=disable_tp,
        )

        # All the linear layer supports quant method.
        assert self.quant_method is not None
        self.quant_method.create_weights(
            self,
            self.input_size,
            self.output_partition_sizes,
            self.input_size,
            self.output_size,
            self.params_dtype,
            weight_loader=self.weight_loader,
        )

        if bias:
            self.bias = Parameter(
                torch.empty(self.output_size, dtype=self.params_dtype)
            )
            set_weight_attrs(
                self.bias,
                {
                    "output_dim": 0,
                    "weight_loader": self.weight_loader,
                },
            )
        else:
            self.register_parameter("bias", None)

    def weight_loader(self, param: Parameter, loaded_weight: torch.Tensor):
        # If the weight on disk does not have a shape, give it one
        # (such scales for AutoFp8).
        # Special case for GGUF

        is_gguf_weight = getattr(param, "is_gguf_weight", False)
        is_gguf_weight_type = getattr(param, "is_gguf_weight_type", False)
        if is_gguf_weight_type:
            param.weight_type = loaded_weight.item()

        # Materialize GGUF UninitializedParameter
        if is_gguf_weight and isinstance(param, UninitializedParameter):
            param.materialize(loaded_weight.shape, dtype=loaded_weight.dtype)

        if len(loaded_weight.shape) == 0:
            loaded_weight = loaded_weight.reshape(1)

        assert param.size() == loaded_weight.size(), (
            f"Tried to load weights of size {loaded_weight.size()}"
            f"to a parameter of size {param.size()}"
        )
        param.data.copy_(loaded_weight)

    def forward(
        self,
        x: torch.Tensor,
    ) -> Union[torch.Tensor, tuple[torch.Tensor, Optional[Parameter]]]:
        bias = self.bias if not self.skip_bias_add else None
        assert self.quant_method is not None

        output = self.quant_method.apply(self, x, bias)
        output_bias = self.bias if self.skip_bias_add else None

        if not self.return_bias:
            return output
        return output, output_bias

    def extra_repr(self) -> str:
        s = f"in_features={self.input_size}"
        s += f", output_features={self.output_size}"
        s += f", bias={self.bias is not None}"
        return s


@CustomOp.register("column_parallel_linear")
class ColumnParallelLinear(LinearBase):
    """Linear layer with column parallelism.

    The linear layer is defined as Y = XA + b. A is parallelized along
    its second dimension as A = [A_1, ..., A_p].

    Args:
        input_size: first dimension of matrix A.
        output_size: second dimension of matrix A.
        bias: If true, add bias.
        gather_output: If true, call all-gather on output and make Y available
                       to all GPUs, otherwise, every GPU will have its output
                       which is Y_i = XA_i
        skip_bias_add: This was added to enable performance optimizations where
                       bias can be fused with other element-wise operations. we
                       skip adding bias but instead return it.
        params_dtype: Data type for the parameters.
        quant_config: Quantization configure.
        output_sizes: list of output sizes packed into one output, like for QKV
                       the list would be size 3.
        prefix: The name of the layer in the state dict, including all parents
                        (e.g. model.layers.0.qkv_proj)
        return_bias: If true, return bias together with outputs in forward pass.
        disable_tp: If true, weights matrix won't be sharded through tp rank.
    """

    def __init__(
        self,
        input_size: int,
        output_size: int,
        bias: bool = True,
        gather_output: bool = False,
        skip_bias_add: bool = False,
        params_dtype: Optional[torch.dtype] = None,
        quant_config: Optional[QuantizationConfig] = None,
        output_sizes: Optional[list[int]] = None,
        prefix: str = "",
        *,
        return_bias: bool = True,
        disable_tp: bool = False,
    ):
        # Divide the weight matrix along the last dimension.
        self.tp_rank = get_tensor_model_parallel_rank() if not disable_tp else 0
        self.tp_size = get_tensor_model_parallel_world_size() if not disable_tp else 1
        self.input_size_per_partition = input_size
        self.output_size_per_partition = divide(output_size, self.tp_size)
        self.output_partition_sizes = [self.output_size_per_partition]
        # If QKV or MergedColumn, use output size of each partition.
        if hasattr(self, "output_sizes"):
            self.output_partition_sizes = [
                divide(output_size, self.tp_size) for output_size in self.output_sizes
            ]

        super().__init__(
            input_size,
            output_size,
            skip_bias_add,
            params_dtype,
            quant_config,
            prefix,
            return_bias=return_bias,
            disable_tp=disable_tp,
        )

        self.gather_output = gather_output

        if output_sizes is None:
            output_sizes = [output_size]

        assert self.quant_method is not None
        self.quant_method.create_weights(
            layer=self,
            input_size_per_partition=self.input_size_per_partition,
            output_partition_sizes=self.output_partition_sizes,
            input_size=self.input_size,
            output_size=self.output_size,
            params_dtype=self.params_dtype,
            weight_loader=(
                self.weight_loader_v2
                if self.quant_method.__class__.__name__ in WEIGHT_LOADER_V2_SUPPORTED
                else self.weight_loader
            ),
        )
        if bias:
            self.bias = Parameter(
                torch.empty(self.output_size_per_partition, dtype=params_dtype)
            )
            set_weight_attrs(
                self.bias,
                {
                    "output_dim": 0,
                    "weight_loader": self.weight_loader,
                },
            )
        else:
            self.register_parameter("bias", None)
        self.update_param_tp_status()

    def weight_loader(self, param: Parameter, loaded_weight: torch.Tensor):
        output_dim = getattr(param, "output_dim", None)

        is_sharded_weight = getattr(param, "is_sharded_weight", False)
        use_bitsandbytes_4bit = getattr(param, "use_bitsandbytes_4bit", False)
        # bitsandbytes loads the weights of the specific portion
        # no need to narrow
        is_sharded_weight = is_sharded_weight or use_bitsandbytes_4bit

        # Special case for GGUF
        is_gguf_weight = getattr(param, "is_gguf_weight", False)
        is_gguf_weight_type = getattr(param, "is_gguf_weight_type", False)
        if is_gguf_weight_type:
            param.weight_type = loaded_weight.item()

        # Materialize GGUF UninitializedParameter
        if is_gguf_weight and isinstance(param, UninitializedParameter):
            final_shape = list(loaded_weight.shape)
            if output_dim is not None:
                assert final_shape[output_dim] % self.tp_size == 0
                final_shape[output_dim] = final_shape[output_dim] // self.tp_size
            param.materialize(final_shape, dtype=loaded_weight.dtype)

        param_data = param.data
        if output_dim is not None and not is_sharded_weight:
            shard_size = param_data.shape[output_dim]
            start_idx = self.tp_rank * shard_size
            loaded_weight = loaded_weight.narrow(output_dim, start_idx, shard_size)

        # Special case for loading scales off disk, which often do not
        # have a shape (such as in the case of AutoFP8).
        if len(loaded_weight.shape) == 0:
            loaded_weight = loaded_weight.reshape(1)

        assert param_data.shape == loaded_weight.shape
        param_data.copy_(loaded_weight)

    def weight_loader_v2(self, param: BasevLLMParameter, loaded_weight: torch.Tensor):
        # Special case for loading scales off disk, which often do not
        # have a shape (such as in the case of AutoFP8).
        if len(loaded_weight.shape) == 0:
            assert loaded_weight.numel() == 1
            loaded_weight = loaded_weight.reshape(1)
        param.load_column_parallel_weight(loaded_weight=loaded_weight)

    def forward(
        self,
        input_,
    ) -> Union[torch.Tensor, tuple[torch.Tensor, Optional[Parameter]]]:
        bias = self.bias if not self.skip_bias_add else None

        # Matrix multiply.
        assert self.quant_method is not None
        output_parallel = self.quant_method.apply(self, input_, bias)

        if self.gather_output and self.tp_size > 1:
            # All-gather across the partitions.
            output = tensor_model_parallel_all_gather(output_parallel)
        else:
            output = output_parallel
        output_bias = self.bias if self.skip_bias_add else None
        if not self.return_bias:
            return output
        return output, output_bias

    def extra_repr(self) -> str:
        s = f"in_features={self.input_size}"
        s += f", output_features={self.output_size_per_partition}"
        s += f", bias={self.bias is not None}"
        s += f", tp_size={self.tp_size}"
        s += f", gather_output={self.gather_output}"
        return s


class MergedColumnParallelLinear(ColumnParallelLinear):
    """Packed linear layers with column parallelism.

    Similar to ColumnParallelLinear, but the weight matrix is concatenated
    along the output dimension. When the weight matrix is loaded, the
    different partitions are sharded separately.

    Args:
        input_size: input dimension of the linear layer.
        output_sizes: list of output dimensions of the linear layer.
        bias: If true, add bias.
        gather_output: If true, call all-gather on output and make the output
                       available to all GPUs, otherwise, every GPU will have
                       its own output.
        skip_bias_add: This was added to enable performance optimizations where
                       bias can be fused with other element-wise operations. we
                       skip adding bias but instead return it.
        params_dtype: Data type for the parameters.
        quant_config: Quantization configure.
        prefix: The name of the layer in the state dict, including all parents
                        (e.g. model.layers.0.qkv_proj)
        return_bias: If true, return bias together with outputs in forward pass.
        disable_tp: If true, all weights matrix won't be sharded, this layer
                    will be treated as a "Replicated" MergedLinear.
    """

    def __init__(
        self,
        input_size: int,
        output_sizes: list[int],
        bias: bool = True,
        gather_output: bool = False,
        skip_bias_add: bool = False,
        params_dtype: Optional[torch.dtype] = None,
        quant_config: Optional[QuantizationConfig] = None,
        prefix: str = "",
        *,
        return_bias: bool = True,
        disable_tp: bool = False,
    ):
        self.output_sizes = output_sizes
        self.tp_size = get_tensor_model_parallel_world_size() if not disable_tp else 1
        self.tp_rank = get_tensor_model_parallel_rank() if not disable_tp else 0

        assert all(output_size % self.tp_size == 0 for output_size in output_sizes)
        super().__init__(
            input_size=input_size,
            output_size=sum(output_sizes),
            bias=bias,
            gather_output=gather_output,
            skip_bias_add=skip_bias_add,
            params_dtype=params_dtype,
            quant_config=quant_config,
            prefix=prefix,
            return_bias=return_bias,
            disable_tp=disable_tp,
        )

    def weight_loader(
        self,
        param: Parameter,
        loaded_weight: torch.Tensor,
        loaded_shard_id: Optional[int] = None,
    ):
        # Special case for GGUF
        # initialize GGUF param after we know the quantize type
        is_gguf_weight = getattr(param, "is_gguf_weight", False)
        is_gguf_weight_type = getattr(param, "is_gguf_weight_type", False)
        if is_gguf_weight_type:
            if loaded_shard_id is not None:
                param.data[loaded_shard_id].copy_(loaded_weight)
                param.shard_weight_type[loaded_shard_id] = loaded_weight.item()
            else:
                param.shard_weight_type = {
                    i: loaded_weight.item() for i, _ in enumerate(self.output_sizes)
                }
            return

        if is_gguf_weight:
            output_dim = getattr(param, "output_dim", None)
            shard_size = loaded_weight.size(output_dim) // self.tp_size
            start_idx = self.tp_rank * shard_size

            if loaded_shard_id is not None:
                loaded_weight = loaded_weight.narrow(output_dim, start_idx, shard_size)
                param.shard_id.append(loaded_shard_id)
                param.shard_id_map[loaded_shard_id] = len(param.data_container)
                param.data_container.append(loaded_weight)
                return

        param_data = param.data
        output_dim = getattr(param, "output_dim", None)
        # Special case for per-tensor scale to load scalar into fused array.
        needs_scalar_to_array = getattr(param, "needs_scalar_to_array", False)

        if loaded_shard_id is None:
            # Loaded weight is already fused on disk (mlp).
            # (e.g., Phi-3's gate_up_proj).
            if output_dim is None:
                if needs_scalar_to_array:
                    param_data, loaded_weight = adjust_scalar_to_fused_array(
                        param_data, loaded_weight, 0
                    )

                assert param_data.shape == loaded_weight.shape
                param_data.copy_(loaded_weight)
                return
            current_shard_offset = 0
            use_bitsandbytes_4bit = getattr(param, "use_bitsandbytes_4bit", False)
            shard_offsets: list[tuple[int, int, int]] = []
            for i, output_size in enumerate(self.output_sizes):
                shard_offsets.append((i, current_shard_offset, output_size))
                current_shard_offset += output_size
            packed_dim = getattr(param, "packed_dim", None)
            for shard_id, shard_offset, shard_size in shard_offsets:
                # Special case for Quantization.
                # If quantized, we need to adjust the offset and size to account
                # for the packing.
                if packed_dim == output_dim:
                    shard_size = shard_size // param.packed_factor
                    shard_offset = shard_offset // param.packed_factor
                    # Special case for Marlin.
                    shard_size, shard_offset = adjust_marlin_shard(
                        param, shard_size, shard_offset
                    )

                shard_size, shard_offset = adjust_bitblas_shard(
                    param, shard_size, shard_offset
                )

                if use_bitsandbytes_4bit:
                    index = list(itertools.accumulate([0] + self.output_sizes))
                    orig_offsets = {
                        str(i): (index[i], size)
                        for i, size in enumerate(self.output_sizes)
                    }
                    orig_offsets["total"] = (self.output_size, 0)
                    shard_size, shard_offset = adjust_bitsandbytes_4bit_shard(
                        param, orig_offsets, str(shard_id)
                    )

                loaded_weight_shard = loaded_weight.narrow(
                    output_dim, shard_offset, shard_size
                )
                self.weight_loader(param, loaded_weight_shard, shard_id)
            return

        assert loaded_shard_id < len(self.output_sizes)
        if output_dim is not None:
            shard_offset = sum(self.output_sizes[:loaded_shard_id]) // self.tp_size
            shard_size = self.output_sizes[loaded_shard_id] // self.tp_size
            # Special case for quantization.
            # If quantized, we need to adjust the offset and size to account
            # for the packing.
            packed_dim = getattr(param, "packed_dim", None)
            if packed_dim == output_dim:
                shard_size = shard_size // param.packed_factor
                shard_offset = shard_offset // param.packed_factor
                # Special case for Marlin.
                shard_size, shard_offset = adjust_marlin_shard(
                    param, shard_size, shard_offset
                )
            shard_size, shard_offset = adjust_bitblas_shard(
                param, shard_size, shard_offset
            )

            use_bitsandbytes_4bit = getattr(param, "use_bitsandbytes_4bit", False)
            is_sharded_weight = getattr(param, "is_sharded_weight", False)
            # bitsandbytes loads the weights of the specific portion
            # no need to narrow
            is_sharded_weight = is_sharded_weight or use_bitsandbytes_4bit

            if use_bitsandbytes_4bit:
                shard_size = loaded_weight.shape[output_dim]
                shard_offset = loaded_weight.shape[output_dim] * loaded_shard_id

            param_data = param_data.narrow(output_dim, shard_offset, shard_size)
            start_idx = self.tp_rank * shard_size
            if not is_sharded_weight:
                loaded_weight = loaded_weight.narrow(output_dim, start_idx, shard_size)
        # Special case for per-tensor scales in fused case.
        elif needs_scalar_to_array:
            param_data, loaded_weight = adjust_scalar_to_fused_array(
                param_data, loaded_weight, loaded_shard_id
            )

        else:
            ignore_warning = getattr(param, "ignore_warning", False)
            if not ignore_warning:
                logger.warning(
                    "Loading a weight without `output_dim` attribute in "
                    "MergedColumnParallelLinear, assume the weight is "
                    "the same for all partitions."
                )

        assert param_data.shape == loaded_weight.shape
        param_data.copy_(loaded_weight)

    def _load_fused_module_from_checkpoint(
        self, param: BasevLLMParameter, loaded_weight: torch.Tensor
    ):
        """
        Handle special case for models where MLP layers are already
        fused on disk. In this case, we have no shard id. This function
        determines the shard id by splitting these layers and then calls
        the weight loader using the shard id.

        An example of a model with these fused layers:
        https://huggingface.co/microsoft/Phi-3-mini-4k-instruct
        """

        current_shard_offset = 0
        shard_offsets: list[tuple[int, int, int]] = []
        for i, output_size in enumerate(self.output_sizes):
            shard_offsets.append((i, current_shard_offset, output_size))
            current_shard_offset += output_size

        for shard_id, shard_offset, shard_size in shard_offsets:
            # Special case for Quantization.
            # If quantized, we need to adjust the offset and size to account
            # for the packing.
            if (
                isinstance(param, (PackedColumnParameter, PackedvLLMParameter))
                and param.packed_dim == param.output_dim
            ):
                shard_size, shard_offset = param.adjust_shard_indexes_for_packing(
                    shard_size=shard_size, shard_offset=shard_offset
                )

            loaded_weight_shard = loaded_weight.narrow(
                param.output_dim, shard_offset, shard_size
            )
            self.weight_loader_v2(param, loaded_weight_shard, shard_id)

    def weight_loader_v2(
        self,
        param: BasevLLMParameter,
        loaded_weight: torch.Tensor,
        loaded_shard_id: Optional[int] = None,
    ):
        if loaded_shard_id is None:
            if isinstance(param, PerTensorScaleParameter):
                param.load_merged_column_weight(loaded_weight=loaded_weight, shard_id=0)
                return
            elif type(param) in (RowvLLMParameter, BasevLLMParameter):
                param.load_merged_column_weight(loaded_weight=loaded_weight)
                return
            # TODO: @dsikka - move to parameter.py
            self._load_fused_module_from_checkpoint(param, loaded_weight)
            return

        assert loaded_shard_id < len(self.output_sizes)

        if isinstance(param, BlockQuantScaleParameter):
            assert self.quant_method is not None
            # Assume the weight block size has been set by quant method
            assert hasattr(self, "weight_block_size")
            weight_block_size = self.weight_block_size
            assert weight_block_size is not None
            block_n, _ = weight_block_size[0], weight_block_size[1]
            shard_offset = (
                (sum(self.output_sizes[:loaded_shard_id]) + block_n - 1) // block_n
            ) // self.tp_size
            shard_size = (
                (self.output_sizes[loaded_shard_id] + block_n - 1)
                // block_n
                // self.tp_size
            )
        else:
            shard_offset = sum(self.output_sizes[:loaded_shard_id]) // self.tp_size
            shard_size = self.output_sizes[loaded_shard_id] // self.tp_size

        param.load_merged_column_weight(
            loaded_weight=loaded_weight,
            shard_id=loaded_shard_id,
            shard_offset=shard_offset,
            shard_size=shard_size,
            tp_rank=self.tp_rank,
        )


class QKVParallelLinear(ColumnParallelLinear):
    """Linear layers for the attention's QKV transformation.

    Linear layers for the linear transformation of the query, key, and value
    vectors in the attention layer. The weight matrix is concatenated along
    the output dimension. The layer is parallelized along the head dimension.
    When the number of key/value heads is smaller than the number of query
    heads (e.g., multi-query/grouped-query attention), the key/value head may
    be replicated while the query heads are partitioned.

    Args:
        hidden_size: input hidden state size of the transformer.
        head_size: size of each attention head.
        total_num_heads: total number of attention query heads.
        total_num_kv_heads: total number of attention key/value heads. If
                            None, assume total_num_kv_heads = total_num_heads.
        bias: If true, add bias.
        skip_bias_add: This was added to enable performance optimizations where
                       bias can be fused with other element-wise operations. we
                       skip adding bias but instead return it.
        params_dtype: Data type for the parameters.
        quant_config: Quantization configure.
        prefix: The name of the layer in the state dict, including all parents
                        (e.g. model.layers.0.qkv_proj)
        return_bias: If true, return bias together with outputs in forward pass.
        disable_tp: If true, weights matrix won't be sharded through tp rank.
    """

    def __init__(
        self,
        hidden_size: int,
        head_size: int,
        total_num_heads: int,
        total_num_kv_heads: Optional[int] = None,
        bias: bool = True,
        skip_bias_add: bool = False,
        params_dtype: Optional[torch.dtype] = None,
        quant_config: Optional[QuantizationConfig] = None,
        prefix: str = "",
        *,
        return_bias: bool = True,
        disable_tp: bool = False,
    ):
        self.hidden_size = hidden_size
        self.head_size = head_size
        self.total_num_heads = total_num_heads
        if total_num_kv_heads is None:
            total_num_kv_heads = total_num_heads
        self.total_num_kv_heads = total_num_kv_heads
        # Divide the weight matrix along the last dimension.
        tp_size = get_tensor_model_parallel_world_size() if not disable_tp else 1
        self.num_heads = divide(self.total_num_heads, tp_size)
        if tp_size >= self.total_num_kv_heads:
            self.num_kv_heads = 1
            self.num_kv_head_replicas = divide(tp_size, self.total_num_kv_heads)
        else:
            self.num_kv_heads = divide(self.total_num_kv_heads, tp_size)
            self.num_kv_head_replicas = 1
        input_size = self.hidden_size
        output_size = (
            (self.num_heads + 2 * self.num_kv_heads) * tp_size * self.head_size
        )
        self.output_sizes = [
            self.num_heads * self.head_size * tp_size,  # q_proj
            self.num_kv_heads * self.head_size * tp_size,  # k_proj
            self.num_kv_heads * self.head_size * tp_size,  # v_proj
        ]

        super().__init__(
            input_size=input_size,
            output_size=output_size,
            bias=bias,
            gather_output=False,
            skip_bias_add=skip_bias_add,
            params_dtype=params_dtype,
            quant_config=quant_config,
            prefix=prefix,
            return_bias=return_bias,
            disable_tp=disable_tp,
        )

    def _get_shard_offset_mapping(self, loaded_shard_id: str):
        shard_offset_mapping = {
            "q": 0,
            "k": self.num_heads * self.head_size,
            "v": (self.num_heads + self.num_kv_heads) * self.head_size,
            "total": (self.num_heads + 2 * self.num_kv_heads) * self.head_size,
        }
        return shard_offset_mapping.get(loaded_shard_id)

    def _get_shard_size_mapping(self, loaded_shard_id: str):
        shard_size_mapping = {
            "q": self.num_heads * self.head_size,
            "k": self.num_kv_heads * self.head_size,
            "v": self.num_kv_heads * self.head_size,
        }
        return shard_size_mapping.get(loaded_shard_id)

    def _load_fused_module_from_checkpoint(
        self, param: BasevLLMParameter, loaded_weight: torch.Tensor
    ):
        """
        Handle special case for models where QKV layers are already
        fused on disk. In this case, we have no shard id. This function
        determines the shard id by splitting these layers and then calls
        the weight loader using the shard id.

        An example of a model with these fused layers:
        https://huggingface.co/microsoft/Phi-3-mini-4k-instruct
        """
        shard_offsets = [
            # (shard_id, shard_offset, shard_size)
            ("q", 0, self.total_num_heads * self.head_size),
            (
                "k",
                self.total_num_heads * self.head_size,
                self.total_num_kv_heads * self.head_size,
            ),
            (
                "v",
                (self.total_num_heads + self.total_num_kv_heads) * self.head_size,
                self.total_num_kv_heads * self.head_size,
            ),
        ]

        for shard_id, shard_offset, shard_size in shard_offsets:
            # Special case for Quantization.
            # If quantized, we need to adjust the offset and size to account
            # for the packing.
            if (
                isinstance(param, (PackedColumnParameter, PackedvLLMParameter))
                and param.packed_dim == param.output_dim
            ):
                shard_size, shard_offset = param.adjust_shard_indexes_for_packing(
                    shard_size=shard_size, shard_offset=shard_offset
                )

            loaded_weight_shard = loaded_weight.narrow(
                param.output_dim, shard_offset, shard_size
            )
            self.weight_loader_v2(param, loaded_weight_shard, shard_id)

    def weight_loader_v2(
        self,
        param: BasevLLMParameter,
        loaded_weight: torch.Tensor,
        loaded_shard_id: Optional[str] = None,
    ):
        if loaded_shard_id is None:  # special case for certain models
            if isinstance(param, PerTensorScaleParameter):
                param.load_qkv_weight(
                    loaded_weight=loaded_weight, shard_id=0, tp_rank=self.tp_rank
                )
                return
            elif type(param) in (RowvLLMParameter, BasevLLMParameter):
                param.load_qkv_weight(loaded_weight=loaded_weight, tp_rank=self.tp_rank)
                return
            # TODO: @dsikka - move to parameter.py
            self._load_fused_module_from_checkpoint(param, loaded_weight)
            return

        assert loaded_shard_id in ["q", "k", "v"]

        shard_offset = self._get_shard_offset_mapping(loaded_shard_id)
        shard_size = self._get_shard_size_mapping(loaded_shard_id)

        # Note(simon): This is needed for Qwen3's fp8 quantization.
        if isinstance(param, BlockQuantScaleParameter):
            assert self.quant_method is not None
            # Assume the weight block size has been set by quant method
            assert hasattr(self, "weight_block_size")
            weight_block_size = self.weight_block_size
            assert weight_block_size is not None
            block_n, _ = weight_block_size[0], weight_block_size[1]
            shard_offset = (shard_offset + block_n - 1) // block_n
            shard_size = (shard_size + block_n - 1) // block_n

        param.load_qkv_weight(
            loaded_weight=loaded_weight,
            num_heads=self.num_kv_head_replicas,
            shard_id=loaded_shard_id,
            shard_offset=shard_offset,
            shard_size=shard_size,
            tp_rank=self.tp_rank,
        )

    def weight_loader(
        self,
        param: Parameter,
        loaded_weight: torch.Tensor,
        loaded_shard_id: Optional[str] = None,
    ):
        # Special case for GGUF
        # initialize GGUF param after we know the quantize type
        is_gguf_weight = getattr(param, "is_gguf_weight", False)
        is_gguf_weight_type = getattr(param, "is_gguf_weight_type", False)
        if is_gguf_weight_type:
            idx_map = {"q": 0, "k": 1, "v": 2}
            if loaded_shard_id is not None:
                param.data[idx_map[loaded_shard_id]].copy_(loaded_weight)
                param.shard_weight_type[loaded_shard_id] = loaded_weight.item()
            else:
                param.shard_weight_type = {k: loaded_weight.item() for k in idx_map}
            return

        if is_gguf_weight:
            output_dim = getattr(param, "output_dim", None)
            shard_size = loaded_weight.size(output_dim) // self.tp_size
            start_idx = self.tp_rank * shard_size

            if loaded_shard_id is not None:
                loaded_weight = loaded_weight.narrow(output_dim, start_idx, shard_size)
                param.shard_id.append(loaded_shard_id)
                param.shard_id_map[loaded_shard_id] = len(param.data_container)
                param.data_container.append(loaded_weight)
                return

        param_data = param.data
        output_dim = getattr(param, "output_dim", None)

        # Special case for per-tensor scales in fused case.
        needs_scalar_to_array = getattr(param, "needs_scalar_to_array", False)

        if loaded_shard_id is None:
            # Loaded weight is already fused on disk (qkv).
            # (e.g., Phi-3's qkv_proj).
            if output_dim is None:
                if needs_scalar_to_array:
                    param_data, loaded_weight = adjust_scalar_to_fused_array(
                        param_data, loaded_weight, 0
                    )

                assert param_data.shape == loaded_weight.shape
                param_data.copy_(loaded_weight)
                return
            shard_offsets = [
                # (shard_id, shard_offset, shard_size)
                ("q", 0, self.total_num_heads * self.head_size),
                (
                    "k",
                    self.total_num_heads * self.head_size,
                    self.total_num_kv_heads * self.head_size,
                ),
                (
                    "v",
                    (self.total_num_heads + self.total_num_kv_heads) * self.head_size,
                    self.total_num_kv_heads * self.head_size,
                ),
            ]
            use_bitsandbytes_4bit = getattr(param, "use_bitsandbytes_4bit", False)

            packed_dim = getattr(param, "packed_dim", None)
            for shard_id, shard_offset, shard_size in shard_offsets:
                # Special case for Quantized Weights.
                # If quantized, we need to adjust the offset and size to account
                # for the packing.
                if packed_dim == output_dim:
                    shard_size = shard_size // param.packed_factor
                    shard_offset = shard_offset // param.packed_factor

                    # Special case for Marlin.
                    shard_size, shard_offset = adjust_marlin_shard(
                        param, shard_size, shard_offset
                    )

                if use_bitsandbytes_4bit:
                    orig_qkv_offsets = {
                        "q": (0, self.total_num_heads * self.head_size),
                        "k": (
                            self.total_num_heads * self.head_size,
                            self.total_num_kv_heads * self.head_size,
                        ),
                        "v": (
                            (self.total_num_heads + self.total_num_kv_heads)
                            * self.head_size,
                            self.total_num_kv_heads * self.head_size,
                        ),
                        "total": (
                            (self.total_num_heads + 2 * self.total_num_kv_heads)
                            * self.head_size,
                            0,
                        ),
                    }

                    shard_size, shard_offset = adjust_bitsandbytes_4bit_shard(
                        param, orig_qkv_offsets, shard_id
                    )

                loaded_weight_shard = loaded_weight.narrow(
                    output_dim, shard_offset, shard_size
                )
                self.weight_loader(param, loaded_weight_shard, shard_id)
            return

        assert loaded_shard_id in ["q", "k", "v"]

        # If output dim is defined, use the default loading process.
        if output_dim is not None:
            if loaded_shard_id == "q":
                shard_offset = 0
                shard_size = self.num_heads * self.head_size
            elif loaded_shard_id == "k":
                shard_offset = self.num_heads * self.head_size
                shard_size = self.num_kv_heads * self.head_size
            elif loaded_shard_id == "v":
                shard_offset = (self.num_heads + self.num_kv_heads) * self.head_size
                shard_size = self.num_kv_heads * self.head_size
            # Special case for Quantized Weights.
            # If quantized, we need to adjust the offset and size to account
            # for the packing.
            packed_dim = getattr(param, "packed_dim", None)
            if packed_dim == output_dim:
                shard_size = shard_size // param.packed_factor
                shard_offset = shard_offset // param.packed_factor

                # Special case for Marlin.
                shard_size, shard_offset = adjust_marlin_shard(
                    param, shard_size, shard_offset
                )

            use_bitsandbytes_4bit = getattr(param, "use_bitsandbytes_4bit", False)
            is_sharded_weight = getattr(param, "is_sharded_weight", False)
            # bitsandbytes loads the weights of the specific portion
            # no need to narrow
            is_sharded_weight = is_sharded_weight or use_bitsandbytes_4bit

            if use_bitsandbytes_4bit:
                orig_qkv_offsets = {
                    "q": (0, self.num_heads * self.head_size),
                    "k": (
                        self.num_heads * self.head_size,
                        self.num_kv_heads * self.head_size,
                    ),
                    "v": (
                        (self.num_heads + self.num_kv_heads) * self.head_size,
                        self.num_kv_heads * self.head_size,
                    ),
                    "total": (
                        (self.num_heads + 2 * self.num_kv_heads) * self.head_size,
                        0,
                    ),
                }
                shard_size, shard_offset = adjust_bitsandbytes_4bit_shard(
                    param, orig_qkv_offsets, loaded_shard_id
                )

            param_data = param_data.narrow(output_dim, shard_offset, shard_size)
            if loaded_shard_id == "q":
                shard_id = self.tp_rank
            else:
                shard_id = self.tp_rank // self.num_kv_head_replicas
            start_idx = shard_id * shard_size

            if not is_sharded_weight:
                loaded_weight = loaded_weight.narrow(output_dim, start_idx, shard_size)

        # Special case for per-tensor scales in fused case.
        elif needs_scalar_to_array:
            param_data, loaded_weight = adjust_scalar_to_fused_array(
                param_data, loaded_weight, loaded_shard_id
            )
        else:
            ignore_warning = getattr(param, "ignore_warning", False)
            if not ignore_warning:
                logger.warning(
                    "Loading a weight without `output_dim` attribute in "
                    "QKVParallelLinear, assume the weight is the same "
                    "for all partitions."
                )

        assert param_data.shape == loaded_weight.shape
        param_data.copy_(loaded_weight)


@CustomOp.register("row_parallel_linear")
class RowParallelLinear(LinearBase):
    """Linear layer with row parallelism.

    The linear layer is defined as Y = XA + b. A is parallelized along
    its first dimension and X along its second dimension as:
               -   -
              | A_1 |
              | .   |
          A = | .   |        X = [X_1, ..., X_p]
              | .   |
              | A_p |
               -   -
    Arguments:
        input_size: first dimension of matrix A.
        output_size: second dimension of matrix A.
        bias: If true, add bias. Note that bias is not parallelized.
        input_is_parallel: If true, we assume that the input is already
                           split across the GPUs and we do not split
                           again.
        skip_bias_add: This was added to enable performance optimization where
                       bias can be fused with other element-wise operations.
                       We skip adding bias but instead return it.
        params_dtype: Data type for the parameters.
        reduce_results: If true, call all-reduce on output and make Y available
                       to all GPUs, otherwise, every GPU will have its output
                       which is Y = X_iA_i
        quant_config: Quantization configure.
        prefix: The name of the layer in the state dict, including all parents
                        (e.g. model.layers.0.down_proj)
        return_bias: If true, return bias together with outputs in forward pass.
        disable_tp: If true, weights matrix won't be sharded through tp rank.
    """

    def __init__(
        self,
        input_size: int,
        output_size: int,
        bias: bool = True,
        input_is_parallel: bool = True,
        skip_bias_add: bool = False,
        params_dtype: Optional[torch.dtype] = None,
        reduce_results: bool = True,
        quant_config: Optional[QuantizationConfig] = None,
        prefix: str = "",
        *,
        return_bias: bool = True,
        disable_tp: bool = False,
    ):
        # Divide the weight matrix along the first dimension.
        self.tp_rank = get_tensor_model_parallel_rank() if not disable_tp else 0
        self.tp_size = get_tensor_model_parallel_world_size() if not disable_tp else 1
        self.input_size_per_partition = divide(input_size, self.tp_size)
        self.output_size_per_partition = output_size
        self.output_partition_sizes = [output_size]

        super().__init__(
            input_size,
            output_size,
            skip_bias_add,
            params_dtype,
            quant_config,
            prefix,
            return_bias=return_bias,
            disable_tp=disable_tp,
        )

        self.input_is_parallel = input_is_parallel
        self.reduce_results = reduce_results

        assert self.quant_method is not None
        self.quant_method.create_weights(
            layer=self,
            input_size_per_partition=self.input_size_per_partition,
            output_partition_sizes=self.output_partition_sizes,
            input_size=self.input_size,
            output_size=self.output_size,
            params_dtype=self.params_dtype,
            weight_loader=(
                self.weight_loader_v2
                if self.quant_method.__class__.__name__ in WEIGHT_LOADER_V2_SUPPORTED
                else self.weight_loader
            ),
        )
        if not reduce_results and (bias and not skip_bias_add):
            raise ValueError(
                "When not reduce the results, adding bias to the "
                "results can lead to incorrect results"
            )

        if bias:
            self.bias = Parameter(torch.empty(self.output_size, dtype=params_dtype))
            set_weight_attrs(
                self.bias,
                {
                    "output_dim": 0,
                    "weight_loader": self.weight_loader,
                },
            )
        else:
            self.register_parameter("bias", None)
        self.update_param_tp_status()

    def weight_loader(self, param: Parameter, loaded_weight: torch.Tensor):
        input_dim = getattr(param, "input_dim", None)
        use_bitsandbytes_4bit = getattr(param, "use_bitsandbytes_4bit", False)
        is_sharded_weight = getattr(param, "is_sharded_weight", False)
        # bitsandbytes loads the weights of the specific portion
        # no need to narrow
        is_sharded_weight = is_sharded_weight or use_bitsandbytes_4bit

        # Special case for GGUF
        is_gguf_weight = getattr(param, "is_gguf_weight", False)
        is_gguf_weight_type = getattr(param, "is_gguf_weight_type", False)
        if is_gguf_weight_type:
            param.weight_type = loaded_weight.item()

        # Materialize GGUF UninitializedParameter
        if is_gguf_weight and isinstance(param, UninitializedParameter):
            weight_shape = list(loaded_weight.shape)
            if input_dim:
                weight_shape[input_dim] = weight_shape[input_dim] // self.tp_size
            param.materialize(tuple(weight_shape), dtype=loaded_weight.dtype)

        param_data = param.data
        if input_dim is not None and not is_sharded_weight:
            shard_size = param_data.shape[input_dim]
            start_idx = self.tp_rank * shard_size
            loaded_weight = loaded_weight.narrow(input_dim, start_idx, shard_size)

        # Special case for loading scales off disk, which often do not
        # have a shape (such as in the case of AutoFP8).
        if len(loaded_weight.shape) == 0:
            loaded_weight = loaded_weight.reshape(1)

        assert param_data.shape == loaded_weight.shape
        param_data.copy_(loaded_weight)

    def weight_loader_v2(self, param: BasevLLMParameter, loaded_weight: torch.Tensor):
        # Special case for loading scales off disk, which often do not
        # have a shape (such as in the case of AutoFP8).
        if len(loaded_weight.shape) == 0:
            assert loaded_weight.numel() == 1
            loaded_weight = loaded_weight.reshape(1)

        param.load_row_parallel_weight(loaded_weight=loaded_weight)

    def forward(
        self,
        input_,
    ) -> Union[torch.Tensor, tuple[torch.Tensor, Optional[Parameter]]]:
        if self.input_is_parallel:
            input_parallel = input_
        else:
            splitted_input = split_tensor_along_last_dim(
                input_, num_partitions=self.tp_size
            )
            input_parallel = splitted_input[self.tp_rank].contiguous()

        # Matrix multiply.
        assert self.quant_method is not None
        # Only fuse bias add into GEMM for rank 0 (this ensures that
        # bias will not get added more than once in TP>1 case)
        bias_ = None if (self.tp_rank > 0 or self.skip_bias_add) else self.bias
        output_parallel = self.quant_method.apply(self, input_parallel, bias_)

        if self.reduce_results and self.tp_size > 1:
            output = tensor_model_parallel_all_reduce(output_parallel)
        else:
            output = output_parallel

        output_bias = self.bias if self.skip_bias_add else None

        if not self.return_bias:
            return output
        return output, output_bias

    def extra_repr(self) -> str:
        s = f"in_features={self.input_size_per_partition}"
        s += f", output_features={self.output_size}"
        s += f", bias={self.bias is not None}"
        s += f", tp_size={self.tp_size}"
        s += f", reduce_results={self.reduce_results}"
        return s


@CustomOp.register("qkv_cross_parallel_linear")
class QKVCrossParallelLinear(LinearBase):
    """Linear layers for efficient cross-attention's QKV transformation.

    Args:
        hidden_size: input hidden state size of the transformer.
        head_size: size of each attention head.
        total_num_heads: total number of attention query heads.
        total_num_kv_heads: total number of attention key/value heads. If
                            None, assume total_num_kv_heads = total_num_heads.
        bias: If true, add bias.
        skip_bias_add: This was added to enable performance optimizations where
                       bias can be fused with other element-wise operations. we
                       skip adding bias but instead return it.
        params_dtype: Data type for the parameters.
        quant_config: Quantization configure.
        prefix: The name of the layer in the state dict, including all parents
                        (e.g. model.layers.0.qkv_proj)
    """

    def __init__(
        self,
        hidden_size: int,
        head_size: int,
        total_num_heads: int,
        total_num_kv_heads: Optional[int] = None,
        bias: bool = True,
        skip_bias_add: bool = False,
        params_dtype: Optional[torch.dtype] = None,
        quant_config: Optional[QuantizationConfig] = None,
        prefix: str = "",
    ):
        # input_size and output_size are not used, just for alignment
        input_size = hidden_size
        output_size = (total_num_heads + (total_num_kv_heads or 0)) * head_size
        super().__init__(
            input_size=input_size,
            output_size=output_size,
            skip_bias_add=skip_bias_add,
            params_dtype=params_dtype,
            quant_config=quant_config,
            prefix=prefix,
        )

        self.quant_config = quant_config

        # Empty placeholders for loading as a single module.
        placeholder_size = 0
        assert self.quant_method is not None
        self.quant_method.create_weights(
            self,
            placeholder_size,
            [placeholder_size],
            placeholder_size,
            placeholder_size,
            self.params_dtype,
            weight_loader=self.weight_loader,
        )

        # Use a dictionary to avoid submodules parameters auto-registration:
        # drop-in replacement for a `QKVParallelLinear` module.
        self.proj = dict()
        self.proj["q_proj_decoder"] = ColumnParallelLinear(
            input_size=hidden_size,
            output_size=total_num_heads * head_size,
            bias=bias,
            quant_config=quant_config,
            skip_bias_add=skip_bias_add,
            params_dtype=params_dtype,
            prefix=f"{prefix}.q_proj_decoder",
        )

        self.proj["kv_proj_encoder"] = QKVParallelLinear(
            hidden_size=hidden_size,
            head_size=head_size,
            total_num_heads=0,
            total_num_kv_heads=total_num_kv_heads,
            bias=bias,
            quant_config=quant_config,
            skip_bias_add=skip_bias_add,
            params_dtype=params_dtype,
            prefix=f"{prefix}.kv_proj_encoder",
        )

        # `kv_proj_encoder.num_kv_heads` accounts for sharding with tp>1.
        self.q_size = self.q_proj_decoder.output_size_per_partition
        self.kv_size = self.kv_proj_encoder.num_kv_heads * head_size

        if bias:
            self.bias = torch.nn.Parameter()
            set_weight_attrs(
                self.bias,
                {
                    "output_dim": 0,
                    "weight_loader": self.weight_loader_v1,
                },
            )
        else:
            self.bias = None

    def process_weights_after_loading(self):
        for layer in self.proj.values():
            if self.quant_method is not None:
                self.quant_method.process_weights_after_loading(layer)

    @property
    def q_proj_decoder(self) -> ColumnParallelLinear:
        layer = self.proj["q_proj_decoder"]
        for name, param in self.named_parameters():
            target_param = getattr(layer, name, None)
            if target_param is not None:
                self.sync_weight_attrs(param, target_param, mode="q_proj_decoder")
        return layer

    @property
    def kv_proj_encoder(self) -> QKVParallelLinear:
        layer = self.proj["kv_proj_encoder"]
        for name, param in self.named_parameters():
            target_param = getattr(layer, name, None)
            if target_param is not None:
                self.sync_weight_attrs(param, target_param, mode="kv_proj_encoder")
        return layer

    def sync_weight_attrs(
        self,
        src_param: nn.Parameter,
        tgt_param: nn.Parameter,
        mode: Literal["q_proj_decoder", "kv_proj_encoder"],
    ):
        missing_attrs_dict = {
            k: getattr(src_param, k)
            for k in (set(vars(src_param).keys()) - set(vars(tgt_param).keys()))
        }
        # TODO(Isotr0py): handle bitsandbytes 8bit
        use_bitsandbytes_4bit = getattr(src_param, "use_bitsandbytes_4bit", False)
        if missing_attrs_dict and use_bitsandbytes_4bit:
            q_proj_attrs, kv_proj_attrs = left_shift_bitsandbytes_4bit_shard(
                missing_attrs_dict
            )
            if mode == "q_proj_decoder":
                set_weight_attrs(tgt_param, q_proj_attrs)
            elif mode == "kv_proj_encoder":
                set_weight_attrs(tgt_param, kv_proj_attrs)
        else:
            set_weight_attrs(tgt_param, missing_attrs_dict)

    def _is_same_param(
        self,
        src_param: torch.nn.Parameter,
        map_param: torch.nn.Parameter,
    ) -> bool:
        """Check if two parameters are exactly pointing to same things."""
        # ignore weight_loader because it's always different
        key_to_ignore = ["weight_loader", "_weight_loader"]
        has_same_type_name = type(src_param) is type(map_param)
        src_param_attrs = {
            k: v for k, v in src_param.__dict__.items() if k not in key_to_ignore
        }
        map_param_attrs = {
            k: v for k, v in map_param.__dict__.items() if k not in key_to_ignore
        }
        has_same_attrs = src_param_attrs == map_param_attrs
        return has_same_type_name and has_same_attrs

    def select_proj_params(
        self,
        layer: nn.Module,
        param: nn.Parameter,
    ) -> nn.Parameter:
        """
        Given the placeholder param,
        return the corresponding param in the proj layers.
        """
        target_param_list = [
            v for _, v in layer.named_parameters() if self._is_same_param(param, v)
        ]
        assert len(target_param_list) == 1
        target_param = target_param_list[0]
        return target_param

    def forward(  # type: ignore[override]
        self,
        decoder_hidden_states: torch.Tensor,
        encoder_hidden_states: torch.Tensor,
    ) -> tuple[torch.Tensor, ...]:
        q, _ = self.q_proj_decoder(decoder_hidden_states)
        if encoder_hidden_states is None:
            # Encoder KV already cached.
            k = None
            v = None
        else:
            # Prefill phase, encoder KV cached here.
            kv_enc, _ = self.kv_proj_encoder(encoder_hidden_states)
            # Split kv in half
            k, v = kv_enc.split(self.kv_size, dim=-1)
        return q, k, v

    def weight_loader_v1(
        self,
        param: torch.nn.Parameter,
        loaded_weight: torch.Tensor,
        loaded_shard_id: Optional[str] = None,
    ):
        # just like all other parameters, does not yet
        # support loading bias with weight_loader_v2
        layer = self.q_proj_decoder if loaded_shard_id == "q" else self.kv_proj_encoder
        target_param = self.select_proj_params(layer, param)
        shard_id_args = (loaded_shard_id,) if loaded_shard_id != "q" else ()
        layer.weight_loader(target_param, loaded_weight, *shard_id_args)

    def weight_loader(
        self,
        param: torch.nn.Parameter,
        loaded_weight: torch.Tensor,
        loaded_shard_id: Optional[str] = None,
    ):
        layer = self.q_proj_decoder if loaded_shard_id == "q" else self.kv_proj_encoder
        target_param = self.select_proj_params(layer, param)
        shard_id_args = (loaded_shard_id,) if loaded_shard_id != "q" else ()
        if self.quant_method.__class__.__name__ in WEIGHT_LOADER_V2_SUPPORTED:
            layer.weight_loader_v2(target_param, loaded_weight, *shard_id_args)
        else:
            layer.weight_loader(target_param, loaded_weight, *shard_id_args)

    def extra_repr(self) -> str:
        s = f"in_features={self.input_size}"
        s += f", q_size={self.q_size}"
        s += f", kv_size={self.kv_size}"
        s += f", bias={self.bias is not None}"
        s += f", tp_size={get_tensor_model_parallel_world_size()}"
        s += ", gather_output=False"
        return s