topk_topp_sampler.py

# SPDX-License-Identifier: Apache-2.0

from typing import Optional

import torch
import torch.nn as nn

from vllm import envs
from vllm.logger import init_logger
from vllm.platforms import current_platform

logger = init_logger(__name__)

try:
    import flashinfer.sampling
    is_flashinfer_available = True
except ImportError:
    is_flashinfer_available = False


class TopKTopPSampler(nn.Module):
    """
    Module that performs optional top-k and top-p filtering followed by
    weighted random sampling of logits.

    Implementations may update the logits tensor in-place.
    """

    def __init__(self):
        super().__init__()
        if current_platform.is_cuda():
            if is_flashinfer_available:
                flashinfer_version = flashinfer.__version__
                if flashinfer_version >= "0.2.3":
                    # FIXME(DefTruth): Currently, we have errors when using
                    # FlashInfer>=v0.2.3 for top-p & top-k sampling. As a
                    # workaround, we disable FlashInfer for top-p & top-k
                    # sampling by default while FlashInfer>=v0.2.3.
                    # The sampling API removes the success return value
                    # of all sampling API, which is not compatible with
                    # earlier design.
                    # https://github.com/flashinfer-ai/flashinfer/releases/
                    # tag/v0.2.3
                    logger.info(
                        "Currently, FlashInfer top-p & top-k sampling sampler "
                        "is disabled because FlashInfer>=v0.2.3 is not "
                        "backward compatible. Falling back to the PyTorch-"
                        "native implementation of top-p & top-k sampling.")
                    self.forward = self.forward_native
                elif envs.VLLM_USE_FLASHINFER_SAMPLER is not False:
                    # NOTE(woosuk): The V0 sampler doesn't use FlashInfer for
                    # sampling unless VLLM_USE_FLASHINFER_SAMPLER=1 (i.e., by
                    # default it is unused). For backward compatibility, we set
                    # `VLLM_USE_FLASHINFER_SAMPLER` as None by default and
                    # interpret it differently in V0 and V1 samplers: In V0,
                    # None means False, while in V1, None means True. This is
                    # why we use the condition
                    # `envs.VLLM_USE_FLASHINFER_SAMPLER is not False` here.
                    logger.info("Using FlashInfer for top-p & top-k sampling.")
                    self.forward = self.forward_cuda
                else:
                    logger.warning(
                        "FlashInfer is available, but it is not enabled. "
                        "Falling back to the PyTorch-native implementation of "
                        "top-p & top-k sampling. For the best performance, "
                        "please set VLLM_USE_FLASHINFER_SAMPLER=1.")
                    self.forward = self.forward_native
            else:
                logger.warning(
                    "FlashInfer is not available. Falling back to the PyTorch-"
                    "native implementation of top-p & top-k sampling. For the "
                    "best performance, please install FlashInfer.")
                self.forward = self.forward_native
        elif current_platform.is_tpu():
            if envs.VLLM_TPU_DISABLE_TOPK_TOPP_OPTIMIZATION:
                logger.warning(
                    "TPU-specific optimization for top-k & top-p sampling are "
                    "disabled, falling back to PyTorch-native implementation "
                    "which could be very slow.")
                self.forward = self.forward_native
            else:
                self.forward = self.forward_tpu
        else:
            self.forward = self.forward_native

    def forward_native(
        self,
        logits: torch.Tensor,
        generators: dict[int, torch.Generator],
        k: Optional[torch.Tensor],
        p: Optional[torch.Tensor],
    ) -> torch.Tensor:
        """
        PyTorch-native implementation of top-k and top-p sampling.

        The logits tensor may be updated in-place.
        """
        logits = apply_top_k_top_p(logits, k, p)
        probs = logits.softmax(dim=-1, dtype=torch.float32)
        return random_sample(probs, generators)

    def forward_cuda(
        self,
        logits: torch.Tensor,
        generators: dict[int, torch.Generator],
        k: Optional[torch.Tensor],
        p: Optional[torch.Tensor],
    ) -> torch.Tensor:
        """More optimized implementation for top-k and top-p sampling."""
        probs = logits.softmax(dim=-1, dtype=torch.float32)
        if k is None and p is None:
            # We prefer `random_sample` over `flashinfer_sample` when sorting is
            # not needed. This is because `random_sample` does not require
            # CPU-GPU synchronization while `flashinfer_sample` does.
            return random_sample(probs, generators)
        return flashinfer_sample(probs, k, p, generators)

    def forward_tpu(
        self,
        logits: torch.Tensor,
        generators: dict[int, torch.Generator],
        k: Optional[torch.Tensor],
        p: Optional[torch.Tensor],
    ) -> torch.Tensor:
        logits = apply_top_k_top_p_tpu(logits, k, p)
        probs = logits.softmax(dim=-1, dtype=torch.float32)
        return random_sample(probs, generators)


def apply_top_k_top_p_tpu(
    logits: torch.Tensor,
    k: torch.Tensor,
    p: torch.Tensor,
) -> torch.Tensor:
    """
    Apply top-k and top-p optimized for TPU.

    This algorithm avoids using torch.scatter which is extremely slow on TPU.
    This is achieved by finding a "cut-off" element in the original logit, and
    after thresholding the logit using this cut-off, the remaining elements
    shall constitute the top-p set.

    Note: in the case of tie (i.e. multipple cut-off elements present in the
    logit), all tie elements are included in the top-p set. In other words,
    this function does not break ties. Instead, these tie tokens have equal
    chance of being chosen during final sampling, so we can consider the tie
    being broken then.
    """
    if k is not None:
        logits = apply_top_k_only(logits, k)

    if p is not None:
        probs = logits.softmax(dim=-1)
        probs_sort, _ = probs.sort(dim=-1, descending=False)
        cumprob = torch.cumsum(probs_sort, dim=-1)
        top_p_mask = cumprob <= 1 - p.unsqueeze(dim=1)
        top_p_mask[:, -1] = False  # at least one

        top_p_count = top_p_mask.sum(dim=-1).unsqueeze(1)
        top_p_cutoff = probs_sort.gather(-1, top_p_count)
        elements_to_discard = probs < top_p_cutoff
        logits.masked_fill_(elements_to_discard, -float("inf"))

    return logits


def apply_top_k_top_p(
    logits: torch.Tensor,
    k: Optional[torch.Tensor],
    p: Optional[torch.Tensor],
) -> torch.Tensor:
    """Apply top-k and top-p masks to the logits.

    If a top-p is used, this function will sort the logits tensor,
    which can be slow for large batches.

    The logits tensor may be updated in-place.
    """
    if p is None:
        if k is None:
            return logits

        # Avoid sorting vocab for top-k only case.
        return apply_top_k_only(logits, k)

    logits_sort, logits_idx = logits.sort(dim=-1, descending=False)

    if k is not None:
        # Apply top-k.
        top_k_mask = logits_sort.size(1) - k.to(torch.long)  # shape: B
        # Get all the top_k values.
        top_k_mask = logits_sort.gather(1, top_k_mask.unsqueeze(dim=1))
        top_k_mask = logits_sort < top_k_mask
        logits_sort.masked_fill_(top_k_mask, -float("inf"))

    if p is not None:
        # Apply top-p.
        probs_sort = logits_sort.softmax(dim=-1)
        probs_sum = torch.cumsum(probs_sort, dim=-1, out=probs_sort)
        top_p_mask = probs_sum <= 1 - p.unsqueeze(dim=1)
        # at least one
        top_p_mask[:, -1] = False
        logits_sort.masked_fill_(top_p_mask, -float("inf"))

    # Re-sort the probabilities.
    logits = logits_sort.scatter(dim=-1, index=logits_idx, src=logits_sort)
    return logits


def apply_top_k_only(
    logits: torch.Tensor,
    k: torch.Tensor,
) -> torch.Tensor:
    """
    Apply top-k mask to the logits.

    This implementation doesn't involve sorting the entire vocab.

    The logits tensor may be updated in-place.
    """
    no_top_k_mask = k == logits.shape[1]
    # Set non-top-k rows to 1 so that we can gather.
    k = k.masked_fill(no_top_k_mask, 1)
    max_top_k = k.max()
    # topk.values tensor has shape [batch_size, max_top_k].
    # Convert top k to 0-based index in range [0, max_top_k).
    k_index = k.sub_(1).unsqueeze(1).expand(logits.shape[0], 1)
    top_k_mask = logits.topk(max_top_k, dim=1).values.gather(1, k_index.long())
    # Handle non-topk rows.
    top_k_mask.masked_fill_(no_top_k_mask.unsqueeze(1), -float("inf"))
    logits.masked_fill_(logits < top_k_mask, -float("inf"))
    return logits


def random_sample(
    probs: torch.Tensor,
    generators: dict[int, torch.Generator],
) -> torch.Tensor:
    """Randomly sample from the probabilities.

    We use this function instead of torch.multinomial because torch.multinomial
    causes CPU-GPU synchronization.
    """
    q = torch.empty_like(probs)
    # NOTE(woosuk): To batch-process the requests without their own seeds,
    # which is the common case, we first assume that every request does
    # not have its own seed. Then, we overwrite the values for the requests
    # that have their own seeds.
    if len(generators) != probs.shape[0]:
        q.exponential_()
    if generators:
        # TODO(woosuk): This can be slow because we handle each request
        # one by one. Optimize this.
        for i, generator in generators.items():
            q[i].exponential_(generator=generator)
    return probs.div_(q).argmax(dim=-1).view(-1)


def flashinfer_sample(
    probs: torch.Tensor,
    k: Optional[torch.Tensor],
    p: Optional[torch.Tensor],
    generators: dict[int, torch.Generator],
) -> torch.Tensor:
    """Sample from the probabilities using FlashInfer.

    Statistically, this function is equivalent to the `random_sample` function.
    However, this function is faster because it avoids sorting the logits tensor
    via rejection sampling.
    
    NOTE: The outputs of this function do not necessarily match the outputs of
    the `random_sample` function. It only guarantees that the outputs are
    statistically equivalent.

    NOTE: This function includes CPU-GPU synchronization, while `random_sample`
    does not. Call this function at the end of the forward pass to minimize
    the synchronization overhead.
    """
    assert not (k is None and p is None)
    max_top_k_round = 32
    batch_size = probs.shape[0]
    uniform_samples = torch.empty((max_top_k_round, batch_size),
                                  device=probs.device)
    if len(generators) != batch_size:
        uniform_samples.uniform_()
    if generators:
        for i, generator in generators.items():
            uniform_samples[:, i].uniform_(generator=generator)

    if k is None:
        # Top-p only.
        next_token_ids, success = flashinfer.sampling.top_p_sampling_from_probs(
            probs, uniform_samples, p, deterministic=True)
    elif p is None:
        # Top-k only.
        next_token_ids, success = flashinfer.sampling.top_k_sampling_from_probs(
            probs, uniform_samples, k, deterministic=True)
    else:
        # Both top-k and top-p.
        next_token_ids, success = (
            flashinfer.sampling.top_k_top_p_sampling_from_probs(
                probs, uniform_samples, k, p, deterministic=True))

    # NOTE: CPU-GPU synchronization happens here.
    if not success.all():
        if k is not None:
            probs = flashinfer.sampling.top_k_renorm_prob(probs, k)
        if p is not None:
            probs = flashinfer.sampling.top_p_renorm_prob(probs, p)
        next_token_ids = flashinfer.sampling.sampling_from_probs(
            probs, uniform_samples[0], deterministic=True)
    return next_token_ids.view(-1)