gpu_model_runner.py

from typing import Any, Optional, Union
import torch
import numpy as np
from vllm import envs
from vllm.distributed.kv_transfer.kv_transfer_state import get_kv_transfer_group, has_kv_transfer_group
from vllm.distributed.parallel_state import get_pp_group, get_tp_group
from vllm.forward_context import set_forward_context
from vllm.sequence import IntermediateTensors
from vllm.utils import async_tensor_h2d, round_up
from vllm.v1.attention.backends.utils import CommonAttentionMetadata
from vllm.v1.core.sched.output import SchedulerOutput
from vllm.v1.outputs import EMPTY_MODEL_RUNNER_OUTPUT, ModelRunnerOutput
from vllm.v1.sample.metadata import SamplingMetadata
from vllm.v1.spec_decode.eagle import EagleProposer
from vllm.v1.spec_decode.medusa import MedusaProposer
from vllm.v1.spec_decode.metadata import SpecDecodeMetadata
from vllm.v1.spec_decode.ngram_proposer import NgramProposer
from vllm.v1.worker.block_table import BlockTable
from vllm.v1.worker.gpu_model_runner import GPUModelRunner
from vllm.zero_overhead.v1.eagle import V1ZeroEagleProposer
from vllm.zero_overhead.v1.outputs import ZeroV1ModelRunnerOutput
from vllm.profiler.prof import profile
from vllm.two_batch_overlap.v1.model_input_split_v1 import tbo_split_and_execute_model
from vllm.v1.spec_decode.utils import DraftProbs

import triton
import triton.language as tl

@triton.jit
def fused_last_valid_scatter_kernel(
    last_ids_ptr,        # [B, T]
    input_ids_ptr,       # [N]
    update_req_ptr,      # [U]
    input_pos_ptr,       # [U]
    stride0,
    stride1,
    T,
    BLOCK_T: tl.constexpr,
):
    pid = tl.program_id(0)

    # indices
    req_idx = tl.load(update_req_ptr + pid)
    input_pos = tl.load(input_pos_ptr + pid)

    # load row
    offs = tl.arange(0, BLOCK_T)
    mask = offs < T

    row_ptr = last_ids_ptr + req_idx * stride0 + offs * stride1
    vals = tl.load(row_ptr, mask=mask, other=-1)

    # ✅ 正确做法：index reduction
    idx = tl.where(vals != -1, offs, -1)
    last_idx = tl.max(idx, axis=0)

    # load last token
    last_val = tl.load(
        last_ids_ptr + req_idx * stride0 + last_idx * stride1,
        mask=last_idx >= 0,
        other=0,
    )

    # scatter
    tl.store(input_ids_ptr + input_pos, last_val)

class V1ZeroModelRunner(GPUModelRunner):
    def __init__(self, vllm_config, device):
        super().__init__(vllm_config, device)
        self.last_sampled_token_ids = None
        self.last_sampled_req_ids = []
        self.last_sampled_token_lens = []
        self.last_sampler_event = torch.cuda.Event(enable_timing=False)
        self.last_sampler_host_tokens = None
        self.token_ids_cpu_fix_record = []
        self.last_draft_token_ids = None
        self.last_draft_host_tokens = None
        self.last_draft_event = torch.cuda.Event(enable_timing=False)
        self.spec_sampler_event = torch.cuda.Event(enable_timing=False)
        self.spec_scheduler_max_num_tokens = 0
        self.fix_req_ids = None
        self.fix_sampled_token_ids = None

        if hasattr(self, 'drafter') and isinstance(self.drafter, EagleProposer):
            self.drafter = V1ZeroEagleProposer(self.vllm_config, self.device,
                                                self) 
    
    def _prepare_inputs(
        self,
        scheduler_output: "SchedulerOutput",
    ) -> tuple[dict[str, Any], bool, torch.Tensor,
               Optional[SpecDecodeMetadata], np.ndarray]:
        """
        :return: tuple[
            attn_metadata: layer-to-attention_metadata mapping,
            attention_cuda_graphs: whether attention can run in cudagraph
            logits_indices, spec_decode_metadata
        ]
        """
        total_num_scheduled_tokens = scheduler_output.total_num_scheduled_tokens
        assert total_num_scheduled_tokens > 0
        num_reqs = self.input_batch.num_reqs
        assert num_reqs > 0

        # OPTIMIZATION: Start copying the block table first.
        # This way, we can overlap the copy with the following CPU operations.
        self.input_batch.block_table.commit(num_reqs)

        # Get the number of scheduled tokens for each request.
        req_ids = self.input_batch.req_ids
        tokens = [scheduler_output.num_scheduled_tokens[i] for i in req_ids]
        num_scheduled_tokens = np.array(tokens, dtype=np.int32)
        max_num_scheduled_tokens = max(tokens)
        self.spec_scheduler_max_num_tokens = max_num_scheduled_tokens

        # Get request indices.
        # E.g., [2, 5, 3] -> [0, 0, 1, 1, 1, 1, 1, 2, 2, 2]
        req_indices = np.repeat(self.arange_np[:num_reqs],
                                num_scheduled_tokens)

        # cu_num_tokens: [2, 5, 3] -> [2, 7, 10]
        # arange: [0, 1, 0, 1, 2, 3, 4, 0, 1, 2]
        cu_num_tokens, arange = self._get_cumsum_and_arange(
            num_scheduled_tokens)

        self.fix_req_ids = self.last_sampled_req_ids
        if self.last_sampler_host_tokens != None:
            self.last_sampler_event.synchronize()  # 等上一轮主模型结束
            if self.speculative_config:  # 处理上一轮mtp
                num_gen_tokens = self.last_sampler_host_tokens.shape[-1]
                if num_gen_tokens == 1:
                    self.fix_sampled_token_ids = self.last_sampler_host_tokens.tolist()
                else:
                    # Includes spec decode tokens.
                    self.fix_sampled_token_ids = self.rejection_sampler.parse_output(
                        self.last_sampler_host_tokens,
                        self.input_batch.vocab_size,
                    )

            for req_idx, start_idx, end_idx in self.token_ids_cpu_fix_record:
                if start_idx == -1:
                    self.fix_sampled_token_ids[req_idx].clear()
                else:
                    num_accepted_tokens = len(self.fix_sampled_token_ids[req_idx])
                    req_id = self.fix_req_ids[req_idx]
                    if req_id in self.input_batch.req_ids:
                        new_req_idx = self.input_batch.req_ids.index(req_id)
                        new_end_idx = start_idx + num_accepted_tokens
                        # # 更新token统计数据
                        self.input_batch.num_tokens_no_spec[new_req_idx] = new_end_idx
                        self.input_batch.num_tokens[new_req_idx] = new_end_idx
                        self.input_batch.token_ids_cpu[new_req_idx, start_idx:new_end_idx] = self.fix_sampled_token_ids[
                            req_idx]
                        self.input_batch.num_computed_tokens_cpu[new_req_idx] -= (end_idx - new_end_idx)
                    if req_id in self.requests:
                        req_state = self.requests[req_id]
                        req_state.output_token_ids.extend(self.fix_sampled_token_ids[req_idx])

        # Get positions.
        positions_np = self.positions_np[:total_num_scheduled_tokens]
        np.add(self.input_batch.num_computed_tokens_cpu[req_indices],
               arange,
               out=positions_np)

        # Calculate M-RoPE positions.
        # Only relevant for models using M-RoPE (e.g, Qwen2-VL)
        if self.uses_mrope:
            self._calc_mrope_positions(scheduler_output)

        # Get token indices.
        # E.g., [0, 1, 0, 1, 2, 3, 4, 0, 1, 2]
        # -> [0, 1, M, M + 1, M + 2, M + 3, M + 4, 2 * M, 2 * M + 1, 2 * M + 2]
        # where M is the max_model_len.
        token_indices = (positions_np +
                         req_indices * self.input_batch.token_ids_cpu.shape[1])

        # NOTE(woosuk): We use torch.index_select instead of np.take here
        # because torch.index_select is much faster than np.take for large
        # tensors.
        torch.index_select(self.input_batch.token_ids_cpu_tensor.flatten(),
                           0,
                           torch.from_numpy(token_indices),
                           out=self.input_ids_cpu[:total_num_scheduled_tokens])

        # Calculate the slot mapping for each KV cache group.
        for kv_cache_group_id, kv_cache_group_spec in enumerate(
                self.kv_cache_config.kv_cache_groups):
            block_size = kv_cache_group_spec.kv_cache_spec.block_size
            block_table: BlockTable = self.input_batch.block_table[
                kv_cache_group_id]
            # E.g., [0, 1, 0, 1, 2, 3, 4, 0, 1, 2]
            # -> [0, 0, K, K, K + 1, K + 1, K + 2, 2 * K, 2 * K, 2 * K + 1]
            # where K is the max_num_blocks_per_req and the block size is 2.
            # NOTE(woosuk): We can't simply use `token_indices // block_size`
            # here because M (max_model_len) is not necessarily divisible by
            # block_size.
            block_table_indices = (
                req_indices * block_table.max_num_blocks_per_req +
                positions_np // block_size)
            block_table_cpu = block_table.get_cpu_tensor()
            block_numbers = block_table_cpu.flatten(
            )[block_table_indices].numpy()
            block_offsets = positions_np % block_size
            np.add(
                block_numbers * block_size,
                block_offsets,
                out=block_table.slot_mapping_np[:total_num_scheduled_tokens])

        # Prepare the attention metadata.
        self.query_start_loc_np[0] = 0
        self.query_start_loc_np[1:num_reqs + 1] = cu_num_tokens

        self.seq_lens_np[:num_reqs] = (
            self.input_batch.num_computed_tokens_cpu[:num_reqs] +
            num_scheduled_tokens)

        # Copy the tensors to the GPU.
        self.input_ids[:total_num_scheduled_tokens].copy_(
            self.input_ids_cpu[:total_num_scheduled_tokens], non_blocking=True)
        
        self.zero_prepare_inputs(scheduler_output, self.input_ids)

        if self.uses_mrope:
            # Only relevant for models using M-RoPE (e.g, Qwen2-VL)
            self.mrope_positions[:, :total_num_scheduled_tokens].copy_(
                self.mrope_positions_cpu[:, :total_num_scheduled_tokens],
                non_blocking=True)
        else:
            # Common case (1D positions)
            self.positions[:total_num_scheduled_tokens].copy_(
                self.positions_cpu[:total_num_scheduled_tokens],
                non_blocking=True)

        self.query_start_loc[:num_reqs + 1].copy_(
            self.query_start_loc_cpu[:num_reqs + 1], non_blocking=True)
        self.seq_lens[:num_reqs].copy_(self.seq_lens_cpu[:num_reqs],
                                       non_blocking=True)

        # Fill unused with -1. Needed for reshape_and_cache
        self.seq_lens[num_reqs:].fill_(0)
        # Note: pad query_start_loc to be non-decreasing, as kernels
        # like FlashAttention requires that
        self.query_start_loc[num_reqs + 1:].fill_(
            self.query_start_loc_cpu[num_reqs].item())

        query_start_loc = self.query_start_loc[:num_reqs + 1]
        seq_lens = self.seq_lens[:num_reqs]

        common_attn_metadata = CommonAttentionMetadata(
            query_start_loc=query_start_loc,
            seq_lens=seq_lens,
            num_reqs=num_reqs,
            num_actual_tokens=total_num_scheduled_tokens,
            max_query_len=max_num_scheduled_tokens,
        )

        attn_metadata: dict[str, Any] = {}
        # Prepare the attention metadata for each KV cache group and make layers
        # in the same group share the same metadata.
        for kv_cache_group_id, kv_cache_group_spec in enumerate(
                self.kv_cache_config.kv_cache_groups):

            # Prepare for cascade attention if enabled & beneficial.
            common_prefix_len = 0
            builder = self.attn_metadata_builders[kv_cache_group_id]
            if self.cascade_attn_enabled:
                common_prefix_len = self._compute_cascade_attn_prefix_len(
                    num_scheduled_tokens,
                    scheduler_output.
                    num_common_prefix_blocks[kv_cache_group_id],
                    kv_cache_group_spec.kv_cache_spec,
                    builder,
                )

            attn_metadata_i = (builder.build(
                common_prefix_len=common_prefix_len,
                common_attn_metadata=common_attn_metadata,
            ))

            for layer_name in kv_cache_group_spec.layer_names:
                attn_metadata[layer_name] = attn_metadata_i

        attention_cuda_graphs = all(
            b.can_run_in_cudagraph(common_attn_metadata)
            for b in self.attn_metadata_builders)

        use_spec_decode = len(
            scheduler_output.scheduled_spec_decode_tokens) > 0
        if not use_spec_decode:
            # NOTE(woosuk): Due to chunked prefills, the batch may contain
            # partial requests. While we should not sample any token
            # from these partial requests, we do so for simplicity.
            # We will ignore the sampled tokens from the partial requests.
            # TODO: Support prompt logprobs.
            logits_indices = query_start_loc[1:] - 1
            spec_decode_metadata = None
        else:
            # Get the number of draft tokens for each request.
            # Iterate over the dictionary rather than all requests since not all
            # requests have draft tokens.
            num_draft_tokens = np.zeros(num_reqs, dtype=np.int32)
            for req_id, draft_token_ids in (
                    scheduler_output.scheduled_spec_decode_tokens.items()):
                req_idx = self.input_batch.req_id_to_index[req_id]
                num_draft_tokens[req_idx] = len(draft_token_ids)

            spec_decode_ids = None
            if envs.VLLM_REJECT_SAMPLE_OPT:
                spec_decode_ids = scheduler_output.scheduled_spec_decode_tokens.keys()

            spec_decode_metadata = self._calc_spec_decode_metadata(
                num_draft_tokens, cu_num_tokens, spec_decode_ids)
            logits_indices = spec_decode_metadata.logits_indices

        # Hot-Swap lora model
        if self.lora_config:
            self.set_active_loras(self.input_batch, num_scheduled_tokens)

        return (attn_metadata, attention_cuda_graphs, logits_indices,
                spec_decode_metadata, num_scheduled_tokens)

    def zero_prepare_inputs(self, scheduler_output, input_ids):
        req_ids = self.input_batch.req_ids
        update_req_indices = []
        input_ids_indices = []
        token_idx = 0
        if self.last_draft_token_ids is not None:
            draft_tokens_num = self.last_draft_token_ids.shape[1]
            for req_id in req_ids:
                if req_id in self.last_sampled_req_ids:
                    req_idx = self.last_sampled_req_ids.index(req_id) * draft_tokens_num
                    for num_idx in range(draft_tokens_num):
                        update_req_indices.append(req_idx + num_idx)
                        input_ids_indices.append(token_idx + num_idx + 1)
                token_idx += draft_tokens_num + 1
            if len(update_req_indices) > 0:
                update_req_indices_tensor = async_tensor_h2d(update_req_indices, torch.int32,
                                                            self.device,
                                                            True)
                input_ids_indices_tensor = async_tensor_h2d(input_ids_indices, torch.int32,
                                                            self.device,
                                                            True)
                last_draft_token_ids = self.last_draft_token_ids.flatten().to(torch.int)
                input_ids[input_ids_indices_tensor] = last_draft_token_ids[update_req_indices_tensor]

        def fused_update_input_ids(
                last_sampled_token_ids,
                input_ids,
                update_req_indices,
                input_ids_indices,
        ):
            B, T = last_sampled_token_ids.shape
            U = update_req_indices.numel()

            BLOCK_T = 1024
            assert T <= BLOCK_T

            grid = (U,)
            fused_last_valid_scatter_kernel[grid](
                last_sampled_token_ids,
                input_ids,
                update_req_indices,
                input_ids_indices,
                last_sampled_token_ids.stride(0),
                last_sampled_token_ids.stride(1),
                T,
                BLOCK_T=BLOCK_T,
            )

        update_req_indices = []
        input_ids_indices = []
        token_idx = 0
        if self.last_sampled_token_ids is not None:
            for req_id in req_ids:
                if req_id in self.last_sampled_req_ids:
                    req_idx = self.last_sampled_req_ids.index(req_id)
                    update_req_indices.append(req_idx)
                    input_ids_indices.append(token_idx)
                token_idx += scheduler_output.num_scheduled_tokens[req_id]
            if len(update_req_indices) > 0:
                update_req_indices_tensor = async_tensor_h2d(update_req_indices, torch.int32,
                                                            self.device,
                                                            True)
                input_ids_indices_tensor = async_tensor_h2d(input_ids_indices, torch.int32,
                                                            self.device,
                                                            True)
                fused_update_input_ids(
                    self.last_sampled_token_ids,
                    input_ids,
                    update_req_indices_tensor,
                    input_ids_indices_tensor)



    def propose_draft_token_ids(
        self,
        scheduler_output: "SchedulerOutput",
        num_accepted_tokens_tensor: torch.Tensor,
        sampled_token_ids: torch.Tensor,
        sampling_metadata: SamplingMetadata,
        hidden_states: torch.Tensor,
        sample_hidden_states: torch.Tensor,
        aux_hidden_states: Optional[torch.Tensor],
        spec_decode_metadata: Optional[SpecDecodeMetadata],
        attn_metadata: dict[str, Any],
    ) -> list[list[int]]:
        num_scheduled_tokens = scheduler_output.total_num_scheduled_tokens
        if self.speculative_config.method == "ngram":
            assert isinstance(self.drafter, NgramProposer)
            spec_token_ids = self.propose_ngram_draft_token_ids(
                sampled_token_ids)
        elif self.speculative_config.method == "medusa":
            assert isinstance(self.drafter, MedusaProposer)
            if sample_hidden_states.shape[0] == len(sampled_token_ids):
                # The input to the target model does not include draft tokens.
                hidden_states = sample_hidden_states
            else:
                indices = []
                offset = 0
                for num_draft, tokens in zip(
                        spec_decode_metadata.num_draft_tokens,
                        sampled_token_ids):
                    indices.append(offset + len(tokens) - 1)
                    offset += num_draft + 1
                indices = torch.tensor(indices, device=self.device)
                hidden_states = sample_hidden_states[indices]

            spec_token_ids = self.drafter.propose(
                target_hidden_states=hidden_states,
                sampling_metadata=sampling_metadata,
            )
        elif self.speculative_config.use_eagle():
            assert isinstance(self.drafter, EagleProposer)
            # TODO(woosuk): Refactor the loop.
            row_indices = torch.arange(sampled_token_ids.size(0), device=sampled_token_ids.device)
            next_token_ids = sampled_token_ids[row_indices, num_accepted_tokens_tensor].flatten()
            # At this moment, we assume all eagle layers belong to the same KV
            # cache group, thus using the same attention metadata.
            eagle_attn_metadata = attn_metadata[
                self.drafter.attn_layer_names[0]]

            # NOTE: deepseek_mtp uses MLA which does not have `block_table`
            if hasattr(eagle_attn_metadata, "block_table"):
                block_table = eagle_attn_metadata.block_table
            else:
                block_table = None

            spec_scheduler_max_num_tokens = self.spec_scheduler_max_num_tokens
            if spec_decode_metadata is None:
                # input_ids can be None for multimodal models.
                target_token_ids = self.input_ids[:num_scheduled_tokens]
                # TODO(woosuk): Support M-RoPE.
                target_positions = self.positions[:num_scheduled_tokens]
                if self.use_aux_hidden_state_outputs:
                    target_hidden_states = torch.cat(
                        [h[:num_scheduled_tokens] for h in aux_hidden_states],
                        dim=-1)
                else:
                    target_hidden_states = hidden_states[:num_scheduled_tokens]
                target_slot_mapping = eagle_attn_metadata.slot_mapping
                cu_num_tokens = eagle_attn_metadata.query_start_loc
            else:
                # TODO(woosuk): Refactor this.
                cu_num_tokens, token_indices = self.drafter.prepare_inputs(
                    eagle_attn_metadata.query_start_loc,
                    num_accepted_tokens_tensor,
                )
                spec_scheduler_max_num_tokens = 1
                target_token_ids = self.input_ids[token_indices]
                # TODO(woosuk): Support M-RoPE.
                target_positions = self.positions[token_indices]
                if self.use_aux_hidden_state_outputs:
                    target_hidden_states = torch.cat(
                        [h[token_indices] for h in aux_hidden_states], dim=-1)
                else:
                    target_hidden_states = hidden_states[token_indices]
                target_slot_mapping = eagle_attn_metadata.slot_mapping[
                    token_indices]
            self.drafter.spec_scheduler_max_num_tokens = spec_scheduler_max_num_tokens
            draft_result = self.drafter.propose(
                target_token_ids=target_token_ids,
                target_positions=target_positions,
                target_hidden_states=target_hidden_states,
                target_slot_mapping=target_slot_mapping,
                next_token_ids=next_token_ids,
                cu_num_tokens=cu_num_tokens,
                block_table=block_table,
                sampling_metadata=sampling_metadata,
                decoding=spec_decode_metadata is not None,
            )

            if not envs.VLLM_REJECT_SAMPLE_OPT:
                draft_token_ids = draft_result
            else:
                draft_token_ids, draft_probs = draft_result
                draft_req_ids = list(scheduler_output.num_scheduled_tokens.keys())
                
                if self.draft_probs is None:
                    self.draft_probs = DraftProbs(
                        draft_probs, draft_req_ids)
                else:
                    self.draft_probs.update(draft_probs, draft_req_ids)

            spec_token_ids = np.ones(draft_token_ids.shape, dtype=int).tolist()
            self.last_draft_token_ids = draft_token_ids
            self.last_draft_host_tokens = draft_token_ids.to('cpu', non_blocking=True)
            self.last_draft_event.record()
        return spec_token_ids

    @torch.inference_mode()
    def execute_model(
        self,
        scheduler_output: "SchedulerOutput",
        intermediate_tensors: Optional[IntermediateTensors] = None,
    ) -> Union[ModelRunnerOutput, IntermediateTensors]:
        self._update_states(scheduler_output)
        if not scheduler_output.total_num_scheduled_tokens:
            if not has_kv_transfer_group():
                # Return empty ModelRunnerOutput if there's no work to do.
                return EMPTY_MODEL_RUNNER_OUTPUT

            return self.kv_connector_no_forward(scheduler_output)

        # Prepare the decoder inputs.
        (attn_metadata, attention_cuda_graphs, logits_indices,
         spec_decode_metadata,
         num_scheduled_tokens_np) = (self._prepare_inputs(scheduler_output))
        num_scheduled_tokens = scheduler_output.total_num_scheduled_tokens
        
        # make sure that the padded length is divisible by attn_tp_size because we may need reduce-scatter across attn_tp dim.
        if self.ep_sp:
            num_input_tokens = round_up(num_scheduled_tokens, tp_size)
            if (self.use_cuda_graph
                    and num_input_tokens <= self.cudagraph_batch_sizes[-1]):
                # Use piecewise CUDA graphs.
                # Add padding to the batch size.
                num_input_tokens = self.vllm_config.pad_for_cudagraph(
                    num_input_tokens)
        else:
            if (self.use_cuda_graph
                    and num_scheduled_tokens <= self.cudagraph_batch_sizes[-1]):
                # Use piecewise CUDA graphs.
                # Add padding to the batch size.
                num_input_tokens = self.vllm_config.pad_for_cudagraph(
                    num_scheduled_tokens)
            else:
                # Eager mode.
                # Pad tokens to multiple of tensor_parallel_size when
                # enabled collective fusion for SP
                tp_size = self.vllm_config.parallel_config.tensor_parallel_size
                if self.compilation_config.pass_config. \
                        enable_sequence_parallelism and tp_size > 1:
                    num_input_tokens = round_up(num_scheduled_tokens, tp_size)
                else:
                    num_input_tokens = num_scheduled_tokens

        # Padding for DP
        num_pad, num_tokens_across_dp = self.get_dp_padding(num_input_tokens)
        num_input_tokens += num_pad

        # _prepare_inputs may reorder the batch, so we must gather multi
        # modal outputs after that to ensure the correct order
        if self.is_multimodal_model:
            # Run the multimodal encoder if any.
            self._execute_mm_encoder(scheduler_output)
            mm_embeds = self._gather_mm_embeddings(scheduler_output)
        else:
            mm_embeds = []

        if self.is_multimodal_model and get_pp_group().is_first_rank:
            # NOTE(woosuk): To unify token ids and soft tokens (vision
            # embeddings), we always use embeddings (rather than token ids)
            # as input to the multimodal model, even when the input is text.
            input_ids = self.input_ids[:num_scheduled_tokens]
            if mm_embeds:
                inputs_embeds = self.model.get_input_embeddings(
                    input_ids, mm_embeds)
            else:
                inputs_embeds = self.model.get_input_embeddings(input_ids)
            # TODO(woosuk): Avoid the copy. Optimize.
            self.inputs_embeds[:num_scheduled_tokens].copy_(inputs_embeds)
            inputs_embeds = self.inputs_embeds[:num_input_tokens]
            input_ids = None
        else:
            # For text-only models, we use token ids as input.
            # While it is possible to use embeddings as input just like the
            # multimodal models, it is not desirable for performance since
            # then the embedding layer is not included in the CUDA graph.
            input_ids = self.input_ids[:num_input_tokens]
            inputs_embeds = None
        if self.uses_mrope:
            positions = self.mrope_positions[:, :num_input_tokens]
        else:
            positions = self.positions[:num_input_tokens]

        if get_pp_group().is_first_rank:
            intermediate_tensors = None
        else:
            intermediate_tensors = self.sync_and_slice_intermediate_tensors(
                num_input_tokens, intermediate_tensors, True)

        # Some attention backends only support CUDA Graphs in pure decode.
        # If attention doesn't support CUDA Graphs for this batch, but we
        # compiled with full CUDA graphs, we have to skip them entirely.
        skip_cuda_graphs = self.full_cuda_graph and not attention_cuda_graphs
        if envs.VLLM_ENABLE_TBO and (not self.use_cuda_graph or skip_cuda_graphs):
            model_output, finished_sending, finished_recving = \
                 tbo_split_and_execute_model(self, attn_metadata, num_input_tokens,
                                             num_tokens_across_dp, input_ids, positions,
                                             inputs_embeds, scheduler_output, intermediate_tensors, 
                                             skip_cuda_graphs)
        else:
            # Run the model.
            # Use persistent buffers for CUDA graphs.
            with set_forward_context(
                    attn_metadata,
                    self.vllm_config,
                    num_tokens=num_input_tokens,
                    num_tokens_across_dp=num_tokens_across_dp,
                    skip_cuda_graphs=skip_cuda_graphs,
            ):
                self.maybe_setup_kv_connector(scheduler_output)

                model_output = self.model(
                    input_ids=input_ids,
                    positions=positions,
                    intermediate_tensors=intermediate_tensors,
                    inputs_embeds=inputs_embeds,
                )

                self.maybe_wait_for_kv_save()
                finished_sending, finished_recving = (
                    self.get_finished_kv_transfers(scheduler_output))
        if self.use_aux_hidden_state_outputs:
            hidden_states, aux_hidden_states = model_output
        else:
            hidden_states = model_output
            aux_hidden_states = None

        # Broadcast PP output for external_launcher (torchrun)
        # to make sure we are synced across pp ranks
        # TODO: Support overlapping mirco-batches
        # https://github.com/vllm-project/vllm/issues/18019
        broadcast_pp_output = \
            self.parallel_config.distributed_executor_backend \
            == "external_launcher" and len(get_pp_group().ranks) > 0
        if not get_pp_group().is_last_rank:
            # For mid-pipeline stages, return the hidden states.
            if not broadcast_pp_output:
                return hidden_states
            assert isinstance(hidden_states, IntermediateTensors)
            get_pp_group().send_tensor_dict(hidden_states.tensors,
                                            all_gather_group=get_tp_group())
            logits = None
        else:
            if self.input_batch.pooling_params:
                return self._pool(hidden_states, num_scheduled_tokens,
                                  num_scheduled_tokens_np, finished_sending,
                                  finished_recving)

            sample_hidden_states = hidden_states[logits_indices]
            logits = self.model.compute_logits(sample_hidden_states, None)
        if broadcast_pp_output:
            model_output_broadcast_data = {
                "logits": logits.contiguous(),
            } if logits is not None else {}
            model_output_broadcast_data = get_pp_group().broadcast_tensor_dict(
                model_output_broadcast_data, src=len(get_pp_group().ranks) - 1)
            assert model_output_broadcast_data is not None
            logits = model_output_broadcast_data["logits"]

        # Apply structured output bitmasks if present
        if scheduler_output.grammar_bitmask is not None:
            self.apply_grammar_bitmask(scheduler_output, logits)

        # Sample the next token and get logprobs if needed.
        sampling_metadata = self.input_batch.sampling_metadata
        if spec_decode_metadata is None:
            sampler_output = self.sampler(
                logits=logits,
                sampling_metadata=sampling_metadata,
            )
        else:
            # When indexing with a tensor (bonus_logits_indices), PyTorch
            # creates a new tensor with separate storage from the original
            # logits tensor. This means any in-place operations on bonus_logits
            # won't affect the original logits tensor.
            assert logits is not None
            if not envs.VLLM_REJECT_SAMPLE_OPT:
                bonus_logits = logits[spec_decode_metadata.bonus_logits_indices]
                sampler_output = self.sampler(
                    logits=bonus_logits,
                    sampling_metadata=sampling_metadata,
                )
                bonus_token_ids = sampler_output.sampled_token_ids

                # Just like `bonus_logits`, `target_logits` is a new tensor with
                # separate storage from the original `logits` tensor. Therefore,
                # it is safe to update `target_logits` in place.
                target_logits = logits[spec_decode_metadata.target_logits_indices]
                output_token_ids = self.rejection_sampler(
                    spec_decode_metadata,
                    None,  # draft_probs
                    target_logits,
                    bonus_token_ids,
                    sampling_metadata,
                )
                sampler_output.sampled_token_ids = output_token_ids
            else:
                # sampling_metadata.all_greedy = True
                # sampling_metadata.all_random = False
                sampler_output = self.sampler(
                    logits=logits,
                    sampling_metadata=sampling_metadata,
                )
                target_token_ids = sampler_output.sampled_token_ids[spec_decode_metadata.target_logits_indices]
                target_logits = logits[spec_decode_metadata.target_logits_indices]

                bonus_token_ids = sampler_output.sampled_token_ids[spec_decode_metadata.bonus_logits_indices]

                output_token_ids = self.rejection_sampler(
                    spec_decode_metadata,
                    self.draft_probs.get_probs(spec_decode_metadata.spec_decode_ids),
                    target_logits,
                    target_token_ids,
                    bonus_token_ids,
                    sampling_metadata,
                )
                sampler_output.sampled_token_ids = output_token_ids

        num_nans_in_logits = {}
        if envs.VLLM_COMPUTE_NANS_IN_LOGITS:
            num_nans_in_logits = self._get_nans_in_logits(logits)

        # TODO(woosuk): The following loop can be slow since it iterates over
        # the requests one by one. Optimize.
        discard_sampled_tokens_req_indices = []
        for i, req_id in enumerate(self.input_batch.req_ids):
            req_state = self.requests[req_id]
            seq_len = (req_state.num_computed_tokens +
                       scheduler_output.num_scheduled_tokens[req_id])
            if seq_len < req_state.num_tokens:
                # Ignore the sampled token for partial prefills.
                # Rewind the generator state as if the token was not sampled.
                # This relies on cuda-specific torch-internal impl details
                generator = self.input_batch.generators.get(i)
                if generator is not None:
                    generator.set_offset(generator.get_offset() - 4)
                # Record the index of the request that should not be sampled,
                # so that we could clear the sampled tokens before returning.
                discard_sampled_tokens_req_indices.append(i)

        # NOTE: GPU -> CPU Sync happens here.
        # Move as many CPU operations as possible before this sync point.
        logprobs_tensors = sampler_output.logprobs_tensors
        logprobs_lists = logprobs_tensors.tolists() \
            if logprobs_tensors is not None else None

        # Compute prompt logprobs if needed.
        prompt_logprobs_dict = self._get_prompt_logprobs_dict(
            hidden_states[:num_scheduled_tokens],
            scheduler_output,
        )

        fix_draft_token_ids = None
        fix_draft_req_ids = self.last_sampled_req_ids
        is_output_valid = False
        # Get the valid generated tokens.
        sampled_token_ids = sampler_output.sampled_token_ids

        self.last_sampler_host_tokens = None
        self.last_sampled_token_ids = None
        self.last_sampler_host_tokens = sampled_token_ids.to('cpu', non_blocking=True)
        self.last_sampler_event.record()
        self.last_sampled_token_ids = sampled_token_ids
        valid_sampled_token_ids = np.ones(sampled_token_ids.shape, dtype=int).tolist()

        # Mask out the sampled tokens that should not be sampled.
        for i in discard_sampled_tokens_req_indices:
            valid_sampled_token_ids[i].clear()

        # Cache the sampled tokens in the model runner, so that the scheduler
        # doesn't need to send them back.
        # NOTE(woosuk): As an exception, when using PP, the scheduler sends
        # the sampled tokens back, because there's no direct communication
        # between the first-stage worker and the last-stage worker.
        self.token_ids_cpu_fix_record.clear()
        self.last_sampled_req_ids = []
        self.last_sampled_token_lens = []
        for req_idx, sampled_ids in enumerate(valid_sampled_token_ids):
            req_id = self.input_batch.req_ids[req_idx]
            self.last_sampled_req_ids.append(req_id)
            cache_output_len = -1
            if not sampled_ids:
                self.last_sampled_token_lens.append(-1)
                self.token_ids_cpu_fix_record.append([req_idx, -1, -1])
                continue

            start_idx = self.input_batch.num_tokens_no_spec[req_idx]
            end_idx = start_idx + len(sampled_ids)
            assert end_idx <= self.max_model_len, (
                "Sampled token IDs exceed the max model length. "
                f"Total number of tokens: {end_idx} > max_model_len: "
                f"{self.max_model_len}")

            self.input_batch.token_ids_cpu[req_idx,
                                            start_idx:end_idx] = sampled_ids
            self.token_ids_cpu_fix_record.append([req_idx, start_idx, end_idx])
            self.input_batch.num_tokens_no_spec[req_idx] = end_idx
            self.input_batch.num_tokens[req_idx] = end_idx

        if not self.speculative_config:
            # Speculative decoding is not enabled.
            spec_token_ids = None
            fix_draft_req_ids = None
        else:
            sampled_token_ids_cpu = sampled_token_ids.to('cpu', non_blocking=True)
            if self.last_draft_host_tokens is not None:
                self.last_draft_event.synchronize()
                fix_draft_token_ids = self.last_draft_host_tokens.tolist()

            mask = (sampled_token_ids == -1)
            mask_int = mask.int()
            first_neg_one_indices = torch.argmax(mask_int, dim=1)
            num_accepted_tokens_tensor = torch.where(torch.any(mask, dim=1), first_neg_one_indices, sampled_token_ids.size(1)) - 1
            spec_token_ids = self.propose_draft_token_ids(
                scheduler_output,
                num_accepted_tokens_tensor,
                sampled_token_ids,
                sampling_metadata,
                hidden_states,
                sample_hidden_states,
                aux_hidden_states,
                spec_decode_metadata,
                attn_metadata,
            )

        # Clear KVConnector state after all KVs are generated.
        if has_kv_transfer_group():
            get_kv_transfer_group().clear_connector_metadata()

        self.eplb_step()

        model_output = ZeroV1ModelRunnerOutput(
            req_ids=self.input_batch.req_ids,
            req_id_to_index=self.input_batch.req_id_to_index,
            sampled_token_ids=valid_sampled_token_ids,
            spec_token_ids=spec_token_ids,
            logprobs=logprobs_lists,
            prompt_logprobs_dict=prompt_logprobs_dict,
            pooler_output=[],
            finished_sending=finished_sending,
            finished_recving=finished_recving,
            num_nans_in_logits=num_nans_in_logits,
            fix_req_ids=self.fix_req_ids,
            fix_sampled_token_ids=self.fix_sampled_token_ids,
            fix_draft_tokens_ids=fix_draft_token_ids,
            fix_draft_req_ids=fix_draft_req_ids,
            is_output_valid=is_output_valid
        )
        return model_output