ChatTTS/model/velocity/model_runner.py

import time
from typing import Dict, List, Optional, Tuple, Union

import numpy as np
import torch
import torch.nn as nn

from .configs import ModelConfig, ParallelConfig, SchedulerConfig
from vllm.logger import init_logger
from .model_loader import get_model
from vllm.model_executor import InputMetadata, SamplingMetadata
from vllm.model_executor.parallel_utils.communication_op import (
    broadcast,
    broadcast_object_list,
)
from .sampling_params import SamplingParams, SamplingType
from .sequence import (
    SamplerOutput,
    SequenceData,
    SequenceGroupMetadata,
    SequenceGroupOutput,
    SequenceOutput,
)
from vllm.utils import in_wsl
from ..embed import Embed
from .sampler import Sampler
from safetensors.torch import safe_open

logger = init_logger(__name__)

KVCache = Tuple[torch.Tensor, torch.Tensor]
_PAD_SLOT_ID = -1
# Capture graphs for batch size 1, 2, 4, 8, 16, 24, 32, 40, ..., 256.
# NOTE: _get_graph_batch_size needs to be updated if this list is changed.
_BATCH_SIZES_TO_CAPTURE = [1, 2, 4] + [8 * i for i in range(1, 33)]


class ModelRunner:

    def __init__(
        self,
        model_config: ModelConfig,
        parallel_config: ParallelConfig,
        scheduler_config: SchedulerConfig,
        is_driver_worker: bool = False,
        post_model_path: str = None,
    ):
        self.model_config = model_config
        self.parallel_config = parallel_config
        self.scheduler_config = scheduler_config
        self.is_driver_worker = is_driver_worker
        self.post_model_path = post_model_path

        # model_config can be None in tests/samplers/test_sampler.py.
        # FIXME(woosuk): This is a hack to make the tests work. Refactor this.
        self.sliding_window = (
            model_config.get_sliding_window() if model_config is not None else None
        )
        self.model = None
        self.block_size = None  # Set after initial profiling.

        self.graph_runners: Dict[int, CUDAGraphRunner] = {}
        self.graph_memory_pool = None  # Set during graph capture.

        self.max_context_len_to_capture = (
            self.model_config.max_context_len_to_capture
            if self.model_config is not None
            else 0
        )
        # When using CUDA graph, the input block tables must be padded to
        # max_context_len_to_capture. However, creating the block table in
        # Python can be expensive. To optimize this, we cache the block table
        # in numpy and only copy the actual input content at every iteration.
        # The shape of the cached block table will be
        # (max batch size to capture, max context len to capture / block size).
        self.graph_block_tables = None  # Set after initial profiling.
        # cache in_wsl result
        self.in_wsl = in_wsl()

    def load_model(self) -> None:
        self.model = get_model(self.model_config)
        self.post_model = Embed(
            self.model_config.get_hidden_size(),
            self.model_config.num_audio_tokens,
            self.model_config.num_text_tokens,
        )
        state_dict_tensors = {}
        with safe_open(self.post_model_path, framework="pt", device=0) as f:
            for k in f.keys():
                state_dict_tensors[k] = f.get_tensor(k)
        self.post_model.load_state_dict(state_dict_tensors)
        self.post_model.to(next(self.model.parameters())).eval()
        self.sampler = Sampler(self.post_model, self.model_config.num_audio_tokens, 4)

    def set_block_size(self, block_size: int) -> None:
        self.block_size = block_size

        max_num_blocks = (
            self.max_context_len_to_capture + block_size - 1
        ) // block_size
        self.graph_block_tables = np.zeros(
            (max(_BATCH_SIZES_TO_CAPTURE), max_num_blocks), dtype=np.int32
        )

    def _prepare_prompt(
        self,
        seq_group_metadata_list: List[SequenceGroupMetadata],
    ) -> Tuple[torch.Tensor, torch.Tensor, InputMetadata, List[int]]:
        assert len(seq_group_metadata_list) > 0
        input_tokens: List[List[int]] = []
        input_positions: List[List[int]] = []
        slot_mapping: List[List[int]] = []

        prompt_lens: List[int] = []
        for seq_group_metadata in seq_group_metadata_list:
            assert seq_group_metadata.is_prompt
            seq_ids = list(seq_group_metadata.seq_data.keys())
            assert len(seq_ids) == 1
            seq_id = seq_ids[0]

            seq_data = seq_group_metadata.seq_data[seq_id]
            prompt_tokens = seq_data.get_token_ids()
            prompt_len = len(prompt_tokens)
            prompt_lens.append(prompt_len)

            input_tokens.append(prompt_tokens)
            # NOTE(woosuk): Here we assume that the first token in the prompt
            # is always the first token in the sequence.
            input_positions.append(list(range(prompt_len)))

            if seq_group_metadata.block_tables is None:
                # During memory profiling, the block tables are not initialized
                # yet. In this case, we just use a dummy slot mapping.
                slot_mapping.append([_PAD_SLOT_ID] * prompt_len)
                continue

            # Compute the slot mapping.
            slot_mapping.append([])
            block_table = seq_group_metadata.block_tables[seq_id]
            # Mask the [0, start_idx) tokens of the prompt with _PAD_SLOT_ID,
            # where start_idx is max(0, prompt_len - sliding_window).
            # For example, if the prompt len is 10, sliding window is 8, and
            # block size is 4, the first two tokens are masked and the slot
            # mapping will be [-1, -1, 2, 3, 4, 5, 6, 7, 0, 1].
            start_idx = 0
            if self.sliding_window is not None:
                start_idx = max(0, prompt_len - self.sliding_window)
            for i in range(prompt_len):
                if i < start_idx:
                    slot_mapping[-1].append(_PAD_SLOT_ID)
                    continue

                block_number = block_table[i // self.block_size]
                block_offset = i % self.block_size
                slot = block_number * self.block_size + block_offset
                slot_mapping[-1].append(slot)

        max_prompt_len = max(prompt_lens)
        input_tokens = _make_tensor_with_pad(
            input_tokens, max_prompt_len, pad=0, dtype=torch.long
        )
        input_positions = _make_tensor_with_pad(
            input_positions, max_prompt_len, pad=0, dtype=torch.long
        )
        slot_mapping = _make_tensor_with_pad(
            slot_mapping, max_prompt_len, pad=_PAD_SLOT_ID, dtype=torch.long
        )

        input_metadata = InputMetadata(
            is_prompt=True,
            slot_mapping=slot_mapping,
            max_context_len=None,
            context_lens=None,
            block_tables=None,
            use_cuda_graph=False,
        )
        return input_tokens, input_positions, input_metadata, prompt_lens

    def _prepare_decode(
        self,
        seq_group_metadata_list: List[SequenceGroupMetadata],
    ) -> Tuple[torch.Tensor, torch.Tensor, InputMetadata]:
        assert len(seq_group_metadata_list) > 0
        input_tokens: List[List[int]] = []
        input_positions: List[List[int]] = []
        slot_mapping: List[List[int]] = []
        context_lens: List[int] = []
        block_tables: List[List[int]] = []

        for seq_group_metadata in seq_group_metadata_list:
            assert not seq_group_metadata.is_prompt

            seq_ids = list(seq_group_metadata.seq_data.keys())
            for seq_id in seq_ids:
                seq_data = seq_group_metadata.seq_data[seq_id]
                generation_token = seq_data.get_last_token_id()
                input_tokens.append([generation_token])

                seq_len = seq_data.get_len()
                position = seq_len - 1
                input_positions.append([position])

                context_len = (
                    seq_len
                    if self.sliding_window is None
                    else min(seq_len, self.sliding_window)
                )
                context_lens.append(context_len)

                block_table = seq_group_metadata.block_tables[seq_id]
                block_number = block_table[position // self.block_size]
                block_offset = position % self.block_size
                slot = block_number * self.block_size + block_offset
                slot_mapping.append([slot])

                if self.sliding_window is not None:
                    sliding_window_blocks = self.sliding_window // self.block_size
                    block_table = block_table[-sliding_window_blocks:]
                block_tables.append(block_table)

        batch_size = len(input_tokens)
        max_context_len = max(context_lens)
        use_captured_graph = (
            not self.model_config.enforce_eager
            and batch_size <= _BATCH_SIZES_TO_CAPTURE[-1]
            and max_context_len <= self.max_context_len_to_capture
        )
        if use_captured_graph:
            # Pad the input tokens, positions, and slot mapping to match the
            # batch size of the captured graph.
            graph_batch_size = _get_graph_batch_size(batch_size)
            assert graph_batch_size >= batch_size
            for _ in range(graph_batch_size - batch_size):
                input_tokens.append([])
                input_positions.append([])
                slot_mapping.append([])
                context_lens.append(1)
                block_tables.append([])
            batch_size = graph_batch_size

        input_tokens = _make_tensor_with_pad(
            input_tokens, max_len=1, pad=0, dtype=torch.long, device="cuda"
        )
        input_positions = _make_tensor_with_pad(
            input_positions, max_len=1, pad=0, dtype=torch.long, device="cuda"
        )
        slot_mapping = _make_tensor_with_pad(
            slot_mapping, max_len=1, pad=_PAD_SLOT_ID, dtype=torch.long, device="cuda"
        )
        context_lens = torch.tensor(context_lens, dtype=torch.int, device="cuda")

        if use_captured_graph:
            # The shape of graph_block_tables is
            # [max batch size, max context len // block size].
            input_block_tables = self.graph_block_tables[:batch_size]
            for i, block_table in enumerate(block_tables):
                if block_table:
                    input_block_tables[i, : len(block_table)] = block_table
            block_tables = torch.tensor(input_block_tables, device="cuda")
        else:
            block_tables = _make_tensor_with_pad(
                block_tables,
                max_len=max_context_len,
                pad=0,
                dtype=torch.int,
                device="cuda",
            )

        input_metadata = InputMetadata(
            is_prompt=False,
            slot_mapping=slot_mapping,
            max_context_len=max_context_len,
            context_lens=context_lens,
            block_tables=block_tables,
            use_cuda_graph=use_captured_graph,
        )
        return input_tokens, input_positions, input_metadata

    def _prepare_sample(
        self,
        seq_group_metadata_list: List[SequenceGroupMetadata],
        prompt_lens: List[int],
    ) -> SamplingMetadata:
        seq_groups: List[Tuple[List[int], SamplingParams]] = []
        selected_token_indices: List[int] = []
        selected_token_start_idx = 0
        categorized_sample_indices = {t: [] for t in SamplingType}
        categorized_sample_indices_start_idx = 0

        max_prompt_len = max(prompt_lens) if prompt_lens else 1
        for i, seq_group_metadata in enumerate(seq_group_metadata_list):
            seq_ids = list(seq_group_metadata.seq_data.keys())
            sampling_params = seq_group_metadata.sampling_params
            seq_groups.append((seq_ids, sampling_params))

            if seq_group_metadata.is_prompt:
                assert len(seq_ids) == 1
                prompt_len = prompt_lens[i]
                if sampling_params.prompt_logprobs is not None:
                    # NOTE: prompt token positions do not need sample, skip
                    categorized_sample_indices_start_idx += prompt_len - 1

                categorized_sample_indices[sampling_params.sampling_type].append(
                    categorized_sample_indices_start_idx
                )
                categorized_sample_indices_start_idx += 1

                if sampling_params.prompt_logprobs is not None:
                    selected_token_indices.extend(
                        range(
                            selected_token_start_idx,
                            selected_token_start_idx + prompt_len - 1,
                        )
                    )
                selected_token_indices.append(selected_token_start_idx + prompt_len - 1)
                selected_token_start_idx += max_prompt_len
            else:
                num_seqs = len(seq_ids)
                selected_token_indices.extend(
                    range(selected_token_start_idx, selected_token_start_idx + num_seqs)
                )
                selected_token_start_idx += num_seqs

                categorized_sample_indices[sampling_params.sampling_type].extend(
                    range(
                        categorized_sample_indices_start_idx,
                        categorized_sample_indices_start_idx + num_seqs,
                    )
                )
                categorized_sample_indices_start_idx += num_seqs

        selected_token_indices = _async_h2d(
            selected_token_indices, dtype=torch.long, pin_memory=not self.in_wsl
        )
        categorized_sample_indices = {
            t: _async_h2d(seq_ids, dtype=torch.int, pin_memory=not self.in_wsl)
            for t, seq_ids in categorized_sample_indices.items()
        }

        seq_data: Dict[int, SequenceData] = {}
        for seq_group_metadata in seq_group_metadata_list:
            seq_data.update(seq_group_metadata.seq_data)

        sampling_metadata = SamplingMetadata(
            seq_groups=seq_groups,
            seq_data=seq_data,
            prompt_lens=prompt_lens,
            selected_token_indices=selected_token_indices,
            categorized_sample_indices=categorized_sample_indices,
        )
        return sampling_metadata

    def prepare_input_tensors(
        self,
        seq_group_metadata_list: Optional[List[SequenceGroupMetadata]],
    ) -> Tuple[torch.Tensor, torch.Tensor, InputMetadata, SamplingMetadata]:
        if self.is_driver_worker:
            # NOTE: We assume that all sequences in the group are all prompts or
            # all decodes.
            is_prompt = seq_group_metadata_list[0].is_prompt
            # Prepare input tensors.
            if is_prompt:
                (input_tokens, input_positions, input_metadata, prompt_lens) = (
                    self._prepare_prompt(seq_group_metadata_list)
                )
            else:
                (input_tokens, input_positions, input_metadata) = self._prepare_decode(
                    seq_group_metadata_list
                )
                prompt_lens = []
            sampling_metadata = self._prepare_sample(
                seq_group_metadata_list, prompt_lens
            )

            def get_size_or_none(x: Optional[torch.Tensor]):
                return x.size() if x is not None else None

            # Broadcast the input data. For input tensors, we first broadcast
            # its shape and then broadcast the tensor to avoid high
            # serialization cost.
            py_data = {
                "input_tokens_size": input_tokens.size(),
                "input_positions_size": input_positions.size(),
                "is_prompt": input_metadata.is_prompt,
                "slot_mapping_size": get_size_or_none(input_metadata.slot_mapping),
                "max_context_len": input_metadata.max_context_len,
                "context_lens_size": get_size_or_none(input_metadata.context_lens),
                "block_tables_size": get_size_or_none(input_metadata.block_tables),
                "use_cuda_graph": input_metadata.use_cuda_graph,
                "selected_token_indices_size": sampling_metadata.selected_token_indices.size(),
            }
            broadcast_object_list([py_data], src=0)
            # TODO(zhuohan): Combine the broadcasts or set async_op=True.
            broadcast(input_tokens, src=0)
            broadcast(input_positions, src=0)
            if input_metadata.slot_mapping is not None:
                broadcast(input_metadata.slot_mapping, src=0)
            if input_metadata.context_lens is not None:
                broadcast(input_metadata.context_lens, src=0)
            if input_metadata.block_tables is not None:
                broadcast(input_metadata.block_tables, src=0)
            broadcast(sampling_metadata.selected_token_indices, src=0)
        else:
            receving_list = [None]
            broadcast_object_list(receving_list, src=0)
            py_data = receving_list[0]
            input_tokens = torch.empty(
                *py_data["input_tokens_size"], dtype=torch.long, device="cuda"
            )
            broadcast(input_tokens, src=0)
            input_positions = torch.empty(
                *py_data["input_positions_size"], dtype=torch.long, device="cuda"
            )
            broadcast(input_positions, src=0)
            if py_data["slot_mapping_size"] is not None:
                slot_mapping = torch.empty(
                    *py_data["slot_mapping_size"], dtype=torch.long, device="cuda"
                )
                broadcast(slot_mapping, src=0)
            else:
                slot_mapping = None
            if py_data["context_lens_size"] is not None:
                context_lens = torch.empty(
                    *py_data["context_lens_size"], dtype=torch.int, device="cuda"
                )
                broadcast(context_lens, src=0)
            else:
                context_lens = None
            if py_data["block_tables_size"] is not None:
                block_tables = torch.empty(
                    *py_data["block_tables_size"], dtype=torch.int, device="cuda"
                )
                broadcast(block_tables, src=0)
            else:
                block_tables = None
            selected_token_indices = torch.empty(
                *py_data["selected_token_indices_size"], dtype=torch.long, device="cuda"
            )
            broadcast(selected_token_indices, src=0)
            input_metadata = InputMetadata(
                is_prompt=py_data["is_prompt"],
                slot_mapping=slot_mapping,
                max_context_len=py_data["max_context_len"],
                context_lens=context_lens,
                block_tables=block_tables,
                use_cuda_graph=py_data["use_cuda_graph"],
            )
            sampling_metadata = SamplingMetadata(
                seq_groups=None,
                seq_data=None,
                prompt_lens=None,
                selected_token_indices=selected_token_indices,
                categorized_sample_indices=None,
                perform_sampling=False,
            )

        return input_tokens, input_positions, input_metadata, sampling_metadata

    @torch.inference_mode()
    def execute_model(
        self,
        seq_group_metadata_list: Optional[List[SequenceGroupMetadata]],
        kv_caches: List[Tuple[torch.Tensor, torch.Tensor]],
    ) -> Optional[SamplerOutput]:
        input_tokens, input_positions, input_metadata, sampling_metadata = (
            self.prepare_input_tensors(seq_group_metadata_list)
        )
        # print(sampling_metadata.seq_data)
        seq_groups = []
        input_tokens_history = []
        for i, rtn in enumerate(sampling_metadata.seq_groups):
            seq_groups.append(rtn[0][0])
            tokens_history = sampling_metadata.seq_data[rtn[0][0]].output_token_ids
            if len(tokens_history) >= 1:
                if len(tokens_history[0]) == 1:
                    tokens_history = [token[0] for token in tokens_history]
                else:
                    tokens_history = [list(token) for token in tokens_history]
            input_tokens_history.append(tokens_history)
        input_tokens_history = torch.tensor(input_tokens_history).to(
            input_tokens.device
        )
        # token_ids = rtn.outputs[0].token_ids
        # for j, token_id in enumerate(token_ids):
        #     if len(token_id) == 1:
        #         token_ids[j] = token_id[0]
        #     else:
        #         token_ids[j] = list(token_id)

        # Execute the model.
        # print("it1",input_tokens)
        if len(input_tokens.shape) == 2:
            input_tokens = input_tokens.unsqueeze(2).repeat(1, 1, 4)
        if len(input_tokens_history.shape) == 2:
            input_tokens_history = input_tokens_history.unsqueeze(2).repeat(1, 1, 4)
        # print(input_tokens_history.shape)
        # print("it2",input_tokens.shape)
        text_mask = input_tokens != 0
        text_mask = text_mask[:, :, 0]

        if input_metadata.use_cuda_graph:
            graph_batch_size = input_tokens.shape[0]
            model_executable = self.graph_runners[graph_batch_size]
        else:
            model_executable = self.model

        infer_text = sampling_metadata.seq_groups[0][1].infer_text
        temperture = sampling_metadata.seq_groups[0][1].temperature
        if not infer_text:
            temperture = torch.tensor(temperture).to(input_tokens.device)
        logits_processors, logits_warpers = sampling_metadata.seq_groups[0][
            1
        ].logits_processors
        # print(logits_processors, logits_warpers)
        min_new_token = sampling_metadata.seq_groups[0][1].min_new_token
        eos_token = sampling_metadata.seq_groups[0][1].eos_token
        start_idx = sampling_metadata.seq_groups[0][1].start_idx
        if input_tokens.shape[-2] == 1:
            if infer_text:
                input_emb: torch.Tensor = self.post_model.emb_text(
                    input_tokens[:, :, 0]
                )
            else:
                code_emb = [
                    self.post_model.emb_code[i](input_tokens[:, :, i])
                    for i in range(self.post_model.num_vq)
                ]
                input_emb = torch.stack(code_emb, 3).sum(3)
                start_idx = (
                    input_tokens_history.shape[-2] - 1
                    if input_tokens_history.shape[-2] > 0
                    else 0
                )
        else:
            input_emb = self.post_model(input_tokens, text_mask)
        # print(input_emb.shape)
        hidden_states = model_executable(
            input_emb=input_emb,
            positions=input_positions,
            kv_caches=kv_caches,
            input_metadata=input_metadata,
        )
        # print(hidden_states.shape)
        # print(input_tokens)
        B_NO_PAD = input_tokens_history.shape[0]
        input_tokens = input_tokens[:B_NO_PAD, :, :]
        hidden_states = hidden_states[:B_NO_PAD, :, :]
        idx_next, logprob, finish = self.sampler.sample(
            inputs_ids=(
                input_tokens
                if input_tokens_history.shape[-2] == 0
                else input_tokens_history
            ),
            hidden_states=hidden_states,
            infer_text=infer_text,
            temperature=temperture,
            logits_processors=logits_processors,
            logits_warpers=logits_warpers,
            min_new_token=min_new_token,
            now_length=1,
            eos_token=eos_token,
            start_idx=start_idx,
        )
        # print(logprob.shape, idx_next.shape)
        if len(logprob.shape) == 2:
            logprob = logprob[:, None, :]
        logprob = torch.gather(logprob, -1, idx_next.transpose(-1, -2))[:, :, 0]
        # print("测试",idx_next.shape, logprob.shape)
        # Sample the next token.
        # output = self.model.sample(
        #     hidden_states=hidden_states,
        #     sampling_metadata=sampling_metadata,
        # )
        results = []
        for i in range(idx_next.shape[0]):
            idx_next_i = idx_next[i, 0, :].tolist()
            logprob_i = logprob[i].tolist()
            tmp_hidden_states = hidden_states[i]
            if input_tokens[i].shape[-2] != 1:
                tmp_hidden_states = tmp_hidden_states[-1:, :]
            result = SequenceGroupOutput(
                samples=[
                    SequenceOutput(
                        parent_seq_id=seq_groups[i],
                        logprobs={tuple(idx_next_i): logprob_i},
                        output_token=tuple(idx_next_i),
                        hidden_states=tmp_hidden_states,
                        finished=finish[i].item(),
                    ),
                ],
                prompt_logprobs=None,
            )
            results.append(result)
        # print(results)
        # print(idx_next, idx_next.shape, logprob.shape)
        return results

    @torch.inference_mode()
    def profile_run(self) -> None:
        # Enable top-k sampling to reflect the accurate memory usage.
        vocab_size = self.model_config.get_vocab_size()
        sampling_params = SamplingParams(
            top_p=0.99, top_k=vocab_size - 1, infer_text=True
        )
        max_num_batched_tokens = self.scheduler_config.max_num_batched_tokens
        max_num_seqs = self.scheduler_config.max_num_seqs

        # Profile memory usage with max_num_sequences sequences and the total
        # number of tokens equal to max_num_batched_tokens.
        seqs: List[SequenceGroupMetadata] = []
        for group_id in range(max_num_seqs):
            seq_len = max_num_batched_tokens // max_num_seqs + (
                group_id < max_num_batched_tokens % max_num_seqs
            )
            seq_data = SequenceData([0] * seq_len)
            seq = SequenceGroupMetadata(
                request_id=str(group_id),
                is_prompt=True,
                seq_data={group_id: seq_data},
                sampling_params=sampling_params,
                block_tables=None,
            )
            seqs.append(seq)

        # Run the model with the dummy inputs.
        num_layers = self.model_config.get_num_layers(self.parallel_config)
        kv_caches = [(None, None)] * num_layers
        self.execute_model(seqs, kv_caches)
        torch.cuda.synchronize()
        return

    @torch.inference_mode()
    def capture_model(self, kv_caches: List[KVCache]) -> None:
        assert not self.model_config.enforce_eager
        logger.info(
            "Capturing the model for CUDA graphs. This may lead to "
            "unexpected consequences if the model is not static. To "
            "run the model in eager mode, set 'enforce_eager=True' or "
            "use '--enforce-eager' in the CLI."
        )
        logger.info(
            "CUDA graphs can take additional 1~3 GiB memory per GPU. "
            "If you are running out of memory, consider decreasing "
            "`gpu_memory_utilization` or enforcing eager mode."
        )
        start_time = time.perf_counter()

        # Prepare dummy inputs. These will be reused for all batch sizes.
        max_batch_size = max(_BATCH_SIZES_TO_CAPTURE)
        input_emb = torch.zeros(
            max_batch_size,
            1,
            self.model_config.get_hidden_size(),
            dtype=next(self.model.parameters()).dtype,
        ).cuda()
        input_positions = torch.zeros(max_batch_size, 1, dtype=torch.long).cuda()
        slot_mapping = torch.empty(max_batch_size, 1, dtype=torch.long).cuda()
        slot_mapping.fill_(_PAD_SLOT_ID)
        context_lens = torch.ones(max_batch_size, dtype=torch.int32).cuda()
        block_tables = torch.from_numpy(self.graph_block_tables).cuda()

        # NOTE: Capturing the largest batch size first may help reduce the
        # memory usage of CUDA graph.
        for batch_size in reversed(_BATCH_SIZES_TO_CAPTURE):
            # Create dummy input_metadata.
            input_metadata = InputMetadata(
                is_prompt=False,
                slot_mapping=slot_mapping[:batch_size],
                max_context_len=self.max_context_len_to_capture,
                context_lens=context_lens[:batch_size],
                block_tables=block_tables[:batch_size],
                use_cuda_graph=True,
            )

            graph_runner = CUDAGraphRunner(self.model)
            graph_runner.capture(
                input_emb[:batch_size],
                input_positions[:batch_size],
                kv_caches,
                input_metadata,
                memory_pool=self.graph_memory_pool,
            )
            self.graph_memory_pool = graph_runner.graph.pool()
            self.graph_runners[batch_size] = graph_runner

        end_time = time.perf_counter()
        elapsed_time = end_time - start_time
        # This usually takes < 10 seconds.
        logger.info(f"Graph capturing finished in {elapsed_time:.0f} secs.")


class CUDAGraphRunner:

    def __init__(self, model: nn.Module):
        self.model = model
        self.graph = None
        self.input_buffers: Dict[str, torch.Tensor] = {}
        self.output_buffers: Dict[str, torch.Tensor] = {}

    def capture(
        self,
        input_emb: torch.Tensor,
        positions: torch.Tensor,
        kv_caches: List[KVCache],
        input_metadata: InputMetadata,
        memory_pool,
    ) -> None:
        assert self.graph is None
        # Run the model once without capturing the graph.
        # This is to make sure that the captured graph does not include the
        # kernel launches for initial benchmarking (e.g., Triton autotune).
        self.model(
            input_emb,
            positions,
            kv_caches,
            input_metadata,
        )
        torch.cuda.synchronize()

        # Capture the graph.
        self.graph = torch.cuda.CUDAGraph()
        with torch.cuda.graph(self.graph, pool=memory_pool):
            hidden_states = self.model(
                input_emb,
                positions,
                kv_caches,
                input_metadata,
            )
        torch.cuda.synchronize()

        # Save the input and output buffers.
        self.input_buffers = {
            "input_emb": input_emb,
            "positions": positions,
            "kv_caches": kv_caches,
            "slot_mapping": input_metadata.slot_mapping,
            "context_lens": input_metadata.context_lens,
            "block_tables": input_metadata.block_tables,
        }
        self.output_buffers = {"hidden_states": hidden_states}
        return

    def forward(
        self,
        input_emb: torch.Tensor,
        positions: torch.Tensor,
        kv_caches: List[Tuple[torch.Tensor, torch.Tensor]],
        input_metadata: InputMetadata,
    ) -> torch.Tensor:
        # KV caches are fixed tensors, so we don't need to copy them.
        del kv_caches

        # Copy the input tensors to the input buffers.
        self.input_buffers["input_emb"].copy_(input_emb, non_blocking=True)
        self.input_buffers["positions"].copy_(positions, non_blocking=True)
        self.input_buffers["slot_mapping"].copy_(
            input_metadata.slot_mapping, non_blocking=True
        )
        self.input_buffers["context_lens"].copy_(
            input_metadata.context_lens, non_blocking=True
        )
        self.input_buffers["block_tables"].copy_(
            input_metadata.block_tables, non_blocking=True
        )

        # Run the graph.
        self.graph.replay()

        # Return the output tensor.
        return self.output_buffers["hidden_states"]

    def __call__(self, *args, **kwargs):
        return self.forward(*args, **kwargs)


def _pad_to_max(x: List[int], max_len: int, pad: int) -> List[int]:
    assert len(x) <= max_len
    if len(x) == max_len:
        return list(x)
    return list(x) + [pad] * (max_len - len(x))


def _make_tensor_with_pad(
    x: List[List[int]],
    max_len: int,
    pad: int,
    dtype: torch.dtype,
    device: Union[str, torch.device] = "cuda",
    pin_memory: bool = False,
) -> torch.Tensor:
    padded_x = []
    for x_i in x:
        pad_i = pad
        if isinstance(x[0][0], tuple):
            pad_i = (0,) * len(x[0][0])
        padded_x.append(_pad_to_max(x_i, max_len, pad_i))

    return torch.tensor(
        padded_x,
        dtype=dtype,
        device=device,
        pin_memory=pin_memory and str(device) == "cpu",
    )


def _get_graph_batch_size(batch_size: int) -> int:
    if batch_size <= 2:
        return batch_size
    elif batch_size <= 4:
        return 4
    else:
        return (batch_size + 7) // 8 * 8


def _async_h2d(data: list, dtype, pin_memory):
    t = torch.tensor(data, dtype=dtype, pin_memory=pin_memory)
    return t.to(device="cuda", non_blocking=True)