xc-llm-ascend/vllm_ascend/worker/worker_v1.py

#
# Copyright (c) 2025 Huawei Technologies Co., Ltd. All Rights Reserved.
# Copyright 2023 The vLLM team.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# This file is a part of the vllm-ascend project.
# Adapted from vllm-project/vllm/vllm/worker/gpu_worker.py
#

import copy
from typing import Optional, Union

import torch
import torch.nn as nn
import torch_npu
import vllm.envs as envs_vllm
from torch_npu.op_plugin.atb._atb_ops import _register_atb_extensions
from torch_npu.profiler import dynamic_profile as dp
from vllm.config import VllmConfig
from vllm.distributed import (ensure_model_parallel_initialized,
                              init_distributed_environment)
from vllm.distributed.kv_transfer import ensure_kv_transfer_initialized
from vllm.distributed.parallel_state import get_pp_group, get_tp_group
from vllm.logger import logger
from vllm.lora.request import LoRARequest
from vllm.sequence import IntermediateTensors
from vllm.tasks import SupportedTask
from vllm.utils import STR_DTYPE_TO_TORCH_DTYPE, GiB_bytes
from vllm.v1.core.sched.output import SchedulerOutput
from vllm.v1.kv_cache_interface import KVCacheConfig, KVCacheSpec
from vllm.v1.outputs import (EMPTY_MODEL_RUNNER_OUTPUT, AsyncModelRunnerOutput,
                             DraftTokenIds, ModelRunnerOutput)
from vllm.v1.worker.worker_base import WorkerBase

import vllm_ascend.envs as envs_ascend
from vllm_ascend.ascend_config import get_ascend_config, init_ascend_config
from vllm_ascend.cpu_binding import bind_cpus
from vllm_ascend.device_allocator.camem import CaMemAllocator
from vllm_ascend.distributed.parallel_state import init_ascend_model_parallel
from vllm_ascend.platform import NPUPlatform
from vllm_ascend.utils import (init_ascend_soc_version, is_enable_nz,
                               register_ascend_customop, sleep_mode_enabled,
                               try_register_lib)
from vllm_ascend.worker.model_runner_v1 import NPUModelRunner

torch._dynamo.trace_rules.clear_lru_cache()  # noqa: E402
from torch._dynamo.variables import TorchInGraphFunctionVariable  # noqa: E402

torch_non_c_binding_in_graph_functions_npu = dict.fromkeys(
    ["torch.npu.current_stream"],
    TorchInGraphFunctionVariable,
)  # noqa: E402
torch_non_c_binding_in_graph_functions_npu[
    "torch.npu.stream"] = TorchInGraphFunctionVariable  # noqa: E402
torch._dynamo.trace_rules.torch_name_rule_map.append(
    torch_non_c_binding_in_graph_functions_npu)  # noqa: E402


class NPUWorker(WorkerBase):

    def __init__(
            self,
            vllm_config: VllmConfig,
            local_rank: int,
            rank: int,
            distributed_init_method: str,
            is_driver_worker: bool = False,
            # Additional parameters for compatibility with vllm
            **kwargs):
        """Initialize the worker for Ascend."""
        # register patch for vllm
        from vllm_ascend.utils import adapt_patch
        adapt_patch()
        is_enable_nz(vllm_config=vllm_config)
        # Register ops when worker init.
        from vllm_ascend import ops
        ops.register_dummy_fusion_op()
        _register_atb_extensions()
        register_ascend_customop(vllm_config)
        # init ascend config and soc version
        init_ascend_config(vllm_config)
        init_ascend_soc_version()
        use_sparse = False
        if vllm_config.model_config is not None:
            use_sparse = hasattr(vllm_config.model_config.hf_config,
                                 "index_topk")
        if use_sparse:
            # Direct import instead of using try_register_lib to ensure proper error handling when
            # custom_ops is necessary but not available (e.g., in DeepSeek v3.2 deployments)
            # yapf: disable
            import custom_ops  # type: ignore # noqa

            # yapf: enable
            logger.info(
                "custom_ops module loaded successfully. Custom operators like "
                "torch.ops.custom.npu_sparse_flash_attention are now available."
            )

        super().__init__(vllm_config=vllm_config,
                         local_rank=local_rank,
                         rank=rank,
                         distributed_init_method=distributed_init_method,
                         is_driver_worker=is_driver_worker)

        # binding cpu
        if get_ascend_config().enable_cpu_binding:
            try:
                bind_cpus(self.local_rank, ratio=1.0)
            except RuntimeError as e:
                logger.error(f"{e} in {self.local_rank}")
            except ValueError as e:
                logger.error(f"{e} in {self.local_rank}")
            except Exception:
                logger.info("Skip binding cpu.")

        # Try to import mindie_turbo to accelerate vLLM inference.
        try_register_lib(
            "mindie_turbo",
            "MindIE Turbo is installed. vLLM inference will be accelerated with MindIE Turbo."
        )
        if self.cache_config.cache_dtype == "auto":
            self.cache_dtype = self.model_config.dtype
        else:
            self.cache_dtype = STR_DTYPE_TO_TORCH_DTYPE[
                self.cache_config.cache_dtype]

        if self.model_config.trust_remote_code:
            # note: lazy import to avoid importing torch before initializing
            from vllm.utils import init_cached_hf_modules
            init_cached_hf_modules()

        self.profiler = self._init_profiler()
        if sleep_mode_enabled():
            # Buffers saved before sleep
            self._sleep_saved_buffers: dict[str, torch.Tensor] = {}

        # FixMe: this is a patch to fix the issue cause by https://github.com/vllm-project/vllm/commit/de94289a98d7ec52a5ef02719e01a1db8b505170
        from vllm.model_executor.layers.linear import \
            WEIGHT_LOADER_V2_SUPPORTED
        if "UnquantizedLinearMethod" in WEIGHT_LOADER_V2_SUPPORTED:
            WEIGHT_LOADER_V2_SUPPORTED.remove("UnquantizedLinearMethod")

    def sleep(self, level: int = 1) -> None:
        if not sleep_mode_enabled():
            raise ValueError(
                "Sleep mode is not enabled. Please compile vllm-ascend with COMPILE_CUSTOM_KERNELS=1."
            )
        free_bytes_before_sleep = NPUPlatform.mem_get_info()[0]
        # Save the buffers before level 2 sleep
        if level == 2:
            model = self.model_runner.model
            self._sleep_saved_buffers = {
                name: buffer.cpu().clone()
                for name, buffer in model.named_buffers()
            }
        allocator = CaMemAllocator.get_instance()
        allocator.sleep(offload_tags=("weights", ) if level == 1 else tuple())
        free_bytes_after_sleep, total = NPUPlatform.mem_get_info()
        freed_bytes = free_bytes_after_sleep - free_bytes_before_sleep
        used_bytes = total - free_bytes_after_sleep
        assert freed_bytes >= 0, "Memory usage increased after sleeping."
        logger.info(
            "Sleep mode freed %.2f GiB memory, "
            "%.2f GiB memory is still in use.", freed_bytes / GiB_bytes,
            used_bytes / GiB_bytes)

    def wake_up(self, tags: Optional[list[str]] = None) -> None:
        if not sleep_mode_enabled():
            raise ValueError(
                "Sleep mode is not enabled. Please compile vllm-ascend with COMPILE_CUSTOM_KERNELS=1."
            )

        if is_enable_nz():
            raise ValueError(
                "FRACTAL_NZ mode is enabled. This may cause model parameter precision issues "
                "in the RL scenarios. Please set VLLM_ASCEND_ENABLE_NZ=0.")
        allocator = CaMemAllocator.get_instance()
        allocator.wake_up(tags=tags)

        # Restore the buffers after level 2 sleep
        if len(self._sleep_saved_buffers):
            model = self.model_runner.model
            for name, buffer in model.named_buffers():
                if name in self._sleep_saved_buffers:
                    buffer.data.copy_(self._sleep_saved_buffers[name].data)
            self._sleep_saved_buffers = {}

    def initialize_cache(self, num_gpu_blocks: int,
                         num_cpu_blocks: int) -> None:
        self.cache_config.num_gpu_blocks = num_gpu_blocks
        self.cache_config.num_cpu_blocks = num_cpu_blocks

    def _init_device(self):
        device = torch.device(f"npu:{self.local_rank}")
        NPUPlatform.set_device(device)
        NPUPlatform.empty_cache()
        self.init_npu_memory = NPUPlatform.mem_get_info()[0]
        # Initialize the distributed environment.
        self._init_worker_distributed_environment()
        # Set random seed.
        NPUPlatform.seed_everything(self.model_config.seed)
        return device

    def init_device(self):
        device = self._init_device()
        # Init ModelRunner here, so that we have access to self.device.
        self.model_runner = NPUModelRunner(self.vllm_config, device)

    def determine_available_memory(self) -> int:
        # Profile the memory usage of the model and get the maximum number of
        # cache blocks that can be allocated with the remaining free memory.
        NPUPlatform.clear_npu_memory()

        # Execute a forward pass with dummy inputs to profile the memory usage
        # of the model.
        _, total_npu_memory = NPUPlatform.mem_get_info()
        self.model_runner.profile_run()

        # Calculate the number of blocks that can be allocated with the
        # profiled peak memory.
        free_npu_memory, _ = NPUPlatform.mem_get_info()
        # NOTE(woosuk): Here we assume that the other processes using the same
        # GPU did not change their memory usage during the profiling.
        assert self.init_npu_memory > free_npu_memory, (
            "Error in memory profiling. "
            f"Initial free memory {self.init_npu_memory}, current free memory"
            f" {free_npu_memory}. This happens when the NPU memory was "
            "not properly cleaned up before initializing the vLLM instance.")

        # Get the peak memory allocation recorded by torch
        peak_memory = torch_npu.npu.memory_stats()["allocated_bytes.all.peak"]
        # TODO: don`t need impl this func after empty_cache in
        # Worker.determine_num_available_blocks() unified`
        NPUPlatform.empty_cache()
        torch_allocated_bytes = torch_npu.npu.memory_stats(
        )["allocated_bytes.all.current"]
        total_allocated_bytes = torch_npu.npu.mem_get_info(
        )[1] - torch_npu.npu.mem_get_info()[0]
        non_torch_allocations = total_allocated_bytes - torch_allocated_bytes
        if non_torch_allocations > 0:
            peak_memory += non_torch_allocations
        available_kv_cache_memory = int(
            total_npu_memory * self.cache_config.gpu_memory_utilization -
            peak_memory)
        available_kv_cache_memory = int(max(available_kv_cache_memory, 0))
        logger.info(
            f"Available memory: {available_kv_cache_memory}, total memory: {total_npu_memory}"
        )
        return available_kv_cache_memory

    def determine_available_memory_idle_offload_mode(self) -> int:
        allocator = CaMemAllocator.get_instance()
        free, total = allocator.get_pool_mem_info()
        available_kv_cache_memory = int(
            total * self.cache_config.gpu_memory_utilization - (total - free))
        available_kv_cache_memory = int(max(available_kv_cache_memory, 0))
        logger.info(
            f"Available memory (idle offload mode): {available_kv_cache_memory}, total memory: {total}"
        )
        return available_kv_cache_memory

    def execute_model(
        self,
        scheduler_output: "SchedulerOutput",
    ) -> Optional[Union[ModelRunnerOutput, AsyncModelRunnerOutput]]:
        # enable msMonitor to monitor the performance of vllm-ascend
        if envs_ascend.MSMONITOR_USE_DAEMON:
            dp.step()

        intermediate_tensors = None
        forward_pass = scheduler_output.total_num_scheduled_tokens > 0
        if forward_pass and not get_pp_group().is_first_rank:
            intermediate_tensors = IntermediateTensors(
                get_pp_group().recv_tensor_dict(
                    all_gather_group=get_tp_group()))

        output = self.model_runner.execute_model(scheduler_output,
                                                 intermediate_tensors)
        if isinstance(output, (ModelRunnerOutput, AsyncModelRunnerOutput)):
            return output

        assert isinstance(output, IntermediateTensors)
        parallel_config = self.vllm_config.parallel_config
        assert parallel_config.distributed_executor_backend != (
            "external_launcher") and not get_pp_group().is_last_rank

        get_pp_group().send_tensor_dict(output.tensors,
                                        all_gather_group=get_tp_group())

        kv_connector_output = output.kv_connector_output
        if not kv_connector_output:
            return None

        # In case of PP with kv transfer, we need to pass through the
        # kv_connector_output
        if (not kv_connector_output.finished_sending
                and not kv_connector_output.finished_recving):
            return EMPTY_MODEL_RUNNER_OUTPUT
        output = copy.copy(EMPTY_MODEL_RUNNER_OUTPUT)
        output.kv_connector_output = kv_connector_output
        return output

    def load_model(self) -> None:
        if self.vllm_config.model_config.enable_sleep_mode:
            allocator = CaMemAllocator.get_instance()
            assert allocator.get_current_usage() == 0, (
                "Sleep mode can only be "
                "used for one instance per process.")
            context = allocator.use_memory_pool(tag="weights")
        elif envs_ascend.VLLM_ASCEND_ENABLE_IDLE_OFFLOAD:
            if not sleep_mode_enabled():
                raise ValueError(
                    "Sleep mode is not enabled. Please compile vllm-ascend with COMPILE_CUSTOM_KERNELS=1."
                )
            if is_enable_nz():
                raise ValueError(
                    "FRACTAL_NZ mode is enabled. This may cause model parameter precision issues "
                    "in the RL scenarios. Please set VLLM_ASCEND_ENABLE_NZ=0.")
            allocator = CaMemAllocator.get_instance()
            assert allocator.get_current_usage() == 0, (
                "Idle offload mode can only be "
                "used for one instance per process.")
            context = allocator.use_memory_pool(tag="weights")
        else:
            from contextlib import nullcontext
            context = nullcontext()  # type: ignore
        with context:
            self.model_runner.load_model()
            if envs_ascend.VLLM_ASCEND_ENABLE_IDLE_OFFLOAD:
                # save memory to host with lock
                self.offload_vram()
                succ, _ = self.try_reload_vram()
                assert succ, "Failed to reload model weights after offloading."

    def offload_vram(self) -> None:
        # free_bytes_before_offload = NPUPlatform.mem_get_info()[0]
        allocator = CaMemAllocator.get_instance()
        allocator.offload_vram(offload_tags=("weights",))
        # free_bytes_after_offload, total = NPUPlatform.mem_get_info()
        # freed_bytes = free_bytes_after_offload - free_bytes_before_offload
        # used_bytes = total - free_bytes_after_offload
        # assert freed_bytes >= 0, "Memory usage increased after sleeping."
        # logger.info(
        #     "Offloading freed %.2f GiB memory, "
        #     "%.2f GiB memory is still in use.", freed_bytes / GiB_bytes,
        #     used_bytes / GiB_bytes)

    def try_reload_vram(self) -> tuple[bool, bool]:
        allocator = CaMemAllocator.get_instance()
        return allocator.try_reload_vram(tags=None)

    def vnpu_unlock_gpu(self) -> None:
        allocator = CaMemAllocator.get_instance()
        allocator.vnpu_unlock_gpu()

    def compile_or_warm_up_model(self) -> None:
        # Note: need to adapt for graph mode.
        self.model_runner.eplb_warmup()
        warmup_sizes = (self.vllm_config.compilation_config.compile_sizes
                        or []).copy()
        if not self.model_config.enforce_eager:
            warmup_sizes = [
                x for x in warmup_sizes if x not in
                self.vllm_config.compilation_config.cudagraph_capture_sizes
            ]
        for size in sorted(warmup_sizes, reverse=True):
            logger.info("Compile and warming up model for size %d", size)
            self.model_runner._dummy_run(size)
        if not self.model_config.enforce_eager:
            self.model_runner.capture_model()
        # Call ATB matmul to warm up; otherwise, the first operation (ReshapeAndCache)
        # may cause performance degradation at runtime.
        self._warm_up_atb()
        # Reset the seed to ensure that the random state is not affected by
        # the model initialization and profiling.
        NPUPlatform.seed_everything(self.model_config.seed)

    def _warm_up_atb(self):
        x = torch.rand((2, 4), dtype=torch.float16).npu()
        weight = torch.rand((2, 4), dtype=torch.float16).npu()
        c = torch.rand((4, 4), dtype=torch.float32).npu()
        torch_npu._npu_matmul_add_fp32(x, weight, c)

    def get_model(self) -> nn.Module:
        return self.model_runner.get_model()

    def get_kv_cache_spec(self) -> dict[str, KVCacheSpec]:
        return self.model_runner.get_kv_cache_spec()

    def initialize_from_config(self, kv_cache_config: KVCacheConfig) -> None:
        """Allocate NPU KV cache with the specified kv_cache_config."""
        if self.vllm_config.model_config.enable_sleep_mode:
            allocator = CaMemAllocator.get_instance()
            context = allocator.use_memory_pool(tag="kv_cache")
        elif envs_ascend.VLLM_ASCEND_ENABLE_IDLE_OFFLOAD:
            allocator = CaMemAllocator.get_instance()
            context = allocator.use_memory_pool(tag="kv_cache")
        else:
            from contextlib import nullcontext
            context = nullcontext()  # type: ignore
        with context:
            self.model_runner.initialize_kv_cache(kv_cache_config)

    def profile(self, is_start: bool = True):
        if self.profiler is None:
            raise RuntimeError("Profiler is not enabled.")
        if is_start:
            self.profiler.start()
        else:
            self.profiler.stop()

    def add_lora(self, lora_request: LoRARequest) -> bool:
        return self.model_runner.add_lora(lora_request)

    def remove_lora(self, lora_id: int) -> bool:
        return self.model_runner.remove_lora(lora_id)

    def list_loras(self) -> set[int]:
        return self.model_runner.list_loras()

    def pin_lora(self, lora_id: int) -> bool:
        return self.model_runner.pin_lora(lora_id)

    def execute_dummy_batch(self) -> None:
        self.model_runner._dummy_run(
            num_tokens=self.model_runner.decode_token_per_req,
            uniform_decode=True)

    def _init_worker_distributed_environment(self) -> None:
        """Initialize the distributed environment."""
        init_distributed_environment(self.parallel_config.world_size,
                                     self.rank, self.distributed_init_method,
                                     self.local_rank, "hccl")
        ensure_model_parallel_initialized(
            self.parallel_config.tensor_parallel_size,
            self.parallel_config.pipeline_parallel_size)
        init_ascend_model_parallel(self.parallel_config)
        ensure_kv_transfer_initialized(self.vllm_config)

    def _init_profiler(self):
        # Torch profiler. Enabled and configured through env vars:
        # VLLM_TORCH_PROFILER_DIR=/path/to/save/trace
        if envs_vllm.VLLM_TORCH_PROFILER_DIR:
            if envs_ascend.MSMONITOR_USE_DAEMON:
                raise RuntimeError(
                    "MSMONITOR_USE_DAEMON and VLLM_TORCH_PROFILER_DIR cannot be both set at the same time."
                )
            torch_profiler_trace_dir = envs_vllm.VLLM_TORCH_PROFILER_DIR
            logger.info("Profiling enabled. Traces will be saved to: %s",
                        torch_profiler_trace_dir)

            experimental_config = torch_npu.profiler._ExperimentalConfig(
                export_type=torch_npu.profiler.ExportType.Text,
                profiler_level=torch_npu.profiler.ProfilerLevel.Level1,
                msprof_tx=False,
                aic_metrics=torch_npu.profiler.AiCMetrics.AiCoreNone,
                l2_cache=False,
                op_attr=False,
                data_simplification=False,
                record_op_args=False,
                gc_detect_threshold=None,
            )

            return torch_npu.profiler.profile(
                activities=[
                    torch_npu.profiler.ProfilerActivity.CPU,
                    torch_npu.profiler.ProfilerActivity.NPU,
                ],
                with_stack=envs_vllm.VLLM_TORCH_PROFILER_WITH_STACK,
                profile_memory=envs_vllm.\
                    VLLM_TORCH_PROFILER_WITH_PROFILE_MEMORY,
                with_modules=False,
                experimental_config=experimental_config,
                on_trace_ready=torch_npu.profiler.tensorboard_trace_handler(
                    torch_profiler_trace_dir))
        else:
            return None

    def get_supported_pooling_tasks(self):
        return self.model_runner.get_supported_pooling_tasks()

    def get_supported_tasks(self) -> "tuple[SupportedTask, ...]":
        return self.model_runner.get_supported_tasks()

    def take_draft_token_ids(self) -> Optional[DraftTokenIds]:
        return self.model_runner.take_draft_token_ids()