e7e1a7dc05
### What this PR does / why we need it?
#5051 only implement a basic framework for model runner v2, but there
are still some bugs for e2e functionality, this PR aim to enable basic
functionality.
model runner v2 plans:
https://github.com/vllm-project/vllm-ascend/issues/5208
- vLLM version: release/v0.13.0
- vLLM main:
ad32e3e19c
---------
Signed-off-by: Ronald1995 <ronaldautomobile@163.com>
378 lines
16 KiB
Python
378 lines
16 KiB
Python
# Adapt from https://github.com/vllm-project/vllm/blob/main/vllm/v1/worker/gpu/model_runner.py
|
|
# SPDX-License-Identifier: Apache-2.0
|
|
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
|
|
# Copyright (c) 2025 Huawei Technologies Co., Ltd. All Rights Reserved.
|
|
#
|
|
# 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.
|
|
#
|
|
|
|
import numpy as np
|
|
import torch
|
|
from vllm.config import VllmConfig
|
|
from vllm.logger import init_logger
|
|
from vllm.v1.core.sched.output import GrammarOutput, SchedulerOutput
|
|
from vllm.v1.worker.gpu.input_batch import (InputBatch,
|
|
combine_sampled_and_draft_tokens,
|
|
prepare_pos_seq_lens,
|
|
prepare_prefill_inputs)
|
|
from vllm.v1.worker.gpu.model_runner import GPUModelRunner
|
|
from vllm.v1.worker.gpu.sample.metadata import SamplingMetadata
|
|
from vllm.v1.worker.gpu.sample.output import SamplerOutput
|
|
|
|
from vllm_ascend.worker.v2.aclgraph_utils import AclGraphManager
|
|
from vllm_ascend.worker.v2.attn_utils import (build_attn_metadata,
|
|
build_attn_state,
|
|
make_attention_mask)
|
|
from vllm_ascend.worker.v2.input_batch import AscendInputBuffers
|
|
from vllm_ascend.worker.v2.sample.sampler import AscendSampler
|
|
from vllm_ascend.worker.v2.states import AscendRequestState, uva_wrapper
|
|
from vllm_ascend.worker.v2.utils import torch_cuda_wrapper
|
|
|
|
logger = init_logger(__name__)
|
|
|
|
|
|
class NPUModelRunner(GPUModelRunner):
|
|
"""Model runner for Ascend NPUs."""
|
|
|
|
def __init__(self, vllm_config: VllmConfig, device: torch.device):
|
|
with (torch_cuda_wrapper(), uva_wrapper()):
|
|
super().__init__(vllm_config, device)
|
|
|
|
# because we will override these attribute, delete these attribute to
|
|
# make sure it's collected by python gc immediately.
|
|
del self.cudagraph_manager
|
|
del self.req_states
|
|
del self.input_buffers
|
|
del self.sampler
|
|
|
|
# NPU specific initializations can be added below.
|
|
self.cudagraph_manager: AclGraphManager = AclGraphManager(
|
|
vllm_config,
|
|
device,
|
|
)
|
|
# AscendRequestState has extra `num_computed_tokens_cpu` attribute.
|
|
# so reinitialize req_states here.
|
|
self.req_states: AscendRequestState = AscendRequestState(
|
|
max_num_reqs=self.max_num_reqs,
|
|
max_model_len=self.max_model_len,
|
|
max_num_batched_tokens=self.max_num_tokens,
|
|
num_speculative_steps=self.num_speculative_steps,
|
|
vocab_size=self.vocab_size,
|
|
device=self.device,
|
|
pin_memory=self.pin_memory,
|
|
)
|
|
# AscendInputBuffers has extra `seq_lens_cpu` attribute.
|
|
# so reinitialize input_buffers here.
|
|
self.input_buffers: AscendInputBuffers = AscendInputBuffers(
|
|
max_num_reqs=self.max_num_reqs,
|
|
max_num_tokens=self.max_num_tokens,
|
|
inputs_embeds_size=self.inputs_embeds_size,
|
|
vocab_size=self.vocab_size,
|
|
dtype=self.dtype,
|
|
device=self.device,
|
|
pin_memory=self.pin_memory,
|
|
)
|
|
# we need to adjust triton operators in sampler,
|
|
# so reinitialize sampler here.
|
|
self.sampler: AscendSampler = AscendSampler(
|
|
logprobs_mode=self.model_config.logprobs_mode, )
|
|
|
|
# actual seq lengths for query (used in attention backends).
|
|
self.actual_seq_lengths_q: list[int] = []
|
|
# decode token per request (used in attention backends).
|
|
self.decode_token_per_req = 1
|
|
|
|
# there attributes are for async scheduling with speculative decoding.
|
|
# because npu attention backend still need to use seq_lens_cpu,
|
|
# we need to copy num_rejected_tokens back to cpu to help
|
|
# update actual seq_lens_cpu. gpu attention backend do not need these
|
|
# attributes, cause their attention backends do not use seq_lens_cpu.
|
|
# and seq_lens_cpu is deprecated in gpu_model_runner_v2.
|
|
self.num_rejected_tokens_event = None
|
|
self.num_rejectd_tokens_cpu = None
|
|
self.num_rejected_token_stream = None
|
|
if self.use_async_scheduling and self.do_spec_decode:
|
|
self.num_rejected_tokens_event = torch.npu.Event()
|
|
self.num_rejected_token_stream = torch.npu.Stream()
|
|
self.num_rejectd_tokens_cpu = torch.empty(
|
|
self.max_num_reqs,
|
|
dtype=torch.int64,
|
|
device="cpu",
|
|
pin_memory=self.pin_memory,
|
|
)
|
|
|
|
def prepare_inputs(
|
|
self,
|
|
scheduler_output: SchedulerOutput,
|
|
num_tokens_after_padding: int,
|
|
) -> InputBatch:
|
|
"""Override GPUModelRunner.prepare_inputs for Ascend NPUs.
|
|
npu attention backends need seq_lens_cpu to work.
|
|
so we need to prepare seq_lens_cpu here.
|
|
"""
|
|
num_tokens = scheduler_output.total_num_scheduled_tokens
|
|
assert num_tokens > 0
|
|
num_reqs = len(scheduler_output.num_scheduled_tokens)
|
|
|
|
# Decode first, then prefill.
|
|
# batch_idx -> req_id
|
|
req_ids = sorted(
|
|
scheduler_output.num_scheduled_tokens.keys(),
|
|
key=lambda k: scheduler_output.num_scheduled_tokens[k],
|
|
)
|
|
|
|
self._update_seq_lens_cpu(scheduler_output, req_ids)
|
|
|
|
num_scheduled_tokens = np.array(
|
|
[scheduler_output.num_scheduled_tokens[i] for i in req_ids],
|
|
dtype=np.int32)
|
|
num_valid_tokens = num_scheduled_tokens
|
|
if scheduler_output.scheduled_spec_decode_tokens:
|
|
num_valid_tokens = np.array(
|
|
[
|
|
num_tokens - len(
|
|
scheduler_output.scheduled_spec_decode_tokens.get(
|
|
i, []))
|
|
for num_tokens, i in zip(num_scheduled_tokens, req_ids)
|
|
],
|
|
dtype=np.int32,
|
|
)
|
|
attn_state = build_attn_state(
|
|
self.vllm_config,
|
|
self.input_buffers.seq_lens_np,
|
|
num_reqs,
|
|
num_scheduled_tokens,
|
|
num_valid_tokens,
|
|
)
|
|
attn_mask = make_attention_mask(
|
|
self.vllm_config,
|
|
attn_state,
|
|
self.dtype,
|
|
self.device,
|
|
)
|
|
|
|
idx_mapping_list = [
|
|
self.req_states.req_id_to_index[req_id] for req_id in req_ids
|
|
]
|
|
idx_mapping = self.input_buffers.idx_mapping
|
|
idx_mapping.np[:num_reqs] = idx_mapping_list
|
|
idx_mapping_np = idx_mapping.np[:num_reqs]
|
|
# add `idx_mapping_cpu` here, because vllm-ascend's self.req_states.
|
|
# num_computed_tokens_cpu is actually cpu's tensor, while it's a gpu's
|
|
# tensor in vllm gpu_model_runner_v2.
|
|
idx_mapping_cpu = idx_mapping.cpu[:num_reqs]
|
|
idx_mapping_npu = idx_mapping.copy_to_gpu(num_reqs)
|
|
|
|
# Get the number of draft tokens for each request.
|
|
if not scheduler_output.scheduled_spec_decode_tokens:
|
|
# No draft token scheduled (common case).
|
|
total_num_draft_tokens = 0
|
|
total_num_logits = num_reqs
|
|
cu_num_logits = torch.arange(num_reqs + 1,
|
|
device=self.device,
|
|
dtype=torch.int32)
|
|
else:
|
|
draft_tokens = scheduler_output.scheduled_spec_decode_tokens
|
|
num_draft_tokens = np.array(
|
|
[
|
|
len(draft_tokens[req_id]) if req_id in draft_tokens else 0
|
|
for req_id in req_ids
|
|
],
|
|
dtype=np.int32,
|
|
)
|
|
total_num_draft_tokens = int(num_draft_tokens.sum())
|
|
total_num_logits = num_reqs + total_num_draft_tokens
|
|
|
|
np.cumsum(
|
|
num_draft_tokens + 1,
|
|
out=self.input_buffers.cu_num_logits.np[1:num_reqs + 1],
|
|
)
|
|
cu_num_logits = self.input_buffers.cu_num_logits.copy_to_gpu(
|
|
num_reqs + 1)
|
|
|
|
# Block tables: num_kv_cache_groups x [num_reqs, max_num_blocks]
|
|
block_tables = self.block_tables.gather_block_tables(idx_mapping_npu)
|
|
|
|
# Get query_start_loc.
|
|
np.cumsum(
|
|
num_scheduled_tokens,
|
|
out=self.input_buffers.query_start_loc.np[1:num_reqs + 1],
|
|
)
|
|
# Pad for full CUDA graph mode.
|
|
# Some attention backends like FA3 require query_start_loc to be non-decreasing.
|
|
self.input_buffers.query_start_loc.np[num_reqs + 1:] = num_tokens
|
|
self.input_buffers.query_start_loc.copy_to_gpu()
|
|
query_start_loc_gpu = self.input_buffers.query_start_loc.gpu[:
|
|
num_reqs +
|
|
1]
|
|
query_start_loc_cpu = self.input_buffers.query_start_loc.cpu[:
|
|
num_reqs +
|
|
1]
|
|
query_start_loc_np = self.input_buffers.query_start_loc.np[:num_reqs +
|
|
1]
|
|
|
|
# Get prefill tokens.
|
|
prepare_prefill_inputs(
|
|
self.input_buffers.input_ids,
|
|
self.req_states.next_prefill_tokens,
|
|
idx_mapping_npu,
|
|
query_start_loc_gpu,
|
|
# use prefill_token_ids.copy_to_gpu() because npu doesn't
|
|
# support uva buffer.
|
|
self.req_states.prefill_token_ids.copy_to_gpu(),
|
|
self.req_states.prefill_len.gpu,
|
|
self.req_states.num_computed_tokens,
|
|
)
|
|
|
|
# Prepare positions and seq_lens.
|
|
prepare_pos_seq_lens(
|
|
idx_mapping_npu,
|
|
query_start_loc_gpu,
|
|
self.req_states.num_computed_tokens,
|
|
self.input_buffers.positions,
|
|
self.input_buffers.seq_lens,
|
|
)
|
|
seq_lens = self.input_buffers.seq_lens[:num_reqs]
|
|
|
|
# Some input token ids are directly read from the last sampled tokens
|
|
# and draft tokens. Also, get the logits indices to sample tokens from.
|
|
logits_indices = combine_sampled_and_draft_tokens(
|
|
self.input_buffers.input_ids,
|
|
idx_mapping_npu,
|
|
self.req_states.last_sampled_tokens,
|
|
query_start_loc_gpu,
|
|
seq_lens,
|
|
self.req_states.prefill_len.gpu,
|
|
self.req_states.draft_tokens,
|
|
cu_num_logits,
|
|
total_num_logits,
|
|
)
|
|
|
|
# Compute slot mappings: [num_kv_cache_groups, num_tokens]
|
|
slot_mappings = self.block_tables.compute_slot_mappings(
|
|
query_start_loc_gpu, self.input_buffers.positions[:num_tokens])
|
|
|
|
# Layer name -> attention metadata.
|
|
# TODO(Ronald1995): try to add a new method `build_attn_metadata` in
|
|
# vllm gpu_model_runner_v2, maybe we don't overwrite `prepare_inputs`
|
|
# method like this.
|
|
attn_metadata = build_attn_metadata(
|
|
attn_metadata_builders=self.attn_metadata_builders,
|
|
num_reqs=num_reqs,
|
|
num_tokens=num_tokens,
|
|
query_start_loc_gpu=query_start_loc_gpu,
|
|
query_start_loc_cpu=query_start_loc_cpu,
|
|
seq_lens=self.input_buffers.seq_lens,
|
|
seq_lens_cpu=self.input_buffers.seq_lens_cpu,
|
|
actual_seq_lengths_q=self.actual_seq_lengths_q,
|
|
num_computed_tokens_cpu=self.req_states.
|
|
num_computed_tokens_cpu[idx_mapping_cpu],
|
|
block_tables=block_tables,
|
|
# torch_npu._reshape_and_cache operator requires slot_mappings to
|
|
# be torch.int32.
|
|
slot_mappings=slot_mappings.to(torch.int32),
|
|
kv_cache_config=self.kv_cache_config,
|
|
decode_token_per_req=self.decode_token_per_req,
|
|
attn_mask=attn_mask,
|
|
attn_state=attn_state,
|
|
)
|
|
|
|
input_ids = self.input_buffers.input_ids[:num_tokens_after_padding]
|
|
positions = self.input_buffers.positions[:num_tokens_after_padding]
|
|
return InputBatch(
|
|
req_ids=req_ids,
|
|
num_reqs=num_reqs,
|
|
idx_mapping=idx_mapping_npu,
|
|
idx_mapping_np=idx_mapping_np,
|
|
num_scheduled_tokens=num_scheduled_tokens,
|
|
num_tokens=num_tokens,
|
|
num_tokens_after_padding=num_tokens_after_padding,
|
|
num_draft_tokens=total_num_draft_tokens,
|
|
query_start_loc=query_start_loc_gpu,
|
|
query_start_loc_np=query_start_loc_np,
|
|
seq_lens=seq_lens,
|
|
seq_lens_np=self.input_buffers.seq_lens_np,
|
|
input_ids=input_ids,
|
|
positions=positions,
|
|
attn_metadata=attn_metadata,
|
|
logits_indices=logits_indices,
|
|
cu_num_logits=cu_num_logits,
|
|
)
|
|
|
|
def sample(
|
|
self,
|
|
hidden_states: torch.Tensor,
|
|
input_batch: InputBatch,
|
|
sampling_metadata: SamplingMetadata,
|
|
grammar_output: GrammarOutput | None,
|
|
) -> tuple[SamplerOutput, torch.Tensor, torch.Tensor]:
|
|
"""Override GPUModelRunner.sample for Ascend NPUs.
|
|
when using async scheduling with speculative decoding,
|
|
we need to copy mpu's num_rejected tensor to cpu.
|
|
these operations aren't needed in gpu_model_runner_v2,
|
|
because gpu attention backends do not use seq_lens_cpu anymore.
|
|
"""
|
|
sampler_output, num_sampled, num_rejected = super().sample(
|
|
hidden_states,
|
|
input_batch,
|
|
sampling_metadata,
|
|
grammar_output,
|
|
)
|
|
if self.num_rejected_tokens_event is not None:
|
|
# npu attention backend still need to use seq_lens_cpu,
|
|
# when doing speculative decoding with async_scheduling,
|
|
# we need to copy num_rejected_tokens back to cpu.
|
|
default_stream = torch.cuda.current_stream()
|
|
assert self.num_rejected_token_stream is not None
|
|
assert self.num_rejectd_tokens_cpu is not None
|
|
with torch.npu.stream(self.num_rejected_token_stream):
|
|
self.num_rejected_token_stream.wait_stream(default_stream)
|
|
self.num_rejectd_tokens_cpu.copy_(
|
|
num_rejected,
|
|
non_blocking=True,
|
|
)
|
|
self.num_rejected_tokens_event.record()
|
|
return sampler_output, num_sampled, num_rejected
|
|
|
|
def _update_seq_lens_cpu(
|
|
self,
|
|
scheduler_output: SchedulerOutput,
|
|
req_ids: list[str],
|
|
):
|
|
num_scheduled_tokens = scheduler_output.num_scheduled_tokens
|
|
|
|
# update num_computed_tokens_cpu
|
|
# TODO(Ronald1995): update num_computed_tokens_cpu by considering
|
|
# num_rejectd_tokens.
|
|
for req_id, num_computed_token in zip(
|
|
scheduler_output.scheduled_cached_reqs.req_ids,
|
|
scheduler_output.scheduled_cached_reqs.num_computed_tokens,
|
|
):
|
|
req_index = self.req_states.req_id_to_index[req_id]
|
|
self.req_states.num_computed_tokens_cpu[
|
|
req_index] = num_computed_token
|
|
|
|
# update seq_lens_cpu
|
|
for i, req_id in enumerate(req_ids):
|
|
req_index = self.req_states.req_id_to_index[req_id]
|
|
num_computed_tokens = self.req_states.num_computed_tokens_cpu[
|
|
req_index]
|
|
self.input_buffers.seq_lens_cpu[
|
|
i] = num_computed_tokens + num_scheduled_tokens[req_id]
|
|
|
|
def eplb_warmup(self):
|
|
# TODO(Ronald1995): just define the method in case calling error in
|
|
# worker, implement it in the future.
|
|
pass
|