EngineX/xc-llm-ascend
3
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

[long_seq] remove long_seq env (#4660)

Browse Source 
### What this PR does / why we need it?
remove env VLLM_ASCEND_ENABLE_CONTEXT_PARALLEL 

- vLLM version: v0.12.0

---------

Signed-off-by: LookAround <lixushi@huawei.com>
Signed-off-by: ZhangMingWei716 <2894054457@qq.com>
Co-authored-by: ZhangMingWei716 <2894054457@qq.com>
Co-authored-by: wangxiyuan <wangxiyuan1007@gmail.com>
This commit is contained in:
LookAround0301
2025-12-05 10:31:49 +08:00
committed by GitHub
parent ea54388e19
commit b32ef53b3b
16 changed files with 230 additions and 176 deletions
Download Patch File Download Diff File Expand all files Collapse all files

17
vllm_ascend/worker/worker_v1.py
View File

@@ -53,7 +53,6 @@ from vllm_ascend.distributed.parallel_state import init_ascend_model_parallel
from vllm_ascend.ops.triton.triton_utils import init_device_properties_triton
from vllm_ascend.platform import NPUPlatform
from vllm_ascend.utils import (check_ascend_device_type, is_enable_nz,
prefill_context_parallel_enable,
register_ascend_customop, sleep_mode_enabled,
try_register_lib)
from vllm_ascend.worker.model_runner_v1 import NPUModelRunner
@@ -405,17 +404,11 @@ class NPUWorker(WorkerBase):
init_distributed_environment(self.parallel_config.world_size,
self.rank, self.distributed_init_method,
self.local_rank, "hccl")
if prefill_context_parallel_enable():
ensure_model_parallel_initialized(
self.parallel_config.tensor_parallel_size,
self.parallel_config.pipeline_parallel_size,
self.parallel_config.prefill_context_parallel_size,
self.parallel_config.decode_context_parallel_size)
else:
ensure_model_parallel_initialized(
self.parallel_config.tensor_parallel_size,
self.parallel_config.pipeline_parallel_size,
self.parallel_config.decode_context_parallel_size)
ensure_model_parallel_initialized(
self.parallel_config.tensor_parallel_size,
self.parallel_config.pipeline_parallel_size,
self.parallel_config.prefill_context_parallel_size,
self.parallel_config.decode_context_parallel_size)
init_ascend_model_parallel(self.parallel_config)
ensure_kv_transfer_initialized(self.vllm_config)
ensure_ec_transfer_initialized(self.vllm_config)