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

[Lint]Style: Convert vllm-ascend/ to ruff format(Batch #6) (#6001)

Browse Source 
### What this PR does / why we need it?
| File Path |
| :--- |
| ` vllm_ascend/eplb/adaptor/abstract_adaptor.py` |
| ` vllm_ascend/eplb/adaptor/vllm_adaptor.py` |
| ` vllm_ascend/eplb/core/eplb_device_transfer_loader.py` |
| ` vllm_ascend/eplb/core/eplb_utils.py` |
| ` vllm_ascend/eplb/core/eplb_worker.py` |
| ` vllm_ascend/eplb/core/policy/policy_abstract.py` |
| ` vllm_ascend/eplb/core/policy/policy_default_eplb.py` |
| ` vllm_ascend/eplb/core/policy/policy_factory.py` |
| ` vllm_ascend/eplb/core/policy/policy_flashlb.py` |
| ` vllm_ascend/eplb/core/policy/policy_random.py` |
| ` vllm_ascend/eplb/core/policy/policy_swift_balancer.py` |
| ` vllm_ascend/eplb/eplb_updator.py` |
| ` vllm_ascend/eplb/utils.py` |
| ` vllm_ascend/model_loader/netloader/executor/elastic_load.py` |
| ` vllm_ascend/model_loader/netloader/executor/netloader_pg.py` |
| ` vllm_ascend/model_loader/netloader/interaction/elastic.py` |
| ` vllm_ascend/model_loader/netloader/load.py` |
| ` vllm_ascend/model_loader/netloader/netloader.py` |
| ` vllm_ascend/model_loader/netloader/utils.py` |
| ` vllm_ascend/patch/platform/__init__.py` |
| ` vllm_ascend/patch/platform/patch_balance_schedule.py` |
| ` vllm_ascend/patch/platform/patch_ec_connector.py` |
| ` vllm_ascend/patch/platform/patch_mamba_config.py` |
| ` vllm_ascend/patch/platform/patch_multiproc_executor.py` |
| ` vllm_ascend/patch/platform/patch_sched_yield.py` |


- vLLM version: v0.13.0
- vLLM main:
2c24bc6996

---------

Signed-off-by: MrZ20 <2609716663@qq.com>
This commit is contained in:
SILONG ZENG
2026-01-24 22:08:33 +08:00
committed by GitHub
parent 153da1a669
commit 4e53c1d900
26 changed files with 894 additions and 1148 deletions
Download Patch File Download Diff File Expand all files Collapse all files

46
vllm_ascend/model_loader/netloader/executor/elastic_load.py
View File

@@ -18,8 +18,7 @@ import torch
import torch_npu
from vllm.logger import logger
from .netloader_pg import (destroy_stateless_process_group,
stateless_init_process_group)
from .netloader_pg import destroy_stateless_process_group, stateless_init_process_group
class P2PLoad:
@@ -56,9 +55,7 @@ class P2PLoad:
- The model if loading is successful, otherwise None.
"""
model_device = next(model.parameters()).device
logger.info(
f"Start init_process_group, name: {self.world_name}, addr: {self.source_ip}:{self.source_port}"
)
logger.info(f"Start init_process_group, name: {self.world_name}, addr: {self.source_ip}:{self.source_port}")
receiver_pg = None
loaded_model = None
try:
@@ -67,15 +64,13 @@ class P2PLoad:
port=self.source_port,
rank=0,
world_size=2,
group_name='netloader',
group_name="netloader",
)
logger.info(
f"Finish init_process_group, name: {self.world_name}, addr: {self.source_ip}:{self.source_port}"
)
logger.info(
f"Start recv, name: {self.world_name}, addr: {self.source_ip}:{self.source_port}"
)
logger.info(f"Start recv, name: {self.world_name}, addr: {self.source_ip}:{self.source_port}")
logger.info(f"Model device: {model_device}")
trans_stream = torch_npu.npu.Stream()
@@ -84,14 +79,11 @@ class P2PLoad:
if len(param.shape) == 0:
continue
receiver_pg.recv([param], 1, 0).wait()
torch.distributed.barrier(group=receiver_pg,
device_ids=[model_device.index])
torch.distributed.barrier(group=receiver_pg, device_ids=[model_device.index])
torch_npu.npu.synchronize(trans_stream)
logger.info(
f"Finish recv, name: {self.world_name}, addr: {self.source_ip}:{self.source_port}"
)
logger.info(f"Finish recv, name: {self.world_name}, addr: {self.source_ip}:{self.source_port}")
loaded_model = model
except Exception as e:
logger.error("Failed to recv model: {}".format(e))
@@ -129,9 +121,7 @@ class P2PSend:
"""
model_device = next(model.parameters()).device
torch.npu.set_device(model_device)
logger.info(
f"Start init_process_group, name: {self.comm_name}, addr: {self.listen_ip}:{self.listen_port}"
)
logger.info(f"Start init_process_group, name: {self.comm_name}, addr: {self.listen_ip}:{self.listen_port}")
sender_pg = None
try:
sender_pg = stateless_init_process_group(
@@ -139,14 +129,10 @@ class P2PSend:
port=self.listen_port,
rank=1,
world_size=2,
group_name='netloader',
)
logger.info(
f"Finish init_process_group, name: {self.comm_name}, addr: {self.listen_ip}:{self.listen_port}"
)
logger.info(
f"Start send, name: {self.comm_name}, addr: {self.listen_ip}:{self.listen_port}"
group_name="netloader",
)
logger.info(f"Finish init_process_group, name: {self.comm_name}, addr: {self.listen_ip}:{self.listen_port}")
logger.info(f"Start send, name: {self.comm_name}, addr: {self.listen_ip}:{self.listen_port}")
logger.info(f"Model device: {model_device}")
trans_stream = torch_npu.npu.Stream()
@@ -155,16 +141,12 @@ class P2PSend:
if "aclnn_input_scale" in name:
continue
if name in int8_params:
sender_pg.send([int8_params[name].to(model_device)], 0,
0).wait()
sender_pg.send([int8_params[name].to(model_device)], 0, 0).wait()
else:
sender_pg.send([param.contiguous()], 0, 0).wait()
torch.distributed.barrier(group=sender_pg,
device_ids=[model_device.index])
torch.distributed.barrier(group=sender_pg, device_ids=[model_device.index])
torch_npu.npu.synchronize(trans_stream)
logger.info(
f"Finish send, name: {self.comm_name}, addr: {self.listen_ip}:{self.listen_port}"
)
logger.info(f"Finish send, name: {self.comm_name}, addr: {self.listen_ip}:{self.listen_port}")
finally:
if sender_pg:
destroy_stateless_process_group(sender_pg)
destroy_stateless_process_group(sender_pg)