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

[Bugfix] Fix the bug in initializing the shared_weight communication domain in sfa-cp, and fix the mtp weight load in pp>1 situation (#4913)

Browse Source 
### What this PR does / why we need it?
In PR #4188, a small bug was introduced that caused sfa-cp to be unable
to find the global_pp_size parameter during initialization, and this PR
fixed the issue.

- vLLM version: v0.12.0
- vLLM main:
ad32e3e19c

Signed-off-by: zzhx1 <zzh_201018@outlook.com>
Co-authored-by: Jade Zheng <zheng.shoujian@outlook.com>
This commit is contained in:
zzhxxx
2025-12-15 16:21:49 +08:00
committed by GitHub
parent 70606e0bb9
commit e16444f21f
2 changed files with 10 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
vllm_ascend/distributed/parallel_state.py
View File

@@ -182,9 +182,8 @@ def init_ascend_model_parallel(parallel_config: ParallelConfig, ):
global _SHARED_WEIGHT global _SHARED_WEIGHT
# TODO: Check if the model is Deepseek V3.2 with enabled SFA CP and activated shared weights. It will then be normalized within the PCP parameters. -- clrs97 # TODO: Check if the model is Deepseek V3.2 with enabled SFA CP and activated shared weights. It will then be normalized within the PCP parameters. -- clrs97
is_ds_v32 = hasattr(vllm_config.model_config.hf_config, "index_topk") is_ds_v32 = hasattr(vllm_config.model_config.hf_config, "index_topk")
if enable_sp() and is_ds_v32: if enable_sp() and is_ds_v32 and _SHARED_WEIGHT is None:
_SHARED_WEIGHT = _create_shared_weight_group("CP_shared_weight") _SHARED_WEIGHT = _create_shared_weight_group("CP_shared_weight")
# TODO: Extract and unify the logic across different communication group. # TODO: Extract and unify the logic across different communication group.
if flashcomm2_enable(): if flashcomm2_enable():
flashcomm2_otp_size = get_ascend_config( flashcomm2_otp_size = get_ascend_config(
@@ -240,7 +239,9 @@ def init_ascend_model_parallel(parallel_config: ParallelConfig, ):
# Create shared weight group for flashcomm2 oproj # Create shared weight group for flashcomm2 oproj
if flashcomm2_o_shared_enabled(): if flashcomm2_o_shared_enabled():
assert flashcomm2_otp_size == 1, "flashcomm2_o_shared is only supported when flashcomm2_otp_size is 1" assert flashcomm2_otp_size == 1, "flashcomm2_o_shared is only supported when flashcomm2_otp_size is 1"
_SHARED_WEIGHT = _create_shared_weight_group("flashcomm2_o_shared") if _SHARED_WEIGHT is None:
_SHARED_WEIGHT = _create_shared_weight_group(
"flashcomm2_o_shared")
if get_ascend_config().multistream_overlap_gate: if get_ascend_config().multistream_overlap_gate:
global _FC3_QUANT_X global _FC3_QUANT_X

9
vllm_ascend/spec_decode/mtp_proposer.py
View File

@@ -8,6 +8,7 @@ import torch.nn.functional as F
from vllm.config import (CUDAGraphMode, VllmConfig, from vllm.config import (CUDAGraphMode, VllmConfig,
get_layers_from_vllm_config, set_current_vllm_config) get_layers_from_vllm_config, set_current_vllm_config)
from vllm.distributed import get_pcp_group from vllm.distributed import get_pcp_group
from vllm.distributed.parallel_state import get_pp_group
from vllm.forward_context import get_forward_context from vllm.forward_context import get_forward_context
from vllm.logger import init_logger from vllm.logger import init_logger
from vllm.model_executor.layers.attention_layer_base import AttentionLayerBase from vllm.model_executor.layers.attention_layer_base import AttentionLayerBase
@@ -205,9 +206,11 @@ class MtpProposer(Proposer):
if self.vllm_config.model_config.is_deepseek_mla: if self.vllm_config.model_config.is_deepseek_mla:
# check if mtp model use main model's embedding and LMhead # check if mtp model use main model's embedding and LMhead
main_model = model main_model = model
if torch.equal(self.model.model.embed_tokens.weight, if get_pp_group().world_size == 1:
main_model.model.embed_tokens.weight): # If pp>1, the weights of mtp and the main model's embedding are not on the same device.
self.model.model.embed_tokens = main_model.model.embed_tokens if torch.equal(self.model.model.embed_tokens.weight,
main_model.model.embed_tokens.weight):
self.model.model.embed_tokens = main_model.model.embed_tokens
for _, layer_module in self.model.model.layers.items(): for _, layer_module in self.model.model.layers.items():
if torch.equal(layer_module.shared_head.head.weight, if torch.equal(layer_module.shared_head.head.weight,
main_model.lm_head.weight): main_model.lm_head.weight):