xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2025-12-11 13:15:39 +08:00
Code Issues Packages Projects Releases Wiki Activity

[Model][Bugfix]fix ernie45 load failed due to ernie45 eplb code (#26684)

Browse Source 
Signed-off-by: wangyafeng <wangyafeng@baidu.com>
This commit is contained in:
CSWYF3634076 2025-10-14 14:55:23 +08:00 committed by GitHub
parent 481545b397
commit 01ad27faff
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 22 additions and 12 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

34
vllm/model_executor/models/ernie45_moe.py
View File

@ -23,7 +23,8 @@
# limitations under the License. # limitations under the License.
"""Inference-only ErineMoE model compatible with HuggingFace weights.""" """Inference-only ErineMoE model compatible with HuggingFace weights."""
from collections.abc import Iterable import typing
from collections.abc import Callable, Iterable
from itertools import islice from itertools import islice
from typing import Any from typing import Any
@ -139,10 +140,10 @@ class Ernie4_5_MoeMoE(nn.Module):
# Load balancing settings. # Load balancing settings.
vllm_config = get_current_vllm_config() vllm_config = get_current_vllm_config()
parallel_config = vllm_config.parallel_config eplb_config = vllm_config.parallel_config.eplb_config
self.enable_eplb = enable_eplb self.enable_eplb = enable_eplb
self.n_redundant_experts = parallel_config.num_redundant_experts self.n_redundant_experts = eplb_config.num_redundant_experts
self.n_logical_experts = self.n_routed_experts self.n_logical_experts = self.n_routed_experts
self.n_physical_experts = self.n_logical_experts + self.n_redundant_experts self.n_physical_experts = self.n_logical_experts + self.n_redundant_experts
self.n_local_physical_experts = self.n_physical_experts // self.ep_size self.n_local_physical_experts = self.n_physical_experts // self.ep_size
@ -426,8 +427,10 @@ class Ernie4_5_MoeModel(nn.Module):
self.vocab_size = config.vocab_size self.vocab_size = config.vocab_size
self.config = config self.config = config
parallel_config = vllm_config.parallel_config parallel_config = vllm_config.parallel_config
eplb_config = parallel_config.eplb_config
enable_eplb = parallel_config.enable_eplb enable_eplb = parallel_config.enable_eplb
self.num_redundant_experts = parallel_config.num_redundant_experts
self.num_redundant_experts = eplb_config.num_redundant_experts
if get_pp_group().is_first_rank: if get_pp_group().is_first_rank:
self.embed_tokens = VocabParallelEmbedding( self.embed_tokens = VocabParallelEmbedding(
@ -570,20 +573,27 @@ class Ernie4_5_MoeModel(nn.Module):
# Skip loading extra bias for GPTQ models. # Skip loading extra bias for GPTQ models.
if ( if (
name.endswith(".bias") or name.endswith("_bias") name_mapped.endswith(".bias") or name_mapped.endswith("_bias")
) and name not in params_dict: ) and name_mapped not in params_dict:
continue continue
param = params_dict[name] param = params_dict[name_mapped]
# We should ask the weight loader to return success or not
weight_loader = param.weight_loader # here since otherwise we may skip experts with other
weight_loader( # available replicas.
weight_loader = typing.cast(
Callable[..., bool], param.weight_loader
)
success = weight_loader(
param, param,
loaded_weight, loaded_weight,
name, name_mapped,
shard_id=shard_id, shard_id=shard_id,
expert_id=expert_id, expert_id=expert_id,
return_success=True,
) )
break if success:
name = name_mapped
break
else: else:
if is_expert_weight: if is_expert_weight:
# We've checked that this is an expert weight # We've checked that this is an expert weight