apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-12-24 13:28:13 +08:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

[vl]remove duplicated load logic (#2744)
Some checks failed
Deploy GitHub Pages / deploy (push) Has been cancelled
Details

Browse Source
This commit is contained in:
bukejiyu
2025-07-13 07:36:26 +08:00
committed by GitHub
parent 16940822a7
commit bad53c6b6e
11 changed files with 510 additions and 632 deletions
Download Patch File Download Diff File Expand all files Collapse all files

30
fastdeploy/model_executor/models/ernie4_5_moe.py
View File

@@ -26,6 +26,7 @@ from paddleformers.transformers import PretrainedModel
from paddleformers.utils.log import logger
from fastdeploy.config import FDConfig, ModelConfig
from fastdeploy.model_executor.forward_meta import ForwardMeta
from fastdeploy.model_executor.graph_optimization.decorator import \
support_graph_optimization
from fastdeploy.model_executor.layers.activation import SiluAndMul
@@ -41,7 +42,6 @@ from fastdeploy.model_executor.models.tp_utils import TensorSplitMode as tsm
from fastdeploy.model_executor.models.utils import \
LayerIdPlaceholder as layerid
from fastdeploy.model_executor.models.utils import WeightMeta
from fastdeploy.model_executor.forward_meta import ForwardMeta
class Ernie4_5_MLP(nn.Layer):
@@ -599,27 +599,13 @@ class Ernie4_5_PretrainedModel(PretrainedModel):
start_layer = (moe_layer_start_index
if moe_layer_start_index > 0 else num_layers)
final_actions = build_expanded_keys(
num_layers,
moe_num_experts,
start_layer,
base_actions,
base_actions, num_layers, start_layer, moe_num_experts
)
return final_actions
moe_num_experts = 0
if isinstance(config.moe_num_experts, list):
moe_num_experts = sum(config.moe_num_experts)
elif isinstance(config.moe_num_experts, int):
moe_num_experts = config.moe_num_experts
moe_layer_start_index = -1
if isinstance(config.moe_layer_start_index, list):
moe_layer_start_index = min(config.moe_layer_start_index)
elif isinstance(config.moe_layer_start_index, int):
moe_layer_start_index = config.moe_layer_start_index
mappings = get_tensor_parallel_split_mappings(config.num_layers,
moe_num_experts,
moe_layer_start_index,
config.prefix_name)
mappings = get_tensor_parallel_split_mappings(
config.num_layers,
config.moe_num_experts,
config.moe_layer_start_index,
config.prefix_name,
)
return mappings