apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-09-27 04:46:16 +08:00
Code Issues Actions 5 Packages Projects Releases Wiki Activity

fix ernie vl distributed attr. (#4217)
Some checks failed
CE Compile Job / ce_job_pre_check (push) Has been cancelled
Details
CE Compile Job / print_ce_job_pre_check_outputs (push) Has been cancelled
Details
CE Compile Job / FD-Clone-Linux (push) Has been cancelled
Details
CE Compile Job / Show Code Archive Output (push) Has been cancelled
Details
CE Compile Job / BUILD_SM8090 (push) Has been cancelled
Details
CE Compile Job / BUILD_SM8689 (push) Has been cancelled
Details
CE Compile Job / CE_UPLOAD (push) Has been cancelled
Details

Browse Source
This commit is contained in:
Zhong Hui
2025-09-23 19:37:38 +08:00
committed by GitHub
parent cb8d87b945
commit a460462d2a
1 changed files with 1 additions and 13 deletions
Download Patch File Download Diff File Expand all files Collapse all files

14
fastdeploy/model_executor/models/ernie4_5_vl/modeling_resampler.py
View File

@@ -23,7 +23,7 @@ from paddle import nn
from paddle.autograd import PyLayer from paddle.autograd import PyLayer
from paddle.distributed.fleet.utils import recompute from paddle.distributed.fleet.utils import recompute
from fastdeploy.model_executor.layers.utils import _set_var_distributed, get_tensor from fastdeploy.model_executor.layers.utils import get_tensor
from fastdeploy.model_executor.models.ernie4_5_vl.dist_utils import ( from fastdeploy.model_executor.models.ernie4_5_vl.dist_utils import (
RowSequenceParallelLinear, RowSequenceParallelLinear,
all_gather_group, all_gather_group,
@@ -197,19 +197,7 @@ class VariableResolutionResamplerModel(nn.Layer):
self.after_norm = RMSNorm(out_config) self.after_norm = RMSNorm(out_config)
if self.tensor_parallel_degree > 1: if self.tensor_parallel_degree > 1:
for idx in [2, 3]:
mark_as_sequence_parallel_parameter(self.spatial_linear[idx].weight)
mark_as_sequence_parallel_parameter(self.spatial_linear[idx].bias)
_set_var_distributed(self.spatial_linear[idx].weight, split_axis=0)
_set_var_distributed(self.spatial_linear[idx].bias, split_axis=0)
if self.use_temporal_conv:
for idx in [0, 2, 3]:
mark_as_sequence_parallel_parameter(self.temporal_linear[idx].weight)
mark_as_sequence_parallel_parameter(self.temporal_linear[idx].bias)
mark_as_sequence_parallel_parameter(self.mlp.weight)
mark_as_sequence_parallel_parameter(self.mlp.bias)
mark_as_sequence_parallel_parameter(self.after_norm.weight)
set_weight_attrs(self.spatial_linear[0].weight, {"output_dim": False}) set_weight_attrs(self.spatial_linear[0].weight, {"output_dim": False})
def spatial_conv_reshape(self, x, spatial_conv_size): def spatial_conv_reshape(self, x, spatial_conv_size):