apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2026-04-24 01:29:57 +08:00
Code Issues Actions 11 Packages Projects Releases Wiki Activity

[Metax] support default_v1 loader & thinking model (#4956)

Browse Source 
Co-authored-by: plusNew001 <95567040+plusNew001@users.noreply.github.com>
This commit is contained in:
MingkunZhang
2025-11-12 16:32:26 +08:00
committed by GitHub
parent bde6e2f931
commit 9d9f5df8d0
4 changed files with 23 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files
fastdeploy/model_executor/layers/embeddings.py
+3 -1
View File
@@ -24,6 +24,7 @@ from paddle.distributed import fleet
from fastdeploy.config import FDConfig
from fastdeploy.model_executor.utils import h2d_copy, set_weight_attrs, slice_fn
from fastdeploy.platforms import current_platform
from .utils import (
DEFAULT_VOCAB_PADDING_SIZE,
@@ -274,7 +275,8 @@ class VocabParallelEmbedding(nn.Layer):
if output_dim == 0:
h2d_copy(param[: shard_weight.shape[0]], shard_weight)
param[shard_weight.shape[0] :].fill_(0)
if not current_platform.is_maca():
param[shard_weight.shape[0] :].fill_(0)
else:
h2d_copy(param[:, : shard_weight.shape[1]], shard_weight)
param[:, shard_weight.shape[1] :].fill_(0)