apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-10-05 16:48:03 +08:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

fix mask_offset in append_attn (#3745)
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
Deploy GitHub Pages / deploy (push) Has been cancelled
Details

Browse Source 
* fix mask_offset in append_attn

* fix test
This commit is contained in:
lizhenyun01
2025-08-31 15:03:16 +08:00
committed by GitHub
parent 753772ace8
commit bed09ae8f8
8 changed files with 46 additions and 44 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
fastdeploy/model_executor/layers/attention/attention.py
View File

@@ -163,14 +163,14 @@ class Attention(nn.Layer):
def init_weight(self):
self.q_norm_weight = self.create_parameter(
shape=[self.qk_head_dim],
dtype=self._dtype,
dtype="float32",
is_bias=False,
default_initializer=paddle.nn.initializer.Constant(0),
)
self.k_norm_weight = self.create_parameter(
shape=[self.qk_head_dim],
dtype=self._dtype,
dtype="float32",
is_bias=False,
default_initializer=paddle.nn.initializer.Constant(0),
)
@@ -184,8 +184,8 @@ class Attention(nn.Layer):
if self.use_qk_norm:
q_norm_weight_tensor = paddle.to_tensor(get_tensor(state_dict.pop(self.q_norm_key + ".weight")))
k_norm_weight_tensor = paddle.to_tensor(get_tensor(state_dict.pop(self.k_norm_key + ".weight")))
self.q_norm_weight.set_value(q_norm_weight_tensor)
self.k_norm_weight.set_value(k_norm_weight_tensor)
self.q_norm_weight.set_value(q_norm_weight_tensor.astype("float32"))
self.k_norm_weight.set_value(k_norm_weight_tensor.astype("float32"))
def forward(
self,