apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2026-04-23 17:11:21 +08:00
Code Issues Actions 23 Packages Projects Releases Wiki Activity

[Models] Add forward_meta to moe models' forward function (#5138)

Browse Source 
* [Models] Add forward_meta to moe models' forward function

* fix missing param

* fix

* fix

* fix forward_meta

* fix test and remove chunked MoE releated in config

* fix test

* fix

* fix
This commit is contained in:
Longzhi Wang
2025-12-04 13:26:58 +08:00
committed by GitHub
parent f5bdb36e9b
commit 5cd17fd662
21 changed files with 131 additions and 87 deletions
Download Patch File Download Diff File Expand all files Collapse all files
fastdeploy/model_executor/models/qwen2.py
+2 -2
View File
@@ -89,7 +89,7 @@ class Qwen2MLP(nn.Layer):
self.up_gate_proj.load_state_dict(state_dict)
self.down_proj.load_state_dict(state_dict)
def forward(self, x):
def forward(self, x, forward_meta):
""" """
gate_up_out = self.up_gate_proj(x)
act_out = self.act_fn(gate_up_out)
@@ -205,7 +205,7 @@ class Qwen2DecoderLayer(nn.Layer):
# Fully Connected
hidden_states, residual = self.post_attention_layernorm(hidden_states, residual)
hidden_states = self.mlp(hidden_states)
hidden_states = self.mlp(hidden_states, forward_meta)
return hidden_states, residual