apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2026-04-23 08:21:53 +08:00
Code Issues Actions 19 Packages Projects Releases Wiki Activity

[RL] Support Rollout Routing Replay (#5321)

Browse Source 
* [RL] Support Rollout Routing Replay

* add routing indices cache

* fix config bug and moe forward bug

* R3 Support GLM

* support eb4.5

* fix merge bug

* Apply suggestion from @Copilot

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

* Apply suggestion from @Copilot

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

* Apply suggestion from @Copilot

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

* Apply suggestion from @Copilot

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

* add routing replay ci

* support glm topk

* support orther top_k

* fix ci bug

* pre-commit

* only support chatcmpl

---------

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
Co-authored-by: Yuanle Liu <yuanlehome@163.com>
This commit is contained in:
RAM
2025-12-05 20:01:33 +08:00
committed by GitHub
parent 8545b705ed
commit 96d2d4877b
24 changed files with 592 additions and 24 deletions
Download Patch File Download Diff File Expand all files Collapse all files
fastdeploy/model_executor/models/glm4_moe.py
+2 -5
View File
@@ -161,7 +161,7 @@ class Glm4Moe(nn.Layer):
reduce_results=False,
)
def forward(self, x, forward_meta):
def forward(self, x, forward_meta: ForwardMeta = None):
shared_experts_out = self.shared_experts(x)
out = self.experts(x, self.gate, forward_meta)
out = out + shared_experts_out
@@ -306,10 +306,7 @@ class Glm4MoeDecoderLayer(nn.Layer):
# Fully Connected
hidden_states, residual = self.post_attention_layernorm(hidden_states, residual)
hidden_states = self.mlp(
hidden_states,
forward_meta,
)
hidden_states = self.mlp(hidden_states, forward_meta)
return hidden_states, residual