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 9 Packages Projects Releases Wiki Activity

[V1 Loader] Support Ernie text（moe and dense） (#3110)

Browse Source 
* new loader support 0.3B

* fix weight

* support parallel load

* support parallel load

* fix slice

* support moe

* delete code

* perfect code

* perfect code
This commit is contained in:
YuanRisheng
2025-08-14 20:25:28 +08:00
committed by GitHub
parent ab60292f89
commit 09c979f3dd
6 changed files with 218 additions and 85 deletions
Download Patch File Download Diff File Expand all files Collapse all files
fastdeploy/model_executor/models/qwen3moe.py
+1 -1
View File
@@ -317,12 +317,12 @@ class Qwen3MoeForCausalLM(ModelForCasualLM):
) -> list[tuple[str, str, int, str]]:
# (param_name, weight_name, expert_id, shard_id)
return FusedMoE.make_expert_params_mapping(
num_experts=self.fd_config.model_config.num_experts,
ckpt_gate_proj_name="gate_proj",
ckpt_down_proj_name="down_proj",
ckpt_up_proj_name="up_proj",
param_gate_up_proj_name="experts.up_gate_proj_",
param_down_proj_name="experts.down_proj_",
num_experts=self.fd_config.model_config.num_experts,
)
@paddle.no_grad()