apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-10-05 08:37:06 +08:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

refactor rl get_name_mappings_to_training (#2847)
Some checks failed
Deploy GitHub Pages / deploy (push) Has been cancelled
Details

Browse Source 
* refactor rl get_name_mappings_to_training

* fix tp>1

* change variable name(ffn1->up_gate_proj/ffn2->down_proj)

* change variable name(linear_weight->weight/linear_bias->bias)

* add rl names mapping for vl

* fix ernie 0.3B error

* fix develop code

* fix
This commit is contained in:
Yuanle Liu
2025-07-15 22:31:42 +08:00
committed by GitHub
parent e7bcbbab52
commit 61b3997b85
47 changed files with 1591 additions and 1629 deletions
Download Patch File Download Diff File Expand all files Collapse all files

24
fastdeploy/model_executor/load_weight_utils.py
View File

@@ -48,22 +48,22 @@ def load_ep_checkpoint(model_path: str,
config.num_experts_start_offset,
config.num_experts_start_offset + config.num_experts_per_rank,
):
ffn1_key = f"ernie.layers.{i}.mlp.experts.{j}.up_gate_proj.weight"
ffn2_key = (f"ernie.layers.{i}.mlp.experts.{j}.down_proj.weight")
up_gate_proj_key = f"ernie.layers.{i}.mlp.experts.{j}.up_gate_proj.weight"
down_proj_key = (f"ernie.layers.{i}.mlp.experts.{j}.down_proj.weight")
ffn1_quant_key = f"ernie.layers.{i}.mlp.experts.{j}.up_gate_proj.quant_weight"
ffn2_quant_key = (
up_gate_proj_quant_key = f"ernie.layers.{i}.mlp.experts.{j}.up_gate_proj.quant_weight"
down_proj_quant_key = (
f"ernie.layers.{i}.mlp.experts.{j}.down_proj.quant_weight")
ffn1_scale_key = f"ernie.layers.{i}.mlp.experts.{j}.up_gate_proj.weight_scale"
ffn2_scale_key = (
up_gate_proj_scale_key = f"ernie.layers.{i}.mlp.experts.{j}.up_gate_proj.weight_scale"
down_proj_scale_key = (
f"ernie.layers.{i}.mlp.experts.{j}.down_proj.weight_scale")
num_local_ffn_keys.append(ffn1_key)
num_local_ffn_keys.append(ffn2_key)
num_local_ffn_keys.append(ffn1_quant_key)
num_local_ffn_keys.append(ffn2_quant_key)
num_local_ffn_keys.append(ffn1_scale_key)
num_local_ffn_keys.append(ffn2_scale_key)
num_local_ffn_keys.append(up_gate_proj_key)
num_local_ffn_keys.append(down_proj_key)
num_local_ffn_keys.append(up_gate_proj_quant_key)
num_local_ffn_keys.append(down_proj_quant_key)
num_local_ffn_keys.append(up_gate_proj_scale_key)
num_local_ffn_keys.append(down_proj_scale_key)
for k in num_local_ffn_keys:
if k in weight_list: