ayo/ipex-llm
1
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions 1

optimize moonlight again (#12909)

Browse source
This commit is contained in:
Yishuo Wang 2025-03-03 09:21:15 +08:00 committed by GitHub
parent 35e5fa851c
commit b6f33d5c4d
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
1 changed files with 2 additions and 0 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

2
python/llm/src/ipex_llm/transformers/convert.py
View file

@ -2029,10 +2029,12 @@ def _optimize_post(model):
modeling_module_name = model.__class__.__module__
module = importlib.import_module(modeling_module_name)
from ipex_llm.transformers.models.common import rms_norm_forward
from ipex_llm.transformers.models.common import mlp_silu_forward
from ipex_llm.transformers.models.deepseek import deepseek_model_forward
from ipex_llm.transformers.models.deepseek import deepseek_attention_forward
from ipex_llm.transformers.models.deepseek import deepseek_moe_forward
convert_forward(model, module.DeepseekV3RMSNorm, rms_norm_forward)
convert_forward(model, module.DeepseekV3MLP, mlp_silu_forward)
convert_forward(model, module.DeepseekV3Model, deepseek_model_forward)
convert_forward(model, module.DeepseekV3Attention, deepseek_attention_forward)
convert_forward(model, module.DeepseekV3MoE, deepseek_moe_forward)