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

update for bge-m3 (#13138)

Browse source
This commit is contained in:
Wang, Jian4 2025-05-07 16:59:52 +08:00 committed by GitHub
parent e88a2aa65b
commit f2598b119e
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
1 changed files with 2 additions and 1 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

3
python/llm/src/ipex_llm/vllm/xpu/model_convert.py
View file

@ -65,13 +65,14 @@ def _model_sample_convert():
def _ipex_llm_convert(load_in_low_bit):
# import pdb
# pdb.set_trace()
from vllm.worker.xpu_model_runner import XPUModelRunner
from vllm.worker.xpu_model_runner import XPUModelRunner, XPUModelRunnerBase
from ipex_llm.vllm.xpu.ipex_llm_wrapper import get_ipex_llm_wrapper
from ipex_llm.vllm.xpu.ipex_llm_v1_wrapper import get_ipex_llm_v1_wrapper
import vllm.executor.ray_utils as ray_utils_v0
import vllm.v1.executor.ray_utils as ray_utils_v1
from vllm.v1.worker.gpu_model_runner import GPUModelRunner
setattr(XPUModelRunner, "load_model", get_load_function(load_in_low_bit))
setattr(XPUModelRunnerBase, "load_model", get_load_function(load_in_low_bit))
setattr(GPUModelRunner, "load_model", get_load_function(load_in_low_bit))
setattr(ray_utils_v0, "RayWorkerWrapper", get_ipex_llm_wrapper(load_in_low_bit))
setattr(ray_utils_v1, "RayWorkerWrapper", get_ipex_llm_v1_wrapper(load_in_low_bit))