xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-07-10 15:27:22 +08:00
Code Issues Packages Projects Releases Wiki Activity

[Misc] add process_weights_after_loading for DummyLoader (#8969)

Browse Source
This commit is contained in:
Divakar Verma 2024-09-30 22:46:41 -05:00 committed by GitHub
parent 7da2487591
commit 82f3937e59
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 12 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

12
vllm/model_executor/model_loader/loader.py
View File

@ -441,6 +441,18 @@ class DummyModelLoader(BaseModelLoader):
# NOTE(woosuk): For accurate performance evaluation, we assign # NOTE(woosuk): For accurate performance evaluation, we assign
# random values to the weights. # random values to the weights.
initialize_dummy_weights(model) initialize_dummy_weights(model)
for _, module in model.named_modules():
quant_method = getattr(module, "quant_method", None)
if quant_method is not None:
# When quant methods need to process weights after loading
# (for repacking, quantizing, etc), they expect parameters
# to be on the global target device. This scope is for the
# case where cpu offloading is used, where we will move the
# parameters onto device for processing and back off after.
with device_loading_context(
module, torch.device(device_config.device)):
quant_method.process_weights_after_loading(module)
return model.eval() return model.eval()