mmarch/text-generation-webui
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

Add comments

Browse source
This commit is contained in:
oobabooga 2023-03-10 10:54:08 -03:00
parent 113b791aa5
commit de7dd8b6aa
1 changed files with 3 additions and 0 deletions
Download patch file Download diff file Expand all files Collapse all files

3
modules/models.py
View file

@ -111,6 +111,7 @@ def load_model(model_name):
model = load_quant(path_to_model, Path(f"models/{pt_model}"), 4) model = load_quant(path_to_model, Path(f"models/{pt_model}"), 4)
# Multi-GPU setup
if shared.args.gpu_memory: if shared.args.gpu_memory:
import accelerate import accelerate
@ -121,6 +122,8 @@ def load_model(model_name):
device_map = accelerate.infer_auto_device_map(model, max_memory=max_memory, no_split_module_classes=["LLaMADecoderLayer"]) device_map = accelerate.infer_auto_device_map(model, max_memory=max_memory, no_split_module_classes=["LLaMADecoderLayer"])
model = accelerate.dispatch_model(model, device_map=device_map) model = accelerate.dispatch_model(model, device_map=device_map)
# Single GPU
else: else:
model = model.to(torch.device('cuda:0')) model = model.to(torch.device('cuda:0'))