Mirrors/text-generation-webui
1
0
Fork 0
mirror of https://github.com/oobabooga/text-generation-webui.git synced 2024-10-30 14:10:14 +01:00
Code Issues Packages Projects Releases Wiki Activity

Merge pull request #219 from deepdiffuser/4bit-multigpu

Browse Source 
add multi-gpu support for 4bit gptq LLaMA
This commit is contained in:
oobabooga 2023-03-10 10:52:45 -03:00 committed by GitHub
commit 113b791aa5
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 13 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

14
modules/models.py
View File

@ -110,7 +110,19 @@ def load_model(model_name):
exit()
model = load_quant(path_to_model, Path(f"models/{pt_model}"), 4)
model = model.to(torch.device('cuda:0'))
if shared.args.gpu_memory:
import accelerate
max_memory = {}
for i in range(len(shared.args.gpu_memory)):
max_memory[i] = f"{shared.args.gpu_memory[i]}GiB"
max_memory['cpu'] = f"{shared.args.cpu_memory or '99'}GiB"
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)
else:
model = model.to(torch.device('cuda:0'))
# Custom
else: