squall/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity

[Bugfix] Fix LoRA with PP (#7292)

Browse Source
This commit is contained in:
Murali Andoorveedu 2024-08-08 00:02:27 -07:00 committed by GitHub
parent 48abee9e54
commit 6dffa4b0a6
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 3 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
vllm/lora/models.py
View File

@ -25,6 +25,7 @@ from vllm.lora.punica import PunicaWrapper
from vllm.lora.utils import (from_layer, from_layer_logits_processor, from vllm.lora.utils import (from_layer, from_layer_logits_processor,
parse_fine_tuned_lora_name, replace_submodule) parse_fine_tuned_lora_name, replace_submodule)
from vllm.model_executor.models.interfaces import SupportsLoRA from vllm.model_executor.models.interfaces import SupportsLoRA
from vllm.model_executor.models.utils import PPMissingLayer
from vllm.utils import is_pin_memory_available from vllm.utils import is_pin_memory_available
logger = init_logger(__name__) logger = init_logger(__name__)
@ -432,6 +433,8 @@ class LoRAModelManager(AdapterModelManager):
def _create_lora_modules(self): def _create_lora_modules(self):
for module_name, module in self.model.named_modules( for module_name, module in self.model.named_modules(
remove_duplicate=False): remove_duplicate=False):
if isinstance(module, PPMissingLayer):
continue
if not self._match_target_modules(module_name): if not self._match_target_modules(module_name):
continue continue
parts = module_name.split(".")[-1] parts = module_name.split(".")[-1]