Skip to content

GitLab

Unverified Commit 6dffa4b0 authored by Murali Andoorveedu's avatar Murali Andoorveedu Committed by GitHub
Browse files

[Bugfix] Fix LoRA with PP (#7292)

parent 48abee9e
Show whitespace changes
Inline Side-by-side
Showing with 3 additions and 0 deletions
vllm/lora/models.py
View file @ 6dffa4b0
...@@ -25,6 +25,7 @@ from vllm.lora.punica import PunicaWrapper ...@@ -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): ...@@ -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]
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment