biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

Convert formatting to use ruff instead of yapf + isort (#26247)

Browse Source 
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
This commit is contained in:
Harry Mellor
2025-10-05 15:06:22 +01:00
committed by GitHub
parent 17edd8a807
commit d6953beb91
1508 changed files with 115244 additions and 94146 deletions
Download Patch File Download Diff File Expand all files Collapse all files

11
vllm/plugins/__init__.py
View File

@@ -8,7 +8,7 @@ import vllm.envs as envs
logger = logging.getLogger(__name__)
DEFAULT_PLUGINS_GROUP = 'vllm.general_plugins'
DEFAULT_PLUGINS_GROUP = "vllm.general_plugins"
# make sure one process only loads plugins once
plugins_loaded = False
@@ -16,6 +16,7 @@ plugins_loaded = False
def load_plugins_by_group(group: str) -> dict[str, Callable[[], Any]]:
import sys
if sys.version_info < (3, 10):
from importlib_metadata import entry_points
else:
@@ -29,7 +30,7 @@ def load_plugins_by_group(group: str) -> dict[str, Callable[[], Any]]:
return {}
# Check if the only discovered plugin is the default one
is_default_group = (group == DEFAULT_PLUGINS_GROUP)
is_default_group = group == DEFAULT_PLUGINS_GROUP
# Use INFO for non-default groups and DEBUG for the default group
log_level = logger.debug if is_default_group else logger.info
@@ -38,8 +39,10 @@ def load_plugins_by_group(group: str) -> dict[str, Callable[[], Any]]:
log_level("- %s -> %s", plugin.name, plugin.value)
if allowed_plugins is None:
log_level("All plugins in this group will be loaded. "
"Set `VLLM_PLUGINS` to control which plugins to load.")
log_level(
"All plugins in this group will be loaded. "
"Set `VLLM_PLUGINS` to control which plugins to load."
)
plugins = dict[str, Callable[[], Any]]()
for plugin in discovered_plugins:

17
vllm/plugins/io_processors/__init__.py
View File

@@ -15,8 +15,8 @@ logger = logging.getLogger(__name__)
def get_io_processor(
vllm_config: VllmConfig,
plugin_from_init: Optional[str] = None) -> IOProcessor | None:
vllm_config: VllmConfig, plugin_from_init: Optional[str] = None
) -> IOProcessor | None:
# Input.Output processors are loaded as plugins under the
# 'vllm.io_processor_plugins' group. Similar to platform
# plugins, these plugins register a function that returns the class
@@ -39,8 +39,9 @@ def get_io_processor(
logger.debug("IOProcessor plugin to be loaded %s", model_plugin)
# Load all installed plugin in the group
multimodal_data_processor_plugins = \
load_plugins_by_group('vllm.io_processor_plugins')
multimodal_data_processor_plugins = load_plugins_by_group(
"vllm.io_processor_plugins"
)
loadable_plugins = {}
for name, func in multimodal_data_processor_plugins.items():
@@ -54,14 +55,16 @@ def get_io_processor(
num_available_plugins = len(loadable_plugins.keys())
if num_available_plugins == 0:
raise ValueError("No IOProcessor plugins installed"
f" but one is required ({model_plugin}).")
raise ValueError(
f"No IOProcessor plugins installed but one is required ({model_plugin})."
)
if model_plugin not in loadable_plugins:
raise ValueError(
f"The model requires the '{model_plugin}' IO Processor plugin "
"but it is not installed. "
f"Available plugins: {list(loadable_plugins.keys())}")
f"Available plugins: {list(loadable_plugins.keys())}"
)
activated_plugin_cls = loadable_plugins[model_plugin]

23
vllm/plugins/io_processors/interface.py
View File

@@ -10,12 +10,11 @@ from vllm.entrypoints.openai.protocol import IOProcessorResponse
from vllm.inputs.data import PromptType
from vllm.outputs import PoolingRequestOutput
IOProcessorInput = TypeVar('IOProcessorInput')
IOProcessorOutput = TypeVar('IOProcessorOutput')
IOProcessorInput = TypeVar("IOProcessorInput")
IOProcessorOutput = TypeVar("IOProcessorOutput")
class IOProcessor(ABC, Generic[IOProcessorInput, IOProcessorOutput]):
def __init__(self, vllm_config: VllmConfig):
self.vllm_config = vllm_config
@@ -37,10 +36,12 @@ class IOProcessor(ABC, Generic[IOProcessorInput, IOProcessorOutput]):
return self.pre_process(prompt, request_id, **kwargs)
@abstractmethod
def post_process(self,
model_output: Sequence[PoolingRequestOutput],
request_id: Optional[str] = None,
**kwargs) -> IOProcessorOutput:
def post_process(
self,
model_output: Sequence[PoolingRequestOutput],
request_id: Optional[str] = None,
**kwargs,
) -> IOProcessorOutput:
raise NotImplementedError
async def post_process_async(
@@ -52,8 +53,9 @@ class IOProcessor(ABC, Generic[IOProcessorInput, IOProcessorOutput]):
# We cannot guarantee outputs are returned in the same order they were
# fed to vLLM.
# Let's sort them by id before post_processing
sorted_output = sorted([(i, item) async for i, item in model_output],
key=lambda output: output[0])
sorted_output = sorted(
[(i, item) async for i, item in model_output], key=lambda output: output[0]
)
collected_output = [output[1] for output in sorted_output]
return self.post_process(collected_output, request_id, **kwargs)
@@ -63,5 +65,6 @@ class IOProcessor(ABC, Generic[IOProcessorInput, IOProcessorOutput]):
@abstractmethod
def output_to_response(
self, plugin_output: IOProcessorOutput) -> IOProcessorResponse:
self, plugin_output: IOProcessorOutput
) -> IOProcessorResponse:
raise NotImplementedError

35
vllm/plugins/lora_resolvers/filesystem_resolver.py
View File

@@ -10,25 +10,29 @@ from vllm.lora.resolver import LoRAResolver, LoRAResolverRegistry
class FilesystemResolver(LoRAResolver):
def __init__(self, lora_cache_dir: str):
self.lora_cache_dir = lora_cache_dir
async def resolve_lora(self, base_model_name: str,
lora_name: str) -> Optional[LoRARequest]:
async def resolve_lora(
self, base_model_name: str, lora_name: str
) -> Optional[LoRARequest]:
lora_path = os.path.join(self.lora_cache_dir, lora_name)
if os.path.exists(lora_path):
adapter_config_path = os.path.join(self.lora_cache_dir, lora_name,
"adapter_config.json")
adapter_config_path = os.path.join(
self.lora_cache_dir, lora_name, "adapter_config.json"
)
if os.path.exists(adapter_config_path):
with open(adapter_config_path) as file:
adapter_config = json.load(file)
if adapter_config["peft_type"] == "LORA" and adapter_config[
"base_model_name_or_path"] == base_model_name:
lora_request = LoRARequest(lora_name=lora_name,
lora_int_id=abs(
hash(lora_name)),
lora_path=lora_path)
if (
adapter_config["peft_type"] == "LORA"
and adapter_config["base_model_name_or_path"] == base_model_name
):
lora_request = LoRARequest(
lora_name=lora_name,
lora_int_id=abs(hash(lora_name)),
lora_path=lora_path,
)
return lora_request
return None
@@ -38,13 +42,12 @@ def register_filesystem_resolver():
lora_cache_dir = envs.VLLM_LORA_RESOLVER_CACHE_DIR
if lora_cache_dir:
if not os.path.exists(lora_cache_dir) or not os.path.isdir(
lora_cache_dir):
if not os.path.exists(lora_cache_dir) or not os.path.isdir(lora_cache_dir):
raise ValueError(
"VLLM_LORA_RESOLVER_CACHE_DIR must be set to a valid directory \
for Filesystem Resolver plugin to function")
for Filesystem Resolver plugin to function"
)
fs_resolver = FilesystemResolver(lora_cache_dir)
LoRAResolverRegistry.register_resolver("Filesystem Resolver",
fs_resolver)
LoRAResolverRegistry.register_resolver("Filesystem Resolver", fs_resolver)
return